diff options
| author | Berke Güzel <wenekar1@gmail.com> | 2026-10-12 17:49:34 +0300 |
|---|---|---|
| committer | Berke Güzel <wenekar1@gmail.com> | 2026-10-12 17:49:34 +0300 |
| commit | 4384435941b8c7658856b381d071aee2e0035e89 (patch) | |
| tree | 5ece99c3dd1decb8ec3fd404c0ece2b5865a552a /tests | |
term: move the term core to subprojects/term
The core now builds, tests, and benchmarks alone, as a Meson subproject
with no dependency on the main tree.
- include/term/ holds term.h and the header-only unicode/. src/ holds
the core and its private headers. The main tree gets only
include/term through term_dep, so it cannot reach a private header.
- The term tests, the term benchmarks, the Kitty parser suite, and
tools/amalgamate.py and tools/generate_unicode.py move with it. The
tests run as term:<name>.
- The main build passes its -march (arch_flag) and b_ndebug to the
subproject. A directory configured before this commit needs
meson configure -Dterm:arch_flag=-march=... once.
- The term tests compile with -UNDEBUG. They do work inside assert.
- The subproject exports term_test_dep with term_api.h, for the main
tree's render, thumbnail, color, and selection tests.
- The Kitty bridge used struct vector2i and has not compiled since the
public API redesign. It builds again. 2 of its 14 tests fail: REP
after a tab, and a cursor report after a resize.
- dirgraph.py drops the term and unicode layers. Its self-test uses pty.
- The symbols test ignores the __odr_asan symbols that ASan adds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Diffstat (limited to 'tests')
32 files changed, 12391 insertions, 0 deletions
diff --git a/tests/cross_test.sh b/tests/cross_test.sh new file mode 100644 index 0000000..6c0efa2 --- /dev/null +++ b/tests/cross_test.sh @@ -0,0 +1,24 @@ +#!/usr/bin/env bash +# The term core compiles as strict C11 for the platforms a host may embed +# it on. zig cc supplies the cross compilers. The aarch64 targets build the +# NEON paths, and windows-gnu builds the Windows path of compat.h. The +# amalgamation, one term.c, compiles the same way. +set -euo pipefail +root=${1:-$(dirname "$0")/..} +cd "$root/src" +command -v zig >/dev/null || { echo "zig not found; skipping"; exit 77; } + +out=$(mktemp -d) +trap 'rm -rf "$out"' EXIT +python3 "$root/tools/amalgamate.py" "$root" "$out/amalgamation" +status=0 +for target in aarch64-linux-gnu x86_64-macos aarch64-macos x86_64-windows-gnu; do + for src in *.c osc/*.c; do + zig cc -target "$target" -std=c11 -O2 -Wall -Wextra -Wpedantic -Werror \ + -I. -I../include/term -c "$src" -o "$out/obj.o" || { echo "$target: $src fails" >&2; status=1; } + done + zig cc -target "$target" -std=c11 -O2 -Wall -Wextra -Wpedantic -Werror \ + -c "$out/amalgamation/term.c" -o "$out/obj.o" || + { echo "$target: the amalgamation fails" >&2; status=1; } +done +exit $status diff --git a/tests/dcs_test.c b/tests/dcs_test.c new file mode 100644 index 0000000..0e67f78 --- /dev/null +++ b/tests/dcs_test.c @@ -0,0 +1,475 @@ +#include "term_api.h" +#include <assert.h> +#include <stdbool.h> +#include <stdio.h> +#include <string.h> + +static char response[16384]; +static size_t response_len; +static void respond(void *ud, const char *data, size_t len) +{ + (void)ud; + if (len >= sizeof(response)) len = sizeof(response) - 1; + memcpy(response, data, len); + response[len] = 0; + response_len = len; +} +static void write_seq(struct term *t, const char *s) +{ + for (; *s; s++) tt_write(t, s, 1); +} +static const char *row_text(struct term *t, int r, char *buf) +{ + int n = 0; + for (int c = 0; c < tt_cols(t); c++) + buf[n++] = (char)tt_view_cell(t, c, r)->text; + while (n > 0 && buf[n - 1] == ' ') n--; + buf[n] = 0; + return buf; +} +static bool row_is(struct term *t, int r, const char *want) +{ + char got[TERM_MAX_COLS + 1]; + if (strcmp(row_text(t, r, got), want)) { + fprintf(stderr, "row %d: got \"%s\" want \"%s\"\n", r, got, want); + return false; + } + return true; +} +static struct term *new_term(int cols, int rows) +{ + struct term *t = tt_new_opts(cols, rows, &(struct term_callbacks){ .reply = respond }); + assert(t); + response_len = 0; + response[0] = 0; + return t; +} +static void expect_response(const char *want) +{ + if (response_len != strlen(want) || memcmp(response, want, response_len)) { + fprintf(stderr, "response: got len %zu \"%s\" want \"%s\"\n", + response_len, response, want); + assert(!"response mismatch"); + } +} +static void expect_no_response(void) +{ + assert(response_len == 0); +} + +static void hex_encode(char *out, const char *s) +{ + for (; *s; s++) { + snprintf(out, 3, "%02x", (unsigned char)*s); + out += 2; + } + *out = 0; +} + +static void assert_rqss(struct term *t, const char *spec, const char *payload) +{ + char req[128], want[256]; + snprintf(req, sizeof(req), "\033P$q%s\033\\", spec); + snprintf(want, sizeof(want), "\033P1$r%s\033\\", payload); + response_len = 0; + write_seq(t, req); + expect_response(want); +} +static void assert_rqss_invalid(struct term *t, const char *spec) +{ + char req[128]; + snprintf(req, sizeof(req), "\033P$q%s\033\\", spec); + response_len = 0; + write_seq(t, req); + expect_response("\033P0$r\033\\"); +} + +static void test_decrqss_sgr(void) +{ + struct term *t = new_term(80, 24); + assert_rqss(t, "m", "m"); + write_seq(t, "\033[1;31m"); + assert_rqss(t, "m", "0;1;31m"); + term_free(t); + + t = new_term(80, 24); + write_seq(t, "\033[4:3m"); + assert_rqss(t, "m", "0;4:3m"); + term_free(t); + + t = new_term(80, 24); + write_seq(t, "\033[38;2;10;20;30m"); + assert_rqss(t, "m", "0;38:2:10:20:30m"); + term_free(t); +} + +static void test_decrqss_margins_and_page(void) +{ + struct term *t = new_term(80, 24); + assert_rqss(t, "r", "1;24r"); + assert_rqss(t, "t", "24t"); + assert_rqss(t, "$|", "80$|"); + assert_rqss(t, "*|", "24*|"); + write_seq(t, "\033[3;10r"); + assert_rqss(t, "r", "3;10r"); + term_free(t); + + t = new_term(80, 24); + assert_rqss(t, "s", "1;80s"); + write_seq(t, "\033[?69h\033[5;20s"); + assert_rqss(t, "s", "5;20s"); + term_free(t); +} + +static void test_decrqss_cursor_style(void) +{ + struct term *t = new_term(80, 24); + assert_rqss(t, " q", "1 q"); + write_seq(t, "\033[5 q"); + assert_rqss(t, " q", "5 q"); + write_seq(t, "\033[4 q"); + assert_rqss(t, " q", "4 q"); + term_free(t); +} + +static void test_decrqss_misc(void) +{ + struct term *t = new_term(80, 24); + assert_rqss(t, "\"p", "64;1\"p"); + assert_rqss(t, "\"q", "0\"q"); + assert_rqss(t, "$}", "0$}"); + assert_rqss(t, "$~", "0$~"); + assert_rqss(t, "*x", "0*x"); + write_seq(t, "\033[2*x"); + assert_rqss(t, "*x", "2*x"); + term_free(t); +} + +static void test_decrqss_invalid(void) +{ + struct term *t = new_term(80, 24); + static const char *bad[] = { + "){", ",|", ",}", ">0f", ">0m", ">0t", "z", "ABC", + }; + for (size_t i = 0; i < sizeof(bad) / sizeof(bad[0]); i++) + assert_rqss_invalid(t, bad[i]); + + response_len = 0; + write_seq(t, "\033P$x\033\\"); + expect_no_response(); + term_free(t); +} + +static void assert_tcap(struct term *t, const char *name, const char *value) +{ + char hn[128], hv[512], req[256], want[1024]; + hex_encode(hn, name); + hex_encode(hv, value); + snprintf(req, sizeof(req), "\033P+q%s\033\\", hn); + snprintf(want, sizeof(want), "\033P1+r%s=%s\033\\", hn, hv); + response_len = 0; + write_seq(t, req); + expect_response(want); +} + +static void test_xtgettcap(void) +{ + struct term *t = new_term(80, 24); + assert_tcap(t, "kind", "\033[1;2B"); + assert_tcap(t, "TN", "wsap_term"); + assert_tcap(t, "name", "wsap_term"); + assert_tcap(t, "Co", "16777216"); + assert_tcap(t, "colors", "16777216"); + assert_tcap(t, "RGB", "8/8/8"); + assert_tcap(t, "kbs", "\177"); + assert_tcap(t, "kcbt", "\033[Z"); + assert_tcap(t, "kcuu1", "\033OA"); + assert_tcap(t, "kcud1", "\033OB"); + assert_tcap(t, "kcuf1", "\033OC"); + assert_tcap(t, "kcub1", "\033OD"); + assert_tcap(t, "khome", "\033OH"); + assert_tcap(t, "kend", "\033OF"); + assert_tcap(t, "kent", "\033OM"); + assert_tcap(t, "kich1", "\033[2~"); + assert_tcap(t, "kdch1", "\033[3~"); + assert_tcap(t, "kpp", "\033[5~"); + assert_tcap(t, "knp", "\033[6~"); + assert_tcap(t, "kf1", "\033OP"); + assert_tcap(t, "kf2", "\033OQ"); + assert_tcap(t, "kf3", "\033OR"); + assert_tcap(t, "kf4", "\033OS"); + assert_tcap(t, "kf5", "\033[15~"); + assert_tcap(t, "kf6", "\033[17~"); + assert_tcap(t, "kf7", "\033[18~"); + assert_tcap(t, "kf8", "\033[19~"); + assert_tcap(t, "kf9", "\033[20~"); + assert_tcap(t, "kf10", "\033[21~"); + assert_tcap(t, "kf11", "\033[23~"); + assert_tcap(t, "kf12", "\033[24~"); + assert_tcap(t, "kf13", "\033[1;2P"); + assert_tcap(t, "kf24", "\033[24;2~"); + assert_tcap(t, "kUP", "\033[1;2A"); + assert_tcap(t, "kDN", "\033[1;2B"); + assert_tcap(t, "kLFT", "\033[1;2D"); + assert_tcap(t, "kRIT", "\033[1;2C"); + assert_tcap(t, "kmous", "\033[<"); + assert_tcap(t, "kri", "\033[1;2A"); + assert_tcap(t, "kxIN", "\033[I"); + assert_tcap(t, "kxOUT", "\033[O"); + term_free(t); +} + +static void test_xtgettcap_multi(void) +{ + struct term *t = new_term(80, 24); + char hk[16], htn[16], hb[16], hv[32], req[256], want[1024]; + hex_encode(hk, "kind"); + hex_encode(htn, "TN"); + hex_encode(hb, "BOGUS"); + hex_encode(hv, "\033[1;2B"); + + snprintf(req, sizeof(req), "\033P+q%s;%s\033\\", hk, htn); + snprintf(want, sizeof(want), "\033P1+r%s=%s;%s=%s\033\\", + hk, hv, htn, "777361705f7465726d"); + response_len = 0; + write_seq(t, req); + expect_response(want); + + snprintf(req, sizeof(req), "\033P+q%s;%s;%s\033\\", hk, hb, htn); + snprintf(want, sizeof(want), "\033P1+r%s=%s;%s\033\\", hk, hv, hb); + response_len = 0; + write_seq(t, req); + expect_response(want); + assert(!strstr(response, htn)); + + snprintf(req, sizeof(req), "\033P+q%s\033\\", hb); + snprintf(want, sizeof(want), "\033P0+r%s\033\\", hb); + response_len = 0; + write_seq(t, req); + expect_response(want); + term_free(t); +} + +static void test_xtgettcap_malformed(void) +{ + struct term *t = new_term(80, 24); + static const char *bad[] = {"6b696", "zz", "6g", "6b69 6e"}; + for (size_t i = 0; i < sizeof(bad) / sizeof(bad[0]); i++) { + char req[128]; + snprintf(req, sizeof(req), "\033P+q%s\033\\", bad[i]); + response_len = 0; + write_seq(t, req); + assert(response_len > 4); + assert(!memcmp(response, "\033P0+r", 5)); + assert(response[response_len - 2] == '\033'); + assert(response[response_len - 1] == '\\'); + } + term_free(t); +} + +static void test_xtsettcap(void) +{ + struct term *t = new_term(80, 24); + char hn[32]; + hex_encode(hn, "myterm"); + char req[128]; + snprintf(req, sizeof(req), "\033P+p%s\033\\", hn); + response_len = 0; + write_seq(t, req); + expect_no_response(); + assert_tcap(t, "TN", "myterm"); + term_free(t); +} + +static void test_xtgetxres(void) +{ + struct term *t = new_term(80, 24); + char hn[64], req[128], want[256]; + hex_encode(hn, "allowTcapOps"); + snprintf(req, sizeof(req), "\033P+Q%s\033\\", hn); + snprintf(want, sizeof(want), "\033P0+R%s\033\\", hn); + response_len = 0; + write_seq(t, req); + expect_response(want); + term_free(t); +} + +static void test_decrsps_cir(void) +{ + struct term *t = new_term(80, 24); + write_seq(t, "\033P1$t3;5;1;I;@;H;00;@;BBBB\033\\"); + assert(tt_cursor_y(t) == 2 && tt_cursor_x(t) == 4); + + write_seq(t, "Z"); + const struct term_cell *cell = tt_cell(t, 4, 2); + assert(cell->text == 'Z'); + assert(cell->attrs & TERM_ATTR_BOLD); + assert(cell->fg == TERM_COLOR_BG); + term_free(t); + + t = new_term(80, 24); + write_seq(t, "\033P1$t3;5;1;@;@;H;00;@;BBBB\033\\"); + write_seq(t, "\033[1;80Hab"); + assert(tt_cell(t, 79, 0)->text == 'a'); + assert(tt_cell(t, 0, 1)->text == 'b'); + term_free(t); + + t = new_term(80, 24); + write_seq(t, "\033[5;15r"); + write_seq(t, "\033P1$t3;5;1;@;@;I;00;@;BBBB\033\\"); + write_seq(t, "\033[1;1H"); + assert(tt_cursor_y(t) == 4 && tt_cursor_x(t) == 0); + term_free(t); +} + +static void test_decrsps_tabsr(void) +{ + struct term *t = new_term(80, 24); + write_seq(t, "\033P2$t4/10\033\\"); + write_seq(t, "\033[1;1H"); + write_seq(t, "\t"); + assert(tt_cursor_x(t) == 3); + write_seq(t, "\t"); + assert(tt_cursor_x(t) == 9); + write_seq(t, "\t"); + assert(tt_cursor_x(t) == 79); + term_free(t); +} + +static void test_synchronized_output(void) +{ + struct term *t = new_term(80, 24); + write_seq(t, "\033P=1s\033\\"); + assert(term_modes(t).synchronized); + write_seq(t, "\033P=2s\033\\"); + assert(!term_modes(t).synchronized); + + write_seq(t, "\033P=10s\033\\"); + assert(!term_modes(t).synchronized); + write_seq(t, "\033P=1;s\033\\"); + assert(term_modes(t).synchronized); + write_seq(t, "\033P=xs\033\\"); + assert(term_modes(t).synchronized); + write_seq(t, "\033P=2;s\033\\"); + assert(!term_modes(t).synchronized); + term_free(t); +} + +static void test_ignored_dcs(void) +{ + static const char *seqs[] = { + "\033P0;1|17/0d\033\\", + "\033P0!u%5\033\\", + "\033P1!uA\033\\", + "\033P1;0;1;10;0;2;0;1{~\033\\", + "\033P1;0;0q#0~~\033\\", + "\033P1pS(E)\033\\", + "\033P?~junk\033\\", + }; + for (size_t i = 0; i < sizeof(seqs) / sizeof(seqs[0]); i++) { + struct term *t = new_term(80, 24); + write_seq(t, "A"); + write_seq(t, seqs[i]); + write_seq(t, "B"); + assert(row_is(t, 0, "AB")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + expect_no_response(); + term_free(t); + } +} + +static void test_xtversion(void) +{ + struct term *t = new_term(80, 24); + response_len = 0; + write_seq(t, "\033[>0q"); + expect_response("\033P>|wsap_term\033\\"); + term_free(t); +} + +static void test_bel_terminator(void) +{ + struct term *t = new_term(80, 24); + response_len = 0; + write_seq(t, "\033P+q6b696e64\007"); + expect_response("\033P1+r6b696e64=1b5b313b3242\033\\"); + + response_len = 0; + write_seq(t, "\033P$qm\007"); + expect_response("\033P1$rm\033\\"); + term_free(t); +} + +/* A query far past the old 256-byte DCS buffer is answered in full, + * whether it arrives byte by byte or in one write. */ +static void test_dcs_long_query(void) +{ + enum { NAMES = 60 }; + char hk[16], hv[32]; + hex_encode(hk, "kind"); + hex_encode(hv, "\033[1;2B"); + static char req[4096], want[8192]; + int rn = snprintf(req, sizeof(req), "\033P+q"); + int wn = snprintf(want, sizeof(want), "\033P1+r"); + for (int i = 0; i < NAMES; i++) { + rn += snprintf(req + rn, sizeof(req) - (size_t)rn, "%s%s", i ? ";" : "", hk); + wn += snprintf(want + wn, sizeof(want) - (size_t)wn, "%s%s=%s", + i ? ";" : "", hk, hv); + } + snprintf(req + rn, sizeof(req) - (size_t)rn, "\033\\"); + snprintf(want + wn, sizeof(want) - (size_t)wn, "\033\\"); + assert(strlen(req) > 256); + + struct term *t = new_term(80, 24); + write_seq(t, req); + expect_response(want); + response_len = 0; + tt_write(t, req, strlen(req)); + expect_response(want); + write_seq(t, "X"); + assert(row_is(t, 0, "X")); + term_free(t); +} + +static void test_dcs_abort_c0(void) +{ + struct term *t = new_term(80, 24); + response_len = 0; + write_seq(t, "\033P+q6b69\nX"); + expect_no_response(); + assert(row_is(t, 1, "X")); + term_free(t); +} + +static void test_dcs_abort_esc(void) +{ + struct term *t = new_term(80, 24); + response_len = 0; + write_seq(t, "\033P$qm\033[5n"); + expect_response("\033[0n"); + term_free(t); +} + +int main(void) +{ + test_decrqss_sgr(); + test_decrqss_margins_and_page(); + test_decrqss_cursor_style(); + test_decrqss_misc(); + test_decrqss_invalid(); + test_xtgettcap(); + test_xtgettcap_multi(); + test_xtgettcap_malformed(); + test_xtsettcap(); + test_xtgetxres(); + test_decrsps_cir(); + test_decrsps_tabsr(); + test_synchronized_output(); + test_ignored_dcs(); + test_xtversion(); + test_bel_terminator(); + test_dcs_long_query(); + test_dcs_abort_c0(); + test_dcs_abort_esc(); + return 0; +} diff --git a/tests/extern/README.md b/tests/extern/README.md new file mode 100644 index 0000000..cff42ed --- /dev/null +++ b/tests/extern/README.md @@ -0,0 +1,43 @@ +# Imported test suites + +This directory holds test suites imported from external projects. These are +kept for reference and cross-checking against our own terminal emulator's +behavior. On Linux, `make test` builds and runs the adapted Kitty parser suite. + +## kitty/ + +Vendored from <https://github.com/kovidgoyal/kitty>, commit `f13c8cd4`. + +### Running the adapted suite + +```sh +meson test -C build --suite kitty +``` + +This builds `kitty_bridge` (a shared library compiled from +the project's terminal sources), then runs the +Kitty parser test suite against it via a Python compatibility layer. + +### Files + +- `kitty/parser.py` — Kitty's VT/CSI parser test suite (`kitty_tests/parser.py`), + with imports patched to use the local compatibility modules below. +- `kitty_parser.py` — Pure-Python VT parser that emits Kitty-style callback + events while feeding every byte to the C terminal via `screen.write()`. + Maintains state across calls for split-write (threading) tests. +- `screen.py` — `ctypes` wrapper around `libterm.so` providing the Kitty + `Screen`, `Cursor`, `Line`, `CellCursor`, and `Callbacks` API. +- `fast_data_types.py` — Pure-Python replacements for Kitty's C-extension + helpers (base64, UTF-8 decode, constants). +- `kitty_harness.py` — `BaseTest`, `CmdDump`, and `parse_bytes` re-export. +- `__init__.py` — Package marker. + +### Test status (14 tests) + +All 14 tests pass. Features implemented for Kitty parity include DEC +Special Graphics charset translation (G0/G1, SO/SI), literal `\t` cells, +LNM-default-off linefeed semantics, CSI insert-character and REP edge +cases, DECRQM mode reports, DECRQSS (DCS `$q`) and XTGETTCAP (DCS `+q`) +responses, DECCARA/DECSACE (CSI `$r` / `*x`), secondary device attributes +(CSI `=c`), XTWINOPS cell-size report (CSI `14t`), underline styles 0–5, +and SGR underline color (58/59) including colon-form subparameters. diff --git a/tests/extern/kitty/.gitignore b/tests/extern/kitty/.gitignore new file mode 100644 index 0000000..79a6f1c --- /dev/null +++ b/tests/extern/kitty/.gitignore @@ -0,0 +1,2 @@ +__pycache__/ +libterm.so diff --git a/tests/extern/kitty/__init__.py b/tests/extern/kitty/__init__.py new file mode 100644 index 0000000..5dddbaf --- /dev/null +++ b/tests/extern/kitty/__init__.py @@ -0,0 +1,9 @@ +"""Package marker for the Kitty parser test harness. + +Re-exports :class:`BaseTest`, :class:`CmdDump`, and ``parse_bytes`` from +:mod:`kitty_harness`. +""" + +from kitty_harness import BaseTest, CmdDump, parse_bytes + +__all__ = ['BaseTest', 'CmdDump', 'parse_bytes'] diff --git a/tests/extern/kitty/fast_data_types.py b/tests/extern/kitty/fast_data_types.py new file mode 100644 index 0000000..c8c519d --- /dev/null +++ b/tests/extern/kitty/fast_data_types.py @@ -0,0 +1,237 @@ +"""Stubs for Kitty's ``kitty.fast_data_types`` C extension module. + +These pure-Python implementations provide the subset of the API used by +Kitty's parser test-suite. The SIMD-accelerated variants are not available, +so ``has_avx2`` and ``has_sse4_2`` are ``False`` and the decoder falls back +to a scalar implementation that matches the WHATWG Encoding Standard. +""" + +import base64 as _base64 + +# --------------------------------------------------------------------------- +# Constants +# --------------------------------------------------------------------------- + +CURSOR_BLOCK = 0 +VT_PARSER_BUFFER_SIZE = 8192 # Large enough that buffer-full tests don't trigger + +has_avx2 = False +has_sse4_2 = False + +# --------------------------------------------------------------------------- +# Base64 helpers (padding-free, matching Kitty's convention) +# --------------------------------------------------------------------------- + +def base64_decode(data) -> bytes: + """Decode base64, accepting both padded and unpadded input.""" + if isinstance(data, str): + data = data.encode('utf-8') + data = bytes(data) + # Strip any existing padding and whitespace. + data = data.rstrip(b'=').strip() + # Re-add the minimum padding required by the stdlib decoder. + missing = len(data) % 4 + if missing: + data += b'=' * (4 - missing) + return _base64.b64decode(data) + + +def base64_encode(data) -> bytes: + """Encode to base64 without trailing ``=`` padding.""" + if isinstance(data, str): + data = data.encode('utf-8') + return _base64.b64encode(bytes(data)).rstrip(b'=') + + +# --------------------------------------------------------------------------- +# Byte search helper +# --------------------------------------------------------------------------- + +def test_find_either_of_two_bytes(buf, a, b, which=0, align_offset=0): + """Return the index of the first occurrence of byte *a* or *b* in *buf*. + + Returns -1 when neither byte is found. The *which* and *align_offset* + parameters select SIMD variants in Kitty's native code; the pure-Python + fallback ignores them. + """ + if isinstance(buf, str): + buf = buf.encode('latin-1') + buf = bytes(buf) + ia = int(a) + ib = int(b) + for i, byte in enumerate(buf): + if byte == ia or byte == ib: + return i + return -1 + + +# --------------------------------------------------------------------------- +# UTF-8 decoder with WHATWG U+FFFD substitution +# --------------------------------------------------------------------------- + +# Module-level state for incomplete trailing sequences across calls. +_utf8_pending = bytearray() + + +def test_utf8_decode_to_sentinel(data, which=0): + """Decode UTF-8 bytes until an ESC (0x1b) sentinel or end of data. + + Returns ``(found_sentinel: bool, decoded_text: str, num_consumed: int)``. + + *num_consumed* counts only the bytes consumed from *data* (not any + pending bytes carried over from a previous call). Invalid sequences are + replaced with U+FFFD using the maximal-subpart rules from the WHATWG + Encoding Standard. + + Passing *which* = -1 resets the internal pending-state buffer. + """ + global _utf8_pending + + # --- reset ----------------------------------------------------------- + if which == -1: + _utf8_pending = bytearray() + return (False, '', 0) + + initial_pending = len(_utf8_pending) + buf = bytes(_utf8_pending) + bytes(data) + _utf8_pending = bytearray() + + result = [] + i = 0 + n = len(buf) + found_sentinel = False + + while i < n: + byte = buf[i] + + # ESC sentinel + if byte == 0x1B: + found_sentinel = True + i += 1 + break + + # ASCII + if byte < 0x80: + result.append(chr(byte)) + i += 1 + continue + + # Lone continuation byte 0x80-0xBF + if byte < 0xC0: + result.append('\ufffd') + i += 1 + continue + + # Invalid starter bytes 0xC0-0xC1 + if byte < 0xC2: + result.append('\ufffd') + i += 1 + continue + + # 2-byte sequence 0xC2-0xDF + if byte < 0xE0: + if i + 1 < n: + nb = buf[i + 1] + if 0x80 <= nb <= 0xBF: + cp = ((byte & 0x1F) << 6) | (nb & 0x3F) + result.append(chr(cp)) + i += 2 + continue + # Bad continuation: starter alone is one subpart + result.append('\ufffd') + i += 1 + continue + # Incomplete at end of input + _utf8_pending.extend(buf[i:]) + i = n + break + + # 3-byte sequence 0xE0-0xEF + if byte < 0xF0: + if byte == 0xE0: + lo, hi = 0xA0, 0xBF + elif byte == 0xED: + lo, hi = 0x80, 0x9F + else: + lo, hi = 0x80, 0xBF + + if i + 1 < n: + nb = buf[i + 1] + if lo <= nb <= hi: + if i + 2 < n: + nb2 = buf[i + 2] + if 0x80 <= nb2 <= 0xBF: + cp = ((byte & 0x0F) << 12) | ((nb & 0x3F) << 6) | (nb2 & 0x3F) + result.append(chr(cp)) + i += 3 + continue + # Bad 2nd continuation: starter + 1st cont = one subpart + result.append('\ufffd') + i += 2 + continue + # Incomplete: save starter + 1st continuation + _utf8_pending.extend(buf[i:]) + i = n + break + # Bad 1st continuation: starter alone is one subpart + result.append('\ufffd') + i += 1 + continue + # Incomplete: save starter only + _utf8_pending.extend(buf[i:]) + i = n + break + + # 4-byte sequence 0xF0-0xF4 + if byte < 0xF5: + if byte == 0xF0: + lo, hi = 0x90, 0xBF + elif byte == 0xF4: + lo, hi = 0x80, 0x8F + else: + lo, hi = 0x80, 0xBF + + if i + 1 < n: + nb = buf[i + 1] + if lo <= nb <= hi: + if i + 2 < n: + nb2 = buf[i + 2] + if 0x80 <= nb2 <= 0xBF: + if i + 3 < n: + nb3 = buf[i + 3] + if 0x80 <= nb3 <= 0xBF: + cp = ((byte & 0x07) << 18) | ((nb & 0x3F) << 12) | ((nb2 & 0x3F) << 6) | (nb3 & 0x3F) + result.append(chr(cp)) + i += 4 + continue + # Bad 3rd continuation + result.append('\ufffd') + i += 3 + continue + # Incomplete: save starter + 2 continuations + _utf8_pending.extend(buf[i:]) + i = n + break + # Bad 2nd continuation: starter + 1st cont = one subpart + result.append('\ufffd') + i += 2 + continue + # Incomplete: save starter + 1st continuation + _utf8_pending.extend(buf[i:]) + i = n + break + # Bad 1st continuation: starter alone is one subpart + result.append('\ufffd') + i += 1 + continue + # Incomplete: save starter only + _utf8_pending.extend(buf[i:]) + i = n + break + + # Invalid starter bytes 0xF5-0xFF + result.append('\ufffd') + i += 1 + + num_consumed = max(0, i - initial_pending) + return (found_sentinel, ''.join(result), num_consumed) diff --git a/tests/extern/kitty/kitty_harness.py b/tests/extern/kitty/kitty_harness.py new file mode 100644 index 0000000..d528478 --- /dev/null +++ b/tests/extern/kitty/kitty_harness.py @@ -0,0 +1,46 @@ +"""Test harness for running Kitty's parser.py test suite. + +Provides :class:`BaseTest`, :class:`CmdDump`, and re-exports ``parse_bytes`` +from ``kitty_parser`` (written by a separate subagent). +""" + +import unittest + +from kitty_parser import parse_bytes + + +class CmdDump(list): + """Collects parser events, matching Kitty's ``CmdDump``.""" + + def __call__(self, window_id, *a): + if a and a[0] == 'bytes': + return + if a and a[0] == 'error': + a = a[1:] + self.append(tuple(a)) + + def get_result(self): + current = '' + q = [] + for args in self: + if args[0] == 'draw': + current += args[1] + else: + if current: + q.append(('draw', current)) + current = '' + q.append(args) + if current: + q.append(('draw', current)) + return tuple(q) + + +class BaseTest(unittest.TestCase): + """Base test class matching Kitty's test API.""" + + def ae(self, a, b, msg=None): + self.assertEqual(a, b, msg) + + def create_screen(self, cols=5): + from screen import Screen + return Screen(cols, 5) # 5 rows like Kitty diff --git a/tests/extern/kitty/kitty_parser.py b/tests/extern/kitty/kitty_parser.py new file mode 100644 index 0000000..f34a608 --- /dev/null +++ b/tests/extern/kitty/kitty_parser.py @@ -0,0 +1,1068 @@ +"""Pure-Python VT parser that emits Kitty-style callback events. + +This parser mirrors the event vocabulary that Kitty's ``parser.py`` test +suite expects (``draw``, ``screen_cursor_position``, +``select_graphic_rendition``, ``set_title``, …) while feeding every byte +to the C terminal through ``screen.write(data)`` so that observable +screen state is driven by the real implementation. + +The parser maintains state across calls so that partial escape sequences +split across ``parse_bytes`` invocations are handled correctly — this is +exercised by ``test_parser_threading``. +""" + +# Parser states +_GROUND = 0 +_ESC = 1 +_CSI = 2 +_OSC = 3 +_DCS = 4 +_APC = 5 +_PM = 6 +_SOS = 7 +_STR_ESC = 8 +_CHARSET = 9 +_ESC_PCT = 10 + +# CSI byte classes +_CSI_PARAM = 0 # 0x30-0x3a (digits, ';', ':') +_CSI_PREFIX = 1 # 0x3c-0x3f (< = > ?) +_CSI_INTER = 2 # 0x20-0x2f +_CSI_FINAL = 3 # 0x40-0x7e +_CSI_OTHER = 4 + +_MAX_PARAMS = 256 + + +class _CSIState: + """CSI parse state, mirroring kitty's vt-parser.c grammar. + + Params accumulate into ``accum`` with ``num_digits`` digits seen; + ``commit`` pushes mult*accum. A param with no digits is never + committed, so a trailing ';' does not create an empty final param. + ``phase``: 0=start, 1=body, 2=post-secondary (after intermediate). + """ + + __slots__ = ('prefix', 'intermediate', 'final', 'params', 'nparams', 'subparams', + 'num_digits', 'mult', 'accum', 'phase') + + def __init__(self): + self.reset() + + def reset(self): + self.prefix = 0 + self.intermediate = 0 + self.final = 0 + self.params = [0] * _MAX_PARAMS + self.nparams = 0 + self.subparams = 0 + self.num_digits = 0 + self.mult = 1 + self.accum = 0 + self.phase = 0 + + def add_digit(self, d): + if self.num_digits >= 17: + return + self.accum = self.accum * 10 + d + self.num_digits += 1 + + def commit(self): + """Commit the pending param. Returns False on param overflow.""" + if not self.num_digits: + return True + if self.nparams >= _MAX_PARAMS: + return False + v = self.mult * self.accum + self.params[self.nparams] = ((v + 2**31) % 2**32) - 2**31 + self.nparams += 1 + self.num_digits = 0 + self.mult = 1 + self.accum = 0 + return True + + +class KittyParser: + """Stateful VT parser that emits Kitty-style events.""" + + def __init__(self): + self.state = _GROUND + self.text = '' # accumulated UTF-8 text for draw events + self.utf8_pending = b'' # incomplete UTF-8 sequence + self.csi = _CSIState() + self.str_buf = bytearray() # OSC/DCS/APC/PM/SOS payload + self.str_type = 0 # which string mode we're in + self.charset_slot = 0 + + # -- text / draw -- + + def _flush_text(self, cb): + if self.text: + cb(0, 'draw', self.text) + self.text = '' + + def _emit(self, cb, *args): + self._flush_text(cb) + cb(0, *args) + + def _kill_pending(self): + """A control byte kills a pending incomplete UTF-8 sequence.""" + if self.utf8_pending: + self.utf8_pending = b'' + self.text += '\ufffd' + + @staticmethod + def _control_event(cb, b): + if b == 0x0a: + cb(0, 'screen_linefeed') + elif b == 0x0d: + cb(0, 'screen_carriage_return') + elif b == 0x08: + cb(0, 'screen_backspace') + elif b == 0x09: + cb(0, 'screen_tab') + + def _decode_utf8(self, data, cb): + """Decode UTF-8 bytes, appending to self.text. Uses U+FFFD for errors.""" + # Combine pending bytes with new data + buf = self.utf8_pending + bytes(data) + self.utf8_pending = b'' + i = 0 + n = len(buf) + while i < n: + b = buf[i] + # C0 controls (including ESC) terminate text and are handled by caller + if b < 0x20: + # Do NOT save to utf8_pending — C0 controls are not incomplete UTF-8 + return i # Return position of the control byte + if b < 0x80: + self.text += chr(b) + i += 1 + elif b < 0xC2: + # Invalid starter (0x80-0xC1) or continuation byte + self.text += '\ufffd' + i += 1 + elif b < 0xE0: + # 2-byte sequence + if i + 1 < n: + b2 = buf[i + 1] + if 0x80 <= b2 <= 0xBF: + self.text += chr((b & 0x1F) << 6 | (b2 & 0x3F)) + i += 2 + else: + # Bad continuation + self.text += '\ufffd' + i += 1 + else: + # Incomplete — save for next call + self.utf8_pending = buf[i:] + return n + elif b < 0xF0: + # 3-byte sequence + if i + 2 < n: + b2, b3 = buf[i + 1], buf[i + 2] + # Check validity + if b == 0xE0: + valid = 0xA0 <= b2 <= 0xBF + elif b == 0xED: + valid = 0x80 <= b2 <= 0x9F + else: + valid = 0x80 <= b2 <= 0xBF + if valid and 0x80 <= b3 <= 0xBF: + self.text += chr((b & 0x0F) << 12 | (b2 & 0x3F) << 6 | (b3 & 0x3F)) + i += 3 + elif 0x80 <= b2 <= 0xBF: + # First continuation is valid, second is not + self.text += '\ufffd' + i += 2 + # b3 will be processed next (may be another U+FFFD) + else: + self.text += '\ufffd' + i += 1 + elif i + 1 < n: + b2 = buf[i + 1] + if 0x80 <= b2 <= 0xBF: + # Could be valid — need more data + self.utf8_pending = buf[i:] + return n + else: + self.text += '\ufffd' + i += 1 + else: + self.utf8_pending = buf[i:] + return n + elif b < 0xF5: + # 4-byte sequence + if i + 3 < n: + b2, b3, b4 = buf[i + 1], buf[i + 2], buf[i + 3] + if b == 0xF0: + valid2 = 0x90 <= b2 <= 0xBF + elif b == 0xF4: + valid2 = 0x80 <= b2 <= 0x8F + else: + valid2 = 0x80 <= b2 <= 0xBF + if valid2 and 0x80 <= b3 <= 0xBF and 0x80 <= b4 <= 0xBF: + cp = (b & 0x07) << 18 | (b2 & 0x3F) << 12 | (b3 & 0x3F) << 6 | (b4 & 0x3F) + self.text += chr(cp) + i += 4 + elif 0x80 <= b2 <= 0xBF: + if 0x80 <= b3 <= 0xBF: + # b4 is bad + self.text += '\ufffd' + i += 3 + else: + self.text += '\ufffd' + i += 2 + else: + self.text += '\ufffd' + i += 1 + else: + # Not enough bytes — save for next call + self.utf8_pending = buf[i:] + return n + else: + # 0xF5-0xFF: never valid + self.text += '\ufffd' + i += 1 + return n + + # -- CSI dispatch -- + + def _csi_count(self, idx=0): + """Return param[idx] or 1 if zero/missing.""" + if idx < self.csi.nparams and self.csi.params[idx] > 0: + return self.csi.params[idx] + return 1 + + def _format_sgr(self): + """Format SGR params into Kitty-style colon/semicolon groups.""" + if self.csi.nparams == 0: + return ['0'] + params = self.csi.params[:self.csi.nparams] + subparams = self.csi.subparams + groups = [] + current = [] + i = 0 + while i < len(params): + p = params[i] + next_has_colon = (i + 1 < len(params)) and bool(subparams & (1 << (i + 1))) + if p in (38, 48, 58): + # Extended color — flush current, consume following params + if current: + groups.append(';'.join(current)) + current = [] + sub = [str(p)] + i += 1 + while i < len(params) and params[i] not in (38, 48, 58): + sub.append(str(params[i])) + i += 1 + groups.append(':'.join(sub)) + elif next_has_colon: + # This param starts a colon group + if current: + groups.append(';'.join(current)) + current = [] + sub = [str(p)] + i += 1 + while i < len(params) and (subparams & (1 << i)): + sub.append(str(params[i])) + i += 1 + groups.append(':'.join(sub)) + else: + current.append(str(p)) + i += 1 + if current: + groups.append(';'.join(current)) + return groups + + def _dispatch_csi(self, screen, cb): + c = self.csi + final = chr(c.intermediate) if c.intermediate else '' + # Handle intermediate-based dispatches first + if c.intermediate: + if c.intermediate == ord(' ') and c.prefix == 0: + if c.final == ord('q'): + # DECSCUSR: CSI Ps SP q + if c.nparams > 0 and c.params[0] in (3, 4): + self._emit(cb, 'screen_set_cursor', c.params[0], ord(' ')) + elif c.nparams > 0 and c.params[0] in (1, 2, 5, 6): + self._emit(cb, 'screen_set_cursor', c.params[0], ord(' ')) + elif c.nparams == 0: + self._emit(cb, 'screen_set_cursor', 0, ord(' ')) + elif c.nparams > 0 and c.params[0] == 3: + # Shift codes + if c.final == ord('@'): + self._emit(cb, 'Shift left escape code not implemented') + elif c.final == ord('A'): + self._emit(cb, 'Shift right escape code not implemented') + elif c.final == ord('S'): + self._emit(cb, 'Select presentation directions escape code not implemented') + return + if c.intermediate == ord('!') and c.prefix == 0: + # Soft reset: CSI ! p — no event in Kitty's test + return + if c.intermediate == ord('$'): + if c.final == ord('p'): + # DECRQM: CSI ? Ps $ p / CSI Ps $ p + mode = c.params[0] if c.nparams > 0 else 0 + private = 1 if c.prefix == ord('?') else 0 + self._emit(cb, 'report_mode_status', mode, private) + elif c.final == ord('r') and c.prefix == 0: + # DECCARA: CSI $ r + for event_str in self._format_deccara(): + self._emit(cb, 'deccara', event_str) + return + if c.intermediate == ord('*') and c.prefix == 0: + # DECSACE: CSI Ps * x + if c.nparams > 0: + self._emit(cb, 'screen_decsace', c.params[0]) + else: + self._emit(cb, 'screen_decsace', 0) + return + if c.intermediate == ord('+') and c.final == ord('T'): + self._emit(cb, 'screen_reverse_scroll_and_fill_from_scrollback', + self._csi_count()) + return + return + + # No intermediate + f = c.final + pfx = c.prefix + + if f == ord('@'): + if c.nparams > 1: + self._emit(cb, f'CSI code @ has {c.nparams} > 1 parameters') + return + if c.nparams == 1 and c.params[0] < 0: + self._emit(cb, f'CSI code @ is not allowed to have negative parameter ({c.params[0]})') + return + self._emit(cb, 'screen_insert_characters', self._csi_count()) + return + + if f == ord('H') or f == ord('f'): + row = c.params[0] if c.nparams > 0 else 1 + col = c.params[1] if c.nparams > 1 else 1 + self._emit(cb, 'screen_cursor_position', row, col) + return + + if f == ord('J'): + mode = c.params[0] if c.nparams > 0 else 0 + private = 1 if pfx == ord('?') else 0 + self._emit(cb, 'screen_erase_in_display', mode, private) + return + + if f == ord('K'): + mode = c.params[0] if c.nparams > 0 else 0 + private = 1 if pfx == ord('?') else 0 + self._emit(cb, 'screen_erase_in_line', mode, private) + return + + if f == ord('h'): + private = 1 if pfx == ord('?') else 0 + if c.nparams == 0: + return + for i in range(c.nparams): + self._emit(cb, 'screen_set_mode', c.params[i], private) + return + + if f == ord('l'): + private = 1 if pfx == ord('?') else 0 + if c.nparams == 0: + return + for i in range(c.nparams): + self._emit(cb, 'screen_reset_mode', c.params[i], private) + return + + if f == ord('c'): + if pfx == ord('>'): + self._emit(cb, 'report_device_attributes', c.params[0] if c.nparams > 0 else 0, 62) + elif pfx == ord('='): + self._emit(cb, 'report_device_attributes', c.params[0] if c.nparams > 0 else 0, 61) + elif pfx == 0: + self._emit(cb, 'report_device_attributes', c.params[0] if c.nparams > 0 else 0, 0) + return + + if f == ord('m'): + for grp in self._format_sgr(): + self._emit(cb, 'select_graphic_rendition', grp) + return + + if f == ord('n'): + mode = c.params[0] if c.nparams > 0 else 0 + private = 1 if pfx == ord('?') else 0 + self._emit(cb, 'report_device_status', mode, private) + return + + if f == ord('r'): + if pfx: + return + if c.nparams == 0: + self._emit(cb, 'screen_set_margins', 0, 0) + else: + top = c.params[0] if c.params[0] > 0 else 1 + bottom = c.params[1] if c.nparams > 1 and c.params[1] > 0 else 0 + self._emit(cb, 'screen_set_margins', top, bottom) + return + + if f == ord('b'): + if c.nparams > 1: + return + self._emit(cb, 'screen_repeat_character', self._csi_count()) + return + + if f == ord('T'): + if pfx == 0 and c.nparams <= 1: + self._emit(cb, 'screen_reverse_scroll', self._csi_count()) + return + + if f == ord('t'): + if pfx == 0 and c.nparams > 0: + self._emit(cb, 'screen_report_size', c.params[0], c.params[1] if c.nparams > 1 else 0) + return + + if f == ord('p'): + return + + # Unhandled CSI — silently ignore (no event) + + def _format_deccara(self): + """Format DECCARA params (CSI $ r) into a list of event strings. + + Each event is: 'top;left;bottom;right;sgr_group' + """ + c = self.csi + if c.nparams == 0: + return ['0;0;0;0;0'] + params = list(c.params[:c.nparams]) + while len(params) < 4: + params.append(0) + subparams = c.subparams + rect = ';'.join(str(p) for p in params[:4]) + # Format SGR groups from params[4:] + sgr_params = params[4:] + if not sgr_params: + return [f'{rect};0'] + groups = [] + current = [] + i = 0 + while i < len(sgr_params): + p = sgr_params[i] + next_has_colon = (i + 1 < len(sgr_params)) and bool(subparams & (1 << (i + 5))) + if p in (38, 48, 58): + if current: + groups.append(';'.join(current)) + current = [] + sub = [str(p)] + i += 1 + # Consume following params that are subparams (preceded by ':') + while i < len(sgr_params) and (subparams & (1 << (i + 4))): + sub.append(str(sgr_params[i])) + i += 1 + groups.append(':'.join(sub)) + elif next_has_colon: + if current: + groups.append(';'.join(current)) + current = [] + sub = [str(p)] + i += 1 + while i < len(sgr_params) and (subparams & (1 << (i + 4))): + sub.append(str(sgr_params[i])) + i += 1 + groups.append(':'.join(sub)) + else: + current.append(str(p)) + i += 1 + if current: + groups.append(';'.join(current)) + return [f'{rect};{g}' for g in groups] + + # -- OSC dispatch -- + + def _dispatch_osc(self, screen, cb): + payload = bytes(self.str_buf).decode('utf-8', errors='replace') + # Parse OSC code: leading digits form the code, rest is data. + # If there's a ';', it separates code from data. + sep = payload.find(';') + if sep < 0: + # No ';' — leading digits are the code, rest is data + i = 0 + while i < len(payload) and payload[i].isdigit(): + i += 1 + if i == 0: + # No leading digits — treat as OSC 0 with payload as title + code = 0 + rest = payload + elif i == len(payload): + # Entire payload is digits — code only, no data + code = int(payload) + rest = '' + else: + code = int(payload[:i]) + rest = payload[i:] + else: + code_str = payload[:sep] + rest = payload[sep + 1:] + try: + code = int(code_str) if code_str else 0 + except ValueError: + code = -1 + + if code == 0: + self._emit(cb, 'set_title', rest) + self._emit(cb, 'set_icon', rest) + elif code == 1: + self._emit(cb, 'set_icon', rest) + elif code == 2: + self._emit(cb, 'set_title', rest) + elif code == 8: + # Hyperlink + if not rest: + self._emit(cb, 'set_active_hyperlink', None, None) + elif ';' not in rest: + self._emit(cb, 'Ignoring malformed OSC 8 code') + else: + params, uri = rest.split(';', 1) + if not uri: + uri = None + link_id = None + if params: + for kv in params.split(':'): + if kv.startswith('id=') and link_id is None: + link_id = kv[3:] + self._emit(cb, 'set_active_hyperlink', link_id, uri) + elif code == 9: + self._emit(cb, 'desktop_notify', 9, rest) + elif code == 99: + self._emit(cb, 'desktop_notify', 99, rest) + elif code == 52: + self._emit(cb, 'clipboard_control', 52, rest) + elif code == 110: + self._emit(cb, 'set_dynamic_color', 110, rest) + elif code == 22: + self._emit(cb, 'set_dynamic_color', 22, rest) + else: + # Unknown OSC — pass through as set_dynamic_color for some, + # or ignore. Kitty's test only uses specific codes. + pass + + # -- DCS dispatch -- + + def _dispatch_dcs(self, screen, cb): + payload = bytes(self.str_buf) + if not payload: + return + + # DCS +q — request capabilities + if payload.startswith(b'+q'): + data = payload[2:].decode('ascii', errors='replace') + self._emit(cb, 'screen_request_capabilities', 43, data) + return + + # DCS $q — request status + if payload.startswith(b'$q'): + data = payload[2:].decode('ascii', errors='replace') + self._emit(cb, 'screen_request_capabilities', ord('$'), data) + return + + # DCS @kitty-cmd + if payload.startswith(b'@kitty-cmd'): + data = payload[len(b'@kitty-cmd'):].decode('utf-8', errors='replace') + self._emit(cb, 'handle_remote_cmd', data) + return + + # DCS @kitty-print + if payload.startswith(b'@kitty-print|'): + data = payload[len(b'@kitty-print|'):].decode('ascii', errors='replace') + self._emit(cb, 'handle_remote_print', data) + return + + # DCS =1s / =2s — pending mode + if payload == b'=1s': + self._emit(cb, 'screen_start_pending_mode') + return + if payload == b'=2s': + self._emit(cb, 'screen_stop_pending_mode') + return + + # -- Graphics command (APC _ G) -- + + def _dispatch_graphics(self, screen, cb): + payload = bytes(self.str_buf) + if not payload: + return + + # Short key → full name mapping (from Kitty's graphics protocol) + key_map = { + 'a': 'action', 'd': 'delete_action', 't': 'transmission_type', + 'c': 'compressed', 'f': 'format', 'm': 'more', 'i': 'id', + 's': 'data_width', 'o': 'data_offset', 'w': 'width', + 'h': 'height', 'x': 'x_offset', 'y': 'y_offset', + 'v': 'data_height', 'u': 'data_sz', 'C': 'cursor_movement', + 'n': 'num_cells', 'r': 'num_lines', 'X': 'cell_x_offset', + 'Y': 'cell_y_offset', 'z': 'z_index', 'p': 'placement_id', + 'I': 'image_number', 'q': 'quiet', 'U': 'unicode_placement', + 'P': 'parent_id', 'Q': 'parent_placement_id', + 'O': 'offset_from_parent_x', 'N': 'offset_from_parent_y', + } + + # Parse the graphics command + # Format: _G{params};{base64_payload}ESC\ + # params are key=value pairs separated by commas + semi = payload.find(b';') + if semi >= 0: + cmd_str = payload[:semi] + b64_payload = payload[semi + 1:] + else: + cmd_str = payload + b64_payload = b'' + + # Parse key=value pairs + result = {} + # Defaults + for f in 'action delete_action transmission_type compressed'.split(): + result[f] = b'\0' + for f in ('format more id data_sz data_offset width height x_offset y_offset ' + 'data_height data_width cursor_movement num_cells num_lines ' + 'cell_x_offset cell_y_offset z_index placement_id image_number ' + 'quiet unicode_placement parent_id parent_placement_id ' + 'offset_from_parent_x offset_from_parent_y').split(): + result[f] = 0 + try: + from fast_data_types import base64_decode + result[''] = base64_decode(b64_payload).decode('utf-8', errors='replace') + except Exception: + result[''] = b64_payload.decode('ascii', errors='replace') + + i = 0 + n = len(cmd_str) + while i < n: + # Parse key + key_start = i + while i < n and cmd_str[i] != ord('='): + ch = cmd_str[i] + if ch == ord(','): + self._emit(cb, 'Malformed GraphicsCommand control block, invalid key character: 0x%02x' % ch) + return + # Only accept known graphics command keys + if chr(ch) not in key_map: + self._emit(cb, 'Malformed GraphicsCommand control block, invalid key character: 0x%02x' % ch) + return + i += 1 + if i >= n: + self._emit(cb, 'Malformed GraphicsCommand control block, no = after key') + return + key = cmd_str[key_start:i].decode('ascii') + i += 1 # skip '=" + + # Parse value + val_start = i + full_key = key_map.get(key, key) + is_char_key = full_key in ('action', 'delete_action', 'transmission_type', 'compressed') + while i < n and cmd_str[i] != ord(','): + ch = cmd_str[i] + if ch == ord('='): + # = after value — check if we have a value yet + if i == val_start: + # No value parsed yet + self._emit(cb, 'Malformed GraphicsCommand control block, expecting an integer value for key: %s' % key) + else: + self._emit(cb, 'Malformed GraphicsCommand control block, expecting a , or semi-colon after a value, found: 0x%02x' % ch) + return + if is_char_key: + # Character keys accept a single character + if i > val_start: + # Already got one char — only one expected + self._emit(cb, 'Malformed GraphicsCommand control block, expecting a , or semi-colon after a value, found: 0x%02x' % ch) + return + elif not (ord('0') <= ch <= ord('9') or ch == ord('-')): + self._emit(cb, 'Malformed GraphicsCommand control block, expecting an integer value for key: %s' % key) + return + i += 1 + + val_str = cmd_str[val_start:i].decode('ascii') + if not val_str: + self._emit(cb, 'Malformed GraphicsCommand control block, expecting an integer value') + return + if is_char_key: + result[full_key] = val_str.encode('ascii') + else: + try: + val = int(val_str) + except ValueError: + self._emit(cb, 'Malformed GraphicsCommand control block, expecting an integer value for key: %s' % key) + return + if val > 0xFFFFFFFF: + self._emit(cb, 'Malformed GraphicsCommand control block, number is too large') + return + result[full_key] = val + + if i < n and cmd_str[i] == ord(','): + i += 1 + + self._emit(cb, 'graphics_command', result) + + # -- ESC dispatch -- + + def _dispatch_esc(self, ch, cb): + if ch == ord('D'): + self._emit(cb, 'screen_index') + elif ch == ord('E'): + self._emit(cb, 'screen_nel') + elif ch == ord('c'): + self._emit(cb, 'screen_reset') + elif ch == ord('M'): + self._emit(cb, 'screen_reverse_index') + elif ch == ord('7'): + self._emit(cb, 'screen_save_cursor') + elif ch == ord('8'): + self._emit(cb, 'screen_restore_cursor') + elif ch == ord('H'): + self._emit(cb, 'screen_set_tab') + elif ch == ord('['): + pass # handled by state transition + elif ch == ord(']'): + pass # handled by state transition + elif ch == ord('P'): + pass # handled by state transition + elif ch == ord('_'): + pass # handled by state transition + elif ch == ord('^'): + pass # handled by state transition + elif ch == ord('X'): + pass # handled by state transition + elif ch == ord('(') or ch == ord(')'): + pass # handled by state transition + elif ch == ord('='): + self._emit(cb, 'screen_application_keypad', True) + elif ch == ord('>'): + self._emit(cb, 'screen_application_keypad', False) + elif ch == ord('.'): + # Unhandled charset escape — needs the next ESC byte + pass + else: + self._emit(cb, 'Unknown char after ESC: 0x%x' % ch) + + # -- Main parse loop -- + + def parse(self, screen, data, cb): + """Parse *data*, emitting events to *cb* and feeding bytes to screen.""" + if isinstance(data, str): + data = data.encode('utf-8') + data = bytes(data) + + # Feed ALL bytes to the C terminal for screen state + screen.write(data) + + i = 0 + n = len(data) + while i < n: + b = data[i] + + if self.state == _GROUND: + if b == 0x1b: + self._kill_pending() + self._flush_text(cb) + self.state = _ESC + i += 1 + elif b < 0x20: + # C0 control + self._kill_pending() + self._flush_text(cb) + self._control_event(cb, b) + # Other C0 controls: no event + i += 1 + else: + # Text byte — decode UTF-8 + consumed = self._decode_utf8(data[i:], cb) + i += consumed + + elif self.state == _ESC: + if b == 0x1b: + # ESC after ESC + self._emit(cb, 'Unknown char after ESC: 0x1b') + # Stay in ESC state + i += 1 + elif b < 0x20: + # C0 control after ESC — execute it + self._flush_text(cb) + self._control_event(cb, b) + self.state = _GROUND + i += 1 + elif b == ord('['): + self.csi.reset() + self.state = _CSI + i += 1 + elif b == ord(']'): + self.str_buf.clear() + self.str_type = _OSC + self.state = _OSC + i += 1 + elif b == ord('P'): + self.str_buf.clear() + self.str_type = _DCS + self.state = _DCS + i += 1 + elif b == ord('_'): + self.str_buf.clear() + self.str_type = _APC + self.state = _APC + i += 1 + elif b == ord('^'): + self.str_buf.clear() + self.str_type = _PM + self.state = _PM + i += 1 + elif b == ord('X'): + self.str_buf.clear() + self.str_type = _SOS + self.state = _SOS + i += 1 + elif b == ord('(') or b == ord(')'): + self.charset_slot = 0 if b == ord('(') else 1 + self.state = _CHARSET + i += 1 + elif b == ord('.'): + # Unhandled charset escape — needs the next ESC byte + # Emit special message when we see the next ESC + self.state = _CHARSET + self.charset_slot = -1 # special marker + i += 1 + elif b == ord('%'): + # ESC % — codepage selector; needs one more byte + self.state = _ESC_PCT + i += 1 + else: + self._dispatch_esc(b, cb) + self.state = _GROUND + i += 1 + + elif self.state == _ESC_PCT: + self.state = _GROUND + if b in (ord('G'), ord('@'), ord('/')): + # DOCS selectors (ESC % G etc.) — consume silently + i += 1 + else: + # Unknown — kitty draws the sequence literally + self.text += '%' + # Reprocess this byte in GROUND (don't advance) + + elif self.state == _CHARSET: + if self.charset_slot == -1: + # Unhandled charset escape (ESC .) + if b == 0x1b: + self._emit(cb, 'Unhandled charset related escape code: 0x2e 0x1b') + self.state = _GROUND + else: + self._emit(cb, 'Unhandled charset related escape code: 0x2e 0x%02x' % b) + self.state = _GROUND + i += 1 + else: + self._emit(cb, 'screen_designate_charset', self.charset_slot, b) + if self.charset_slot == 1: + # G1 charset — also emit screen_change_charset + self._emit(cb, 'screen_change_charset', 1) + self.state = _GROUND + i += 1 + + elif self.state == _CSI: + c = self.csi + if b == 0x1b: + # ESC interrupts CSI + self._flush_text(cb) + self.state = _ESC + i += 1 + elif b < 0x20: + # C0 control embedded inside CSI — dispatch, keep parsing + self._flush_text(cb) + self._control_event(cb, b) + i += 1 + elif c.phase == 2: + # Post-secondary: only a final byte is valid + if 0x40 <= b <= 0x7e: + c.final = b + self._dispatch_csi(screen, cb) + else: + self._emit(cb, 'Invalid character in CSI: %s (0x%x), ignoring the sequence' + % (chr(b) if 0x21 <= b <= 0x7e else '0x%x' % b, b)) + self.state = _GROUND + i += 1 + elif c.phase == 0: + if b == ord(';'): + if c.nparams < _MAX_PARAMS: + c.params[c.nparams] = 0 + c.nparams += 1 + c.phase = 1 + elif 0x30 <= b <= 0x39: + c.add_digit(b - 0x30) + c.phase = 1 + elif b in (ord('?'), ord('>'), ord('<'), ord('=')): + c.prefix = b + c.phase = 1 + elif 0x20 <= b <= 0x2f: + if b == ord('-'): + c.mult = -1 + c.num_digits = 1 + c.phase = 1 + else: + c.intermediate = b + c.phase = 2 + elif 0x40 <= b <= 0x7e: + c.final = b + self._dispatch_csi(screen, cb) + self.state = _GROUND + else: + self._emit(cb, 'Invalid character in CSI: %s (0x%x), ignoring the sequence' + % (chr(b) if 0x21 <= b <= 0x7e else '0x%x' % b, b)) + self.state = _GROUND + i += 1 + else: + # CSI_BODY + if 0x30 <= b <= 0x39: + c.add_digit(b - 0x30) + elif b == ord(';'): + if not c.num_digits: + c.num_digits = 1 # empty means zero + if not c.commit(): + self._emit(cb, 'CSI escape code has too many parameters, ignoring it') + self.state = _GROUND + elif b == ord(':'): + if c.commit(): + c.subparams |= 1 << c.nparams + else: + self._emit(cb, 'CSI escape code has too many parameters, ignoring it') + self.state = _GROUND + elif 0x20 <= b <= 0x2f: + if b == ord('-') and c.num_digits == 0: + c.mult = -1 + c.num_digits = 1 + elif c.commit(): + c.intermediate = b + c.phase = 2 + else: + self._emit(cb, 'CSI escape code has too many parameters, ignoring it') + self.state = _GROUND + elif 0x40 <= b <= 0x7e: + if c.num_digits == 1 and c.intermediate == 0 and c.mult == -1: + # A bare '-' before the final is an intermediate + c.num_digits = 0 + c.intermediate = ord('-') + if c.commit(): + c.final = b + self._dispatch_csi(screen, cb) + else: + self._emit(cb, 'CSI escape code has too many parameters, ignoring it') + self.state = _GROUND + else: + self._emit(cb, 'Invalid character in CSI: %s (0x%x), ignoring the sequence' + % (chr(b) if 0x21 <= b <= 0x7e else '0x%x' % b, b)) + self.state = _GROUND + i += 1 + + elif self.state in (_OSC, _DCS, _APC, _PM, _SOS): + if b == 0x07: + # BEL terminates string + self._finish_string(screen, cb) + self.state = _GROUND + i += 1 + elif b == 0x1b: + self.state = _STR_ESC + i += 1 + elif b < 0x20: + # C0 control — abort string (except for BEL handled above) + self.str_buf.clear() + self.state = _GROUND + # Process the control byte in GROUND + self._control_event(cb, b) + i += 1 + else: + self.str_buf.append(b) + i += 1 + + elif self.state == _STR_ESC: + if b == ord('\\'): + # ST (ESC \) terminates string + self._finish_string(screen, cb) + self.state = _GROUND + i += 1 + else: + # Other byte after ESC in string — reprocess + self.str_buf.clear() + self.state = _GROUND + # Don't consume — reprocess in GROUND/ESC + # Actually, the ESC should be reprocessed + self.state = _ESC + i += 0 # Don't advance — reprocess this byte via ESC + # But we already consumed the ESC, so process this byte in ESC state + if b == 0x1b: + self._emit(cb, 'Unknown char after ESC: 0x1b') + # Stay in ESC + elif b < 0x20: + self._control_event(cb, b) + self.state = _GROUND + else: + self._dispatch_esc(b, cb) + self.state = _GROUND + i += 1 + + # End of data — flush any pending text + self._flush_text(cb) + + def _finish_string(self, screen, cb): + """Dispatch the completed string (OSC/DCS/APC/PM/SOS).""" + st = self.str_type + if st == _OSC: + self._dispatch_osc(screen, cb) + elif st == _DCS: + self._dispatch_dcs(screen, cb) + elif st == _APC: + # Check for graphics command (_G) + payload = bytes(self.str_buf) + if payload.startswith(b'G'): + self.str_buf = bytearray(payload[1:]) + self._dispatch_graphics(screen, cb) + else: + code = payload[0] if payload else 0 + self._emit(cb, 'Unrecognized APC code: 0x%02x' % code) + elif st == _PM: + code = self.str_buf[0] if self.str_buf else 0 + self._emit(cb, 'Unrecognized PM code: 0x%02x' % code) + elif st == _SOS: + code = self.str_buf[0] if self.str_buf else 0 + self._emit(cb, 'Unrecognized SOS code: 0x%02x' % code) + self.str_buf.clear() + + +# Module-level parse_bytes function +def parse_bytes(screen, data, callback=None): + """Parse *data* and emit events to *callback*. + + Maintains parser state on the screen object so partial sequences + split across ``parse_bytes`` invocations are handled correctly. + Also updates screen.callbacks for OSC events (titlebuf, etc.). + """ + if callback is None: + callback = lambda *args: None + if not hasattr(screen, '_kitty_parser'): + screen._kitty_parser = KittyParser() + + def wrapped_cb(wid, *args): + # Update screen callbacks for OSC/DCS events + if args and args[0] == 'set_title': + screen.callbacks.titlebuf.append(args[1]) + elif args and args[0] == 'set_icon': + screen.callbacks.iconbuf += args[1] if isinstance(args[1], str) else str(args[1], 'utf-8') + elif args and args[0] == 'desktop_notify': + screen.callbacks.notifications.append((args[1], args[2])) + elif args and args[0] == 'set_dynamic_color': + val = args[2] if isinstance(args[2], str) else str(args[2], 'utf-8') + screen.callbacks.colorbuf += val + elif args and args[0] == 'handle_remote_print': + try: + from fast_data_types import base64_decode + screen.callbacks.printbuf.append(base64_decode(args[1]).decode('utf-8')) + except Exception: + screen.callbacks.printbuf.append(args[1]) + # Forward to original callback + callback(wid, *args) + + screen._kitty_parser.parse(screen, data, wrapped_cb) diff --git a/tests/extern/kitty/parser.py b/tests/extern/kitty/parser.py new file mode 100644 index 0000000..537e1c0 --- /dev/null +++ b/tests/extern/kitty/parser.py @@ -0,0 +1,995 @@ +#!