1059498 - Scrolling filtered stuff into view that extends past the unfiltered bounds doesn't paint properly

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Markus Stange [:mstange]]

Attachment: testcase

I thought this used to work, so this may be a recent regression.

STR:
 1. Load the testcase.
 2. Scroll down.
 3. Slowly scroll back up.

The blue shadow should be painted completely as it gets visible. Instead, it's only painted when you've scrolled up far enough that parts of the green square are also visible.

Comment 1

[Markus Stange [:mstange]]

This regressed between July 18th and July 20th, so it's probably due to bug 1022612.

Comment 2

[Markus Stange [:mstange]]

Attachment: v1

When building the display list for the filter's stacking context, we were initializing the display items with the post-filter dirty rect instead of the pre-filter dirty rect because the call to nsSVGIntegrationUtils::GetRequiredSourceForInvalidArea was happening after we already passed the post-filter dirtyRect to the AutoBuildingDisplayList thing. So the filtered background color item's mVisibleRect, which is initialized using aBuilder->GetDirtyRect(), ended up outside the item's bounds, which meant that it didn't contribute any visible area to the basic thebes layer that's built for the filter's inactive layer manager later.

Comment 3

[Markus Stange [:mstange]]

https://tbpl.mozilla.org/?tree=Try&rev=197a555fccb5

Comment 4

[Markus Stange [:mstange]]

The reftest failures in that try push are from a different patch (bug 1009786).

Comment 5

[Markus Stange [:mstange]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/39a8d9b2b639

Comment 6

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/39a8d9b2b639

Comment 7

[Markus Stange [:mstange]]

[Tracking Requested - why for this release]: regression, caused by bug 1022612

Comment 8

[Markus Stange [:mstange]]

Comment on attachment 8482349 [details] [diff] [review]
v1

Approval Request Comment
[Feature/regressing bug #]: caused by bug 1022612
[User impact if declined]: incomplete drawing on pages that use filters
[Describe test coverage new/current, TBPL]: this patch adds a test, green on m-c
[Risks and why]: low to moderate, test coverage is fairly good in this area
[String/UUID change made/needed]: none

Comment 9

[Daniel Holbert [:dholbert] (AFK from Oct 16-23)]

(This seems to have caused bug 1059498; we probably should hold off on backporting until we have a fix over there.)

Comment 10

[Daniel Holbert [:dholbert] (AFK from Oct 16-23)]

(er, "caused bug 1062870")

Comment 11

[Markus Stange [:mstange]]

Comment on attachment 8482349 [details] [diff] [review]
v1

Agreed.

Comment 12

[Markus Stange [:mstange]]

Attachment: aurora patch

This patch includes the regression fix from bug 1062870.

Approval Request Comment
[Feature/regressing bug #]: caused by bug 1022612
[User impact if declined]: incomplete drawing on pages that use filters
[Describe test coverage new/current, TBPL]: this patch adds a test, green on m-c
[Risks and why]: low to moderate
[String/UUID change made/needed]: none

Comment 13

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8488737 [details] [diff] [review]
aurora patch

Aurora+

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/abd84750b5e6