This adds a set of Wasm convenience functions to ease memory management.
These are all prefixed with `ghostty_wasm` and are documented as part of
the standard Doxygen docs.
I also added a very simple single-page HTML example that demonstrates
how to use the Wasm module for key encoding.
This also adds a bunch of safety checks to the C API to verify that
valid values are actually passed to the function. This is an easy to hit
bug.
**AI disclosure:** The example is AI-written with Amp. I read through
all the code and understand it but I can't claim there isn't a better
way, I'm far from a JS expert. It is simple and works currently though.
Happy to see improvements if anyone wants to contribute.
This makes `libghostty-vt` build for freestanding wasm targets (aka a
browser) and produce a `ghostty-vt.wasm` file. This exports the same C
API that libghostty-vt does.
This commit specifically makes the changes necessary for the build to
build properly and for us to run the build in CI. We don't yet actually
try using it...
Fixes a crash when NSLocale returns nil for languageCode or countryCode
properties. This can happen when the app launches without locale
environment variables set.
The crash occurs at `src/os/locale.zig:87-88` when trying to call
`getProperty()` on a nil object. The fix adds a null check and falls
back to `en_US.UTF-8` instead of dereferencing null.
## Testing
Tested by running with locale variables unset:
```bash
unset LC_ALL && ./zig-out/Ghostty.app/Contents/MacOS/ghostty
```
Before: segmentation fault
After: launches successfully with fallback locale
Fixes#8900
Our xterm modify other keys state 2 encoding was stripped consumed mods
from the keyboard event. This doesn't match xterm or other popular
terminal emulators (but most importantly: xterm). Use the full set of
mods and add a test to verify this.
Reproduction:
```
printf '\033[>4;2m'
cat
```
Then press `ctrl+shift+h` and compare across terminals.
Closes#8430
A few questions:
* Should I set a default keybind for `toggle-mouse-reporting`? The issue
mentioned one, it's currently unset.
* Am I handling the `toggle-mouse-reporting` action properly in
`performAction` (gtk) / `action` (macos)?
Copilot was used to understand the codebase, but code was authored
manually.
This cleans up some of our termio exec code by unifying process launch
state into a single union type. This makes it easier to distinguish
between the current two mutually exclusive modes of launching a process:
fork/exec and flatpak dbus commands.
It also ensures everyplace we touch related to process launching is
forced to address every case (exhaustive switch handling). I did find
one resource cleanup bug based on this cleanup, which is also fixed
here. This just improves memory slightly so it's not a big deal.
If we add future ways to launch processes, we can add a new union case.
For example, I originally had a `posix_spawn` option while I was
experimenting with that before abandoning it (see #9274).
The host validation code previously expected IPv6 addresses to be
enclosed in [brackets], but that's not how ssh(1) expects them.
This change removes that requirement and reimplements the host
validation routine to check for valid hostnames and IP addresses (IPv4
and IPv6) using standard routines rather than custom logic.
std.net.isValidHostname is currently too generous. It considers strings
like ".example.com", "exa..mple.com", and "-example.com" to be valid
hostnames, which is incorrect according to RFC 1123 (the currently
accepted standard).
Until the standard library function is improved, we can use this local
implementation that does follow the RFC 1123 standard.
I asked Claude to perform an audit of the code based on its understand
of the RFC. It suggested some additional test cases and considers the
overall implementation to be robust (its words) and standards compliant.
Ref: https://www.rfc-editor.org/rfc/rfc1123
`fish_add_path` by default updates the `fish_user_paths` universal
variable which makes the modification persist across shell sessions.
The integration also tries to update the `fish_user_paths` when the
desired path already appears in the `PATH` environment variable but
not in `fish_user_paths`. Because `fish_user_paths` will always be
inserted before the inherited `PATH` env. This makes the added path
unintentionally has a higher priority.
This patch makes the above issues by adding `--global` and `--path`
options to `fish_user_paths` which limits the modification scope and
ensures that the path won't be added if it already exists in `PATH`.
This fixes the source of a deadlock that some tip users have hit. If our
surface mailbox is full and there is a dirty scrollbar state, then
drawFrame would block forever trying to queue to the surface mailbox.
We now fail instantly if the queue is full and keep the scrollbar state
dirty. We can try again on the next frame, it's not a critical thing to
get updated.
This fixes the source of a deadlock that some tip users have hit. If our
surface mailbox is full and there is a dirty scrollbar state, then
drawFrame would block forever trying to queue to the surface mailbox.
We now fail instantly if the queue is full and keep the scrollbar state
dirty. We can try again on the next frame, it's not a critical thing to
get updated.
