commit a96c12ebc2df6ca08a8c94f847d4b78d20ed6117
author David Baron <dbaron@chromium.org> Tue Mar 22 21:11:21 2022
committer Blink WPT Bot <blink-w3c-test-autoroller@chromium.org> Tue Mar 22 21:41:14 2022
tree f7a96a9020e6b4ddd5f7521671dd2758383816b8
parent 9dd86c4ac1a7262723bbb9ed56b194bb85a511a6

Revert "Create property tree nodes for will-change only when relevant property has will-change."

This reverts commit 5d3dd81ee4f41ffc7fd40fda745cb097cd50e9f5.

Reason for revert: failure of fast/sub-pixel/transformed-iframe-copy-on-scroll.html on mac-mac11-arm64.

Original change's description:
> Create property tree nodes for will-change only when relevant property has will-change.
>
> Prior to this change, will-change: transform, opacity, filter, or
> backdrop-filter would cause the creation of property tree nodes (in both
> the paint property tree and the cc property tree) for all of them
> (though only a single effect property node for both opacity and
> backdrop-filter).
>
> With this change, will-change: transform only causes the creation of a
> transform node (in the transform tree), will-change: filter only causes
> the creation of a filter node (in the effect tree), and will-change:
> opacity or backdrop-filter only causes the creation of an effect node
> (in the effect tree).  However, *if* the nodes are created by something
> else, the presence of a different will-change still causes any node that
> is created in the paint property tree to be composited (i.e., created in
> the cc property tree).
>
> This provides the basis on which we can separate the transform node into
> multiple transform nodes, for efficient implementation and animation of
> individual transform properties, without causing a performance
> regression.
>
> Earlier versions of this CL (prior to the change that kept nodes being
> composited for an unrelated will-change) caused the test
> external/wpt/css/filter-effects/effect-reference-feimage-002.html to
> start failing (in the same way as -001 and -003) because the filter
> effect is no longer composited.  It creates a copy of the test as -004
> with will-change: filter rather than will-change: transform that
> continues to have a composited filter effect node and thus continues
> passing even with that earlier change.  This test still seems worth
> adding despite no longer strictly being needed.
>
> Bug: 900241
> Change-Id: I18efc630668ea7eac2a8c0a0417b766762e71d9f
> Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/3328730
> Reviewed-by: Xianzhu Wang <wangxianzhu@chromium.org>
> Commit-Queue: David Baron <dbaron@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#983821}

Bug: 900241
Change-Id: I8143ff91fa0d856ab189414e44f12ce1bcdf77e8
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/3544106
Auto-Submit: David Baron <dbaron@chromium.org>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Owners-Override: David Baron <dbaron@chromium.org>
Commit-Queue: David Baron <dbaron@chromium.org>
Cr-Commit-Position: refs/heads/main@{#984017}