This causes linker issues for some libghostty users. I don't know why we
never saw these issues with Ghostty release builds, but generally
speaking I think its fine to do this for 3rd party code unless we've
witnessed an issue. And these deps have been stable for a long, long
time.
These now properly match the FreeType API- compared directly in the unit
tests against the values provided by the FreeType header itself.
This was ridiculously wrong before, like... wow.
FcLangSetHasLang returns FcLangResult enum values:
- FcLangEqual (0): Exact match
- FcLangDifferentTerritory (1): Same language, different territory
- FcLangDifferentLang (2): Different language
The previous comparison to FcTrue (1) caused:
- Exact matches (0) to incorrectly return false
- Partial matches (1) to incorrectly return true
This fix changes the comparison to FcLangEqual (0) so hasLang()
correctly returns true only for exact language matches.
Fixes emoji font detection which relies on checking for 'und-zsye'
language tag support.
Fixes#9579
Protect against panics caused by integer overflows by using functions
that allow integer overflows to be caught instead of causing a panic.
Also protect against DOS from images that might not cause an
overflow but do consume an absurd amount of memory by limiting
images to a maximum size of 4GiB (which is the maximum size of
`image-storage-limit`).
The solution we had before worked in most cases but there were some
which caused problems still. This is what HarfBuzz's CoreText shaper
backend does, it uses a CTTypesetter with the forced embedding level
attribute. This fixes the failure case I found that was causing non-
monotonic outputs which can have all sorts of unexpected results, and
causes a crash in Debug modes because we assert the monotonicity while
rendering.
Reduce potential allocation while processing glyphs by ensuring capacity
in the buffer ahead of time and also using CTRunGet*Ptr functions first
and only allocating for those if that didn't work (it should almost
always work in practice.)
Follow-up to #8720 adding
* Two improvements to FreeType glyph measurements:
- Ensuring that glyphs are measured with the same hinting as they are
rendered, ref
[#8720#issuecomment-3305408157](https://github.com/ghostty-org/ghostty/pull/8720#issuecomment-3305408157);
- For outline glyphs, using the outline bbox instead of the built-in
metrics, like `renderGlyph()`.
* Basic unit tests for face metrics and their estimators, using the
narrowest and widest fonts from the resource directory, Cozette Vector
and Geist Mono.
---
I also made one unrelated change to `freetype.zig`, replacing
`@alignCast(@ptrCast(...))` with `@ptrCast(@alignCast(...))` on line
173. Autoformatting has been making this change on every save for weeks,
and reverting the hunk before each commit is getting old, so I hope it's
OK that I use this PR to upstream this decree from the formatter.
This makes more dependencies lazy. This has a practical effect of
reducing the number of dependencies that need to be downloaded when
running certain zig build steps.
This is all limited because we're blocked on an upstream Zig issue:
https://github.com/ziglang/zig/issues/21525 This prevents us from
fully avoiding downloading many dependencies, but at least they're
relatively small.
One major improvement here is the usage of `lazyImport` for
`zig-wayland` that prevents downloading `zig_wayland` unconditionally on
all platforms. On macOS, we don't download this at all anymore.
Another, weirder change is that all our transitive dependencies are now
marked lazy (e.g. glslang's upstream source) even if the glslang build
always requires it. This was necessary because without this, even if we
simply referenced glslang in the root build.zig, it would force the
source package to download unconditionally. This no longer happens.
The GLSL to MSL conversion process uses a passed-in sampler state for
the `iChannel0` parameter and we weren't providing it. This magically
worked on Apple Silicon for unknown reasons but failed on Intel GPUs.
In normal, hand-written MSL, we'd explicitly create the sampler state as
a normal variable (we do this in `shaders.metal` already!), but the
Shadertoy conversion stuff doesn't do this, probably because the exact
sampler parameters can't be safely known.
This fixes a Metal validation error when using custom shaders:
```
-[MTLDebugRenderCommandEncoder validateCommonDrawErrors:]:5970: failed
assertion `Draw Errors Validation Fragment Function(main0): missing Sampler
binding at index 0 for iChannel0Smplr[0].
```
Hi there, this is just a low-hanging fruit and it also prepares the way
for the future 0.15, which removes addStaticLibrary.
Please, let me know what to do on the `// TODO` comments.