/usr/bin/env python +# License: GPL v3 Copyright: 2016, Kovid Goyal <kovid at kovidgoyal.net> + +from binascii import hexlify +from functools import partial + +from fast_data_types import ( + CURSOR_BLOCK, + VT_PARSER_BUFFER_SIZE, + base64_decode, + base64_encode, + has_avx2, + has_sse4_2, + test_find_either_of_two_bytes, + test_utf8_decode_to_sentinel, +) + +from kitty_harness import BaseTest, parse_bytes + + +def cnv(x): + if isinstance(x, memoryview): + x = str(x, 'utf-8') + return x + + +class CmdDump(list): + + def __call__(self, window_id, *a): + if a and a[0] == 'bytes': + return + if a and a[0] == 'error': + a = a[1:] + self.append(tuple(map(cnv, a))) + + def get_result(self): + current = '' + q = [] + for args in self: + if args[0] == 'draw': + current += args[1] + else: + if current: + q.append(('draw', current)) + current = '' + q.append(args) + if current: + q.append(('draw', current)) + return tuple(q) + + +class TestParser(BaseTest): + + def create_write_buffer(self, screen): + return screen.test_create_write_buffer() + + def write_bytes(self, screen, write_buf, data): + if isinstance(data, str): + data = data.encode('utf-8') + s = screen.test_commit_write_buffer(data, write_buf) + return data[s:] + + def parse_written_data(self, screen, *cmds): + cd = CmdDump() + screen.test_parse_written_data(cd) + cmds = tuple(('draw', x) if isinstance(x, str) else tuple(map(cnv, x)) for x in cmds) + self.ae(cmds, cd.get_result()) + + def parse_bytes_dump(self, s, x, *cmds): + cd = CmdDump() + if isinstance(x, str): + x = x.encode('utf-8') + cmds = tuple(('draw', x) if isinstance(x, str) else tuple(map(cnv, x)) for x in cmds) + parse_bytes(s, x, cd) + self.ae(cmds, cd.get_result()) + + def test_charsets(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb(b'\xc3') + pb(b'\xa1', ('draw', b'\xc3\xa1'.decode('utf-8'))) + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb('\033)0\x0e/_', ('screen_designate_charset', 1, ord('0')), ('screen_change_charset', 1), '/_') + self.ae(str(s.line(0)), '/\xa0') + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb('\033(0/_', ('screen_designate_charset', 0, ord('0')), '/_') + self.ae(str(s.line(0)), '/\xa0') + + def test_parser_threading(self): + s = self.create_screen() + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), 'a\x1b]2;some title')) + b = self.create_write_buffer(s) + self.parse_written_data(s, 'a') + self.assertFalse(self.write_bytes(s, b, ' full\x1b\\')) + self.parse_written_data(s, ('set_title', 'some title full')) + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), 'a\x1b]')) + b = self.create_write_buffer(s) + self.parse_written_data(s, 'a') + self.assertFalse(self.write_bytes(s, b, '2;title\x1b\\')) + self.parse_written_data(s, ('set_title', 'title')) + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), 'a\x1b')) + b = self.create_write_buffer(s) + self.parse_written_data(s, 'a') + self.assertFalse(self.write_bytes(s, b, ']2;title\x1b\\')) + self.parse_written_data(s, ('set_title', 'title')) + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), 'a\x1b]2;some title\x1b')) + b = self.create_write_buffer(s) + self.parse_written_data(s, 'a') + self.assertFalse(self.write_bytes(s, b, '\\b')) + self.parse_written_data(s, ('set_title', 'some title'), 'b') + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '1\x1b')) + self.parse_written_data(s, '1') + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), 'E2')) + self.parse_written_data(s, ('screen_nel',), ('draw', '2')) + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '1\x1b[2')) + self.parse_written_data(s, '1') + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '3mx')) + self.parse_written_data(s, ('select_graphic_rendition', '23'), 'x') + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '1\x1b')) + self.parse_written_data(s, '1') + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '[23mx')) + self.parse_written_data(s, ('select_graphic_rendition', '23'), 'x') + + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '1\x1b[')) + self.parse_written_data(s, '1') + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), '23mx')) + self.parse_written_data(s, ('select_graphic_rendition', '23'), 'x') + + # test full write + sz = VT_PARSER_BUFFER_SIZE // 3 + 7 + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), b'a' * sz)) + self.assertFalse(self.write_bytes(s, self.create_write_buffer(s), b'b' * sz)) + left = self.write_bytes(s, self.create_write_buffer(s), b'c' * sz) + self.assertTrue(len(left), 3 * sz - VT_PARSER_BUFFER_SIZE) + self.assertFalse(self.create_write_buffer(s)) + s.test_parse_written_data() + b = self.create_write_buffer(s) + self.assertTrue(b) + self.write_bytes(s, b, b'') + + def test_base64(self): + for src, expected in { + 'bGlnaHQgdw==': 'light w', + 'bGlnaHQgd28=': 'light wo', + 'bGlnaHQgd29y': 'light wor', + }.items(): + self.ae(base64_decode(src.encode()), expected.encode(), f'Decoding of {src} failed') + self.ae(base64_decode(src.replace('=', '').encode()), expected.encode(), f'Decoding of {src} failed') + self.ae(base64_encode(expected.encode()), src.replace('=', '').encode(), f'Encoding of {expected} failed') + + def test_simple_parsing(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + + pb('12', '12') + self.ae(str(s.line(0)), '12') + self.ae(s.cursor.x, 2) + pb('3456', '3456') + self.ae(str(s.line(0)), '12345') + self.ae(str(s.line(1)), '6') + pb(b'\n123\n\r45', ('screen_linefeed',), '123', ('screen_linefeed',), ('screen_carriage_return',), '45') + self.ae(str(s.line(1)), '6') + self.ae(str(s.line(2)), ' 123') + self.ae(str(s.line(3)), '45') + pb(b'\rabcde', ('screen_carriage_return',), 'abcde') + self.ae(str(s.line(3)), 'abcde') + pb('\rßxyz1', ('screen_carriage_return',), 'ßxyz1') + self.ae(str(s.line(3)), 'ßxyz1') + pb('ニチ ', 'ニチ ') + self.ae(str(s.line(4)), 'ニチ ') + s.reset() + self.assertFalse(str(s.line(1)) + str(s.line(2)) + str(s.line(3))) + c1_controls = '\x84\x85\x88\x8d\x8e\x8f\x90\x96\x97\x98\x9a\x9b\x9c\x9d\x9e\x9f' + pb(c1_controls, c1_controls) + self.assertFalse(str(s.line(1)) + str(s.line(2)) + str(s.line(3))) + pb('😀'.encode()[:-1]) + pb('\x1b\x1b%a', '\ufffd', ('Unknown char after ESC: 0x1b',), ('draw', '%a')) + + def test_utf8_simd_decode(self): + def unsupported(which): + return (which == 2 and not has_sse4_2) or (which == 3 and not has_avx2) + + def reset_state(): + test_utf8_decode_to_sentinel(b'', -1) + + def asbytes(x): + if isinstance(x, str): + x = x.encode() + return x + + def t(*a, which=2): + if unsupported(which): + return + + def parse_parts(which): + total_consumed = 0 + esc_found = False + parts = [] + for x in a: + found_sentinel, x, num_consumed = test_utf8_decode_to_sentinel(asbytes(x), which) + total_consumed += num_consumed + if found_sentinel: + esc_found = found_sentinel + parts.append(x) + return esc_found, ''.join(parts), total_consumed + + reset_state() + expected = parse_parts(1) + reset_state() + actual = parse_parts(which) + self.ae(expected, actual, msg=f'Failed for {a} with {which=}\n{expected!r} !=\n{actual!r}') + return actual + + def double_test(x): + for which in (2, 3): + t(x, which=which) + t(x*2, which=3) + reset_state() + + # incomplete trailer at end of vector + t("a"*10 + "😸😸" + "b"*15) + + x = double_test + x('2:α3') + x('2:α\x1b3') + x('2:α3:≤4:😸|') + x('abcd1234efgh5678') + x('abc\x1bd1234efgh5678') + x('abcd1234efgh5678ijklABCDmnopEFGH') + + for which in (2, 3): + x = partial(t, which=which) + x('abcdef', 'ghijk') + x('2:α3', ':≤4:😸|') + # trailing incomplete sequence + for prefix in (b'abcd', '😸'.encode()): + for suffix in (b'1234', '😸'.encode()): + x(prefix + b'\xf0\x9f', b'\x98\xb8' + suffix) + x(prefix + b'\xf0\x9f\x9b', b'\xb8' + suffix) + x(prefix + b'\xf0', b'\x9f\x98\xb8' + suffix) + x(prefix + b'\xc3', b'\xa4' + suffix) + x(prefix + b'\xe2', b'\x89\xa4' + suffix) + x(prefix + b'\xe2\x89', b'\xa4' + suffix) + + def test_expected(src, expected, which=2): + if unsupported(which): + return + reset_state() + _, actual, _ = t(b'filler' + asbytes(src), which=which) + expected = 'filler' + expected + self.ae(expected, actual, f'Failed for: {src!r} with {which=}') + + for which in (1, 2, 3): + pb = partial(test_expected, which=which) + pb('ニチ', 'ニチ') + pb('\x84\x85', '\x84\x85') + pb('\x84\x85', '\x84\x85') + pb('\uf4df', '\uf4df') + pb('\uffff', '\uffff') + pb('\0', '\0') + pb(chr(0x10ffff), chr(0x10ffff)) + # Kitty's UTF-8 decoding uses `U+FFFD substitution of maximal subparts + # <https://www.unicode.org/versions/Unicode16.0.0/core-spec/chapter-3/#G66453>`_, + # same as in the WHATWG Encoding Standard. + # This means that ill-formed sequences may be replaced by multiple + # U+FFFD REPLACEMENT CHARACTERs. + pb(b'abcd\xf51234', 'abcd\ufffd1234') # bytes > 0xf4 + pb(b'abcd\xff1234', 'abcd\ufffd1234') # bytes > 0xf4 + pb(b'"\xbf"', '"\ufffd"') + pb(b'"\x80"', '"\ufffd"') + pb(b'"\x80\xbf"', '"\ufffd\ufffd"') + pb(b'"\x80\xbf\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xc0 "', '"\ufffd "') + pb(b'"\xfe"', '"\ufffd"') + pb(b'"\xff"', '"\ufffd"') + pb(b'"\xff\xfe"', '"\ufffd\ufffd"') + pb(b'"\xfe\xfe\xff\xff"', '"\ufffd\ufffd\ufffd\ufffd"') + pb(b'"\xef\xbf"', '"\ufffd"') + pb(b'"\xe0\xa0"', '"\ufffd"') + pb(b'"\xf0\x9f\x98"', '"\ufffd"') + pb(b'"\xef\x93\x94\x95"', '"\uf4d4\ufffd"') + + # Lone continuation bytes with no leading starts + pb(b'"\xbf"', '"\ufffd"') + pb(b'"\x80"', '"\ufffd"') + + # Multiple lone continuation bytes + pb(b'"\x80\xbf"', '"\ufffd\ufffd"') + pb(b'"\x80\xbf\x80"', '"\ufffd\ufffd\ufffd"') + + # Lone starter byte of 2-byte sequence + pb(b'"\xc0 "', '"\ufffd "') + + # Single never-valid bytes + pb(b'"\xfe"', '"\ufffd"') + pb(b'"\xff"', '"\ufffd"') + + # Multiple never-valid bytes + pb(b'"\xff\xfe"', '"\ufffd\ufffd"') + pb(b'"\xfe\xfe\xff\xff"', '"\ufffd\ufffd\ufffd\ufffd"') + + # Truncated 2-byte sequence (only 1 byte) + pb(b'"\xc2"', '"\ufffd"') + + # Truncated 3-byte sequences (only 2 bytes) + pb(b'"\xef\xbf"', '"\ufffd"') + pb(b'"\xe0\xa0"', '"\ufffd"') + + # Truncated 4-byte sequence (only 2 or 3 bytes) + pb(b'"\xf0\x9f"', '"\ufffd"') + pb(b'"\xf0\x9f\x98"', '"\ufffd"') + + # Bad continuation byte (restored as ASCII) + pb(b'"\xe1\x28\xa1"', '"\ufffd(\ufffd"') # ) + + # Overlong 2-byte sequence for U+0000 (should be `0x00`) + pb(b'"\xc0\x80"', '"\ufffd\ufffd"') + + # Overlong 3-byte sequence for U+0000 (violates boundary) + pb(b'"\xe0\x80\x80"', '"\ufffd\ufffd\ufffd"') + + # Overlong 4-byte sequence for U+0000 (violates boundary) + pb(b'"\xf0\x80\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + + # High surrogate code point + pb(b'"\xed\xa0\x80"', '"\ufffd\ufffd\ufffd"') + + # Low surrogate code point + pb(b'"\xed\xb0\x80"', '"\ufffd\ufffd\ufffd"') + + # Too large starter byte + pb(b'"\xff\x80\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + + # The following boundary cases come from the table of well-formed UTF-8 byte sequences + # <https://www.unicode.org/versions/Unicode16.0.0/core-spec/chapter-3/#G27506>`_. + # For continuation bytes, both 0xC0 and 0xC2 are tested as values that exceed the valid maximum. + # This is because 0xC0 is an invalid starter byte, but 0xC2 is also a starter byte for 2-byte sequences. + # simd-string-impl.h prefers classifying bytes as starter bytes when possible (e.g., in "\xf0\x90\xc2\x80"). + # The tests need to check that simd-string-impl.h correctly detects + # starter bytes that are actually invalid continution bytes, like 0xC2. + + # Boundary cases: 2-byte sequences + pb(b'"\xc1\x7f"', '"\ufffd\x7f"') + pb(b'"\xc1\x80"', '"\ufffd\ufffd"') + pb(b'"\xc1\xbf"', '"\ufffd\ufffd"') + pb(b'"\xc1\xc0"', '"\ufffd\ufffd"') + pb(b'"\xc1\xc2"', '"\ufffd\ufffd"') + pb(b'"\xc2\x7f"', '"\ufffd\x7f"') + pb(b'"\xc2\x80"', '"\x80"') + pb(b'"\xc2\xbf"', '"\xbf"') + pb(b'"\xc2\xc0"', '"\ufffd\ufffd"') + pb(b'"\xc2\xc2"', '"\ufffd\ufffd"') + pb(b'"\xdf\x7f"', '"\ufffd\x7f"') + pb(b'"\xdf\x80"', '"\u07c0"') + pb(b'"\xdf\xbf"', '"\u07ff"') + pb(b'"\xdf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xdf\xc2"', '"\ufffd\ufffd"') + + # Boundary cases: 3-byte sequences starting with 0xE0 + pb(b'"\xe0\x9f\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xe0\xa0\x7f"', '"\ufffd\x7f"') + pb(b'"\xe0\xa0\x80"', '"\u0800"') + pb(b'"\xe0\xa0\xbf"', '"\u083f"') + pb(b'"\xe0\xa0\xc0"', '"\ufffd\ufffd"') + pb(b'"\xe0\xa0\xc2"', '"\ufffd\ufffd"') + pb(b'"\xe0\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xe0\xbf\x80"', '"\u0fc0"') + pb(b'"\xe0\xbf\xbf"', '"\u0fff"') + pb(b'"\xe0\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xe0\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xe0\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xe0\xc2\x80"', '"\ufffd\x80"') + + # Boundary cases: 3-byte sequences starting with 0xE1..0xEC + pb(b'"\xe1\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xe1\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xe1\x80\x80"', '"\u1000"') + pb(b'"\xe1\x80\xbf"', '"\u103f"') + pb(b'"\xe1\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xe1\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xe1\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xe1\xbf\x80"', '"\u1fc0"') + pb(b'"\xe1\xbf\xbf"', '"\u1fff"') + pb(b'"\xe1\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xe1\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xe1\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xe1\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xec\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xec\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xec\x80\x80"', '"\uc000"') + pb(b'"\xec\x80\xbf"', '"\uc03f"') + pb(b'"\xec\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xec\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xec\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xec\xbf\x80"', '"\ucfc0"') + pb(b'"\xec\xbf\xbf"', '"\ucfff"') + pb(b'"\xec\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xec\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xec\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xec\xc2\x80"', '"\ufffd\x80"') + + # Boundary cases: 3-byte sequences starting with 0xED + pb(b'"\xed\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xed\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xed\x80\x80"', '"\ud000"') + pb(b'"\xed\x80\xbf"', '"\ud03f"') + pb(b'"\xed\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xed\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xed\x9f\x7f"', '"\ufffd\x7f"') + pb(b'"\xed\x9f\x80"', '"\ud7c0"') + pb(b'"\xed\x9f\xbf"', '"\ud7ff"') + pb(b'"\xed\x9f\xc0"', '"\ufffd\ufffd"') + pb(b'"\xed\x9f\xc2"', '"\ufffd\ufffd"') + + # Boundary cases: 3-byte sequences starting with 0xEE..0xEF + pb(b'"\xed\xa0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xee\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xee\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xee\x80\x80"', '"\ue000"') + pb(b'"\xee\x80\xbf"', '"\ue03f"') + pb(b'"\xee\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xee\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xee\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xee\xbf\x80"', '"\uefc0"') + pb(b'"\xee\xbf\xbf"', '"\uefff"') + pb(b'"\xee\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xee\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xee\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xee\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xef\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xef\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xef\x80\x80"', '"\uf000"') + pb(b'"\xef\x80\xbf"', '"\uf03f"') + pb(b'"\xef\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xef\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xef\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xef\xbf\x80"', '"\uffc0"') + pb(b'"\xef\xbf\xbf"', '"\uffff"') + pb(b'"\xef\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xef\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xef\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xef\xc2\x80"', '"\ufffd\x80"') + + # Boundary cases: 4-byte sequences starting with 0xF0 + pb(b'"\xf0\x8f\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + pb(b'"\xf0\x90\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf0\x90\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf0\x90\x80\x80"', '"\U00010000"') + pb(b'"\xf0\x90\x80\xbf"', '"\U0001003f"') + pb(b'"\xf0\x90\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf0\x90\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf0\x90\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf0\x90\xbf\x80"', '"\U00010fc0"') + pb(b'"\xf0\x90\xbf\xbf"', '"\U00010fff"') + pb(b'"\xf0\x90\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf0\x90\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf0\x90\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf0\x90\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\x90\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\x90\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\x90\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\x90\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf0\x90\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf0\x90\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf0\x90\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\x90\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xbf\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf0\xbf\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf0\xbf\x80\x80"', '"\U0003f000"') + pb(b'"\xf0\xbf\x80\xbf"', '"\U0003f03f"') + pb(b'"\xf0\xbf\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf0\xbf\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf0\xbf\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf0\xbf\xbf\x80"', '"\U0003ffc0"') + pb(b'"\xf0\xbf\xbf\xbf"', '"\U0003ffff"') + pb(b'"\xf0\xbf\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf0\xbf\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf0\xbf\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf0\xbf\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xbf\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xbf\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xbf\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xbf\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf0\xbf\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf0\xbf\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf0\xbf\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xbf\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf0\xc0\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + + # Boundary cases: 4-byte sequences starting with 0xF1..0xF3 + pb(b'"\xf1\x7f\x80\x80"', '"\ufffd\x7f\ufffd\ufffd"') + pb(b'"\xf1\x80\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf1\x80\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf1\x80\x80\x80"', '"\U00040000"') + pb(b'"\xf1\x80\x80\xbf"', '"\U0004003f"') + pb(b'"\xf1\x80\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf1\x80\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf1\x80\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf1\x80\xbf\x80"', '"\U00040fc0"') + pb(b'"\xf1\x80\xbf\xbf"', '"\U00040fff"') + pb(b'"\xf1\x80\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf1\x80\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf1\x80\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf1\x80\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\x80\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\x80\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\x80\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf1\x80\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf1\x80\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf1\x80\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xbf\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf1\xbf\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf1\xbf\x80\x80"', '"\U0007f000"') + pb(b'"\xf1\xbf\x80\xbf"', '"\U0007f03f"') + pb(b'"\xf1\xbf\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf1\xbf\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf1\xbf\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf1\xbf\xbf\x80"', '"\U0007ffc0"') + pb(b'"\xf1\xbf\xbf\xbf"', '"\U0007ffff"') + pb(b'"\xf1\xbf\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf1\xbf\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf1\xbf\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf1\xbf\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xbf\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xbf\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xbf\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xbf\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf1\xbf\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf1\xbf\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf1\xbf\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xbf\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xc0\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + pb(b'"\xf1\xc2\x80\x80"', '"\ufffd\x80\ufffd"') + pb(b'"\xf3\x7f\x80\x80"', '"\ufffd\x7f\ufffd\ufffd"') + pb(b'"\xf3\x80\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf3\x80\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf3\x80\x80\x80"', '"\U000c0000"') + pb(b'"\xf3\x80\x80\xbf"', '"\U000c003f"') + pb(b'"\xf3\x80\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf3\x80\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf3\x80\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf3\x80\xbf\x80"', '"\U000c0fc0"') + pb(b'"\xf3\x80\xbf\xbf"', '"\U000c0fff"') + pb(b'"\xf3\x80\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf3\x80\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf3\x80\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf3\x80\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\x80\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\x80\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\x80\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\x80\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf3\x80\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf3\x80\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf3\x80\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\x80\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xbf\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf3\xbf\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf3\xbf\x80\x80"', '"\U000ff000"') + pb(b'"\xf3\xbf\x80\xbf"', '"\U000ff03f"') + pb(b'"\xf3\xbf\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf3\xbf\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf3\xbf\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf3\xbf\xbf\x80"', '"\U000fffc0"') + pb(b'"\xf3\xbf\xbf\xbf"', '"\U000fffff"') + pb(b'"\xf3\xbf\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf3\xbf\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf3\xbf\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf3\xbf\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xbf\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xbf\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xbf\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xbf\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf3\xbf\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf3\xbf\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf3\xbf\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xbf\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xc0\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + pb(b'"\xf3\xc2\x80\x80"', '"\ufffd\x80\ufffd"') + + # Boundary cases: 4-byte sequences starting with 0xF4 + pb(b'"\xf4\x7f\x80\x80"', '"\ufffd\x7f\ufffd\ufffd"') + pb(b'"\xf4\x80\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf4\x80\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf4\x80\x80\x80"', '"\U00100000"') + pb(b'"\xf4\x80\x80\xbf"', '"\U0010003f"') + pb(b'"\xf4\x80\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf4\x80\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf4\x80\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf4\x80\xbf\x80"', '"\U00100fc0"') + pb(b'"\xf4\x80\xbf\xbf"', '"\U00100fff"') + pb(b'"\xf4\x80\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf4\x80\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf4\x80\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf4\x80\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x80\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x80\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x80\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x80\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf4\x80\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf4\x80\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf4\x80\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x80\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x8f\x7f\x80"', '"\ufffd\x7f\ufffd"') + pb(b'"\xf4\x8f\x80\x7f"', '"\ufffd\x7f"') + pb(b'"\xf4\x8f\x80\x80"', '"\U0010f000"') + pb(b'"\xf4\x8f\x80\xbf"', '"\U0010f03f"') + pb(b'"\xf4\x8f\x80\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf4\x8f\x80\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf4\x8f\xbf\x7f"', '"\ufffd\x7f"') + pb(b'"\xf4\x8f\xbf\x80"', '"\U0010ffc0"') + pb(b'"\xf4\x8f\xbf\xbf"', '"\U0010ffff"') + pb(b'"\xf4\x8f\xbf\xc0"', '"\ufffd\ufffd"') + pb(b'"\xf4\x8f\xbf\xc2"', '"\ufffd\ufffd"') + pb(b'"\xf4\x8f\xc0\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf4\x8f\xc0\x80"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x8f\xc0\xbf"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x8f\xc0\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x8f\xc0\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x8f\xc2\x7f"', '"\ufffd\ufffd\x7f"') + pb(b'"\xf4\x8f\xc2\x80"', '"\ufffd\x80"') + pb(b'"\xf4\x8f\xc2\xbf"', '"\ufffd\xbf"') + pb(b'"\xf4\x8f\xc2\xc0"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x8f\xc2\xc2"', '"\ufffd\ufffd\ufffd"') + pb(b'"\xf4\x90\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + pb(b'"\xf5\x80\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + + # Boundary case: too large codepoint (> U+10FFFF) + pb(b'"\xf5\x80\x80\x80"', '"\ufffd\ufffd\ufffd\ufffd"') + + + def test_find_either_of_two_bytes(self): + sizes = [] + if has_sse4_2: + sizes.append(2) + if has_avx2: + sizes.append(3) + sizes.append(0) + + def test(buf, a, b, align_offset=0): + a_, b_ = ord(a), ord(b) + expected = test_find_either_of_two_bytes(buf, a_, b_, 1, 0) + for sz in sizes: + actual = test_find_either_of_two_bytes(buf, a_, b_, sz, align_offset) + self.ae(expected, actual, f'Failed for: {buf!r} {a=} {b=} at {sz=} and {align_offset=}') + + q = 'abc' + for off in range(32): + test(q, '<', '>', off) + test(q, ' ', 'b', off) + test(q, '<', 'a', off) + test(q, '<', 'b', off) + test(q, 'c', '>', off) + + def tests(buf, a, b): + for sz in (0, 16, 32, 64, 79): + buf = (' ' * sz) + buf + for align_offset in range(32): + test(buf, a, b, align_offset) + tests("", '<', '>') + tests("a", '\0', '\0') + tests("a", '<', '>') + tests("dsdfsfa", '1', 'a') + tests("xa", 'a', 'a') + tests("bbb", 'a', '1') + tests("bba", 'a', '<') + tests("baa", '>', 'a') + + def test_esc_codes(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb('12\033Da', '12', ('screen_index',), 'a') + self.ae(str(s.line(0)), '12') + self.ae(str(s.line(1)), ' a') + pb('\033xa', ('Unknown char after ESC: 0x%x' % ord('x'),), 'a') + pb('\033c123', ('screen_reset', ), '123') + self.ae(str(s.line(0)), '123') + pb('\033.\033a', ('Unhandled charset related escape code: 0x2e 0x1b',), 'a') + + def test_csi_codes(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb('abcde', 'abcde') + s.cursor_move(5) + pb('x\033[2@y', 'x', ('screen_insert_characters', 2), 'y') + self.ae(str(s.line(0)), 'xy bc') + pb('x\033[2;7@y', 'x', ('CSI code @ has 2 > 1 parameters',), 'y') + pb('x\033[2;-7@y', 'x', ('CSI code @ has 2 > 1 parameters',), 'y') + pb('x\033[-0001234567890@y', 'x', ('CSI code @ is not allowed to have negative parameter (-1234567890)',), 'y') + pb('x\033[2-3@y', 'x', ('Invalid character in CSI: 3 (0x33), ignoring the sequence',), '@y') + pb('x\033[@y', 'x', ('screen_insert_characters', 1), 'y') + pb('x\033[345@y', 'x', ('screen_insert_characters', 345), 'y') + pb('x\033[345;@y', 'x', ('screen_insert_characters', 345), 'y') + pb('\033[H', ('screen_cursor_position', 1, 1)) + self.ae(s.cursor.x, 0), self.ae(s.cursor.y, 0) + pb('\033[4H', ('screen_cursor_position', 4, 1)) + pb('\033[4;0H', ('screen_cursor_position', 4, 0)) + pb('\033[3;2H', ('screen_cursor_position', 3, 2)) + pb('\033[3;2;H', ('screen_cursor_position', 3, 2)) + pb('\033[00000000003;0000000000000002H', ('screen_cursor_position', 3, 2)) + self.ae(s.cursor.x, 1), self.ae(s.cursor.y, 2) + pb('\033[0001234567890H', ('screen_cursor_position', 1234567890, 1)) + pb('\033[J', ('screen_erase_in_display', 0, 0)) + pb('\033[?J', ('screen_erase_in_display', 0, 1)) + pb('\033[?2J', ('screen_erase_in_display', 2, 1)) + pb('\033[h') + pb('\033[20;4h', ('screen_set_mode', 20, 0), ('screen_set_mode', 4, 0)) + pb('\033[?1000;1004h', ('screen_set_mode', 1000, 1), ('screen_set_mode', 1004, 1)) + pb('\033[20;4;20l', ('screen_reset_mode', 20, 0), ('screen_reset_mode', 4, 0), ('screen_reset_mode', 20, 0)) + pb('\033[=c', ('report_device_attributes', 0, 61)) + s.reset() + + def sgr(*params): + return (('select_graphic_rendition', f'{x}') for x in params) + + pb('\033[1;2;3;4;7;9;34;44m', *sgr('1;2;3;4;7;9;34;44')) + for attr in 'bold italic reverse strikethrough dim'.split(): + self.assertTrue(getattr(s.cursor, attr), attr) + self.ae(s.cursor.decoration, 1) + self.ae(s.cursor.fg, 4 << 8 | 1) + self.ae(s.cursor.bg, 4 << 8 | 1) + pb('\033[38;5;1;48;5;7m', ('select_graphic_rendition', '38:5:1'), ('select_graphic_rendition', '48:5:7')) + self.ae(s.cursor.fg, 1 << 8 | 1) + self.ae(s.cursor.bg, 7 << 8 | 1) + pb('\033[38;2;1;2;3;48;2;7;8;9m', ('select_graphic_rendition', '38:2:1:2:3'), ('select_graphic_rendition', '48:2:7:8:9')) + self.ae(s.cursor.fg, 1 << 24 | 2 << 16 | 3 << 8 | 2) + self.ae(s.cursor.bg, 7 << 24 | 8 << 16 | 9 << 8 | 2) + pb('\033[0;2m', *sgr('0;2')) + pb('\033[;2m', *sgr('0;2')) + pb('\033[m', *sgr('0')) + pb('\033[1;;2m', *sgr('1;0;2')) + pb('\033[38;5;1m', ('select_graphic_rendition', '38:5:1')) + pb('\033[58;2;1;2;3m', ('select_graphic_rendition', '58:2:1:2:3')) + pb('\033[38;2;1;2;3m', ('select_graphic_rendition', '38:2:1:2:3')) + pb('\033[1001:2:1:2:3m', ('select_graphic_rendition', '1001:2:1:2:3')) + pb('\033[38:2:1:2:3;48:5:9;58;5;7m', ( + 'select_graphic_rendition', '38:2:1:2:3'), ('select_graphic_rendition', '48:5:9'), ('select_graphic_rendition', '58:5:7')) + s.reset() + pb('\033[1;2;3;4:5;7;9;34;44m', *sgr('1;2;3', '4:5', '7;9;34;44')) + for attr in 'bold italic reverse strikethrough dim'.split(): + self.assertTrue(getattr(s.cursor, attr), attr) + self.ae(s.cursor.decoration, 5) + c = s.callbacks + pb('\033[5n', ('report_device_status', 5, 0)) + self.ae(c.wtcbuf, b'\033[0n') + c.clear() + pb('\033[6n', ('report_device_status', 6, 0)) + self.ae(c.wtcbuf, b'\033[1;1R') + pb('12345', '12345') + c.clear() + pb('\033[6n', ('report_device_status', 6, 0)) + self.ae(c.wtcbuf, b'\033[2;1R') + c.clear() + s.cursor_key_mode = True + pb('\033[?1$p', ('report_mode_status', 1, 1)) + self.ae(c.wtcbuf, b'\033[?1;1$y') + pb('\033[?1l', ('screen_reset_mode', 1, 1)) + self.assertFalse(s.cursor_key_mode) + c.clear() + pb('\033[?1$p', ('report_mode_status', 1, 1)) + self.ae(c.wtcbuf, b'\033[?1;2$y') + pb('\033[2;4r', ('screen_set_margins', 2, 4)) + c.clear() + pb('\033[14t', ('screen_report_size', 14, 0)) + pb('\033[14;2t', ('screen_report_size', 14, 2)) + self.ae(c.wtcbuf, b'\033[4;100;50t\033[4;100;50t') + self.ae(s.margin_top, 1), self.ae(s.margin_bottom, 3) + pb('\033[r', ('screen_set_margins', 0, 0)) + self.ae(s.margin_top, 0), self.ae(s.margin_bottom, 4) + pb('\033[1 q', ('screen_set_cursor', 1, ord(' '))) + self.assertTrue(s.cursor.blink) + self.ae(s.cursor.shape, CURSOR_BLOCK) + + s.reset() + pb('\033[3 @', ('Shift left escape code not implemented',)) + pb('\033[3 A', ('Shift right escape code not implemented',)) + pb('\033[3;4 S', ('Select presentation directions escape code not implemented',)) + pb('\033[1T', ('screen_reverse_scroll', 1)) + pb('\033[T', ('screen_reverse_scroll', 1)) + pb('\033[+T', ('screen_reverse_scroll_and_fill_from_scrollback', 1)) + + c.clear() + pb('\033[?2026$p', ('report_mode_status', 2026, 1)) + self.ae(c.wtcbuf, b'\x1b[?2026;2$y') + c.clear() + pb('\033[?2026h', ('screen_set_mode', 2026, 1)) + pb('\033[?2026$p', ('report_mode_status', 2026, 1)) + self.ae(c.wtcbuf, b'\x1b[?2026;1$y') + pb('\033[?2026l', ('screen_reset_mode', 2026, 1)) + c.clear() + pb('\033[?2026$p', ('report_mode_status', 2026, 1)) + self.ae(c.wtcbuf, b'\x1b[?2026;2$y') + + def test_csi_code_rep(self): + s = self.create_screen(8) + pb = partial(self.parse_bytes_dump, s) + pb('\033[1b', ('screen_repeat_character', 1)) + self.ae(str(s.line(0)), '') + pb('x\033[7b', 'x', ('screen_repeat_character', 7)) + self.ae(str(s.line(0)), 'xxxxxxxx') + pb('\033[1;3H', ('screen_cursor_position', 1, 3)) + pb('\033[byz\033[b', ('screen_repeat_character', 1), 'yz', ('screen_repeat_character', 1)) + # repeat 'x' at 3, then 'yz' at 4-5, then repeat 'z' at 6 + self.ae(str(s.line(0)), 'xxxyzzxx') + s.reset() + pb(' \033[3b', ' ', ('screen_repeat_character', 3)) + self.ae(str(s.line(0)), ' ') + s.reset() + pb('\t\033[b', ('screen_tab',), ('screen_repeat_character', 1)) + # The terminal includes blank tab padding in its line text. + self.ae(str(s.line(0)).rstrip(' '), '\t') + self.ae(s.cursor.x, 7) + self.ae(s.cursor.y, 0) + s.reset() + b']]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]' + + def test_osc_codes(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + c = s.callbacks + pb('a\033]2;x\\ryz\033\\bcde', 'a', ('set_title', 'x\\ryz'), 'bcde') + self.ae(str(s.line(0)), 'abcde') + self.ae(c.titlebuf, ['x\\ryz']) + c.clear() + pb('\033]\x07', ('set_title', ''), ('set_icon', '')) + self.ae(c.titlebuf, ['']), self.ae(c.iconbuf, '') + pb('1\033]ab\x072', '1', ('set_title', 'ab'), ('set_icon', 'ab'), '2') + self.ae(c.titlebuf, ['', 'ab']), self.ae(c.iconbuf, 'ab') + c.clear() + pb('\033]2;;;;\x07', ('set_title', ';;;')) + self.ae(c.titlebuf, [';;;']) + c.clear() + pb('\033]2;\x07', ('set_title', '')) + self.ae(c.titlebuf, ['']) + pb('\033]110\x07', ('set_dynamic_color', 110, '')) + self.ae(c.colorbuf, '') + c.clear() + pb('\033]9;\x07', ('desktop_notify', 9, '')) + pb('\033]9;test it with a nice long string\x07', ('desktop_notify', 9, 'test it with a nice long string')) + pb('\033]99;moo=foo;test it\x07', ('desktop_notify', 99, 'moo=foo;test it')) + self.ae(c.notifications, [(9, ''), (9, 'test it with a nice long string'), (99, 'moo=foo;test it')]) + c.clear() + pb('\033]8;;\x07', ('set_active_hyperlink', None, None)) + pb('\033]8moo\x07', ('Ignoring malformed OSC 8 code',)) + pb('\033]8;moo\x07', ('Ignoring malformed OSC 8 code',)) + pb('\033]8;id=xyz;\x07', ('set_active_hyperlink', 'xyz', None)) + pb('\033]8;moo:x=z:id=xyz:id=abc;http://yay;.com\x07', ('set_active_hyperlink', 'xyz', 'http://yay;.com')) + c.clear() + payload = '1' * 1024 + pb(f'\033]52;p;{payload}\x07', ('clipboard_control', 52, f'p;{payload}')) + c.clear() + pb('\033]52;p;xyz\x07', ('clipboard_control', 52, 'p;xyz')) + c.clear() + pb('\033]22;?__current__\x07', ('set_dynamic_color', 22, '?__current__')) + + def test_dcs_codes(self): + s = self.create_screen() + c = s.callbacks + pb = partial(self.parse_bytes_dump, s) + q = hexlify(b'kind').decode('ascii') + pb(f'a\033P+q{q}\033\\bcde', 'a', ('screen_request_capabilities', 43, q), 'bcde') + self.ae(str(s.line(0)), 'abcde') + self.ae(c.wtcbuf, '\033P1+r{}={}\033\\'.format(q, '1b5b313b3242').encode('ascii')) + c.clear() + pb('\033P$q q\033\\', ('screen_request_capabilities', ord('$'), ' q')) + self.ae(c.wtcbuf, b'\033P1$r1 q\033\\') + c.clear() + pb('\033P$qm\033\\', ('screen_request_capabilities', ord('$'), 'm')) + self.ae(c.wtcbuf, b'\033P1$rm\033\\') + for sgr in '0;34;102;1;2;3;4 0;38:5:200;58:2:10:11:12'.split(): + expected = set(sgr.split(';')) + c.clear() + parse_bytes(s, f'\033[{sgr}m\033P$qm\033\\'.encode('ascii')) + r = c.wtcbuf.decode('ascii').partition('r')[2].partition('m')[0] + self.ae(expected, set(r.split(';'))) + c.clear() + pb('\033P$qr\033\\', ('screen_request_capabilities', ord('$'), 'r')) + self.ae(c.wtcbuf, f'\033P1$r{s.margin_top + 1};{s.margin_bottom + 1}r\033\\'.encode('ascii')) + pb('\033P@kitty-cmd{abc\033\\', ('handle_remote_cmd', '{abc')) + p = base64_encode('abcd').decode() + pb(f'\033P@kitty-print|{p}\033\\', ('handle_remote_print', p)) + self.ae(['abcd'], s.callbacks.printbuf) + + c.clear() + pb('\033[?2026$p', ('report_mode_status', 2026, 1)) + self.ae(c.wtcbuf, b'\x1b[?2026;2$y') + pb('\033P=1s\033\\', ('screen_start_pending_mode',)) + c.clear() + pb('\033[?2026$p', ('report_mode_status', 2026, 1)) + self.ae(c.wtcbuf, b'\x1b[?2026;1$y') + pb('\033P=2s\033\\', ('screen_stop_pending_mode',)) + c.clear() + pb('\033[?2026$p', ('report_mode_status', 2026, 1)) + self.ae(c.wtcbuf, b'\x1b[?2026;2$y') + + + def test_oth_codes(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb('a\033_+\\+\033\\bcde', ('draw', 'a'), ('Unrecognized APC code: 0x2b',), ('draw', 'bcde')) + pb('a\033^+\\+\033\\bcde', ('draw', 'a'), ('Unrecognized PM code: 0x2b',), ('draw', 'bcde')) + pb('a\033X+\\+\033\\bcde', ('draw', 'a'), ('Unrecognized SOS code: 0x2b',), ('draw', 'bcde')) + + def test_graphics_command(self): + from base64 import standard_b64encode + + def enc(x): + return standard_b64encode(x.encode('utf-8') if isinstance(x, str) else x).decode('ascii') + + def c(**k): + for p, v in tuple(k.items()): + if isinstance(v, str) and p != 'payload': + k[p] = v.encode('ascii') + for f in 'action delete_action transmission_type compressed'.split(): + k.setdefault(f, b'\0') + for f in ('format more id data_sz data_offset width height x_offset y_offset data_height data_width cursor_movement' + ' num_cells num_lines cell_x_offset cell_y_offset z_index placement_id image_number quiet unicode_placement' + ' parent_id parent_placement_id offset_from_parent_x offset_from_parent_y' + ).split(): + k.setdefault(f, 0) + p = k.pop('payload', '') + k[''] = p + return ('graphics_command', k) + + def t(cmd, **kw): + pb('\033_G{};{}\033\\'.format(cmd, enc(kw.get('payload', ''))), c(**kw)) + + def e(cmd, err): + pb(f'\033_G{cmd}\033\\', (err,)) + + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + uint32_max = 2**32 - 1 + t('i=%d' % uint32_max, id=uint32_max) + t('i=3,p=4', id=3, placement_id=4) + e('i=%d' % (uint32_max + 1), 'Malformed GraphicsCommand control block, number is too large') + pb('\033_Gi=12\033\\', c(id=12)) + t('a=t,t=d,s=100,z=-9', payload='X', action='t', transmission_type='d', data_width=100, z_index=-9) + t('a=t,t=d,s=100,z=9', payload='payload', action='t', transmission_type='d', data_width=100, z_index=9) + t('a=t,t=d,s=100,z=9,q=2', action='t', transmission_type='d', data_width=100, z_index=9, quiet=2) + e(',s=1', 'Malformed GraphicsCommand control block, invalid key character: 0x2c') + e('W=1', 'Malformed GraphicsCommand control block, invalid key character: 0x57') + e('1=1', 'Malformed GraphicsCommand control block, invalid key character: 0x31') + e('a=t,,w=2', 'Malformed GraphicsCommand control block, invalid key character: 0x2c') + e('s', 'Malformed GraphicsCommand control block, no = after key') + e('s=', 'Malformed GraphicsCommand control block, expecting an integer value') + e('s==', 'Malformed GraphicsCommand control block, expecting an integer value for key: s') + e('s=1=', 'Malformed GraphicsCommand control block, expecting a , or semi-colon after a value, found: 0x3d') + + def test_deccara(self): + s = self.create_screen() + pb = partial(self.parse_bytes_dump, s) + pb('\033[$r', ('deccara', '0;0;0;0;0')) + pb('\033[;;;;4:3;38:5:10;48:2:1:2:3;1$r', + ('deccara', '0;0;0;0;4:3'), ('deccara', '0;0;0;0;38:5:10'), ('deccara', '0;0;0;0;48:2:1:2:3'), ('deccara', '0;0;0;0;1')) + for y in range(s.lines): + line = s.line(y) + for x in range(s.columns): + c = line.cursor_from(x) + self.ae(c.bold, True) + self.ae(c.italic, False) + self.ae(c.decoration, 3) + self.ae(c.fg, (10 << 8) | 1) + self.ae(c.bg, (1 << 24 | 2 << 16 | 3 << 8 | 2)) + self.ae(s.line(0).cursor_from(0).bold, True) + pb('\033[1;2;2;3;22;39$r', ('deccara', '1;2;2;3;22;39')) + self.ae(s.line(0).cursor_from(0).bold, True) + line = s.line(0) + for x in range(1, s.columns): + c = line.cursor_from(x) + self.ae(c.bold, False) + self.ae(c.fg, 0) + line = s.line(1) + for x in range(0, 3): + c = line.cursor_from(x) + self.ae(c.bold, False) + self.ae(line.cursor_from(3).bold, True) + pb('\033[2*x\033[3;2;4;3;34$r\033[*x', ('screen_decsace', 2), ('deccara', '3;2;4;3;34'), ('screen_decsace', 0)) + for y in range(2, 4): + line = s.line(y) + for x in range(s.columns): + self.ae(line.cursor_from(x).fg, (10 << 8 | 1) if x < 1 or x > 2 else (4 << 8) | 1) + + +if __name__ == '__main__': + import unittest + unittest.main() diff --git a/tests/extern/kitty/screen.py b/tests/extern/kitty/screen.py new file mode 100644 index 0000000..80f8e71 --- /dev/null +++ b/tests/extern/kitty/screen.py @@ -0,0 +1,476 @@ +"""ctypes wrapper providing the Kitty ``Screen`` API on top of our C terminal. + +The shared library ``libterm.so`` is loaded via :mod:`ctypes`. Python calls +only the ``kitty_*`` functions of ``screen_bridge.c``, which wrap the public +API (declared in ``src/term/term.h``). Python classes on top of them match the +subset of Kitty's test API used by ``parser.py``. +""" + +import ctypes +import os + +# --------------------------------------------------------------------------- +# Shared library loading +# --------------------------------------------------------------------------- + +_lib = ctypes.CDLL(os.environ.get('WSAP_KITTY_LIB', os.path.join(os.path.dirname(__file__), 'libterm.so'))) + +# struct term_cell { uint32_t text, fg, bg; uint16_t attrs; uint8_t width, reserved; } +class TermCell(ctypes.Structure): + _fields_ = [ + ('text', ctypes.c_uint32), + ('fg', ctypes.c_uint32), + ('bg', ctypes.c_uint32), + ('attrs', ctypes.c_uint16), + ('width', ctypes.c_uint8), + ('reserved', ctypes.c_uint8), + ] + + +assert ctypes.sizeof(TermCell) == 16 + +# Reply callback type: void (*)(void *ud, const char *bytes, size_t len). +# The bytes are not NUL-terminated, so the callback takes a raw pointer. +REPLY_FN = ctypes.CFUNCTYPE(None, ctypes.c_void_p, ctypes.POINTER(ctypes.c_char), ctypes.c_size_t) + +# Function signatures +_lib.kitty_term_new.restype = ctypes.c_void_p +_lib.kitty_term_new.argtypes = [ctypes.c_int, ctypes.c_int, REPLY_FN, ctypes.c_void_p] +_lib.kitty_term_free.restype = None +_lib.kitty_term_free.argtypes = [ctypes.c_void_p] +_lib.kitty_term_resize.restype = ctypes.c_bool +_lib.kitty_term_resize.argtypes = [ctypes.c_void_p, ctypes.c_int, ctypes.c_int] +_lib.kitty_term_write.restype = None +_lib.kitty_term_write.argtypes = [ctypes.c_void_p, ctypes.c_char_p, ctypes.c_size_t] +_lib.kitty_term_cell.restype = ctypes.c_bool +_lib.kitty_term_cell.argtypes = [ctypes.c_void_p, ctypes.c_int, ctypes.c_int, + ctypes.POINTER(TermCell)] +_lib.kitty_term_cell_text.restype = ctypes.c_int +_lib.kitty_term_cell_text.argtypes = [ctypes.c_void_p, ctypes.c_int, ctypes.c_int, + ctypes.POINTER(ctypes.c_uint32), ctypes.c_int] +for _name, _type in { + 'cols': ctypes.c_int, + 'rows': ctypes.c_int, + 'cursor_x': ctypes.c_int, + 'cursor_y': ctypes.c_int, + 'cursor_style': ctypes.c_int, + 'cursor_blink': ctypes.c_bool, + 'application_cursor': ctypes.c_bool, +}.items(): + _fn = getattr(_lib, 'kitty_term_' + _name) + _fn.argtypes = [ctypes.c_void_p] + _fn.restype = _type + +# --------------------------------------------------------------------------- +# Constants +# --------------------------------------------------------------------------- + +TERM_DEFAULT_FG = 0xFFFFFFFF +TERM_DEFAULT_BG = 0xFF1E1E2E + +# Cell attributes (must match enum term_attr in term.h) +TERM_ATTR_BOLD = 1 << 0 +TERM_ATTR_ITALIC = 1 << 1 +TERM_ATTR_FAINT = 1 << 2 +TERM_ATTR_STRIKE = 1 << 3 +TERM_ATTR_OVERLINE = 1 << 4 +TERM_ATTR_LINK = 1 << 5 +TERM_ATTR_EMOJI = 1 << 6 +TERM_ATTR_TEXT = 1 << 7 +TERM_ATTR_UNDERLINE_SHIFT = 8 +TERM_ATTR_UNDERLINE_MASK = 7 << TERM_ATTR_UNDERLINE_SHIFT + +# The longest grapheme cluster that a cell's text is read with. +_CLUSTER_CAP = 32 + +# Test-only accessors compiled against the current C terminal definition. +for _name, _type in { + 'scroll_top': ctypes.c_int, + 'scroll_bottom': ctypes.c_int, + 'cur_fg': ctypes.c_uint32, + 'cur_bg': ctypes.c_uint32, + 'bold': ctypes.c_bool, + 'faint': ctypes.c_bool, + 'italic': ctypes.c_bool, + 'strike': ctypes.c_bool, + 'inverse': ctypes.c_bool, + 'underline': ctypes.c_uint8, +}.items(): + _fn = getattr(_lib, 'kitty_screen_' + _name) + _fn.argtypes = [ctypes.c_void_p] + _fn.restype = _type + +_lib.kitty_screen_row_length.argtypes = [ctypes.c_void_p, ctypes.c_int] +_lib.kitty_screen_row_length.restype = ctypes.c_int +_lib.kitty_screen_cell_is_tab.argtypes = [ctypes.c_void_p, ctypes.c_int, ctypes.c_int] +_lib.kitty_screen_cell_is_tab.restype = ctypes.c_bool + +# ANSI color table from escape.c — used for reverse-mapping to Kitty indices. +# Must match the built-in default palette in src/term/palette.c. +_ANSI_COLORS = [ + 0xFF585B70, 0xFFCC6666, 0xFFB5BD68, 0xFFF0C674, + 0xFF81A2BE, 0xFFB294BB, 0xFF8ABEB7, 0xFFC5C8C6, + 0xFF666666, 0xFFD54E53, 0xFFB9CA4A, 0xFFE7C547, + 0xFF7AA6DA, 0xFFC397D8, 0xFF70C0B1, 0xFFEAEAEA, +] + + +def _build_color_table(): + """Build a reverse lookup from RGB (0xFFRRGGBB) to Kitty indexed color.""" + table = {} + for i, rgb in enumerate(_ANSI_COLORS): + table[rgb] = (i << 8) | 1 + levels = [0, 95, 135, 175, 215, 255] + for i in range(16, 232): + n = i - 16 + r, g, b = levels[n // 36], levels[(n // 6) % 6], levels[n % 6] + rgb = 0xFF000000 | (r << 16) | (g << 8) | b + table[rgb] = (i << 8) | 1 + for i in range(232, 256): + gray = 8 + (i - 232) * 10 + rgb = 0xFF000000 | (gray << 16) | (gray << 8) | gray + table[rgb] = (i << 8) | 1 + return table + + +_COLOR_TO_KITTY = _build_color_table() + + +TERM_COLOR_REF = 0x01000000 + + +def _convert_color(color): + """Convert our cell color to Kitty's format. + + Ours: a palette reference (top byte 0x01; slot 0-255, 0x100 default fg, + 0x101 default bg) or a literal 0xFFRRGGBB color. + + Kitty format: + 0 = default + index<<8|1 = indexed (0-255) + r<<24|g<<16|b<<8|2 = RGB + """ + if (color >> 24) == (TERM_COLOR_REF >> 24): + slot = color & 0x1FF + return 0 if slot >= 0x100 else (slot << 8) | 1 + if color == TERM_DEFAULT_FG or color == TERM_DEFAULT_BG: + return 0 + kitty = _COLOR_TO_KITTY.get(color) + if kitty is not None: + return kitty + r = (color >> 16) & 0xFF + g = (color >> 8) & 0xFF + b = color & 0xFF + return (r << 24) | (g << 16) | (b << 8) | 2 + + +# --------------------------------------------------------------------------- +# Screen property readers +# --------------------------------------------------------------------------- + +def _read_property(ptr, name): + return getattr(_lib, 'kitty_screen_' + name)(ptr) + + +# --------------------------------------------------------------------------- +# Cursor — wraps the terminal's current cursor SGR state + position +# --------------------------------------------------------------------------- + +class Cursor: + """Wraps the terminal's current cursor state (SGR attributes + position).""" + + def __init__(self, screen): + self._screen = screen + + @property + def _ptr(self): + return self._screen._ptr + + # -- position -- + @property + def x(self): + return _lib.kitty_term_cursor_x(self._screen._term) + + @property + def y(self): + return _lib.kitty_term_cursor_y(self._screen._term) + + # -- SGR boolean attributes -- + @property + def bold(self): + return _read_property(self._ptr, 'bold') + + @property + def italic(self): + return _read_property(self._ptr, 'italic') + + @property + def dim(self): + return _read_property(self._ptr, 'faint') + + @property + def reverse(self): + return _read_property(self._ptr, 'inverse') + + @property + def strikethrough(self): + return _read_property(self._ptr, 'strike') + + @property + def decoration(self): + # Kitty: 0=none, 1=straight, 2=double, 3=curly, 4=dotted, 5=dashed + return _read_property(self._ptr, 'underline') + + # -- colors -- + @property + def fg(self): + return _convert_color(_read_property(self._ptr, 'cur_fg')) + + @property + def bg(self): + return _convert_color(_read_property(self._ptr, 'cur_bg')) + + # -- cursor style -- + @property + def blink(self): + return _lib.kitty_term_cursor_blink(self._screen._term) + + @property + def shape(self): + return _lib.kitty_term_cursor_style(self._screen._term) + + +# --------------------------------------------------------------------------- +# CellCursor — wraps a single cell's attributes +# --------------------------------------------------------------------------- + +class CellCursor: + """Wraps a single cell's attributes for SGR inspection.""" + + def __init__(self, cell): + self._cell = cell + + @property + def bold(self): + return bool(self._cell.attrs & TERM_ATTR_BOLD) + + @property + def italic(self): + return bool(self._cell.attrs & TERM_ATTR_ITALIC) + + @property + def dim(self): + return bool(self._cell.attrs & TERM_ATTR_FAINT) + + @property + def reverse(self): + # Our terminal swaps fg/bg when inverse is set, so cells don't + # carry a separate inverse flag. + return False + + @property + def strikethrough(self): + return bool(self._cell.attrs & TERM_ATTR_STRIKE) + + @property + def decoration(self): + # enum term_underline uses Kitty's numbers: 0=none, 1=straight, + # 2=double, 3=curly, 4=dotted, 5=dashed. + return (self._cell.attrs & TERM_ATTR_UNDERLINE_MASK) >> TERM_ATTR_UNDERLINE_SHIFT + + @property + def fg(self): + return _convert_color(self._cell.fg) + + @property + def bg(self): + return _convert_color(self._cell.bg) + + +# --------------------------------------------------------------------------- +# Line — wraps a screen line, providing cell access +# --------------------------------------------------------------------------- + +class Line: + """Wraps a screen line, providing cell access and text content.""" + + def __init__(self, screen, y): + self._screen = screen + self._y = y + + def cursor_from(self, x): + """Return a :class:`CellCursor` for cell (x, y).""" + cell = TermCell() + if not _lib.kitty_term_cell(self._screen._term, x, self._y, ctypes.byref(cell)): + return CellCursor(TermCell()) + return CellCursor(cell) + + def __str__(self): + """Return the text content of the line up to the written length.""" + length = min(_lib.kitty_screen_row_length(self._screen._term, self._y), + self._screen.columns) + if length == 0: + return '' + result = [] + cell = TermCell() + cps = (ctypes.c_uint32 * _CLUSTER_CAP)() + for x in range(length): + if not _lib.kitty_term_cell(self._screen._term, x, self._y, ctypes.byref(cell)): + result.append(' ') + continue + # Width 0 is the right half of a wide glyph. + if cell.width == 0: + continue + if _lib.kitty_screen_cell_is_tab(self._screen._term, x, self._y): + result.append('\t') + continue + n = _lib.kitty_term_cell_text(self._screen._term, x, self._y, cps, _CLUSTER_CAP) + n = min(n, _CLUSTER_CAP) + result.append(''.join(chr(cp) for cp in cps[:n])) + return ''.join(result) + + +# --------------------------------------------------------------------------- +# Callbacks — mock callbacks object for OSC/DCS responses +# --------------------------------------------------------------------------- + +class Callbacks: + """Mock callbacks object for OSC/DCS responses.""" + + def __init__(self, screen=None): + self._screen = screen + self.wtcbuf = b'' + self.titlebuf = [] + self.iconbuf = '' + self.colorbuf = '' + self.notifications = [] + self.printbuf = [] + + def clear(self): + self.wtcbuf = b'' + self.titlebuf = [] + self.iconbuf = '' + self.colorbuf = '' + self.notifications = [] + self.printbuf = [] + if self._screen is not None: + self._screen._response_buf.clear() + + +# --------------------------------------------------------------------------- +# Screen — main screen wrapper using ctypes +# --------------------------------------------------------------------------- + +class Screen: + """Main screen wrapper using ctypes to drive the C terminal.""" + + def __init__(self, cols=80, rows=5): + self._cols = cols + self._rows = rows + self._response_buf = bytearray() + self._write_buf = bytearray() + self._callbacks = Callbacks(self) + # Keep a reference to the callback function so it isn't GC'd + self._resp_fn = REPLY_FN(self._on_response) + self._term = _lib.kitty_term_new(cols, rows, self._resp_fn, None) + if not self._term: + raise RuntimeError('kitty_term_new failed') + + @property + def _ptr(self): + return ctypes.cast(self._term, ctypes.c_void_p).value + + # -- response callback -- + def _on_response(self, ud, data, length): + if data and length: + self._response_buf.extend(data[:length]) + self._callbacks.wtcbuf = bytes(self._response_buf) + + # -- writing -- + def write(self, data): + if data: + _lib.kitty_term_write(self._term, bytes(data), len(data)) + + # -- test write buffer API (matches Kitty's test_parse interface) -- + def test_create_write_buffer(self): + """Return the available space in the write buffer (int, falsy when full).""" + from fast_data_types import VT_PARSER_BUFFER_SIZE + return VT_PARSER_BUFFER_SIZE - len(self._write_buf) + + def test_commit_write_buffer(self, data, write_buf): + """Append *data* to the internal write buffer, return bytes consumed.""" + from fast_data_types import VT_PARSER_BUFFER_SIZE + available = VT_PARSER_BUFFER_SIZE - len(self._write_buf) + n = min(len(data), available) + if n > 0: + self._write_buf.extend(data[:n]) + return n + + def test_parse_written_data(self, cd=None): + """Parse the write buffer, dispatching to the C terminal and callback.""" + data = bytes(self._write_buf) + self._write_buf.clear() + from kitty_parser import parse_bytes + if cd is not None: + parse_bytes(self, data, cd) + elif data: + # No callback — still parse to keep parser state in sync + parse_bytes(self, data, lambda *a: None) + + # -- properties -- + @property + def cursor(self): + return Cursor(self) + + @property + def callbacks(self): + return self._callbacks + + @property + def columns(self): + return _lib.kitty_term_cols(self._term) + + @property + def lines(self): + return _lib.kitty_term_rows(self._term) + + @property + def margin_top(self): + return _read_property(self._ptr, 'scroll_top') + + @property + def margin_bottom(self): + return _read_property(self._ptr, 'scroll_bottom') + + @property + def cursor_key_mode(self): + return _lib.kitty_term_application_cursor(self._term) + + @cursor_key_mode.setter + def cursor_key_mode(self, value): + if value: + self.write(b'\x1b[?1h') + else: + self.write(b'\x1b[?1l') + + # -- line access -- + def line(self, n): + return Line(self, n) + + # -- operations -- + def reset(self): + self.write(b'\x1bc') + + def cursor_move(self, n): + """Move cursor back n cells (kitty cursor_move: direction -1, + allowed to move to previous lines).""" + self.write(b'\x1b[' + str(n).encode() + b'D') + + def close(self): + """Destroy the underlying C terminal.""" + if getattr(self, '_term', None): + _lib.kitty_term_free(self._term) + self._term = None + + def __del__(self): + self.close() diff --git a/tests/extern/kitty/screen_bridge.c b/tests/extern/kitty/screen_bridge.c new file mode 100644 index 0000000..63a3583 --- /dev/null +++ b/tests/extern/kitty/screen_bridge.c @@ -0,0 +1,113 @@ +/* Test-only entry points for the Kitty screen wrapper in screen.py. + * Python calls only these functions, so it needs no struct layouts other + * than the public struct term_cell. The kitty_screen_* accessors read + * properties that term.h does not expose. They compile against the current + * terminal layout rather than duplicating byte offsets in Python. */ +#include "internal.h" +#include "screen.h" + +typedef void (*kitty_reply_fn)(void *ud, const char *bytes, size_t len); + +struct term *kitty_term_new(int cols, int rows, kitty_reply_fn reply, void *ud) +{ + return term_new(&(struct term_options){ + .size = { .x = cols, .y = rows }, + .callbacks = { .ud = ud, .reply = reply }, + }); +} + +/* One snapshot serves every cell read. */ +static struct term_snapshot *snap; + +void kitty_term_free(struct term *t) +{ + /* A new terminal can get the address of the freed one. The snapshot + * would then keep its rows, so it goes away too. */ + term_snapshot_free(snap); + snap = NULL; + term_free(t); +} + +void kitty_term_write(struct term *t, const char *bytes, size_t len) +{ + term_feed(t, bytes, len, 0); +} + +bool kitty_term_resize(struct term *t, int cols, int rows) +{ + return term_resize(t, (struct term_vector2i){ .x = cols, .y = rows }); +} + +int kitty_term_cols(struct term *t) { return term_size(t).x; } +int kitty_term_rows(struct term *t) { return term_size(t).y; } +int kitty_term_cursor_x(struct term *t) { return term_cursor(t).pos.x; } +int kitty_term_cursor_y(struct term *t) { return term_cursor(t).pos.y; } +int kitty_term_cursor_style(struct term *t) { return (int)term_cursor(t).style; } +bool kitty_term_cursor_blink(struct term *t) { return term_cursor(t).blink; } +bool kitty_term_application_cursor(struct term *t) { return term_modes(t).app_cursor; } + +/* Update the snapshot from the live screen, as a renderer reads it. + * Returns NULL when the cell is out of range or memory runs out. */ +static struct term_snapshot *cell_snapshot(struct term *t, int col, int row) +{ + struct term_vector2i size = term_size(t); + if (col < 0 || col >= size.x || row < 0 || row >= size.y) + return NULL; + if (!snap) + snap = term_snapshot_new(); + term_scroll_to(t, 0); + if (!snap || !term_snapshot_update(snap, t)) + return NULL; + return snap; +} + +bool kitty_term_cell(struct term *t, int col, int row, struct term_cell *out) +{ + struct term_snapshot *s = cell_snapshot(t, col, row); + if (!s) + return false; + *out = term_snapshot_row(s, row)[col]; + return true; +} + +/* The code points of a cell's text: one, or a whole grapheme cluster. + * Returns the number of code points and writes at most cap of them. */ +int kitty_term_cell_text(struct term *t, int col, int row, uint32_t *cps, int cap) +{ + struct term_snapshot *s = cell_snapshot(t, col, row); + if (!s) + return 0; + return term_snapshot_cluster(s, term_snapshot_row(s, row)[col].text, cps, cap); +} + +#define PROPERTY(type, name) \ + type kitty_screen_##name(struct term *t) { return t->name; } + +PROPERTY(int, scroll_top) +PROPERTY(int, scroll_bottom) +PROPERTY(uint32_t, cur_fg) +PROPERTY(uint32_t, cur_bg) +PROPERTY(bool, bold) +PROPERTY(bool, faint) +PROPERTY(bool, italic) +PROPERTY(bool, strike) +PROPERTY(bool, inverse) +PROPERTY(uint8_t, underline) + +int kitty_screen_row_length(struct term *t, int row) +{ + if (!t || row < 0 || row >= t->rows) return 0; + const struct srow *r = term_screen_row(t, row); + return r ? (int)r->len : 0; +} + +/* A tab stores a marker cell that a snapshot shows as a blank. Kitty keeps + * the tab in its line text, so the wrapper asks for the marker. */ +bool kitty_screen_cell_is_tab(struct term *t, int col, int row) +{ + if (col < 0 || col >= t->cols || row < 0 || row >= t->rows) + return false; + term_scroll_to(t, 0); + const struct cell *cells = term_screen_view_row(t, row); + return cells && cells[col].c == '\t'; +} diff --git a/tests/fuzz_gen.h b/tests/fuzz_gen.h new file mode 100644 index 0000000..166d729 --- /dev/null +++ b/tests/fuzz_gen.h @@ -0,0 +1,188 @@ +#ifndef WSAP_TERM_FUZZ_GEN_H +#define WSAP_TERM_FUZZ_GEN_H + +/* + * The input generator that write-fuzz-test and model-test share. gen() + * makes whole sequences, malformed UTF-8, and raw bytes. The model handles + * every byte that gen() makes. + */ + +#include <stdarg.h> +#include <stdint.h> +#include <stdio.h> +#include <string.h> + +static uint64_t rng; + +static uint32_t rnd(void) +{ + rng ^= rng << 13; + rng ^= rng >> 7; + rng ^= rng << 17; + return (uint32_t)rng; +} + +static int rn(int n) +{ + return (int)(rnd() % (uint32_t)n); +} + +static char buf[1 << 16]; +static size_t blen; + +static void putb(char c) +{ + if (blen + 1 < sizeof(buf)) + buf[blen++] = c; +} + +static void put(const char *s) +{ + size_t n = strlen(s); + if (blen + n < sizeof(buf)) { + memcpy(buf + blen, s, n); + blen += n; + } +} + +static void __attribute__((format(printf, 1, 2))) putf(const char *fmt, ...) +{ + char tmp[64]; + va_list args; + va_start(args, fmt); + vsnprintf(tmp, sizeof(tmp), fmt, args); + va_end(args); + put(tmp); +} + +#define PICK(list) (list)[rn((int)(sizeof(list) / sizeof((list)[0])))] + +static void gen_sgr(void) +{ + static const int plain[] = { + 0, 1, 2, 3, 4, 5, 7, 8, 9, 21, 22, 23, 24, 27, 29, 39, 49, 53, 55, 59, + }; + put("\033["); + int n = 1 + rn(3); + for (int i = 0; i < n; i++) { + if (i) put(";"); + switch (rn(9)) { + case 0: putf("38;5;%d", rn(300)); break; + case 1: putf("48;5;%d", rn(300)); break; + case 2: putf("38;2;%d;%d;%d", rn(256), rn(256), rn(256)); break; + case 3: putf("48:2::%d:%d:%d", rn(256), rn(256), rn(256)); break; + case 4: putf("4:%d", rn(7)); break; + case 5: putf("%d", 30 + rn(8) + (rn(2) ? 60 : 0)); break; + case 6: putf("%d", 40 + rn(8) + (rn(2) ? 60 : 0)); break; + case 7: putf("58;5;%d", rn(256)); break; + default: putf("%d", PICK(plain)); break; + } + } + put("m"); +} + +static inline void gen(int ops, int cols, int rows) +{ + static const char *words[] = { + "a", "hello", "x y", "0123456789", "日本", "é", "中文字", + "e\xcc\x81", "👍", " ", + }; + /* Include valid text, malformed UTF-8, and incomplete UTF-8. */ + static const char *utf8_3[] = { + "中", "文", "字", "€", "ｱ", "\xe2\x83\x90", "\xed\xa0\x80", + "\xe0\x80\xaf", "\xef\xbf\xbd", "한", "a", "\xe2\x80\x8d", "\xe3\x80", + }; + /* Emoji, flags, modifiers, ZWJ sequences, presentation selectors, + * stacked marks, and C1 controls in UTF-8 form. */ + static const char *clusters[] = { + "😀", "\xf0\x9f\x87\xa9\xf0\x9f\x87\xaa", "👍🏽", "👨👩👧", "❤️", "☺︎", + "a\xcc\x81\xcc\x82", "\xe2\x80\x8b", "\xc2\x85", "\xc2\x84", "\xc2\x8d", + "\xc2\x88", "\xf0\x9f\x87\xa9", "\xcc\x81", + }; + static const char *modes[] = { + "\033[4h", "\033[4l", "\033[?7l", "\033[?7h", + "\033[?1049h", "\033[?1049l", "\033[?6h", "\033[?6l", + "\033[20h", "\033[20l", "\033[?25l", "\033[?25h", + "\033[?5h", "\033[?5l", + }; + static const char *escs[] = { + "\033D", "\033M", "\033E", "\0337", "\0338", "\033[s", "\033[u", "\033c", + "\033[?1048h", "\033[?1048l", "\033H", "\033=", "\033>", "\033%G", + }; + static const char *charsets[] = { + "\033(0", "\033(B", "\033)0", "\033)B", "\x0e", "\x0f", + }; + static const char *queries[] = { + "\033[5n", "\033[6n", "\033[c", "\033[0c", "\033[>c", "\033[1c", + "\033[14t", "\033P$qm\033\\", "\033P$qr\033\\", "\033[?6$p", + "\033[4$p", "\033[20$p", + }; + /* C0 controls inside CSI, cancelled sequences, and ignored strings. */ + static const char *odd[] = { + "\033[2\n;3H", "\033[1\bC", "\033[3\x18m", "\033[1\x1a;2H", + "\033[\1772C", "\x0b", "\x0c", "\x7f", "\x18", "\x1a", + "\033]0;title\007", "\033]2;x\033\\", "\033_apc\033\\", "\033^pm\007", + "\033X\033\\", "\033]133;A\007", "\033]133;A\033\\", "\033]133;D\007", + "\033#8", + }; + static const char csi_final[] = "ABCDEFGHIJKLMPSTXZ@`abdefghlmnrsu"; + + for (int i = 0; i < ops; i++) { + switch (rn(36)) { + case 0: case 1: case 2: put(PICK(words)); break; + case 3: { + int n = rn(cols * 2 + 2); + for (int k = 0; k < n; k++) + putb((char)(0x21 + rn(94))); + break; + } + case 4: put("\r\n"); break; + case 5: put("\n"); break; + case 6: put("\r"); break; + case 7: put("\t"); break; + case 8: put("\b"); break; + case 9: putf("\033[%d;%dH", rn(rows + 2), rn(cols + 2)); break; + case 10: putf("\033[%d%c", rn(4), "JK"[rn(2)]); break; + case 11: putf("\033[%d%c", rn(cols), "XP@LMST"[rn(7)]); break; + case 12: putf("\033[%d;%dm", 30 + rn(8), 40 + rn(8)); break; + case 13: put(rn(2) ? "\033[m" : "\033[0m"); break; + case 14: putf("\033[48;2;%d;%d;1m", rn(256), rn(256)); break; + case 15: putf("\033[%d%c", rn(cols + 2), csi_final[rn((int)sizeof(csi_final) - 1)]); break; + case 16: putf("\033[%d;%dr", rn(rows) + 1, rn(rows) + 1); break; + case 17: put(PICK(modes)); break; + case 18: put(PICK(escs)); break; + case 19: putf("\033[%d%c", rn(5), "IZb"[rn(3)]); break; + case 20: putf("\033[%d;%d;1;1$r", rn(rows) + 1, rn(cols) + 1); break; + case 21: put("\033[1;7m"); break; + case 22: case 23: { + /* Keep ESC out of raw bytes so later input reaches the fast paths. */ + int n = 64 + rn(400); + for (int k = 0; k < n; k++) { + unsigned char b = (unsigned char)rnd(); + putb((char)(b == 0x1b ? 'e' : b)); + } + break; + } + case 24: case 25: { + int n = 4 + rn(40); + for (int k = 0; k < n; k++) + put(utf8_3[rn(20) < 14 ? rn(3) : rn(13)]); + break; + } + case 26: case 27: gen_sgr(); break; + case 28: put(PICK(clusters)); break; + case 29: put(PICK(charsets)); break; + case 30: put(PICK(queries)); break; + case 31: put(PICK(odd)); break; + case 32: putf("\033[%d q", rn(7)); break; + case 33: putf("\033[%dg", rn(4)); break; + case 34: + putf("\033[%d;%d;%d;%d;%d$r", rn(rows) + 1, rn(cols) + 1, + rn(rows) + 1, rn(cols) + 1, PICK(((const int[]){ 0, 1, 4, 7, 22, 24, 27, 31, 42 }))); + break; + case 35: putf("\033[%d*x", rn(3)); break; + } + } +} + +#endif diff --git a/tests/gstore_test.c b/tests/gstore_test.c new file mode 100644 index 0000000..637b136 --- /dev/null +++ b/tests/gstore_test.c @@ -0,0 +1,227 @@ +/* + * Tests for the per-terminal grapheme store (gstore). + * + * Verifies: + * - Two terminals do not share clusters: freeing one leaves the other intact. + * - Clusters that remain visible on screen survive GC sweeps. + * - More than 65,536 distinct clusters can be interned as old ones scroll + * out of history, without reaching any cap or exhausting memory. + * - Direct unit tests of intern, expand, deduplication, and sweep. + */ +#include "gstore.h" +#include "internal.h" +#include "term/term.h" +#include "term/unicode/utf8.h" + +#include <assert.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static void test_direct_unit(void) +{ + struct gstore *g = term_gstore_new(NULL); + assert(g != NULL); + assert(term_gstore_count(g) == 0); + + /* Single code point returns the code point itself. */ + uint32_t single = 'A'; + assert(term_gstore_intern(g, &single, 1) == 'A'); + assert(term_gstore_count(g) == 0); + + /* Cluster returns a handle. */ + uint32_t c1[2] = { 'e', 0x0301 }; + uint32_t h1 = term_gstore_intern(g, c1, 2); + assert(grapheme_is_handle(h1)); + assert(term_gstore_count(g) == 1); + + /* Identical cluster deduplicates to the same handle. */ + uint32_t h1_dup = term_gstore_intern(g, c1, 2); + assert(h1_dup == h1); + assert(term_gstore_count(g) == 1); + + /* Different cluster gets a distinct handle. */ + uint32_t c2[2] = { 'a', 0x0300 }; + uint32_t h2 = term_gstore_intern(g, c2, 2); + assert(grapheme_is_handle(h2)); + assert(h2 != h1); + assert(term_gstore_count(g) == 2); + + /* Expand both handles. */ + uint32_t out[GRAPHEME_MAX_LEN] = {0}; + int n = term_gstore_expand(g, h1, out, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(out[0] == 'e' && out[1] == 0x0301); + + n = term_gstore_expand(g, h2, out, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(out[0] == 'a' && out[1] == 0x0300); + + /* Sweep with h1 marked and h2 unmarked. */ + term_gstore_mark(g, h1); + term_gstore_sweep(g, NULL); + assert(term_gstore_count(g) == 1); + + /* h1 is still valid. */ + n = term_gstore_expand(g, h1, out, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(out[0] == 'e' && out[1] == 0x0301); + + /* h2 was reclaimed. */ + n = term_gstore_expand(g, h2, out, GRAPHEME_MAX_LEN); + assert(n == 1 && out[0] == 0xFFFD); + + /* New cluster can reuse the reclaimed slot. */ + uint32_t c3[3] = { 'o', 0x0308, 0x0301 }; + uint32_t h3 = term_gstore_intern(g, c3, 3); + assert(grapheme_is_handle(h3)); + assert(term_gstore_count(g) == 2); + n = term_gstore_expand(g, h3, out, GRAPHEME_MAX_LEN); + assert(n == 3); + assert(out[0] == 'o' && out[1] == 0x0308 && out[2] == 0x0301); + + term_gstore_free(g); +} + +static void test_two_terminals_isolation(void) +{ + struct term_options opt = { .size = { .x = 20, .y = 4 } }; + struct term *t1 = term_new(&opt); + struct term *t2 = term_new(&opt); + assert(t1 && t2); + + const char *cluster_text = "e\xcc\x81"; /* 'e' with acute */ + term_feed(t1, cluster_text, strlen(cluster_text), 0); + term_feed(t2, cluster_text, strlen(cluster_text), 0); + + /* Free terminal 1. */ + term_free(t1); + + /* Terminal 2 must still expand its cluster correctly. */ + struct term_snapshot *snap = term_snapshot_new(); + assert(snap); + assert(term_snapshot_update(snap, t2)); + + const struct term_cell *row = term_snapshot_row(snap, 0); + assert(row[0].text & TERM_TEXT_CLUSTER); + + uint32_t cps[GRAPHEME_MAX_LEN] = {0}; + int n = term_snapshot_cluster(snap, row[0].text, cps, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(cps[0] == 'e' && cps[1] == 0x0301); + + term_snapshot_free(snap); + term_free(t2); +} + +static void test_screen_clusters_survive_sweep(void) +{ + struct term_options opt = { .size = { .x = 40, .y = 10 }, .scrollback = 10 }; + struct term *t = term_new(&opt); + assert(t); + + /* Write 5 clusters on the top row. */ + term_feed(t, "a\xcc\x80" "b\xcc\x81" "c\xcc\x82" "d\xcc\x83" "e\xcc\x84", 15, 0); + + /* Write ephemeral clusters on row 1, then overwrite them. */ + for (int i = 0; i < 50; i++) { + term_feed(t, "\033[2;1H\033[2K", 8, 0); + char buf[32]; + int len = snprintf(buf, sizeof(buf), "z\xcc\x80\xcc\x81%d", i); + term_feed(t, buf, (size_t)len, 0); + } + term_feed(t, "\033[2;1H\033[2K", 8, 0); + + /* Explicit sweep. The clusters on row 0 are on screen and must survive. */ + term_gstore_sweep(t->gstore, t); + + struct term_snapshot *snap = term_snapshot_new(); + assert(snap); + assert(term_snapshot_update(snap, t)); + + /* Row 0 of screen has "a\xcc\x80...". */ + const struct term_cell *row = term_snapshot_row(snap, 0); + assert(row[0].text & TERM_TEXT_CLUSTER); + + uint32_t cps[GRAPHEME_MAX_LEN] = {0}; + int n = term_snapshot_cluster(snap, row[0].text, cps, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(cps[0] == 'a' && cps[1] == 0x0300); + + n = term_snapshot_cluster(snap, row[1].text, cps, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(cps[0] == 'b' && cps[1] == 0x0301); + + n = term_snapshot_cluster(snap, row[2].text, cps, GRAPHEME_MAX_LEN); + assert(n == 2); + assert(cps[0] == 'c' && cps[1] == 0x0302); + + term_snapshot_free(snap); + term_free(t); +} + +static void test_more_than_65536_clusters(void) +{ + /* Bounded history so old lines scroll out. */ + struct term_options opt = { .size = { .x = 80, .y = 24 }, .scrollback = 10 }; + struct term *t = term_new(&opt); + assert(t); + + /* + * Generate 66,000 distinct clusters. + * Base 'a' (0x61) + mark1 (0x0300..0x034F, 80 options) + mark2 (0x0350..0x036F + 0xFE00..0xFE0F, 830 options) + * 256 * 260 = 66,560 distinct clusters. + */ + char buf[64]; + uint32_t total = 0; + const uint32_t target = 66000; + + for (uint32_t i = 0; total < target; i++) { + char base = (char)('a' + (i % 26)); + uint32_t cp1 = 0x0300 + ((i / 26) % 55); + uint32_t cp2 = 0x0340 + ((i / (26 * 55)) % 47); + + size_t len = 0; + buf[len++] = base; + len += (size_t)utf8_encode(cp1, buf + len); + len += (size_t)utf8_encode(cp2, buf + len); + buf[len++] = '\r'; + buf[len++] = '\n'; + + term_feed(t, buf, len, 0); + total++; + } + + assert(total >= 65536); + + /* Verify the latest cluster on screen reads back correctly. */ + struct term_snapshot *snap = term_snapshot_new(); + assert(snap); + assert(term_snapshot_update(snap, t)); + + struct term_cursor c = term_cursor(t); + int y = c.pos.y > 0 ? c.pos.y - 1 : 0; + const struct term_cell *row = term_snapshot_row(snap, y); + assert(row[0].text & TERM_TEXT_CLUSTER); + + uint32_t cps[GRAPHEME_MAX_LEN] = {0}; + int n = term_snapshot_cluster(snap, row[0].text, cps, GRAPHEME_MAX_LEN); + assert(n == 3); + + /* Live cluster count must be small (only bounded screen and history). */ + uint32_t count = term_gstore_count(t->gstore); + assert(count < 2000); + + term_snapshot_free(snap); + term_free(t); +} + +int main(void) +{ + test_direct_unit(); + test_two_terminals_isolation(); + test_screen_clusters_survive_sweep(); + test_more_than_65536_clusters(); + puts("PASS: gstore tests"); + return 0; +} diff --git a/tests/header_test.sh b/tests/header_test.sh new file mode 100644 index 0000000..6636418 --- /dev/null +++ b/tests/header_test.sh @@ -0,0 +1,16 @@ +#!/usr/bin/env bash +# A host includes term.h and nothing else. The header compiles alone as +# strict C11 and as C++, with only include/term on the include path. +set -euo pipefail +cd "${1:-$(dirname "$0")/..}" +cc=${CC:-cc} +cxx=${CXX:-c++} + +status=0 +echo '#include "term.h"' | "$cc" -std=c11 -Wall -Wextra -Wpedantic -Werror \ + -Iinclude/term -x c -fsyntax-only - || { echo "term.h: not valid C11" >&2; status=1; } +if command -v "$cxx" >/dev/null; then + echo '#include "term.h"' | "$cxx" -std=c++17 -Wall -Wextra -Wpedantic -Werror \ + -Iinclude/term -x c++ -fsyntax-only - || { echo "term.h: not valid C++" >&2; status=1; } +fi +exit $status diff --git a/tests/helpers/term_api.h b/tests/helpers/term_api.h new file mode 100644 index 0000000..bfa8d71 --- /dev/null +++ b/tests/helpers/term_api.h @@ -0,0 +1,182 @@ +#ifndef WSAP_TEST_TERM_API_H +#define WSAP_TEST_TERM_API_H + +/* + * Short forms of the public terminal API for tests. Cell reads go through + * a snapshot, the same way a renderer reads, so tests check only what the + * terminal shows. + */ + +#include "term/term.h" + +#include <stdbool.h> +#include <stddef.h> +#include <stdlib.h> +#include <string.h> + +/* The history size the tests assume. */ +#define TEST_SCROLLBACK 4096 + +static inline struct term *tt_new_opts(int cols, int rows, const struct term_callbacks *cb) +{ + struct term_options opt = { + .size = { .x = cols, .y = rows }, + .scrollback = TEST_SCROLLBACK, + }; + if (cb) + opt.callbacks = *cb; + return term_new(&opt); +} + +static inline struct term *tt_new(int cols, int rows) +{ + return tt_new_opts(cols, rows, NULL); +} + +static inline void tt_write(struct term *t, const char *s, size_t len) +{ + term_feed(t, s, len, 0); +} + +static inline void tt_puts(struct term *t, const char *s) +{ + term_feed(t, s, strlen(s), 0); +} + +static inline int tt_cols(const struct term *t) { return term_size(t).x; } +static inline int tt_rows(const struct term *t) { return term_size(t).y; } +static inline int tt_cursor_x(const struct term *t) { return term_cursor(t).pos.x; } +static inline int tt_cursor_y(const struct term *t) { return term_cursor(t).pos.y; } +static inline int tt_history(const struct term *t) { return term_scrollback(t).total; } +static inline int tt_view_offset(const struct term *t) { return term_scrollback(t).offset; } + +/* One snapshot serves every read. It follows whichever terminal a test + * reads last. */ +static struct term_snapshot *tt_snapshot_for(struct term *t) +{ + static struct term_snapshot *snap; + if (!snap) + snap = term_snapshot_new(); + term_snapshot_update(snap, t); + return snap; +} + +static const struct term_cell tt_blank_cell = { + .text = ' ', .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG, .width = 1, +}; + +/* A cell of the viewport. Out-of-range columns read as a blank cell. */ +static inline const struct term_cell *tt_view_cell(struct term *t, int col, int row) +{ + struct term_snapshot *s = tt_snapshot_for(t); + const struct term_cell *cells = term_snapshot_row(s, row); + if (!cells) + return NULL; + if (col < 0 || col >= term_snapshot_view(s)->size.x) + return &tt_blank_cell; + return &cells[col]; +} + +/* A cell of the live screen, whatever the viewport shows. */ +static inline const struct term_cell *tt_cell(struct term *t, int col, int row) +{ + int offset = tt_view_offset(t); + if (offset) + term_scroll_to(t, 0); + const struct term_cell *cell = tt_view_cell(t, col, row); + if (offset) + term_scroll_to(t, offset); + return cell; +} + +/* The code points of a cell's text: one, or a whole cluster. The snapshot + * that the last read used must still describe the terminal. */ +static inline int tt_cell_cps(struct term *t, const struct term_cell *cell, + uint32_t *cps, int cap) +{ + return term_snapshot_cluster(tt_snapshot_for(t), cell->text, cps, cap); +} + +/* The first code point of a cell's text. */ +static inline uint32_t tt_cell_cp(struct term *t, const struct term_cell *cell) +{ + uint32_t cp = 0; + if (!(cell->text & TERM_TEXT_CLUSTER)) + return cell->text; + tt_cell_cps(t, cell, &cp, 1); + return cp; +} + +static inline bool tt_is_cluster(const struct term_cell *cell) +{ + return (cell->text & TERM_TEXT_CLUSTER) != 0; +} + +/* The selected columns [begin, end) of a viewport row. */ +static inline struct term_span tt_row_selection(struct term *t, int row) +{ + return term_snapshot_row_selection(tt_snapshot_for(t), row); +} + +static inline bool tt_selected(struct term *t, int col, int row) +{ + struct term_span span = tt_row_selection(t, row); + return col >= span.begin && col < span.end; +} + +/* Select from one viewport cell to another, both inclusive. */ +static inline void tt_select(struct term *t, int x1, int y1, int x2, int y2) +{ + bool forward = y1 < y2 || (y1 == y2 && x1 <= x2); + struct term_vector2i a = { .x = x1, .y = y1 }, b = { .x = x2, .y = y2 }; + term_select_start(t, forward ? a : b, false, TERM_SELECT_CELL); + term_select_extend(t, forward ? b : a, true); +} + +/* The helpers below return new strings. The caller frees them. */ +/* NULL when nothing is selected. */ +static inline char *tt_selection_text(const struct term *t) +{ + size_t n = term_selection_text(t, NULL, 0); + if (!n) + return NULL; + char *text = malloc(n + 1); + if (text) + term_selection_text(t, text, n + 1); + return text; +} + +static inline char *tt_history_text(struct term *t, int max_rows) +{ + size_t n = term_history_text(t, max_rows, NULL, 0); + char *text = malloc(n + 1); + if (text) + term_history_text(t, max_rows, text, n + 1); + return text; +} + +/* NULL when there is no link. */ +static inline char *tt_link_at(struct term *t, int col, int row) +{ + struct term_vector2i cell = { .x = col, .y = row }; + size_t n = term_link_at(t, cell, NULL, 0); + if (!n) + return NULL; + char *url = malloc(n + 1); + if (url) + term_link_at(t, cell, url, n + 1); + return url; +} + +static inline bool tt_resize(struct term *t, int cols, int rows) +{ + return term_resize(t, (struct term_vector2i){ .x = cols, .y = rows }); +} + +/* The ARGB value of a color in the terminal's current palette. */ +static inline uint32_t tt_color(struct term *t, term_color c) +{ + return term_view_color(term_snapshot_view(tt_snapshot_for(t)), c); +} + +#endif diff --git a/tests/hist_test.c b/tests/hist_test.c new file mode 100644 index 0000000..be9fd3f --- /dev/null +++ b/tests/hist_test.c @@ -0,0 +1,765 @@ +#include "hist.h" +#include "hist_view.h" +#include "term/unicode/grapheme.h" + +#include <assert.h> +#include <limits.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static uint64_t lines(const struct hist *h) +{ + return hist_end_id(h) - hist_first_id(h); +} + +static struct cell dc(uint32_t c) +{ + return (struct cell){ .c = c, .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG }; +} + +static void test_init_free(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + assert(lines(&h) == 0); + assert(hist_first_id(&h) == 0); + assert(!hist_is_open(&h)); + term_hist_free(&h); +} + +static void test_ascii_roundtrip(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + + const char *msg = "Hello, world!"; + term_hist_push_ascii(&h, msg, (int)strlen(msg), NULL, false); + assert(lines(&h) == 1); + assert(term_hist_line_rows(&h, 0, 80) == 1); + assert(term_hist_line_rows(&h, 0, 5) == 3); + + struct cell out[80]; + int n = term_hist_decode_segment(&h, 0, 0, 80, out, NULL); + assert(n == (int)strlen(msg)); + for (int i = 0; i < n; i++) { + assert(out[i].c == (uint32_t)msg[i]); + assert(out[i].fg == TERM_COLOR_FG); + assert(out[i].bg == TERM_COLOR_BG); + assert(out[i].flags == 0); + } + + assert(term_hist_decode_segment(&h, 0, 0, 5, out, NULL) == 5); + assert(out[0].c == 'H' && out[4].c == 'o'); + assert(term_hist_decode_segment(&h, 0, 1, 5, out, NULL) == 5); + assert(out[0].c == ',' && out[4].c == 'r'); + assert(term_hist_decode_segment(&h, 0, 2, 5, out, NULL) == 3); + assert(out[0].c == 'l' && out[2].c == '!'); + + term_hist_free(&h); +} + +static void test_attributes_roundtrip(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + + struct cell row[10]; + for (int i = 0; i < 10; i++) { + row[i] = (struct cell){ + .c = 'A' + i, + .fg = term_color_rgb(i * 10, i * 20, i * 25), + .bg = term_color_rgb(255 - i * 10, 0, 0), + .flags = (i % 2 == 0) ? CELL_BOLD : CELL_ITALIC, + .pad = (uint16_t)(100 + i), + }; + } + term_hist_push_row(&h, row, 10, false, NULL); + assert(lines(&h) == 1); + + struct cell out[10]; + assert(term_hist_decode_segment(&h, 0, 0, 10, out, NULL) == 10); + for (int i = 0; i < 10; i++) { + assert(out[i].c == row[i].c); + assert(out[i].fg == row[i].fg); + assert(out[i].bg == row[i].bg); + assert(out[i].flags == row[i].flags); + assert(out[i].pad == row[i].pad); + } + term_hist_free(&h); +} + +static void test_wide_and_wrap(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + + struct cell row[4] = { dc('A'), dc(0x4e16), dc(' '), dc('B') }; + row[1].flags = CELL_WIDE; + row[2].flags = CELL_TRAIL; + term_hist_push_row(&h, row, 4, false, NULL); + + assert(term_hist_line_rows(&h, 0, 4) == 1); + /* Width 2: 'A' and a pad, then the wide cell, then 'B'. */ + assert(term_hist_line_rows(&h, 0, 2) == 3); + + struct cell out[4]; + assert(term_hist_decode_segment(&h, 0, 0, 2, out, NULL) == 2); + assert(out[0].c == 'A'); + assert(out[1].flags & CELL_WRAP_PAD); + assert(term_hist_decode_segment(&h, 0, 1, 2, out, NULL) == 2); + assert(out[0].c == 0x4e16 && (out[0].flags & CELL_WIDE)); + assert(out[1].flags & CELL_TRAIL); + assert(term_hist_decode_segment(&h, 0, 2, 2, out, NULL) == 1); + assert(out[0].c == 'B'); + term_hist_free(&h); +} + +static void test_clusters_roundtrip(void) +{ + struct gstore *g = term_gstore_new(NULL); + assert(g); + uint32_t cps[3] = { 'e', 0x0301, 0x0300 }; + uint32_t handle = term_gstore_intern(g, cps, 3); + + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + struct cell row[2] = { dc(handle), dc('X') }; + term_hist_push_row(&h, row, 2, false, g); + + struct cell out[2]; + assert(term_hist_decode_segment(&h, 0, 0, 2, out, g) == 2); + assert(grapheme_is_handle(out[0].c)); + uint32_t got[GRAPHEME_MAX_LEN]; + assert(term_gstore_expand(g, out[0].c, got, GRAPHEME_MAX_LEN) == 3); + assert(got[0] == cps[0] && got[1] == cps[1] && got[2] == cps[2]); + assert(out[1].c == 'X'); + + term_hist_free(&h); + term_gstore_free(g); +} + +static void test_open_lines_and_pop(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + + struct cell part1[3] = { dc('1'), dc('2'), dc('3') }; + term_hist_push_row(&h, part1, 3, true, NULL); + assert(lines(&h) == 1 && hist_is_open(&h)); + struct cell part2[2] = { dc('4'), dc('5') }; + term_hist_push_row(&h, part2, 2, false, NULL); + assert(lines(&h) == 1 && !hist_is_open(&h)); + + struct cell out[5]; + assert(term_hist_decode_segment(&h, 0, 0, 10, out, NULL) == 5); + for (int i = 0; i < 5; i++) + assert(out[i].c == (uint32_t)('1' + i)); + + struct cell popped[2]; + assert(term_hist_pop_tail_cells(&h, 2, popped, NULL) == 2); + assert(popped[0].c == '4' && popped[1].c == '5'); + assert(term_hist_decode_segment(&h, 0, 0, 10, out, NULL) == 3); + assert(out[0].c == '1' && out[2].c == '3'); + + /* Pop from an open line with a pen. */ + term_hist_close_open_line(&h); + struct cell color[2] = { + { .c = 'A', .fg = 0xff0000, .bg = 0x00ff00, .flags = CELL_BOLD }, + { .c = 'B', .fg = 0xff0000, .bg = 0x00ff00, .flags = CELL_BOLD }, + }; + term_hist_push_row(&h, color, 2, true, NULL); + assert(lines(&h) == 2); + struct cell pop_color[1]; + assert(term_hist_pop_tail_cells(&h, 1, pop_color, NULL) == 1); + assert(pop_color[0].c == 'B' && pop_color[0].fg == 0xff0000 && pop_color[0].bg == 0x00ff00); + assert(pop_color[0].flags & CELL_BOLD); + assert(term_hist_decode_segment(&h, 1, 0, 10, out, NULL) == 1); + assert(out[0].c == 'A' && out[0].fg == 0xff0000 && (out[0].flags & CELL_BOLD)); + + /* Popping a whole line returns its cells and removes it. */ + struct cell all[4]; + assert(term_hist_pop_line(&h, all, 4, NULL) == 1); + assert(all[0].c == 'A'); + assert(lines(&h) == 1 && !hist_is_open(&h)); + term_hist_free(&h); +} + +static void test_open_line_joins(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + + struct cell line1[2] = { dc('X'), dc('Y') }; + term_hist_push_row(&h, line1, 2, true, NULL); + term_hist_close_open_line(&h); + struct cell line2[2] = { dc('Z'), dc('W') }; + term_hist_push_row(&h, line2, 2, false, NULL); + assert(lines(&h) == 2); + + struct cell out[16]; + assert(term_hist_decode_segment(&h, 0, 0, 10, out, NULL) == 2); + assert(out[0].c == 'X' && out[1].c == 'Y'); + assert(term_hist_decode_segment(&h, 1, 0, 10, out, NULL) == 2); + assert(out[0].c == 'Z' && out[1].c == 'W'); + + /* ASCII joins a wide open line. */ + struct cell wide[2] = { { .c = 0x4e2d, .fg = 0x123456, .bg = TERM_COLOR_BG, .flags = CELL_WIDE }, + { .c = 0, .fg = 0x123456, .bg = TERM_COLOR_BG, .flags = CELL_TRAIL } }; + term_hist_push_row(&h, wide, 2, true, NULL); + term_hist_push_ascii(&h, "hello", 5, NULL, false); + assert(lines(&h) == 3); + assert(term_hist_decode_segment(&h, 2, 0, 10, out, NULL) == 7); + assert(out[0].c == 0x4e2d && out[0].fg == 0x123456); + assert(out[2].c == 'h' && out[2].fg == TERM_COLOR_FG); + term_hist_free(&h); +} + +static void test_eviction(void) +{ + struct hist h; + assert(term_hist_init(&h, 5, 64 * 1024 * 2)); + for (int i = 0; i < 10; i++) { + char buf[16]; + snprintf(buf, sizeof(buf), "line%d", i); + term_hist_push_ascii(&h, buf, (int)strlen(buf), NULL, false); + } + assert(lines(&h) == 5); + assert(hist_first_id(&h) == 5 && hist_end_id(&h) == 10); + + struct cell out[16]; + assert(term_hist_decode_segment(&h, 5, 0, 16, out, NULL) == 5); + assert(out[0].c == 'l' && out[4].c == '5'); + assert(term_hist_decode_segment(&h, 9, 0, 16, out, NULL) == 5); + assert(out[4].c == '9'); + assert(term_hist_decode_segment(&h, 4, 0, 16, out, NULL) == 0); + term_hist_free(&h); +} + +static void test_emoji_and_text_presentation(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + struct cell row[3] = { dc(0x2764), dc(' '), dc(0x2600) }; + row[0].flags = CELL_WIDE | CELL_EMOJI; + row[1].flags = CELL_TRAIL; + row[2].flags = CELL_TEXT; + term_hist_push_row(&h, row, 3, false, NULL); + + struct cell out[3]; + assert(term_hist_decode_segment(&h, 0, 0, 10, out, NULL) == 3); + assert(out[0].c == 0x2764 && (out[0].flags & CELL_WIDE) && (out[0].flags & CELL_EMOJI)); + assert(out[1].flags & CELL_TRAIL); + assert(out[2].c == 0x2600 && (out[2].flags & CELL_TEXT)); + term_hist_free(&h); +} + +static void test_width_one(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + term_hist_push_ascii(&h, "ABC", 3, NULL, false); + assert(term_hist_line_rows(&h, 0, 1) == 3); + + struct cell out[2]; + assert(term_hist_decode_segment(&h, 0, 0, 1, out, NULL) == 1 && out[0].c == 'A'); + assert(term_hist_decode_segment(&h, 0, 1, 1, out, NULL) == 1 && out[0].c == 'B'); + assert(term_hist_decode_segment(&h, 0, 2, 1, out, NULL) == 1 && out[0].c == 'C'); + + struct cell wide[2] = { dc(0x4e16), dc(' ') }; + wide[0].flags = CELL_WIDE; + wide[1].flags = CELL_TRAIL; + term_hist_push_row(&h, wide, 2, false, NULL); + assert(term_hist_line_rows(&h, 1, 1) == 1); + assert(term_hist_decode_segment(&h, 1, 0, 1, out, NULL) == 1); + assert(out[0].c == 0x4e16 && (out[0].flags & CELL_WIDE)); + term_hist_free(&h); +} + +static void test_blank_lines(void) +{ + struct hist h; + assert(term_hist_init(&h, 10, 64 * 1024 * 4)); + const char *first = "First line on page 0"; + term_hist_push_ascii(&h, first, (int)strlen(first), NULL, false); + term_hist_push_blank(&h, 10); + assert(lines(&h) == 10); + assert(hist_first_id(&h) == 1); + + for (int i = 0; i < 20; i++) { + char buf[32]; + snprintf(buf, sizeof(buf), "line %d", i); + term_hist_push_ascii(&h, buf, (int)strlen(buf), NULL, false); + } + struct cell out[32]; + assert(term_hist_decode_segment(&h, hist_end_id(&h) - 1, 0, 32, out, NULL) == 7); + assert(out[5].c == '1' && out[6].c == '9'); + + /* A blank row closes an open line and adds no line. */ + term_hist_push_ascii(&h, "open", 4, NULL, true); + uint64_t before = hist_end_id(&h); + term_hist_push_blank(&h, 1); + assert(!hist_is_open(&h) && hist_end_id(&h) == before); + + /* More blank rows than the limit keep the id count going. */ + term_hist_push_blank(&h, 1000); + assert(lines(&h) == 10 && hist_end_id(&h) == before + 1000); + term_hist_push_blank(&h, 5); + assert(lines(&h) == 10 && hist_end_id(&h) == before + 1005); + assert(term_hist_decode_segment(&h, hist_end_id(&h) - 1, 0, 32, out, NULL) == 0); + assert(term_hist_line_rows(&h, hist_end_id(&h) - 1, 32) == 1); + term_hist_free(&h); +} + +static void test_open_line_attrs_continuation(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + struct cell r1[5], r2[10]; + for (int i = 0; i < 5; i++) + r1[i] = (struct cell){ .c = '1' + i, .fg = term_color_rgb(255, 0, 0), .bg = TERM_COLOR_BG }; + for (int i = 0; i < 10; i++) + r2[i] = (struct cell){ .c = 'a' + i, .fg = i < 5 ? term_color_rgb(255, 0, 0) : term_color_rgb(0, 0, 255), + .bg = TERM_COLOR_BG }; + term_hist_push_row(&h, r1, 5, true, NULL); + term_hist_push_row(&h, r2, 10, false, NULL); + assert(lines(&h) == 1); + + struct cell out[15]; + assert(term_hist_decode_segment(&h, 0, 0, 15, out, NULL) == 15); + for (int i = 0; i < 10; i++) + assert(out[i].fg == term_color_rgb(255, 0, 0)); + for (int i = 10; i < 15; i++) + assert(out[i].fg == term_color_rgb(0, 0, 255)); + term_hist_free(&h); +} + +static void mark_link(void *ud, uint16_t link) +{ + ((int *)ud)[link]++; +} + +static void test_hyperlinks(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + struct pen_attr pen = pen_attr_default(); + pen.link = 42; + term_hist_push_ascii(&h, "Link text", 9, &pen, false); + struct cell row[2] = { dc('a'), dc('b') }; + row[1].pad = 7; + row[1].flags = CELL_LINK; + term_hist_push_row(&h, row, 2, true, NULL); + + struct cell out[16]; + assert(term_hist_decode_segment(&h, 0, 0, 16, out, NULL) == 9); + for (int i = 0; i < 9; i++) + assert(out[i].pad == 42); + + int marks[64] = { 0 }; + term_hist_mark_links(&h, mark_link, marks); + assert(marks[42] && marks[7]); + term_hist_free(&h); +} + +static void test_long_lines(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + + /* A line past the part limit keeps one id. */ + char chunk[1024]; + memset(chunk, 'X', sizeof(chunk)); + for (int i = 0; i < 64; i++) + term_hist_push_ascii(&h, chunk, sizeof(chunk), NULL, true); + term_hist_push_ascii(&h, chunk, 100, NULL, false); + assert(lines(&h) == 1); + assert(term_hist_line_cells(&h, 0) == 65636); + assert(term_hist_line_rows(&h, 0, 1000) == 66); + + struct cell out[1000]; + assert(term_hist_decode_segment(&h, 0, 64, 1000, out, NULL) == 1000); + for (int i = 0; i < 1000; i++) + assert(out[i].c == 'X'); + assert(term_hist_decode_segment(&h, 0, 65, 1000, out, NULL) == 636); + + memset(chunk, 'Y', sizeof(chunk)); + for (int i = 0; i < 64; i++) + term_hist_push_ascii(&h, chunk, sizeof(chunk), NULL, true); + term_hist_push_ascii(&h, chunk, 100, NULL, false); + term_hist_push_ascii(&h, "tail", 4, NULL, false); + assert(lines(&h) == 3); + assert(term_hist_line_rows(&h, 1, 1000) == 66); + assert(term_hist_decode_segment(&h, 1, 0, 1000, out, NULL) == 1000 && out[0].c == 'Y'); + assert(term_hist_decode_segment(&h, 2, 0, 1000, out, NULL) == 4 && out[0].c == 't'); + assert(term_hist_rows(&h, 1000) == 133); + + /* One long ASCII push splits into parts but reads back whole. */ + term_hist_free(&h); + assert(term_hist_init(&h, 100, 64 * 1024 * 8)); + int big_len = 150000; + char *big = malloc((size_t)big_len); + assert(big); + for (int i = 0; i < big_len; i++) + big[i] = (char)('A' + (i % 26)); + term_hist_push_ascii(&h, big, big_len, NULL, false); + assert(lines(&h) == 1); + assert(term_hist_line_rows(&h, 0, 1000) == 150); + for (int s = 0; s < 150; s++) { + assert(term_hist_decode_segment(&h, 0, s, 1000, out, NULL) == 1000); + for (int x = 0; x < 1000; x++) + assert(out[x].c == (uint32_t)big[s * 1000 + x]); + } + free(big); + term_hist_free(&h); +} + +static void test_page_limits(void) +{ + struct hist h; + struct cell out[1000]; + + /* One page: old lines leave to make room. */ + assert(term_hist_init(&h, 10000, HIST_PAGE_SIZE)); + char line64[64]; + memset(line64, 'M', sizeof(line64)); + for (int i = 0; i < 1200; i++) + term_hist_push_ascii(&h, line64, sizeof(line64), NULL, false); + assert(lines(&h) > 0 && lines(&h) < 1200); + assert(term_hist_decode_segment(&h, hist_end_id(&h) - 1, 0, 64, out, NULL) == 64); + assert(out[0].c == 'M'); + term_hist_free(&h); + + /* A line limit of one keeps the long line as one line. */ + assert(term_hist_init(&h, 1, 64 * 1024 * 4)); + char chunk[1024]; + memset(chunk, 'Z', sizeof(chunk)); + for (int i = 0; i < 65; i++) + term_hist_push_ascii(&h, chunk, sizeof(chunk), NULL, true); + assert(lines(&h) == 1 && hist_is_open(&h)); + term_hist_push_ascii(&h, "new", 3, NULL, false); + term_hist_push_ascii(&h, "next", 4, NULL, false); + assert(lines(&h) == 1); + assert(term_hist_decode_segment(&h, hist_first_id(&h), 0, 10, out, NULL) == 4); + term_hist_free(&h); + + /* A line longer than the arena drops its head but keeps its id. */ + assert(term_hist_init(&h, 10, HIST_PAGE_SIZE)); + term_hist_push_ascii(&h, "PREFIX", 6, NULL, true); + for (int i = 0; i < 100; i++) + term_hist_push_ascii(&h, chunk, sizeof(chunk), NULL, true); + term_hist_close_open_line(&h); + assert(lines(&h) == 1 && hist_first_id(&h) == 0); + assert(term_hist_line_cells(&h, 0) > 0 && term_hist_line_cells(&h, 0) < 100 * 1024); + assert(term_hist_decode_segment(&h, 0, 0, 6, out, NULL) == 6 && out[0].c == 'Z'); + term_hist_free(&h); + + /* A small open line moves to a new page instead of splitting. */ + assert(term_hist_init(&h, 10, HIST_PAGE_SIZE)); + for (int i = 0; i < 60; i++) + term_hist_push_ascii(&h, chunk, sizeof(chunk), NULL, false); + term_hist_push_ascii(&h, "PREFIX", 6, NULL, true); + for (int i = 0; i < 10; i++) + term_hist_push_ascii(&h, chunk, sizeof(chunk), NULL, true); + term_hist_close_open_line(&h); + uint64_t id = hist_end_id(&h) - 1; + assert(term_hist_decode_segment(&h, id, 0, 6, out, NULL) == 6); + assert(out[0].c == 'P' && out[5].c == 'X'); + assert(term_hist_line_rows(&h, id, 1000) == 11); + term_hist_free(&h); +} + +static void test_nul_and_wide_pop(void) +{ + struct hist h; + struct cell out[8]; + assert(term_hist_init(&h, 10, 64 * 1024)); + struct cell nul_row[4] = { dc('A'), dc(0), dc('B'), dc('C') }; + term_hist_push_row(&h, nul_row, 4, false, NULL); + assert(term_hist_decode_segment(&h, 0, 0, 4, out, NULL) == 4); + assert(out[0].c == 'A' && out[1].c == ' ' && out[2].c == 'B' && out[3].c == 'C'); + term_hist_free(&h); + + /* A pop does not split a wide cell. */ + assert(term_hist_init(&h, 10, 64 * 1024 * 4)); + struct cell wide[4] = { dc('A'), dc(0x4E2D), dc(0), dc('B') }; + wide[1].flags = CELL_WIDE; + wide[2].flags = CELL_TRAIL; + term_hist_push_row(&h, wide, 