## Summary
This fixes macOS live-resize behavior for the `eframe`/`egui-wgpu` path
when using the low-latency wgpu surface configuration.
The problem I was seeing is that native window resize can look visibly
below the baseline expected from a desktop GUI: stale or stretched
frames (manifesting as wobble/jitter), or severe lag while dragging a
window edge.
The fix has three parts:
- use `CAMetalLayer.presentsWithTransaction` during live resize to avoid
stale/stretched frames
- temporarily use at least `desired_maximum_frame_latency = 2` while
live resize is active, so transaction presentation does not stall when
the app normally uses `SurfaceConfig::LOW_LATENCY`
- treat macOS `WindowEvent::Moved` as part of the live-resize event
stream, since resizing from the top or left edge changes the window
origin
This PR depends on the winit-side AppKit live-resize timing fix in
[rust-windowing/winit#4588](https://github.com/rust-windowing/winit/pull/4588)
A renderer-only frame-latency change is not enough by itself. The
temporary latency bump only solves the drawable starvation caused by
combining `presentsWithTransaction` with `SurfaceConfig::LOW_LATENCY`.
It does not change when winit emits resize/redraw events, whether
redraws are delivered during AppKit's live-resize event-tracking loop,
or whether the surface size is derived from the current backing rect.
That is why the winit fix is needed first: it makes the windowing layer
report the current AppKit backing size and request redraws from the
live-resize/display callbacks. egui-wgpu still needs this PR on top
because winit does not own the wgpu `Surface` or the underlying
`CAMetalLayer` presentation policy.
In other words: winit fixes when the windowing layer reports
resize/redraw work, while this PR fixes how egui-wgpu presents
Metal-backed wgpu frames during that resize.
## Why change the existing feature?
The existing `macos-window-resize-jitter-fix` feature addresses one
symptom by enabling transaction presentation during resize, but it is
not enough for the low-latency wgpu path.
In particular, `presentsWithTransaction` and
`SurfaceConfig::LOW_LATENCY` interact poorly during AppKit live resize.
The old code avoids that by [skipping transaction presentation when
latency is
`1`](71c4ff3c33/crates/egui-wgpu/src/winit.rs (L417)),
but that means low-latency users get the resize jitter/wobble back.
This PR keeps the low-latency path normally, but temporarily bumps frame
latency only while live resize is active. That gives the resize path
enough drawable slack without changing normal interaction latency.
I removed the `macos-window-resize-jitter-fix` feature because this
seems like the behavior the macOS wgpu path should have by default, not
a separate opt-in. If keeping the feature as a no-op compatibility alias
is preferred, I can adjust the PR.
## Validation
I created a small demo app that somewhat resembles the layout of my
actual app and highlights both horizontal and vertical resize jitter:
- a borderless macOS window
- a simple toolbar
- a scrolling side list
- `SurfaceConfig::LOW_LATENCY`
The toolbar and list make stale or stretched frames easy to see during
native resize. The jitter is visible even on the traffic light buttons.
Recordings:
### Before 1: no transaction presentation, low latency
Shows jitter/wobble and stale/stretched frames during live resize.
https://github.com/user-attachments/assets/2cf4467b-e14c-4f41-8021-0b8c23f41004
### Before 2: transaction presentation with low latency
Shows the other failure mode: live resize can become severely laggy when
transaction presentation is used while keeping
`SurfaceConfig::LOW_LATENCY`.
https://github.com/user-attachments/assets/2f866790-f472-4ede-a3c0-480e8f0f041a
### After: patched egui-wgpu + patched winit, low latency
No visible wobble/jitter and no severe live-resize lag.
https://github.com/user-attachments/assets/59e46e9f-7906-4b5c-a6c7-1d09eae644cd
---------
Co-authored-by: lucasmerlin <hi@lucasmerlin.me>
## Symptom
Fix this long-standing, occasional bug, that can cause text to look
compressed and "drunk":
<img width="552" height="226" alt="Screenshot 2026-06-22 at 13 12 56"
src="https://github.com/user-attachments/assets/9b1abad4-5ef6-4771-8168-f201afc341ab"
/>
## Root cause
`epaint::TextureAtlas::take_delta` is fire-and-forget: it resets the
dirty region as soon as it hands out a delta, assuming the delta will be
uploaded. Atlas growth always emits a **full** `ImageDelta` (`pos:
None`) which recreates the GPU texture at the new size — *as long as it
is applied*. But both native integrations applied `textures_delta`
inside skippable code paths:
- **wgpu** (`egui-wgpu/src/winit.rs`): textures were uploaded only
*after* surface-dependent early-returns (`render_state` /
`surfaces.get_mut(viewport_id)` missing). Texture uploads are
device-level and don't need a surface.
