Introduce terminal/sys.zig which provides runtime-swappable function
pointers for operations that depend on external implementations. This
allows embedders of the terminal package to swap out implementations
at startup without hard dependencies on specific libraries.
The first function exposed is decode_png, which defaults to a wuffs
implementation. The kitty graphics image loader now calls through
sys.decode_png instead of importing wuffs directly.
This allows us to enable Kitty graphics support in libghostty-vt
for all targets except wasm32-freestanding.
Wire up the APC handler to `terminal.TerminalStream` to process APC
sequences, enabling support for kitty graphics commands in libghostty,
in theory.
The "in theory" is because we still don't export a way to actually
enable Kitty graphics in libghostty because we have some other things in
the way: PNG decoding and OS filesystem access that need to be more
conditionally compiled before we can enable the feature. However, this
is a step in the right direction, and we can at least verify that the
APC handler works via a test in Ghostty GUI.
Wire up the APC handler to `terminal.TerminalStream` to process
APC sequences, enabling support for kitty graphics commands in
libghostty, in theory.
The "in theory" is because we still don't export a way to actually
enable Kitty graphics in libghostty because we have some other things in
the way: PNG decoding and OS filesystem access that need to be more
conditionally compiled before we can enable the feature. However, this
is a step in the right direction, and we can at least verify that the
APC handler works via a test in Ghostty GUI.
Add a new GhosttySelection C API type (selection.h / c/selection.zig)
that pairs two GhosttyGridRef endpoints with a rectangle flag. This maps
directly to the internal Selection type using untracked pins.
The formatter terminal options gain an optional selection pointer. When
non-null the formatter restricts output to the specified range instead
of emitting the entire screen. When null the existing behavior of
formatting the full screen is preserved.
Add ghostty_grid_ref_hyperlink_uri to extract the OSC 8 hyperlink URI
from a cell at a grid reference position. Follows the same buffer
pattern as ghostty_grid_ref_graphemes: callers pass a buffer and get
back the byte length, or GHOSTTY_OUT_OF_SPACE with the required size if
the buffer is too small. Cells without a hyperlink return success with
length 0.
Add a new GhosttySelection C API type (selection.h / c/selection.zig)
that pairs two GhosttyGridRef endpoints with a rectangle flag. This
maps directly to the internal Selection type using untracked pins.
The formatter terminal options gain an optional selection pointer.
When non-null the formatter restricts output to the specified range
instead of emitting the entire screen. When null the existing
behavior of formatting the full screen is preserved.
Add ghostty_grid_ref_hyperlink_uri to extract the OSC 8 hyperlink
URI from a cell at a grid reference position. Follows the same
buffer pattern as ghostty_grid_ref_graphemes: callers pass a buffer
and get back the byte length, or GHOSTTY_OUT_OF_SPACE with the
required size if the buffer is too small. Cells without a hyperlink
return success with length 0.
The libghostty-vt pkg-config file was missing Libs.private, so
pkg-config --libs --static returned the same flags as the shared case,
omitting the C++ standard library needed by the SIMD code.
Additionally, the static archive did not bundle the vendored SIMD
dependencies (simdutf, highway, utfcpp), leaving consumers with
unresolved symbols when linking. If we're choosing to vendor (no -fsys)
then we should produce a fat static archive that includes them. If
`-fsys` is used, then we should not bundle them and instead reference
them via Requires.private, letting pkg-config chain to their own .pc
files.
Add Libs.private with the C++ runtime (-lc++ on Darwin, -lstdc++ on
Linux) and Requires.private for any SIMD deps provided via system
integration. When SIMD deps are vendored (the default), produce a fat
static archive that bundles them using libtool on Darwin and ar on
Linux. When they come from the system (-fsys=), reference them via
Requires.private instead, letting pkg-config chain to their own .pc
files.
The libghostty-vt pkg-config file was missing Libs.private, so
pkg-config --libs --static returned the same flags as the shared
case, omitting the C++ standard library needed by the SIMD code.
Additionally, the static archive did not bundle the vendored SIMD
dependencies (simdutf, highway, utfcpp), leaving consumers with
unresolved symbols when linking. If we're choosing to vendor (no -fsys)
then we should produce a fat static archive that includes them. If `-fsys`
is used, then we should not bundle them and instead reference them via
Requires.private, letting pkg-config chain to their own .pc files.
Add Libs.private with the C++ runtime (-lc++ on Darwin, -lstdc++
on Linux) and Requires.private for any SIMD deps provided via
system integration. When SIMD deps are vendored (the default),
produce a fat static archive that bundles them using libtool on
Darwin and ar on Linux. When they come from the system (-fsys=),
reference them via Requires.private instead, letting pkg-config
chain to their own .pc files.
