This changes headless operation to strictly be a runtime option, rather
than a compile-time one. Note that the old headless version still relied
on a display server to support WebGL, while it now requires one all the
time.
Fixes#8573
Previously, the flow for ticking animations was:
Compositor -> Constellation -> Layout -> Script
However, this means that the compositor <-> layout messages can thrash, meaning layout thread is very rarely idle.
This means that the script thread (which joins on the layout thread during reflow) was unable to execute and run rAF callbacks.
With this change, the flow is now:
Compositor -> Constellation -> Script (when rAF is active).
Compositor -> Constellation -> Layout (when transitions / animations are active and no rAF is present).
This makes rAF based animation *much* smoother.
Instead, schedule a delayed composite after each frame of an animation.
The previous code would cause jank, because the following sequence
frequently occurred:
1. The page uses `requestAnimationFrame()` to request a frame.
2. The compositor receives the message, schedules a composite,
dispatches the rAF message to the script thread, composites, and goes to
sleep waiting for vblank (frame 1).
3. The script makes a change and sends it through the pipeline.
Eventually it gets painted and is sent to the compositor, but the
compositor is sleeping.
4. The compositor wakes up, sees the new painted content, page flips,
and goes to sleep (frame 2). Repeat from step 1.
The problem is that we have two composition frames, not just one. This
halves Web apps' framerate!
This commit fixes the problem by scheduling the composite in step 2 to
12 ms in the future. We already have this delayed-composition
functionality in the form of the scrolling timer, which I repurposed and
renamed to the "delayed composition timer" for this task. This change
gives the page 12 ms to prepare the frame, which seems to usually be
enough, especially with WebRender.
Note that simply removing the scheduled composite after rAF is not the
correct solution. If this is done, then pages that call rAF and don't
modify the page won't receive future rAFs, since the compositor will be
sleeping and won't be notified of vblank.
Fixes a bunch of jank in browser.html. The remaining jank seems to be a
problem with browser.html itself.
WebRender is an experimental GPU accelerated rendering backend for Servo.
The WebRender backend can be specified by running Servo with the -w option (otherwise the default rendering backend will be used).
WebRender has many bugs, and missing features - but it is usable to browse most websites - please report any WebRender specific rendering bugs you encounter!
Moved CompositorMsg enum into compositing crate.
moved CompositorMsg enum into compositing crate.
moved from components/msg/constellation_msg.rs
to components/compositing/lib.rs
fixes#8832
<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/servo/servo/8898)
<!-- Reviewable:end -->
To actually make the multiprocess communication work, we'll need to
reroute the task creation to the pipeline or the compositor. But this
works as a first step.
We currently store LayerBuffers, because previously NativeSurfaces did
not record their own size. Now we can store NativeSurfaces directly,
which saves a bit of space in the surface cache and allows us to create
LayerBuffers only in the PaintTask.
This also means that instead of sending cached LayerBuffers, the
compositor can just send cached NativeSurfaces to the PaintTask.
an IPC channel instead.
Because this used a boxed trait object to invoke messages across a
process boundary, and boxed trait objects are not supported across IPC,
we spawn a helper thread inside the compositor to perform the marshaling
for us.
Now that NativeDisplay can be shared between the compositor and the
paint task, we can move the LayerBuffer cache to the compositor. This
allows surfaces to be potentially reused between different paint tasks
and will eventually allow OpenGL contexts to be preserved between
instances of GL rasterization.
I've done a bit of job to get this done. Right now readback is still used, but we have a `LayerId` -> `CanvasRenderer` map on the paint task, that we can use to get rid of that.
I'd want review, to see if this is a good approach (I know it's not the initial `CanvasId` -> renderer approach, but it's pretty similar, since a canvas involves a `PaintLayer`).
I had to do a bit of refactoring to avoid cyclic dependencies between canvas and gfx. I'd want you to review them too.
It's mergeable and doesn't break any tests :P
Some of my main concerns:
* Does the canvas render really need to be behind an `Arc<Mutex<T>>`?
* I can't clone a `NativeSurface` right now (that's why the `SendNativeSurface()` msg is unimplemented in the WebGL task). It should be easy to add that to rust-layers, supposing the caller is responsible to mark it as non-leaking, any reason to not do it?
cc @jdm @pcwalton
<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/servo/servo/6083)
<!-- Reviewable:end -->