2029043 - Use PureRelative scroll position update in more situations to fix bug 2009679

Status: NEW

(Core :: Panning and Zooming, task, P3)

Description

[Hiroyuki Ikezoe (:hiro)]

This is an alternative approach to fix bug 2009679. This approach should be independent from bug 1946610 (scroll position rounding/pixel alignment), of course even with this approach there may be appearing scroll jittering due to scroll position rounding, but the behavior of bug 2009679 will be significantly improved, I would say scroll stuck will be resolved.

This bug mainly consists of three parts; 1) bug 1740164, 2) a subset of 1662022, i.e. changing this NewRelativeScroll to NewPureRelativeScroll, 3) ScrollOrigin::Clamp becomes PureRelative rather than Absolute.

Bug 2029033 might also be mandatory in some edge cases to make this bug more robust, such as multiple PureRelative happens at once while paint skipping.

Comment 1

[Hiroyuki Ikezoe (:hiro)]

It turned out that bug 1740164 is unnecessary since bug 2009679's case is actually involving a smooth animation being run by APZ.

Comment 2

[Hiroyuki Ikezoe (:hiro)]

There are two small but important differences between FrameMetrics::ApplyPureRelativeScrollUpdateFrom and FrameMetrics::ApplyRelativeScrollUpdateFrom;

1. ApplyRelativeScrollUpdateFrom uses the update source which was introduced in bug 1978682
2. ApplyRelativeScrollUpdateFrom doesn't clamp the destination, which was introduced in bug 1871760

For 1) after-scrollable-range-shrinkage-001.html and other after-scrollable-range-shrinkage-00X.html can be test cases for pure relative scroll position update once after we migrated this NewRelativeScroll to NewPureRelativeScroll.

But for 2) I haven't succeeded in writing automated tests, bug 1978682 landed a gtest but modifying the test doesn't catch the bug in pure relative cases.

Meanwhile I realized that

(In reply to Hiroyuki Ikezoe (:hiro) from comment #0)

3. ScrollOrigin::Clamp becomes PureRelative rather than Absolute.

Just changing ScrollOrigin::Clamp from ScrollUpdateType::Absolute to ScrollUpdateType::Relative fixes bug 2009679. I am going to file a new bug for the change.

Comment 3

[Hiroyuki Ikezoe (:hiro)]

Filed bug 2029551. This bug is no longer blocking bug 2009679.

Comment 4

[Hiroyuki Ikezoe (:hiro)]

Attachment: WIP: Bug 2029043 - Skip clamping the scroll offset while applying a pure relative ScrollPositionUpdate, as done for relative ScrollPositionUpdates in bug 1871760.

As of now clamping while applying the update is not a problem, but it
will be a problem when we merge pure relative into relative one.

There are some test cases in wpts which will be affected by this change
when we have merged those scroll position updates.

testing/web-platform/tests/css/css-scroll-anchoring/after-scrollable-range-shrinkage-00[1-4].html

Comment 5

[Hiroyuki Ikezoe (:hiro)]

Attachment: WIP: Bug 2029043 - Use ScrollPositionUpdate::mSource as the scroll origin of pure relative scroll position updates if the APZC is default, as done for relative/instant ScrollPositionUpdates in bug 1978682.

Comment 6

[Hiroyuki Ikezoe (:hiro)]

Attachment: WIP: Bug 2029043 - Handle pure relative scroll position updates in FrameMetrics::UpdatePendingScrollInfo.

Comment 7

[Hiroyuki Ikezoe (:hiro)]

Attachment: WIP: Bug 2029043 - Use pure relative scroll position update for ScrollOrigin::Clamp.

Comment 8

[Hiroyuki Ikezoe (:hiro)]

Attachment: WIP: Bug 2029043 - Use pure relative scroll position update for non zero delta ScrollOrigin::Relative.

Comment 9

[Hiroyuki Ikezoe (:hiro)]

For references in future I uploaded the changes for this bug.