Fixes#2731 (again)
This regressed in 1.2 due to the renderer rework missing porting this. I
believe this issue is still valid even with the rework since the font
grid changes the atlas and if there are still cached cells that
reference the old atlas coordinates it will produce garbage.
Related to #111
This adds the necessary logic and data for the `PageList` data structure
to keep track of **total length** of the screen, **offset** into the
viewport, and **length** of the viewport. These three values are
necessary to _render_ a scrollbar. This PR updates the renderer to grab
this information but stops short of actually drawing a scrollbar (which
we'll do with native UI), in the interest of having a PR that doesn't
contain too many changes.
**This doesn't yet draw a scrollbar, these are just the internal changes
necessary to support it.**
## Background
The `PageList` structure is very core to how we represent terminal
state. It maintains a doubly linked list of "pages" (not literally
virtual memory pages, but close). Each page stores cell information,
styles, hyperlinks, etc fully self-contained in a contiguous sets of VM
pages using offset addresses rather than full pointers. **Pages are not
guaranteed to be equal sizes.** (This is where scrollbars get difficult)
Because it is a linked list structure of non-equal sized nodes, it isn't
amenable to typical scrollbar behavior. A scrollbar needs to know: full
size, offset, and length in order to draw the scrollbar properly.
Getting these values naively is `O(N)` within the data structure that is
on the hottest IO performance path in all of Ghostty.
## Implementation
### PageList
We now maintain two cached values for **total length** and **viewport
offset**.
The total length is relatively straightforward, we just have to be
careful to update it in every operation that could add or remove rows.
I've done this and ensured that every place we update it is covered with
unit test coverage.
The viewport offset is nasty, but I came up with what I believe is a
good solution. The viewport when arbitrarily scrolled is defined as a
direct pointer to the linked list node plus a row offset into that node.
The only way to calculate offset from the top is `O(N)`.
But we have a couple shortcuts:
1. If the viewport is at the bottom (most common) or top, calculating
the offset is `O(1)`: bottom is `total_rows - active_rows`, both readily
available. And top is `0` by definition.
2. Operations on the PageList typically add or remove rows. We don't do
arbitrary linked list surgery. If we instrument those areas with delta
updates to our cache, we can avoid the `O(N)` cost for most operations,
including scrolling a scrollbar. The only expensive operation is a full,
arbitrary jump (new node pointer).
Point 1 was quick to implement, so I focused all the complexity on point
2. Whenever we have an operation that adds or removes rows (for example
pruning the scroll back, adding more, erase rows within the active area,
etc.) then I do the math to calculate the delta change required for the
offset if we've already calculated it, and apply that directly.
### Renderer
The other issue was how to notify the apprts of scrollbar state. Sending
messages on any terminal change within the IO thread is a non-option
because (1) sending messages is slow (2) the terminal changes a lot and
(3) any slowness in the IO thread slows down overall terminal
throughput.
The solution was to **trigger scrollbar notifications with the renderer
vsync**. We read the scrollbar information when we render a frame,
compare it to renderer previous state, and if the scrollbar changed,
send a message to the apprt _after the frame is GPU-renderer_.
The renderer spends _most_ of its time sleeping compared to the IO
thread, and has more opportunities for optimizing its awake time.
Additionally, there's no reason to update the scrollbar information if
the renderer hasn't rendered the new frames because the user can't even
see the stuff the scrollbar wants to scroll to. We're talking about
millisecond scale stuff here at worst but it adds up.
## Performance
No noticeable performance impact for the additional metrics:
<img width="1012" height="738" alt="image"
src="https://github.com/user-attachments/assets/4ed0a3e8-6d76-40c1-b249-e34041c2f6fd"
/>
## AI Usage
I used Amp to help audit the codebase and write tests. I wrote all the
main implementation code manually. I came up with the main design
myself. Relevant threads:
-
https://ampcode.com/threads/T-95fff686-75bb-4553-a2fb-e41fe4cd4b77#message-0-block-0
-
https://ampcode.com/threads/T-48e9a288-b280-4eec-83b7-ca73d029b4ef#message-91-block-0
## Future
This is just the internal changes necessary to _draw_ a scrollbar. There
will be other changes we'll need to add to handle grabbing and actually
jumping the scrollbar. I have a good idea of how to implement those
performantly as well.
Fixes#9191
This changes our color change operations from writing to the renderer
mailbox directly to using our `rendererMailboxWriter` function which
handles the scenario where the mailbox is full by yielding the lock,
waking up the renderer, and retrying later.
This is a known deadlock scenario we've worked around since the private
beta days, but unfortunately this slipped through and I didn't catch it
in review.
What happens here is it's possible with certain escape sequences for the
IO thread to saturate other mailboxes with messages while holding the
terminal state lock. This can happen to any thread. This ultimately
leads to starvation and all threads deadlock.