4, true, NULL); + struct cell popped[2]; + assert(term_hist_pop_tail_cells(&h, 2, popped, NULL) == 1); + assert(popped[0].c == 'B'); + assert(term_hist_decode_segment(&h, 0, 0, 4, out, NULL) == 3); + assert(out[0].c == 'A' && out[1].c == 0x4E2D && (out[2].flags & CELL_TRAIL)); + term_hist_free(&h); +} + +static void test_default_prefix_attributes(void) +{ + struct hist h; + assert(term_hist_init(&h, 100, 64 * 1024 * 4)); + struct cell row[8]; + for (int i = 0; i < 8; i++) { + row[i] = dc('a' + i); + if (i >= 4) { + row[i].fg = term_color_rgb(10, 20, 30); + row[i].bg = term_color_rgb(40, 50, 60); + row[i].flags = CELL_BOLD; + } + } + term_hist_push_row(&h, row, 8, false, NULL); + struct cell out[8]; + assert(term_hist_decode_segment(&h, 0, 0, 8, out, NULL) == 8); + for (int i = 0; i < 8; i++) { + assert(out[i].c == (uint32_t)('a' + i)); + assert(out[i].fg == row[i].fg && out[i].bg == row[i].bg && out[i].flags == row[i].flags); + } + + struct pen_attr colored = pen_attr_default(); + colored.fg = term_color_rgb(200, 100, 50); + term_hist_push_ascii(&h, "0123", 4, NULL, true); + term_hist_push_ascii(&h, "4567", 4, &colored, false); + assert(term_hist_decode_segment(&h, 1, 0, 8, out, NULL) == 8); + for (int i = 0; i < 8; i++) + assert(out[i].fg == (i < 4 ? TERM_COLOR_FG : term_color_rgb(200, 100, 50))); + term_hist_free(&h); +} + +static void test_utf8_and_disabled(void) +{ + struct hist h; + assert(term_hist_init(&h, 2, 64 * 1024 * 2)); + term_hist_push_ascii(&h, "first", 5, NULL, false); + struct cell p1[2] = { dc('A'), dc('B') }; + struct cell p2[2] = { dc(0x00e9), dc('C') }; + term_hist_push_row(&h, p1, 2, true, NULL); + term_hist_push_row(&h, p2, 2, false, NULL); + assert(lines(&h) == 2 && hist_first_id(&h) == 0); + struct cell out[8]; + assert(term_hist_decode_segment(&h, 1, 0, 8, out, NULL) == 4); + assert(out[0].c == 'A' && out[1].c == 'B' && out[2].c == 0x00e9 && out[3].c == 'C'); + term_hist_free(&h); + + assert(term_hist_init(&h, 0, 0)); + term_hist_push_ascii(&h, "ignored", 7, NULL, false); + term_hist_push_row(&h, p1, 2, false, NULL); + term_hist_push_blank(&h, 3); + assert(lines(&h) == 0); + term_hist_free(&h); +} + +/* Random rows of every kind, pushed and compared with a plain copy. */ +static void test_fuzz_rows(void) +{ + enum { LIMIT = 40, COLS = 24, ITER = 4000 }; + struct gstore *g = term_gstore_new(NULL); + uint32_t cps[2] = { 'a', 0x0301 }; + uint32_t cluster = term_gstore_intern(g, cps, 2); + struct hist h; + assert(term_hist_init(&h, LIMIT, 64 * 1024 * 2)); + + static struct cell ref[LIMIT + 1][COLS * 8]; + static int ref_len[LIMIT + 1]; + uint64_t ref_first = 0, ref_end = 0; + bool open = false; + unsigned state = 7; + for (int it = 0; it < ITER; it++) { + state = state * 1103515245u + 12345u; + int len = (int)((state >> 16) % (COLS + 1)); + bool wrapped = ((state >> 8) & 3) == 0; + struct cell row[COLS]; + int n = 0; + while (n < len) { + state = state * 1103515245u + 12345u; + unsigned k = (state >> 16) % 10; + struct cell c = dc('a' + (state >> 20) % 26); + if (k == 1) + c.fg = term_color_index((int)(state >> 24) % 16); + if (k == 2) + c.flags = CELL_BOLD; + if (k == 3) + c.c = 0xe9; + if (k == 4) + c.c = cluster; + if (k == 5 && n + 1 < len) { + c.c = 0x4e00 + (state >> 20) % 100; + c.flags = CELL_WIDE; + row[n++] = c; + c.c = 0; + c.flags = CELL_TRAIL; + } + row[n++] = c; + } + if (!open) { + if (ref_end - ref_first == LIMIT) + ref_first++; + ref_len[ref_end % (LIMIT + 1)] = 0; + ref_end++; + } + int slot = (int)((ref_end - 1) % (LIMIT + 1)); + if (!open && len == 0 && !wrapped) { + /* A blank line. */ + } else { + memcpy(&ref[slot][ref_len[slot]], row, (size_t)n * sizeof(row[0])); + ref_len[slot] += n; + } + term_hist_push_row(&h, row, n, wrapped, g); + open = wrapped; + assert(hist_first_id(&h) == ref_first && hist_end_id(&h) == ref_end); + } + + struct cell out[COLS * 8]; + for (uint64_t id = ref_first; id < ref_end; id++) { + int slot = (int)(id % (LIMIT + 1)); + int want = ref_len[slot]; + assert((int)term_hist_line_cells(&h, id) == want); + if (!want) + continue; + assert(term_hist_decode_segment(&h, id, 0, COLS * 8, out, g) == want); + for (int x = 0; x < want; x++) { + const struct cell *a = &ref[slot][x], *b = &out[x]; + if (a->flags & CELL_TRAIL) { + assert(b->flags & CELL_TRAIL); + continue; + } + assert(a->fg == b->fg && a->flags == b->flags); + if (grapheme_is_handle(a->c)) { + uint32_t got[GRAPHEME_MAX_LEN]; + assert(term_gstore_expand(g, b->c, got, GRAPHEME_MAX_LEN) == 2 && got[1] == 0x0301); + } else { + assert(a->c == b->c); + } + } + } + term_hist_free(&h); + term_gstore_free(g); +} + +static void test_large_wide_count(void) +{ + struct hist h; + struct hist_view view; + assert(term_hist_init(&h, 600, 4 * HIST_PAGE_SIZE)); + term_hist_view_init(&view); + struct cell row[4] = { dc('A'), dc(0x65e5), dc(0), dc('B') }; + row[1].flags = CELL_WIDE; + row[2].flags = CELL_TRAIL; + for (int i = 0; i < 400; i++) + term_hist_push_row(&h, row, 4, true, NULL); + term_hist_close_open_line(&h); + uint64_t exact = term_hist_rows(&h, 3); + term_hist_prepare_rows(&h, 3); + assert(term_hist_estimated_rows(&h, 3) < exact); + assert(term_hist_view_sync(&view, &h, 3)); + term_hist_view_scroll_to(&view, &h, 3, (struct line_pos){0}, INT_MAX); + term_hist_view_refine(&view, &h, 3, 3); + assert(exact > 254); + assert(term_hist_estimated_rows(&h, 3) == exact); + for (int i = 0; i < 300; i++) + term_hist_push_ascii(&h, "x", 1, NULL, false); + assert(term_hist_estimated_rows(&h, 3) == exact + 300); + assert(term_hist_view_sync(&view, &h, 3)); + for (int i = 0; i < 300; i++) + term_hist_push_ascii(&h, "x", 1, NULL, false); + assert(term_hist_estimated_rows(&h, 3) == 600); + term_hist_view_free(&view); + term_hist_free(&h); +} + +static void test_view_after_prefix_eviction(void) +{ + struct hist h; + struct hist_view view; + assert(term_hist_init(&h, 10, 2 * HIST_PAGE_SIZE)); + term_hist_view_init(&view); + char bytes[1000]; + for (int i = 0; i < 100; i++) { + memset(bytes, 'A' + i % 26, sizeof(bytes)); + term_hist_push_ascii(&h, bytes, sizeof(bytes), NULL, true); + } + term_hist_view_scroll_to(&view, &h, 1000, (struct line_pos){0}, 10); + uint64_t id; + int seg, screen_y; + struct cell before[1000], after[1000]; + assert(term_hist_view_row_pos(&view, &h, 1000, 3, (struct line_pos){0}, + 0, &id, &seg, &screen_y)); + assert(term_hist_decode_segment(&h, id, seg, 1000, before, NULL) == 1000); + for (int i = 100; i < 140; i++) { + memset(bytes, 'A' + i % 26, sizeof(bytes)); + term_hist_push_ascii(&h, bytes, sizeof(bytes), NULL, true); + term_hist_view_note_push(&view, &h, 1000); + } + assert(term_hist_line_cells(&h, 0) < 140000); + assert(term_hist_view_row_pos(&view, &h, 1000, 3, (struct line_pos){0}, + 0, &id, &seg, &screen_y)); + assert(term_hist_decode_segment(&h, id, seg, 1000, after, NULL) == 1000); + for (int i = 0; i < 1000; i++) + assert(before[i].c == after[i].c); + term_hist_view_rewrap(&view, &h, 500); + assert(term_hist_view_row_pos(&view, &h, 500, 3, (struct line_pos){0}, + 0, &id, &seg, &screen_y)); + assert(term_hist_decode_segment(&h, id, seg, 500, after, NULL) == 500); + for (int i = 0; i < 500; i++) + assert(before[i].c == after[i].c); + term_hist_view_free(&view); + term_hist_free(&h); +} + +static void test_short_lines_after_eviction(void) +{ + struct hist h; + assert(term_hist_init(&h, 64, 64 * 1024 * 4)); + term_hist_prepare_rows(&h, 80); + const char text[] = "abc\n\nx\npqrs\n\nz\n"; + const int offsets[] = {0, 4, 5, 7, 12, 13}; + const int lengths[] = {3, 0, 1, 4, 0, 1}; + struct hist_ascii_line batch[6]; + for (int i = 0; i < 6; i++) + batch[i] = (struct hist_ascii_line){ + .ptr = text + offsets[i], .len = (uint16_t)lengths[i], + }; + for (int pass = 0; pass < 20000; pass++) + term_hist_push_ascii_lines(&h, batch, 6, NULL); + assert(lines(&h) == 64); + assert(term_hist_rows(&h, 80) == 64); + for (uint64_t id = hist_first_id(&h); id < hist_end_id(&h); id++) { + int k = (int)(id % 6); + struct cell out[80]; + assert(term_hist_decode_segment(&h, id, 0, 80, out, NULL) == lengths[k]); + for (int x = 0; x < lengths[k]; x++) { + assert(out[x].c == (uint32_t)text[offsets[k] + x]); + assert(out[x].fg == TERM_COLOR_FG); + assert(out[x].bg == TERM_COLOR_BG); + } + } + term_hist_free(&h); +} + +int main(void) +{ + test_short_lines_after_eviction(); + test_large_wide_count(); + test_view_after_prefix_eviction(); + test_init_free(); + test_ascii_roundtrip(); + test_attributes_roundtrip(); + test_wide_and_wrap(); + test_clusters_roundtrip(); + test_open_lines_and_pop(); + test_open_line_joins(); + test_eviction(); + test_width_one(); + test_emoji_and_text_presentation(); + test_blank_lines(); + test_open_line_attrs_continuation(); + test_hyperlinks(); + test_long_lines(); + test_page_limits(); + test_nul_and_wide_pop(); + test_default_prefix_attributes(); + test_utf8_and_disabled(); + test_fuzz_rows(); + + printf("PASS: hist tests\n"); + return 0; +} diff --git a/tests/history_text_test.c b/tests/history_text_test.c new file mode 100644 index 0000000..88ca6aa --- /dev/null +++ b/tests/history_text_test.c @@ -0,0 +1,135 @@ +#include "term/term.h" +#include "term_api.h" + +#include <assert.h> +#include <pthread.h> +#include <stdatomic.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static struct term *new_term(int cols) +{ + struct term *term = term_new(&(struct term_options){ + .size = { .x = cols, .y = 2 }, + .scrollback = 100, + }); + assert(term); + return term; +} + +static void feed(struct term *term, const char *text) +{ + term_feed(term, text, strlen(text), 0); +} + +static void expect_history(struct term *term, int rows, const char *expected) +{ + char *text = tt_history_text(term, rows); + assert(text); + if (strcmp(text, expected)) { + fprintf(stderr, "history: got \"%s\" want \"%s\"\n", text, expected); + assert(0); + } + free(text); +} + +static void unicode_and_selection(void) +{ + struct term *term = new_term(16); + expect_history(term, 100, ""); + feed(term, "é👩💻界\r\nnext\r\nlast"); + expect_history(term, 100, "é👩💻界\n"); + expect_history(term, 0, ""); + term_select_start(term, (struct term_vector2i){0}, false, TERM_SELECT_CELL); + term_select_extend(term, (struct term_vector2i){ .x = 4 }, false); + char *before = tt_selection_text(term); + assert(before); + expect_history(term, 100, "é👩💻界\n"); + char *after = tt_selection_text(term); + assert(after && !strcmp(before, after)); + free(before); + free(after); + feed(term, "\033[?1049h"); + expect_history(term, 100, ""); + feed(term, "\033[?1049l"); + expect_history(term, 100, "é👩💻界\n"); + term_free(term); +} + +static void limits_and_scrolled_view(void) +{ + struct term *term = new_term(8); + feed(term, "one\r\ntwo\r\nthree\r\nfour\r\nfive"); + expect_history(term, 100, "one\ntwo\nthree\n"); + expect_history(term, 2, "two\nthree\n"); + term_scroll(term, 2); + expect_history(term, 100, "one\n"); + term_scroll(term, 100); + expect_history(term, 100, ""); + term_scroll_to(term, 0); + assert(term_resize(term, (struct term_vector2i){ .x = 12, .y = 2 })); + expect_history(term, 100, "one\ntwo\nthree\n"); + term_free(term); +} + +static void wrapped_rows(void) +{ + struct term *term = new_term(4); + feed(term, "abcdefghi\r\nx\r\ny"); + expect_history(term, 100, "abcdefghi\n"); + expect_history(term, 2, "efghi\n"); + term_free(term); + + term = new_term(4); + feed(term, "abcdefghijklm"); + expect_history(term, 100, "abcdefgh"); + term_free(term); +} + +struct writer { + struct term *term; + _Atomic bool done; +}; + +static const char unicode_line[] = "é👩💻界\r\n"; +static const char history_line[] = "é👩💻界\n"; + +static void *write_history(void *userdata) +{ + struct writer *writer = userdata; + for (int i = 0; i < 1000; i++) + feed(writer->term, unicode_line); + atomic_store(&writer->done, true); + return NULL; +} + +static void concurrent_history(void) +{ + struct term *term = new_term(16); + struct writer writer = { .term = term }; + pthread_t thread; + assert(pthread_create(&thread, NULL, write_history, &writer) == 0); + do { + char *text = tt_history_text(term, 10); + assert(text); + size_t n = strlen(text); + size_t line_len = sizeof(history_line) - 1; + assert(n % line_len == 0 && n / line_len <= 10); + for (size_t offset = 0; offset < n; offset += line_len) + assert(!memcmp(text + offset, history_line, line_len)); + free(text); + } while (!atomic_load(&writer.done)); + assert(pthread_join(thread, NULL) == 0); + term_free(term); +} + +int main(void) +{ + unicode_and_selection(); + limits_and_scrolled_view(); + wrapped_rows(); + concurrent_history(); + puts("PASS: history text"); + return 0; +} diff --git a/tests/input_test.c b/tests/input_test.c new file mode 100644 index 0000000..6ee2caa --- /dev/null +++ b/tests/input_test.c @@ -0,0 +1,353 @@ +#include "term/term.h" +#include <assert.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static struct term_keyboard_state st; +static struct term_modes m; +static char out[128]; +static size_t out_len; + +static void mode(unsigned kitty_flags) +{ + st = (struct term_keyboard_state){0}; + m = (struct term_modes){ .keyboard = kitty_flags, .grid = { .x = 80, .y = 24 } }; +} + +static void enc(struct term_key_event ev) +{ + if (!ev.action) ev.action = TERM_KEY_PRESS; + term_keyboard_update(&st, &ev); + struct term_input_result r = + term_input_encode_key(&st, &ev, &m, out, sizeof(out)); + out_len = r.len; +} +#define K(...) enc((struct term_key_event){ __VA_ARGS__ }) + +static void typed(uint32_t cp, unsigned mods) +{ + K(.codepoint = cp, .base = cp, .text = cp, .mods = mods); +} + +static bool consumed; +static void enc_mouse(struct term_mouse_event ev) +{ + if (ev.pixel.x == 0 && ev.pixel.y == 0) + ev.pixel = ev.pos; + struct term_input_result r = + term_input_encode_mouse(&ev, &m, out, sizeof(out)); + out_len = r.len; + consumed = r.status != TERM_INPUT_UNHANDLED; +} +#define MOUSE(x_, y_, ...) enc_mouse((struct term_mouse_event){ .pos = { .x = (x_), .y = (y_) }, __VA_ARGS__ }) + +static void expect_at(const char *s, int line) +{ + if (out_len != strlen(s) || memcmp(out, s, out_len)) { + fprintf(stderr, "line %d: expected %s, got %.*s (%zu bytes)\n", line, s, + (int)out_len, out, out_len); + abort(); + } +} +#define expect(s) expect_at(s, __LINE__) + +int main(void) +{ + mode(0); + K(.key = TERM_KEY_UP); expect("\033[A"); + K(.key = TERM_KEY_HOME); expect("\033[H"); + K(.key = TERM_KEY_DELETE); expect("\033[3~"); + K(.key = TERM_KEY_PAGE_UP); expect("\033[5~"); + K(.key = TERM_KEY_F1); expect("\033OP"); + K(.key = TERM_KEY_F5); expect("\033[15~"); + K(.key = TERM_KEY_F12); expect("\033[24~"); + K(.key = TERM_KEY_RETURN); expect("\r"); + K(.key = TERM_KEY_ESCAPE); expect("\033"); + K(.key = TERM_KEY_TAB); expect("\t"); + K(.key = TERM_KEY_TAB, .mods = TERM_MOD_SHIFT); expect("\033[Z"); + K(.key = TERM_KEY_BACKSPACE); expect("\177"); + K(.key = TERM_KEY_RIGHT, .mods = TERM_MOD_CTRL); expect("\033[1;5C"); + K(.key = TERM_KEY_F5, .mods = TERM_MOD_SHIFT); expect("\033[15;2~"); + K(.key = TERM_KEY_DELETE, .mods = TERM_MOD_CTRL); expect("\033[3;5~"); + + m.backarrow_bs = true; + K(.key = TERM_KEY_BACKSPACE); expect("\010"); + m.backarrow_bs = false; + + m.app_cursor = true; + K(.key = TERM_KEY_UP); expect("\033OA"); + K(.key = TERM_KEY_END); expect("\033OF"); + + K(.key = TERM_KEY_UP, .mods = TERM_MOD_CTRL); expect("\033[1;5A"); + m.app_cursor = false; + + typed('a', 0); expect("a"); + typed(0x00E9, 0); expect("\xc3\xa9"); + K(.codepoint = 'a', .shifted = 'A', .base = 'a', .text = 'A', + .mods = TERM_MOD_SHIFT); expect("A"); + K(.action = TERM_KEY_RELEASE, .codepoint = 'a', .text = 'a'); expect(""); + + typed('c', TERM_MOD_CTRL); expect("\003"); + typed('a', TERM_MOD_CTRL); expect("\001"); + typed('\\', TERM_MOD_CTRL); expect("\034"); + typed(' ', TERM_MOD_CTRL); assert(out_len == 1 && out[0] == 0); + + K(.codepoint = '2', .shifted = '@', .base = '2', .text = '@', + .mods = TERM_MOD_CTRL | TERM_MOD_SHIFT); + assert(out_len == 1 && out[0] == 0); + + typed('b', TERM_MOD_ALT); expect("\033b"); + typed('b', TERM_MOD_ALT | TERM_MOD_CTRL); expect("\033\002"); + K(.key = TERM_KEY_BACKSPACE, .mods = TERM_MOD_ALT); expect("\033\177"); + K(.key = TERM_KEY_RETURN, .mods = TERM_MOD_ALT); expect("\033\r"); + + K(.key = TERM_KEY_LEFT, .mods = TERM_MOD_ALT); expect("\033[1;3D"); + + K(.codepoint = 'b', .base = 'b', .text = 0x222B, .text_deferred = true); + expect(""); + + K(.key = TERM_KEY_KP_DELETE, .keypad = TERM_KEY_KP_DECIMAL); expect("\033[3~"); + K(.key = TERM_KEY_KP_ENTER, .keypad = TERM_KEY_KP_ENTER); expect("\r"); + K(.key = TERM_KEY_KP_LEFT, .keypad = TERM_KEY_KP_4); expect("\033[D"); + + m.app_keypad = true; + K(.key = TERM_KEY_KP_END, .keypad = TERM_KEY_KP_1); expect("\033Oq"); + K(.key = TERM_KEY_KP_ENTER, .keypad = TERM_KEY_KP_ENTER); expect("\033OM"); + K(.key = TERM_KEY_KP_ADD, .keypad = TERM_KEY_KP_ADD); expect("\033Ok"); + m.app_keypad = false; + + mode(1); + K(.key = TERM_KEY_ESCAPE); expect("\033[27u"); + K(.key = TERM_KEY_DELETE); expect("\033[3~"); + K(.key = TERM_KEY_F3); expect("\033[13~"); + K(.key = TERM_KEY_KP_ENTER, .keypad = TERM_KEY_KP_ENTER); expect("\033[57414u"); + typed('i', TERM_MOD_CTRL); expect("\033[105;5u"); + K(.key = TERM_KEY_RETURN, .mods = TERM_MOD_CTRL); expect("\033[13;5u"); + + typed('a', 0); expect("a"); + K(.key = TERM_KEY_RETURN); expect("\r"); + + K(.key = TERM_KEY_UP, .action = TERM_KEY_RELEASE); expect(""); + K(.key = TERM_KEY_LSHIFT); expect(""); + + mode(1 | 4); + K(.codepoint = 'a', .shifted = 'A', .base = 'a', .text = 'A', + .mods = TERM_MOD_CTRL | TERM_MOD_SHIFT); expect("\033[97:65;6u"); + + mode(1 | 2 | 8); + K(.key = TERM_KEY_ESCAPE); expect("\033[27;1u"); + K(.key = TERM_KEY_ESCAPE, .action = TERM_KEY_REPEAT); expect("\033[27;1:2u"); + K(.key = TERM_KEY_ESCAPE, .action = TERM_KEY_RELEASE); expect("\033[27;1:3u"); + K(.key = TERM_KEY_UP, .action = TERM_KEY_REPEAT); expect("\033[1;1:2A"); + + mode(31); + K(.codepoint = 1092, .base = 'a', .text = 1092); + expect("\033[1092::97;1;1092u"); + K(.codepoint = 'a', .shifted = 'A', .base = 'a', .text = 'A', + .mods = TERM_MOD_SHIFT); expect("\033[97:65;2;65u"); + + K(.action = TERM_KEY_RELEASE, .codepoint = 'a', .base = 'a', .text = 'a'); + expect("\033[97;1:3u"); + + mode(1 | 2 | 8); + K(.key = TERM_KEY_LSHIFT, .mods = TERM_MOD_SHIFT); expect("\033[57441;2u"); + K(.key = TERM_KEY_RSHIFT, .mods = TERM_MOD_SHIFT); expect("\033[57447;2u"); + K(.key = TERM_KEY_LSHIFT, .action = TERM_KEY_RELEASE, .mods = TERM_MOD_SHIFT); + expect("\033[57441;2:3u"); + K(.key = TERM_KEY_RSHIFT, .action = TERM_KEY_RELEASE, .mods = TERM_MOD_SHIFT); + expect("\033[57447;1:3u"); + + K(.key = TERM_KEY_ISO_LEVEL3_SHIFT); expect("\033[57453;1u"); + K(.key = TERM_KEY_ISO_LEVEL5_SHIFT); expect("\033[57454;1u"); + K(.key = TERM_KEY_CAPS_LOCK); expect("\033[57358;65u"); + + mode(1 | 2 | 4); + K(.key = TERM_KEY_CAPS_LOCK); expect(""); + K(.key = TERM_KEY_CAPS_LOCK, .action = TERM_KEY_RELEASE); expect(""); + + mode(1 | 2 | 4 | 8); + K(.key = TERM_KEY_CAPS_LOCK); expect("\033[57358;65u"); + K(.key = TERM_KEY_CAPS_LOCK, .action = TERM_KEY_RELEASE, + .mods = TERM_MOD_CAPS); expect("\033[57358;65:3u"); + + mode(1); + K(.key = TERM_KEY_LEFT, .mods = TERM_MOD_NUM); expect("\033[1;129D"); + K(.key = TERM_KEY_LEFT); expect("\033[D"); + K(.key = TERM_KEY_F1, .mods = TERM_MOD_CAPS); expect("\033[1;65P"); + K(.codepoint = 'a', .base = 'a', .mods = TERM_MOD_CTRL | TERM_MOD_CAPS); + expect("\033[97;69u"); + K(.codepoint = 'a', .base = 'a', .mods = TERM_MOD_CTRL); + expect("\033[97;5u"); + K(.codepoint = 'a', .base = 'a', .text = 'A', .mods = TERM_MOD_CAPS); + expect("A"); + K(.key = TERM_KEY_RETURN, .mods = TERM_MOD_CAPS); expect("\r"); + + mode(1 | 8); + K(.codepoint = 'a', .base = 'a', .text = 'A', .mods = TERM_MOD_CAPS); + expect("\033[97;65u"); + K(.key = TERM_KEY_RETURN, .mods = TERM_MOD_CAPS); expect("\033[13;65u"); + + mode(0); + K(.key = TERM_KEY_LEFT, .mods = TERM_MOD_NUM); expect("\033[D"); + K(.key = TERM_KEY_F1, .mods = TERM_MOD_CAPS); expect("\033OP"); + K(.codepoint = 'a', .base = 'a', .mods = TERM_MOD_CTRL | TERM_MOD_CAPS); + expect("\001"); + + mode(0); + MOUSE(2, 1, .kind = TERM_MOUSE_PRESS); + assert(out_len == 0 && !consumed); + + m.mouse = TERM_MOUSE_NORMAL; + MOUSE(2, 1, .kind = TERM_MOUSE_PRESS); expect("\033[M\040\043\042"); + MOUSE(2, 1, .kind = TERM_MOUSE_RELEASE); expect("\033[M\043\043\042"); + MOUSE(2, 1, .kind = TERM_MOUSE_MOTION); expect(""); + + MOUSE(2, 1, .mods = TERM_MOD_SHIFT); + assert(out_len == 0 && !consumed); + + MOUSE(200, 99, .kind = TERM_MOUSE_PRESS); + assert(out_len == 0 && consumed); + + m.mouse_encoding = TERM_MOUSE_ENC_SGR; + MOUSE(2, 1, .kind = TERM_MOUSE_PRESS); expect("\033[<0;3;2M"); + MOUSE(2, 1, .kind = TERM_MOUSE_RELEASE); expect("\033[<0;3;2m"); + MOUSE(2, 1, .mods = TERM_MOD_CTRL); expect("\033[<16;3;2M"); + MOUSE(2, 1, .button = TERM_MOUSE_WHEEL_UP); expect("\033[<64;3;2M"); + m.mouse_encoding = TERM_MOUSE_ENC_SGR_PIXELS; + MOUSE(2, 1, .pixel = { .x = 25, .y = 17 }); expect("\033[<0;26;18M"); + m.mouse_encoding = TERM_MOUSE_ENC_URXVT; + MOUSE(2, 1, .kind = TERM_MOUSE_PRESS); expect("\033[32;3;2M"); + m.mouse_encoding = TERM_MOUSE_ENC_X10; + + m.mouse = TERM_MOUSE_BUTTON; + MOUSE(2, 1, .kind = TERM_MOUSE_MOTION, .button = 3); expect(""); + MOUSE(2, 1, .kind = TERM_MOUSE_MOTION, .button = 0); + expect("\033[M\100\043\042"); + m.mouse_encoding = TERM_MOUSE_ENC_SGR; + MOUSE(2, 1, .kind = TERM_MOUSE_MOTION, .button = 0, .mods = TERM_MOD_CTRL); + expect("\033[<48;3;2M"); + m.mouse = TERM_MOUSE_ANY; + MOUSE(2, 1, .kind = TERM_MOUSE_MOTION, .button = 3); expect("\033[<35;3;2M"); + m.mouse_encoding = TERM_MOUSE_ENC_X10; + MOUSE(2, 1, .button = TERM_MOUSE_WHEEL_UP); expect("\033[M\140\043\042"); + + m.mouse = TERM_MOUSE_X10; + MOUSE(2, 1, .kind = TERM_MOUSE_PRESS); expect("\033[M\040\043\042"); + MOUSE(2, 1, .kind = TERM_MOUSE_RELEASE); expect(""); + + { + struct term_keyboard_state s2 = {0}; + struct term_modes m2 = { .keyboard = 1 | 2 | 8, .grid = { .x = 80, .y = 24 } }; + char b2[64]; + + struct term_key_event sh = { + .key = TERM_KEY_LSHIFT, .action = TERM_KEY_PRESS, + .mods = TERM_MOD_SHIFT, + }; + term_keyboard_update(&s2, &sh); + struct term_key_event rsh = { + .key = TERM_KEY_RSHIFT, .action = TERM_KEY_PRESS, + .mods = TERM_MOD_SHIFT, + }; + term_keyboard_update(&s2, &rsh); + + struct term_key_event rel = { + .key = TERM_KEY_LSHIFT, .action = TERM_KEY_RELEASE, + .mods = TERM_MOD_SHIFT, + }; + term_keyboard_update(&s2, &rel); + struct term_input_result r = + term_input_encode_key(&s2, &rel, &m2, b2, sizeof(b2)); + assert(r.status == TERM_INPUT_BYTES); + assert(!memcmp(b2, "\033[57441;2:3u", r.len)); + + struct term_key_event rrel = { + .key = TERM_KEY_RSHIFT, .action = TERM_KEY_RELEASE, + .mods = TERM_MOD_SHIFT, + }; + term_keyboard_update(&s2, &rrel); + r = term_input_encode_key(&s2, &rrel, &m2, b2, sizeof(b2)); + assert(!memcmp(b2, "\033[57447;1:3u", r.len)); + } + + { + char tiny[4]; + struct term_input_result r = term_input_encode_key( + &st, + &(struct term_key_event){ + .key = TERM_KEY_DELETE, .action = TERM_KEY_PRESS, + .mods = TERM_MOD_CTRL | TERM_MOD_ALT | TERM_MOD_SHIFT, + }, + &(struct term_modes){ .grid = { .x = 80, .y = 24 } }, + tiny, sizeof(tiny)); + assert(r.status == TERM_INPUT_BUFFER_FULL && r.len == 0); + } + + { + struct term_keyboard_state s5 = {0}; + struct term_key_event bad = { + .key = UINT16_MAX, .action = TERM_KEY_PRESS, + }; + term_keyboard_update(&s5, &bad); + assert(!term_keyboard_key_down(&s5, TERM_KEY_LSHIFT)); + struct term_input_result r = + term_input_encode_key(&s5, &bad, &m, out, sizeof(out)); + assert(r.status == TERM_INPUT_UNHANDLED && r.len == 0); + + struct term_key_event bad_kp = { + .key = TERM_KEY_KP_1, .keypad = UINT16_MAX, + .action = TERM_KEY_PRESS, + }; + m.app_keypad = true; + r = term_input_encode_key(&s5, &bad_kp, &m, out, sizeof(out)); + assert(r.len == 0); + m.app_keypad = false; + } + + { + struct term_key_event deferred = { + .action = TERM_KEY_PRESS, .codepoint = 'a', .text = 'a', + .text_deferred = true, + }; + struct term_input_result r = term_input_encode_key( + &st, &deferred, &(struct term_modes){0}, out, sizeof(out)); + assert(r.status == TERM_INPUT_UNHANDLED && r.len == 0); + struct term_modes kitty = { .keyboard = TERM_KEYBOARD_ALL_KEYS }; + r = term_input_encode_key(&st, &deferred, &kitty, out, sizeof(out)); + assert(r.status == TERM_INPUT_BYTES && r.len == strlen("\033[97u")); + assert(!memcmp(out, "\033[97u", r.len)); + deferred.action = TERM_KEY_RELEASE; + r = term_input_encode_key(&st, &deferred, &kitty, out, sizeof(out)); + assert(r.status == TERM_INPUT_CONSUMED && r.len == 0); + } + m.mouse = TERM_MOUSE_ANY; + m.mouse_encoding = TERM_MOUSE_ENC_SGR; + MOUSE(2, 1, .button = TERM_MOUSE_BUTTON_EXTRA_1); expect("\033[<8;3;2M"); + MOUSE(2, 1, .button = TERM_MOUSE_WHEEL_LEFT); expect("\033[<66;3;2M"); + MOUSE(2, 1, .button = TERM_MOUSE_WHEEL_RIGHT); expect("\033[<67;3;2M"); + + { + /* A host sizes the paste buffer from the text alone. */ + static const char text[] = "ls\r\n\033[31m\u202erm\x7f\t\xe4\xb8\xad"; + static const char want[] = "\033[200~ls\r[31mrm\t\xe4\xb8\xad\033[201~"; + char buf[sizeof(text) - 1 + TERM_PASTE_EXTRA]; + struct term_modes bracketed = { .bracketed_paste = true }; + struct term_input_result r = + term_input_encode_paste(&bracketed, text, sizeof(text) - 1, buf, sizeof(buf)); + assert(r.status == TERM_INPUT_BYTES && r.len == sizeof(want) - 1); + assert(!memcmp(buf, want, r.len)); + + r = term_input_encode_paste(&bracketed, text, sizeof(text) - 1, buf, r.len - 1); + assert(r.status == TERM_INPUT_BUFFER_FULL && r.len == 0); + + r = term_input_encode_paste(&bracketed, "\033\a\u2066", 5, buf, sizeof(buf)); + assert(r.status == TERM_INPUT_CONSUMED && r.len == 0); + + struct term_modes plain = {0}; + r = term_input_encode_paste(&plain, "a\nb", 3, buf, 3); + assert(r.status == TERM_INPUT_BYTES && r.len == 3 && !memcmp(buf, "a\rb", 3)); + } + puts("PASS: terminal key, mouse, and paste encoding"); + return 0; +} diff --git a/tests/meson.build b/tests/meson.build new file mode 100644 index 0000000..42c442e --- /dev/null +++ b/tests/meson.build @@ -0,0 +1,128 @@ +thread_dep = dependency('threads') +dl_dep = cc.find_library('dl', required: false) +python = find_program('python3', native: true, version: '>=3.10') +bash = find_program('bash', native: true) + +# The tests check their steps with assert, so NDEBUG must stay off. +test_c_args = ['-UNDEBUG'] + +# The host terminal may preload a library into every child. Keep it out. +test_env = environment() +test_env.set('LD_PRELOAD', '') +test_env.set('DYLD_INSERT_LIBRARIES', '') + +# [name, sources, dependencies, suites] +term_tests = [ + ['unicode-test', files('unicode/unicode_test.c'), [], []], + ['term-input-test', files('input_test.c'), [], []], + ['term-test', files('term_test.c'), [thread_dep], ['isa']], + ['write-fuzz-test', files('write_fuzz_test.c'), [], ['isa']], + ['spec-test', files('spec_test.c'), [], []], + ['model-test', files('model_test.c', 'model.c'), [], ['isa']], + ['model-row-test', files('model_row_test.c', 'model.c'), [], []], + ['gstore-test', files('gstore_test.c'), [], []], + ['hist-test', files('hist_test.c'), [], []], + ['history-text-test', files('history_text_test.c'), [thread_dep], []], + ['view-test', files('view_test.c'), [], []], + ['resize-property-test', files('resize_property_test.c'), [], []], + ['term-resize-failure-test', files('resize_failure_test.c'), [dl_dep], ['extlib']], + ['dcs-test', files('dcs_test.c'), [], []], + ['osc-test', files('osc_test.c'), [], []], +] +test_helpers_inc = include_directories('helpers') + +term_test_exes = {} +foreach entry : term_tests + exe = executable(entry[0], entry[1], + c_args: test_c_args, + include_directories: test_helpers_inc, + dependencies: [term_internal_dep] + entry[2], + ) + term_test_exes += {entry[0]: exe} + test(entry[0], exe, + args: entry[0] == 'write-fuzz-test' ? [get_option('write_fuzz_seeds').to_string()] : [], + env: test_env, + suite: ['unit'] + entry[3], + timeout: 180, + ) +endforeach + +# Without -march, the core runs the best kernels the CPU supports. Run the +# tests again on the lower levels, so each one is covered. +if host_machine.cpu_family() == 'x86_64' and cc.get_argument_syntax() == 'gcc' and \ + cc.get_define('__AVX2__', args: get_option('arch_flag') != '' ? [get_option('arch_flag')] : []) == '' + foreach level : ['base', 'avx2'] + level_env = test_env + level_env.set('TERM_HEADLESS_ISA', level) + foreach name : ['term-test', 'spec-test', 'model-test', 'view-test', 'term-input-test', + 'hist-test', 'history-text-test'] + test(name + '-' + level, term_test_exes[name], env: level_env, + suite: ['unit', 'isa'], timeout: 180) + endforeach + endforeach +endif + +# The amalgamation is the core as one term.c, the way a host builds it. The +# tests that use only term.h run again against it. +term_amalgamation = custom_target( + 'term-amalgamation', + output: ['term.c', 'term.h'], + depfile: 'term-amalgamation.d', + command: [python, files('../tools/amalgamate.py'), + meson.project_source_root(), '@OUTDIR@', '@DEPFILE@'], +) +term_amalgamation_lib = static_library( + 'term-amalgamation', + term_amalgamation[0], + dependencies: [m_dep], + override_options: ['c_std=c11', 'warning_level=3', 'werror=true'], + build_by_default: false, +) +foreach entry : term_tests + if entry[0] in ['term-test', 'spec-test', 'view-test', 'dcs-test', 'osc-test', + 'history-text-test', 'term-input-test', 'model-test', 'model-row-test', + 'resize-property-test'] + test(entry[0] + '-amalgamation', + executable(entry[0] + '-amalgamation', entry[1], + c_args: test_c_args, + include_directories: [term_inc, test_helpers_inc], + link_with: term_amalgamation_lib, + dependencies: [m_dep] + entry[2], + build_by_default: false), + env: test_env, suite: ['unit', 'amalgamation'], timeout: 180) + endif +endforeach + +test('term-symbols-test', bash, + args: [files('symbols_test.sh'), term_headless_lib.get_static_lib(), + meson.project_source_root() / 'include' / 'term', + term_headless_lib.get_shared_lib()], + suite: 'unit', +) +test('term-header-test', bash, + args: [files('header_test.sh'), meson.project_source_root()], + suite: 'unit', +) +test('term-cross-test', bash, + args: [files('cross_test.sh'), meson.project_source_root()], + suite: 'portability', + timeout: 600, +) + +if get_option('kitty_tests') + if host_machine.system() != 'linux' or get_option('b_sanitize') != 'none' + error('Kitty tests require an unsanitized Linux build') + endif + kitty_bridge = shared_library('term', 'extern/kitty/screen_bridge.c', + dependencies: term_internal_dep, + ) + kitty_env = environment() + kitty_env.set('WSAP_KITTY_LIB', kitty_bridge.full_path()) + test('kitty-test', python, + args: files('extern/kitty/parser.py'), + depends: kitty_bridge, + env: kitty_env, + suite: 'kitty', + is_parallel: false, + ) +endif diff --git a/tests/model.c b/tests/model.c new file mode 100644 index 0000000..4fd3737 --- /dev/null +++ b/tests/model.c @@ -0,0 +1,2105 @@ +/* + * Reference model of the terminal. See model.h. + * + * Sources of truth, in order: + * - xterm and ECMA-48 for controls, escape sequences, and modes. + * - Unicode for UTF-8: each maximal invalid subpart becomes one U+FFFD. + * - This terminal's design where xterm has no rule: reflow on resize, the + * row length that reflow keeps, history, and grapheme clusters. + * + * The model is slow on purpose. It has no lazy rows, no rings, and no SIMD. + * + * Known differences from xterm follow. For each one the model matches the + * current terminal, and the macro names the phase that removes it. + */ +#include "model.h" +#include "term/unicode/unicode.h" + +#include <limits.h> +#include <stdarg.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +/* Phase 6: the history limit counts rows. It must count logical lines. */ +#define KNOWN_HISTORY_COUNTS_ROWS 0 +/* Phase 6: a column resize on the alternate screen also cuts the saved + * normal screen and the history to the new width. */ +#define KNOWN_ALT_RESIZE_CUTS_NORMAL 0 + +#define MAX_TABS 4096 +#define STR_KEEP 1024 + +struct cell { + uint32_t cp[GRAPHEME_MAX_LEN]; + uint8_t n, width; + uint16_t attrs; + term_color fg, bg; + bool pad; /* the blank that a wide glyph left at a row end */ +}; + +/* + * A row. len is the length that reflow keeps: it covers every cell that + * was written or erased in a color, and a default erase at the end of the + * row shortens it. Cells at or after len are default blanks. + */ +struct row { + int len; + bool wrapped; /* autowrap continued this row on the next one */ + struct cell c[]; +}; + +struct pen { + term_color fg, bg, ul; + uint16_t attrs; /* TERM_ATTR_* style bits and the underline style */ + bool inverse; +}; + +/* What DECSC saves. */ +struct saved { + int x, y; + bool pending, origin; + struct pen pen; + char g[2]; + int gl; +}; + +struct keyboard { + unsigned flags, saved[32], depth; +}; + +enum { GROUND, ESC, CSI, STR, STR_ESC, CHARSET, HASH, DOCS }; +enum { OSC, DCS, IGNORE }; + +struct model { + int cols, rows, limit; + struct row **scr; + struct row **hist; + int hist_len, hist_cap, hist_lines; + + /* The normal screen while the alternate screen is active. */ + bool alternate; + struct row **normal; + int normal_cols, normal_rows; + struct saved normal_cursor, normal_saved; + bool normal_visible, normal_blink; + int normal_style; + + int x, y; + bool pending; + struct pen pen; + char g[2]; + int gl; + struct saved saved; + bool autowrap, origin, insert, lnm, reverse, app_cursor, app_keypad; + bool backarrow, bracketed, focus, sync, visible, blink; + int style, mouse, mouse_enc, decsace; + int top, bottom; + bool tabs[MAX_TABS]; + struct keyboard keyboard[2]; + /* The last printed glyph, for REP and cluster joins. */ + uint32_t last[GRAPHEME_MAX_LEN]; + int last_n; + int prompt; /* screen row of the last OSC 133;A, or -1 */ + unsigned flags; + + int state, str_kind; + char slot; + int param[16], nparam, csi_phase; + uint16_t sub; + char prefix, inter; + char str[STR_KEEP]; + size_t str_len; + + int need; + uint32_t acc; + uint8_t lo, hi; + + char *reply; + size_t reply_len, reply_cap; + struct model_cell **out; + int out_cap, out_cols; +}; + +/* ---- cells and rows ---- */ + +static struct cell blank(term_color fg, term_color bg) +{ + return (struct cell){ .cp = { ' ' }, .n = 1, .width = 1, .fg = fg, .bg = bg }; +} + +static struct cell default_blank(void) +{ + return blank(TERM_COLOR_FG, TERM_COLOR_BG); +} + +static struct row *row_new(int cols) +{ + struct row *r = malloc(sizeof(*r) + (size_t)cols * sizeof(struct cell)); + if (!r) abort(); + r->len = 0; + r->wrapped = false; + for (int x = 0; x < cols; x++) r->c[x] = default_blank(); + return r; +} + +/* Cut len to n. The cells after it become default blanks. */ +static void row_cut(struct row *r, int n, int cols) +{ + if (n >= r->len) return; + for (int x = n; x < r->len && x < cols; x++) r->c[x] = default_blank(); + r->len = n; +} + +static bool pristine(const struct model *m) +{ + return m->pen.fg == TERM_COLOR_FG && m->pen.bg == TERM_COLOR_BG; +} + +static bool is_ri(uint32_t cp) +{ + return cp - 0x1F1E6u <= 0x19u && grapheme_regional_indicator(cp); +} + +/* ---- replies ---- */ + +static void __attribute__((format(printf, 2, 3))) reply(struct model *m, const char *fmt, ...) +{ + char tmp[256]; + va_list args; + va_start(args, fmt); + int n = vsnprintf(tmp, sizeof(tmp), fmt, args); + va_end(args); + if (n < 0) return; + if (m->reply_len + (size_t)n + 1 > m->reply_cap) { + m->reply_cap = (m->reply_len + (size_t)n + 1) * 2; + m->reply = realloc(m->reply, m->reply_cap); + if (!m->reply) abort(); + } + memcpy(m->reply + m->reply_len, tmp, (size_t)n); + m->reply_len += (size_t)n; +} + +/* ---- history ---- */ + +static int history_limit(int requested) +{ + return requested == 0 ? TERM_SCROLLBACK_DEFAULT : requested < 0 ? 0 : requested; +} + +static void hist_reflow_tail(struct model *m); + +static void hist_push(struct model *m, struct row *r) +{ + if (!m->limit) { + free(r); + return; + } + if (KNOWN_HISTORY_COUNTS_ROWS) { + if (m->hist_len == m->limit) { + free(m->hist[0]); + memmove(m->hist, m->hist + 1, (size_t)(m->hist_len - 1) * sizeof(*m->hist)); + m->hist_len--; + } + } else { + bool new_line = (m->hist_len == 0 || !m->hist[m->hist_len - 1]->wrapped); + if (new_line) { + if (m->hist_lines == m->limit) { + int evict_rows = 1; + while (evict_rows < m->hist_len && m->hist[evict_rows - 1]->wrapped) + evict_rows++; + for (int i = 0; i < evict_rows; i++) + free(m->hist[i]); + memmove(m->hist, m->hist + evict_rows, + (size_t)(m->hist_len - evict_rows) * sizeof(*m->hist)); + m->hist_len -= evict_rows; + m->hist_lines--; + } + m->hist_lines++; + } + } + if (m->hist_len == m->hist_cap) { + m->hist_cap = m->hist_cap ? m->hist_cap * 2 : 64; + m->hist = realloc(m->hist, (size_t)m->hist_cap * sizeof(*m->hist)); + if (!m->hist) abort(); + } + m->hist[m->hist_len++] = r; + if (!KNOWN_HISTORY_COUNTS_ROWS) + hist_reflow_tail(m); +} + +static void term_hist_clear(struct model *m) +{ + for (int i = 0; i < m->hist_len; i++) free(m->hist[i]); + m->hist_len = 0; + m->hist_lines = 0; +} + +/* ---- scrolling ---- */ + +static void erase_row(struct model *m, int y); + +/* Scroll the whole screen up one row. On the normal screen the top row + * goes to history. */ +static void scroll_up_full(struct model *m) +{ + struct row *top = m->scr[0]; + memmove(m->scr, m->scr + 1, (size_t)(m->rows - 1) * sizeof(*m->scr)); + m->scr[m->rows - 1] = row_new(m->cols); + erase_row(m, m->rows - 1); + if (m->alternate) { + free(top); + return; + } + if (m->prompt > 0) m->prompt--; + hist_push(m, top); +} + +static void scroll_region(struct model *m, int top, int bottom, int n, bool down, bool history) +{ + int height = bottom - top + 1; + if (n > height) n = height; + if (!down && history && top == 0 && bottom == m->rows - 1) { + for (int i = 0; i < n; i++) scroll_up_full(m); + return; + } + /* A region at the top of the normal screen saves the rows that leave + * it. The rows below the region stay in place. */ + if (!down && history && top == 0 && !m->alternate) { + for (int i = 0; i < n; i++) { + scroll_up_full(m); + scroll_region(m, bottom, m->rows - 1, 1, true, false); + } + return; + } + for (int i = 0; i < n; i++) { + if (down) { + free(m->scr[bottom]); + memmove(m->scr + top + 1, m->scr + top, (size_t)(height - 1) * sizeof(*m->scr)); + m->scr[top] = row_new(m->cols); + erase_row(m, top); + } else { + free(m->scr[top]); + memmove(m->scr + top, m->scr + top + 1, (size_t)(height - 1) * sizeof(*m->scr)); + m->scr[bottom] = row_new(m->cols); + erase_row(m, bottom); + } + } + if (!m->alternate && m->prompt >= top && m->prompt <= bottom) { + m->prompt += down ? n : -n; + if (m->prompt < top || m->prompt > bottom) m->prompt = -1; + } +} + +static void line_feed(struct model *m) +{ + if (m->y == m->bottom) { + scroll_region(m, m->top, m->bottom, 1, false, true); + } else if (m->y < m->rows - 1) { + m->y++; + } +} + +static void reverse_index(struct model *m) +{ + if (m->y == m->top) + scroll_region(m, m->top, m->bottom, 1, true, false); + else if (m->y > 0) + m->y--; +} + +/* ---- erasing ---- */ + +/* An erase that covers half of a wide glyph erases all of it. Erased cells + * take the pen colors without inverse and lose their attributes. */ +static void erase_cells(struct model *m, int y, int from, int to) +{ + struct row *r = m->scr[y]; + if (to > m->cols) to = m->cols; + bool plain = pristine(m); + if (plain && to > r->len) to = r->len; + if (from >= to) return; + if (from > 0 && r->c[from].width == 0) from--; + if (to < r->len && r->c[to].width == 0) to++; + if (to >= m->cols) r->wrapped = false; + if (!plain) { + for (int x = from; x < to; x++) r->c[x] = blank(m->pen.fg, m->pen.bg); + if (to > r->len) r->len = to; + return; + } + if (to < r->len) { + for (int x = from; x < to; x++) r->c[x] = default_blank(); + return; + } + row_cut(r, from, m->cols); + if (from == 0) r->wrapped = false; +} + +static void erase_row(struct model *m, int y) +{ + erase_cells(m, y, 0, m->cols); +} + +/* The last screen row that shows a glyph other than a space. */ +static int last_text_row(const struct model *m) +{ + for (int y = m->rows - 1; y >= 0; y--) + for (int x = 0; x < m->cols; x++) { + const struct cell *c = &m->scr[y]->c[x]; + if (c->n > 1 || (c->n == 1 && c->cp[0] > ' ')) return y; + } + return -1; +} + +static void erase_display(struct model *m, int mode) +{ + if (mode == 3) { + term_hist_clear(m); + return; + } + if (mode == 2) { + /* The normal screen keeps its text: rows up to the last one with + * text move to history first. */ + int last = m->alternate ? -1 : last_text_row(m); + for (int i = 0; i <= last; i++) scroll_up_full(m); + for (int y = 0; y < m->rows; y++) erase_row(m, y); + } else if (mode == 0) { + erase_cells(m, m->y, m->x, m->cols); + for (int y = m->y + 1; y < m->rows; y++) erase_row(m, y); + } else if (mode == 1) { + for (int y = 0; y < m->y; y++) erase_row(m, y); + erase_cells(m, m->y, 0, m->x + 1); + } +} + +static void erase_line(struct model *m, int mode) +{ + if (mode == 0) erase_cells(m, m->y, m->x, m->cols); + else if (mode == 1) erase_cells(m, m->y, 0, m->x + 1); + else if (mode == 2) erase_cells(m, m->y, 0, m->cols); +} + +/* ---- wide glyph repair ---- */ + +/* A write to [from, to) splits the wide glyphs at its edges. Their other + * halves become blanks that keep their colors. */ +static void split_edges(struct model *m, struct row *r, int from, int to) +{ + if (from >= to) return; + if (from > 0 && from < r->len && r->c[from].width == 0) + r->c[from - 1] = blank(r->c[from - 1].fg, r->c[from - 1].bg); + if (to < r->len && r->c[to].width == 0) + r->c[to] = blank(r->c[to].fg, r->c[to].bg); + (void)m; +} + +/* Blank every half of a wide glyph in [from, to) that lost its partner. */ +static void repair_row(struct model *m, struct row *r, int from, int to) +{ + if (from < 0) from = 0; + if (to > m->cols) to = m->cols; + for (int x = from; x < to; x++) { + struct cell *c = &r->c[x]; + if (c->width == 0 && (x == 0 || r->c[x - 1].width != 2)) + *c = blank(c->fg, c->bg); + if (c->width == 2 && m->cols > 1 && (x + 1 == m->cols || r->c[x + 1].width != 0)) + *c = blank(c->fg, c->bg); + } +} + +/* Writing at x past len first fills the gap with default blanks. */ +static void fill_gap(struct row *r, int x) +{ + if (x > r->len) r->len = x; +} + +/* ---- printing ---- */ + +static uint16_t presentation(const uint32_t *cps, int n) +{ + bool pictographic = unicode_extended_pictographic(cps[0]); + bool emoji = false; + for (int i = 1; i < n; i++) { + if (cps[i] == 0xFE0F) return TERM_ATTR_EMOJI; + if (cps[i] == 0xFE0E) return TERM_ATTR_TEXT; + if (cps[i] == 0x20E3 || grapheme_regional_indicator(cps[i]) || + (pictographic && (cps[i] == 0x200D || grapheme_emoji_modifier(cps[i]) || + (cps[i] >= 0xE0020 && cps[i] <= 0xE007F)))) + emoji = true; + } + return emoji ? TERM_ATTR_EMOJI : 0; +} + +static void wrap_to_next_row(struct model *m) +{ + m->scr[m->y]->wrapped = true; + m->x = 0; + line_feed(m); + m->pending = false; +} + +/* Put one glyph at the cursor with the pen colors and the given style. */ +static void put_glyph(struct model *m, const uint32_t *cps, int n, uint16_t attrs) +{ + if (n > 1) + attrs = (uint16_t)((attrs & ~(TERM_ATTR_EMOJI | TERM_ATTR_TEXT)) | presentation(cps, n)); + int width = unicode_width(cps[0]) == 2 ? 2 : 1; + if (attrs & TERM_ATTR_EMOJI) width = 2; + if (attrs & TERM_ATTR_TEXT) width = 1; + if (m->pending) wrap_to_next_row(m); + int right = m->cols - 1; + struct row *r = m->scr[m->y]; + /* A wide glyph that does not fit wraps and leaves a pad blank. */ + if (width == 2 && m->cols > 1 && m->x == right && m->autowrap) { + fill_gap(r, m->x); + split_edges(m, r, m->x, right + 1); + r->c[m->x] = default_blank(); + r->c[m->x].pad = true; + r->len = m->cols; + wrap_to_next_row(m); + r = m->scr[m->y]; + } + int x = m->x; + int span = width > right + 1 - x ? 1 : width; + fill_gap(r, x); + if (m->insert) { + int move = right + 1 - x - span; + if (move > 0) memmove(r->c + x + span, r->c + x, (size_t)move * sizeof(struct cell)); + r->len = r->len + span > m->cols ? m->cols : r->len + span; + repair_row(m, r, right, right + 1); + } + split_edges(m, r, x, x + span); + struct cell c = { + .n = (uint8_t)n, .width = (uint8_t)span, .attrs = attrs, + .fg = m->pen.inverse ? m->pen.bg : m->pen.fg, + .bg = m->pen.inverse ? m->pen.fg : m->pen.bg, + }; + memcpy(c.cp, cps, (size_t)n * sizeof(*cps)); + r->c[x] = c; + if (span == 2) { + r->c[x + 1] = c; + r->c[x + 1].n = 0; + r->c[x + 1].width = 0; + memset(r->c[x + 1].cp, 0, sizeof(c.cp)); + } + if (x + span > r->len) r->len = x + span; + if (x + span <= right) { + m->x = x + span; + } else { + m->x = right; + m->pending = m->autowrap; + } +} + +/* Marks, emoji modifiers, and spacing marks extend the glyph before them. */ +static bool joins(uint32_t cp, int width) +{ + if (cp < 0x300) return false; + return width == 0 || grapheme_emoji_modifier(cp) || + (width == 1 && unicode_spacing_mark(cp)); +} + +static bool cluster_joins(const uint32_t *cps, int n, uint32_t cp, int width) +{ + if (joins(cp, width)) return true; + if (cps[n - 1] == 0x200D) return unicode_extended_pictographic(cp); + return n == 1 && is_ri(cp) && is_ri(cps[0]); +} + +/* + * Try to add cp to the glyph left of the cursor. True when cp was used: + * joined, or dropped because it has no glyph to join. The glyph keeps its + * colors and style and is placed again, so it can grow to two cells. + */ +static bool extend_cluster(struct model *m, uint32_t cp, int width) +{ + bool after_join = m->last_n > 1 || (m->last_n == 1 && m->last[0] == 0x200D); + if (!(width == 0 || after_join || is_ri(cp) || joins(cp, width))) + return false; + struct row *r = m->scr[m->y]; + int x = m->pending ? m->cols - 1 : m->x - 1; + struct cell *prev = NULL; + if (x >= 0 && x < r->len) { + if (x > 0 && r->c[x].width == 0) x--; + prev = &r->c[x]; + } + if (!prev || !(prev->n > 1 || (prev->n == 1 && prev->cp[0] > ' '))) + return joins(cp, width); + if (!cluster_joins(prev->cp, prev->n, cp, width)) + return false; + if (prev->n >= GRAPHEME_MAX_LEN) + return true; + struct cell old = *prev; + uint32_t cps[GRAPHEME_MAX_LEN]; + memcpy(cps, old.cp, (size_t)old.n * sizeof(*cps)); + cps[old.n] = cp; + int n = old.n + 1; + int span = old.width == 2 && x + 1 < m->cols ? 2 : 1; + for (int i = 0; i < span; i++) r->c[x + i] = blank(old.fg, old.bg); + if (r->len == x + span) row_cut(r, x, m->cols); + m->x = x; + m->pending = false; + memcpy(m->last, cps, (size_t)n * sizeof(*cps)); + m->last_n = n; + put_glyph(m, cps, n, (uint16_t)(old.attrs & ~(TERM_ATTR_EMOJI | TERM_ATTR_TEXT))); + r = m->scr[m->y]; + x = m->pending ? m->cols - 1 : m->x - 1; + if (x > 0 && r->c[x].width == 0) x--; + r->c[x].fg = old.fg; + r->c[x].bg = old.bg; + if (r->c[x].width == 2 && x + 1 < m->cols) { + r->c[x + 1].fg = old.fg; + r->c[x + 1].bg = old.bg; + } + return true; +} + +static void esc_dispatch(struct model *m, unsigned char ch); + +static void emit(struct model *m, uint32_t cp) +{ + static const uint32_t dec_graphics[] = { + 0x00a0, + 0x25c6, 0x2592, 0x2409, 0x240c, 0x240d, 0x240a, 0x00b0, 0x00b1, + 0x2424, 0x240b, 0x2518, 0x2510, 0x250c, 0x2514, 0x253c, 0x23ba, + 0x23bb, 0x2500, 0x23bc, 0x23bd, 0x251c, 0x2524, 0x2534, 0x252c, + 0x2502, 0x2264, 0x2265, 0x03c0, 0x2260, 0x00a3, 0x00b7, + }; + /* C1 controls in UTF-8 form act as their ESC forms. */ + if (cp >= 0x80 && cp <= 0x9f) { + static const char c1[32] = { + [4] = 'D', [5] = 'E', [8] = 'H', [13] = 'M', + [16] = 'P', [27] = '[', [28] = '\\', [29] = ']', + }; + if (c1[cp - 0x80]) { + m->state = GROUND; + esc_dispatch(m, (unsigned char)c1[cp - 0x80]); + } + return; + } + if (m->g[m->gl] == '0' && cp >= '_' && cp <= '~') + cp = dec_graphics[cp - '_']; + m->last[0] = cp; + m->last_n = 1; + put_glyph(m, &cp, 1, m->pen.attrs); +} + +static void print_cp(struct model *m, uint32_t cp) +{ + if (cp < 0x80) { + emit(m, cp); + return; + } + int width = unicode_width(cp); + if (!extend_cluster(m, cp, width)) + emit(m, cp); +} + +/* ---- cursor ---- */ + +static int top_limit(const struct model *m) { return m->origin ? m->top : 0; } +static int bottom_limit(const struct model *m) { return m->origin ? m->bottom : m->rows - 1; } + +static void move_to(struct model *m, int x, int y) +{ + m->pending = false; + m->x = x < 0 ? 0 : x >= m->cols ? m->cols - 1 : x; + m->y = y < 0 ? 0 : y >= m->rows ? m->rows - 1 : y; +} + +/* CUU, CUD, CNL, CPL, and VPR stop at a margin when they start inside the + * scroll region (xterm, DEC STD 070). */ +static void move_v(struct model *m, int n, bool col0) +{ + int lo = 0, hi = m->rows - 1; + if (m->origin || (m->y >= m->top && m->y <= m->bottom)) { + lo = m->top; + hi = m->bottom; + } else if (m->y < m->top) { + hi = m->bottom; + } else { + lo = m->top; + } + long y = (long)m->y + n; + if (n < 0 && m->y >= lo && y < lo) y = lo; + if (n > 0 && m->y <= hi && y > hi) y = hi; + move_to(m, col0 ? 0 : m->x, (int)(y < 0 ? 0 : y > m->rows - 1 ? m->rows - 1 : y)); +} + +static void move_h(struct model *m, int n) +{ + long x = (long)m->x + n; + move_to(m, (int)(x < 0 ? 0 : x > m->cols - 1 ? m->cols - 1 : x), m->y); +} + +static struct saved save_cursor(const struct model *m) +{ + return (struct saved){ + .x = m->x, .y = m->y, .pending = m->pending, .origin = m->origin, + .pen = m->pen, .g = { m->g[0], m->g[1] }, .gl = m->gl, + }; +} + +static void restore_cursor(struct model *m, const struct saved *s) +{ + m->pen = s->pen; + m->origin = s->origin; + m->g[0] = s->g[0]; + m->g[1] = s->g[1]; + m->gl = s->gl; + move_to(m, s->x, s->y); + m->pending = s->pending; +} + +static struct saved initial_saved(void) +{ + return (struct saved){ + .pen = { .