- **glow** (`eframe/src/native/glow_integration.rs`): textures were
uploaded only inside `if is_visible { … }` (and after a viewport-missing
early-return), while `integration.update` still ran and grew the atlas.
The root window even starts hidden on purpose (`with_visible(false)`, to
avoid a startup white flash), so the very first frames hit this.
When the delta was dropped, the GPU font texture stayed smaller than the
CPU-side atlas; every glyph UV (normalized by the CPU atlas size) then
sampled the wrong rows until the next full atlas recreation. wgpu/Metal
can't detect this — the read is in-bounds, just the wrong row.
## Fixes
- **wgpu**: apply `textures_delta.set` right after `render_state` is
obtained, **before** any surface-dependent early-return. `free` still
runs after submit (unchanged).
- **glow**: apply `textures_delta.set` (and `free`) regardless of
`is_visible`, making the GL context current when there's anything to
upload; only tessellation/paint/swap stay gated on visibility.
- **debug assert** in `egui-wgpu`'s `Renderer::update_texture`: a full
delta must (re)create the GPU texture at exactly the delta size —
catches any future CPU/GPU size desync at the source.
## wgpu ruled out
Confirmed the desync is **not** inside wgpu: Metal `create_texture` uses
the exact descriptor size, and `queue.write_texture` validates against
the texture's own live `desc` — a single texture can't have CPU/GPU
sizes disagree. The mismatch is born at the egui boundary (atlas size
for UVs vs. last-applied upload), which wgpu cannot see.
## Testing note
A headless regression test of `paint_and_update_textures` isn't
practical (it needs a real winit window; `render_state` is private with
no surface-less setter). I verified the failure *mechanism* separately
on macOS/Metal (texture lagging the atlas → silent wrong-row sampling,
no wgpu error), but that demo did not exercise the fixed code path, so
it's not included. The fixes rest on the reasoning above.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
### Related
* Closes#8134.
* Related to #5136.
Possibly fixes:
* #8123
* #5145
### What
We did not properly handle the variants of
[`CurrentSurfaceTexture`](https://docs.rs/wgpu/latest/wgpu/enum.CurrentSurfaceTexture.html)
and always returned `SkipFrame`.
Because of this `egui` could end up in a state where frames are always
skipped after observing `Outdated`, without the chance to recover
(unless an event arrives from the outside).
> [!NOTE]
> This is not Wayland-specific, but could happen on all platforms. It
just happens frequently for Wayland compositors that directly resize a
window after creation (such as tiling/scrolling compositors like
`hyprland` and `niri`).
This PR improves this by separating the code paths for `Outdated` and
`Lost`, to help recover from those events.
Let apps change present_mode and desired_maximum_frame_latency at
runtime instead of only at startup.
API changes (egui-wgpu):
- New SurfaceConfig { present_mode, desired_maximum_frame_latency }.
- WgpuConfiguration now nests these as pub surface: SurfaceConfig (was
two top-level fields).
- RenderState gains pub surface_config: SurfaceConfig — the
currently-requested value.
API additions (eframe):
- Frame::wgpu_surface_config() / Frame::set_wgpu_surface_config(...) for
get/set.
- SurfaceConfig re-exported as eframe::SurfaceConfig.
How it works:
The wgpu painter compares render_state.surface_config to its
currently-applied values each paint. If they differ it updates its
config and flips
needs_reconfigure on every surface, piggybacking on the existing
deferred-reconfigure pathway.
Demo:
The backend panel (egui_demo_app) gets dropdowns for present mode and
desired max frame latency, wired through the new Frame accessors.
<img width="282" height="172" alt="image"
src="https://github.com/user-attachments/assets/0b1274b2-7e4e-4413-969b-0a014c415f79"
/>
* [x] I have followed the instructions in the PR template
This updates wgpu to v29 across the egui crate stack.
There a a few API changes due to the requirement to provide a display
handle up front to properly support GLES on linux. I have done my best
to make the api changes as reasonable as possible, but I don't have all
the greater project context, so lmk if things should be done a bit
differently.
I've also updated glow to 0.17 to make cargo deny happy, there are no
source changes. I'm not sure how you want to land these.
---------
Co-authored-by: lucasmerlin <hi@lucasmerlin.me>
Hey! I am not sure if this is something that's been considered before
and decided against (I couldn't find any PR's or issues).