Our IO thread is the only thread that produces this kind of massive
stream of events while holding the lock, so we have helpers in it to
attempt to queue (cheap, fast) and if that fails then to yield the lock,
wakeup the target thread, requeue, and grab the lock again (expensive,
slow).
This hasn't caused any known bugs but leads to selection memory
corruption and assertion failures in runtime safe modes. When the
terminal screen changes (primary to secondary) and we have an active
dragging mode going either by moving the mouse or our selection tick
timer, we should halt.
We still keep the mouse state active which lets selection continue once
the screen switches back.
- update nixpkgs now that Zig 0.15.2 is available in nixpkgs
- drop hack that worked around compile failures on systems with more
than 32 cores
- enforce patch version of Zig
This reimplements the MAC address-aware URI parsing logic used by the
OSC 7 handler and adds an additional .raw_path option that returns the
full, unencoded path string (including query and fragment values), which
is needed for compliant kitty-shell-cwd:// handling.
Notably, this implementation takes an options-based approach that allows
these additional behaviors to be enabled at runtime. It also leverages
two std.Uri.parse guarantees:
1. Return slices point into the original text string.
2. .raw components don't require unescaping (.percent_encoded does).
The implementation is in a new 'os.uri' module because its now generic
enough to not be hostname-oriented.
We use os.uri.parseUri and its parsing options to reimplement our OSC 7
file-style URI handling. This has two advantages:
First, it fixes kitty-shell-cwd scheme handling. This scheme expects the
full, unencoded path string, whereas the file scheme expects normal URI
percent encoding. This was preventing paths containing "special" URI
characters (like "path?") from working correctly in our bash, zsh, and
elvish shell integrations, which report working directories using the
kitty-shell-cwd scheme. (fish uses file URIs, which work as expected.)
Second, we can greatly simplify our hostname and path string handling
because we can now rely on the "raw" std.Uri component form to always
provide the correct representation.
Lastly, this lets us remove the previous URI-related code from the
os.hostname module, restoring its focus to hostname-related functions.
See: #5289
This might fix#9191, but since I don't have a reproduction I can't be
sure. In any case, this is a bad bug that should be fixed.
The issue is that we weren't checking our scroll timer completion state.
This meant that if `start_scroll_timer` was called multiple times within
a single loop tick, we'd enqueue our completion multiple times, leading
to various undefined behaviors.
If we don't enqueue anything else in the interim, this is safe by
chance. But if we enqueue something else, then we'd hit a queue
assertion failure and honestly I'm not totally sure what would happen.
I wasn't able to trigger the "bad" case, but I was able to trigger the
benign case very easily. Our other timers such as the renderer cursor
timer already have this protection.
Let's fix this and continue looking...
Update libvaxis. The latest commit of libvaxis includes `uucode` as the
unicode
library. `uucode` has a much cleaner API and is actively developed by a
Ghostty
maintainer (@jacobsandlund). This also has the advantage of removing the
last
transitive dependency Ghostty has that is hosted on codeberg, which
separates us
from the frequent outages.
Disclosures: I used AI to debug the import issue in `boo.zig`
---------
Co-authored-by: Mitchell Hashimoto <m@mitchellh.com>
Resolves#8689
For various reason, ghostty wants to have a unique file extension for
the config files. The name was settled on `config.ghostty`. This will
help with tooling. See #8438 (original discussion) for more details.
This PR introduces the preferred default of `.ghostty` while still
supporting the previous `config` file. If both files exist, a warning
log is sent.
The docs / website will need to be updated to reflect this change.
> [!NOTE]
> Only tested on macOS 26.0.
---------
Co-authored-by: Mitchell Hashimoto <m@mitchellh.com>
Previously, the fish shell integration interfered with fish's builtin vi
mode cursor switching configurations such as `$fish_cursor_default` and
`$fish_cursor_insert`.
```console
$ ghostty --config-default-files=false -e fish --no-config --init-command 'source "$GHOSTTY_RESOURCES_DIR"/shell-integration/fish/vendor_conf.d/ghostty-shell-integration.fish; fish_vi_key_bindings'
```
The above command starts fish in vi mode with Ghostty shell
integrations. Manually loading the integration is necessary due to
`--no-config` blocking auto injection.
1. At the prompt, fish is in insert mode, and the cursor is a blinking
beam. However, press escape and then "i" to exit then re-enter insert
mode, and the cursor will be a solid beam due to the
`$fish_cursor_unknown` setting. Without the shell integration, insert
mode always uses a solid beam cursor.