Found by Valgrind:
```
==265734== 320 bytes in 8 blocks are definitely lost in loss record 13,786 of 15,141
==265734== at 0x5A65810: malloc (in /nix/store/l431jn8ahj09g5c1arrl7q6wcxngg21q-valgrind-3.24.0/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==265734== by 0x3D799EB: onig_region_resize (.cache/zig/p/N-V-__8AAHjwMQDBXnLq3Q2QhaivE0kE2aD138vtX2Bq1g7c/src/regexec.c:923)
==265734== by 0x3D81BCA: onig_region_resize_clear (.cache/zig/p/N-V-__8AAHjwMQDBXnLq3Q2QhaivE0kE2aD138vtX2Bq1g7c/src/regexec.c:949)
==265734== by 0x3DA814F: search_in_range (.cache/zig/p/N-V-__8AAHjwMQDBXnLq3Q2QhaivE0kE2aD138vtX2Bq1g7c/src/regexec.c:5454)
==265734== by 0x3DA7F25: onig_search (.cache/zig/p/N-V-__8AAHjwMQDBXnLq3Q2QhaivE0kE2aD138vtX2Bq1g7c/src/regexec.c:5414)
==265734== by 0x382E7E8: regex.Regex.searchAdvanced (regex.zig:61)
==265734== by 0x382E974: regex.Regex.search (regex.zig:48)
==265734== by 0x382EC08: terminal.StringMap.SearchIterator.next (StringMap.zig:40)
==265734== by 0x39ADDCD: renderer.link.Set.matchSetFromLinks (link.zig:320)
==265734== by 0x39AE5C7: renderer.link.Set.matchSet (link.zig:95)
==265734== by 0x39BCCE1: renderer.generic.Renderer(renderer.OpenGL).rebuildCells (generic.zig:2307)
==265734== by 0x39C3BDB: renderer.generic.Renderer(renderer.OpenGL).updateFrame (generic.zig:1228)
==265734== by 0x3993E88: renderer.Thread.renderCallback (Thread.zig:607)
==265734== by 0x3993CFF: renderer.Thread.wakeupCallback (Thread.zig:524)
==265734== by 0x39C522E: callback (async.zig:679)
==265734== by 0x39C522E: watcher.async.AsyncEventFd(api.Xev(.io_uring,backend.io_uring)).waitPoll__anon_592685__struct_596870.callback (async.zig:181)
==265734== by 0x3970EAE: backend.io_uring.Completion.invoke (io_uring.zig:804)
==265734== by 0x3973AD8: backend.io_uring.Loop.tick___anon_586861 (io_uring.zig:193)
==265734== by 0x3973BCD: backend.io_uring.Loop.run (io_uring.zig:84)
==265734== by 0x3978673: dynamic.Xev(&.{ .io_uring, .epoll }[0..2]).Loop.run (dynamic.zig:172)
==265734== by 0x3978972: renderer.Thread.threadMain_ (Thread.zig:263)
==265734== by 0x3954580: renderer.Thread.threadMain (Thread.zig:202)
==265734== by 0x39279CA: Thread.callFn__anon_573552 (Thread.zig:488)
==265734== by 0x38F4594: Thread.PosixThreadImpl.spawn__anon_570448.Instance.entryFn (Thread.zig:757)
==265734== by 0x6E567EA: start_thread (pthread_create.c:448)
==265734== by 0x6ED9FB3: clone (clone.S:100)
==265734==
```
Related to #7879
This commit updates `zig build test` to run Xcode tests, too. These run
in parallel to the Zig tests, so they don't add any time to the test.
The Xcode tests will _not_ run when: (1) the target is not macOS, or (2)
the `-Dtest-filter` option is non-empty. This makes it so that this
change doesn't affect non-macOS and doesn't affect the general dev cycle
because you usually will run `-Dtest-filter` when developing a core
feature.
I didn't add a step to only run Xcode tests because I find that when I'm
working in Xcode I'm probably going to run the tests from there anyways.
The integration with `zig build test` is just a convenience, especially
around CI.
Speaking of CI, this change also makes it so this will run in CI.
This sets up for a couple improvments (see TODO comments) and also sets
the glyph atlas textures to nearest neighbor sampling since we can do
that now that we never scale glyphs.
This is a big'un.
- **Glyph constraint logic is now done fully on the CPU** at the
rasterization stage, so it only needs to be done once per glyph instead
of every frame. This also lets us eliminate padding between glyphs on
the atlas because we're doing nearest-neighbor sampling instead of
interpolating-- which ever so slightly increases our packing efficiency.
