This is part of the future work of implementing LazyDOMString as
outlined in issue #39479.
We use str() method or direct implementations on DOMString for these
methods. We also change some types.
This is independent of https://github.com/servo/servo/pull/39480
Signed-off-by: Narfinger Narfinger@users.noreply.github.com
Testing: This is essentially just renaming a method and a type and
should not change functionality.
Signed-off-by: Narfinger <Narfinger@users.noreply.github.com>
Vello has updated to wgpu v26 recently. It might be a good time for
servo to update as well. This PR should wait for #39015 and #38717
Testing: WebGPU CTS
Fixes: None
---------
Signed-off-by: Wu Yu Wei <yuweiwu@pm.me>
There are many important changes here:
- Generalize the presentation buffer into standalone staging buffers
that hold their own state. This allow them to be used by getImage.
- Move all clear handling to the ScriptThread and send the configuration
on each request present/getimage, thus avoiding any recreate/clearing
messages. This means that we prepare staging buffers lazily, on the
first request.
Try run for this change:
https://github.com/sagudev/servo/actions/runs/17341982368
Testing: This is covered by existing WebGPU CTS tests. There are some
bad expectations updates, but they are also on main (presumably from
last update the rendering work) although I think CTS is actually wrong
(see https://github.com/gpuweb/cts/issues/4440).
Fixes: #36820Fixes: #37705Fixes: #33368 (we now keep reference alive in hashmap)
---------
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
Any RenderingContext/OffscreenRenderingContext type has readonly
"canvas" attribute
and associated native-code DOM context objects have reference to target
DOM canvas objects.
https://html.spec.whatwg.org/multipage/canvas.html#renderingcontexthttps://html.spec.whatwg.org/multipage/canvas.html#offscreenrenderingcontext
And currently the reference to DOM canvas object is the rooting pointer
on the stack,
which leads to the circular reference problem.
The SpiderMonkey's (SM) garbage collector will not be able to free the
DOM canvas and context
objects (unreacheble from JS) because of the rooting pointer on stack
(see STACK_ROOTS).
And these objects will be stored until the associated script
runtime/thread will be terminated.
SM -> JS Roots -> DOM Canvas* (on heap) -> DOM Context (on heap)
SM -> Rust Roots -> Dom Canvas* (on stack) <- as "canvas" member field
Let's replace the rooting pointer to the traceble pointer (DomRoot ->
Dom)
in the "canvas" member field of DOM context object, which allows to
broke circular referencing problem.
Testing: No changes in existed tests
Signed-off-by: Andrei Volykhin <volykhin.andrei@huawei.com>
Co-authored-by: Andrei Volykhin <volykhin.andrei@huawei.com>
According to the WebGPU specification there are the dedicated task
source
which is used to queue a global task for a GPUDevice on content
timeline.
https://gpuweb.github.io/gpuweb/#-webgpu-task-source
Tasks on content timeline:
- to fire "uncaptureevent" event
- to resolve GPUDevice.lost promise
Also fixed the "isTrusted" attribute status (false -> true) of the
"uncaptureevent" event by using non JS version of event dispatching.
Testing: No changes in WebGPU CTS expectations
- webgpu:api,operation,uncapturederror:*
- webgpu:api,operation,device,lost:*
- webgpu:api,validation,state,device_lost,destroy:*
Signed-off-by: Andrei Volykhin <volykhin.andrei@huawei.com>
Co-authored-by: Andrei Volykhin <volykhin.andrei@huawei.com>
Adds epoch to each WR image op command that is sent to compositor. The
renderer now has a `FrameDelayer` data structure that is responsible for
tracking when a frame is ready to be displayed. When asking canvases to
update their rendering, they are given an optional `Epoch` which denotes
the `Document`'s canvas epoch. When all image updates for that `Epoch`
are seen in the renderer, the frame can be displayed.