2. A similar problem shows if we start fish with `fish_vi_key_bindings
default`. The cursor ends up as a blinking beam in normal mode only due
to the shell integration interfering. This glitch can also be reset away
by entering then exiting insert mode.
3. Also, `$fish_cursor_external` has no effect when used with shell
integration. After `fish_vi_key_bindings`, set it to `line`, run cat(1),
and shell integration will give you a blinking block, not the asked for
line/beam.
I verified that this patch makes the shell integration stop interfering
in three scenarios above, and it still changes the cursor when not using
fish's vi mode.
Note that `$fish_cursor_*` variables can be set when fish isn't in vi
mode, so they're not great signals for the shell integration hooks.
As pointed out in #9156, an unintended consequence of all the work to
get icon sizing right is that `adjust-icon-height` now only applies to
the small icons you get when the next cell is not whitespace. Large
icons are unaffected.
With this PR, `adjust-icon-height` affects the maximum height of every
symbol specifying the `.icon` constraint height, regardless of
constraint width. This includes most Nerd Font icons, but excludes emoji
and other unicode symbols, and also excludes terminal graphics-oriented
Nerd Font symbols such as Powerline symbols.
In the following screenshots, **Baseline** is without
`adjust-icon-height`, while **Before** and **After** are with
`adjust-icon-height = -25%`.
**Baseline**
<img width="711" height="95" alt="Screenshot 2025-10-11 at 23 28 20"
src="https://github.com/user-attachments/assets/7499db4d-75a4-4dbd-b107-8cb5849e31a3"
/>
**Before** (only small icons affected)
<img width="711" height="95" alt="Screenshot 2025-10-11 at 23 20 12"
src="https://github.com/user-attachments/assets/9afd9fbf-ef25-44cc-9d8e-c39a69875163"
/>
**After** (both small and large icons affected, but not emoji)
<img width="711" height="95" alt="Screenshot 2025-10-11 at 23 21 05"
src="https://github.com/user-attachments/assets/90999f59-3b43-4684-9c8e-2c3c1edd6d18"
/>
Of course #9142 would require a minor follow-up!
* Scale groups can cut across patch sets, but not across fonts. We had
some scale group mixing between Font Awesome and the weather symbols,
which is removed by this PR.[^cp_table_full]
* There's one case where a scale group includes a glyph that's not part
of any patch sets, just for padding out the group bounding box.
Previously, an unrelated glyph from a different font would be pulled in.
Now we use an appropriate stand-in. (See code comment for details.)
* I noticed overlaps weren't being split between each side of the
bounding box, they were added to both sides, resulting in twice as much
padding as specified.
Screenshots showing the extra vertical padding for progress bar elements
due to the second bullet point:
**Before**
<img width="191" height="42" alt="Screenshot 2025-10-11 at 15 33 54"
src="https://github.com/user-attachments/assets/cf288cce-86d3-46fd-ae86-18e5c274b0e4"
/>
**After**
<img width="191" height="42" alt="Screenshot 2025-10-11 at 15 33 20"
src="https://github.com/user-attachments/assets/7ac799c7-bf50-4e65-a74a-f8a2c42d2441"
/>
[^cp_table_full]: Forming and using the merged `cp_table_full` table
should have been a red flag. Such a table doesn't make sense, it would
be a one-to-many map. You need the names of the original fonts to
disambiguate.
Fixes#9076
**Before**
<img width="128" height="57" alt="Screenshot 2025-10-11 at 00 07 09"
src="https://github.com/user-attachments/assets/a6b416d5-dae1-4cea-a836-00640ceaf39b"
/>
**After**
<img width="128" height="57" alt="Screenshot 2025-10-11 at 00 07 31"
src="https://github.com/user-attachments/assets/7d2df7b1-4767-4e2d-84d2-8301da5c6602"
/>
These screenshots show the chevrons mentioned in
https://github.com/ghostty-org/ghostty/discussions/7820#discussioncomment-14617170,
which should be scaled as a group but were not until this PR.
The added code downloads each individual symbol font file from the Nerd
Fonts github repo (making sure to get the version corresponding to the
vendored `font-patcher.py`) and iterates over all of them to build the
correct and complete codepoint mapping table. The table is saved to
`nerd_font_codepoint_tables.py`, which `nerd_font_codegen.py` will reuse
if possible instead of downloading the font files again.
I'm not going to utter any famous last words or anything, but... after
this, I don't think the number of remaining issues with icon
scaling/alignment is _large._
Clicking on desktop notifications sent by Ghostty _should_ cause the
window that sent the notification to come to the top. However, because
the notification that was sent targeted the wrong surface (apprt surface
vs core surface) and the window did not call `present()` on itself the
window would never be brought to the surface, the correct tab would not
be selected, etc.
Fixes#9145
