Fixes#8487
I did this by setting up a publisher on `BaseTerminalController` for any
bell state change on any surfaces in the tree (including removing
surfaces). By listening to this event at AppDelegate and reinspecting
all our windows, we can reliably set the badge.
**This also includes a change to show the number of terminals with an
active bell!** We can now determine the number, so we show it!
The C struct Palette.C declared colors as [265]Color.C, but the terminal
palette is 256 colors (terminal.color.Palette = [256]RGB) and the C
header ghostty_config_palette_s correctly uses colors[256].
The mismatch causes ghostty_config_get to write 265×3 = 795 bytes
through a pointer sized for 256×3 = 768 bytes, producing a 27-byte
buffer overflow. On macOS Release builds with stack protector enabled,
this triggers __stack_chk_fail → SIGABRT on launch.
Discovered this while working on https://github.com/markhuot/watchtower.
Builds were succeeding but the app was crashing on launch because we use
the new C bindings to get the foreground/background colors to set window
theming like "selection" color.
The C struct Palette.C declared colors as [265]Color.C, but the
terminal palette is 256 colors (terminal.color.Palette = [256]RGB)
and the C header ghostty_config_palette_s correctly uses colors[256].
The mismatch causes ghostty_config_get to write 265×3 = 795 bytes
through a pointer sized for 256×3 = 768 bytes, producing a 27-byte
buffer overflow. On macOS Release builds with stack protector enabled,
this triggers __stack_chk_fail → SIGABRT on launch.
Fixes#10406
ImGui_ImplOpenGL3_Shutdown() calls imgl3wShutdown() which dlcloses the
GL library handles but does not zero out the imgl3w function pointer
table (imgl3wProcs). When a GLArea is re-realized (e.g. during
reparenting), ImGui_ImplOpenGL3_Init() calls
ImGui_ImplOpenGL3_InitLoader() which checks "if (glGetIntegerv ==
nullptr)". Since the stale pointers are non-null, it skips
re-initialization. The next GL call through a dangling function pointer
causes a SIGSEGV.
Fix this by introducing ImGui_ImplOpenGL3_ShutdownWithLoaderCleanup()
which calls the normal shutdown and then zeroes the imgl3wProcs table,
forcing the next Init to reload GL function pointers via imgl3wInit().
Also properly destroy the ImGui context and reset widget state in
glAreaUnrealize so re-realize starts clean. This was extra but was
probably leaking memory.
Fixes#10406
ImGui_ImplOpenGL3_Shutdown() calls imgl3wShutdown() which dlcloses the
GL library handles but does not zero out the imgl3w function pointer
table (imgl3wProcs). When a GLArea is re-realized (e.g. during
reparenting), ImGui_ImplOpenGL3_Init() calls ImGui_ImplOpenGL3_InitLoader()
which checks "if (glGetIntegerv == nullptr)". Since the stale pointers
are non-null, it skips re-initialization. The next GL call through a
dangling function pointer causes a SIGSEGV.
Fix this by introducing ImGui_ImplOpenGL3_ShutdownWithLoaderCleanup()
which calls the normal shutdown and then zeroes the imgl3wProcs table,
forcing the next Init to reload GL function pointers via imgl3wInit().
Also properly destroy the ImGui context and reset widget state in
glAreaUnrealize so re-realize starts clean. This was extra but was
probably leaking memory.
Fixes#10935
This is a more robust way to detect "is my surface focused" because that
question usually means "is my surface the last focused surface" if a
_different_ surface is not focused. We already have used this pattern
all over but we should extend it to SwiftUI too.
Fixes#10935
This is a more robust way to detect "is my surface focused" because that
question usually means "is my surface the last focused surface" if a
_different_ surface is not focused. We already have used this pattern
all over but we should extend it to SwiftUI too.
(moved from #11008)
I have this branch up to fix#10252. It was written with AI (in Cursor),
but only after I made sure I thoroughly understood what was happening
(almost to an excessive extent). I had already determined that an
Objective-C helper was necessary, I just asked GPT 5.3 Codex in Cursor
to write it for me.
**TL;DR: deep within AppKit, there is an Objective-C exception that is
_always_ thrown when opening a new tab from the visual tab picker ("tab
overview"). [Objective-C exceptions *cannot* be safely recovered from in
Swift.](http://developer.apple.com/documentation/swift/handling-cocoa-errors-in-swift#Handle-Exceptions-in-Objective-C-Only)
As Ghostty is primarily Swift, we must introduce some Objective-C
wrapper around tab creation to safely swallow this exception.**
There is a lot more I know about this than the above, and can discuss it
at length if desired. Interestingly, it seems debug builds of Ghostty
(`zig build run`) *do* gracefully recover and don't crash. Release
builds (`zig build run -Doptimize=ReleaseFast`), however, *do* crash.
The crashing seems to be expected behavior and **_I don't think there's
any feasible way to get release builds to recover as debug builds do._**
The debug builds do, arguably, have better animation behavior. Not sure
how I can approach that part.
Release build off of my commit:
https://github.com/user-attachments/assets/c81927be-b2d2-48b3-a18f-30b389a90f04
Debug build off `db1e31c7a69924913e8faafcedb290de3cb4a8b6` (current
`main`, as of writing):
https://github.com/user-attachments/assets/76367154-b039-4453-8d39-8a0465973deb
* Using MAP prohibited `NSDockTilePlugIn` to update custom icons more
reliably. And it also fixes the corner radius issue on older os: #8870
* Changes in AboutWindow cc @jparise
* Start cycling with current icon
* Stop cycling icons when AboutWindow is closed
* Add menu to copy icon config
<img width="166" height="63" alt="Xnip2025-12-20_18-40-58"
src="https://github.com/user-attachments/assets/52fc1215-909e-49c7-a37a-b7c73eef61f1"
/>
> [!WARNING]
> Upgrading from `macOS-custom-icon` needs to manually open the app once
to update the icon, since this plugin is running under
`com.apple.dock.external.extra.arm64`, which has sandbox enabled.
>
> When first upgraded to this commit, a notification about the dock
plugin will pop up. The user has to enable this to change the icon
smoothly.
> <img width="389" height="159" alt="image"
src="https://github.com/user-attachments/assets/a883ac6b-0b4d-4794-8c61-50b60707f6a2"
/>
Here are the testing results on
[Tahoe](https://github.com/user-attachments/assets/e5fc8354-5f5c-4280-805f-88f043ceadca)
and
[Sequoia](https://github.com/user-attachments/assets/633d9a07-7d9d-4806-8496-82ddaffb8833):
> When you see some pause in the recording, that's when I rebuild or
replace the older version with the latest.
This also fixes some issues when changing between different styles,
consistency issues, and resetting from others to `official`.
### Developer's Note
This shouldn't affect current CI flow, since this new target is just a
bundle not runnable, and I tested with archiving, exporting and signing
in Xcode, nothing big changed.
> [!NOTE]
> AI helped me to write the typo ignore-re and proofread my comments