This change removes the internal profiling macros in library crates and
the `puffin` feature and replaces it with similar functions in the
[profiling](https://github.com/aclysma/profiling) crate. This crate
provides a layer of abstraction over various profiler instrumentation
crates and allows library users to pick their favorite (supported)
profiler.
An additional benefit for puffin users is that dependencies of egui are
included in the instrumentation output too (mainly wgpu which uses the
profiling crate), so more details might be available when profiling.
A breaking change is that instead of using the `puffin` feature on egui,
users that want to profile the crate with puffin instead have to enable
the `profile-with-puffin` feature on the profiling crate. Similarly they
could instead choose to use `profile-with-tracy` etc.
I tried to add a 'tracy' feature to egui_demo_app in order to showcase ,
however the /scripts/check.sh currently breaks on mutually exclusive
features (which this introduces), so I decided against including it for
the initial PR. I'm happy to iterate more on this if there is interest
in taking this PR though.
Screenshot showing the additional info for wgpu now available when using
puffin
This implements web support for taking screenshots in an eframe app (and
adds a nice demo).
It also updates the native screenshot implementation to work with the
wgpu gl backend.
The wgpu implementation is quite different than the native one because
we can't block to wait for the screenshot result, so instead I use a
channel to pass the result to a future frame asynchronously.
* Closes <https://github.com/emilk/egui/issues/5425>
* [x] I have followed the instructions in the PR template
https://github.com/user-attachments/assets/67cad40b-0384-431d-96a3-075cc3cb98fb
When mixing and matching eframe with other wgpu applications
(https://github.com/tracel-ai/burn in my case), it can be helpful to use
an existing wgpu setup to initialize eframe with. This PR changes the
WpuConfiguration (in a non-backwards compat way :/), to either take some
options how to create a wgpu setup, or an existing wgpu setup
(consisting of an instance, adapter, device and queue).
* [x] I have followed the instructions in the PR template
---------
Co-authored-by: Andreas Reich <r_andreas2@web.de>
* Closes#5224
I'm unfamiliar with wgpu, so I'd like someone to confirm, that calling
`wgpu::Texture` _after_ `wgpu::Queue::submit` is in fact the right thing
to do.
---------
Co-authored-by: Andreas Reich <r_andreas2@web.de>
A very common usability issue on egui-wgpu callbacks is that `paint`
can't access any data that doesn't strictly outlive the callback
resources' data. E.g. if the callback resources have an `Arc` to some
resource manager, you can't easily pull out resources since you
statically needed to ensure that those resource references outlived the
renderpass, whose lifetime was only constrained to the callback
resources themselves.
Wgpu 22 no longer has this restriction! Its (render/compute-)passes take
care of the lifetime of any passed resource internally. The lifetime
constraint is _still_ opt-out since it protects from a common runtime
error of adding commands/passes on the parent encoder while a previously
created pass wasn't closed yet.
This is not a concern in egui-wgpu since the paint method where we have
to access the render pass doesn't even have access to the encoder!
This should help slightly with CPU/GPU parallelism when vsync is on.
I also return the time spent on vsync, which can help users figure out
how much CPU wall-time was used on non-vsync stuff
Setting `desired_maximum_frame_latency` to a low value should
theoretically lead to lower latency in winit apps using `egui-wgpu`
(e.g. in `eframe` with `wgpu` backend).
* Replaces https://github.com/emilk/egui/pull/3714
* See also https://github.com/gfx-rs/wgpu/pull/4899
----
It seems like `desired_maximum_frame_latency` has no effect on my Mac. I
lowered my monitor refresh-rate to 30Hz to test, and can see no
difference between `desired_maximum_frame_latency` of `0` or `3`.
Before when experimenting with changing the global `DESIRED_NUM_FRAMES`
in `wgpu` I saw a huge difference, so I wonder what has changed.
I verified that `set_maximum_drawable_count` is being called with either
`1` or `2`, but I perceive no difference between the two.
Introduced in the recent multi-viewports work, we accidentally recreated
the wgpu surfaces every frame. This is now fixed.
I found this while improving the profiling of `eframe`
* Closes#1044
---
(new PR description written by @emilk)
## Overview
This PR introduces the concept of `Viewports`, which on the native
eframe backend corresponds to native OS windows.
You can spawn a new viewport using `Context::show_viewport` and
`Cotext::show_viewport_immediate`.
These needs to be called every frame the viewport should be visible.
This is implemented by the native `eframe` backend, but not the web one.
## Viewport classes
The viewports form a tree of parent-child relationships.