I don’t know why the search-related commands were added as performable
keybinds in 240d5e0fc5, but **I asked
Claude to add some tests for that**
> This won't fix cmd+g/G not working when the search bar is focused.
This reverts commit 20cfaae2e5, reversing
changes made to 3509ccf78e.
This breaks some behaviours when there are multiple splits, which
requires another click to focus to another split in the same window
This is known issues before key-related PRs, tested on
fa9265636b.
The following config is mapped incorrectly to the menu shortcut:
```
keybind=A=goto_split:left
```
<img width="223" height="106" alt="image"
src="https://github.com/user-attachments/assets/b80da251-9cff-4b29-b143-64854a5c4271"
/>
Surfaces only accept `a` as a trigger to select left split, not
`shift+a`
- Expose that ID as the environment variable GHOSTTY_SURFACE_ID to
processes running in Ghostty surfaces.
- Add a function to the core app to search for surfaces by ID.
- ID is randomly generated, it has no other meaning other than as a
unique identifier for the surface. The ID also cannot be zero as that is
used to indicate a null ID in some situations.
- Expose that ID as the environment variable GHOSTTY_SURFACE_ID to
processes running in Ghostty surfaces.
- Add a function to the core app to search for surfaces by ID.
- ID is randomly generated, it has no other meaning other than as a
unique identifier for the surface. The ID also cannot be zero as that
is used to indicate a null ID in some situations.
Fixes#12020
The C header declared ghostty_surface_free_text with both a
ghostty_surface_t and ghostty_text_s* parameter, but the Zig
implementation only accepted a *Text parameter. This caused the surface
pointer to be interpreted as the text pointer, so the actual text
allocation was never freed.
I opted to keep the surface parameter to minimize the diff here. I'm not
sure why I thought I would need access to that surface pointer but just
want to fix the leak first.
Fixes#12020
The C header declared ghostty_surface_free_text with both a
ghostty_surface_t and ghostty_text_s* parameter, but the Zig
implementation only accepted a *Text parameter. This caused the
surface pointer to be interpreted as the text pointer, so the
actual text allocation was never freed.
- Fixes: https://github.com/ghostty-org/ghostty/issues/11964
Made a private enum type `VersionConfig` to reference whether the
release is a semver or tip, makes it easier for later in the view to
`switch` between cases.
I do think there could be a better place for this enum or we can get rid
of it, open to opinions. Right now version parsing is kind of duplicated
between `AboutView` and `UpdateModalView` so we can also extract to a
common helper if wanted.
Tested by manually setting `Marketing Version` in build settings to
`1.3.1`
<img width="412" height="532" alt="Screenshot 2026-03-30 at 18 31 15"
src="https://github.com/user-attachments/assets/285bb94d-138b-4169-bb66-684eb04b6ca3"
/>
`332b2aefc`
<img width="412" height="532" alt="Screenshot 2026-03-30 at 18 32 48"
src="https://github.com/user-attachments/assets/fea30d39-bea7-4885-8221-1696e148f45e"
/>
### AI Disclosure
I used Sonnet 4.6 to understand where the version strings came from and
in what format, it read release yml files to see what's going on. Then
it proposed really bad code so I manually went in and cleaned up the
view.
## Summary
- Move `sys/types.h` include into the `#else` branch of the existing
`_MSC_VER` guard
- MSVC does not ship `sys/types.h` (POSIX header), and already gets
`ssize_t` from `BaseTsd.h`
## Test plan
- [x] `zig build -Dapp-runtime=none` -- clean build
- [x] `zig build test -Dapp-runtime=none` on Windows (2606/2660 passed,
54 skipped)
- [x] `zig build test` on Linux (2658/2684 passed, 26 skipped)
- [x] `zig build test` on macOS (2658/2668 passed, 10 skipped)
- [x] `zig build test-lib-vt` on all 3 platforms
- [x] Zig examples build on all 3 platforms
- [x] CMake examples build on Windows (c-vt-cmake pass,
c-vt-cmake-static pass)
`GHOSTTY_KEY_DELETE` should be mapped to `KeyEquivalent.deleteForward`.
This fixes the correct symbol showing in the menu. Previously, both
`GHOSTTY_KEY_DELETE` and `GHOSTTY_KEY_BACKSPACE` were showing `⌫`, but
`GHOSTTY_KEY_DELETE` only worked for `fn+delete`.
Add the following keybind and observe the symbol in the menu:
```
keybind=delete=new_tab
```
<img width="535" height="318" alt="image"
src="https://github.com/user-attachments/assets/67ed7b5d-f848-42ee-a382-fe364d86cb2c"
/>