- **Special constraints for nerd font glyphs** are applied based roughly
on the constraints they use in their patcher. It's a simplification of
what they do, the largest difference being that they scale groups of
glyphs based on a shared bounding box so that they maintain relative
size to one another, but that would require loading all glyphs on the
group and I'd want to do that on font load TBH and at that point I'd
basically be re-implementing the nerd fonts patcher in Zig to patch
fonts at load time which is way beyond the scope I want to have. (Fixes
#7069)
- These constraints allow for **perfectly sized and centered emojis**,
this is very nice.
- **Changed the default embedded fonts** from 4 copies (regular, italic,
bold, bold italic) of a patched (and outdated) JetBrains Mono to a
single JetBrains Mono variable font and a single Nerd Fonts Symbols Only
font. This cuts the weight of those down from 9MB to 3MB!
- **FreeType's `renderGlyph` is significantly reworked**, and the new
code is, IMO, much cleaner- although there are probably some edge case
behavior differences I've introduced.
> [!NOTE]
> One breaking change I definitely introduced is changing the
`monochrome` freetype load flag config from its previous completely
backwards meaning to instead the correct one (I also changed the
default, so this won't affect any user who hasn't touched it, but users
who set the `monochrome` flag will find their fonts quite crispy after
this change because they will have no anti-aliasing anymore)
### Future work
Following this change I want to get to work on automatic font size
matching (a la CSS
[`font-size-adjust`](https://developer.mozilla.org/en-US/docs/Web/CSS/font-size-adjust)).
I set the stage for that quite some time ago so it shouldn't be too much
work and it will be a big benefit for users who regularly use multiple
writing systems and so have multiple fonts for them that aren't
necessarily size-compatible.
This is in preparation to move constraint off the GPU to simplify our
shaders, instead we only need to constrain once at raster time and never
again.
This also significantly reworks the freetype renderGlyph function to be
generally much cleaner and more straightforward.
This commit doesn't actually apply the constraints to anything yet, that
will be in following commits.
This deletes the GLFW apprt from the Ghostty codebase.
The GLFW apprt was the original apprt used by Ghostty (well, before
Ghostty even had the concept of an "apprt" -- it was all just a single
application then). It let me iterate on the core terminal features,
rendering, etc. without bothering about the UI. It was a good way to get
started. But it has long since outlived its usefulness.
We've had a stable GTK apprt for Linux (and Windows via WSL) and a
native macOS app via libghostty for awhile now. The GLFW apprt only
remained within the tree for a few reasons:
1. Primarily, it provided a faster feedback loop on macOS because
building the macOS app historically required us to hop out of the
zig build system and into Xcode, which is slow and cumbersome.
2. It was a convenient way to narrow whether a bug was in the
core Ghostty codebase or in the apprt itself. If a bug was in both
the glfw and macOS app then it was likely in the core.
3. It provided us a way on macOS to test OpenGL.
All of these reasons are no longer valid. Respectively:
1. Our Zig build scripts now execute the `xcodebuild` CLI directly and
can open the resulting app, stream logs, etc. This is the same
experience we have on Linux. (Xcode has always been a dependency of
building on macOS in general, so this is not cumbersome.)
2. We have a healthy group of maintainers, many of which have access
to both macOS and Linux, so we can quickly narrow down bugs
regardless of the apprt.
3. Our OpenGL renderer hasn't been compatible with macOS for some time
now, so this is no longer a useful feature.
At this point, the GLFW apprt is just a burden. It adds complexity
across the board, and some people try to run Ghostty with it in the real
world and get confused when it doesn't work (it's always been lacking in
features and buggy compared to the other apprts).
So, it's time to say goodbye. Its bittersweet because it is a big part
of Ghostty's history, but we've grown up now and it's time to move on.
Thank you, goodbye.
(NOTE: If you are a user of the GLFW apprt, then please fork the project
prior to this commit or start a new project based on it. We've warned
against using it for a very, very long time now.)
Supersedes #7154
In gtk4-layer-shell versions < 1.0.4, the app could crash upon opening
a quick terminal window on certain compositors that implement the
`xdg_wm_dialog_v1` protocol. The exact reason is a bit complicated,
but is nicely summarized in the upstream issue (wmww/gtk4-layer-shell#50).