There are different classes of viewports.
### Root vieport
The root viewport is the original viewport, and cannot be closed without
closing the application.
### Deferred viewports
These are created with `Context::show_viewport`.
Deferred viewports take a closure that is called by the integration at a
later time, perhaps multiple times.
Deferred viewports are repainted independenantly of the parent viewport.
This means communication with them need to done via channels, or
`Arc/Mutex`.
This is the most performant type of child viewport, though a bit more
cumbersome to work with compared to immediate viewports.
### Immediate viewports
These are created with `Context::show_viewport_immediate`.
Immediate viewports take a `FnOnce` closure, similar to other egui
functions, and is called immediately. This makes communication with them
much simpler than with deferred viewports, but this simplicity comes at
a cost: whenever tha parent viewports needs to be repainted, so will the
child viewport, and vice versa. This means that if you have `N`
viewports you are poentially doing `N` times as much CPU work. However,
if all your viewports are showing animations, and thus are repainting
constantly anyway, this doesn't matter.
In short: immediate viewports are simpler to use, but can waste a lot of
CPU time.
### Embedded viewports
These are not real, independenant viewports, but is a fallback mode for
when the integration does not support real viewports. In your callback
is called with `ViewportClass::Embedded` it means you need to create an
`egui::Window` to wrap your ui in, which will then be embedded in the
parent viewport, unable to escape it.
## Using the viewports
Only one viewport is active at any one time, identified wth
`Context::viewport_id`.
You can send commands to other viewports using
`Context::send_viewport_command_to`.
There is an example in
<https://github.com/emilk/egui/tree/master/examples/multiple_viewports/src/main.rs>.
## For integrations
There are several changes relevant to integrations.
* There is a [`crate::RawInput::viewport`] with information about the
current viewport.
* The repaint callback set by `Context::set_request_repaint_callback`
now points to which viewport should be repainted.
* `Context::run` now returns a list of viewports in `FullOutput` which
should result in their own independant windows
* There is a new `Context::set_immediate_viewport_renderer` for setting
up the immediate viewport integration
* If you support viewports, you need to call
`Context::set_embed_viewports(false)`, or all new viewports will be
embedded (the default behavior).
## Future work
* Make it easy to wrap child viewports in the same chrome as
`egui::Window`
* Automatically show embedded viewports using `egui::Window`
* Use the new `ViewportBuilder` in `eframe::NativeOptions`
* Automatically position new viewport windows (they currently cover each
other)
* Add a `Context` method for listing all existing viewports
Find more at https://github.com/emilk/egui/issues/3556
---
<details>
<summary>
Outdated PR description by @konkitoman
</summary>
## Inspiration
- Godot because the app always work desktop or single_window because of
embedding
- Dear ImGui viewport system
## What is a Viewport
A Viewport is a egui isolated component!
Can be used by the egui integration to create native windows!
When you create a Viewport is possible that the backend do not supports
that!
So you need to check if the Viewport was created or you are in the
normal egui context!
This is how you can do that:
```rust
if ctx.viewport_id() != ctx.parent_viewport_id() {
// In here you add the code for the viewport context, like
egui::CentralPanel::default().show(ctx, |ui|{
ui.label("This is in a native window!");
});
}else{
// In here you add the code for when viewport cannot be created!
// You cannot use CentralPanel in here because you will override the app CentralPanel
egui::Window::new("Virtual Viewport").show(ctx, |ui|{
ui.label("This is without a native window!\nThis is in a embedded viewport");
});
}
```
This PR do not support for drag and drop between Viewports!
After this PR is accepted i will begin work to intregrate the Viewport
system in `egui::Window`!
The `egui::Window` i want to behave the same on desktop and web
The `egui::Window` will be like Godot Window
## Changes and new
These are only public structs and functions!
<details>
<summary>
## New
</summary>
- `egui::ViewportId`
- `egui::ViewportBuilder`
This is like winit WindowBuilder
- `egui::ViewportCommand`
With this you can set any winit property on a viewport, when is a native
window!
