diff options
| author | Ogulcan Celik <ogulcancelik@gmail.com> | 2026-07-16 23:28:05 +0300 |
|---|---|---|
| committer | Ogulcan Celik <ogulcancelik@gmail.com> | 2026-07-16 23:29:28 +0300 |
| commit | 4a3302d19a2baa3a33e910e0088c24e780905b90 (patch) | |
| tree | dcaf76fae1c82102d933c9225b17ca7ac5d3a847 /vendor/libghostty-vt.patches.md | |
| parent | 550e9f8cc085f428429f5cf2da0bb2dba897e8b8 (diff) | |
chore: update vendored libghostty-vt
refs #276
Diffstat (limited to 'vendor/libghostty-vt.patches.md')
| -rw-r--r-- | vendor/libghostty-vt.patches.md | 76 |
1 files changed, 17 insertions, 59 deletions
diff --git a/vendor/libghostty-vt.patches.md b/vendor/libghostty-vt.patches.md index 94bb998b..c01fa9b6 100644 --- a/vendor/libghostty-vt.patches.md +++ b/vendor/libghostty-vt.patches.md @@ -2,81 +2,39 @@ This file tracks intentional local changes applied on top of the vendored `libghostty-vt` source. Remove a patch only when the vendored source commit -contains the upstream fix and the listed verification still passes. +contains the upstream behavior and the listed verification still passes. -## 0001 backport resizeCols cursor subtraction saturation +## 0001 default lib-vt panes to grapheme clustering status: active -patch: `vendor/patches/libghostty-vt/0001-backport-resizecols-cursor-subtraction.patch` +patch: `vendor/patches/libghostty-vt/0001-default-grapheme-cluster-mode.patch` -herdr issue: https://github.com/ogulcancelik/herdr/issues/465 +herdr issue: https://github.com/ogulcancelik/herdr/issues/243 -upstream discussion: https://github.com/ghostty-org/ghostty/discussions/12905 +upstream discussion: not opened; libghostty-vt currently exposes current mode mutation but no C API for configuring terminal default modes -upstream pr: https://github.com/ghostty-org/ghostty/pull/12907 +upstream pr: not opened -introduced upstream: `c44afa625` - -vendored base: `0f7cd84b880b203c98683e520e84b9db0c5938d8` +vendored base: `c5a21edfcbc2d5b46540ad91b7980aca31f5f1f3` local files: -- `vendor/libghostty-vt/src/terminal/PageList.zig` - `vendor/libghostty-vt/src/terminal/c/terminal.zig` -reason: shrinking rows and columns in one resize can leave the pre-resize -cursor row past the new row count. `PageList.resizeCols` then computed rows -below the cursor with checked unsigned subtraction and aborted in safety builds. - -remove when: the vendored source commit contains upstream PR #12907 and the -local ReleaseSafe resize regression tests pass without this patch. - -verification: - -```sh -zig build test-lib-vt -Demit-lib-vt -Doptimize=ReleaseSafe -Dtest-filter="resize shrinks both axes with cursor at bottom" -zig build test-lib-vt -Demit-lib-vt -Doptimize=ReleaseSafe -Dtest-filter="PageList resize less rows and cols cursor at bottom" -``` - -## 0002 expose kitty image transmit time in the C API - -status: active - -patch: `vendor/patches/libghostty-vt/0002-expose-kitty-image-transmit-time-ns.patch` - -herdr issue: https://github.com/ogulcancelik/herdr/issues/947 - -upstream discussion: https://github.com/ghostty-org/ghostty/discussions/13177 -(proposes extending the kitty graphics inspection C API from -https://github.com/ghostty-org/ghostty/pull/12145, which has no transmit -time/serial accessor yet) - -introduced upstream: not yet - -vendored base: `0f7cd84b880b203c98683e520e84b9db0c5938d8` - -local files: - -- `vendor/libghostty-vt/include/ghostty/vt/kitty_graphics.h` -- `vendor/libghostty-vt/src/terminal/c/kitty_graphics.zig` - -reason: herdr fingerprints kitty image data to decide when to re-encode an -image for render clients. Hashing the full payload on every render is too -expensive for multi-megabyte images, and sampling windows misses small -changes, freezing streaming sources. The image's transmit time already -refreshes on every (re)transmission, so exposing it as -`GHOSTTY_KITTY_IMAGE_DATA_TRANSMIT_TIME_NS` gives herdr an exact, O(1) change -serial to invalidate a cached full-data fingerprint. +reason: Herdr renders terminal cells directly and requires DEC private mode +2027 to store flags, ZWJ emoji, and other multi-codepoint grapheme clusters in +one cell. This patch makes clustering active for new terminals and keeps it as +the reset default so RIS (`ESC c`) does not disable it. -remove when: the vendored source commit exposes the image transmit time (or an -equivalent transmission serial) in the C API and -`ghostty::tests::kitty_image_fingerprint_refreshes_on_retransmission` passes -without this patch. +remove when: libghostty-vt exposes a C API for setting default mode 2027, or +upstream makes grapheme clustering the lib-vt default, and the reset-survival +regression passes without this patch. verification: ```sh -zig build test-lib-vt -Dtest-filter="image_get transmit_time_ns changes on retransmission" -cargo nextest run kitty_image_fingerprint +cargo nextest run --locked grapheme_cluster_mode_is_default_and_survives_full_reset +cargo nextest run --locked grapheme_cluster_mode_renders_flag_emoji_in_single_wide_cell +cargo nextest run --locked grapheme_cluster_mode_renders_zwj_family_in_single_wide_cell ``` |