The circumstances that could cause this crash to occur should gradually
diminish as distros update to newer gtk4-layer-shell versions, but this
is known to crash on Fedora 41 and Hyprland, which could be a sizable
chunk of our userbase given that this would also occur on GNOME/Mutter
and KDE/KWin. The diff should be minimal enough that this can be removed
or reverted once this band-aid fix is no longer necessary.
BPTC is required to be available OpenGL >= 4.2 and our minimum is 4.3 so
this is safe in terms of support. I tested briefly in a VM and didn't
encounter any problems so this should just be a complete win.
(Note: texture data is already automatically compressed on Metal)
The code in metal/image.zig and opengl/image.zig was virtually identical
save for the texture options, so I've moved that to the GraphicsAPI and
unified them in to renderer/image.zig
It's here, the long-foretold and long-procrastinated renderer rework!
Hopefully this makes it easier to adapt and modify the renderer in the
future and ensures feature parity between Metal and OpenGL. Despite
having been a lot of work to write initially, with the abstraction layer
in place I feel like working on the renderer will be a much more
pleasant experience going forward.
## Key points
- CPU-side renderer logic is now mostly unified via a generic
`Renderer`.
- A graphics API abstraction layer over OpenGL and Metal has been
introduced.
- Minimum OpenGL version bumped to `4.3`, so can no longer be run on
macOS; I used the nix VM stuff for my testing during development. (Edit
by @mitchellh: Note for readers that Ghostty still works on macOS, but
the OpenGL backend doesn't, only the Metal one)
- The OpenGL backend now supports linear blending! Woohoo! The default
`alpha-blending` has been updated to `linear-corrected` since it's
essentially a strict improvement over `native`. The default on macOS is
still `native` though to match other mac apps in appearance, since macOS
users are more sensitive to text appearance.
- Custom shaders can now be hot reloaded.
- The background color is once again drawn by us, so custom shaders can
interact with it properly. In general, custom shaders should be a little
more robust.
## The abstraction layer
The general hierarchy of the abstraction layer is as such:
```
[ GraphicsAPI ] - Responsible for configuring the runtime surface
| | and providing render `Target`s that draw to it,
| | as well as `Frame`s and `Pipeline`s.
| V
| [ Target ] - Represents an abstract target for rendering, which
| could be a surface directly but is also used as an
| abstraction for off-screen frame buffers.
V
[ Frame ] - Represents the context for drawing a given frame,
| provides `RenderPass`es for issuing draw commands
| to, and reports the frame health when complete.
V
[ RenderPass ] - Represents a render pass in a frame, consisting of
: one or more `Step`s applied to the same target(s),
[ Step ] - - - - each describing the input buffers and textures and
: the vertex/fragment functions and geometry to use.
:_ _ _ _ _ _ _ _ _ _/
v
[ Pipeline ] - Describes a vertex and fragment function to be used
for a `Step`; the `GraphicsAPI` is responsible for
these and they should be constructed and cached
ahead of time.
[ Buffer ] - An abstraction over a GPU buffer.
[ Texture ] - An abstraction over a GPU texture.
```
More specific documentation can be found on the relevant structures.
## Miscellany
Renderers (which effectively just means the generic renderer) are now
expected to only touch GPU resources in `init`, certain lifecycle
functions such as the `displayRealized`/`displayUnrealized` callbacks
from GTK-- and `drawFrame`; and are also expected to be thread-safe.
This allows the renderer thread to build the CPU-side buffers
(`updateFrame`) even if we can only *draw* from the app thread.
Because of this change, we can draw synchronously from the main thread
on macOS when necessary to always have a frame of the correct size
during a resize animation. This was necessary to allow the background to
be drawn by our GPU code (instead of setting a background color on the
layer) while still avoiding holes during resize.
The OpenGL backend now theoretically has access to multi-buffering, but
it's disabled (by setting the buffer count to 1) because it
synchronously waits for frames to complete anyway which means that the
extra buffers were just a waste of memory.
## Validation
To validate that there are no significant or obvious problems, I
exercised both backends with a variety of configurations, and visually
inspected the results. Everything looks to be in order.
The images are available in a gist here:
https://gist.github.com/qwerasd205/c1bd3e4c694d888e41612e53c0560179
## Memory
Here's a comparison of memory usage for ReleaseFast builds on macOS,
between `main` and this branch.