- `egui::Context::new`
- `egui::Context::create_viewport`
- `egui::Context::create_viewport_sync`
- `egui::Context::viewport_id`
- `egui::Context::parent_viewport_id`
- `egui::Context::viewport_id_pair`
- `egui::Context::set_render_sync_callback`
- `egui::Context::is_desktop`
- `egui::Context::force_embedding`
- `egui::Context::set_force_embedding`
- `egui::Context::viewport_command`
- `egui::Context::send_viewport_command_to`
- `egui::Context::input_for`
- `egui::Context::input_mut_for`
- `egui::Context::frame_nr_for`
- `egui::Context::request_repaint_for`
- `egui::Context::request_repaint_after_for`
- `egui::Context::requested_repaint_last_frame`
- `egui::Context::requested_repaint_last_frame_for`
- `egui::Context::requested_repaint`
- `egui::Context::requested_repaint_for`
- `egui::Context::inner_rect`
- `egui::Context::outer_rect`
- `egui::InputState::inner_rect`
- `egui::InputState::outer_rect`
- `egui::WindowEvent`
</details>
<details>
<summary>
## Changes
</summary>
- `egui::Context::run`
Now needs the viewport that we want to render!
- `egui::Context::begin_frame`
Now needs the viewport that we want to render!
- `egui::Context::tessellate`
Now needs the viewport that we want to render!
- `egui::FullOutput`
```diff
- repaint_after
+ viewports
+ viewport_commands
```
- `egui::RawInput`
```diff
+ inner_rect
+ outer_rect
```
- `egui::Event`
```diff
+ WindowEvent
```
</details>
### Async Viewport
Async means that is independent from other viewports!
Is created by `egui::Context::create_viewport`
To be used you will need to wrap your state in `Arc<RwLock<T>>`
Look at viewports example to understand how to use it!
### Sync Viewport
Sync means that is dependent on his parent!
Is created by `egui::Context::create_viewport_sync`
This will pause the parent then render itself the resumes his parent!
#### ⚠️ This currently will make the fps/2 for every sync
viewport
### Common
#### ⚠️ Attention
You will need to do this when you render your content
```rust
ctx.create_viewport(ViewportBuilder::new("Simple Viewport"), | ctx | {
let content = |ui: &mut egui::Ui|{
ui.label("Content");
};
// This will make the content a popup if cannot create a native window
if ctx.viewport_id() != ctx.parent_viewport_id() {
egui::CentralPanel::default().show(ctx, content);
} else {
egui::Area::new("Simple Viewport").show(ctx, |ui| {
egui::Frame::popup(ui.style()).show(ui, content);
});
};
});
````
## What you need to know as egui user
### If you are using eframe
You don't need to change anything!
### If you have a manual implementation
Now `egui::run` or `egui::begin` and `egui::tessellate` will need the
current viewport id!
You cannot create a `ViewportId` only `ViewportId::MAIN`
If you make a single window app you will set the viewport id to be
`egui::ViewportId::MAIN` or see the `examples/pure_glow`
If you want to have multiples window support look at `crates/eframe`
glow or wgpu implementations!
## If you want to try this
- cargo run -p viewports
## This before was wanted to change
This will probably be in feature PR's
### egui::Window
To create a native window when embedded was set to false
You can try that in viewports example before:
[78a0ae8](78a0ae879e)
### egui popups, context_menu, tooltip
To be a native window
</details>
---------
Co-authored-by: Konkitoman <konkitoman@users.noreply.github.com>
Co-authored-by: Emil Ernerfeldt <emil.ernerfeldt@gmail.com>
Co-authored-by: Pablo Sichert <mail@pablosichert.com>
Tested on M1 Mac:
* native
* webgl, firefox
* webgpu, chrome
all looking normal
Updated minor ahash version because 0.8.1 got yanked. Added some deny
exceptions for now - we'll have to update winit soon to resolve glow
related cargo deny errors (not a big issue though since we don't expect
wgpu and glow backends to be used at the same time)
* Add puffin profile scopes to the startup and running of eframe
* puffin_profiler example: start puffin right away
* cargo format let-else statements
* More profile scopes
* Add some `#[inline]`
* Standardize puffin profile scope definitions
* standardize again
* Silence warning when puffin is disabled
* Triage for GL backend
* And cargo-fmt
* Changelog update with PR and issue
* Update crates/eframe/src/epi/mod.rs
Co-authored-by: Andreas Reich <r_andreas2@web.de>
* Update crates/egui-wgpu/src/winit.rs
Co-authored-by: Andreas Reich <r_andreas2@web.de>
* Add "supports_screenshot" to surface state
* Cranky fix
* fmt
---------
Co-authored-by: Andreas Reich <r_andreas2@web.de>
* Log warning instead of error when failing to decode RON in storage
* New web demo
* Clean up some warn/error logging
* Avoid deadlock that could happen on crash
* Log errors using console.warn, because console.error can cause crashes
* Use patched version of wasm-bindgen-cli, allowing >2GB memory
* New web demo