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG }, + .g = { 'B', 'B' }, + }; +} + +static void tabs_default(struct model *m) +{ + memset(m->tabs, 0, sizeof(m->tabs)); + for (int c = 8; c < MAX_TABS; c += 8) m->tabs[c] = true; +} + +/* ---- insert and delete ---- */ + +static void delete_chars(struct model *m, int n) +{ + struct row *r = m->scr[m->y]; + int x = m->x, cols = m->cols; + if (n > cols - x) n = cols - x; + int len = r->len; + split_edges(m, r, x, x + n); + if (len > x + n) { + memmove(r->c + x, r->c + x + n, (size_t)(len - x - n) * sizeof(struct cell)); + for (int i = len - n; i < len; i++) r->c[i] = default_blank(); + r->len = len - n; + } else if (len > x) { + row_cut(r, x, cols); + } + if (!pristine(m)) { + int from = cols - n < x ? x : cols - n; + fill_gap(r, from); + for (int i = from; i < cols; i++) r->c[i] = blank(m->pen.fg, m->pen.bg); + r->len = cols; + } + r->wrapped = false; + repair_row(m, r, x - 1, len); +} + +static void insert_chars(struct model *m, int n) +{ + struct row *r = m->scr[m->y]; + int x = m->x, cols = m->cols; + if (n > cols - x) n = cols - x; + fill_gap(r, x); + int len = r->len; + if (x > 0 && r->c[x].width == 0) { + r->c[x - 1] = blank(m->pen.fg, m->pen.bg); + r->c[x] = blank(m->pen.fg, m->pen.bg); + } + if (len > x) { + int move = len - x; + if (move > cols - x - n) move = cols - x - n; + if (move > 0) memmove(r->c + x + n, r->c + x, (size_t)move * sizeof(struct cell)); + } + for (int i = x; i < x + n; i++) r->c[i] = blank(m->pen.fg, m->pen.bg); + int new_len = len + n > cols ? cols : len + n; + if (pristine(m) && len <= x) new_len = len; + else if (new_len < x + n) new_len = x + n; + for (int i = new_len; i < cols; i++) r->c[i] = default_blank(); + r->len = new_len; + r->wrapped = false; + int end = x + n + 1 > len + 1 ? x + n + 1 : len + 1; + repair_row(m, r, x - 1, end); +} + +/* ---- SGR ---- */ + +static void pen_reset(struct pen *p) +{ + *p = (struct pen){ .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG }; +} + +static void set_underline(uint16_t *attrs, int style) +{ + if (style > 5) style = 1; + *attrs = (uint16_t)((*attrs & ~TERM_ATTR_UNDERLINE_MASK) | + (unsigned)style << TERM_ATTR_UNDERLINE_SHIFT); +} + +/* 38, 48, and 58. Returns the parameters it used after the first. */ +static int extended_color(const struct model *m, int i, term_color *color, bool *ok) +{ + *ok = false; + if (i + 1 >= m->nparam) return 0; + if (i + 2 < m->nparam && m->param[i + 1] == 5) { + int index = m->param[i + 2]; + if (index >= 0 && index <= 255) { + *color = term_color_index(index); + *ok = true; + } + return 2; + } + if (m->param[i + 1] != 2) return 0; + int colon = 0; + while (i + 2 + colon < m->nparam && i + 2 + colon < 16 && + (m->sub & (1u << (i + 2 + colon)))) + colon++; + int skip = colon >= 4 ? 1 : 0; + if (i + 4 + skip >= m->nparam) return 0; + *color = term_color_rgb((uint8_t)m->param[i + 2 + skip], (uint8_t)m->param[i + 3 + skip], + (uint8_t)m->param[i + 4 + skip]); + *ok = true; + return 4 + skip; +} + +static void sgr(struct model *m) +{ + struct pen *p = &m->pen; + if (m->nparam == 0) { + pen_reset(p); + return; + } + for (int i = 0; i < m->nparam; i++) { + int v = m->param[i]; + term_color color; + bool ok; + switch (v) { + case 0: pen_reset(p); break; + case 1: p->attrs |= TERM_ATTR_BOLD; break; + case 2: p->attrs |= TERM_ATTR_FAINT; break; + case 3: p->attrs |= TERM_ATTR_ITALIC; break; + case 4: { + int style = 1; + if (i + 1 < m->nparam && (m->sub & (1u << (i + 1)))) + style = m->param[++i]; + set_underline(&p->attrs, style); + break; + } + case 7: p->inverse = true; break; + case 9: p->attrs |= TERM_ATTR_STRIKE; break; + case 21: set_underline(&p->attrs, 2); break; + case 22: p->attrs &= (uint16_t)~(TERM_ATTR_BOLD | TERM_ATTR_FAINT); break; + case 23: p->attrs &= (uint16_t)~TERM_ATTR_ITALIC; break; + case 24: set_underline(&p->attrs, 0); break; + case 27: p->inverse = false; break; + case 29: p->attrs &= (uint16_t)~TERM_ATTR_STRIKE; break; + case 38: i += extended_color(m, i, &color, &ok); if (ok) p->fg = color; break; + case 39: p->fg = TERM_COLOR_FG; break; + case 48: i += extended_color(m, i, &color, &ok); if (ok) p->bg = color; break; + case 49: p->bg = TERM_COLOR_BG; break; + case 53: p->attrs |= TERM_ATTR_OVERLINE; break; + case 55: p->attrs &= (uint16_t)~TERM_ATTR_OVERLINE; break; + case 58: i += extended_color(m, i, &color, &ok); if (ok) p->ul = color; break; + case 59: p->ul = 0; break; + default: + if (v >= 30 && v <= 37) p->fg = term_color_index(v - 30); + else if (v >= 40 && v <= 47) p->bg = term_color_index(v - 40); + else if (v >= 90 && v <= 97) p->fg = term_color_index(v - 90 + 8); + else if (v >= 100 && v <= 107) p->bg = term_color_index(v - 100 + 8); + break; + } + } +} + +/* The DECRQSS reply body for SGR. */ +static void sgr_report(const struct model *m, char *out, size_t cap) +{ + const struct pen *p = &m->pen; + size_t n = 0; + out[0] = 0; +#define ADD(...) do { int k = snprintf(out + n, cap - n, __VA_ARGS__); if (k > 0) n += (size_t)k; } while (0) + int ul = (int)((p->attrs & TERM_ATTR_UNDERLINE_MASK) >> TERM_ATTR_UNDERLINE_SHIFT); + if (!(p->attrs & (TERM_ATTR_BOLD | TERM_ATTR_FAINT | TERM_ATTR_ITALIC | TERM_ATTR_STRIKE | + TERM_ATTR_OVERLINE | TERM_ATTR_UNDERLINE_MASK)) && + !p->inverse && p->fg == TERM_COLOR_FG && p->bg == TERM_COLOR_BG && !p->ul) + return; + ADD("0"); + if (p->attrs & TERM_ATTR_BOLD) ADD(";1"); + if (p->attrs & TERM_ATTR_FAINT) ADD(";2"); + if (p->attrs & TERM_ATTR_ITALIC) ADD(";3"); + if (ul == 1) ADD(";4"); + else if (ul) ADD(";4:%d", ul); + if (p->inverse) ADD(";7"); + if (p->attrs & TERM_ATTR_STRIKE) ADD(";9"); + if (p->attrs & TERM_ATTR_OVERLINE) ADD(";53"); + const term_color colors[3] = { p->fg, p->bg, p->ul }; + const bool set[3] = { p->fg != TERM_COLOR_FG, p->bg != TERM_COLOR_BG, p->ul != 0 }; + for (int k = 0; k < 3; k++) { + if (!set[k]) continue; + int base = 30 + 10 * k; + term_color c = colors[k]; + if (term_color_is_slot(c) && (c & 0x1FFu) < 256) { + int i = (int)(c & 0xFFu); + if (i < 16 && k < 2) ADD(";%d", i < 8 ? base + i : base + 60 + i - 8); + else ADD(";%d:5:%d", base + 8, i); + } else { + ADD(";%d:2:%d:%d:%d", base + 8, (int)((c >> 16) & 0xff), + (int)((c >> 8) & 0xff), (int)(c & 0xff)); + } + } +#undef ADD +} + +/* DECCARA: change attributes in a rectangle, or in a stream of cells when + * DECSACE is not 2. A rectangle edge that splits a wide glyph changes both + * of its halves. */ +static void cara_cell(struct model *m, struct cell *c) +{ + if (m->nparam <= 4) { + c->fg = TERM_COLOR_FG; + c->bg = TERM_COLOR_BG; + c->attrs &= TERM_ATTR_EMOJI | TERM_ATTR_TEXT | TERM_ATTR_LINK; + return; + } + for (int i = 4; i < m->nparam; i++) { + int v = m->param[i]; + term_color color; + bool ok; + switch (v) { + case 0: + c->fg = TERM_COLOR_FG; + c->bg = TERM_COLOR_BG; + c->attrs &= TERM_ATTR_EMOJI | TERM_ATTR_TEXT | TERM_ATTR_LINK; + break; + case 1: c->attrs |= TERM_ATTR_BOLD; break; + case 2: c->attrs |= TERM_ATTR_FAINT; break; + case 3: c->attrs |= TERM_ATTR_ITALIC; break; + case 4: { + int style = 1; + if (i + 1 < m->nparam && (m->sub & (1u << (i + 1)))) + style = m->param[++i]; + set_underline(&c->attrs, style); + break; + } + case 7: { term_color t = c->fg; c->fg = c->bg; c->bg = t; break; } + case 9: c->attrs |= TERM_ATTR_STRIKE; break; + case 21: set_underline(&c->attrs, 2); break; + case 22: c->attrs &= (uint16_t)~(TERM_ATTR_BOLD | TERM_ATTR_FAINT); break; + case 23: c->attrs &= (uint16_t)~TERM_ATTR_ITALIC; break; + case 24: set_underline(&c->attrs, 0); break; + case 29: c->attrs &= (uint16_t)~TERM_ATTR_STRIKE; break; + case 38: i += extended_color(m, i, &color, &ok); if (ok) c->fg = color; break; + case 39: c->fg = TERM_COLOR_FG; break; + case 48: i += extended_color(m, i, &color, &ok); if (ok) c->bg = color; break; + case 49: c->bg = TERM_COLOR_BG; break; + case 53: c->attrs |= TERM_ATTR_OVERLINE; break; + case 55: c->attrs &= (uint16_t)~TERM_ATTR_OVERLINE; break; + case 58: i += extended_color(m, i, &color, &ok); break; + default: + if (v >= 30 && v <= 37) c->fg = term_color_index(v - 30); + else if (v >= 40 && v <= 47) c->bg = term_color_index(v - 40); + else if (v >= 90 && v <= 97) c->fg = term_color_index(v - 90 + 8); + else if (v >= 100 && v <= 107) c->bg = term_color_index(v - 100 + 8); + break; + } + } +} + +static void cara_row(struct model *m, int y, int x0, int x1) +{ + struct row *r = m->scr[y]; + if (x0 > 0 && r->c[x0].width == 0) x0--; + if (x1 + 1 < m->cols && r->c[x1].width == 2) x1++; + fill_gap(r, x1 + 1); + for (int x = x0; x <= x1; x++) cara_cell(m, &r->c[x]); +} + +static void deccara(struct model *m) +{ + int top = m->param[0] > 0 ? m->param[0] - 1 : 0; + int left = m->nparam > 1 && m->param[1] > 0 ? m->param[1] - 1 : 0; + int bottom = m->nparam > 2 && m->param[2] > 0 ? m->param[2] - 1 : m->rows - 1; + int right = m->nparam > 3 && m->param[3] > 0 ? m->param[3] - 1 : m->cols - 1; + if (top > bottom || left > right || top >= m->rows || left >= m->cols) return; + if (bottom >= m->rows) bottom = m->rows - 1; + if (right >= m->cols) right = m->cols - 1; + if (m->decsace == 2) { + for (int y = top; y <= bottom; y++) cara_row(m, y, left, right); + return; + } + cara_row(m, top, left, top == bottom ? right : m->cols - 1); + for (int y = top + 1; y < bottom; y++) cara_row(m, y, 0, m->cols - 1); + if (bottom > top) cara_row(m, bottom, 0, right); +} + +/* ---- screens and resize ---- */ + +static void free_rows(struct row **rows, int n) +{ + if (!rows) return; + for (int i = 0; i < n; i++) free(rows[i]); + free(rows); +} + +static struct row **rows_new(int n, int cols) +{ + struct row **rows = calloc((size_t)n, sizeof(*rows)); + if (!rows) abort(); + for (int i = 0; i < n; i++) rows[i] = row_new(cols); + return rows; +} + +/* Copy a row into a new width. Cells past the width drop. */ +static struct row *row_resized(const struct row *r, int old_cols, int cols) +{ + struct row *n = row_new(cols); + int keep = old_cols < cols ? old_cols : cols; + memcpy(n->c, r->c, (size_t)keep * sizeof(struct cell)); + n->len = r->len < cols ? r->len : cols; + n->wrapped = r->wrapped; + return n; +} + +struct reflow { + int cols, rows; + struct row **out; + int count, cap; + struct row *cur; + int x; + bool cursor_set, saved_set, prompt_set; + int cursor_x, cursor_y, saved_x, saved_y, prompt; + bool pending; +}; + +static void reflow_end_row(struct reflow *r, bool wrapped) +{ + if (!r->cur) r->cur = row_new(r->cols); + r->cur->len = r->x > r->cols ? r->cols : r->x; + r->cur->wrapped = wrapped; + if (r->count == r->cap) { + r->cap = r->cap ? r->cap * 2 : 64; + r->out = realloc(r->out, (size_t)r->cap * sizeof(*r->out)); + if (!r->out) abort(); + } + r->out[r->count++] = r->cur; + r->cur = NULL; + r->x = 0; +} + +static struct cell *reflow_dst(struct reflow *r) +{ + if (!r->cur) r->cur = row_new(r->cols); + return r->cur->c; +} + +/* + * Append one logical line. Pads drop and wide glyphs wrap again. A cursor + * keeps its cell. A cursor on a pad moves to the next glyph. + */ +static void reflow_append(struct reflow *r, const struct cell *cells, int len, + int cursor, int saved) +{ + for (int i = 0; i < len;) { + const struct cell *c = &cells[i]; + if (c->pad || c->width == 0) { + int at = c->pad ? r->x : r->x - 1; + if (cursor == i) { + r->cursor_x = at < 0 ? 0 : at; + r->cursor_y = r->count; + r->cursor_set = true; + r->pending = false; + } + if (saved == i) { + r->saved_x = at < 0 ? 0 : at; + r->saved_y = r->count; + r->saved_set = true; + } + i++; + continue; + } + int n = c->width == 2 && r->cols > 1 ? 2 : 1; + if (r->x + n > r->cols) { + if (r->x < r->cols) { + reflow_dst(r)[r->x] = default_blank(); + reflow_dst(r)[r->x].pad = true; + r->x++; + } + reflow_end_row(r, true); + } + if (cursor == i) { + r->cursor_x = r->x; + r->cursor_y = r->count; + r->cursor_set = true; + r->pending = false; + } + if (saved == i) { + r->saved_x = r->x; + r->saved_y = r->count; + r->saved_set = true; + } + struct cell *dst = reflow_dst(r); + dst[r->x] = *c; + if (n == 2) { + dst[r->x + 1] = *c; + dst[r->x + 1].n = 0; + dst[r->x + 1].width = 0; + memset(dst[r->x + 1].cp, 0, sizeof(c->cp)); + } + r->x += n; + i++; + } + if (cursor == len) { + r->cursor_y = r->count; + r->cursor_x = r->x == r->cols ? r->cols - 1 : r->x; + r->pending = r->x == r->cols; + r->cursor_set = true; + } + if (saved == len) { + r->saved_y = r->count; + r->saved_x = r->x == r->cols ? r->cols - 1 : r->x; + r->saved_set = true; + } +} + +static void hist_reflow_tail(struct model *m) +{ + if (m->hist_len <= 0) + return; + int start = m->hist_len - 1; + while (start > 0 && m->hist[start - 1]->wrapped) + start--; + + bool line_open = m->hist[m->hist_len - 1]->wrapped; + struct reflow rf = { .cols = m->cols, .rows = m->rows }; + for (int i = start; i < m->hist_len; i++) { + const struct row *src = m->hist[i]; + int len = src->wrapped && src->len < m->cols ? m->cols : src->len; + reflow_append(&rf, src->c, len, -1, -1); + } + if (rf.x || rf.count == 0) + reflow_end_row(&rf, false); + free(rf.cur); + + if (line_open) { + struct row *last = rf.out[rf.count - 1]; + last->wrapped = true; + if (last->len < m->cols) { + for (int x = last->len; x < m->cols; x++) { + last->c[x] = default_blank(); + last->c[x].pad = true; + } + last->len = m->cols; + } + } + + for (int i = start; i < m->hist_len; i++) + free(m->hist[i]); + m->hist_len = start; + while (m->hist_len + rf.count > m->hist_cap) { + m->hist_cap = m->hist_cap ? m->hist_cap * 2 : 64; + m->hist = realloc(m->hist, (size_t)m->hist_cap * sizeof(*m->hist)); + if (!m->hist) abort(); + } + for (int i = 0; i < rf.count; i++) + m->hist[m->hist_len++] = rf.out[i]; + free(rf.out); +} + +/* + * Reflow the normal screen and its history to a new width. Logical lines + * wrap again. Lines at or after the prompt row (OSC 133;A) are cut, not + * wrapped. Blank rows below the cursor drop. Rows more than one screen + * below the cursor drop. + */ +static void reflow(struct model *m, int cols, int rows) +{ + int old_cols = m->cols, old_rows = m->rows; + int active = old_rows; + while (active > m->y + 1 && active > m->saved.y + 1 && + !m->scr[active - 1]->len && !m->scr[active - 1]->wrapped) + active--; + struct reflow r = { .cols = cols, .rows = rows }; + int total_in = m->hist_len + active; + struct cell *line = malloc(((size_t)old_cols + (size_t)m->saved.x + 2) * sizeof(*line)); + if (!line) abort(); + for (int i = 0; i < total_in; i++) { + int y = i - m->hist_len; + const struct row *src = i < m->hist_len ? m->hist[i] : m->scr[y]; + int len = src->len; + bool wrapped = src->wrapped; + if (m->prompt >= 0 && y >= m->prompt) { + if (r.x) reflow_end_row(&r, false); + if (y == m->prompt) { + r.prompt = r.count; + r.prompt_set = true; + } + if (y == m->y) { + r.cursor_y = r.count; + r.cursor_x = m->x < cols ? m->x : cols - 1; + r.pending = false; + r.cursor_set = true; + } + if (y == m->saved.y) { + r.saved_y = r.count; + r.saved_x = m->saved.x < cols ? m->saved.x : cols - 1; + r.saved_set = true; + } + int keep = len > cols ? cols : len; + memcpy(reflow_dst(&r), src->c, (size_t)keep * sizeof(struct cell)); + r.x = keep; + reflow_end_row(&r, wrapped); + continue; + } + if (wrapped && len < old_cols) len = old_cols; + if (y == m->y && len < m->x + m->pending) len = m->x + m->pending; + if (y == m->saved.y && len <= m->saved.x) len = m->saved.x + 1; + for (int x = 0; x < len; x++) + line[x] = x < old_cols ? src->c[x] : default_blank(); + reflow_append(&r, line, len, y == m->y ? m->x + m->pending : -1, + y == m->saved.y ? m->saved.x : -1); + if (!wrapped || i + 1 == total_in) reflow_end_row(&r, false); + } + free(line); + if (r.x) reflow_end_row(&r, false); + free(r.cur); + + int visible = rows - (cols > old_cols ? old_rows - active : 0); + if (visible < 1) visible = 1; + int start = r.count > visible ? r.count - visible : 0; + if (start > r.cursor_y) start = r.cursor_y; + int history = start; + if (KNOWN_HISTORY_COUNTS_ROWS) { + if (history > m->limit) history = m->limit; + } else { + int lines = 0; + int first = start; + for (int i = start - 1; i >= 0; i--) { + if (i == start - 1 || !r.out[i]->wrapped) { + if (lines == m->limit) { + first = i + 1; + break; + } + lines++; + } + if (i == 0) first = 0; + } + history = start - first; + } + + m->cols = cols; + term_hist_clear(m); + for (int i = start - history; i < start; i++) { + hist_push(m, r.out[i]); + r.out[i] = NULL; + } + free_rows(m->scr, old_rows); + m->scr = calloc((size_t)rows, sizeof(*m->scr)); + if (!m->scr) abort(); + for (int y = 0; y < rows; y++) { + if (start + y < r.count) { + m->scr[y] = r.out[start + y]; + r.out[start + y] = NULL; + } else { + m->scr[y] = row_new(cols); + } + } + for (int i = 0; i < r.count; i++) free(r.out[i]); + free(r.out); + + m->rows = rows; + m->x = r.cursor_x < cols ? r.cursor_x : cols - 1; + m->y = r.cursor_y - start; + m->pending = r.pending && m->autowrap; + if (m->prompt >= 0) { + if (!r.prompt_set || r.prompt - start >= rows) m->prompt = -1; + else m->prompt = r.prompt > start ? r.prompt - start : 0; + } + /* A saved cursor on a row below the new screen moves to the last row + * and keeps its column. */ + if (r.saved_set && r.saved_y - start < rows) { + m->saved.x = r.saved_x < cols ? r.saved_x : cols - 1; + m->saved.y = r.saved_y < start ? 0 : r.saved_y - start; + } else { + m->saved.y = rows - 1; + } + if (m->saved.x >= cols) m->saved.x = cols - 1; + if (m->saved.y >= rows) m->saved.y = rows - 1; +} + +/* Resize without reflow: the alternate screen, or the same width. */ +static void resize_plain(struct model *m, int cols, int rows) +{ + int old_cols = m->cols, old_rows = m->rows; + if (cols != old_cols) { + for (int y = 0; y < old_rows; y++) { + struct row *n = row_resized(m->scr[y], old_cols, cols); + free(m->scr[y]); + m->scr[y] = n; + } + if (KNOWN_ALT_RESIZE_CUTS_NORMAL && m->alternate) { + for (int y = 0; y < m->normal_rows; y++) + if (m->normal[y]->len > cols) row_cut(m->normal[y], cols, m->normal_cols); + for (int i = 0; i < m->hist_len; i++) + if (m->hist[i]->len > cols) row_cut(m->hist[i], cols, m->normal_cols); + } + m->cols = cols; + } + if (rows < old_rows && m->y >= rows) { + int shift = m->y - rows + 1; + for (int i = 0; i < shift; i++) scroll_up_full(m); + m->y -= shift; + m->saved.y = m->saved.y >= shift ? m->saved.y - shift : 0; + } + if (rows != old_rows) { + /* Growth pulls rows back from history above the old rows. New + * blank rows take the pen colors. */ + int pull = 0; + if (rows > old_rows && !m->alternate) + pull = m->hist_len < rows - old_rows ? m->hist_len : rows - old_rows; + struct row **scr = calloc((size_t)rows, sizeof(*scr)); + if (!scr) abort(); + int y = 0; + for (int i = 0; i < pull; i++) scr[y++] = m->hist[m->hist_len - pull + i]; + for (int i = 0; i < old_rows; i++) { + if (y < rows) scr[y++] = m->scr[i]; + else free(m->scr[i]); + } + int first_new = y; + while (y < rows) scr[y++] = row_new(cols); + free(m->scr); + m->scr = scr; + m->rows = rows; + for (y = first_new; y < rows; y++) erase_row(m, y); + m->hist_len -= pull; + m->hist_lines = 0; + for (int i = 0; i < m->hist_len; i++) + if (i == 0 || !m->hist[i - 1]->wrapped) + m->hist_lines++; + m->y += pull; + m->saved.y += pull; + if (m->prompt >= 0) m->prompt += pull; + } + /* A shrink that cuts the prompt row off the normal screen ends the + * prompt. */ + if (!m->alternate && m->prompt >= rows) m->prompt = -1; + if (cols != old_cols) m->pending = false; + if (m->x >= cols) m->x = cols - 1; + if (m->y >= rows) m->y = rows - 1; + if (m->saved.x >= cols) m->saved.x = cols - 1; + if (m->saved.y >= rows) m->saved.y = rows - 1; +} + +static bool resize(struct model *m, int cols, int rows) +{ + if (cols > TERM_MAX_COLS) cols = TERM_MAX_COLS; + if (cols == m->cols && rows == m->rows) return true; + if (cols > MAX_TABS) return false; + m->top = 0; + m->bottom = rows - 1; + if (cols != m->cols && !m->alternate) + reflow(m, cols, rows); + else + resize_plain(m, cols, rows); + return true; +} + +static void set_alternate(struct model *m, bool on) +{ + if (m->alternate == on) return; + if (on) { + m->normal = m->scr; + m->normal_cols = m->cols; + m->normal_rows = m->rows; + m->normal_cursor = save_cursor(m); + m->normal_saved = m->saved; + m->normal_visible = m->visible; + m->normal_style = m->style; + m->normal_blink = m->blink; + m->scr = rows_new(m->rows, m->cols); + m->alternate = true; + m->x = m->y = 0; + m->pending = false; + } else { + int cols = m->cols, rows = m->rows; + free_rows(m->scr, m->rows); + m->scr = m->normal; + m->normal = NULL; + m->cols = m->normal_cols; + m->rows = m->normal_rows; + m->alternate = false; + m->saved = m->normal_saved; + m->visible = m->normal_visible; + m->style = m->normal_style; + m->blink = m->normal_blink; + restore_cursor(m, &m->normal_cursor); + m->x = m->normal_cursor.x; + m->y = m->normal_cursor.y; + m->pending = m->normal_cursor.pending; + resize(m, cols, rows); + } + m->top = 0; + m->bottom = m->rows - 1; +} + +/* ---- modes and resets ---- */ + +static void soft_reset(struct model *m) +{ + m->app_cursor = m->app_keypad = m->backarrow = false; + m->insert = m->origin = m->reverse = m->lnm = false; + m->focus = m->sync = false; + m->mouse = m->mouse_enc = 0; + m->bracketed = false; + m->autowrap = m->visible = true; + m->blink = true; + m->style = TERM_CURSOR_BLOCK; + m->top = 0; + m->bottom = m->rows - 1; + m->decsace = 0; + memset(m->keyboard, 0, sizeof(m->keyboard)); + pen_reset(&m->pen); +} + +static void full_reset(struct model *m) +{ + set_alternate(m, false); + soft_reset(m); + m->x = m->y = 0; + m->pending = false; + m->saved = initial_saved(); + m->last_n = 0; + m->g[0] = m->g[1] = 'B'; + m->gl = 0; + m->prompt = -1; + tabs_default(m); + for (int y = 0; y < m->rows; y++) erase_row(m, y); +} + +static int mode_get(const struct model *m, int mode, bool priv) +{ + if (!priv) { + if (mode == 4) return m->insert; + if (mode == 20) return m->lnm; + return -1; + } + switch (mode) { + case 1: return m->app_cursor; + case 12: return m->blink; + case 25: return m->visible; + case 67: return m->backarrow; + case 1004: return m->focus; + case 2004: return m->bracketed; + case 2026: return m->sync; + case 5: return m->reverse; + case 6: return m->origin; + case 7: return m->autowrap; + case 47: case 1047: case 1049: return m->alternate; + case 69: return 0; + case 9: case 1000: case 1001: case 1002: case 1003: return m->mouse == mode; + case 1005: case 1006: case 1015: case 1016: return m->mouse_enc == mode; + default: return -1; + } +} + +static void mode_set(struct model *m, int mode, bool on, bool priv) +{ + if (!priv) { + if (mode == 4) m->insert = on; + else if (mode == 20) m->lnm = on; + return; + } + switch (mode) { + case 1: m->app_cursor = on; break; + case 12: m->blink = on; break; + case 25: m->visible = on; break; + case 67: m->backarrow = on; break; + case 1004: m->focus = on; break; + case 2004: m->bracketed = on; break; + case 2026: m->sync = on; break; + case 5: m->reverse = on; break; + case 6: + m->origin = on; + move_to(m, 0, on ? m->top : 0); + break; + case 7: + m->autowrap = on; + m->pending = false; + break; + case 1048: + if (on) m->saved = save_cursor(m); + else restore_cursor(m, &m->saved); + break; + case 47: case 1047: case 1049: + set_alternate(m, on); + break; + case 9: case 1000: case 1001: case 1002: case 1003: + m->mouse = on ? mode : 0; + break; + case 1005: case 1006: case 1015: case 1016: + m->mouse_enc = on ? mode : 0; + break; + default: + break; + } +} + +/* ---- CSI ---- */ + +static int count(const struct model *m) +{ + return m->param[0] > 0 ? m->param[0] : 1; +} + +static bool csi_allowed(const struct model *m, unsigned char final) +{ + char prefix = m->prefix, inter = m->inter; + if (final == 'm') return !prefix && !inter; + if (final == 'H' || final == 'f') return !prefix && !inter && !m->sub; + if (inter) { + bool plain = (final == 'q' && inter == ' ') || (final == 'p' && inter == '!') || + (final == 'x' && inter == '*'); + if (plain) { + if (prefix || m->sub) return false; + } else if (final == 'p' && inter == '$') { + if (m->sub) return false; + } else if (final == 'r' && inter == '$') { + if (prefix) return false; + } else { + return false; + } + } + if (prefix && + final != 'u' && + !(final == 'c' && (prefix == '>' || prefix == '=')) && + !(final == 'q' && prefix == '>' && !inter) && + !(final == 'p' && inter == '$' && prefix == '?') && + !(final == 'W' && prefix == '?') && + !((final == 'h' || final == 'l') && prefix == '?')) + return false; + if (m->sub && !(inter == '$' && final == 'r')) return false; + return true; +} + +static void keyboard_csi(struct model *m) +{ + if (m->nparam > (m->prefix == '=' ? 2 : 1)) return; + struct keyboard *k = &m->keyboard[m->alternate]; + unsigned flags = (unsigned)m->param[0] & 31; + if (m->prefix == '?') { + reply(m, "\033[?%uu", k->flags); + } else if (m->prefix == '>') { + if (k->depth == 32) { + memmove(k->saved, k->saved + 1, sizeof(k->saved) - sizeof(k->saved[0])); + k->depth--; + } + k->saved[k->depth++] = k->flags; + k->flags = flags; + } else if (m->prefix == '<') { + unsigned n = (unsigned)count(m); + if (n > k->depth) { + k->depth = 0; + k->flags = 0; + } else { + k->depth -= n; + k->flags = k->saved[k->depth]; + } + } else if (m->prefix == '=') { + int mode = m->nparam > 1 && m->param[1] ? m->param[1] : 1; + if (mode == 1) k->flags = flags; + else if (mode == 2) k->flags |= flags; + else if (mode == 3) k->flags &= ~flags; + } +} + +static void csi_dispatch(struct model *m, unsigned char final) +{ + if (!csi_allowed(m, final)) return; + int n = count(m); + switch (final) { + case 'm': sgr(m); break; + case 'H': case 'f': { + long y = (long)(m->param[0] > 0 ? m->param[0] - 1 : 0) + top_limit(m); + if (y > bottom_limit(m)) y = bottom_limit(m); + long x = m->nparam > 1 && m->param[1] > 0 ? m->param[1] - 1 : 0; + move_to(m, (int)(x > m->cols - 1 ? m->cols - 1 : x), (int)y); + break; + } + case 'r': { + if (m->inter == '$') { + deccara(m); + break; + } + int top = n - 1; + int bottom = m->nparam > 1 && m->param[1] ? m->param[1] - 1 : m->rows - 1; + if (top < bottom && bottom < m->rows) { + m->top = top; + m->bottom = bottom; + move_to(m, 0, m->origin ? top : 0); + } + break; + } + case 'L': case 'M': + if (m->y < m->top || m->y > m->bottom) break; + scroll_region(m, m->y, m->bottom, n, final == 'L', false); + move_to(m, 0, m->y); + break; + case 'S': case 'T': + if (m->nparam > 1) break; + m->pending = false; + scroll_region(m, m->top, m->bottom, n, final == 'T', true); + break; + case 'd': { + long y = (long)n - 1 + top_limit(m); + move_to(m, m->x, (int)(y > bottom_limit(m) ? bottom_limit(m) : y)); + break; + } + case 'b': { + int times = n > m->cols * m->rows ? m->cols * m->rows : n; + for (int i = 0; i < times && m->last_n; i++) + put_glyph(m, m->last, m->last_n, m->pen.attrs); + break; + } + case 'A': move_v(m, -n, false); break; + case 'B': case 'e': move_v(m, n, false); break; + case 'E': move_v(m, n, true); break; + case 'F': move_v(m, -n, true); break; + case 'a': case 'C': move_h(m, n); break; + case 'D': move_h(m, -n); break; + case 'G': case '`': { + long x = (long)n - 1; + move_to(m, (int)(x > m->cols - 1 ? m->cols - 1 : x), m->y); + break; + } + case 's': + if (!m->prefix) m->saved = save_cursor(m); + break; + case 'u': + if (!m->prefix) restore_cursor(m, &m->saved); + else keyboard_csi(m); + break; + case 'I': case 'Z': { + int times = n > m->cols ? m->cols : n, dir = final == 'I' ? 1 : -1; + while (times-- > 0) { + int x = m->x + dir; + while (x > 0 && x < m->cols - 1 && !m->tabs[x]) x += dir; + m->x = x < 0 ? 0 : x >= m->cols ? m->cols - 1 : x; + } + m->pending = false; + break; + } + case 'g': + if (m->param[0] == 0) m->tabs[m->x] = false; + else if (m->param[0] == 3) memset(m->tabs, 0, sizeof(m->tabs)); + break; + case 'W': + if (m->prefix == '?' && m->param[0] == 5) tabs_default(m); + break; + case 'n': + if (m->prefix) break; + if (m->param[0] == 5) reply(m, "\033[0n"); + else if (m->param[0] == 6) reply(m, "\033[%d;%dR", m->y - top_limit(m) + 1, m->x + 1); + break; + case 'J': m->pending = false; erase_display(m, m->param[0]); break; + case 'K': m->pending = false; erase_line(m, m->param[0]); break; + case 'X': { + m->pending = false; + int k = n > m->cols - m->x ? m->cols - m->x : n; + erase_cells(m, m->y, m->x, m->x + k); + break; + } + case 'x': + if (m->inter == '*') m->decsace = m->param[0]; + break; + case 'P': m->pending = false; delete_chars(m, n); break; + case '@': m->pending = false; insert_chars(m, n); break; + case 'c': + if (m->prefix == '>') reply(m, "\033[>0;10;0c"); + else if (!m->prefix && m->param[0] == 0) reply(m, "\033[?62;22c"); + break; + case 't': + if (!m->prefix && m->param[0] == 14) + reply(m, "\033[4;%d;%dt", m->rows * 20, m->cols * 10); + break; + case 'h': case 'l': + for (int i = 0; i < m->nparam; i++) + mode_set(m, m->param[i], final == 'h', m->prefix == '?'); + break; + case 'q': + if (m->prefix == '>' && !m->inter && (!m->nparam || !m->param[0])) { + reply(m, "\033P>|wsap_term\033\\"); + } else if (!m->prefix && m->inter == ' ') { + int ps = m->nparam ? m->param[0] : 0; + if (ps == 0) { + m->style = TERM_CURSOR_BLOCK; + m->blink = true; + } else if (ps <= 6) { + m->style = ps < 3 ? TERM_CURSOR_BLOCK : ps < 5 ? TERM_CURSOR_UNDERLINE + : TERM_CURSOR_BAR; + m->blink = ps & 1; + } + } + break; + case 'p': + if (m->inter == '!') { + soft_reset(m); + move_to(m, 0, 0); + } else if (m->inter == '$' && m->nparam >= 1) { + int set = mode_get(m, m->param[0], m->prefix == '?'); + reply(m, "\033[%s%d;%d$y", m->prefix == '?' ? "?" : "", m->param[0], + set < 0 ? 0 : set ? 1 : 2); + } + break; + default: + break; + } +} + +/* ---- strings ---- */ + +static void osc_finish(struct model *m) +{ + const char *s = m->str; + size_t len = m->str_len; + if (!m->alternate && len >= 5 && !strncmp(s, "133;", 4) && + (len == 5 || s[5] == ';')) { + if (s[4] == 'A') m->prompt = m->y; + if (s[4] == 'C' || s[4] == 'D') m->prompt = -1; + } +} + +static void dcs_finish(struct model *m) +{ + const char *s = m->str; + size_t len = m->str_len; + if (len < 2 || s[0] != '$' || s[1] != 'q') return; + const char *spec = s + 2; + size_t n = len - 2; + char body[256]; + if (n == 2 && !memcmp(spec, " q", 2)) { + int shape = m->style == TERM_CURSOR_BLOCK ? (m->blink ? 1 : 2) + : m->style == TERM_CURSOR_UNDERLINE ? (m->blink ? 3 : 4) + : (m->blink ? 5 : 6); + reply(m, "\033P1$r%d q\033\\", shape); + } else if (n == 1 && spec[0] == 'm') { + sgr_report(m, body, sizeof(body)); + reply(m, "\033P1$r%sm\033\\", body); + } else if (n == 1 && spec[0] == 'r') { + reply(m, "\033P1$r%d;%dr\033\\", m->top + 1, m->bottom + 1); + } else if (n == 1 && spec[0] == 's') { + reply(m, "\033P1$r%d;%ds\033\\", 1, m->cols); + } else if (n == 1 && spec[0] == 't') { + reply(m, "\033P1$r%dt\033\\", m->rows); + } else if (n == 2 && !memcmp(spec, "\"p", 2)) { + reply(m, "\033P1$r64;1\"p\033\\"); + } else if (n == 2 && !memcmp(spec, "\"q", 2)) { + reply(m, "\033P1$r0\"q\033\\"); + } else if (n == 2 && !memcmp(spec, "$|", 2)) { + reply(m, "\033P1$r%d$|\033\\", m->cols); + } else if (n == 2 && !memcmp(spec, "$}", 2)) { + reply(m, "\033P1$r0$}\033\\"); + } else if (n == 2 && !memcmp(spec, "$~", 2)) { + reply(m, "\033P1$r0$~\033\\"); + } else if (n == 2 && !memcmp(spec, "*x", 2)) { + reply(m, "\033P1$r%d*x\033\\", m->decsace); + } else if (n == 2 && !memcmp(spec, "*|", 2)) { + reply(m, "\033P1$r%d*|\033\\", m->rows); + } else { + reply(m, "\033P0$r\033\\"); + } +} + +static void str_add(struct model *m, unsigned char c) +{ + if (m->str_len < STR_KEEP) m->str[m->str_len] = (char)c; + m->str_len++; +} + +/* A string longer than STR_KEEP keeps its true length, so it matches no + * short request. */ +static void str_finish(struct model *m) +{ + if (m->str_kind == OSC) osc_finish(m); + else if (m->str_kind == DCS) dcs_finish(m); + m->str_len = 0; +} + +/* ---- the byte parser ---- */ + +static void esc_dispatch(struct model *m, unsigned char ch) +{ + m->state = GROUND; + switch (ch) { + case '=': m->app_keypad = true; break; + case '>': m->app_keypad = false; break; + case 'H': m->tabs[m->x] = true; break; + case 'c': full_reset(m); break; + case '[': + m->state = CSI; + m->nparam = 0; + m->param[0] = 0; + m->sub = 0; + m->prefix = m->inter = 0; + m->csi_phase = 0; + break; + case ']': m->state = STR; m->str_kind = OSC; m->str_len = 0; break; + case 'P': m->state = STR; m->str_kind = DCS; m->str_len = 0; break; + case '_': case '^': case 'X': m->state = STR; m->str_kind = IGNORE; m->str_len = 0; break; + case '#': m->state = HASH; break; + case '%': m->state = DOCS; break; + case '(': case ')': case '*': case '+': case '-': case '.': case '/': + m->slot = (char)ch; + m->state = CHARSET; + break; + case 'D': case 'E': + m->pending = false; + if (ch == 'E') m->x = 0; + line_feed(m); + break; + case 'M': + m->pending = false; + reverse_index(m); + break; + case '7': m->saved = save_cursor(m); break; + case '8': restore_cursor(m, &m->saved); break; + default: break; + } +} + +static void utf8_interrupt(struct model *m) +{ + if (m->need) { + m->need = 0; + print_cp(m, 0xFFFD); + } +} + +static void control(struct model *m, unsigned char c) +{ + switch (c) { + case 0x1b: m->state = ESC; break; + case '\n': case 0x0b: case 0x0c: + m->pending = false; + if (m->lnm || (m->flags & TERM_FEED_ONLCR)) m->x = 0; + line_feed(m); + break; + case '\r': m->pending = false; m->x = 0; break; + case '\b': + m->pending = false; + if (m->x > 0) m->x--; + break; + case '\t': { + m->pending = false; + int x = m->x + 1; + while (x < m->cols - 1 && !m->tabs[x]) x++; + m->x = x < m->cols ? x : m->cols - 1; + break; + } + case 0x0e: m->gl = 1; break; + case 0x0f: m->gl = 0; break; + default: break; + } +} + +/* Ground-state bytes: printable ASCII, C0 controls, DEL, and UTF-8. Each + * maximal invalid UTF-8 subpart prints one U+FFFD. */ +static void ground(struct model *m, unsigned char c) +{ + if (c < 0x20 || c == 0x7f) { + utf8_interrupt(m); + if (c < 0x20) control(m, c); + return; + } + if (m->need) { + if (c >= m->lo && c <= m->hi) { + m->acc = m->acc << 6 | (c & 0x3fu); + m->lo = 0x80; + m->hi = 0xbf; + if (--m->need == 0) print_cp(m, m->acc); + return; + } + utf8_interrupt(m); + } + if (c < 0x80) { + print_cp(m, c); + return; + } + m->lo = 0x80; + m->hi = 0xbf; + if (c >= 0xc2 && c <= 0xdf) { m->need = 1; m->acc = c & 0x1fu; } + else if (c == 0xe0) { m->need = 2; m->acc = c & 0x0fu; m->lo = 0xa0; } + else if (c == 0xed) { m->need = 2; m->acc = c & 0x0fu; m->hi = 0x9f; } + else if (c >= 0xe1 && c <= 0xef) { m->need = 2; m->acc = c & 0x0fu; } + else if (c == 0xf0) { m->need = 3; m->acc = c & 0x07u; m->lo = 0x90; } + else if (c == 0xf4) { m->need = 3; m->acc = c & 0x07u; m->hi = 0x8f; } + else if (c >= 0xf1 && c <= 0xf3) { m->need = 3; m->acc = c & 0x07u; } + else print_cp(m, 0xFFFD); +} + +static void csi_byte(struct model *m, unsigned char c) +{ + if (c == 0x7f) return; + if (c >= 0x40) { + if (m->csi_phase != 3) csi_dispatch(m, c); + m->state = GROUND; + return; + } + if (m->csi_phase == 3) return; + if (c >= '0' && c <= ';') { + if (m->csi_phase == 2) { + m->csi_phase = 3; + return; + } + m->csi_phase = 1; + if (!m->nparam) m->nparam = 1; + int *p = &m->param[m->nparam - 1]; + if (c <= '9') { + *p = *p < 214748364 ? *p * 10 + (c - '0') : 0x7fffffff; + } else { + if (m->nparam == 16) { + m->csi_phase = 3; + return; + } + if (c == ':') m->sub |= (uint16_t)(1u << m->nparam); + m->param[m->nparam++] = 0; + } + } else if (c >= '<') { + if (m->csi_phase != 0 || m->prefix) m->csi_phase = 3; + else m->prefix = (char)c; + } else { + if (m->csi_phase == 2) { + m->csi_phase = 3; + } else { + m->csi_phase = 2; + m->inter = (char)c; + } + } +} + +static void feed_byte(struct model *m, unsigned char c) +{ + /* A byte of 0x80 or more cancels any sequence except a string. */ + if (c >= 0x80 && m->state != GROUND && m->state != STR) { + m->state = GROUND; + ground(m, c); + return; + } + switch (m->state) { + case GROUND: + ground(m, c); + break; + case ESC: + if (c < 0x20) { + m->state = GROUND; + ground(m, c); + } else { + esc_dispatch(m, c); + } + break; + case CSI: + /* ESC restarts, CAN and SUB cancel, other C0 controls run inside + * the sequence (ECMA-48 5.5, xterm). */ + if (c == 0x1b) m->state = ESC; + else if (c == 0x18 || c == 0x1a) m->state = GROUND; + else if (c < 0x20) control(m, c); + else csi_byte(m, c); + break; + case CHARSET: + if (m->slot == '(' || m->slot == ')') m->g[m->slot == ')'] = (char)c; + m->state = GROUND; + break; + case HASH: + m->state = GROUND; + if (c == '8') { + for (int y = 0; y < m->rows; y++) { + struct row *r = m->scr[y]; + for (int x = 0; x < m->cols; x++) { + r->c[x] = default_blank(); + r->c[x].cp[0] = 'E'; + } + r->len = m->cols; + r->wrapped = false; + } + move_to(m, 0, 0); + } + break; + case DOCS: + m->state = GROUND; + if (c != 'G' && c != '@' && c != '/') { + ground(m, '%'); + ground(m, c); + } + break; + case STR: + if (m->str_kind == IGNORE) { + if (c == 0x07 || c == 0x18 || c == 0x1a) m->state = GROUND; + else if (c == 0x1b) m->state = STR_ESC; + break; + } + if (c == 0x07) { + str_finish(m); + m->state = GROUND; + } else if (c == 0x1b) { + m->state = STR_ESC; + } else if (c < 0x20) { + m->str_len = 0; + m->state = GROUND; + ground(m, c); + } else { + str_add(m, c); + } + break; + case STR_ESC: + if (m->str_kind == IGNORE) { + if (c == '\\' || c == 0x07 || c == 0x18 || c == 0x1a) { + m->state = GROUND; + } else { + m->state = ESC; + feed_byte(m, c); + } + break; + } + if (c == '\\') { + str_finish(m); + m->state = GROUND; + } else { + m->str_len = 0; + m->state = ESC; + feed_byte(m, c); + } + break; + } +} + +/* ---- public ---- */ + +struct model *model_new(struct term_vector2i size, int history) +{ + struct model *m = calloc(1, sizeof(*m)); + if (!m) abort(); + m->cols = size.x < 1 ? 80 : size.x; + m->rows = size.y < 1 ? 24 : size.y; + m->limit = history_limit(history); + m->scr = rows_new(m->rows, m->cols); + m->autowrap = m->visible = m->blink = true; + m->style = TERM_CURSOR_BLOCK; + m->bottom = m->rows - 1; + m->g[0] = m->g[1] = 'B'; + m->saved = initial_saved(); + m->prompt = -1; + pen_reset(&m->pen); + tabs_default(m); + (void)KNOWN_HISTORY_COUNTS_ROWS; + return m; +} + +void model_free(struct model *m) +{ + if (!m) return; + free_rows(m->scr, m->rows); + free_rows(m->normal, m->normal_rows); + term_hist_clear(m); + free(m->hist); + free(m->reply); + for (int i = 0; i < m->out_cap; i++) free(m->out[i]); + free(m->out); + free(m); +} + +void model_feed(struct model *m, const char *s, size_t len, unsigned flags) +{ + m->flags = flags; + for (size_t i = 0; i < len; i++) feed_byte(m, (unsigned char)s[i]); +} + +bool model_resize(struct model *m, struct term_vector2i size) +{ + if (size.x < 1 || size.y < 1) return false; + return resize(m, size.x, size.y); +} + +struct term_vector2i model_size(const struct model *m) +{ + return (struct term_vector2i){ .x = m->cols, .y = m->rows }; +} + +int model_history(const struct model *m) +{ + return m->alternate ? 0 : m->hist_len; +} + +const struct model_cell *model_row(const struct model *m, int row) +{ + int hist = model_history(m); + if (row < 0 || (row >= hist && row - hist >= m->rows)) return NULL; + + struct model *mm = (struct model *)m; + const struct row *r = row < hist ? m->hist[row] : m->scr[row - hist]; + if (mm->out_cols != m->cols) { + for (int i = 0; i < mm->out_cap; i++) { + free(mm->out[i]); + mm->out[i] = NULL; + } + mm->out_cols = m->cols; + } + if (row >= mm->out_cap) { + int cap = mm->out_cap ? mm->out_cap : 16; + while (cap <= row) { + if (cap > INT_MAX / 2) abort(); + cap *= 2; + } + struct model_cell **out = realloc(mm->out, (size_t)cap * sizeof(*out)); + if (!out) abort(); + memset(out + mm->out_cap, 0, (size_t)(cap - mm->out_cap) * sizeof(*out)); + mm->out = out; + mm->out_cap = cap; + } + if (!mm->out[row]) { + mm->out[row] = calloc((size_t)m->cols, sizeof(*mm->out[row])); + if (!mm->out[row]) abort(); + } + for (int x = 0; x < m->cols; x++) { + const struct cell *c = &r->c[x]; + struct model_cell *o = &mm->out[row][x]; + memcpy(o->cp, c->cp, sizeof(o->cp)); + o->n = c->n; + o->width = c->width; + o->attrs = c->attrs; + o->fg = c->fg; + o->bg = c->bg; + } + return mm->out[row]; +} + +struct term_cursor model_cursor(const struct model *m) +{ + return (struct term_cursor){ + .pos = { .x = m->x, .y = m->y }, + .style = (enum term_cursor_style)m->style, + .visible = m->visible, + .blink = m->blink, + }; +} + +struct term_modes model_modes(const struct model *m) +{ + return (struct term_modes){ + .keyboard = m->keyboard[m->alternate].flags, + .mouse = (enum term_mouse)m->mouse, + .mouse_encoding = (enum term_mouse_encoding)m->mouse_enc, + .app_cursor = m->app_cursor, + .app_keypad = m->app_keypad, + .backarrow_bs = m->backarrow, + .bracketed_paste = m->bracketed, + .focus_events = m->focus, + .alternate = m->alternate, + .reverse_video = m->reverse, + .synchronized = m->sync, + }; +} + +const char *model_replies(const struct model *m, size_t *len) +{ + *len = m->reply_len; + return m->reply ? m->reply : ""; +} + +void model_take_replies(struct model *m) +{ + m->reply_len = 0; +} diff --git a/tests/model.h b/tests/model.h new file mode 100644 index 0000000..9a4316f --- /dev/null +++ b/tests/model.h @@ -0,0 +1,50 @@ +#ifndef WSAP_TERM_MODEL_H +#define WSAP_TERM_MODEL_H + +/* + * A slow reference model of the terminal. It defines correct behavior for + * the input that tests/term/fuzz_gen.h makes. model.c names its sources: + * xterm and ECMA-48 for controls, Unicode for UTF-8, and this terminal's + * own design where xterm has no rule (reflow, history, clusters). + * + * The model reads nothing from the terminal. Tests compare the two through + * the public API only. + */ + +#include "term/term.h" +#include "term/unicode/grapheme.h" + +#include <stdbool.h> +#include <stddef.h> +#include <stdint.h> + +struct model_cell { + uint32_t cp[GRAPHEME_MAX_LEN]; + uint8_t n; /* code points; 0 for the right half of a wide cell */ + uint8_t width; /* 1, 2 for a wide cell, 0 for its right half */ + uint16_t attrs; /* TERM_ATTR_* */ + term_color fg, bg; +}; + +struct model; + +/* history is the history limit, as in term_options.scrollback. */ +struct model *model_new(struct term_vector2i size, int history); +void model_free(struct model *m); +void model_feed(struct model *m, const char *s, size_t len, unsigned flags); +bool model_resize(struct model *m, struct term_vector2i size); + +struct term_vector2i model_size(const struct model *m); +/* History rows. The alternate screen shows no history, so it reports 0. */ +int model_history(const struct model *m); +/* Row 0 is the oldest history row. Row model_history() is the top screen + * row. Return NULL for an invalid row. Different rows can be read together. + * A feed or resize ends the lifetime of returned row data. */ +const struct model_cell *model_row(const struct model *m, int row); +struct term_cursor model_cursor(const struct model *m); +struct term_modes model_modes(const struct model *m); +/* Replies since the last model_take_replies(). */ +const char *model_replies(const struct model *m, size_t *len); +void model_take_replies(struct model *m); + +#endif diff --git a/tests/model_row_test.c b/tests/model_row_test.c new file mode 100644 index 0000000..7afca05 --- /dev/null +++ b/tests/model_row_test.c @@ -0,0 +1,29 @@ +#include "model.h" + +#include <assert.h> +#include <string.h> + +int main(void) +{ + struct model *m = model_new((struct term_vector2i){ .x = 10, .y = 3 }, 10); + model_feed(m, "A\r\nB\r\nC\r\nD", strlen("A\r\nB\r\nC\r\nD"), 0); + assert(model_history(m) == 1); + assert(model_row(m, -1) == NULL); + assert(model_row(m, 4) == NULL); + + const struct model_cell *oldest = model_row(m, 0); + const struct model_cell *next = model_row(m, 1); + const struct model_cell *last = model_row(m, 3); + assert(oldest[0].cp[0] == 'A'); + assert(next[0].cp[0] == 'B'); + assert(last[0].cp[0] == 'D'); + assert(oldest != next && next != last); + + model_feed(m, "E", 1, 0); + assert(model_row(m, 3)[1].cp[0] == 'E'); + assert(model_resize(m, (struct term_vector2i){ .x = 12, .y = 3 })); + assert(model_row(m, 3)[0].cp[0] == 'D'); + assert(model_row(m, 3)[1].cp[0] == 'E'); + model_free(m); + return 0; +} diff --git a/tests/model_test.c b/tests/model_test.c new file mode 100644 index 0000000..395097e --- /dev/null +++ b/tests/model_test.c @@ -0,0 +1,458 @@ +/* + * Compare the terminal with the reference model (model.c) through the + * public API: every history page and screen row through a snapshot, the + * cursor, the modes, the history size, and the reply bytes. + * + * Each seed makes a random geometry and history limit, then runs four + * steps of generated input. The terminal gets each step in random chunks, + * so the fast paths and the byte parser both run. One step in four ends + * with a random resize. Every step ends with queries (DSR, DECRQSS, + * DECRQM), so hidden state such as the pen, the margins, and the modes + * shows in the replies. + * + * On a mismatch the test shrinks the input, prints a short C string that + * reproduces it, and fails. + * + * Usage: model-test [seeds] [first-seed] + */ +#include "model.h" +#include "fuzz_gen.h" +#include "term/term.h" + +#include <stdio.h> +#include <limits.h> +#include <stdlib.h> +#include <string.h> + +/* Queries that make hidden state visible. They change nothing. */ +static const char probe[] = + "\033[6n\033P$qm\033\\\033P$qr\033\\\033P$q q\033\\" + "\033[?6$p\033[?7$p\033[4$p\033[20$p"; + +struct event { + char *bytes; /* NULL for a resize */ + size_t len; + unsigned flags; + struct term_vector2i size; +}; + +struct script { + struct term_vector2i size; + int history; + struct event *ev; + int n, cap; +}; + +static void script_add(struct script *s, struct event e) +{ + if (s->n == s->cap) { + s->cap = s->cap ? s->cap * 2 : 16; + s->ev = realloc(s->ev, (size_t)s->cap * sizeof(*s->ev)); + if (!s->ev) abort(); + } + s->ev[s->n++] = e; +} + +static void script_free(struct script *s) +{ + for (int i = 0; i < s->n; i++) free(s->ev[i].bytes); + free(s->ev); + *s = (struct script){0}; +} + +static char *dup_bytes(const char *p, size_t n) +{ + char *d = malloc(n ? n : 1); + if (!d) abort(); + memcpy(d, p, n); + return d; +} + +struct replies { + char *s; + size_t len, cap; +}; + +static void on_reply(void *ud, const char *bytes, size_t len) +{ + struct replies *r = ud; + if (r->len + len > r->cap) { + r->cap = (r->len + len) * 2; + r->s = realloc(r->s, r->cap); + if (!r->s) abort(); + } + memcpy(r->s + r->len, bytes, len); + r->len += len; +} + +static void dump(FILE *f, const char *s, size_t n) +{ + fputc('"', f); + for (size_t i = 0; i < n; i++) { + unsigned char c = (unsigned char)s[i]; + if (c == 27) fputs("\\033", f); + else if (c == '\r') fputs("\\r", f); + else if (c == '\n') fputs("\\n", f); + else if (c == '\t') fputs("\\t", f); + else if (c == '"' || c == '\\') fprintf(f, "\\%c", c); + else if (c >= 32 && c < 127) fputc(c, f); + else fprintf(f, "\\x%02x\"\"", c); + } + fputc('"', f); +} + +static char glyph(uint32_t cp) +{ + return cp >= 32 && cp < 127 ? (char)cp : cp ? '?' : '.'; +} + +/* Print every retained row of both, side by side. */ +static void dump_all(struct model *m, struct term *t, struct term_snapshot *snap, FILE *log) +{ + int total = term_scrollback(t).total, mtotal = model_history(m); + struct term_vector2i size = term_size(t), msize = model_size(m); + int rows = total + size.y > mtotal + msize.y ? total + size.y : mtotal + msize.y; + fprintf(log, " term (history %d)%*s model (history %d)\n", total, size.x > 8 ? size.x - 8 : 0, + "", mtotal); + for (int r = 0; r < rows; r++) { + fprintf(log, " %3d |", r); + if (r < total + size.y) { + int offset = r < total ? total - r : 0; + term_scroll_to(t, offset); + term_snapshot_update(snap, t); + const struct term_cell *row = term_snapshot_row(snap, r < total ? 0 : r - total); + for (int x = 0; x < size.x; x++) { + uint32_t cp[GRAPHEME_MAX_LEN] = {0}; + if (row[x].text) term_snapshot_cluster(snap, row[x].text, cp, GRAPHEME_MAX_LEN); + fputc(glyph(cp[0]), log); + } + } else { + fprintf(log, "%*s", size.x, ""); + } + fprintf(log, "| |"); + if (r < mtotal + msize.y) { + const struct model_cell *row = model_row(m, r); + for (int x = 0; x < msize.x; x++) fputc(glyph(row[x].n ? row[x].cp[0] : 0), log); + } + fprintf(log, "|\n"); + } + term_scroll_to(t, 0); + struct term_cursor c = term_cursor(t), mc = model_cursor(m); + fprintf(log, " cursor term %d,%d model %d,%d\n", c.pos.x, c.pos.y, mc.pos.x, mc.pos.y); +} + +/* Returns NULL when the terminal matches the model, else a description. + * Details go to log when it is not NULL. */ +static const char *compare(struct model *m, struct term *t, struct term_snapshot *snap, + struct replies *got, char *why, size_t cap, FILE *log) +{ + size_t want_len; + const char *want = model_replies(m, &want_len); + if (want_len != got->len || (want_len && memcmp(want, got->s, want_len))) { + if (log) { + fprintf(log, "replies: term "); + dump(log, got->s ? got->s : "", got->len); + fprintf(log, "\n model "); + dump(log, want, want_len); + fputc('\n', log); + } + snprintf(why, cap, "replies differ"); + return why; + } + struct term_vector2i size = term_size(t), msize = model_size(m); + if (size.x != msize.x || size.y != msize.y) { + snprintf(why, cap, "size %dx%d, model %dx%d", size.x, size.y, msize.x, msize.y); + return why; + } + struct term_cursor c = term_cursor(t), mc = model_cursor(m); + if (c.pos.x != mc.pos.x || c.pos.y != mc.pos.y || c.style != mc.style || + c.visible != mc.visible || c.blink != mc.blink) { + snprintf(why, cap, "cursor %d,%d style %d visible %d blink %d; model %d,%d %d %d %d", + c.pos.x, c.pos.y, c.style, c.visible, c.blink, + mc.pos.x, mc.pos.y, mc.style, mc.visible, mc.blink); + return why; + } + struct term_modes md = term_modes(t), mm = model_modes(m); + if (md.keyboard != mm.keyboard || md.mouse != mm.mouse || + md.mouse_encoding != mm.mouse_encoding || md.app_cursor != mm.app_cursor || + md.app_keypad != mm.app_keypad || md.backarrow_bs != mm.backarrow_bs || + md.bracketed_paste != mm.bracketed_paste || md.focus_events != mm.focus_events || + md.alternate != mm.alternate || md.reverse_video != mm.reverse_video || + md.synchronized != mm.synchronized) { + snprintf(why, cap, "modes differ (alternate %d/%d reverse %d/%d keypad %d/%d)", + md.alternate, mm.alternate, md.reverse_video, mm.reverse_video, + md.app_keypad, mm.app_keypad); + return why; + } + /* Visit each retained row. Wide-line totals can start as estimates. */ + int total = model_history(m); + term_scroll_to(t, 0); + term_scroll(t, INT_MAX); + for (int offset = total;; offset = offset > size.y ? offset - size.y : 0) { + if (!offset) + term_scroll_to(t, 0); + if (!term_snapshot_update(snap, t)) { + snprintf(why, cap, "snapshot update failed"); + return why; + } + for (int y = 0; y < size.y; y++) { + const struct term_cell *row = term_snapshot_row(snap, y); + int retained = total - offset + y; + const struct model_cell *want_row = model_row(m, retained); + for (int x = 0; x < size.x; x++) { + const struct model_cell *e = &want_row[x]; + uint32_t cps[GRAPHEME_MAX_LEN] = {0}; + int n = row[x].text ? term_snapshot_cluster(snap, row[x].text, cps, + GRAPHEME_MAX_LEN) : 0; + if (n == e->n && row[x].width == e->width && row[x].attrs == e->attrs && + row[x].fg == e->fg && row[x].bg == e->bg && + !memcmp(cps, e->cp, (size_t)n * sizeof(*cps))) + continue; + if (log) { + fprintf(log, "retained row %d (offset %d):\n term: ", retained, offset); + for (int k = 0; k < size.x; k++) { + uint32_t cp[GRAPHEME_MAX_LEN] = {0}; + if (row[k].text) + term_snapshot_cluster(snap, row[k].text, cp, GRAPHEME_MAX_LEN); + fputc(glyph(cp[0]), log); + } + fprintf(log, "\n model: "); + want_row = model_row(m, retained); + for (int k = 0; k < size.x; k++) + fputc(glyph(want_row[k].n ? want_row[k].cp[0] : 0), log); + fputc('\n', log); + e = &want_row[x]; + } + snprintf(why, cap, + "cell %d,%d: term U+%04X (%d cps) w%u attrs %#x fg %#x bg %#x; " + "model U+%04X (%u cps) w%u attrs %#x fg %#x bg %#x", + x, y, cps[0], n, row[x].width, row[x].attrs, row[x].fg, row[x].bg, + e->cp[0], e->n, e->width, e->attrs, e->fg, e->bg); + term_scroll_to(t, 0); + return why; + } + } + if (!offset) break; + term_scroll(t, -(offset > size.y ? size.y : offset)); + } + int observed_total = term_scrollback(t).total; + if (observed_total != total) { + snprintf(why, cap, "visited history %d rows, model %d", observed_total, total); + return why; + } + term_scroll_to(t, 0); + return NULL; +} + +enum chunking { CHUNK_RANDOM, CHUNK_WHOLE, CHUNK_BYTES }; + +/* Run a script. Returns the event index of the first mismatch, or -1. */ +static int run(const struct script *s, enum chunking chunks, uint64_t chunk_seed, + char *why, size_t cap, bool verbose) +{ + struct replies got = {0}; + struct term_options opt = { + .size = s->size, .scrollback = s->history, + .callbacks = { .ud = &got, .reply = on_reply }, + }; + struct term *t = term_new(&opt); + struct model *m = model_new(s->size, s->history); + struct term_snapshot *snap = term_snapshot_new(); + if (!t || !snap) abort(); + uint64_t saved_rng = rng; + rng = chunk_seed | 1; + int bad = -1; + for (int i = 0; i < s->n && bad < 0; i++) { + const struct event *e = &s->ev[i]; + if (e->bytes) { + model_feed(m, e->bytes, e->len, e->flags); + for (size_t k = 0; k < e->len;) { + size_t n = chunks == CHUNK_WHOLE ? e->len - k + : chunks == CHUNK_BYTES ? 