Memory figures given are values from Activity Monitor measuring windows
of the same size, with two tabs with 3 splits each.
||Before|After|
|-:|-|-|
|**Memory**|247.9 MB|224.2 MB|
|**Real Memory**|174.4 MB|172.5 MB|
Happily, the rework has slightly *reduced* the memory footprint- likely
due to removing the overhead of `CAMetalLayer`. (The footprint could be
reduced much further if we got rid of multi-buffering and satisfied
ourselves with blocking for each frame, but that's a discussion for
another day.)
If someone could do a similar comparison for Linux, that'd be much
appreciated!
## Notes / future work
- There are a couple structures that *can* be unified using the
abstraction layer, but I haven't gotten around to unifying yet.
Specifically, in `renderer/(opengl|metal)/`, there's `cell.zig` and
`image.zig`, both of which are substantially identical between the two
backends. `shaders.zig` may also be a candidate for unification, but
that might be *overly* DRY.
- ~~I did not double-check the documentation for config options, which
may mention whether certain options can be hot-reloaded; if it does then
that will need to be updated.~~ Fixed: be5908f
- The `fragCoord` for custom shaders originates at the top left for
Metal, but *bottom* left for OpenGL; fixing this will be a bit annoying,
since the screen texture is likewise vertically flipped between the two.
Some shaders rely on the fragcoord for things like falling particles, so
this does need to be fixed.
- `Target` should be improved to support multiple types of targets right
now it only represents a framebuffer or iosurface, but it should also be
able to represent a texture; right now a kind of messy tagged union is
used so that steps can accept both.
- Custom shader cursor uniforms (#6912) and terminal background images
(#4226, #5233) should be much more straightforward to implement on top
of this rework, and I plan to make follow-up PRs for them once this is
merged.
- I *do* want to do a rework of the pipelines themselves, since the way
we're rendering stuff is a bit messy currently, but this is already a
huge enough PR as it is- so for now the renderer still uses the same
rendering passes that Metal did before.
- We should probably add a system requirements section to the README
where we can note the minimum required OpenGL version of `4.3`, any even
slightly modern Linux system will support this, but it would be good to
document it somewhere user-facing anyway.
# TODO BEFORE MERGE
- [x] Have multiple people test this on both macOS and linux.
- [ ] ~~Have someone with a better dev setup on linux check for memory
leaks and other problems.~~ (Skipped, will merge and let tip users
figure this out, someone should *specifically* look for memory leaks
before the next versioned release though.)
- [x] Address any code review feedback.
This commit is very large, representing about a month of work with many
interdependent changes that don't separate cleanly in to atomic commits.
The main change here is unifying the renderer logic to a single generic
renderer, implemented on top of an abstraction layer over OpenGL/Metal.
I'll write a more complete summary of the changes in the description of
the PR.
Updates all copyright notices to include "Ghostty contributors" to
reflect the fact that Mitchell is not the sole copyright owner.
Also adds "Ghostty contributors" to the author section in the manpages,
linking https://github.com/ghostty-org/ghostty/graphs/contributors for
proper credit.
Even though gtk4-layer-shell's documentation claims that "nobody quite
knows what it's for", some compositors (like Niri) can define custom
rules based on the layer name and it's beneficial in those cases to
define a distinct name just for our quick terminals.
This fixes an issue where Ghostty would not build against the macOS 15.5 SDK.
What was happening was that Zig was adding its embedded libc paths to
the clang command line, which included old headers that were
incompatible with the latest (macOS 15.5) SDK. Ghostty was adding the
newer paths but they were being overridden by the embedded libc paths.
The reason this was happening is because Zig was using its own logic to
find the libc paths and this was colliding with the paths we were
setting manually. To fix this, we now use a `libc.txt` file that
explicitly tells Zig where to find libc, and we base this on our own SDK
search logic.
This commit changes a LOT of areas of the code to use decl literals
instead of redundantly referring to the type.
These changes were mostly driven by some regex searches and then manual
adjustment on a case-by-case basis.
I almost certainly missed quite a few places where decl literals could
be used, but this is a good first step in converting things, and other
instances can be addressed when they're discovered.
I tested GLFW+Metal and building the framework on macOS and tested a GTK
build on Linux, so I'm 99% sure I didn't introduce any syntax errors or
other problems with this. (fingers crossed)
Mitchell Hashimoto