Testing: Existing WPT tests
Fixes: #35733
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Co-authored-by: Martin Robinson <mrobinson@igalia.com>
This shows up sometimes in code reviews, so it makes sense that tidy
enforces it. `rustfmt` supports this via comment normalization, but it
does many other things and is still an unstable feature (with bugs).
Testing: There are new tidy tests for this change.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Follow the HTML speficication and add missing 'transferToImageBitmap'
method to OffscreenCanvas interface.
https://html.spec.whatwg.org/multipage/#dom-offscreencanvas-transfertoimagebitmap
Testing: Improvements in the following tests
- html/canvas/offscreen/compositing/2d.composite.grid*
- html/canvas/offscreen/fill-and-stroke-styles/2d.gradient*
- html/canvas/offscreen/manual/the-offscreen-canvas/offscreencanvas*
- html/canvas/offscreen/reset/2d.reset*
- html/canvas/offscreen/text/2d.text*
Fixes (partially): #34111
Signed-off-by: Andrei Volykhin <andrei.volykhin@gmail.com>
1. The `shared` directory is for the "_traits" crates, which will likely
be moved out of this directly at some point and renamed "_api". These
crates expose the API of crates to avoid circular dependencies.
`Snapshot` isn't really this.
2. `Snapshot` is essentially a specialied kind of `Image` so it makes
sense that it is grouped with other image-related things in `pixels`.
Testing: This should not change any behavior so is covered by existing
tests.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Replaces uses of `euclid::default::Size2D<u64>` with
`euclid::default::Size2D<u32>` for the canvas/context/snapshot.
This PR includes changes to the following components:
- canvas
- pixels
- script
- script_bindings
- shared/canvas
- shared/snapshot
Testing: https://github.com/hashcatHitman/servo/actions/runs/15426115391
(as of 892edc0048)
Fixes: #36706
---------
Signed-off-by: hashcatHitman <155700084+hashcatHitman@users.noreply.github.com>
This follows firefox's implementation:
24d49101ce/dom/webgpu/Instance.h (L68)
It changes the default on most systems to `Bgra8Unorm` but leaves it as
`Rgba8Unorm` on android.
Signed-off-by: Ashwin Naren <arihant2math@gmail.com>
All canvases return `Option<ImageKey>`.
Testing: Just refactor without behavior changes
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
Each canvas context returns snapshot instead of just raw bytes. This
allows as to hold off conversions (BGRA <-> RGBA, (un)premultiply) to
when/if they are actually needed. For example when loading snapshot into
webgl we can load both RGBA and BGRA so no conversion is really needed.
Currently whole thing is designed to be able to be extend on
https://github.com/servo/ipc-channel/pull/356, to make less copies.
Hence some commented out code.
Fixes#35759
There are tests for these changes in WPT
---------
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
Updates wgpu to v25 and remove some verbose logging from CTS (that also
causes OOM).
Testing: WebGPU CTS
---------
Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>
This is the last big change necessary to create the
`constellation_traits` crate. This moves the data structure for messages
that originate from the `ScriptThread` and are sent to the
`Contellation` to `constellation_traits`, effectively splitting
`script_traits` in half. Before, `script_traits` was responsible for
exposing the API of both the `ScriptThread` and the `Constellation` to
the rest of Servo.
- Data structures that are used by `ScriptToConstellationMsg` are moved
to `constellation_traits`. The dependency graph looks a bit like this:
`script_layout_interface` depends on `script_traits` depends on
`constellation_traits` depends on `embedder_traits`.
- Data structures that are used in the embedding layer
(`UntrustedNodeAddress`, `CompositorHitTestResult`, `TouchEventResult`
and `AnimationState`) are moved to embedder_traits, to avoid a
dependency cycle between `webrender_traits` and
`constellation_traits`.
- Types dealing with MessagePorts and serialization are moved to
`constellation_traits::message_port`.
Testing: This is covered by existing tests as it just moves types
around.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Messages that are sent to the `Constellation` have pretty ambiguous
names.
