commit 352fbdae1017811d80b9d0d10a28464e24fae395
author Jean-Philippe Gravel <jpgravel@chromium.org> Fri Apr 04 20:03:16 2025
committer Chromium LUCI CQ <chromium-scoped@luci-project-accounts.iam.gserviceaccount.com> Fri Apr 04 20:03:16 2025
tree 1ab481dde5775a1970b22fb8f62fa83d332891da
parent bcb71ceef7d8202dd2a0af171496e52ac90c67f8

Remove cc_layer_ check in IsResourceValid

This check was introduced by https://codereview.chromium.org/1372463002,
back in 2015. This is when we first started tuning if and when the
canvas should be GPU accelerated. Before this CL, Canvas2DLayerBridge
was always accelerated and always had a valid m_layer attribute. After
this CL, m_layer was created only for GPU accelerated canvases.

Because Canvas2DLayerBridge could now use software mode, the logic had
to be adjusted to skip anything specific to GPU acceleration. In
particular, the logic where we would detect a new GPU context loss
and clear the canvas had to be bypassed because software resources
shouldn't be impacted by a GPU process loss. The CL did this by using
m_layer as a proxy to detect whether GPU acceleration is used (see
line 363 in [1]) and early abort if it isn't. The CL uses this
interpretation of m_layer in other places too, for instance in
isAccelerated (line 165) and getOrCreateSurface (line 179).

The purpose of that CL was to postpone the decision to use GPU
acceleration to when the first flush of the canvas occurs because some
flush operations are better done in CPU (e.g. read backs). Thus, until
the first flush, the bridge had no surface, no layer, and didn't know
whether it was accelerated. For better of for worse, checkSurfaceValid
returned true in this state.

Fast forward to 2025, and the cc_layer_ check no longer serves its
purpose. We no longer need it to skip the GPU context lost detection
and canvas reset logic because it was moved out of IsResourceValid.
The check could even be harmful, because now, the cc_layer_ is no
longer created along with the surface. Thus, it means that if the
canvas suffered from a context lost before a frame is presented to
the compositor, cc_layer_ would still be nullptr and IsResourceValid
would incorrectly return true.

[1]: https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/platform/graphics/canvas_2d_layer_bridge.cc;dlf=third_party%2FWebKit%2FSource%2Fplatform%2Fgraphics%2FCanvas2DLayerBridge.cpp;dlc=e8cac159f27bb9a24eca106cdea07abdf421f29c;drf=third_party%2FWebKit%2FSource%2Fplatform%2Fgraphics%2FCanvas2DLayerBridge.cpp;drc=8046330d6021c156f59a49729ed2308eda912add;bpv=0;bpt=0

Bug: 346766781
Change-Id: Ie84011eaf3e22dd083350436056199faae988873
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/6426512
Reviewed-by: Colin Blundell <blundell@chromium.org>
Commit-Queue: Jean-Philippe Gravel <jpgravel@chromium.org>
Cr-Commit-Position: refs/heads/main@{#1442863}