1 + : 1 + (size_t)rn((int)(e->len - k)); + term_feed(t, e->bytes + k, n, e->flags); + k += n; + } + } else { + bool a = model_resize(m, e->size); + bool b = term_resize(t, e->size); + if (a != b) { + snprintf(why, cap, "resize result differs"); + bad = i; + break; + } + } + if (compare(m, t, snap, &got, why, cap, verbose ? stderr : NULL)) { + bad = i; + if (verbose) dump_all(m, t, snap, stderr); + } + model_take_replies(m); + got.len = 0; + } + rng = saved_rng; + free(got.s); + term_snapshot_free(snap); + model_free(m); + term_free(t); + return bad; +} + +static bool fails(const struct script *s, char *why, size_t cap) +{ + return run(s, CHUNK_WHOLE, 1, why, cap, false) >= 0 || + run(s, CHUNK_BYTES, 1, why, cap, false) >= 0; +} + +/* Remove byte ranges and events while the script still fails. */ +static void shrink(struct script *s) +{ + char why[512]; + for (int i = s->n - 1; i >= 0; i--) { + struct event e = s->ev[i]; + memmove(s->ev + i, s->ev + i + 1, (size_t)(s->n - i - 1) * sizeof(*s->ev)); + s->n--; + if (fails(s, why, sizeof(why))) { + free(e.bytes); + continue; + } + memmove(s->ev + i + 1, s->ev + i, (size_t)(s->n - i) * sizeof(*s->ev)); + s->ev[i] = e; + s->n++; + } + for (int i = 0; i < s->n; i++) { + struct event *e = &s->ev[i]; + if (!e->bytes) continue; + for (size_t step = e->len / 2 ? e->len / 2 : 1; step >= 1; step /= 2) { + for (size_t at = 0; at + step <= e->len;) { + char *keep = dup_bytes(e->bytes, e->len); + size_t keep_len = e->len; + memmove(e->bytes + at, e->bytes + at + step, e->len - at - step); + e->len -= step; + if (fails(s, why, sizeof(why))) { + free(keep); + continue; + } + free(e->bytes); + e->bytes = keep; + e->len = keep_len; + at++; + } + if (step == 1) break; + } + } +} + +static void report(struct script *s, int seed, const char *why) +{ + fprintf(stderr, "seed %d: %s\n", seed, why); + shrink(s); + char buf2[512]; + fprintf(stderr, "shrunk: %dx%d history %d\n", s->size.x, s->size.y, s->history); + for (int i = 0; i < s->n; i++) { + if (s->ev[i].bytes) { + fprintf(stderr, " feed%s ", s->ev[i].flags ? " (ONLCR)" : ""); + dump(stderr, s->ev[i].bytes, s->ev[i].len); + fputc('\n', stderr); + } else { + fprintf(stderr, " resize %dx%d\n", s->ev[i].size.x, s->ev[i].size.y); + } + } + int at = run(s, CHUNK_WHOLE, 1, buf2, sizeof(buf2), true); + if (at < 0) at = run(s, CHUNK_BYTES, 1, buf2, sizeof(buf2), true); + fprintf(stderr, "shrunk mismatch after event %d: %s\n", at, buf2); +} + +/* Fixed cases that the generator reaches rarely. */ +static int fixed_cases(void) +{ + static const struct { + int cols, rows, history; + const char *input; + } cases[] = { + { 12, 3, 5, "a中e\xcc\x81日本 \033[1;3;4;9;38;5;123;48;2;1;2;3mQ\033[0mR" }, + { 8, 4, 5, "\033[2;3r\033[?6hZ\nQ\nR\033[?6l\033[r" }, + { 12, 3, 50, "one two three four five\r\nsix seven eight\r\nnine" }, + { 10, 3, 50, "\033]133;A\007$ ls -la /usr/share\r\nout1\r\n\033]133;A\007$ " }, + { 6, 3, 5, "\xf0\x9f\x87\xa9\xf0\x9f\x87\xaa\xe2\x9d\xa4\xef\xb8\x8f" "ab\xe2\x98\xba\xef\xb8\x8e" }, + { 8, 3, 5, "\033(0lqqk\033(Bx\x0e" "abc\x0f" "abc" }, + { 8, 3, 5, "abcdefgh\033[3;4H\033[31;1m\0337\033[0m\033[H\0338X" }, + { 6, 2, 5, "abcdef\xe3\x80z\xed\xa0\x80\x80" }, + }; + static const struct term_vector2i sizes[] = { + { .x = 5, .y = 3 }, { .x = 16, .y = 4 }, { .x = 7, .y = 3 }, { .x = 20, .y = 5 }, + { .x = 1, .y = 2 }, { .x = 3, .y = 6 }, + }; + for (size_t i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) { + struct script s = { + .size = { .x = cases[i].cols, .y = cases[i].rows }, + .history = cases[i].history, + }; + script_add(&s, (struct event){ + .bytes = dup_bytes(cases[i].input, strlen(cases[i].input)), + .len = strlen(cases[i].input) }); + script_add(&s, (struct event){ .bytes = dup_bytes(probe, sizeof(probe) - 1), + .len = sizeof(probe) - 1 }); + for (size_t k = 0; k < sizeof(sizes) / sizeof(sizes[0]); k++) + script_add(&s, (struct event){ .size = sizes[k] }); + char why[512]; + int at = run(&s, CHUNK_BYTES, 1, why, sizeof(why), true); + if (at >= 0) { + fprintf(stderr, "fixed case %zu event %d: %s\n", i, at, why); + report(&s, -(int)i, why); + script_free(&s); + return 1; + } + script_free(&s); + } + return 0; +} + +int main(int argc, char **argv) +{ + int seeds = argc > 1 ? atoi(argv[1]) : 20000; + int first = argc > 2 ? atoi(argv[2]) : 1; + if (fixed_cases()) return 1; + static const int limits[] = { 0, 5, 50, 4096 }; + for (int seed = first; seed < first + seeds; seed++) { + rng = 0x9e3779b97f4a7c15ull * (uint64_t)seed; + struct script s = { + .size = { .x = 4 + rn(37), .y = 2 + rn(11) }, + .history = limits[rn(4)], + }; + struct term_vector2i size = s.size; + for (int step = 0; step < 4; step++) { + blen = 0; + gen(10 + rn(30), size.x, size.y); + put(probe); + unsigned flags = rn(5) == 0 ? TERM_FEED_ONLCR : 0; + script_add(&s, (struct event){ .bytes = dup_bytes(buf, blen), .len = blen, + .flags = flags }); + if (rn(4) == 0) { + size = (struct term_vector2i){ .x = 1 + rn(44), .y = 1 + rn(14) }; + script_add(&s, (struct event){ .size = size }); + } + } + if (getenv("MODEL_TEST_PRINT")) { + fprintf(stderr, "seed %d: %dx%d history %d\n", seed, s.size.x, s.size.y, s.history); + for (int i = 0; i < s.n; i++) { + if (!s.ev[i].bytes) { + fprintf(stderr, " resize %dx%d\n", s.ev[i].size.x, s.ev[i].size.y); + continue; + } + fprintf(stderr, " feed "); + dump(stderr, s.ev[i].bytes, s.ev[i].len); + fputc('\n', stderr); + } + } + char why[512]; + uint64_t chunk_seed = rng; + if (run(&s, CHUNK_RANDOM, chunk_seed, why, sizeof(why), false) >= 0) { + run(&s, CHUNK_RANDOM, chunk_seed, why, sizeof(why), true); + report(&s, seed, why); + script_free(&s); + return 1; + } + script_free(&s); + } + return 0; +} diff --git a/tests/osc_test.c b/tests/osc_test.c new file mode 100644 index 0000000..ea0ef08 --- /dev/null +++ b/tests/osc_test.c @@ -0,0 +1,400 @@ +#include "term_api.h" +#include "term/unicode/unicode.h" +#include <assert.h> +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static void write_seq(struct term *t, const char *s) +{ + for (; *s; s++) tt_write(t, s, 1); +} + +static void write_buf(struct term *t, const char *s, size_t len) +{ + tt_write(t, s, len); +} + +static char reply[256]; +static void capture_reply(void *ud, const char *s, size_t len) +{ + (void)ud; + snprintf(reply, sizeof(reply), "%.*s", (int)len, s); +} + +static char title[256]; +static void capture_title(void *ud, const char *s) +{ + (void)ud; + snprintf(title, sizeof(title), "%s", s ? s : ""); +} + +static char clip_sel[16]; +static char clip_data[4096]; +static size_t clip_len; +static int clip_count; +static void capture_clip(void *ud, const char *sel, const char *data, size_t len) +{ + (void)ud; + snprintf(clip_sel, sizeof(clip_sel), "%s", sel ? sel : ""); + assert(len < sizeof(clip_data)); + memcpy(clip_data, data, len); + clip_len = len; + clip_count++; +} + +static char notify_title[256]; +static char notify_body[1024]; +static int notify_count; +static void capture_notify(void *ud, const char *t_, const char *b) +{ + (void)ud; + snprintf(notify_title, sizeof(notify_title), "%s", t_ ? t_ : ""); + snprintf(notify_body, sizeof(notify_body), "%s", b ? b : ""); + notify_count++; +} + +static struct term *osc_test_term(void) +{ + struct term *t = tt_new_opts(80, 24, &(struct term_callbacks){ + .reply = capture_reply, + .title = capture_title, + .clipboard = capture_clip, + .notify = capture_notify, + }); + assert(t); + return t; +} + +/* Check the OSC 8 link or the detected URL at a viewport cell. NULL means + * no link. */ +static void assert_link(struct term *t, int col, int row, const char *want) +{ + char *url = tt_link_at(t, col, row); + if (want) + assert(url && !strcmp(url, want)); + else + assert(!url); + free(url); +} + +static uint32_t cursor_color(struct term *t) +{ + return term_snapshot_view(tt_snapshot_for(t))->palette[TERM_PALETTE_CURSOR]; +} + +/* Send a color query and check the reply. */ +static void assert_query(struct term *t, const char *query, const char *want) +{ + reply[0] = 0; + write_seq(t, query); + if (strcmp(reply, want)) { + fprintf(stderr, "reply to %s: got \"%s\" want \"%s\"\n", query, reply, want); + assert(!"reply mismatch"); + } +} + +static void reset_captures(void) +{ + reply[0] = 0; + title[0] = clip_sel[0] = clip_data[0] = notify_title[0] = notify_body[0] = 0; + clip_len = 0; + clip_count = 0; + notify_count = 0; +} + +static void test_osc52(void) +{ + struct term *t = osc_test_term(); + reset_captures(); + write_seq(t, "\033]52;c;aGVsbG8=\007"); + assert(clip_count == 1 && !strcmp(clip_sel, "c")); + assert(clip_len == 5 && !memcmp(clip_data, "hello", 5)); + + reset_captures(); + write_seq(t, "\033]52;c;aGVsbG8=\033\\"); + assert(clip_count == 1 && clip_len == 5 && !memcmp(clip_data, "hello", 5)); + + reset_captures(); + static const char bulk[] = "\033]52;c;aGVsbG8=\007"; + write_buf(t, bulk, sizeof(bulk) - 1); + assert(clip_count == 1 && clip_len == 5 && !memcmp(clip_data, "hello", 5)); + + reset_captures(); + char big[7 + 400 + 1]; + memcpy(big, "\033]52;c;", 7); + for (int i = 0; i < 400; i += 4) + memcpy(big + 7 + i, "QUJD", 4); + big[407] = '\007'; + write_buf(t, big, sizeof(big)); + assert(clip_count == 1 && clip_len == 300); + for (int i = 0; i < 300; i += 3) + assert(!memcmp(clip_data + i, "ABC", 3)); + + reset_captures(); + write_seq(t, "\033]52;c;\007"); + assert(clip_count == 1 && clip_len == 0); + + reset_captures(); + write_seq(t, "\033]52;c;?\007"); + write_seq(t, "\033]52;c;!!!invalid!!!\007"); + write_seq(t, "\033]52;aGVsbG8=\007"); + write_seq(t, "\033]52;;aGVsbG8=\007"); + assert(clip_count == 0); + + write_seq(t, "\033]52;pc;aGVsbG8=\007"); + assert(clip_count == 1 && !strcmp(clip_sel, "pc")); + assert(clip_len == 5 && !memcmp(clip_data, "hello", 5)); + + reset_captures(); + write_seq(t, "\033]52;s;AAH9\007"); + assert(clip_count == 1 && !strcmp(clip_sel, "s") && clip_len == 3); + assert(clip_data[0] == 0 && clip_data[1] == 1 && clip_data[2] == (char)0xfd); + term_free(t); +} + +/* An SGR reset must not drop an open link. + * + * The fast SGR path caches the cell flags, and those flags carry the link id. + * The reset used to set the cached value to zero, which cleared the link from + * every cell written after it. Write the whole sequence at once so the fast + * path runs. */ +static void test_sgr_reset_keeps_link(void) +{ + struct term *t = osc_test_term(); + reset_captures(); + + static const char seq[] = "\033]8;;https://reset.dev\033\\\033[0mX"; + write_buf(t, seq, sizeof(seq) - 1); + assert(tt_cell(t, 0, 0)->attrs & TERM_ATTR_LINK); + assert_link(t, 0, 0, "https://reset.dev"); + + /* The same reset in its own write, which takes the normal parser. */ + static const char seq2[] = "\033]8;;https://second.dev\033\\\033[0m"; + write_buf(t, seq2, sizeof(seq2) - 1); + write_buf(t, "Y", 1); + assert(tt_cell(t, 1, 0)->attrs & TERM_ATTR_LINK); + assert_link(t, 1, 0, "https://second.dev"); + + term_free(t); +} + +static void test_osc8(void) +{ + struct term *t = osc_test_term(); + reset_captures(); + + write_seq(t, "\033]8;;https://example.com\033\\link text\033]8;;\033\\after"); + for (int c = 0; c < 9; c++) { + assert(tt_cell(t, c, 0)->attrs & TERM_ATTR_LINK); + assert_link(t, c, 0, "https://example.com"); + } + for (int c = 9; c < 14; c++) { + assert(!(tt_cell(t, c, 0)->attrs & TERM_ATTR_LINK)); + assert_link(t, c, 0, NULL); + } + + /* Each cell keeps the URI of its own link. */ + write_seq(t, "\033[2;1H\033]8;;https://other.org\033\\bbb\033]8;;\033\\"); + write_seq(t, "\033[3;1H\033]8;;https://example.com\033\\ccc\033]8;;\033\\"); + assert(tt_cell(t, 0, 1)->attrs & TERM_ATTR_LINK); + assert(tt_cell(t, 0, 2)->attrs & TERM_ATTR_LINK); + assert_link(t, 0, 1, "https://other.org"); + assert_link(t, 2, 1, "https://other.org"); + assert_link(t, 0, 2, "https://example.com"); + assert_link(t, 0, 0, "https://example.com"); + + /* An erase removes the link. */ + write_seq(t, "\033[4;1H\033]8;;https://gone.io\033\\zzz\033]8;;\033\\"); + assert(tt_cell(t, 0, 3)->attrs & TERM_ATTR_LINK); + write_seq(t, "\033[4;1H\033[K"); + assert(!(tt_cell(t, 0, 3)->attrs & TERM_ATTR_LINK)); + assert_link(t, 0, 3, NULL); + + write_seq(t, "\033[5;1H\033]8;x\033\\nop\033]8;xyz\033\\qq"); + assert(!(tt_cell(t, 0, 4)->attrs & TERM_ATTR_LINK)); + assert_link(t, 0, 4, NULL); + assert(!(tt_cell(t, 4, 4)->attrs & TERM_ATTR_LINK)); + write_seq(t, "\033]8;;https://stay.net\033\\\033]8;bogus\033\\ST\033]8;;\033\\"); + assert(tt_cell(t, 5, 4)->attrs & TERM_ATTR_LINK); + assert_link(t, 5, 4, "https://stay.net"); + + write_seq(t, "\033[6;1H\033]8;id=grp;https://params.dev\007p1\033]8;;\007"); + assert(tt_cell(t, 0, 5)->attrs & TERM_ATTR_LINK); + assert_link(t, 0, 5, "https://params.dev"); + + /* A link stays on its cells when they scroll into the history. */ + for (int i = 0; i < 30; i++) + write_buf(t, "filler\r\n", 8); + assert(tt_history(t) > 0); + term_scroll(t, tt_history(t)); + assert(tt_view_cell(t, 0, 0)->attrs & TERM_ATTR_LINK); + assert_link(t, 0, 0, "https://example.com"); + term_scroll_to(t, 0); + + write_seq(t, "\033[7;1H"); + for (int i = 0; i < 8300; i++) { + char seq[64]; + int n = snprintf(seq, sizeof(seq), + "\033]8;;https://h/%d\033\\A\010", i); + write_buf(t, seq, (size_t)n); + } + assert_link(t, 0, 6, "https://h/8299"); + assert_link(t, -1, 0, NULL); + assert_link(t, 0, -1, NULL); + assert_link(t, 80, 0, NULL); + assert_link(t, 0, 24, NULL); + + write_seq(t, "\033[?1049h"); + write_seq(t, "\033[1;1H\033]8;;https://altscreen.test\033\\ALT\033]8;;\033\\"); + assert_link(t, 0, 0, "https://altscreen.test"); + write_seq(t, "\033[?1049l"); + term_scroll(t, tt_history(t)); + assert_link(t, 0, 0, "https://example.com"); + term_scroll_to(t, 0); + write_seq(t, "\033[8;1HSee file:///home/user/code/file.c#L10-L20 for info"); + assert_link(t, 10, 7, "file:///home/user/code/file.c#L10-L20"); + + term_free(t); +} + +static bool text_is(size_t (*get)(const struct term *, char *, size_t), + const struct term *t, const char *want) +{ + char buf[256]; + return get(t, buf, sizeof(buf)) == strlen(want) && !strcmp(buf, want); +} + +static void test_osc_misc(void) +{ + struct term *t = osc_test_term(); + reset_captures(); + + write_seq(t, "\033]7;file://host/home/user/proj\007"); + assert(text_is(term_cwd, t, "/home/user/proj")); + + write_seq(t, "\033]7;file:///a%20b\033\\"); + assert(text_is(term_cwd, t, "/a b")); + + const char *bulk = "\033]7;file://h/bulk%20dir\007"; + write_buf(t, bulk, strlen(bulk)); + assert(text_is(term_cwd, t, "/bulk dir")); + + write_seq(t, "\033]7;https://x/y\007"); + write_seq(t, "\033]7;notauri\007"); + assert(text_is(term_cwd, t, "/bulk dir")); + + write_seq(t, "\033]9;hello there\033\\"); + assert(notify_count == 1); + assert(!strcmp(notify_title, "wsap_term")); + assert(!strcmp(notify_body, "hello there")); + write_seq(t, "\033]9;4;3;50\007"); + assert(notify_count == 1); + + write_seq(t, "\033]777;notify;Title;Body text\007"); + assert(notify_count == 2); + assert(!strcmp(notify_title, "Title")); + assert(!strcmp(notify_body, "Body text")); + write_seq(t, "\033]777;other;T;B\007"); + assert(notify_count == 2); + + write_seq(t, "\033]22;pointer\033\\"); + assert(text_is(term_pointer_shape, t, "pointer")); + write_seq(t, "\033]22;text\007"); + assert(text_is(term_pointer_shape, t, "text")); + write_seq(t, "\033]22;\033\\"); + assert(!term_pointer_shape(t, NULL, 0)); + + term_free(t); +} + +/* OSC 4, 10, 11, and 12 change the colors of one terminal. The reset codes + * restore the colors of its theme. */ +static void test_osc_colors(void) +{ + struct term_theme theme; + term_theme_default(&theme); + struct term *t = osc_test_term(); + struct term *other = osc_test_term(); + reset_captures(); + + /* tt_new starts with the built-in theme. */ + assert(tt_color(t, TERM_COLOR_FG) == 0xFFFFFFFFu && theme.fg == 0xFFFFFFFFu); + assert(tt_color(t, TERM_COLOR_BG) == 0xFF1E1E2Eu && theme.bg == 0xFF1E1E2Eu); + assert(cursor_color(t) == 0xFF89B4FAu && theme.cursor == 0xFF89B4FAu); + assert_query(t, "\033]10;?\007", "\033]10;rgb:ffff/ffff/ffff\033\\"); + + write_seq(t, "\033]10;#ff0000\007"); + assert(tt_color(t, TERM_COLOR_FG) == 0xFFFF0000u); + assert_query(t, "\033]10;?\007", "\033]10;rgb:ffff/0000/0000\033\\"); + write_seq(t, "\033]110\007"); + assert(tt_color(t, TERM_COLOR_FG) == theme.fg); + assert_query(t, "\033]10;?\033\\", "\033]10;rgb:ffff/ffff/ffff\033\\"); + + write_seq(t, "\033]11;#00ff00\033\\"); + assert(tt_color(t, TERM_COLOR_BG) == 0xFF00FF00u); + assert_query(t, "\033]11;?\007", "\033]11;rgb:0000/ffff/0000\033\\"); + write_seq(t, "\033]111\007"); + assert(tt_color(t, TERM_COLOR_BG) == theme.bg); + assert_query(t, "\033]11;?\007", "\033]11;rgb:1e1e/1e1e/2e2e\033\\"); + + write_seq(t, "\033]12;#0000ff\007"); + assert(cursor_color(t) == 0xFF0000FFu); + assert_query(t, "\033]12;?\007", "\033]12;rgb:0000/0000/ffff\033\\"); + write_seq(t, "\033]112\033\\"); + assert(cursor_color(t) == theme.cursor); + assert_query(t, "\033]12;?\007", "\033]12;rgb:8989/b4b4/fafa\033\\"); + + uint32_t ansi5 = tt_color(t, term_color_index(5)); + assert(ansi5 == theme.ansi[5]); + write_seq(t, "\033]4;5;#ff0000\007"); + assert(tt_color(t, term_color_index(5)) == 0xFFFF0000u); + assert(tt_color(t, term_color_index(4)) == theme.ansi[4]); + + /* A color change in one terminal does not change another terminal. */ + write_seq(t, "\033]10;#123456\007"); + assert(tt_color(other, term_color_index(5)) == theme.ansi[5]); + assert(tt_color(other, TERM_COLOR_FG) == theme.fg); + assert_query(other, "\033]10;?\007", "\033]10;rgb:ffff/ffff/ffff\033\\"); + assert(tt_color(t, TERM_COLOR_FG) == 0xFF123456u); + + write_seq(t, "\033]104;5\007"); + assert(tt_color(t, term_color_index(5)) == ansi5); + write_seq(t, "\033]4;1;#010203;2;#040506\007"); + assert(tt_color(t, term_color_index(1)) == 0xFF010203u); + assert(tt_color(t, term_color_index(2)) == 0xFF040506u); + write_seq(t, "\033]104\007"); + assert(tt_color(t, term_color_index(1)) == theme.ansi[1]); + assert(tt_color(t, term_color_index(2)) == theme.ansi[2]); + + term_free(other); + term_free(t); +} + +static void test_osc_legacy_titles(void) +{ + struct term *t = osc_test_term(); + reset_captures(); + + write_seq(t, "\033]legacy title\007"); + assert(!strcmp(title, "legacy title")); + write_seq(t, "\033]\007"); + assert(!strcmp(title, "")); + write_seq(t, "\033]2adjacent\033\\"); + assert(!strcmp(title, "adjacent")); + + term_free(t); +} + +int main(void) +{ + test_osc52(); + test_osc8(); + test_sgr_reset_keeps_link(); + test_osc_misc(); + test_osc_colors(); + test_osc_legacy_titles(); + puts("PASS: OSC tests"); + return 0; +} diff --git a/tests/resize_failure_test.c b/tests/resize_failure_test.c new file mode 100644 index 0000000..5d7e63c --- /dev/null +++ b/tests/resize_failure_test.c @@ -0,0 +1,152 @@ +#define _GNU_SOURCE +#include "term_api.h" + +#include <assert.h> +#include <dlfcn.h> +#include <stdbool.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> + +#ifndef __has_feature +#define __has_feature(x) 0 +#endif + +#if defined(__SANITIZE_ADDRESS__) || defined(__SANITIZE_THREAD__) +#define SANITIZED 1 +#elif __has_feature(address_sanitizer) || __has_feature(memory_sanitizer) || \ + __has_feature(thread_sanitizer) +#define SANITIZED 1 +#endif + +#ifndef SANITIZED +static bool fail_next_calloc; +static bool fail_reallocs; + +void *calloc(size_t n, size_t size) +{ + static void *(*real_calloc)(size_t, size_t); + if (!real_calloc) + real_calloc = (void *(*)(size_t, size_t))dlsym(RTLD_NEXT, "calloc"); + if (fail_next_calloc) { + fail_next_calloc = false; + return NULL; + } + return real_calloc(n, size); +} + +void *realloc(void *p, size_t size) +{ + static void *(*real_realloc)(void *, size_t); + if (!real_realloc) + real_realloc = (void *(*)(void *, size_t))dlsym(RTLD_NEXT, "realloc"); + if (fail_reallocs) + return NULL; + return real_realloc(p, size); +} +#endif + +int main(void) +{ +#ifdef SANITIZED + puts("resize failure test skipped under sanitizer"); + return 0; +#else + struct term *term = tt_new(10, 2); + assert(term); + tt_write(term, "abc", 3); + + fail_next_calloc = true; + bool resized = tt_resize(term, 20, 2); + /* The resize must reach the failing allocation. */ + assert(!fail_next_calloc); + + /* A failed resize must leave the old terminal geometry usable as a unit. */ + assert(!resized); + assert(tt_cols(term) == 10); + assert(tt_rows(term) == 2); + assert(tt_cell(term, 0, 0)->text == 'a'); + assert(tt_cell(term, 2, 0)->text == 'c'); + + tt_write(term, "\r12345678901", 12); + assert(tt_cursor_x(term) == 1); + assert(tt_cursor_y(term) == 1); + assert(tt_cell(term, 0, 1)->text == '1'); + + /* The failed lazy allocation must not prevent a later resize retry. */ + resized = tt_resize(term, 20, 2); + assert(resized); + assert(tt_cols(term) == 20); + assert(tt_rows(term) == 2); + assert(tt_cell(term, 0, 0)->text == '1'); + term_free(term); + + term = tt_new(10, 3); + assert(term); + const char margins[] = "\033[?69h\033[1;5s\033[H"; + tt_write(term, margins, sizeof(margins) - 1); + fail_next_calloc = true; + resized = tt_resize(term, 20, 3); + assert(!fail_next_calloc); + + /* A failed resize keeps the margins that fit the old width. */ + assert(!resized); + assert(tt_cols(term) == 10); + tt_write(term, "abcdefg", 7); + assert(tt_cell(term, 4, 0)->text == 'e'); + assert(tt_cell(term, 5, 0)->text != 'f'); + assert(tt_cell(term, 0, 1)->text == 'f'); + term_free(term); + + term = tt_new(10, 2); + assert(term); + tt_write(term, "normal", 6); + fail_next_calloc = true; + tt_write(term, "\033[?1049h", 8); + assert(!fail_next_calloc); + + /* A failed alternate-screen allocation keeps the normal screen active. */ + assert(!term_modes(term).alternate); + assert(tt_cell(term, 0, 0)->text == 'n'); + assert(tt_cell(term, 5, 0)->text == 'l'); + tt_write(term, "\033[?1049l", 8); + assert(tt_cell(term, 0, 0)->text == 'n'); + term_free(term); + + /* A reflow that cannot grow its row buffers fails before it changes the + * history, so the old geometry and text survive the failed resize. The + * first resize sizes the buffers, so only the reflow's growth fails. */ + term = tt_new(20, 40); + assert(term); + for (int i = 0; i < 40; i++) { + char line[32]; + int n = snprintf(line, sizeof line, "row %02d 0123456789\r\n", i); + tt_write(term, line, (size_t)n); + } + resized = tt_resize(term, 21, 40); + assert(resized); + int history = tt_history(term); + char *before = tt_history_text(term, 1000); + assert(before); + + fail_reallocs = true; + resized = tt_resize(term, 10, 40); + fail_reallocs = false; + assert(!resized); + assert(tt_cols(term) == 21); + assert(tt_rows(term) == 40); + assert(tt_history(term) == history); + char *after = tt_history_text(term, 1000); + assert(after && !strcmp(before, after)); + free(before); + free(after); + + resized = tt_resize(term, 10, 40); + assert(resized); + assert(tt_cols(term) == 10); + term_free(term); + + puts("lazy screen allocation failures preserve observable terminal state"); + return 0; +#endif +} diff --git a/tests/resize_property_test.c b/tests/resize_property_test.c new file mode 100644 index 0000000..360e640 --- /dev/null +++ b/tests/resize_property_test.c @@ -0,0 +1,156 @@ +#include "term/term.h" +#include "term_api.h" + +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static uint64_t state; + +static unsigned random_below(unsigned limit) +{ + state ^= state << 13; + state ^= state >> 7; + state ^= state << 17; + return (unsigned)(state % limit); +} + +/* Select from the oldest retained row to the end of the live screen. + * Selection text joins wrapped rows and trims blank cells on other rows. */ +static char *logical_text(struct term *t) +{ + struct term_vector2i size = term_size(t); + term_scroll_to(t, term_scrollback(t).total); + term_select_start(t, (struct term_vector2i){ .x = 0, .y = 0 }, false, TERM_SELECT_CELL); + term_scroll_to(t, 0); + term_select_extend(t, (struct term_vector2i){ .x = size.x - 1, .y = size.y - 1 }, true); + char *text = tt_selection_text(t); + term_select_clear(t); + if (!text) return NULL; + size_t out = 0; + for (size_t i = 0; text[i]; i++) { + if (text[i] == '\n') { + while (out && text[out - 1] == ' ') out--; + } + text[out++] = text[i]; + } + while (out && (text[out - 1] == '\n' || text[out - 1] == ' ')) out--; + text[out] = 0; + return text; +} + +static uint32_t cell_at(struct term *t, struct term_snapshot *s, int x, int y) +{ + if (!term_snapshot_update(s, t)) return 0; + return term_snapshot_row(s, y)[x].text; +} + +/* A height change that pulls rows back from history moves the saved + * cursor with its row. DECRC then prints next to the saved text. */ +static int saved_cursor_follows_history(struct term_snapshot *s) +{ + static const struct { int cols, rows, grow_cols, grow_rows; } cases[] = { + { 8, 3, 8, 5 }, /* same width: rows come back from history */ + { 8, 3, 10, 5 }, /* new width: reflow */ + }; + for (size_t i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) { + struct term_options opt = { + .size = { .x = cases[i].cols, .y = cases[i].rows }, .scrollback = 10, + }; + struct term *t = term_new(&opt); + if (!t) return 1; + const char *text = "l1\r\nl2\r\nl3\r\nl4\r\nab\0337\033[1;1H"; + term_feed(t, text, strlen(text), 0); + struct term_vector2i size = { .x = cases[i].grow_cols, .y = cases[i].grow_rows }; + if (!term_resize(t, size)) return 1; + term_feed(t, "\0338X", 3, 0); + struct term_cursor c = term_cursor(t); + int y = c.pos.y; + if (cell_at(t, s, 0, y) != 'a' || cell_at(t, s, 1, y) != 'b' || + cell_at(t, s, 2, y) != 'X') { + fprintf(stderr, "case %zu: DECRC after a resize lost the saved row\n", i); + return 1; + } + term_free(t); + } + return 0; +} + +/* Narrowing a full screen with the cursor at the top wraps more rows than + * the ring holds. The cursor row must survive and stay on screen. */ +static int narrow_resize_keeps_cursor_row(struct term_snapshot *s) +{ + struct term_options opt = { .size = { .x = 40, .y = 12 }, .scrollback = 5 }; + struct term *t = term_new(&opt); + if (!t) return 1; + for (int y = 0; y < 12; y++) { + char line[64]; + memset(line, 'a' + y, 40); + line[40] = 0; + term_feed(t, line, 40, 0); + } + term_feed(t, "\033[1;1H", 6, 0); + if (!term_resize(t, (struct term_vector2i){ .x = 1, .y = 12 })) return 1; + struct term_cursor c = term_cursor(t); + if (c.pos.x != 0 || c.pos.y != 0 || cell_at(t, s, 0, 0) != 'a') { + fprintf(stderr, "narrow resize: cursor %d,%d, cell %x\n", c.pos.x, c.pos.y, + cell_at(t, s, 0, 0)); + return 1; + } + term_feed(t, "Z", 1, 0); + if (cell_at(t, s, 0, 0) != 'Z') return 1; + term_free(t); + return 0; +} + +int main(int argc, char **argv) +{ + struct term_snapshot *snap = term_snapshot_new(); + if (!snap || saved_cursor_follows_history(snap) || narrow_resize_keeps_cursor_row(snap)) + return 1; + term_snapshot_free(snap); + + int seeds = argc > 1 ? atoi(argv[1]) : 100; + static const char *word[] = { + "alpha", " x", "中", "日", "e\xcc\x81", "🦊", "👍", " beta", "123", "a", + }; + for (int seed = 1; seed <= seeds; seed++) { + state = 0x9e3779b97f4a7c15ull * (uint64_t)seed; + struct term_options opt = { + .size = { .x = 40, .y = 12 }, + .scrollback = 4096, + }; + struct term *t = term_new(&opt); + if (!t) return 1; + for (int line = 0; line < 16; line++) { + int words = 1 + (int)random_below(4); + for (int i = 0; i < words; i++) { + const char *s = word[random_below(sizeof(word) / sizeof(word[0]))]; + term_feed(t, s, strlen(s), 0); + } + if (line != 15) term_feed(t, "\r\n", 2, 0); + } + char *before = logical_text(t); + if (!before) return 1; + for (int step = 0; step < 20; step++) { + struct term_vector2i size = { + .x = 1 + (int)random_below(200), + .y = 1 + (int)random_below(60), + }; + if (!term_resize(t, size)) return 1; + char *after = logical_text(t); + if (!after) return 1; + if (strcmp(before, after)) { + fprintf(stderr, "seed %d step %d resize %dx%d: logical text changed\n" + "before: <%s>\nafter: <%s>\n", + seed, step, size.x, size.y, before, after); + return 1; + } + free(after); + } + free(before); + term_free(t); + } + return 0; +} diff --git a/tests/spec_test.c b/tests/spec_test.c new file mode 100644 index 0000000..ec9d45d --- /dev/null +++ b/tests/spec_test.c @@ -0,0 +1,395 @@ +/* + * Screen specs: one case per control, escape sequence, and CSI final that + * the write core handles. Expected results follow xterm. A case that shows + * a deliberate difference from xterm says so. + * + * Rows read as text: '#' is U+FFFD, '?' is any other code point outside + * ASCII or a cluster, '_' is the right half of a wide glyph. + */ +#include "term/term.h" + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +struct spec { + const char *name, *input; + int cols, rows; + const char *line[4]; + int x, y; + /* Optional checks. attr_x < 0 skips the cell check; history < 0 skips + * the history check; reply NULL skips the reply check. */ + int attr_x, attr_y; + uint16_t attrs; + term_color fg, bg; + int history; + const char *reply; +}; + +#define ROWS3(a, b, c) .line = { a, b, c } +#define AT(x_, y_) .x = x_, .y = y_ +#define NO_EXTRA .attr_x = -1, .history = -1 +#define CASE(name_, input_, a, b, c, x_, y_) \ + { .name = name_, .input = input_, .cols = 8, .rows = 3, ROWS3(a, b, c), AT(x_, y_), NO_EXTRA } +#define RED (TERM_COLOR_SLOT | 1) +#define BLUE (TERM_COLOR_SLOT | 4) + +static const struct spec cases[] = { + /* Printing and C0 controls. */ + CASE("print", "abc", "abc", "", "", 3, 0), + CASE("delayed wrap", "abcdefghZ", "abcdefgh", "Z", "", 1, 1), + CASE("carriage return", "abc\rZ", "Zbc", "", "", 1, 0), + CASE("line feed keeps column", "ab\nZ", "ab", " Z", "", 3, 1), + CASE("vertical tab is a line feed", "ab\x0bZ", "ab", " Z", "", 3, 1), + CASE("form feed is a line feed", "ab\x0cZ", "ab", " Z", "", 3, 1), + CASE("backspace", "abc\bZ", "abZ", "", "", 3, 0), + CASE("backspace at column 0", "\bZ", "Z", "", "", 1, 0), + /* The cursor sits on the last column; BS moves left from there. */ + CASE("backspace clears pending wrap", "abcdefgh\bZ", "abcdefZh", "", "", 7, 0), + CASE("tab", "a\tZ", "a Z", "", "", 7, 0), + CASE("tab stops at the last column", "\t\tZ", " Z", "", "", 7, 0), + CASE("tab clears pending wrap", "abcdefgh\tZ", "abcdefgZ", "", "", 7, 0), + { .name = "tab does not erase", .input = "abcdefghijk\r\tZ", .cols = 12, .rows = 2, + .line = { "abcdefghZjk" }, AT(9, 0), NO_EXTRA }, + CASE("tab clear all", "\033[3g\tZ", " Z", "", "", 7, 0), + { .name = "tab clear at cursor", .input = "\033[9G\033[0g\r\tZ", .cols = 12, .rows = 2, + .line = { " Z" }, AT(11, 0), NO_EXTRA }, + { .name = "tab set", .input = "\033[4G\033H\r\tZ", .cols = 12, .rows = 2, + .line = { " Z" }, AT(4, 0), NO_EXTRA }, + CASE("bell", "a\aZ", "aZ", "", "", 2, 0), + CASE("delete is ignored", "ab\x7f" "c", "abc", "", "", 3, 0), + CASE("NUL, CAN, SUB are ignored", "a\x18" "b\x1a" "c", "abc", "", "", 3, 0), + CASE("scroll", "a\r\nb\r\nc\r\nd", "b", "c", "d", 1, 2), + CASE("shift out to DEC graphics", "\033)0a\x0eq\x0fq", "a?q", "", "", 3, 0), + CASE("DEC graphics in G0", "\033(0x\033(Bx", "?x", "", "", 2, 0), + + /* UTF-8. Each maximal invalid subpart prints U+FFFD (Unicode 3.9). */ + CASE("UTF-8 truncated before ASCII", "a\xe3\x80z", "a#z", "", "", 3, 0), + CASE("UTF-8 lone continuation", "a\x80z", "a#z", "", "", 3, 0), + CASE("UTF-8 surrogate", "\xed\xa0\x80z", "###z", "", "", 4, 0), + CASE("UTF-8 overlong", "\xc0\xafz", "##z", "", "", 3, 0), + CASE("UTF-8 bad second byte", "\xf0\x80z", "##z", "", "", 3, 0), + CASE("UTF-8 ended by a control", "\xc3\rZ", "Z", "", "", 1, 0), + CASE("UTF-8 ended by ESC", "\xc3\033[CZ", "# Z", "", "", 3, 0), + CASE("UTF-8 ended by DEL", "\xc3\x7fZ", "#Z", "", "", 2, 0), + CASE("UTF-8 C1 NEL", "ab\xc2\x85Z", "ab", "Z", "", 1, 1), + CASE("UTF-8 C1 CSI", "\xc2\x9b" "3CZ", " Z", "", "", 4, 0), + CASE("wide glyph", "a中b", "a?_b", "", "", 4, 0), + CASE("wide glyph wraps with a pad", "abcdefg中", "abcdefg", "?_", "", 2, 1), + /* A mark joins the glyph left of the cursor, never the row above. */ + CASE("combining mark at column 0 has no base", "abcdefgh\r\n\xcc\x81Z", + "abcdefgh", "Z", "", 1, 1), + CASE("combining mark at right margin", "\033[?69h\033[2;4s\033[1;2Habc\xcc\x81Z", + " ab?", " Z", "", 2, 1), + CASE("combining mark at left margin has no base", "\033[?69h\033[3;6s\033[1;1Habcd\033[1;3H\xcc\x81Z", + "abZd", "", "", 3, 0), + { .name = "wide glyph on one column", .input = "中", .cols = 1, .rows = 1, + .line = { "?" }, AT(0, 0), .attr_x = -1, .history = 0 }, + + /* CSI cursor moves. */ + CASE("CUP", "abc\033[2;3HZ", "abc", " Z", "", 3, 1), + CASE("CUP clamps", "\033[9;9HZ", "", "", " Z", 7, 2), + CASE("CUU", "\033[3;4H\033[2AZ", " Z", "", "", 4, 0), + CASE("CUD", "\033[1;4H\033[2BZ", "", "", " Z", 4, 2), + CASE("CUF", "\033[3CZ", " Z", "", "", 4, 0), + CASE("CUB", "abc\033[2DZ", "aZc", "", "", 2, 0), + CASE("CNL", "\033[2EZ", "", "", "Z", 1, 2), + CASE("CPL", "\033[3;4H\033[2FZ", "Z", "", "", 1, 0), + CASE("CHA", "abc\033[2GZ", "aZc", "", "", 2, 0), + CASE("HPA", "abc\033[2`Z", "aZc", "", "", 2, 0), + CASE("HPR", "a\033[2aZ", "a Z", "", "", 4, 0), + CASE("VPA", "a\033[3dZ", "a", "", " Z", 2, 2), + CASE("VPR", "a\033[2eZ", "a", "", " Z", 2, 2), + /* Vertical moves stop at a margin when they start inside the region. */ + CASE("CUU stops at the top margin", "\033[2;3r\033[3;1H\033[5AZ", "", "Z", "", 1, 1), + CASE("CUD stops at the bottom margin", "\033[1;2r\033[5BZ", "", "Z", "", 1, 1), + { .name = "CUU from below the region", .input = "\033[2;3r\033[4;1H\033[9AZ", + .cols = 8, .rows = 4, .line = { "", "Z", "", "" }, AT(1, 1), NO_EXTRA }, + { .name = "CUD from above the region", .input = "\033[2;3r\033[1;1H\033[9BZ", + .cols = 8, .rows = 4, .line = { "", "", "Z", "" }, AT(1, 2), NO_EXTRA }, + + /* Erase, insert, delete. */ + CASE("EL 0", "abcdef\033[1;3H\033[K", "ab", "", "", 2, 0), + CASE("EL 1", "abcdef\033[1;3H\033[1K", " def", "", "", 2, 0), + CASE("EL 2", "abcdef\033[2K", "", "", "", 6, 0), + CASE("ED 0", "abcd\r\nefgh\033[1;3H\033[J", "ab", "", "", 2, 0), + CASE("ED 1", "abcd\r\nefgh\033[2;3H\033[1J", "", " h", "", 2, 1), + CASE("ECH", "abcdef\033[1;3H\033[2X", "ab ef", "", "", 2, 0), + CASE("ECH erases a whole wide glyph", "a中b\033[1;3H\033[X", "a b", "", "", 2, 0), + CASE("DCH", "abcdef\033[1;3H\033[2P", "abef", "", "", 2, 0), + CASE("ICH", "abcdef\033[1;3H\033[2@", "ab cdef", "", "", 2, 0), + CASE("REP", "a\033[3b", "aaaa", "", "", 4, 0), + CASE("insert mode", "abcd\r\033[4hZ", "Zabcd", "", "", 1, 0), + CASE("insert mode at the row end", "abc\033[4hZ", "abcZ", "", "", 4, 0), + /* IL and DL also move the cursor to the left margin. */ + CASE("IL", "a\r\nb\r\nc\033[2;3H\033[L", "a", "", "b", 0, 1), + CASE("DL", "a\r\nb\r\nc\033[2;3H\033[M", "a", "c", "", 0, 1), + CASE("SU", "a\r\nb\r\nc\033[S", "b", "c", "", 1, 2), + CASE("SD", "a\r\nb\r\nc\033[T", "", "a", "b", 1, 2), + CASE("CHT", "\033[1IZ", " Z", "", "", 7, 0), + CASE("CBT", "\033[2;3H\033[1ZZ", "", "Z", "", 1, 1), + /* ED 2 moves rows up to the last one with a glyph to history. A row + * that DCH emptied does not count. This keeps rows in history where + * xterm does not. */ + { .name = "ED 2 keeps text in history", .input = "x\r\nabc\r\033[3P\033[2J", + .cols = 8, .rows = 3, ROWS3("", "", ""), AT(0, 1), .attr_x = -1, .history = 1 }, + { .name = "ED 2 after ECH reads cells", .input = "x\r\nabc\r\033[3X\033[2J", + .cols = 8, .rows = 3, ROWS3("", "", ""), AT(0, 1), .attr_x = -1, .history = 1 }, + { .name = "ED 3", .input = "a\r\nb\r\nc\r\nd\033[3J", .cols = 8, .rows = 3, + ROWS3("b", "c", "d"), AT(1, 2), .attr_x = -1, .history = 0 }, + + /* ESC sequences. */ + CASE("IND", "a\033D Z", "a", " Z", "", 3, 1), + CASE("NEL", "abc\033EZ", "abc", "Z", "", 1, 1), + CASE("RI", "\033[2;1HZ\033M", "", "Z", "", 1, 0), + CASE("RI scrolls at the top", "a\033MZ", " Z", "a", "", 2, 0), + CASE("save restore", "a\0337bc\0338Z", "aZc", "", "", 2, 0), + CASE("restore without save goes home", "abc\0338Z", "Zbc", "", "", 1, 0), + CASE("restore keeps the pending wrap", "abcdefgh\0337\r\0338Z", "abcdefgh", "Z", "", 1, 1), + CASE("restore restores origin mode", "\033[2;3r\033[?6h\0337\033[?6l\0338\033[HZ", + "", "Z", "", 1, 1), + CASE("restore restores the character set", "\033(0\0337\033(B\0338x", "?", "", "", 1, 0), + CASE("reset", "abc\033cZ", "Z", "", "", 1, 0), + CASE("screen alignment", "\033#8", "EEEEEEEE", "EEEEEEEE", "EEEEEEEE", 0, 0), + + /* Modes. */ + CASE("autowrap off", "\033[?7labcdefghi", "abcdefgi", "", "", 7, 0), + CASE("alternate", "abc\033[?1049hZ", "Z", "", "", 1, 0), + CASE("alternate restores", "abc\033[?1049hZ\033[?1049lQ", "abcQ", "", "", 4, 0), + CASE("origin", "\033[2;3r\033[?6h\033[HZ", "", "Z", "", 1, 1), + CASE("margin", "\033[2;3r\033[3;1Ha\033D", "", "a", "", 1, 2), + CASE("line feed mode", "ab\033[20h\nZ", "ab", "Z", "", 1, 1), + + /* Parsing. C0 controls run inside a CSI (ECMA-48 5.5, xterm). */ + CASE("C0 inside CSI", "\033[2\n;3HZ", "", " Z", "", 3, 1), + CASE("CAN cancels CSI", "\033[31\x18Z", "Z", "", "", 1, 0), + CASE("high byte cancels CSI", "\033[3\xffZ", "#Z", "", "", 2, 0), + CASE("DEL inside CSI is ignored", "\033[\1772CZ", " Z", "", "", 3, 0), + CASE("OSC is swallowed", "a\033]0;title\007b", "ab", "", "", 2, 0), + CASE("DCS is swallowed", "a\033Pxyz\033\\b", "ab", "", "", 2, 0), + CASE("APC is swallowed", "a\033_x\033\\b", "ab", "", "", 2, 0), + + /* Attributes. */ + { + .name = "SGR", .input = "\033[1;3;4;9;53;38;5;123;48;2;1;2;3mX", + .cols = 8, .rows = 3, ROWS3("X", "", ""), AT(1, 0), + .attr_x = 0, .attr_y = 0, + .attrs = TERM_ATTR_BOLD | TERM_ATTR_ITALIC | TERM_ATTR_STRIKE | + TERM_ATTR_OVERLINE | (TERM_UNDERLINE_SINGLE << TERM_ATTR_UNDERLINE_SHIFT), + .fg = TERM_COLOR_SLOT | 123, .bg = 0xff010203u, .history = -1, + }, + { + .name = "SGR colon RGB with a color space", .input = "\033[38:2::1:2:3mX", + .cols = 8, .rows = 3, ROWS3("X", "", ""), AT(1, 0), + .attr_x = 0, .attr_y = 0, .fg = 0xff010203u, .bg = TERM_COLOR_BG, .history = -1, + }, + { + /* 48:2:: takes the next three fields as RGB; the trailing 0 resets. */ + .name = "SGR colon form is not two semicolon colors", + .input = "\033[48:2::::38;2;;;mX", + .cols = 8, .rows = 3, ROWS3("X", "", ""), AT(1, 0), + .attr_x = 0, .attr_y = 0, .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG, .history = -1, + }, + { + .name = "restore restores the pen", .input = "\033[31m\0337\033[m\0338X", + .cols = 8, .rows = 3, ROWS3("X", "", ""), AT(1, 0), + .attr_x = 0, .attr_y = 0, .fg = RED, .bg = TERM_COLOR_BG, .history = -1, + }, + { + .name = "SCORC restores the pen", .input = "\033[1m\033[s\033[m\033[uX", + .cols = 8, .rows = 3, ROWS3("X", "", ""), AT(1, 0), + .attr_x = 0, .attr_y = 0, .attrs = TERM_ATTR_BOLD, + .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG, .history = -1, + }, + { + .name = "restore without save resets the pen", .input = "\033[31m\0338X", + .cols = 8, .rows = 3, ROWS3("X", "", ""), AT(1, 0), + .attr_x = 0, .attr_y = 0, .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG, .history = -1, + }, + { + /* A rectangle edge on the right half of a wide glyph changes both + * halves. */ + .name = "DECCARA on half a wide glyph", .input = "a中b\033[1;3;1;3;1$r", + .cols = 8, .rows = 3, ROWS3("a?_b", "", ""), AT(4, 0), + .attr_x = 1, .attr_y = 0, .attrs = TERM_ATTR_BOLD, + .fg = TERM_COLOR_FG, .bg = TERM_COLOR_BG, .history = -1, + }, + { + .name = "erase uses the pen background", .input = "ab\033[41m\033[1;1H\033[X", + .cols = 8, .rows = 3, ROWS3(" b", "", ""), AT(0, 0), + .attr_x = 0, .attr_y = 0, .fg = TERM_COLOR_FG, .bg = TERM_COLOR_SLOT | 1, .history = -1, + }, + { + .name = "full-screen scroll BCE", .input = "abc\033[44m\r\n\r\n\r\n", + .cols = 8, .rows = 3, ROWS3("", "", ""), AT(0, 2), + .attr_x = 0, .attr_y = 2, + .fg = TERM_COLOR_FG, .bg = BLUE, .history = -1, + }, + + /* Replies. */ + { .name = "DSR 5", .input = "\033[5n", .cols = 8, .rows = 3, ROWS3("", "", ""), + AT(0, 0), NO_EXTRA, .reply = "\033[0n" }, + { .name = "DSR 6", .input = "ab\033[6n", .cols = 8, .rows = 3, ROWS3("ab", "", ""), + AT(2, 0), NO_EXTRA, .reply = "\033[1;3R" }, + /* The cursor stays on the last column while a wrap is pending. */ + { .name = "DSR 6 with a pending wrap", .input = "abcdefgh\033[6n", .cols = 8, .rows = 3, + ROWS3("abcdefgh", "", ""), AT(7, 0), NO_EXTRA, .reply = "\033[1;8R" }, + { .name = "DSR 6 in origin mode", .input = "\033[2;3r\033[?6h\033[6n", .cols = 8, .rows = 3, + ROWS3("", "", ""), AT(0, 1), NO_EXTRA, .reply = "\033[1;1R" }, + { .name = "DA1", .input = "\033[c\033[0c", .cols = 8, .rows = 3, ROWS3("", "", ""), + AT(0, 0), NO_EXTRA, .reply = "\033[?62;22c\033[?62;22c" }, + { .name = "DA1 with a nonzero parameter", .input = "\033[1c", .cols = 8, .rows = 3, + ROWS3("", "", ""), AT(0, 0), NO_EXTRA, .reply = "" }, + { .name = "DA2", .input = "\033[>c", .cols = 8, .rows = 3, ROWS3("", "", ""), + AT(0, 0), NO_EXTRA, .reply = "\033[>0;10;0c" }, + { .name = "DECRQSS SGR", .input = "\033[1;31m\033P$qm\033\\", .cols = 8, .rows = 3, + ROWS3("", "", ""), AT(0, 0), NO_EXTRA, .reply = "\033P1$r0;1;31m\033\\" }, + /* SGR 58 has no short form for colors 0 to 15. */ + { .name = "DECRQSS underline color", .input = "\033[58;5;3m\033P$qm\033\\", + .cols = 8, .rows = 3, ROWS3("", "", ""), AT(0, 0), NO_EXTRA, + .reply = "\033P1$r0;58:5:3m\033\\" }, + { .name = "DECRQSS margins", .input = "\033[2;3r\033P$qr\033\\", .cols = 8, .rows = 3, + ROWS3("", "", ""), AT(0, 0), NO_EXTRA, .reply = "\033P1$r2;3r\033\\" }, + { .name = "DECRQM", .input = "\033[4h\033[4$p\033[?7$p", .cols = 8, .rows = 3, + ROWS3("", "", ""), AT(0, 0), NO_EXTRA, .reply = "\033[4;1$y\033[?7;1$y" }, +}; + +struct replies { + char s[512]; + size_t len; +}; + +static void on_reply(void *ud, const char *bytes, size_t len) +{ + struct replies *r = ud; + if (r->len + len < sizeof(r->s)) { + memcpy(r->s + r->len, bytes, len); + r->len += len; + } +} + +static char cell_char(const struct term_snapshot *s, const struct term_cell *c) +{ + if (c->width == 0) return '_'; + uint32_t cps[4] = {0}; + int n = term_snapshot_cluster(s, c->text, cps, 4); + if (n != 1) return '?'; + if (cps[0] == 0xFFFD) return '#'; + return cps[0] >= 0x20 && cps[0] < 0x7f ? (char)cps[0] : '?'; +} + +static int run(const struct spec *c, struct term_snapshot *s, bool bytes) +{ + struct replies got = {0}; + struct term_options opt = { + .size = { .x = c->cols, .y = c->rows }, .scrollback = 100, + .callbacks = { .ud = &got, .reply = on_reply }, + }; + struct term *t = term_new(&opt); + if (!t) return 1; + size_t len = strlen(c->input); + if (bytes) + for (size_t i = 0; i < len; i++) term_feed(t, c->input + i, 1, 0); + else + term_feed(t, c->input, len, 0); + const char *mode = bytes ? "byte by byte" : "whole"; + int bad = 0; + struct term_cursor cursor = term_cursor(t); + if (cursor.pos.x != c->x || cursor.pos.y != c->y) { + fprintf(stderr, "%s (%s): cursor %d,%d; expected %d,%d\n", c->name, mode, + cursor.pos.x, cursor.pos.y, c->x, c->y); + bad = 1; + } + if (!term_snapshot_update(s, t)) return 1; + for (int y = 0; y < c->rows && y < 4; y++) { + const struct term_cell *row = term_snapshot_row(s, y); + char text[64]; + int n = 0; + for (int x = 0; x < c->cols && n < 63; x++) text[n++] = cell_char(s, &row[x]); + while (n > 0 && text[n - 1] == ' ') n--; + text[n] = 0; + const char *want = c->line[y] ? c->line[y] : ""; + if (strcmp(text, want)) { + fprintf(stderr, "%s (%s): row %d \"%s\"; expected \"%s\"\n", c->name, mode, + y, text, want); + bad = 1; + } + } + if (c->attr_x >= 0) { + const struct term_cell *cell = &term_snapshot_row(s, c->attr_y)[c->attr_x]; + if (cell->attrs != c->attrs || cell->fg != c->fg || cell->bg != c->bg) { + fprintf(stderr, "%s (%s): attributes %x fg %x bg %x; expected %x %x %x\n", + c->name, mode, cell->attrs, cell->fg, cell->bg, c->attrs, c->fg, c->bg); + bad = 1; + } + } + if (c->history >= 0 && term_scrollback(t).total != c->history) { + fprintf(stderr, "%s (%s): history %d; expected %d\n", c->name, mode, + term_scrollback(t).total, c->history); + bad = 1; + } + if (c->reply && (got.len != strlen(c->reply) || memcmp(got.s, c->reply, got.len))) { + fprintf(stderr, "%s (%s): reply differs\n", c->name, mode); + bad = 1; + } + term_free(t); + return bad; +} + +int main(void) +{ + struct term_snapshot *s = term_snapshot_new(); + if (!s) return 1; + int failed = 0; + for (size_t i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) + failed += run(&cases[i], s, false) | run(&cases[i], s, true); + + /* ONLCR is a feed option. A plain line feed keeps the column. */ + struct term_options opt = { .size = { .x = 8, .y = 2 } }; + struct term *t = term_new(&opt); + if (!t) return 1; + term_feed(t, "ab\nZ", 4, TERM_FEED_ONLCR); + if (term_cursor(t).pos.x != 1 || term_cursor(t).pos.y != 1) { + fprintf(stderr, "ONLCR: line feed did not return to column zero\n"); + failed++; + } + term_free(t); + + /* A long stream can skip old rows. A bare CR must keep the last rows. */ + opt.size = (struct term_vector2i){ .x = 8, .y = 3 }; + opt.scrollback = 5; + t = term_new(&opt); + if (!t) return 1; + char *stream = malloc(13000u * 9u + 8u); + if (!stream) return 1; + for (int i = 0; i < 13000; i++) memcpy(stream + (size_t)i * 9, "1234567\r\n", 9); + memcpy(stream + 13000u * 9u, "broken\rY", 8); + term_feed(t, stream, 13000u * 9u + 8u, 0); + free(stream); + if (term_scrollback(t).total != 5) { + fprintf(stderr, "long stream: history count differs\n"); + failed++; + } + for (int offset = 5;; offset = offset > 3 ? offset - 3 : 0) { + term_scroll_to(t, offset); + if (!term_snapshot_update(s, t)) return 1; + for (int y = 0; y < 3; y++) { + const char *want = offset == 0 && y == 2 ? "Yroken" : "1234567"; + const struct term_cell *row = term_snapshot_row(s, y); + for (int x = 0; x < 8; x++) { + uint32_t cp = x < (int)strlen(want) ? (unsigned char)want[x] : ' '; + if (row[x].text != cp) { + fprintf(stderr, "long stream: offset %d cell %d,%d = %x, expected %x\n", + offset, x, y, row[x].text, cp); + failed++; + break; + } + } + } + if (!offset) break; + } + term_free(t); + term_snapshot_free(s); + if (failed) fprintf(stderr, "%d spec failures\n", failed); + return failed != 0; +} diff --git a/tests/symbols_test.sh b/tests/symbols_test.sh new file mode 100644 index 0000000..4c7aee1 --- /dev/null +++ b/tests/symbols_test.sh @@ -0,0 +1,63 @@ +#!/usr/bin/env bash +# Every global symbol of the term library starts with term_. A host links +# the library next to its own code, so a bare name such as palette_init can +# collide with a host symbol. Every function that term.h declares is also +# defined, so the library cannot leave one out. A shared build exports +# exactly the functions of term.h. +set -euo pipefail +lib=$1 +src=${2:-$(dirname "$0")/../include/term} + +syms=$(nm -g --defined-only -P "$lib") +# ASan adds an __odr_asan.