This change does two renames:
- `ConstellationMsg` → `EmbedderToConstellationMessage`
- `ScriptMsg` → `ScriptToConstellationMessage`
This naming reflects that the `Constellation` stands in between the
embedding layer and the script layer and can receive messages from both.
Soon both of these message types will live in `constellation_traits`,
reflecting the idea that the `_traits` variant for a crate is
responsible for exposing the API for that crate.
Testing: No new tests are necessary here as this just renames two enums.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
After #36320, `script` only depends on `webgpu` for its type aliases to
`wgpu_core` and `wgpu_types`. This change removes the dependency on
`webgpu` in favor of depending directly on the upstream crates. This
makes it so that you can change `webgpu` without recompiling `script`.
Testing: This change is covered by existing WebGPU tests and is mainly
just changing the way dependencies are accessed.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This breaks the `script_traits` dependency on `webgpu`. In general, the
`traits` crates shouldn't depend on Servo non-`traits` crates. This is
necessary to move "script to constellation" messages to the
`constellation_traits` crate, making it the entire API for talking to
the
constellation. This will break a circular dependency when that happens.
Testing: Successfully building is enough of a test for this one as
it is mainly moving types around.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Using the RoutedPromiseListener let us define a different
response type for each promise. This removes unreachable branches
that used to exist when they all shared the same WebGPUResponse.
Signed-off-by: webbeef <me@webbeef.org>
* Migrate to 2024 edition
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Allow unsafe_op_in_unsafe_fn lint
This lint warns by default in the 2024
edition, but is *way* too noisy for servo.
We might enable it in the future, but not now.
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Compile using the 2024 edition
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
---------
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Use 2024 style edition
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
* Reformat all code
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
---------
Signed-off-by: Simon Wülker <simon.wuelker@arcor.de>
No longer hide errors while queueing tasks on the main thread. This
requires creating two types of `TaskSource`s: one for the main thread
and one that can be sent to other threads. This makes queueing a bit
more efficient on the main thread and more importantly, no longer hides
task queue errors.
Fixes#25688.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Co-authored-by: Mukilan Thiyagarajan <mukilan@igalia.com>
This is a simplification of the internal `TaskQueue` API that moves the
`TaskManager` to the `GlobalScope` itself. In addition, the handling of
cancellers is moved to the `TaskManager` as well. This means that no
arguments other than the `task` are necessary for queueing tasks, which
makes the API a lot easier to use and cleaner.
`TaskSource` now also keeps a copy of the canceller with it, so that
they always know the proper way to cancel any tasks queued on them.
There is one complication here. The event loop `sender` for dedicated
workers is constantly changing as it is set to `None` when not handling
messages. This is because this sender keeps a handle to the main
thread's `Worker` object, preventing garbage collection while any
messages are still in flight or being handled. This change allows
setting the `sender` on the `TaskManager` to `None` to allow proper
garbabge collection.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
This puts a few commonly used `Node` helpers into a trait (`NodeTraits`)
and gives them more descriptive names and documentation. The renames:
- `document_from_node` -> `NodeTraits::owner_document`
- `window_from_node` -> `NodeTraits::owner_window`
- `stylesheets_owner_from_node<T:` -> `NodeTraits::stylesheet_list_owner`
- `containing_shadow_root` -> `NodeTraits::containing_shadow_root`
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
Instead of creating a type for each `TaskSource` variety have each `TaskSource`
hold the same kind of sender (this was inconsistent before, but each
sender was effectively the same trait object), a pipeline, and a
`TaskSourceName`. This elminates the need to reimplement the same
queuing code for every task source.
In addition, have workers hold their own `TaskManager`. This allows just
exposing the manager on the `GlobalScope`. Currently the `TaskCanceller`
is different, but this will also be eliminated in a followup change.
This is a the first step toward having a shared set of `Sender`s on
`GlobalScope`.
Signed-off-by: Martin Robinson <mrobinson@igalia.com>
