Prerolling a layer can have side effects. In particular, PlatformViewLayer::Preroll
will call view_embedder->PrerollCompositeEmbeddedView.
Clip layers will check whether the layer's children are all clipped and if so
will skip calling Preroll on the children. However, the Paint implementation in
these layers was always calling Paint on their children.
This could result in a call to PlatformViewLayer::Paint without a corresponding
call to PlatformViewLayer::Preroll. This translates to a CompositeEmbeddedView
call without a PrerollCompositeEmbeddedView call on the affected view_id.
The EmbedderExternalViewEmbedder implementation does not allow that.
With this change, clip layers will only call PaintChildren if the preroll
called PrerollChildren.
See https://github.com/flutter/flutter/issues/46111
* Revert "Add flow test fixtures and tests (#13986)"
This reverts commit 620f5281b8.
* Revert "Dynamically determine whether to use offscreen surface based on need (#13976)"
This reverts commit a86ef94656.
This adds more trace events to more layer operations and enhances the
trace counters for the Fuchsia vulkan surface pool to include retained
surface counts, emit stats on recycle events that might change the
surface count, and by separating counters which measure bytes from
counters which measure counts to make analysis simpler.
This is a duplicate of flutter/engine#13360 with the test switched to use the software backend instead of the GL backend.
After some debugging and testing on another GL embedder I think the issue with the test is some bug having to do with the GL implementation in the test harness specifically.
Fixesflutter/flutter#38903
Previously the cache was disabled on whether or not PlatformViews were
globally enabled. Instead track their existence in the view hierarchy
and only disable RasterCache if a PlatformView is actually present.
This used to only be handled correctly for non-root layer backing stores. This
was mostly a side effect of the fact that we used recording canvases instead of
rendering directly into the backing store. We now use recording canvases
consistently.
Fixes b/143464703
Fixes https://github.com/flutter/flutter/issues/43732
SkiaUnrefQueue should be empty at destruction time. If the queue is nonempty,
then there will be a pending drain task that will hold a reference to the
queue. The queue can only be destructed after the drain completes and the
reference is dropped.
Drains must only be done on the queue's task runner thread, which may not be
the thread where the queue is destructed.
The contents rendered into the backing stores are already correctly scaled.
The initial implementation assumed this also held true for the metrics obtained
via embedded view parameters.
Fixes b/142699417
Jim Graham
liyuqian
Jason Simmons
David Worsham
Chinmay Garde
Kaushik Iska
gaaclarke
James Robinson
Michael Klimushyn
Dan Field
Chinmay Garde