<name> symbol for each global variable. +defined=$(awk 'NF >= 2 && $2 ~ /^[A-Z]$/ && $2 != "U" { print $1 }' <<<"$syms" | + grep -v '^_*odr_asan\.' | sed 's/^_//' | sort -u) + +status=0 +bad=$(grep -v '^term_' <<<"$defined" || true) +if [[ -n $bad ]]; then + echo "$lib: global symbols without the term_ prefix:" >&2 + echo "$bad" >&2 + status=1 +fi + +declared=$("${CC:-cc}" -std=c11 -E -P -I"$src" "$src/term.h" | python3 -c ' +import re, sys +text = re.sub(r"\s+", " ", sys.stdin.read()) +depth, flat = 0, [] +for ch in text: + depth += ch == "{" + if depth == 0: + flat.append(ch) + depth -= ch == "}" + if ch == "}" and depth == 0: + flat.append(";") +for stmt in "".join(flat).split(";"): + if "static" in stmt or "typedef" in stmt: + continue + m = re.search(r"\b(term_\w+)\s*\(", stmt) + if m: + print(m.group(1)) +' | sort -u) +missing=$(comm -23 <(echo "$declared") <(echo "$defined")) +if [[ -n $missing ]]; then + echo "$lib: functions declared in term.h but not defined:" >&2 + echo "$missing" >&2 + status=1 +fi + +# The shared library exports term.h and nothing else. +shared=${3:-} +if [[ -n $shared ]]; then + exported=$(nm -D --defined-only -P "$shared" | awk '$2 ~ /^[TDBRVW]$/ { print $1 }' | + sed 's/^_//' | sort -u) + extra=$(comm -13 <(echo "$declared") <(echo "$exported")) + absent=$(comm -23 <(echo "$declared") <(echo "$exported")) + if [[ -n $extra || -n $absent ]]; then + echo "$shared: exports differ from term.h" >&2 + [[ -n $extra ]] && echo "not in term.h: $extra" >&2 + [[ -n $absent ]] && echo "not exported: $absent" >&2 + status=1 + fi +fi +exit $status diff --git a/tests/term_test.c b/tests/term_test.c new file mode 100644 index 0000000..02c050e --- /dev/null +++ b/tests/term_test.c @@ -0,0 +1,2369 @@ +#include "term_api.h" +#include "term/unicode/unicode.h" +#include "term/unicode/utf8.h" +#include <assert.h> +#include <limits.h> +#include <pthread.h> +#include <stdatomic.h> +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +static char response[64]; +static size_t response_len; +static void respond(void *ud, const char *data, size_t len) +{ + (void)ud; + assert(len < sizeof(response)); + memcpy(response, data, len); + response[len] = 0; + response_len = len; +} +static char title[128]; +static void capture_title(void *ud, const char *s) +{ + (void)ud; + snprintf(title, sizeof(title), "%s", s); +} +static const struct term_callbacks reply_cb = { .reply = respond }; +static const struct term_callbacks title_cb = { .reply = respond, .title = capture_title }; + +static void write_seq(struct term *t, const char *s) +{ + + for (; *s; s++) tt_write(t, s, 1); +} +static void feed_seq(struct term *t, const char *s, unsigned flags) +{ + for (; *s; s++) term_feed(t, s, 1, flags); +} +static const char *row_text(struct term *t, int r, char *buf) +{ + const struct term_cell *cells = term_snapshot_row(tt_snapshot_for(t), r); + int n = 0; + for (int c = 0; c < tt_cols(t); c++) { + uint32_t cp = tt_cell_cp(t, &cells[c]); + /* '?' stands for any code point outside ASCII. */ + buf[n++] = !cp ? ' ' : cp < 0x80 ? (char)cp : '?'; + } + while (n > 0 && buf[n - 1] == ' ') n--; + buf[n] = 0; + return buf; +} +static bool row_eq(struct term *t, int r, const char *want) +{ + char *got = malloc((size_t)tt_cols(t) + 1); + assert(got); + bool equal = !strcmp(row_text(t, r, got), want); + free(got); + return equal; +} +static bool row_is(struct term *t, int r, const char *want) +{ + char *got = malloc((size_t)tt_cols(t) + 1); + assert(got); + bool equal = !strcmp(row_text(t, r, got), want); + if (!equal) + fprintf(stderr, "row %d: got \"%s\" want \"%s\"\n", r, got, want); + free(got); + return equal; +} + +static void write_both(struct term *whole, struct term *bytes, const char *s) +{ + tt_write(whole, s, strlen(s)); + write_seq(bytes, s); +} + +/* The view version of s after it copies t. A renderer redraws when the + * version changes. */ +static uint64_t view_version(struct term_snapshot *s, struct term *t) +{ + term_snapshot_update(s, t); + return term_snapshot_view(s)->version; +} + +/* + * Regression: diff renderers (Claude Code, Ink) skip unchanged spaces with + * CUF. Cells the cursor skips must read as blanks, even on a row whose cells + * still hold old glyphs: a row recycled by scrolling, or a row whose tail + * an erase dropped without filling. Stale glyphs showed up in place of + * every skipped space. + */ +static void test_skipped_cells_read_blank(void) +{ + static const char *const lines[] = { + "ab\033[1Ccd\033[3Cef", + "a…b\033[1Cc", + "\033[1m\033[38;2;1;2;3mab\033[0m\033[2Ccd", + }; + /* row_text() reads U+2026 as '?'. */ + static const char *const want[] = { "ab cd ef", "a?b c", "ab cd" }; + for (size_t k = 0; k < sizeof(lines) / sizeof(lines[0]); k++) { + struct term *whole = tt_new(20, 4); + struct term *bytes = tt_new(20, 4); + for (int i = 0; i < 12; i++) + write_both(whole, bytes, "XXXXXXXXXXXXXXXXXXXX\r\n"); + write_both(whole, bytes, lines[k]); + for (struct term *t = whole; t; t = t == whole ? bytes : NULL) { + char *got = malloc((size_t)tt_cols(t) + 1); + assert(got); + row_text(t, 3, got); + if (strcmp(got, want[k])) { + fprintf(stderr, "line %zu (%s): got \"%s\" want \"%s\"\n", k, + t == whole ? "whole" : "bytes", got, want[k]); + assert(0); + } + free(got); + } + term_free(whole); + term_free(bytes); + } + + struct term *whole = tt_new(20, 4); + struct term *bytes = tt_new(20, 4); + write_both(whole, bytes, "abcdefgh\r\033[2C\033[K\033[2CZ"); + assert(row_is(whole, 0, "ab Z")); + assert(row_is(bytes, 0, "ab Z")); + term_free(whole); + term_free(bytes); +} + +/* Compare the first rows of both viewports cell by cell. Each terminal gets + * its own snapshot. A cluster has a different handle in each snapshot, so + * the test compares its code points. */ +static void assert_same_rows(struct term *a, struct term *b, int rows) +{ + struct term_snapshot *sa = term_snapshot_new(); + struct term_snapshot *sb = term_snapshot_new(); + bool copied = sa && sb && term_snapshot_update(sa, a) && term_snapshot_update(sb, b); + assert(copied); + int cols = term_snapshot_view(sa)->size.x; + assert(cols == term_snapshot_view(sb)->size.x); + for (int y = 0; y < rows; y++) { + const struct term_cell *ra = term_snapshot_row(sa, y); + const struct term_cell *rb = term_snapshot_row(sb, y); + assert(ra && rb); + for (int x = 0; x < cols; x++) { + const struct term_cell *ac = &ra[x], *bc = &rb[x]; + uint32_t acp[32], bcp[32]; + int an = term_snapshot_cluster(sa, ac->text, acp, 32); + int bn = term_snapshot_cluster(sb, bc->text, bcp, 32); + bool same = an == bn && an <= 32 && + !memcmp(acp, bcp, (size_t)an * sizeof(*acp)) && + ac->fg == bc->fg && ac->bg == bc->bg && + ac->attrs == bc->attrs && ac->width == bc->width; + if (!same) { + fprintf(stderr, "x=%d y=%d: a(cp=%x fg=%x bg=%x attrs=%x width=%d) " + "b(cp=%x fg=%x bg=%x attrs=%x width=%d)\n", x, y, + an ? acp[0] : 0, ac->fg, ac->bg, ac->attrs, ac->width, + bn ? bcp[0] : 0, bc->fg, bc->bg, bc->attrs, bc->width); + assert(0); + } + } + } + term_snapshot_free(sa); + term_snapshot_free(sb); +} + +static void assert_same_screen(struct term *a, struct term *b) +{ + assert(tt_cursor_x(a) == tt_cursor_x(b)); + assert(tt_cursor_y(a) == tt_cursor_y(b)); + assert(tt_history(a) == tt_history(b)); + assert(term_modes(a).synchronized == term_modes(b).synchronized); + assert(tt_rows(a) == tt_rows(b)); + assert_same_rows(a, b, tt_rows(a)); +} + +static void test_read_boundaries(void) +{ + static const char input[] = + "ASCII 1234567890\r\nT\xc3\xbc" "rk\xe2\x94\x80" + "\033[31mred\033[m\033(0lqqk\033(B" + "\033[2;3H!\033[2b\033[1D?\033[K" + "\033[?1049halt\033[?1049lZ" + "\033]11;?\007\033Pignored\033\\" + "\xf0\x9f\x98\x80\xc3XQ©️界\033[2G\033[P"; + struct term *expected = tt_new(16, 4); + tt_write(expected, input, sizeof(input) - 1); + for (size_t split = 0; split < sizeof(input); split++) { + struct term *actual = tt_new(16, 4); + tt_write(actual, input, split); + tt_write(actual, input + split, sizeof(input) - 1 - split); + assert_same_screen(actual, expected); + term_free(actual); + } + + /* The byte that ends an unfinished sequence is not lost. */ + struct term *t = tt_new(16, 4); + write_seq(t, "\xc3XQ"); + assert(tt_cell_cp(t, tt_cell(t, 0, 0)) == 0xFFFD); + assert(tt_cell_cp(t, tt_cell(t, 1, 0)) == 'X'); + assert(tt_cell_cp(t, tt_cell(t, 2, 0)) == 'Q'); + term_free(t); + term_free(expected); +} + +static void test_ascii_controls(void) +{ + static const char *separators[] = {"\n", "\r", "\b", "\t", "\033[31m", "\177", "é界"}; + for (int bottom = 0; bottom < 2; bottom++) { + for (size_t c = 0; c < sizeof(separators) / sizeof(separators[0]); c++) { + char input[256]; + memset(input, 'a', 29); + size_t n = strlen(separators[c]); + memcpy(input + 29, separators[c], n); + memset(input + 29 + n, 'b', 173); + input[202 + n] = '\0'; + struct term *bulk = tt_new(120, 6); + struct term *bytes = tt_new(120, 6); + const char *setup = bottom ? "\033[6;1H\033[1;3;4;38;2;12;34;56m" + : "\033[1;3;4;38;2;12;34;56m"; + write_seq(bulk, setup); + write_seq(bytes, setup); + tt_write(bulk, input, strlen(input)); + write_seq(bytes, input); + assert_same_screen(bulk, bytes); + term_scroll(bulk, 6); + term_scroll(bytes, 6); + assert_same_screen(bulk, bytes); + term_free(bulk); + term_free(bytes); + } + } +} + +static void test_esc_control_strings_and_designators(void) +{ + struct term *t = tt_new(32, 2); + assert(t); + + write_seq(t, "A\033_payload ignored\033\\B"); + write_seq(t, "\033^privacy message\007C"); + write_seq(t, "\033Xsos ignored\033\\D"); + assert(row_is(t, 0, "ABCD")); + + write_seq(t, "\033%Gx\033%@y\033%/z"); + assert(row_is(t, 0, "ABCDxyz")); + write_seq(t, "\033%qx"); + assert(row_is(t, 0, "ABCDxyz%qx")); + + write_seq(t, "\033.Am\033*0n"); + assert(row_is(t, 0, "ABCDxyz%qxmn")); + term_free(t); + + t = tt_new(8, 2); + write_seq(t, "A\033_ignored\033DB"); + assert(row_is(t, 0, "A")); + assert(row_is(t, 1, " B")); + term_free(t); +} + +static void test_random_read_boundaries(void) +{ + static const char embedded[] = + "\033\\\033(B\r\nA\xc3\xbc\xe7\x95\x8c" "e\xcc\x81" + "\033[31mR\033[m\tX\bY\033]0;title\007\033Pignored\033\\" + "\033[?2026hS\033[?2026l\xc3XQ\xe0\x80\x80\xf4\x90\x80\x80"; + static const size_t chunks[] = {1, 2, 7, 31, 127}; + for (unsigned seed = 1; seed <= 4; seed++) { + char input[1024]; + unsigned random = seed; + for (size_t i = 0; i < sizeof(input); i++) { + random = random * 1103515245u + 12345u; + input[i] = (char)(random >> 16); + } + memcpy(input + 400, embedded, sizeof(embedded) - 1); + struct term *whole = tt_new(32, 8); + tt_write(whole, input, sizeof(input)); + for (size_t c = 0; c < sizeof(chunks) / sizeof(chunks[0]); c++) { + struct term *a = tt_new(32, 8); + struct term *b = tt_new(32, 8); + for (size_t block = 0; block < sizeof(input); block += 128) { + tt_write(a, input + block, 128); + for (size_t i = block; i < block + 128;) { + size_t n = chunks[c]; + if (n > block + 128 - i) n = block + 128 - i; + tt_write(b, input + i, n); + i += n; + } + assert_same_screen(a, b); + } + assert_same_screen(whole, a); + write_seq(a, "\033\\\033(B\r\nEND"); + write_seq(b, "\033\\\033(B\r\nEND"); + assert_same_screen(a, b); + int history = tt_history(a); + for (int i = 0; i < history; i += 8) { + term_scroll(a, 8); + term_scroll(b, 8); + assert_same_screen(a, b); + } + term_free(a); + term_free(b); + } + term_free(whole); + } +} + +static void test_csi_recovery(void) +{ + static const struct { const char *input, *first, *second; } cases[] = { + /* CR and LF run inside the CSI; B ends it as CUD (ECMA-48 5.5). */ + {"A\033[\r\nB!", "A", ""}, + /* 0xFF cancels the CSI and prints U+FFFD. */ + {"A\033[\xff" "BC", "A?BC", ""}, + {"A\033[\033[2JZ", " Z", ""}, + {"\033[ 31mX", "X", ""}, + {"\033[31$mX", "X", ""}, + {"\033[31;?mX", "X", ""}, + {"\033[31;0;0;0;0;0;0;0;0;0;0;0;0;0;0;0;0mX", "X", ""}, + {"\033[31\030X", "X", ""}, + {"\033[31\032X", "X", ""}, + {"\033[1$ A X", " X", ""}, + {"AB\033[?2JX", "ABX", ""}, + {"AB\r\033[>4hX", "XB", ""}, + }; + for (size_t c = 0; c < sizeof(cases) / sizeof(cases[0]); c++) { + size_t len = strlen(cases[c].input); + for (size_t split = 0; split <= len; split++) { + struct term *t = tt_new(20, 4); + tt_write(t, cases[c].input, split); + tt_write(t, cases[c].input + split, len - split); + assert(row_is(t, 0, cases[c].first)); + assert(row_is(t, 1, cases[c].second)); + for (int x = 0; x < 20; x++) assert(tt_cell(t, x, 0)->fg == TERM_COLOR_FG); + term_free(t); + } + } +} + +static void test_csi_byte_boundaries(void) +{ + static const char *prefixes[] = {"\033[", "\033[31;", "\033[?", "\033[1 ", "\033[1$ "}; + for (size_t p = 0; p < sizeof(prefixes) / sizeof(prefixes[0]); p++) { + for (unsigned byte = 0; byte < 256; byte++) { + char input[64]; + size_t n = strlen(prefixes[p]); + memcpy(input, prefixes[p], n); + input[n++] = (char)byte; + static const char suffix[] = "mX\033[6n\033[?25l\033[?25h\033[0mY"; + memcpy(input + n, suffix, sizeof(suffix) - 1); + n += sizeof(suffix) - 1; + struct term *whole = tt_new_opts(20, 4, &reply_cb); + response_len = 0; + tt_write(whole, input, n); + char expected_response[64]; + size_t expected_len = response_len; + memcpy(expected_response, response, expected_len); + for (size_t split = 0; split <= strlen(prefixes[p]) + 2; split++) { + struct term *parts = tt_new_opts(20, 4, &reply_cb); + response_len = 0; + tt_write(parts, input, split); + tt_write(parts, input + split, n - split); + assert_same_screen(whole, parts); + assert(term_cursor(whole).visible == term_cursor(parts).visible); + assert(response_len == expected_len); + assert(!memcmp(response, expected_response, expected_len)); + term_free(parts); + } + term_free(whole); + } + } +} + +static void test_parameter_limits(void) +{ + struct term *t = tt_new(10, 4); + write_seq(t, "\033[2;4r\033[?6h\033[999999999999999999999;1H"); + assert(tt_cursor_y(t) == 3 && tt_cursor_x(t) == 0); + write_seq(t, "\033[?6l\033[r"); + write_seq(t, "\033[2;2H\033[999999999999999999999B\033[2147483647C!"); + assert(tt_cursor_y(t) == 3 && tt_cursor_x(t) == 9); + assert(tt_cell(t, 9, 3)->text == '!'); + write_seq(t, "\033[999999999999999999999A\033[999999999999999999999D"); + assert(tt_cursor_y(t) == 0 && tt_cursor_x(t) == 0); + write_seq(t, "\033[38;5;999999999999999999999mX"); + assert(tt_cell(t, 0, 0)->fg == TERM_COLOR_FG); + + /* A renderer sees the cursor move to another row. */ + struct term_snapshot *snap = term_snapshot_new(); + uint64_t version = view_version(snap, t); + write_seq(t, "\033[3d"); + assert(tt_cursor_y(t) == 2 && view_version(snap, t) != version); + term_snapshot_free(snap); + + write_seq(t, "\033[4;8H\0337"); + tt_resize(t, 5, 2); + write_seq(t, "\0338Z"); + assert(tt_cursor_y(t) == 1 && tt_cursor_x(t) == 3); + assert(tt_cell(t, 2, 1)->text == 'Z'); + term_free(t); +} + +static void test_unterminated_control_strings_do_not_swallow_text(void) +{ + struct term *t = tt_new(20, 4); + assert(t); + + write_seq(t, "A\033]0;random bytes\tgit\r\nstatus"); + assert(row_is(t, 0, "A git")); + assert(row_is(t, 1, "status")); + + write_seq(t, "\033Prandom bytes\tbranch\r\nnext"); + assert(row_is(t, 1, "status branch")); + assert(row_is(t, 2, "next")); + + term_free(t); + + char overlong[1200]; + for (int intro = 0; intro < 2; intro++) { + int n = snprintf(overlong, sizeof(overlong), intro ? "\033P" : "\033]0;"); + memset(overlong + n, 'T', 1024); + overlong[n + 1024] = '\0'; + + struct term *u = tt_new(20, 4); + write_seq(u, overlong); + write_seq(u, "\r\n[sudo] password:"); + assert(row_is(u, 0, "")); + assert(row_is(u, 1, "[sudo] password:")); + term_free(u); + + struct term *d = tt_new(20, 4); + write_seq(d, overlong); + write_seq(d, "\033\\after"); + assert(row_is(d, 0, "after")); + term_free(d); + } +} + +static void test_c0_controls_after_esc_are_executed(void) +{ + struct term *t = tt_new(20, 4); + assert(t); + + write_seq(t, "A\033\tB"); + assert(row_is(t, 0, "A B")); + + term_free(t); +} + +static void test_resize_reflow(void) +{ + struct term *t = tt_new(8, 4); + write_seq(t, "abcdefghijkl"); + tt_resize(t, 16, 4); + assert(row_is(t, 0, "abcdefghijkl") && row_is(t, 1, "")); + assert(tt_cursor_y(t) == 0 && tt_cursor_x(t) == 12); + write_seq(t, "M"); + assert(row_is(t, 0, "abcdefghijklM")); + tt_resize(t, 5, 4); + assert(row_is(t, 0, "abcde") && row_is(t, 1, "fghij") && row_is(t, 2, "klM")); + assert(tt_cursor_y(t) == 2 && tt_cursor_x(t) == 3); + tt_select(t, 0, 0, 2, 2); + char *text = tt_selection_text(t); + assert(text && !strcmp(text, "abcdefghijklM")); + free(text); + tt_resize(t, 20, 4); + assert(row_is(t, 0, "abcdefghijklM")); + term_free(t); + + t = tt_new(6, 6); + write_seq(t, "\033[31mab cdef\033[m\r\nnext"); + term_color red = tt_cell(t, 0, 0)->fg; + tt_resize(t, 12, 6); + assert(row_is(t, 0, "ab cdef") && row_is(t, 1, "next")); + assert(tt_cell(t, 7, 0)->fg == red); + assert(tt_cell(t, 0, 1)->fg == TERM_COLOR_FG); + term_free(t); + + t = tt_new(4, 2); + write_seq(t, "abcdefghijkl"); + assert(tt_history(t) == 1); + tt_resize(t, 12, 2); + assert(tt_history(t) == 0 && row_is(t, 0, "abcdefghijkl")); + assert(tt_cursor_y(t) == 0 && tt_cursor_x(t) == 11); + write_seq(t, "M"); + assert(row_is(t, 1, "M")); + tt_resize(t, 4, 2); + assert(tt_history(t) == 2); + assert(row_is(t, 0, "ijkl") && row_is(t, 1, "M")); + tt_resize(t, 20, 2); + assert(row_is(t, 0, "abcdefghijklM")); + term_free(t); + + t = tt_new(4, 3); + write_seq(t, "abcd"); + tt_resize(t, 8, 3); + write_seq(t, "E"); + assert(row_is(t, 0, "abcdE")); + tt_resize(t, 5, 3); + write_seq(t, "F"); + assert(row_is(t, 0, "abcdE") && row_is(t, 1, "F")); + term_free(t); + + t = tt_new(4, 4); + write_seq(t, "abcdefghijkl\033[?1049h12345678"); + tt_resize(t, 8, 4); + assert(row_is(t, 0, "1234") && row_is(t, 1, "5678")); + tt_resize(t, 16, 4); + write_seq(t, "\033[?1049lM"); + assert(row_is(t, 0, "abcdefghijklM")); + assert(tt_cursor_y(t) == 0 && tt_cursor_x(t) == 13); + term_free(t); + + t = tt_new(4, 2); + write_seq(t, "abcd"); + tt_resize(t, 4, 3); + write_seq(t, "E"); + assert(row_is(t, 0, "abcd") && row_is(t, 1, "E")); + term_free(t); + + t = tt_new(80, 4); + for (int i = 0; i < 1000; i++) write_seq(t, "abcdefgh"); + tt_resize(t, 1, 4); + assert(tt_history(t) == 1000 * 8 - 4); + assert(tt_cursor_y(t) == 3 && tt_cursor_x(t) == 0); + assert(row_is(t, 0, "e") && row_is(t, 3, "h")); + tt_resize(t, 80, 4); + assert(tt_cursor_x(t) < 80 && tt_cursor_y(t) < 4); + write_seq(t, "Z"); + assert(tt_cell(t, 0, 3)->text == 'Z'); + term_free(t); + +} + +static void test_resize_exposes_blank_rows(void) +{ + struct term *t = tt_new(20, 2); + write_seq(t, "kept"); + for (int pass = 0; pass < 2; pass++) { + tt_resize(t, 20, 6); + assert(row_is(t, 0, "kept")); + for (int y = 2; y < 6; y++) { + for (int x = 0; x < 20; x++) { + const struct term_cell *cell = tt_cell(t, x, y); + assert(cell->text == ' ' && cell->fg == TERM_COLOR_FG); + assert(cell->bg == TERM_COLOR_BG && cell->attrs == 0 && cell->width == 1); + } + } + write_seq(t, "\033[6;1H\tfile/"); + assert(row_is(t, 5, " file/")); + write_seq(t, "\033[1;1H"); + tt_resize(t, 20, 2); + } + term_free(t); +} + +static void test_prompt_resize(void) +{ + struct term *t = tt_new(80, 6); + const char *prompt = "~/Documents/GitHub/clernen_pi main> "; + write_seq(t, "\033]133;A;click_events=1\033\\"); + write_seq(t, prompt); + write_seq(t, "\033]133;B\033\\"); + + const struct { int cols; const char *text; bool newline; } cases[] = { + {40, "~/Documents/GitHub/clernen_pi main> ", false}, + {20, ".b/clernen_pi main> ", true}, + {12, ".n_pi main> ", true}, + {30, "~/GitHub/clernen_pi> ", false}, + {80, "~/Documents/GitHub/clernen_pi main> ", false}, + }; + bool newline = false; + for (size_t i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) { + tt_resize(t, cases[i].cols, 6); + write_seq(t, "\r\r"); + if (newline) write_seq(t, "\033[A"); + write_seq(t, "\033]133;A;click_events=1\033\\"); + write_seq(t, cases[i].text); + write_seq(t, "\033]133;B\033\\"); + if (cases[i].newline) write_seq(t, "\r\n"); + write_seq(t, "\033[J"); + char expected[80]; + snprintf(expected, sizeof(expected), "%s", cases[i].text); + expected[strlen(expected) - 1] = 0; + assert(row_is(t, 0, expected)); + for (int y = 1; y < 6; y++) assert(row_is(t, y, "")); + assert(tt_history(t) == 0); + newline = cases[i].newline; + } + + write_seq(t, "\033]133;C\033\\"); + tt_resize(t, 20, 6); + assert(row_is(t, 0, "~/Documents/GitHub/c")); + assert(row_is(t, 1, "lernen_pi main>")); + term_free(t); + + t = tt_new(20, 6); + write_seq(t, "abcdefghijklmnop\r\n\033]133;A\007prompt> input\033]133;B\007"); + tt_resize(t, 8, 6); + assert(row_is(t, 0, "abcdefgh") && row_is(t, 1, "ijklmnop")); + assert(row_is(t, 2, "prompt>")); + tt_resize(t, 30, 6); + assert(row_is(t, 0, "abcdefghijklmnop") && row_is(t, 1, "prompt>")); + term_free(t); +} + +static void test_cursor_line_motion(void) +{ + struct term *t = tt_new(80, 9); + assert(t); + const char *lines = "one\r\ntwo\r\nthree\r\nfour\r\nfive\r\nsix\r\nseven\r\n"; + tt_write(t, lines, strlen(lines)); + const char *rewrite = "\033[6Ftwo revised\033[K\033[2Efour revised\033[K\033[3Fone revised\033[K"; + tt_write(t, rewrite, strlen(rewrite)); + assert(row_is(t, 1, "two revised")); + assert(row_is(t, 3, "four revised")); + assert(row_is(t, 0, "one revised")); + term_free(t); +} + +static void test_relative_cursor_motion(void) +{ + struct term *t = tt_new(10, 5); + assert(t); + const char *text = "0123456789\r\nabcdefghij\r\nklmnopqrst"; + tt_write(t, text, strlen(text)); + const char *sequence = "\033[3`X\033[2aY\033[1eZ\033[2dW"; + tt_write(t, sequence, strlen(sequence)); + assert(tt_cell(t, 2, 2)->text == 'X'); + assert(tt_cell(t, 5, 2)->text == 'Y'); + assert(tt_cell(t, 6, 3)->text == 'Z'); + assert(tt_cell(t, 7, 1)->text == 'W'); + term_free(t); +} + +static void test_top_region_scrollback(void) +{ + static const struct { + const char *scroll; + int lines; + } cases[] = { + {"\n", 1}, + {"\033D", 1}, + {"\033E", 1}, + {"\033[S", 1}, + {"\033[2S", 2}, + {"\033[99S", 4}, + }; + for (size_t i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) { + struct term *t = tt_new(12, 6); + write_seq(t, "one\r\ntwo\r\nthree\r\nfour\r\nPROMPT\r\nSTATUS"); + write_seq(t, "\033[1;4r\033[4;1H"); + write_seq(t, cases[i].scroll); + assert(tt_history(t) == cases[i].lines); + assert(row_is(t, 0, cases[i].lines == 1 ? "two" : + cases[i].lines == 2 ? "three" : "")); + assert(row_is(t, 3, "")); + assert(row_is(t, 4, "PROMPT") && row_is(t, 5, "STATUS")); + term_scroll(t, cases[i].lines); + assert(tt_view_offset(t) == cases[i].lines); + assert(row_is(t, 0, "one") && row_is(t, 1, "two")); + term_free(t); + } + + /* Retain the same history as a full screen after the ring wraps. */ + struct term *t = tt_new(12, 6); + struct term *reference = tt_new(12, 4); + const char *footer = "\033[5;1H\033[31mPROMPT\r\nSTATUS\033[m\033[1;4r"; + tt_write(t, footer, strlen(footer)); + for (int i = 0; i < 2 * TEST_SCROLLBACK; i++) { + char line[32]; + int len = snprintf(line, sizeof(line), "line %04d\r\n", i); + tt_write(t, line, (size_t)len); + tt_write(reference, line, (size_t)len); + } + assert(tt_history(t) == tt_history(reference)); + assert(tt_cursor_x(t) == tt_cursor_x(reference)); + assert(tt_cursor_y(t) == tt_cursor_y(reference)); + assert(row_is(t, 4, "PROMPT") && row_is(t, 5, "STATUS")); + assert(tt_cell(t, 0, 4)->fg == tt_cell(t, 0, 5)->fg); + assert(tt_cell(t, 0, 4)->fg != TERM_COLOR_FG); + for (int i = 0; i <= tt_history(reference); i++) { + assert_same_rows(t, reference, tt_rows(reference)); + term_scroll(t, 1); + term_scroll(reference, 1); + } + term_free(t); + term_free(reference); +} + +static void test_editor_scrolling(void) +{ + struct term *t = tt_new(80, 24); + write_seq(t, "\033[?1049hTITLE"); + for (int i = 1; i <= 20; i++) { + char line[64]; + snprintf(line, sizeof(line), "\033[%d;1Hline %02d", i + 1, i); + write_seq(t, line); + } + write_seq(t, "\033[23;1HHELP\033[24;1HFOOTER"); + + write_seq(t, "\033[4;22r\033[22;1H\n\033[1;24r" + "\033[21;1Hline 21\033[K\r\033[4d"); + assert(row_is(t, 0, "TITLE") && row_is(t, 2, "line 02")); + assert(row_is(t, 3, "line 04") && row_is(t, 19, "line 20")); + assert(row_is(t, 20, "line 21") && row_is(t, 21, "")); + assert(row_is(t, 22, "HELP") && row_is(t, 23, "FOOTER")); + assert(tt_cursor_x(t) == 0 && tt_cursor_y(t) == 3); + assert(tt_history(t) == 0); + term_free(t); + + t = tt_new(12, 6); + write_seq(t, "old\r\nHEAD\r\nA\r\nB\r\nC\r\nD\r\nFOOT"); + assert(tt_history(t) == 1); + write_seq(t, "\033[2;5r\033[3;2H\033[M"); + assert(row_is(t, 0, "HEAD") && row_is(t, 1, "A")); + assert(row_is(t, 2, "C") && row_is(t, 3, "D") && row_is(t, 4, "")); + assert(row_is(t, 5, "FOOT")); + /* DL moves the cursor to the left margin (xterm). */ + assert(tt_cursor_x(t) == 0 && tt_cursor_y(t) == 2); + write_seq(t, "\033[L"); + assert(row_is(t, 2, "") && row_is(t, 3, "C") && row_is(t, 4, "D")); + write_seq(t, "\033[2;1H\033M"); + assert(row_is(t, 1, "") && row_is(t, 2, "A") && row_is(t, 4, "C")); + write_seq(t, "\033[5;1H\033D"); + assert(row_is(t, 1, "A") && row_is(t, 3, "C") && row_is(t, 4, "")); + write_seq(t, "\033[T\033[S"); + assert(row_is(t, 1, "A") && row_is(t, 3, "C")); + + write_seq(t, "\033[5;2r\033[1;999999999999r\033[44m\033[999999999999S"); + assert(row_is(t, 0, "HEAD") && row_is(t, 5, "FOOT")); + for (int r = 1; r <= 4; r++) { + assert(row_is(t, r, "")); + assert(tt_cell(t, 0, r)->bg == tt_cell(t, 11, r)->bg); + assert(tt_cell(t, 0, r)->bg != TERM_COLOR_BG); + } + assert(tt_history(t) == 1); + + write_seq(t, "\033[6;1H\033[M\033[L"); + assert(row_is(t, 5, "FOOT")); + + tt_resize(t, 12, 4); + write_seq(t, "\033[4;1H\033[2Kend\n"); + assert(tt_cursor_y(t) == 3 && row_is(t, 2, "end")); + term_free(t); +} + +static bool pending_decsca_decsed_decsel(void) +{ + struct term *d = tt_new(20, 3); + write_seq(d, "\033[H" "AB" "\033[1\"q" "CD" "\033[0\"q" "EF"); + write_seq(d, "\033[?2J"); + bool ok = row_eq(d, 0, " CD"); + term_free(d); + + struct term *l = tt_new(20, 3); + write_seq(l, "\033[H" "GH" "\033[1\"q" "IJ" "\033[0\"q" "KL"); + write_seq(l, "\033[H\033[?0K"); + ok = ok && row_eq(l, 0, " IJ"); + term_free(l); + return ok; +} + +static bool pending_decfra_decera(void) +{ + struct term *f = tt_new(20, 3); + write_seq(f, "\033[88;1;2;2;4$x"); + bool ok = row_eq(f, 0, " XXX") && row_eq(f, 1, " XXX"); + term_free(f); + + struct term *e = tt_new(20, 3); + write_seq(e, "\033[Habcde\r\nfghij"); + write_seq(e, "\033[1;2;2;3$z"); + ok = ok && row_eq(e, 0, "a de") && row_eq(e, 1, "f ij"); + term_free(e); + return ok; +} + +static bool pending_deccra(void) +{ + struct term *t = tt_new(20, 3); + write_seq(t, "\033[HAB\r\nCD"); + write_seq(t, "\033[1;1;2;2;1;1;5;1$v"); + bool ok = row_eq(t, 0, "AB AB") && row_eq(t, 1, "CD CD"); + term_free(t); + return ok; +} + +static bool pending_decrara(void) +{ + struct term *t = tt_new(20, 3); + write_seq(t, "\033[H\033[1mAB\033[mCD"); + write_seq(t, "\033[1;1;1;4;1$t"); + bool ok = !(tt_cell(t, 0, 0)->attrs & TERM_ATTR_BOLD) && + (tt_cell(t, 2, 0)->attrs & TERM_ATTR_BOLD); + term_free(t); + return ok; +} + +static bool pending_decsera(void) +{ + struct term *t = tt_new(20, 3); + write_seq(t, "\033[H\033[0\"qAB\033[1\"qCD"); + write_seq(t, "\033[1;1;1;4${"); + bool ok = row_eq(t, 0, " CD"); + term_free(t); + return ok; +} + +static bool pending_deccolm(void) +{ + struct term *t = tt_new(80, 4); + write_seq(t, "\033[?40h" "keep" "\033[?3h"); + bool ok = tt_cols(t) == 132 && row_eq(t, 0, "") && + tt_cursor_x(t) == 0 && tt_cursor_y(t) == 0; + write_seq(t, "\033[?3l"); + ok = ok && tt_cols(t) == 80; + term_free(t); + return ok; +} + +static bool pending_decic_decdc(void) +{ + struct term *i = tt_new(10, 2); + write_seq(i, "\033[Habcdef\033[H\033[2'}"); + bool ok = row_eq(i, 0, " abcdef"); + term_free(i); + + struct term *d = tt_new(10, 2); + write_seq(d, "\033[Habcdef\033[H\033[2'~"); + ok = ok && row_eq(d, 0, "cdef"); + term_free(d); + return ok; +} + +static bool pending_sl_sr(void) +{ + struct term *l = tt_new(10, 2); + write_seq(l, "\033[Habcdef\033[2 @"); + bool ok = row_eq(l, 0, "cdef"); + term_free(l); + + struct term *r = tt_new(10, 2); + write_seq(r, "\033[Habcdef\033[2 A"); + ok = ok && row_eq(r, 0, " abcdef"); + term_free(r); + return ok; +} + +static bool pending_sixel(void) +{ + struct term *t = tt_new(20, 6); + write_seq(t, "\033[H\033Pq#0;2;100;100;100#0~~~~\033\\"); + bool ok = tt_cursor_y(t) > 0; + term_free(t); + return ok; +} + +static void test_unimplemented_da_features(void) +{ + static const struct { + const char *name; + bool (*run)(void); + } cases[] = { + { "DECSCA/DECSED/DECSEL selective erase (DA1 6)", pending_decsca_decsed_decsel }, + { "DECFRA/DECERA rectangle fill, erase (DA1 28)", pending_decfra_decera }, + { "DECCRA rectangle copy (DA1 28)", pending_deccra }, + { "DECRARA rectangle attributes (DA1 28)", pending_decrara }, + { "DECSERA rectangle selective erase (DA1 28)", pending_decsera }, + { "DECCOLM 132 columns (DA1 1)", pending_deccolm }, + { "DECIC/DECDC column insert, delete (DA1 21)", pending_decic_decdc }, + { "SL/SR scroll left, right (DA1 21)", pending_sl_sr }, + { "sixel graphics, cursor advance (DA1 4)", pending_sixel }, + }; + size_t total = sizeof(cases) / sizeof(cases[0]), ready = 0; + for (size_t i = 0; i < total; i++) { + if (!cases[i].run()) continue; + printf(" %s now passes: assert it and add its DA1 parameter\n", cases[i].name); + ready++; + } + printf("unimplemented features: %zu pending, %zu ready to promote\n", + total - ready, ready); +} + +static void test_device_attributes(void) +{ + struct term *t = tt_new_opts(80, 24, &reply_cb); + + write_seq(t, "\033[c"); + assert(!strcmp(response, "\033[?62;22c")); + response_len = 0; + write_seq(t, "\033[0c"); + assert(!strcmp(response, "\033[?62;22c")); + + write_seq(t, "\033[>c"); + assert(!strcmp(response, "\033[>0;10;0c")); + + response_len = 0; + write_seq(t, "\033[=c"); + assert(response_len == 0); + + term_free(t); +} + +static void expect_flags(struct term *t, const char *commands, const char *reply) +{ + write_seq(t, commands); + write_seq(t, "\033[?u"); + assert(!strcmp(response, reply)); +} +static void test_keyboard_negotiation(void) +{ + struct term *t = tt_new_opts(80, 24, &reply_cb); + assert(!term_modes(t).app_cursor); + write_seq(t, "\033[?1h"); + assert(term_modes(t).app_cursor); + write_seq(t, "\033[?1l"); + assert(!term_modes(t).app_cursor); + expect_flags(t, "", "\033[?0u"); + expect_flags(t, "\033[=1u", "\033[?1u"); + expect_flags(t, "\033[>3u\033[>31u", "\033[?31u"); + expect_flags(t, "\033[<u", "\033[?3u"); + expect_flags(t, "\033[=4;2u", "\033[?7u"); + expect_flags(t, "\033[=2;3u", "\033[?5u"); + expect_flags(t, "\033[=31;99u", "\033[?5u"); + expect_flags(t, "\033[<u", "\033[?1u"); + expect_flags(t, "\033[<999999999999999999999999u", "\033[?0u"); + expect_flags(t, "\033[>1:2u\033[=3;1;4u\033[1>31u", "\033[?0u"); + expect_flags(t, "\033[=255u", "\033[?31u"); + expect_flags(t, "\033[>u", "\033[?0u"); + expect_flags(t, "\033[<u", "\033[?31u"); + expect_flags(t, "\033[>1u\033[>2u\033[<2u", "\033[?31u"); + expect_flags(t, "\033[?1049h", "\033[?0u"); + expect_flags(t, "\033[>3u", "\033[?3u"); + expect_flags(t, "\033[?1049l", "\033[?31u"); + + for (int i = 0; i < 1000; i++) write_seq(t, "\033[>1u"); + expect_flags(t, "\033[<u", "\033[?1u"); + expect_flags(t, "\033[<2147483647u", "\033[?0u"); + write_seq(t, "hello"); + assert(row_is(t, 0, "hello")); + term_free(t); +} + +static void expect_selection(struct term *t, int x1, int y1, int x2, int y2, const char *want) +{ + tt_select(t, x1, y1, x2, y2); + char *text = tt_selection_text(t); + if (!text || strcmp(text, want)) + fprintf(stderr, "selection: got '%s', want '%s'\n", text ? text : "NULL", want); + assert(text && !strcmp(text, want)); + free(text); +} + +static void test_wide_characters(void) +{ + struct term *t = tt_new(8, 4); + write_seq(t, "A😀界B"); + assert(tt_cursor_x(t) == 6); + assert(tt_cell(t, 1, 0)->text == 0x1f600 && tt_cell(t, 3, 0)->text == 0x754c); + expect_selection(t, 2, 0, 2, 0, "😀"); + write_seq(t, "\033[3Gx\033[4GX"); + assert(row_is(t, 0, "A xX B")); + write_seq(t, "\033[2J\033[HA😀BC\033[2G\033[@"); + expect_selection(t, 0, 0, 7, 0, "A 😀BC"); + write_seq(t, "\033[3G\033[P"); + expect_selection(t, 0, 0, 7, 0, "A BC"); + write_seq(t, "\033[2J\033[HA😀BC\033[3G\033[X"); + assert(row_is(t, 0, "A BC")); + + write_seq(t, "\033[4;1H😀\r \033[4;1H😀\033[4;2H\033[38;2;1;2;3mX"); + assert(row_is(t, 3, " X")); + term_free(t); + + t = tt_new(5, 8); + write_seq(t, "abcd😀Z"); + assert(tt_cursor_y(t) == 1 && tt_cursor_x(t) == 3); + expect_selection(t, 0, 0, 4, 1, "abcd😀Z"); + tt_resize(t, 3, 8); + assert(tt_cell(t, 1, 1)->text == 0x1f600 && tt_cell(t, 0, 2)->text == 'Z'); + assert(tt_cursor_y(t) == 2 && tt_cursor_x(t) == 1); + expect_selection(t, 0, 0, 2, 2, "abcd😀Z"); + tt_resize(t, 12, 8); + assert(tt_cursor_x(t) == 7 && tt_cursor_y(t) == 0); + expect_selection(t, 0, 0, 11, 0, "abcd😀Z"); + tt_resize(t, 1, 8); + tt_resize(t, 12, 8); + expect_selection(t, 0, 0, 11, 0, "abcd😀Z"); + assert(tt_cursor_x(t) == 7); + term_free(t); + + t = tt_new(4, 3); + write_seq(t, "abc©️)"); + assert(tt_cursor_y(t) == 1 && tt_cursor_x(t) == 3); + expect_selection(t, 0, 0, 3, 1, "abc©️)"); + write_seq(t, "\033[2J\033[H😀︎x"); + assert(tt_cursor_x(t) == 2 && tt_cell(t, 1, 0)->text == 'x'); + expect_selection(t, 0, 0, 3, 0, "😀︎x"); + write_seq(t, "\033[2J\033[H\033[?7labc😀"); + assert(tt_cursor_y(t) == 0 && tt_cursor_x(t) == 3); + term_free(t); + + t = tt_new(4, 2); + write_seq(t, "😀😀界界AB"); + term_scroll(t, 1); + expect_selection(t, 0, 0, 3, 1, "😀😀界界"); + term_scroll_to(t, 0); + tt_resize(t, 10, 2); + expect_selection(t, 0, 0, 9, 0, "😀😀界界AB"); + term_free(t); +} + +static void test_wide_runs(void) +{ + static const char *const runs[] = { + "日本語漢字仮名交じり文中文測試終端輸出性能比較日本語漢字仮名交じり文", + " 日本語漢字仮名交じり文中文測試終端輸出性能比較", + "┌─────────────────────┬─────────────────────┐", + "日本語漢字仮名交じり文abc中文測試終端輸出性能比較", + "日本語漢字仮名交じり文̈中文測試終端輸出性能比較", + "한국어를테스트하는줄을적습니다한국어를테스트하는줄을적습니다", + "！＂＃＄％＆＇（）＊＋，－．／０１２３４５６７８９", + "日本語漢字仮名交じり文😀中文測試終端輸出性能比較", + }; + for (size_t k = 0; k < sizeof(runs) / sizeof(*runs); k++) { + char text[512]; + snprintf(text, sizeof(text), "\033[999;1H%s", runs[k]); + struct term *bulk = tt_new(120, 6); + struct term *split = tt_new(120, 6); + tt_write(bulk, text, strlen(text)); + write_seq(split, text); + assert_same_screen(bulk, split); + + expect_selection(bulk, 0, 5, 119, 5, runs[k]); + term_free(bulk); + term_free(split); + } +} + +static void test_invalid_utf8_runs(void) +{ + static const unsigned char invalid[][3] = { + {0xe0, 0x82, 0xa1}, {0xed, 0xa0, 0x80}, {0xe0, 0x9f, 0xbf}, + }; + for (size_t k = 0; k < sizeof(invalid) / sizeof(*invalid); k++) { + char input[128]; + size_t n = 0; + for (int i = 0; i < 32; i++) { + memcpy(input + n, invalid[k], 3); + n += 3; + } + memcpy(input + n, "END", 3); + n += 3; + struct term *bulk = tt_new(120, 4); + struct term *split = tt_new(120, 4); + write_seq(bulk, "\033[4;1H\n"); + write_seq(split, "\033[4;1H\n"); + tt_write(bulk, input, n); + for (size_t i = 0; i < n; i++) tt_write(split, input + i, 1); + /* Each byte is a maximal invalid subpart: one U+FFFD each. */ + for (int c = 0; c < 96; c++) + assert(tt_cell_cp(bulk, tt_cell(bulk, c, 3)) == 0xFFFD); + for (int c = 0; c < 3; c++) + assert(tt_cell_cp(bulk, tt_cell(bulk, 96 + c, 3)) == (uint32_t)"END"[c]); + assert_same_screen(bulk, split); + term_free(bulk); + term_free(split); + } +} + +static void test_recycled_scrollback_blanks(void) +{ + for (int mode = 0; mode < 3; mode++) { + struct term *t = tt_new(32, 3); + write_seq(t, "\033[3;1H"); + const char *line = mode == 2 ? "abcdefghijklmnopqrstuvwxyz\r\033[S" : + "abcdefghijklmnopqrstuvwxyz\r\n"; + for (int i = 0; i < TEST_SCROLLBACK + 10; i++) { + if (mode == 1) write_seq(t, line); + else tt_write(t, line, strlen(line)); + } + write_seq(t, "\tfile/"); + assert(row_is(t, 2, " file/")); + term_free(t); + } +} + +/* Scrolling the view by any amount lands inside the scrollback, and a + * scroll that cannot move leaves the frame alone. */ +static void test_scroll_view_clamps_extreme_deltas(void) +{ + struct term *t = tt_new(20, 3); + for (int i = 0; i < 10; i++) + write_seq(t, "line\r\n"); + int max = tt_history(t); + assert(max > 1); + struct term_snapshot *snap = term_snapshot_new(); + term_scroll(t, 1); + term_scroll(t, INT_MAX); + assert(tt_view_offset(t) == max); + uint64_t version = view_version(snap, t); + term_scroll(t, 1); + assert(view_version(snap, t) == version); + term_scroll(t, -1); + term_scroll(t, INT_MIN); + assert(tt_view_offset(t) == 0); + version = view_version(snap, t); + term_scroll(t, -1); + assert(view_version(snap, t) == version); + term_snapshot_free(snap); + term_free(t); +} + +static void test_batched_recycled_lines(void) +{ + static const char *const endings[] = { + "short\r\nx\r\nyz\r\n", + "short\r\nx\r\nyz\r\n\tfile/", + "abcdefghijklmnopqrstuvwxyz012345more\r\nx\r\nyz\r\n\033[2CZ", + "short\r\nx\r\nyz\r\n界\xcc\x81", + "short\r\nx\r\nyz\r\npartial\r", + }; + for (size_t k = 0; k < sizeof(endings) / sizeof(endings[0]); k++) { + struct term *bulk = tt_new(32, 3); + struct term *bytes = tt_new(32, 3); + const char *old = "\033[31;44mabcdefghijklmnopqrst界界界\r\n"; + for (int i = 0; i < TEST_SCROLLBACK + 10; i++) { + tt_write(bulk, old, strlen(old)); + write_seq(bytes, old); + } + write_seq(bulk, "\033[0;1;3m"); + write_seq(bytes, "\033[0;1;3m"); + tt_write(bulk, endings[k], strlen(endings[k])); + write_seq(bytes, endings[k]); + + assert_same_screen(bulk, bytes); + for (int i = 0; i < tt_history(bulk); i += 3) { + term_scroll(bulk, 3); + term_scroll(bytes, 3); + assert_same_screen(bulk, bytes); + } + term_free(bulk); + term_free(bytes); + } +} + +static void test_full_width_line_endings(void) +{ + static const char *const feeds[] = { + "aaaa\r\nbbbb\r\ncccc\r\ndddd\r\neeee\r\nffff", + "aaaa\r\nbbbb\r\ncccc\r\ndddd\neeee", + "aaaa\r\nbbbb\r\ncccc\r\ndddd\re", + "aaaa\r\nbbbb\r\ncccc\r\nddddeeeeff", + "aa\r\nbb\r\ncc\r\nd中d\r\nef", + }; + for (size_t k = 0; k < sizeof(feeds) / sizeof(feeds[0]); k++) { + struct term *bulk = tt_new(4, 3); + struct term *bytes = tt_new(4, 3); + tt_write(bulk, feeds[k], strlen(feeds[k])); + write_seq(bytes, feeds[k]); + assert_same_screen(bulk, bytes); + term_free(bulk); + term_free(bytes); + } + + struct term *t = tt_new(4, 3); + tt_write(t, feeds[0], strlen(feeds[0])); + assert(tt_history(t) == 3); + assert(row_is(t, 0, "dddd") && row_is(t, 1, "eeee") && row_is(t, 2, "ffff")); + term_free(t); +} + +/* Compare the screen, cursor, and scrollback for each input mode. + * + * Input mode | Byte-wise | 64 KiB bulk | 512 KiB single write + * Raw LF | Reference | Must match | Must match + * Newline conversion | Reference | Must match | Must match + * + * The input length selects the stream path. The 512 KiB write also has + * enough newlines to test the backward scan when newline conversion is + * active. + */ +static void test_bulk_line_flood(void) +{ + static const struct term_vector2i geoms[] = { + {.x = 80, .y = 24}, {.x = 27, .y = 8}, {.x = 10, .y = 5}, {.x = 120, .y = 80}, + }; + for (size_t g = 0; g < 2 * sizeof(geoms) / sizeof(geoms[0]); g++) { + int cols = geoms[g / 2].x, rows = geoms[g / 2].y; + size_t cap = 1u << 19; + char *buf = malloc(cap); + assert(buf); + size_t n = 0; + unsigned state = 12345; + while (n < cap / 2) { + state = state * 1103515245u + 12345u; + unsigned pick = (state >> 16) % 27; + buf[n++] = pick == 26 ? '\n' : (char)('a' + pick); + } + for (int k = -1; k <= 1; k++) { + memset(buf + n, 'x', (size_t)(cols + k)); + n += (size_t)(cols + k); + buf[n++] = '\n'; + memset(buf + n, 'y', (size_t)(2 * cols + k)); + n += (size_t)(2 * cols + k); + buf[n++] = '\n'; + } + while (n < cap) { + state = state * 1103515245u + 12345u; + unsigned pick = (state >> 16) % 27; + buf[n++] = pick == 26 ? '\n' : (char)('a' + pick); + } + + struct term *bulk = tt_new(cols, rows); + struct term *stream = tt_new(cols, rows); + struct term *bytes = tt_new(cols, rows); + unsigned flags = g % 2 != 0 ? TERM_FEED_ONLCR : 0; + for (size_t i = 0; i < n; i += 65536) + term_feed(bulk, buf + i, n - i < 65536 ? n - i : 65536, flags); + /* One large write has enough newlines to discard old output. */ + term_feed(stream, buf, n, flags); + for (size_t i = 0; i < n; i++) term_feed(bytes, buf + i, 1, flags); + + assert_same_screen(bulk, bytes); + assert_same_screen(stream, bytes); + for (int i = 0; i < tt_history(bulk); i += 7) { + term_scroll(bulk, 7); + term_scroll(stream, 7); + term_scroll(bytes, 7); + assert_same_screen(bulk, bytes); + assert_same_screen(stream, bytes); + } + term_free(bulk); + term_free(stream); + term_free(bytes); + free(buf); + } +} + +/* Dense LF input must keep its text through edits and buffer moves. */ +static void test_ascii_flood_edits(void) +{ + static const int widths[] = {1, 2, 7, 120}; + char input[65536]; + unsigned state = 43; + for (size_t i = 0; i < sizeof(input); i++) { + state = state * 1103515245u + 12345u; + unsigned pick = (state >> 16) % 5; + input[i] = pick == 4 ? '\n' : (char)('a' + pick); + } + memset(input + 1024, '\n', 256); + static const char edit[] = + "\033[1;1Hé漢x̂\033[T\033[2;1HEND" + "\033[?1049hALT\033[?1049l\033[6;1Htail\n"; + for (size_t w = 0; w < sizeof(widths) / sizeof(widths[0]); w++) { + for (size_t chunk = 4096; chunk <= sizeof(input); chunk *= 16) { + struct term *bulk = tt_new(widths[w], 6); + struct term *bytes = tt_new(widths[w], 6); + for (size_t i = 0; i < sizeof(input); i += chunk) + term_feed(bulk, input + i, chunk, TERM_FEED_ONLCR); + for (size_t i = 0; i < sizeof(input); i++) + term_feed(bytes, input + i, 1, TERM_FEED_ONLCR); + assert_same_screen(bulk, bytes); + term_feed(bulk, edit, sizeof(edit) - 1, TERM_FEED_ONLCR); + for (size_t i = 0; i < sizeof(edit) - 1; i++) + term_feed(bytes, edit + i, 1, TERM_FEED_ONLCR); + assert_same_screen(bulk, bytes); + for (int i = 0; i < tt_history(bulk); i += 6) { + term_scroll(bulk, 6); + term_scroll(bytes, 6); + assert_same_screen(bulk, bytes); + } + term_scroll_to(bulk, 0); + term_scroll_to(bytes, 0); + assert(tt_resize(bulk, widths[w] + 3, 7)); + assert(tt_resize(bytes, widths[w] + 3, 7)); + assert_same_screen(bulk, bytes); + term_free(bulk); + term_free(bytes); + } + } +} + +/* Newline modes: 0 is a plain line feed, 1 is LNM (CSI 20 h), and 2 is + * newline conversion on the feed. */ +static void test_margin_linefeed_bursts(void) +{ + static const char *bursts[] = { + "\n\n", "\n\n\n\n\n", "\n\n\n\n\n\n\n\n", + "\r\n\r\n", "\r\n\r\n\r\n\r\n\r\n\r\n", + "\n\n\r\n\r\n\n\n", + }; + for (size_t k = 0; k < sizeof(bursts) / sizeof(bursts[0]); k++) { + for (int mode = 0; mode < 3; mode++) { + struct term *bulk = tt_new(8, 8); + struct term *bytes = tt_new(8, 8); + const char *setup = "\033[1;1Htop\033[2;1Ha\033[3;1Hb\033[4;1Hc" + "\033[5;1Hd\033[6;1He\033[7;1Hf\033[8;1Hfoot" + "\033[2;7r\033[7;3H"; + tt_write(bulk, setup, strlen(setup)); + tt_write(bytes, setup, strlen(setup)); + if (mode == 1) { + write_seq(bulk, "\033[20h"); + write_seq(bytes, "\033[20h"); + } + unsigned flags = mode == 2 ? TERM_FEED_ONLCR : 0; + term_feed(bulk, bursts[k], strlen(bursts[k]), flags); + feed_seq(bytes, bursts[k], flags); + assert_same_screen(bulk, bytes); + assert(row_is(bulk, 0, "top")); + assert(row_is(bulk, 7, "foot")); + term_free(bulk); + term_free(bytes); + } + } +} + +static void test_alt_linefeed_burst_after_resize(void) +{ + struct term *bulk = tt_new(8, 5); + struct term *bytes = tt_new(8, 5); + const char *setup = "\033[?1049h\033[5;1H\n\n\n\n"; + tt_write(bulk, setup, strlen(setup)); + write_seq(bytes, setup); + tt_resize(bulk, 8, 3); + tt_resize(bytes, 8, 3); + const char *burst = "\n\n\n\n\r\n\r\n"; + tt_write(bulk, burst, strlen(burst)); + write_seq(bytes, burst); + assert_same_screen(bulk, bytes); + tt_resize(bulk, 12, 4); + tt_resize(bytes, 12, 4); + assert_same_screen(bulk, bytes); + term_free(bulk); + term_free(bytes); +} + +/* + * Regression: the alternate screen scrolls by moving its top row. A resize + * changed the row count but kept that top row, so rows came back out of + * order, and further line feeds indexed past the row tables. + */ +static void test_alt_resize_keeps_row_order(void) +{ + const char *setup = "\033[?1049h\033[5;1H\n\n\n" + "\033[1;1HA\033[2;1HB\033[3;1HC\033[4;1HD\033[5;1HE"; + + struct term *grow = tt_new(20, 5); + write_seq(grow, setup); + tt_resize(grow, 20, 8); + assert(row_is(grow, 0, "A") && row_is(grow, 4, "E")); + assert(row_is(grow, 5, "") && row_is(grow, 7, "")); + term_free(grow); + + /* The cursor is on the last row, so shrinking scrolls A and B off. */ + struct term *shrink = tt_new(20, 5); + write_seq(shrink, setup); + tt_resize(shrink, 20, 3); + assert(row_is(shrink, 0, "C") && row_is(shrink, 2, "E")); + term_free(shrink); +} + +static void test_alt_shrink_then_flood(void) +{ + struct term *t = tt_new(80, 40); + tt_write(t, "\033[?1049h\033[40;1H", 16); + for (int i = 0; i < 35; i++) + tt_write(t, "x\n", 2); + tt_resize(t, 80, 10); + /* More lines than the ring holds, one write per line. */ + char line[32]; + for (int i = 0; i < 6000; i++) { + int len = snprintf(line, sizeof(line), "\rline %04d\n", i); + tt_write(t, line, (size_t)len); + } + assert(row_is(t, 8, "line 5999") && row_is(t, 0, "line 5991")); + term_free(t); +} + +/* Regression: ED 3 on the alternate screen kept the history count, so the + * normal screen came back with blank history rows. */ +static void test_alt_erase_scrollback(void) +{ + struct term *t = tt_new(20, 5); + for (int i = 0; i < 30; i++) + tt_write(t, "line\r\n", 6); + assert(tt_history(t) > 0); + write_seq(t, "\033[?1049h\033[3J\033[?1049l"); + assert(tt_history(t) == 0); + term_free(t); +} + +/* Regression: a snapshot knew its terminal by address. A new terminal that + * reused a freed one's memory then showed the old terminal's rows. */ +static void test_snapshot_follows_new_terminal(void) +{ + struct term_snapshot *s = term_snapshot_new(); + for (int i = 0; i < 8; i++) { + struct term *t = tt_new(10, 2); + char text[2] = { (char)('a' + i), 0 }; + tt_puts(t, text); + assert(term_snapshot_update(s, t)); + assert(term_snapshot_row(s, 0)[0].text == (uint32_t)text[0]); + term_free(t); + } + term_snapshot_free(s); +} + +static void test_string_payloads(void) +{ + static const char *prefixes[] = {"\033]0;", "\033]7;file://host/", "\033P"}; + static const char *ends[] = {"\007", "\033\\"}; + static const size_t chunks[] = {1, 7, 4096}; + for (size_t p = 0; p < 3; p++) { + for (size_t e = 0; e < 2; e++) { + for (int large = 0; large < 2; large++) { + char input[1024]; + size_t n = (size_t)snprintf(input, sizeof(input), "%s⠋ çalışma/clernen_pi", prefixes[p]); + if (large) { + memset(input + n, 'a', 600); + n += 600; + } + n += (size_t)snprintf(input + n, sizeof(input) - n, "%sOK", ends[e]); + for (size_t c = 0; c < 3; c++) { + title[0] = 0; + response_len = 0; + struct term *t = tt_new_opts(80, 4, &title_cb); + for (size_t i = 0; i < n;) { + size_t take = n - i < chunks[c] ? n - i : chunks[c]; + tt_write(t, input + i, take); + i += take; + } + assert(row_is(t, 0, "OK")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + assert(response_len == 0); + if (p == 0 && !large) + assert(!strcmp(title, "⠋ çalışma/clernen_pi")); + if (large) { + if (p == 0) { + assert(!strncmp(title, "⠋ çalışma/clernen_pi", + strlen("⠋ çalışma/clernen_pi"))); + assert(title[sizeof(title) - 1] == 0); + } else { + assert(title[0] == 0); + } + } + term_free(t); + } + } + } + } +} + +static void test_newline_policy_survives_reset(void) +{ + static const char *const resets[] = { "\033c", "\033[!p", "\033[20l" }; + for (size_t i = 0; i < sizeof(resets) / sizeof(resets[0]); i++) { + struct term *t = tt_new(20, 6); + term_feed(t, "a\n", 2, TERM_FEED_ONLCR); + assert(tt_cursor_x(t) == 0); + write_seq(t, resets[i]); + term_feed(t, "b\nc", 3, TERM_FEED_ONLCR); + assert(tt_cursor_x(t) == 1); + assert(row_is(t, tt_cursor_y(t), "c")); + term_free(t); + } +} + +static void test_lnm_mode_alone_is_not_sticky(void) +{ + struct term *t = tt_new(20, 6); + write_seq(t, "\033[20h\033c"); + tt_write(t, "b\nc", 3); + assert(tt_cursor_x(t) == 2); + term_free(t); +} + +/* A C0 control ends an unfinished UTF-8 sequence, which prints U+FFFD. A + * lone continuation byte prints U+FFFD too. */ +static void test_c0_ends_utf8_sequence(void) +{ + static const char input[] = "\xc3\n\xa9z"; + for (size_t split = 0; split < sizeof(input) - 1; split++) { + struct term *t = tt_new(20, 4); + tt_write(t, input, split); + tt_write(t, input + split, sizeof(input) - 1 - split); + assert(tt_cell_cp(t, tt_cell(t, 0, 0)) == 0xFFFD); + assert(tt_cell_cp(t, tt_cell(t, 1, 1)) == 0xFFFD); + assert(tt_cell_cp(t, tt_cell(t, 2, 1)) == 'z'); + term_free(t); + } +} + +/* A bulk stream of long lines must behave like byte-wise input after every + * later operation that touches those rows: overwriting, scroll-region + * rotation, reflow, alternate-screen round trips, and selection. */ +static void test_bulk_rows_survive_edits(void) +{ + enum { COLS = 40, ROWS = 6 }; + static char flood[1 << 16]; + for (size_t i = 0; i < sizeof(flood); i++) + flood[i] = (char)('a' + (i * 7 + i / COLS) % 26); + + struct term *bulk = tt_new(COLS, ROWS); + struct term *bytes = tt_new(COLS, ROWS); + write_seq(bulk, "\033[1;4;38;2;9;8;7;48;2;1;2;3m"); + write_seq(bytes, "\033[1;4;38;2;9;8;7;48;2;1;2;3m"); + tt_write(bulk, flood, sizeof(flood)); + for (size_t i = 0; i < sizeof(flood); i++) tt_write(bytes, flood + i, 1); + assert_same_screen(bulk, bytes); + + /* Overwrite the middle of a streamed row, then erase part of another. */ + const char *edit = "\033[m\033[2;5HXYZ\033[3;1H\033[K\033[4;10H\033[1K"; + write_seq(bulk, edit); + write_seq(bytes, edit); + assert_same_screen(bulk, bytes); + assert(tt_view_cell(bulk, 4, 1)->text == 'X'); + + /* Scroll region rotation moves row pointers between ring slots. */ + const char *region = "\033[2;5r\033[5;1H\n\n\033[T\033[r"; + write_seq(bulk, region); + write_seq(bytes, region); + assert_same_screen(bulk, bytes); + + /* Alternate screen and back, then view history. */ + write_seq(bulk, "\033[?1049hALT\033[?1049l"); + write_seq(bytes, "\033[?1049hALT\033[?1049l"); + assert_same_screen(bulk, bytes); + for (int i = 0; i < 3; i++) { + term_scroll(bulk, 100); + term_scroll(bytes, 100); + assert_same_screen(bulk, bytes); + } + term_scroll_to(bulk, 0); + term_scroll_to(bytes, 0); + + /* Selection text reads rows straight from history. */ + term_scroll(bulk, 50); + term_scroll(bytes, 50); + tt_select(bulk, 0, 0, COLS - 1, 2); + tt_select(bytes, 0, 0, COLS - 1, 2); + char *a = tt_selection_text(bulk), *b = tt_selection_text(bytes); + assert(a && b && !strcmp(a, b) && strlen(a) >= COLS); + free(a); + free(b); + term_scroll_to(bulk, 0); + term_scroll_to(bytes, 0); + + /* Reflow narrower and wider keeps history identical. */ + tt_resize(bulk, 29, ROWS); + tt_resize(bytes, 29, ROWS); + assert_same_screen(bulk, bytes); + tt_resize(bulk, 61, ROWS + 2); + tt_resize(bytes, 61, ROWS + 2); + assert_same_screen(bulk, bytes); + term_scroll(bulk, 20); + term_scroll(bytes, 20); + assert_same_screen(bulk, bytes); + + term_free(bulk); + term_free(bytes); +} + +static void test_erase_keeps_used_rows(void) +{ + for (int row = 0; row < 40; row++) { + struct term *t = tt_new(20, 40); + char seq[64]; + int len = snprintf(seq, sizeof(seq), "\033[%d;1Hsaved\033[2J", row + 1); + tt_write(t, seq, (size_t)len); + assert(tt_history(t) == row + 1); + for (int y = 0; y < 40; y++) + assert(row_eq(t, y, "")); + term_scroll(t, 1); + assert(row_eq(t, 0, "saved")); + term_free(t); + } +} + +static void test_blank_rows(void) +{ + /* ED 2 pushes used rows to scrollback; the erased screen reads as + * default-color blanks through every accessor. */ + struct term *t = tt_new(10, 4); + write_seq(t, "one\r\ntwo\r\nthree"); + write_seq(t, "\033[2J"); + assert(tt_history(t) == 3); + for (int y = 0; y < 4; y++) + for (int x = 0; x < 10; x++) { + const struct term_cell *c = tt_view_cell(t, x, y); + assert(c->text == ' ' && c->fg == TERM_COLOR_FG && + c->bg == TERM_COLOR_BG); + assert(tt_cell(t, x, y)->text == ' '); + } + term_scroll(t, 3); + assert(row_is(t, 0, "one") && row_is(t, 1, "two") && row_is(t, 2, "three")); + term_scroll_to(t, 0); + + /* A scrolled-in blank row reads as blanks, then accepts new text. */ + write_seq(t, "\033[4;1Hlast\r\n\r\n\r\n"); + for (int x = 0; x < 10; x++) + assert(tt_view_cell(t, x, 3)->text == ' '); + write_seq(t, "\033[4;6Hhi"); + assert(row_is(t, 3, " hi")); + term_free(t); + + /* A spaces-only row does not extend ED 2 history. */ + t = tt_new(10, 6); + write_seq(t, "top\033[6;1H "); + assert(tt_history(t) == 0); + write_seq(t, "\033[2J"); + assert(tt_history(t) == 1); + term_scroll(t, 1); + assert(row_is(t, 0, "top")); + term_scroll_to(t, 0); + + /* Blank rows survive a resize and still read as blanks. */ + tt_resize(t, 20, 6); + for (int x = 0; x < 20; x++) + assert(tt_cell(t, x, 5)->text == ' ' && + tt_cell(t, x, 5)->bg == TERM_COLOR_BG); + term_free(t); + + /* Erasing a blank row with a color still produces a colored row. */ + t = tt_new(10, 4); + write_seq(t, "\033[2J\033[41m\033[2K"); + assert(tt_cell(t, 0, 0)->bg != TERM_COLOR_BG); + assert(tt_cell(t, 9, 0)->bg == tt_cell(t, 0, 0)->bg); + term_free(t); +} + +static void test_csi_text_chunks(void) +{ + static const char *cases[] = { + "\033[2;3Hhello\033[2DXY\033[K\033[1A\033[3C!\r\n", + "\033[H日本語\033[2GXY\033[7mabc\033[m!\033[2Knext", + "\033[Hscreen\033[2J\033[Hnext\033[2K\r\n", + "\033[H012345678901234567890123\033[2DXY\033[J!", + "\033[?7l\033[H012345678901234567890123\033[1CXY", + "\033[41m\033[2J\033[Htext\033[0Jtail\033[1Jend", + "\033[2;3Htext\033[3Jnext\033[0K\033[1K\033[2K!", + "\033[?6h\033[2;3Htext\033[Jnext", + "\033[16;1Ha\033[17;1Hb\033[32;1Hc\033[40;1Hd\033[2J", + "\033[40;1H \033[17;1Ha\033[2J\033[Hnext", + "\033[?1049h\033[40;1H\n\033[17;1Htext\033[2J", + }; + for (size_t c = 0; c < sizeof(cases) / sizeof(cases[0]); c++) { + const char *seq = cases[c]; + size_t len = strlen(seq); + struct term *expected = tt_new(20, 40); + write_seq(expected, seq); + for (size_t split = 0; split <= len; split++) { + struct term *actual = tt_new(20, 40); + tt_write(actual, seq, split); + tt_write(actual, seq + split, len - split); + assert(tt_cursor_x(actual) == tt_cursor_x(expected)); + assert(tt_cursor_y(actual) == tt_cursor_y(expected)); + assert(tt_history(actual) == tt_history(expected)); + int history = tt_history(expected); + term_scroll(actual, history); + term_scroll(expected, history); + for (int page = 0; page <= history; page++) { + assert_same_rows(actual, expected, 40); + term_scroll(actual, -1); + term_scroll(expected, -1); + } + term_free(actual); + } + term_free(expected); + } +} + +/* A terminal grows past its first reservation: taller than 256 rows and + * wider than 2048 columns, on both screens. History and the saved screen + * keep their text. */ +static void test_grow_past_reservation(void) +{ + struct term *t = tt_new(40, 10); + for (int i = 0; i < 30; i++) { + char line[32]; + int len = snprintf(line, sizeof(line), "hist %02d\r\n", i); + tt_write(t, line, (size_t)len); + } + int history = tt_history(t); + assert(history > 0); + + tt_resize(t, 40, 700); + assert(tt_rows(t) == 700); + tt_write(t, "\033[700;1Hbottom", 14); + assert(row_is(t, 699, "bottom")); + + tt_resize(t, 3000, 700); + assert(tt_cols(t) == 3000); + tt_write(t, "\033[698;2995Hedge!", 16); + assert(tt_cell(t, 2994, 697)->text == 'e' && tt_cell(t, 2998, 697)->text == '!'); + + tt_write(t, "\033[700;1H", 8); + for (int i = 0; i < 1000; i++) { + char line[32]; + int len = snprintf(line, sizeof(line), "\r\nfill %04d", i); + tt_write(t, line, (size_t)len); + } + int history_before = tt_history(t); + assert(history_before > 0); + + tt_write(t, "\033[?1049h", 8); + tt_write(t, "alt", 3); + tt_resize(t, 3000, 2500); + assert(tt_rows(t) == 2500); + assert(row_is(t, 0, "alt")); + tt_write(t, "\033[?1049l", 8); + + bool found = false; + term_scroll(t, tt_history(t)); + for (int y = 0; y < tt_rows(t) && !found; y++) + found = row_eq(t, y, "hist 00"); + assert(found); + term_scroll(t, -tt_history(t)); + bool last = false; + for (int y = 0; y < tt_rows(t) && !last; y++) + last = row_eq(t, y, "fill 0999"); + assert(last); + term_free(t); +} + +struct concurrent_worker_args { + struct term *t; + _Atomic bool done; +}; + +static void *concurrent_feed_worker(void *arg) +{ + struct concurrent_worker_args *w = arg; + static const char stream[] = + "The quick brown fox jumps over the lazy dog.\r\n" + "\033[31mRed text\033[0m and \033[32mgreen text\033[0m.\r\n" + "Some unicode: 漢字, ひらがな, and emojis: 😀🚀🎉\r\n" + "\033[2J\033[HCleared and reset\r\n" + "Line 1\r\nLine 2\r\nLine 3\r\nLine 4\r\nLine 5\r\n" + "\033]7;file:///tmp/dir\007\033]22;pointer\007" + "\033]8;;https://example.com\007link\033]8;;\007\r\n"; + size_t len = strlen(stream); + + for (int i = 0; i < 2000; i++) { + term_feed(w->t, stream, len, 0); + } + atomic_store(&w->done, true); + return NULL; +} + +static void test_concurrent_feed_and_snapshot(void) +{ + struct term *t = term_new(&(struct term_options){ + .size = { .x = 80, .y = 24 }, + .scrollback = 1000, + }); + assert(t != NULL); + + struct concurrent_worker_args w = { .t = t, .done = false }; + pthread_t th; + int err = pthread_create(&th, NULL, concurrent_feed_worker, &w); + assert(err == 0); + + struct term_snapshot *snap = term_snapshot_new(); + assert(snap != NULL); + + while (!atomic_load(&w.done)) { + bool ok = term_snapshot_update(snap, t); + assert(ok); + struct term_vector2i sz = term_size(t); + assert(sz.x > 0 && sz.y > 0); + term_scroll(t, 2); + term_scroll_to(t, 0); + struct term_modes m = term_modes(t); + (void)m; + struct term_cursor c = term_cursor(t); + (void)c; + + /* Every reader is safe while another thread feeds. */ + char buf[256]; + term_select_start(t, (struct term_vector2i){ .x = 0, .y = 0 }, false, TERM_SELECT_WORD); + term_select_extend(t, (struct term_vector2i){ .x = 40, .y = 10 }, false); + (void)term_has_selection(t); + (void)term_selection_text(t, buf, sizeof(buf)); + (void)term_link_at(t, (struct term_vector2i){ .x = 1, .y = 1 }, buf, sizeof(buf)); + (void)term_cwd(t, buf, sizeof(buf)); + (void)term_pointer_shape(t, buf, sizeof(buf)); + (void)term_history_rows(t); + (void)term_history_text(t, 5, buf, sizeof(buf)); + struct term_scrollback sb = term_scrollback(t); + assert(sb.offset >= 0 && sb.total >= 0); + } + + pthread_join(th, NULL); + bool ok = term_snapshot_update(snap, t); + assert(ok); + + term_snapshot_free(snap); + term_free(t); +} + +int main(void) +{ + test_concurrent_feed_and_snapshot(); + test_grow_past_reservation(); + test_skipped_cells_read_blank(); + test_csi_text_chunks(); + test_erase_keeps_used_rows(); + test_blank_rows(); + test_lnm_mode_alone_is_not_sticky(); + test_newline_policy_survives_reset(); + test_esc_control_strings_and_designators(); + test_string_payloads(); + test_recycled_scrollback_blanks(); + test_scroll_view_clamps_extreme_deltas(); + test_batched_recycled_lines(); + test_invalid_utf8_runs(); + test_c0_ends_utf8_sequence(); + test_full_width_line_endings(); + test_bulk_line_flood(); + test_ascii_flood_edits(); + test_margin_linefeed_bursts(); + test_alt_linefeed_burst_after_resize(); + test_alt_resize_keeps_row_order(); + test_alt_shrink_then_flood(); + test_alt_erase_scrollback(); + test_snapshot_follows_new_terminal(); + test_bulk_rows_survive_edits(); + + assert(!strcmp(UNICODE_DATA_VERSION, "18.0.0")); + assert(!strcmp(term_unicode_version(), UNICODE_DATA_VERSION)); + assert(unicode_width('A') == 1); + assert(unicode_width(0x0301) == 0); + assert(unicode_width(0x754C) == 2); + assert(unicode_width(0x1FAE0) == 2); + assert(unicode_emoji_presentation(0x1FAE0)); + assert(unicode_width(0x1FAEB) == 2); + assert(unicode_emoji_presentation(0x1FAEB)); + { + /* A Jamo Extended-B vowel draws inside its leading consonant. */ + struct term *t = term_new(&(struct term_options){ .size = { .x = 10, .y = 2 } }); + term_feed(t, "\xe1\x84\x80\xed\x9e\xb0", 6, 0); + assert(term_cursor(t).pos.x == 2); + term_free(t); + } + + test_wide_characters(); + test_wide_runs(); + test_editor_scrolling(); + test_top_region_scrollback(); + test_device_attributes(); + test_unimplemented_da_features(); + test_keyboard_negotiation(); + test_resize_reflow(); + test_resize_exposes_blank_rows(); + test_prompt_resize(); + test_cursor_line_motion(); + test_relative_cursor_motion(); + test_read_boundaries(); + test_ascii_controls(); + test_random_read_boundaries(); + test_csi_recovery(); + test_csi_byte_boundaries(); + test_parameter_limits(); + test_unterminated_control_strings_do_not_swallow_text(); + test_c0_controls_after_esc_are_executed(); + struct term *t = tt_new_opts(10, 4, &reply_cb); + write_seq(t, "\033(Babc\r\nD"); + assert(tt_cell(t, 0, 0)->text == 'a'); + assert(tt_cell(t, 0, 1)->text == 'D'); + write_seq(t, "\033[H/tmp/proj\r\n$ "); + assert(tt_cell(t, 0, 1)->text == '$'); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 1); + write_seq(t, "\r>"); + assert(tt_cell(t, 0, 1)->text == '>'); + assert(tt_cursor_x(t) == 1 && tt_cursor_y(t) == 1); + write_seq(t, "\r\n$"); + assert(tt_cell(t, 0, 2)->text == '$'); + assert(tt_cursor_x(t) == 1 && tt_cursor_y(t) == 2); + write_seq(t, "\033[H\033(0lqqk\033(B"); + assert(tt_cell(t, 0, 0)->text == 0x250c); + assert(tt_cell(t, 3, 0)->text == 0x2510); + write_seq(t, "\033[3d\rX\033[4bY"); + assert(tt_cell(t, 4, 2)->text == 'X'); + assert(tt_cell(t, 5, 2)->text == 'Y'); + write_seq(t, "\033[3;3H\033[2X"); + assert(tt_cell(t, 2, 2)->text == ' ' && tt_cell(t, 3, 2)->text == ' '); + assert(tt_cell(t, 4, 2)->text == 'X' && tt_cursor_x(t) == 2); + + write_seq(t, "\033[2J\033[H1234567890"); + assert(tt_cell(t, 9, 0)->text == '0'); + assert(tt_cursor_x(t) == 9 && tt_cursor_y(t) == 0); + write_seq(t, "A"); + assert(tt_cell(t, 0, 1)->text == 'A'); + assert(tt_cursor_x(t) == 1 && tt_cursor_y(t) == 1); + + write_seq(t, "\033[2J\033[Habcdefghij\r\nX"); + assert(tt_cell(t, 9, 0)->text == 'j'); + assert(tt_cell(t, 0, 1)->text == 'X' && tt_cursor_y(t) == 1); + + write_seq(t, "\033[2J\033[H\033[99GZ"); + assert(tt_cell(t, 9, 0)->text == 'Z' && tt_cursor_x(t) == 9); + write_seq(t, "\033[2J\033[9;99HY"); + assert(tt_cursor_y(t) == 3 && tt_cell(t, 9, 3)->text == 'Y'); + + write_seq(t, "\033[2J\033[H\033[?7l1234567890AB"); + assert(tt_cell(t, 9, 0)->text == 'B' && tt_cursor_y(t) == 0); + write_seq(t, "\033[?7h"); + + write_seq(t, "\033[2J\033[Habcd\033[2G\033[4hX\033[4l"); + assert(row_is(t, 0, "aXbcd")); + + write_seq(t, "\033[2;4r\033[?6h\033[2;3H@"); + assert(tt_cell(t, 2, 2)->text == '@'); + write_seq(t, "\033[?6l\033[r"); + + write_seq(t, "\033[2J\033[H\033[3g\033[4G\033H\r\tT"); + assert(tt_cell(t, 3, 0)->text == 'T'); + + write_seq(t, "\033[3;5H\033[5n"); + assert(!strcmp(response, "\033[0n")); + write_seq(t, "\033[6n"); + assert(!strcmp(response, "\033[3;5R")); + + write_seq(t, "\033[2J\033[?69h\033[3;6s\033[?6h\033[Habcdef"); + assert(tt_cell(t, 2, 0)->text == 'a' && tt_cell(t, 5, 0)->text == 'd'); + assert(tt_cell(t, 2, 1)->text == 'e' && tt_cell(t, 3, 1)->text == 'f'); + write_seq(t, "\033[?6;69l"); + + write_seq(t, "\033[?1;5;1004;2026h\033[!p"); + struct term_modes modes = term_modes(t); + assert(!modes.app_cursor && !modes.focus_events); + assert(!modes.synchronized && !modes.reverse_video); + write_seq(t, "kept\033c"); + assert(row_is(t, 0, "")); + + write_seq(t, "\033[2J\033[Hz\033[11b"); + assert(tt_cell(t, 9, 0)->text == 'z' && tt_cell(t, 1, 1)->text == 'z'); + assert(tt_cell(t, 2, 1)->text == ' '); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 1); + write_seq(t, "\033[2J\033[H1234567890\rA"); + assert(tt_cell(t, 0, 0)->text == 'A'); + write_seq(t, "\033[2;4H\033[2J"); + assert(tt_cursor_x(t) == 3 && tt_cursor_y(t) == 1); + assert(term_modes(t).mouse == TERM_MOUSE_OFF); + write_seq(t, "\033[?1000;1006h"); + assert(term_modes(t).mouse == TERM_MOUSE_NORMAL); + assert(term_modes(t).mouse_encoding == TERM_MOUSE_ENC_SGR); + write_seq(t, "\033[?1002h"); + assert(term_modes(t).mouse == TERM_MOUSE_BUTTON); + assert(term_modes(t).mouse_encoding == TERM_MOUSE_ENC_SGR); + write_seq(t, "\033[?1003h"); + assert(term_modes(t).mouse == TERM_MOUSE_ANY); + write_seq(t, "\033[?1006l"); + assert(term_modes(t).mouse_encoding == TERM_MOUSE_ENC_X10); + write_seq(t, "\033[?1015h"); + assert(term_modes(t).mouse_encoding == TERM_MOUSE_ENC_URXVT); + write_seq(t, "\033[?1016h"); + assert(term_modes(t).mouse_encoding == TERM_MOUSE_ENC_SGR_PIXELS); + write_seq(t, "\033[?1004h"); + term_focus(t, false); + assert(!strcmp(response, "\033[O")); + term_focus(t, true); + assert(!strcmp(response, "\033[I")); + write_seq(t, "\033[?1003l"); + assert(term_modes(t).mouse == TERM_MOUSE_OFF); + response_len = 0; + write_seq(t, "\033]10;?\007"); + assert(!strcmp(response, "\033]10;rgb:ffff/ffff/ffff\033\\")); + write_seq(t, "\033P+q696e646e\033\\"); + assert(!strcmp(response, "\033P0+r696e646e\033\\")); + write_seq(t, "\033P+q6b696e64\033\\"); + assert(!strcmp(response, "\033P1+r6b696e64=1b5b313b3242\033\\")); + term_free(t); + t = tt_new(40, 4); + write_seq(t, "hello 世界 \r\nsecond line\r\nhttps://example.com/path."); + tt_select(t, 9, 0, 0, 0); + char *text = tt_selection_text(t); + assert(text && !strcmp(text, "hello 世界")); + free(text); + tt_select(t, 5, 1, 0, 0); + text = tt_selection_text(t); + assert(text && !strcmp(text, "hello 世界\nsecond")); + free(text); + assert(tt_selected(t, 0, 0) && tt_selected(t, 5, 1)); + assert(!tt_selected(t, 6, 1)); + text = tt_link_at(t, 10, 2); + assert(text && !strcmp(text, "https://example.com/path")); + free(text); + assert(!tt_link_at(t, 10, 1)); + write_seq(t, "\033[?2004h"); + assert(term_modes(t).bracketed_paste); + text = tt_selection_text(t); + assert(text && !strcmp(text, "hello 世界\nsecond")); + free(text); + write_seq(t, "\033[?2004l\r\nlast\r\n"); + assert(!term_modes(t).bracketed_paste); + term_scroll(t, 1); + tt_select(t, 0, 0, 9, 0); + text = tt_selection_text(t); + assert(text && !strcmp(text, "hello 世界")); + free(text); + tt_resize(t, 50, 4); + text = tt_selection_text(t); + assert(text && !strcmp(text, "hello 世界")); + free(text); + term_free(t); + t = tt_new(40, 4); + write_seq(t, "\033[31;44mA\033[>4;0mB\033[>4;1m\033[31mC"); + assert(tt_cell(t, 0, 0)->fg == tt_cell(t, 1, 0)->fg); + assert(tt_cell(t, 0, 0)->bg == tt_cell(t, 1, 0)->bg); + assert(tt_cell(t, 0, 0)->fg == tt_cell(t, 2, 0)->fg); + assert(tt_cell(t, 0, 0)->bg == tt_cell(t, 2, 0)->bg); + assert(tt_cursor_x(t) == 3 && tt_cursor_y(t) == 0); + write_seq(t, "\033[0mD"); + assert(tt_cell(t, 3, 0)->fg == TERM_COLOR_FG); + assert(tt_cell(t, 3, 0)->bg == TERM_COLOR_BG); + term_free(t); + + t = tt_new(10, 4); + write_seq(t, "ABCDEFGHIJ\r\n\r\n\r\n\r\n"); + term_scroll(t, 1); + assert(tt_view_cell(t, 0, 0)->text == 'A'); + assert(tt_view_cell(t, 9, 0)->text == 'J'); + term_scroll_to(t, 0); + + for (int i = 0; i < TEST_SCROLLBACK; i++) + write_seq(t, "xy\r\n"); + term_scroll(t, 1); + assert(tt_view_cell(t, 0, 0)->text == 'x'); + for (int c = 2; c < 10; c++) { + assert(tt_view_cell(t, c, 0)->text == ' '); + assert(tt_view_cell(t, c, 0)->bg == TERM_COLOR_BG); + } + term_free(t); + + t = tt_new(10, 4); + write_seq(t, "\033[41m\033[2K"); + term_color erased_bg = tt_cell(t, 0, 0)->bg; + assert(erased_bg != TERM_COLOR_BG); + assert(tt_cell(t, 9, 0)->bg == erased_bg); + write_seq(t, "\033[m\n\n\n\n"); + term_scroll(t, 1); + assert(tt_view_cell(t, 9, 0)->bg == erased_bg); + assert(tt_view_cell(t, 10, 0)->bg == TERM_COLOR_BG); + term_free(t); + + t = tt_new(10, 4); + write_seq(t, "\033[Hnormal"); + write_seq(t, "\033[?1049h"); + for (int c = 0; c < 10; c++) { + assert(tt_cell(t, c, 0)->text == ' '); + assert(tt_cell(t, c, 0)->bg == TERM_COLOR_BG); + } + write_seq(t, "\033[Halt\n\033[2J"); + write_seq(t, "\033[?1049l"); + assert(tt_cell(t, 0, 0)->text == 'n' && tt_cell(t, 5, 0)->text == 'l'); + term_free(t); + + t = tt_new(20, 8); + write_seq(t, "one\r\ntwo\r\nthree\r\nfour\r\nfive\r\n$ "); + assert(tt_cursor_y(t) == 5 && tt_history(t) == 0); + tt_resize(t, 20, 3); + assert(tt_cursor_y(t) == 2 && tt_history(t) == 3); + assert(row_is(t, 0, "four") && row_is(t, 1, "five") && row_is(t, 2, "$")); + term_scroll(t, 3); + assert(row_is(t, 0, "one") && row_is(t, 1, "two") && row_is(t, 2, "three")); + term_scroll_to(t, 0); + tt_resize(t, 20, 8); + assert(tt_cursor_y(t) == 5 && tt_history(t) == 0); + assert(row_is(t, 0, "one") && row_is(t, 2, "three") && row_is(t, 5, "$")); + assert(row_is(t, 6, "") && row_is(t, 7, "")); + + tt_resize(t, 20, 6); + assert(tt_cursor_y(t) == 5 && tt_history(t) == 0); + assert(row_is(t, 0, "one") && row_is(t, 5, "$")); + term_free(t); + + t = tt_new(20, 6); + write_seq(t, "a\r\nb\r\nc\r\nd\r\ne\r\n"); + write_seq(t, "\033[?1049h\033[6;1Hbottom"); + int before = tt_history(t); + tt_resize(t, 20, 3); + assert(tt_history(t) == before); + tt_resize(t, 20, 6); + assert(tt_history(t) == before); + write_seq(t, "\033[?1049l"); + assert(row_is(t, 0, "a") && row_is(t, 4, "e")); + term_free(t); + + t = tt_new(40, 6); + write_seq(t, "\033[?1049h\033[1;38Hxy"); + tt_resize(t, 12, 6); + write_seq(t, "\033[?1049l"); + assert(tt_cursor_x(t) < tt_cols(t)); + write_seq(t, "Q"); + assert(tt_cell(t, tt_cursor_x(t) - 1, tt_cursor_y(t))->text == 'Q'); + term_free(t); + + t = tt_new(8, 4); + write_seq(t, "abcdefghijkl"); + tt_select(t, 0, 0, 3, 1); + text = tt_selection_text(t); + assert(text && !strcmp(text, "abcdefghijkl")); + free(text); + write_seq(t, "\033[2J\033[Habcd\r\nefgh"); + tt_select(t, 0, 0, 3, 1); + text = tt_selection_text(t); + assert(text && !strcmp(text, "abcd\nefgh")); + free(text); + term_free(t); + + t = tt_new(10, 4); + write_seq(t, "abcdef"); + write_seq(t, "\033[1;3H\033[P"); + assert(row_is(t, 0, "abdef")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + write_seq(t, "\033[2P"); + assert(row_is(t, 0, "abf")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + write_seq(t, "\033[10P"); + assert(row_is(t, 0, "ab")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + term_free(t); + + t = tt_new(8, 4); + write_seq(t, "abcdef"); + write_seq(t, "\033[1;3H\033[@"); + assert(row_is(t, 0, "ab cdef")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + write_seq(t, "\033[2@"); + assert(row_is(t, 0, "ab cde")); + term_free(t); + + t = tt_new(10, 4); + assert(term_cursor(t).style == TERM_CURSOR_BLOCK); + assert(term_cursor(t).blink == true); + + write_seq(t, "\033[2 q"); + assert(term_cursor(t).style == TERM_CURSOR_BLOCK); + assert(term_cursor(t).blink == false); + + write_seq(t, "\033[3 q"); + assert(term_cursor(t).style == TERM_CURSOR_UNDERLINE); + assert(term_cursor(t).blink == true); + + write_seq(t, "\033[4 q"); + assert(term_cursor(t).style == TERM_CURSOR_UNDERLINE); + assert(term_cursor(t).blink == false); + + write_seq(t, "\033[5 q"); + assert(term_cursor(t).style == TERM_CURSOR_BAR); + assert(term_cursor(t).blink == true); + + write_seq(t, "\033[6 q"); + assert(term_cursor(t).style == TERM_CURSOR_BAR); + assert(term_cursor(t).blink == false); + + write_seq(t, "\033[0 q"); + assert(term_cursor(t).style == TERM_CURSOR_BLOCK); + assert(term_cursor(t).blink == true); + + write_seq(t, "\033[5 q"); + write_seq(t, "\033[ q"); + assert(term_cursor(t).style == TERM_CURSOR_BLOCK); + assert(term_cursor(t).blink == true); + + write_seq(t, "\033[6 q"); + assert(term_cursor(t).style == TERM_CURSOR_BAR && !term_cursor(t).blink); + write_seq(t, "\033[?1049h"); + write_seq(t, "\033[3 q"); + assert(term_cursor(t).style == TERM_CURSOR_UNDERLINE && term_cursor(t).blink); + write_seq(t, "\033[?1049l"); + assert(term_cursor(t).style == TERM_CURSOR_BAR && !term_cursor(t).blink); + + term_set_cursor_default(t, TERM_CURSOR_UNDERLINE, true); + + write_seq(t, "\033[5 q"); + assert(term_cursor(t).style == TERM_CURSOR_BAR && term_cursor(t).blink); + write_seq(t, "\033[0 q"); + assert(term_cursor(t).style == TERM_CURSOR_UNDERLINE && !term_cursor(t).blink); + + write_seq(t, "\033[2 q"); + assert(term_cursor(t).style == TERM_CURSOR_BLOCK && !term_cursor(t).blink); + write_seq(t, "\033[ q"); + assert(term_cursor(t).style == TERM_CURSOR_UNDERLINE && !term_cursor(t).blink); + + write_seq(t, "\033[1 q"); + assert(term_cursor(t).style == TERM_CURSOR_BLOCK && term_cursor(t).blink); + write_seq(t, "\033[!p"); + assert(term_cursor(t).style == TERM_CURSOR_UNDERLINE && !term_cursor(t).blink); + term_free(t); + + t = tt_new(20, 4); + write_seq(t, "\033[1;2;3;4;9;53mA\033[21mB\033[4:2mC\033[22;24mD\033[0mE"); + const uint16_t styles = TERM_ATTR_BOLD | TERM_ATTR_FAINT | TERM_ATTR_ITALIC | + TERM_ATTR_STRIKE | TERM_ATTR_OVERLINE; + assert((tt_cell(t, 0, 0)->attrs & styles) == styles); + assert(term_attr_underline(tt_cell(t, 0, 0)->attrs) == TERM_UNDERLINE_SINGLE); + assert(term_attr_underline(tt_cell(t, 1, 0)->attrs) == TERM_UNDERLINE_DOUBLE); + assert(term_attr_underline(tt_cell(t, 2, 0)->attrs) == TERM_UNDERLINE_DOUBLE); + assert(!(tt_cell(t, 3, 0)->attrs & (TERM_ATTR_BOLD | TERM_ATTR_FAINT))); + assert(term_attr_underline(tt_cell(t, 3, 0)->attrs) == TERM_UNDERLINE_NONE); + assert(tt_cell(t, 4, 0)->attrs == 0); + term_free(t); + + t = tt_new(20, 4); + write_seq(t, "\033[30mA\033[0;1;30mB\033[0;90mC"); + assert(tt_cell(t, 0, 0)->fg == tt_cell(t, 1, 0)->fg); + assert(tt_cell(t, 1, 0)->fg != tt_cell(t, 2, 0)->fg); + assert(!(tt_cell(t, 0, 0)->attrs & TERM_ATTR_BOLD)); + assert(tt_cell(t, 1, 0)->attrs & TERM_ATTR_BOLD); + term_free(t); + + const char *colon_rgb = "\033[48;2;145;2;34m\033[38;2;12;34;56mA" + "\033[38:2::12:34:56mB\033[;mC"; + for (size_t split = 0; split <= strlen(colon_rgb); split++) { + t = tt_new(20, 4); + tt_write(t, colon_rgb, split); + tt_write(t, colon_rgb + split, strlen(colon_rgb) - split); + assert(row_is(t, 0, "ABC")); + for (int x = 0; x < 2; x++) { + assert(tt_cell(t, x, 0)->fg == 0xFF0C2238); + assert(tt_cell(t, x, 0)->bg == 0xFF910222); + } + assert(tt_cell(t, 2, 0)->fg == TERM_COLOR_FG); + assert(tt_cell(t, 2, 0)->bg == TERM_COLOR_BG); + term_free(t); + } + + t = tt_new(20, 4); + write_seq(t, "\033[38;2;246;226;183;49mX"); + assert(tt_cell(t, 0, 0)->fg == 0xFFF6E2B7); + assert(tt_cell(t, 0, 0)->bg == TERM_COLOR_BG); + term_free(t); + + t = tt_new(20, 4); + const char *clusters = "(\xc2\xa9\xef\xb8\x8f)\xe2\x98\xba\xef\xb8\x8e" + "\xf0\x9f\x91\xa8\xe2\x80\x8d\xf0\x9f\x8c\xbe"; + write_seq(t, clusters); + + assert(tt_cursor_x(t) == 7); + assert((tt_cell(t, 1, 0)->attrs & TERM_ATTR_EMOJI) && tt_cell(t, 1, 0)->width == 2); + assert(tt_cell(t, 2, 0)->width == 0); + assert((tt_cell(t, 4, 0)->attrs & TERM_ATTR_TEXT) && tt_cell(t, 4, 0)->width == 1); + assert((tt_cell(t, 5, 0)->attrs & TERM_ATTR_EMOJI) && tt_cell(t, 5, 0)->width == 2); + assert(tt_cell(t, 6, 0)->width == 0); + + tt_select(t, 0, 0, 6, 0); + text = tt_selection_text(t); + assert(text && !strcmp(text, clusters)); + free(text); + term_free(t); + + /* BS with a pending wrap clears the wrap and moves left from the last + * column, as in xterm and the VT100. */ + t = tt_new(5, 2); + write_seq(t, "12345"); + assert(row_is(t, 0, "12345")); + assert(tt_cursor_x(t) == 4 && tt_cursor_y(t) == 0); + write_seq(t, "\b"); + assert(tt_cursor_x(t) == 3 && tt_cursor_y(t) == 0); + write_seq(t, "X"); + assert(row_is(t, 0, "123X5")); + assert(tt_cursor_x(t) == 4 && tt_cursor_y(t) == 0); + write_seq(t, "\b \b"); + assert(row_is(t, 0, "123 5")); + assert(tt_cursor_x(t) == 3 && tt_cursor_y(t) == 0); + term_free(t); + + /* A renderer sees each cursor move within the row. */ + t = tt_new(10, 2); + struct term_snapshot *snap = term_snapshot_new(); + write_seq(t, "abc"); + uint64_t version = view_version(snap, t); + write_seq(t, "\b"); + assert(tt_cursor_x(t) == 2 && view_version(snap, t) != version); + version = view_version(snap, t); + write_seq(t, "\r"); + assert(tt_cursor_x(t) == 0 && view_version(snap, t) != version); + version = view_version(snap, t); + write_seq(t, "\t"); + assert(tt_cursor_x(t) == 8 && view_version(snap, t) != version); + term_snapshot_free(snap); + term_free(t); + + t = tt_new(20, 2); + tt_write(t, "hello world", 11); + assert(row_is(t, 0, "hello world")); + assert(tt_cursor_x(t) == 11 && tt_cursor_y(t) == 0); + term_free(t); + + title[0] = 0; + t = tt_new_opts(10, 4, &title_cb); + write_seq(t, "\033]0;My Shell\007"); + assert(!strcmp(title, "My Shell")); + write_seq(t, "\033]2;Second Title\033\\"); + assert(!strcmp(title, "Second Title")); + write_seq(t, "\033]1;Icon Name\007"); + assert(!strcmp(title, "Icon Name")); + term_free(t); + + t = tt_new(20, 4); + write_seq(t, "$ /mnt/work"); + const char *round_trip = "\033[?1049h\033[?1049l"; + tt_write(t, round_trip, strlen(round_trip)); + assert(row_is(t, 0, "$ /mnt/work")); + term_free(t); + + /* A renderer sees both switches between the screens. */ + t = tt_new(20, 4); + snap = term_snapshot_new(); + write_seq(t, "$ /mnt/work"); + version = view_version(snap, t); + write_seq(t, "\033[?1049h"); + assert(view_version(snap, t) != version); + version = view_version(snap, t); + write_seq(t, "\033[?1049l"); + assert(view_version(snap, t) != version); + assert(row_is(t, 0, "$ /mnt/work")); + term_snapshot_free(snap); + term_free(t); + + t = tt_new(20, 4); + write_seq(t, "output"); + write_seq(t, "\r\033[K../"); + assert(row_is(t, 0, "../")); + term_free(t); + + t = tt_new(80, 24); + assert(!term_modes(t).alternate); + write_seq(t, "\033[?1049h"); + assert(term_modes(t).alternate); + write_seq(t, "\033[?1049l"); + assert(!term_modes(t).alternate); + term_free(t); + + const char *color_runs[] = { + "\033[1;1H\033[48;2;10;20;30m\033[38;2;200;210;220ma" + "\033[48;2;100;110;120m\033[38;2;230;240;250mb", + "\033[1;1H\033[38;2;200;210;220m\033[48;2;10;20;30ma" + "\033[38;2;230;240;250m\033[48;2;100;110;120mb", + }; + for (size_t run = 0; run < sizeof(color_runs) / sizeof(color_runs[0]); run++) { + const char *seq = color_runs[run]; + size_t len = strlen(seq); + for (size_t split = 0; split <= len; split++) { + t = tt_new(100, 2); + tt_write(t, seq, split); + tt_write(t, seq + split, len - split); + assert(row_is(t, 0, "ab")); + assert(tt_cursor_x(t) == 2 && tt_cursor_y(t) == 0); + assert(tt_cell(t, 0, 0)->fg == 0xffc8d2dc); + assert(tt_cell(t, 0, 0)->bg == 0xff0a141e); + assert(tt_cell(t, 1, 0)->fg == 0xffe6f0fa); + assert(tt_cell(t, 1, 0)->bg == 0xff646e78); + for (int col = 2; col < 100; col++) + assert(tt_cell(t, col, 0)->bg == TERM_COLOR_BG); + assert(row_is(t, 1, "")); + term_free(t); + } + } + + puts("PASS: terminal regression tests"); + return 0; +} diff --git a/tests/unicode/unicode_test.c b/tests/unicode/unicode_test.c new file mode 100644 index 0000000..7fc6e3e --- /dev/null +++ b/tests/unicode/unicode_test.c @@ -0,0 +1,138 @@ +#include "term/unicode/utf8.h" +#include <assert.h> +#include <stdio.h> +#include <string.h> + +static void test_utf8_decode_one(void) +{ + static const struct { const char *bytes; int len; uint32_t cp; } valid[] = { + {"\0", 1, 0}, {"\x7f", 1, 0x7f}, + {"\xc2\x80", 2, 0x80}, {"\xdf\xbf", 2, 0x7ff}, + {"\xe0\xa0\x80", 3, 0x800}, {"\xed\x9f\xbf", 3, 0xd7ff}, + {"\xee\x80\x80", 3, 0xe000}, {"\xef\xbf\xbf", 3, 0xffff}, + {"\xf0\x90\x80\x80", 4, 0x10000}, + {"\xf4\x8f\xbf\xbf", 4, 0x10ffff}, + }; + for (size_t i = 0; i < sizeof(valid) / sizeof(*valid); i++) { + uint32_t cp = 0xdeadbeef; + for (int n = 0; n < valid[i].len; n++) { + int r = utf8_decode_one(valid[i].bytes, n, &cp); + assert(r == 0); + assert(cp == 0xdeadbeef); + } + int r = utf8_decode_one(valid[i].bytes, valid[i].len, &cp); + assert(r == valid[i].len); + assert(cp == valid[i].cp); + char followed[8]; + memcpy(followed, valid[i].bytes, valid[i].len); + memcpy(followed + valid[i].len, "next", 4); + r = utf8_decode_one(followed, valid[i].len + 4, &cp); + assert(r == valid[i].len); + assert(cp == valid[i].cp); + } + static const char *const invalid[] = { + "\x80", "\xbf", "\xc0\x80", "\xc1\xbf", "\xc2X", + "\xe0\x9f\xbf", "\xed\xa0\x80", "\xed\xbf\xbf", "\xe1\x80X", + "\xf0\x8f\xbf\xbf", "\xf4\x90\x80\x80", "\xf5\x80\x80\x80", + "\xf8\x88\x80\x80\x80", "\xff", "\xf1\x80\x80X", + }; + for (size_t i = 0; i < sizeof(invalid) / sizeof(*invalid); i++) { + uint32_t cp = 0xdeadbeef; + int r = utf8_decode_one(invalid[i], strlen(invalid[i]), &cp); + assert(r == 0); + assert(cp == 0xdeadbeef); + } +} + +static void test_scalar_roundtrips(void) +{ + for (uint32_t cp = 0; cp <= 0x10ffff; cp++) { + if (cp >= 0xd800 && cp <= 0xdfff) continue; + char bytes[UTF8_MAX_BYTES]; + int len = utf8_encode(cp, bytes); + uint32_t scalar = 0, streamed = 0, state = UTF8_ACCEPT; + int r = utf8_decode_one(bytes, (size_t)len, &scalar); + assert(r == len); + assert(scalar == cp); + for (int i = 0; i < len; i++) { + utf8_decode(&state, &streamed, (unsigned char)bytes[i]); + if (i + 1 < len) assert(state != UTF8_ACCEPT && state != UTF8_REJECT); + } + assert(state == UTF8_ACCEPT && streamed == cp); + } + static const uint32_t invalid[] = {0xd800, 0xdfff, 0x110000, UINT32_MAX}; + for (size_t i = 0; i < sizeof(invalid) / sizeof(*invalid); i++) { + char bytes[UTF8_MAX_BYTES]; + int len = utf8_encode(invalid[i], bytes); + uint32_t cp = 0; + int r = utf8_decode_one(bytes, len, &cp); + assert(len == 3 && r == len && cp == 0xfffd); + } +} + +static void test_utf8_sanitize(void) +{ + char dst[64]; + + utf8_sanitize(dst, sizeof(dst), "hello \xe2\x82\xac\xf0\x9f\x98\x80"); + assert(!strcmp(dst, "hello \xe2\x82\xac\xf0\x9f\x98\x80")); + + utf8_sanitize(dst, sizeof(dst), "a\xff\x80" "b\xed\xa0\x80" "c\xe2\x82"); + assert(!strcmp(dst, "abc")); + + utf8_sanitize(dst, sizeof(dst), ""); + assert(!strcmp(dst, "")); + utf8_sanitize(dst, sizeof(dst), "\xff\xfe"); + assert(!strcmp(dst, "")); + + utf8_sanitize(dst, 6, "ab\xe2\x82\xac" "cd"); + assert(!strcmp(dst, "ab\xe2\x82\xac")); + utf8_sanitize(dst, 5, "ab\xe2\x82\xac" "cd"); + assert(!strcmp(dst, "abcd")); + memset(dst, 'x', sizeof(dst)); + utf8_sanitize(dst, 4, "abcdef"); + assert(!strcmp(dst, "abc")); + memset(dst, 'x', sizeof(dst)); + utf8_sanitize(dst, 0, "abc"); + assert(dst[0] == 'x'); +} + +/* Each maximal invalid subpart gives U+FFFD. The byte that ends a subpart + * stays unused. */ +static void test_ground_invalid_subparts(void) +{ + uint32_t state = UTF8_ACCEPT, codep = 0, cp = 0xdead; + int n = utf8_decode_ground(&state, &codep, "\xe2\x80\x1b\xac", 4, &cp); + assert(n == 2); + assert(cp == 0xFFFD); + assert(state == UTF8_ACCEPT); + n = utf8_decode_ground(&state, &codep, "\xac", 1, &cp); + assert(n == 1); + assert(cp == 0xFFFD); + assert(state == UTF8_ACCEPT); + n = utf8_decode_ground(&state, &codep, "\xe2\x82z", 3, &cp); + assert(n == 2); + assert(cp == 0xFFFD); + assert(state == UTF8_ACCEPT); + n = utf8_decode_ground(&state, &codep, "\xe2\x82", 2, &cp); + assert(n == 2); + assert(cp == 0); + assert(state != UTF8_ACCEPT); + + state = UTF8_ACCEPT; + codep = 0; + n = utf8_decode_ground(&state, &codep, "\xe2\x82\xac", 3, &cp); + assert(n == 3); + assert(cp == 0x20ac); + assert(state == UTF8_ACCEPT); +} + +int main(void) +{ + test_utf8_decode_one(); + test_scalar_roundtrips(); + test_utf8_sanitize(); + test_ground_invalid_subparts(); + puts("PASS: Unicode codec tests"); + return 0; +} diff --git a/tests/view_test.c b/tests/view_test.c new file mode 100644 index 0000000..5e17450 --- /dev/null +++ b/tests/view_test.c @@ -0,0 +1,475 @@ +#include <assert.h> +#include "term_api.h" + +#include <limits.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +/* + * The scrollback view: the viewport can leave the live screen, and the + * terminal then shows history rows. These tests cover what the view does + * while the terminal keeps working underneath it. + */ + +/* The tests use small terminals, so a row fits in a short buffer. */ +#define VIEW_TEXT_CAP 256 + +/* The text of a view row, trailing blanks cut. */ +static void view_row_text(struct term *t, int row, char *buf, size_t cap) +{ + const struct term_cell *cells = term_snapshot_row(tt_snapshot_for(t), row); + size_t n = 0; + for (int c = 0; c < tt_cols(t) && n + 1 < cap; c++) { + uint32_t cp = tt_cell_cp(t, &cells[c]); + buf[n++] = !cp ? ' ' : cp < 0x80 ? (char)cp : '?'; + } + while (n > 0 && buf[n - 1] == ' ') + n--; + buf[n] = 0; +} + +static bool view_row_is(struct term *t, int row, const char *want) +{ + char got[VIEW_TEXT_CAP]; + view_row_text(t, row, got, sizeof(got)); + return !strcmp(got, want); +} + +/* Every view row holds the text of the same row in want. A view row that + * resolves to no row at all fails the text compare. */ +static void assert_view_rows_equal(struct term *t, struct term *want) +{ + for (int row = 0; row < tt_rows(t); row++) { + char got[VIEW_TEXT_CAP], expect[VIEW_TEXT_CAP]; + view_row_text(t, row, got, sizeof(got)); + view_row_text(want, row, expect, sizeof(expect)); + assert(!strcmp(got, expect)); + } +} + +/* The view offset, in range, and the same as the snapshot reports. The two + * must come from one field: a second copy of the offset is what let the + * scrollbar and the rows disagree. */ +static void assert_offset_agrees(struct term *t) +{ + const struct term_view *v = term_snapshot_view(tt_snapshot_for(t)); + struct term_scrollback sb = term_scrollback(t); + assert(sb.offset == v->scroll_offset); + assert(sb.offset >= 0); + assert(sb.offset <= sb.total); +} + +/* A terminal whose history is small enough to fill. */ +static struct term *small_history_term(int cols, int rows, int scrollback) +{ + struct term_options opt = { + .size = { .x = cols, .y = rows }, + .scrollback = scrollback, + }; + struct term *t = term_new(&opt); + assert(t); + return t; +} + +/* Fill the history with count lines of "Lnn" and leave the cursor on the + * last screen row. */ +static void feed_lines(struct term *t, const char *tag, int count) +{ + char line[64]; + for (int i = 0; i < count; i++) { + int len = snprintf(line, sizeof(line), "%s%02d\r\n", tag, i); + tt_write(t, line, (size_t)len); + } +} + +/* Output that arrives while the view sits in the history leaves the view + * on the same text. */ +static void view_holds_text_while_output_arrives(void) +{ + struct term *t = tt_new(12, 5); + struct term *reference = tt_new(12, 5); + feed_lines(t, "L", 15); + feed_lines(reference, "L", 15); + term_scroll(t, 2); + term_scroll(reference, 2); + assert(tt_view_offset(t) == 2); + assert_view_rows_equal(t, reference); + + feed_lines(t, "M", 4); + feed_lines(reference, "M", 4); + assert_offset_agrees(t); + /* Nothing was evicted, so the view still shows the rows it showed. */ + assert_view_rows_equal(t, reference); + term_free(t); + term_free(reference); +} + +/* A view that sits at the top of the history keeps the oldest text there + * when more output arrives. The history grows by a row, so the offset + * grows with it. */ +static void view_at_the_top_keeps_the_oldest_text(void) +{ + const int scrollback = 20; + struct term *t = small_history_term(12, 5, scrollback); + struct term *reference = small_history_term(12, 5, scrollback); + feed_lines(t, "L", 12); + feed_lines(reference, "L", 12); + term_scroll_to(t, INT_MAX); + term_scroll_to(reference, INT_MAX); + int total = tt_history(t); + assert(total > 0 && total + tt_rows(t) <= scrollback); + assert(tt_view_offset(t) == total); + assert(view_row_is(t, 0, "L00")); + + /* Three more lines with room in the history: the oldest line stays on + * the top row of the view, and the offset grows with the history. */ + feed_lines(t, "M", 3); + feed_lines(reference, "M", 3); + assert_offset_agrees(t); + assert(tt_history(t) == total + 3); + assert(tt_view_offset(t) == total + 3); + assert(view_row_is(t, 0, "L00")); + assert_view_rows_equal(t, reference); + term_free(t); + term_free(reference); +} + +/* Output larger than the history evicts the rows the view was anchored to. + * The view then shows the oldest text the terminal still keeps, and the + * offset stays inside the history. */ +static void view_survives_output_larger_than_history(void) +{ + const int scrollback = 20; + struct term *t = small_history_term(12, 5, scrollback); + struct term *reference = small_history_term(12, 5, scrollback); + feed_lines(t, "L", 15); + feed_lines(reference, "L", 15); + term_scroll(t, 3); + + feed_lines(t, "M", 40); + feed_lines(reference, "M", 40); + assert_offset_agrees(t); + assert(tt_history(t) == tt_history(reference)); + assert(tt_history(t) == scrollback); + assert(tt_view_offset(t) > 0); + + /* The top of the history is where the view ends up. */ + term_scroll(reference, INT_MAX); + assert(tt_view_offset(reference) == tt_history(reference)); + assert(tt_view_offset(t) == tt_history(t)); + assert_view_rows_equal(t, reference); + term_free(t); + term_free(reference); +} + +/* An eviction can drop a line that wraps to more rows than the push adds. + * The offset then stops at the oldest row the terminal still keeps, and + * never sits above the history. */ +static void view_survives_eviction_of_a_wrapped_line(void) +{ + const int scrollback = 20; + struct term *t = small_history_term(12, 5, scrollback); + struct term *reference = small_history_term(12, 5, scrollback); + /* The oldest line wraps to three rows. */ + const char *wrapped = "0123456789012345678901234\r\n"; + tt_puts(t, wrapped); + tt_puts(reference, wrapped); + feed_lines(t, "s", 22); + feed_lines(reference, "s", 22); + term_scroll_to(t, INT_MAX); + term_scroll_to(reference, INT_MAX); + assert_offset_agrees(t); + + /* The next line evicts the wrapped one: three rows go, one arrives. */ + for (int i = 0; i < 4; i++) { + tt_puts(t, "n\r\n"); + tt_puts(reference, "n\r\n"); + assert_offset_agrees(t); + assert(tt_view_offset(t) <= tt_history(t)); + } + /* The view sits on the oldest text the terminal still keeps. */ + term_scroll_to(reference, INT_MAX); + assert(tt_view_offset(t) == tt_view_offset(reference)); + assert_view_rows_equal(t, reference); + term_free(t); + term_free(reference); +} + +/* A resize keeps the same retained text at the top of the view. */ +static void view_keeps_anchor_on_resize(void) +{ + static const struct { + int cols, rows; + } sizes[] = { { 12, 6 }, { 20, 5 }, { 6, 5 } }; + for (size_t i = 0; i < sizeof(sizes) / sizeof(sizes[0]); i++) { + struct term *t = tt_new(12, 5); + struct term *reference = tt_new(12, 5); + feed_lines(t, "L", 15); + feed_lines(reference, "L", 15); + term_scroll(t, 3); + assert(tt_view_offset(t) == 3); + assert(tt_resize(t, sizes[i].cols, sizes[i].rows)); + assert(tt_resize(reference, sizes[i].cols, sizes[i].rows)); + assert_offset_agrees(t); + assert(tt_view_offset(t) > 0); + assert(view_row_is(t, 0, "L08")); + term_scroll_to(t, 0); + assert_view_rows_equal(t, reference); + term_free(t); + term_free(reference); + } +} + +/* The alternate screen has no history to show. The normal screen comes + * back at its own bottom. */ +static void view_follows_the_alternate_screen(void) +{ + struct term *t = tt_new(12, 5); + feed_lines(t, "L", 15); + term_scroll(t, 3); + assert(tt_view_offset(t) == 3); + + tt_puts(t, "\033[?1049hALT"); + assert_offset_agrees(t); + assert(tt_view_offset(t) == 0); + assert(view_row_is(t, 0, "ALT")); + + tt_puts(t, "\033[?1049l"); + assert_offset_agrees(t); + assert(tt_view_offset(t) == 0); + assert(view_row_is(t, 0, "L11")); + term_free(t); +} + +/* The view offset has one owner. Every operation that moves or clears the + * view leaves the public offset and the snapshot offset equal. */ +static void view_offset_has_one_owner(void) +{ + struct term *t = tt_new(12, 5); + feed_lines(t, "L", 15); + assert_offset_agrees(t); + term_scroll(t, 3); + assert_offset_agrees(t); + term_scroll_to(t, INT_MAX); + assert_offset_agrees(t); + assert(tt_view_offset(t) == tt_history(t)); + term_scroll(t, INT_MIN); + assert_offset_agrees(t); + assert(tt_view_offset(t) == 0); + term_free(t); +} + +/* ED2 and ED3 return the view to the live screen. RIS keeps the offset, as + * it did before the view had an anchor of its own. */ +static void view_returns_to_bottom_on_erase(void) +{ + struct term *t = tt_new(12, 5); + feed_lines(t, "L", 15); + term_scroll(t, 3); + assert(tt_view_offset(t) == 3); + + tt_puts(t, "\033[2J"); + assert_offset_agrees(t); + assert(tt_view_offset(t) == 0); + + term_scroll(t, 2); + assert(tt_view_offset(t) == 2); + tt_puts(t, "\033[3J"); + assert_offset_agrees(t); + assert(tt_view_offset(t) == 0); + assert(tt_history(t) == 0); + term_free(t); + + t = tt_new(12, 5); + feed_lines(t, "L", 15); + term_scroll(t, 3); + int before = tt_view_offset(t); + tt_puts(t, "\033c"); + assert_offset_agrees(t); + assert(tt_view_offset(t) == before); + term_free(t); +} + +/* Rewrap keeps the cell that selected the anchored segment in view. */ +static void wrapped_anchor_survives_resize(void) +{ + struct term *t = tt_new(10, 3); + tt_puts(t, "abcdefghijKLMNOPQRSTuvwxyz\r\n"); + feed_lines(t, "L", 10); + term_scroll_to(t, INT_MAX); + term_scroll(t, -1); + assert(view_row_is(t, 0, "KLMNOPQRST")); + tt_select(t, 0, 0, 2, 0); + char *selected = tt_selection_text(t); + assert(selected && !strcmp(selected, "KLM")); + free(selected); + assert(tt_resize(t, 6, 3)); + assert_offset_agrees(t); + assert(view_row_is(t, 0, "ghijKL")); + selected = tt_selection_text(t); + assert(selected && !strcmp(selected, "KLM")); + free(selected); + feed_lines(t, "N", 5); + assert(view_row_is(t, 0, "ghijKL")); + assert(tt_resize(t, 13, 3)); + assert(view_row_is(t, 0, "abcdefghijKLM")); + assert(tt_resize(t, 6, 3)); + assert(view_row_is(t, 0, "ghijKL")); + assert_offset_agrees(t); + term_free(t); +} + +/* A wide glyph leaves a wrap pad. That pad does not move the anchor. */ +static void wide_anchor_survives_resize(void) +{ + struct term *t = tt_new(5, 3); + tt_puts(t, "abcd日本語XYZ\r\n"); + feed_lines(t, "L", 10); + term_scroll_to(t, INT_MAX); + term_scroll(t, -1); + assert(tt_cell_cp(t, tt_view_cell(t, 0, 0)) == 0x65e5); + assert(tt_resize(t, 3, 3)); + assert_offset_agrees(t); + assert(tt_cell_cp(t, tt_view_cell(t, 0, 0)) == 'd'); + assert(tt_cell_cp(t, tt_view_cell(t, 1, 0)) == 0x65e5); + assert(tt_resize(t, 8, 3)); + assert(tt_cell_cp(t, tt_view_cell(t, 0, 0)) == 'a'); + assert(tt_resize(t, 3, 3)); + assert(tt_cell_cp(t, tt_view_cell(t, 1, 0)) == 0x65e5); + term_free(t); +} + +/* Incremental and new snapshots must show the same retained rows. */ +static void snapshots_agree_after_eviction_and_rewrap(void) +{ + struct term *t = small_history_term(11, 5, 23); + struct term_snapshot *persistent = term_snapshot_new(); + assert(persistent); + for (int n = 0; n < 120; n++) { + char line[120]; + snprintf(line, sizeof(line), "%03d abcdefghijklmnop 日本語 café\r\n", n); + tt_puts(t, line); + if (n % 13 == 0) + assert(tt_resize(t, 6 + n % 17, 5)); + int total = tt_history(t); + term_scroll_to(t, n % (total + 1)); + assert(term_snapshot_update(persistent, t)); + struct term_snapshot *fresh = term_snapshot_new(); + assert(fresh && term_snapshot_update(fresh, t)); + struct term_vector2i size = term_size(t); + for (int y = 0; y < size.y; y++) { + const struct term_cell *a = term_snapshot_row(persistent, y); + const struct term_cell *b = term_snapshot_row(fresh, y); + for (int x = 0; x < size.x; x++) { + assert(a[x].fg == b[x].fg && a[x].bg == b[x].bg); + assert(a[x].width == b[x].width && a[x].attrs == b[x].attrs); + uint32_t acp[32], bcp[32]; + int an = term_snapshot_cluster(persistent, a[x].text, acp, 32); + int bn = term_snapshot_cluster(fresh, b[x].text, bcp, 32); + assert(an == bn); + if (an) + assert(!memcmp(acp, bcp, (size_t)an * sizeof(*acp))); + else + assert(a[x].text == b[x].text); + } + assert(term_snapshot_row_key(persistent, y) == term_snapshot_row_key(fresh, y)); + } + assert_offset_agrees(t); + term_snapshot_free(fresh); + } + term_snapshot_free(persistent); + term_free(t); +} + +static void snapshots_agree_after_prefix_eviction(void) +{ + struct term *t = tt_new(1000, 3); + char line[1000]; + for (int i = 0; i < 9500; i++) { + memset(line, 'A' + i % 26, sizeof(line)); + tt_write(t, line, sizeof(line)); + } + tt_write(t, "\r\n", 2); + term_scroll(t, INT_MAX); + struct term_snapshot *persistent = term_snapshot_new(); + assert(persistent && term_snapshot_update(persistent, t)); + uint64_t old_key = term_snapshot_row_key(persistent, 0); + memset(line, 'z', sizeof(line)); + for (int i = 0; i < 1200; i++) { + tt_write(t, line, sizeof(line)); + tt_write(t, "\r\n", 2); + } + assert(term_snapshot_update(persistent, t)); + assert(term_snapshot_row_key(persistent, 0) != old_key); + struct term_snapshot *fresh = term_snapshot_new(); + assert(fresh && term_snapshot_update(fresh, t)); + for (int y = 0; y < 3; y++) { + const struct term_cell *a = term_snapshot_row(persistent, y); + const struct term_cell *b = term_snapshot_row(fresh, y); + for (int x = 0; x < 1000; x++) + assert(a[x].text == b[x].text && a[x].fg == b[x].fg && a[x].bg == b[x].bg); + assert(term_snapshot_row_key(persistent, y) == term_snapshot_row_key(fresh, y)); + } + term_snapshot_free(fresh); + term_snapshot_free(persistent); + term_free(t); +} + +/* A line that needs more than one storage part stays scannable. Every + * scroll offset names one history row, and every view row shows the text + * of the history row the offset puts there. */ +static void view_scrolls_a_line_that_needs_several_parts(void) +{ + const int cols = 40, rows = 6, cells = 2000000; + struct term_options opt = { .size = { .x = cols, .y = rows }, .scrollback = 60000 }; + struct term *t = term_new(&opt); + assert(t); + char *big = malloc((size_t)cells); + assert(big); + for (int i = 0; i < cells; i++) + big[i] = (char)('0' + i % 10); + tt_write(t, big, (size_t)cells); + tt_write(t, "\r\n", 2); + free(big); + + /* The rows still on the screen are not history rows. */ + int total = tt_history(t); + assert(total == cells / cols - (rows - 1)); + struct term_snapshot *snap = term_snapshot_new(); + assert(snap); + for (int off = 1; off <= total; off += 5) { + term_scroll_to(t, off); + assert(tt_view_offset(t) == off); + assert(term_snapshot_update(snap, t)); + for (int y = 0; y < rows; y++) { + int row = total - off + y; + if (row >= total) + continue; + const struct term_cell *line = term_snapshot_row(snap, y); + for (int x = 0; x < cols; x++) + assert(line[x].text == (uint32_t)('0' + (row * cols + x) % 10)); + } + } + term_snapshot_free(snap); + term_free(t); +} + +int main(void) +{ + snapshots_agree_after_prefix_eviction(); + wrapped_anchor_survives_resize(); + wide_anchor_survives_resize(); + snapshots_agree_after_eviction_and_rewrap(); + view_holds_text_while_output_arrives(); + view_at_the_top_keeps_the_oldest_text(); + view_survives_output_larger_than_history(); + view_survives_eviction_of_a_wrapped_line(); + view_keeps_anchor_on_resize(); + view_follows_the_alternate_screen(); + view_offset_has_one_owner(); + view_scrolls_a_line_that_needs_several_parts(); + view_returns_to_bottom_on_erase(); + printf("PASS: scrollback view tests\n"); + return 0; +} diff --git a/tests/write_fuzz_test.c b/tests/write_fuzz_test.c new file mode 100644 index 0000000..8c1475a --- /dev/null +++ b/tests/write_fuzz_test.c @@ -0,0 +1,119 @@ +/* + * Differential fuzz test for term_feed. Each seed makes random terminal + * output: text, controls, CSI and ESC sequences, raw bytes, and resizes. + * One terminal gets the output in a few large writes; the bulk fast paths + * handle it. A second terminal gets it one byte at a time; the generic + * parser handles it. The two must show the same cursor, screen, and + * scrollback. + */ +#include "term_api.h" +#include "fuzz_gen.h" + +#include <assert.h> +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +/* Each terminal has its own snapshot, so cells of both stay readable at + * the same time. */ +static struct term_snapshot *snap_bulk, *snap_bytes; + +/* The code points of a cell's text. A cluster has handles that differ + * between snapshots, so the test compares its code points. */ +static int cell_text(const struct term_snapshot *s, const struct term_cell *cell, + uint32_t *cps, int cap) +{ + int n = term_snapshot_cluster(s, cell->text, cps, cap); + return n < cap ? n : cap; +} + +static bool same_cell(const struct term_cell *a, const struct term_cell *b) +{ + if (a->fg != b->fg || a->bg != b->bg || a->attrs != b->attrs || a->width != b->width) + return false; + uint32_t ta[16], tb[16]; + int na = cell_text(snap_bulk, a, ta, 16); + int nb = cell_text(snap_bytes, b, tb, 16); + return na == nb && memcmp(ta, tb, (size_t)na * sizeof(ta[0])) == 0; +} + +static void assert_same(struct term *bulk, struct term *bytes, int seed, int step) +{ + int cols = tt_cols(bulk), rows = tt_rows(bulk); + int history = tt_history(bulk); + bool same = tt_cursor_x(bulk) == tt_cursor_x(bytes) && + tt_cursor_y(bulk) == tt_cursor_y(bytes) && + history == tt_history(bytes); + term_scroll(bulk, history); + term_scroll(bytes, history); + for (int page = 0; same && page <= history; page += rows) { + bool copied = term_snapshot_update(snap_bulk, bulk); + copied &= term_snapshot_update(snap_bytes, bytes); + assert(copied); + for (int y = 0; same && y < rows; y++) { + const struct term_cell *ra = term_snapshot_row(snap_bulk, y); + const struct term_cell *rb = term_snapshot_row(snap_bytes, y); + for (int x = 0; x < cols; x++) { + const struct term_cell *a = &ra[x], *b = &rb[x]; + if (!same_cell(a, b)) { + fprintf(stderr, "seed %d step %d: page %d row %d col %d: " + "bulk %x/%x/%x/%x/%u, bytes %x/%x/%x/%x/%u\n", + seed, step, page, y, x, a->text, a->fg, a->bg, a->attrs, + a->width, b->text, b->fg, b->bg, b->attrs, b->width); + same = false; + break; + } + } + } + term_scroll(bulk, -rows); + term_scroll(bytes, -rows); + } + term_scroll_to(bulk, 0); + term_scroll_to(bytes, 0); + if (!same) + fprintf(stderr, "seed %d step %d: cursor %d,%d vs %d,%d, history %d vs %d\n", + seed, step, tt_cursor_x(bulk), tt_cursor_y(bulk), + tt_cursor_x(bytes), tt_cursor_y(bytes), + history, tt_history(bytes)); + assert(same); +} + +int main(int argc, char **argv) +{ + int seeds = argc > 1 ? atoi(argv[1]) : 50000; + for (int seed = 1; seed <= seeds; seed++) { + rng = 0x9e3779b97f4a7c15ull * (uint64_t)seed; + int cols = 4 + rn(30), rows = 2 + rn(10); + struct term *bulk = tt_new(cols, rows); + struct term *bytes = tt_new(cols, rows); + /* A new terminal can get the address of a freed one, so each seed + * uses new snapshots. */ + snap_bulk = term_snapshot_new(); + snap_bytes = term_snapshot_new(); + assert(bulk && bytes && snap_bulk && snap_bytes); + for (int step = 0; step < 6; step++) { + blen = 0; + gen(20 + rn(60), cols, rows); + size_t split = blen ? (size_t)rn((int)blen + 1) : 0; + tt_write(bulk, buf, split); + tt_write(bulk, buf + split, blen - split); + for (size_t i = 0; i < blen; i++) + tt_write(bytes, buf + i, 1); + assert_same(bulk, bytes, seed, step); + if (rn(4) == 0) { + cols = 4 + rn(30); + rows = 2 + rn(10); + bool resized = tt_resize(bulk, cols, rows); + resized &= tt_resize(bytes, cols, rows); + assert(resized); + assert_same(bulk, bytes, seed, step); + } + } + term_snapshot_free(snap_bulk); + term_snapshot_free(snap_bytes); + term_free(bulk); + term_free(bytes); + } + return 0; +} |