Write a mochitest for bug 1647156
Categories
(Core :: Panning and Zooming, task)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox80 | --- | fixed |
People
(Reporter: kats, Assigned: kats)
References
Details
Attachments
(1 file)
Nical wrote a patch that caused a perma-checkerboarding regression. He fixed it in bug 1647156 but we should add a test.
|
Assignee | |
Comment 1•5 years ago
|
||
The relevant points in the scenario that was perma-checkerboarding were:
- there was a scrollframe inside a svg filter, so APZ got a scrollinfo layer for it
- during a repaint request, APZ computed zero margins for it, because scrollinfo
- The main-thread code to apply the margins and alignment and come up with a final displayport had a posAlignment > sizeAlignment. This meant that even though the size was aligned up to the next 128, the position was aligned down to previous 512, so the "increase" due to size alignment wasn't enough to completely cover the "decrease" due to the pos alignment. With zero margins this meant part of the scroller was left uncovered by the displayport.
I constructed a test that reproduces the problem for me locally, here is a try push (with nical's patch reverted) to verify it fails on try too: https://treeherder.mozilla.org/#/jobs?repo=try&revision=37c39626aaeba0d5857d4615ea95ba2ba53888dc
|
|
Assignee | |
Comment 2•5 years ago
|
||
(Note that the test only fails with --enable-webrender because that's the branch on which the posAlignment/sizeAlignment stuff was split).
|
|
Assignee | |
Comment 3•5 years ago
|
||
Ok, so the test fails for me locally because I'm on a retina mac and the devicePixelRatio is 2.0. On my Linux machine that's not the case and so the test passes because the numbers come out differently. In order to reproduce the failure on Linux I basically need to 2x the height of the scrollframe and scrolled content in the test, but then it's also taller than the browser window and the checkerboarding part is not actually user-visible.
Also on Linux the scrollbars take up layout space and so I need to adjust the numbers a bit more to account for that.
|
|
Assignee | |
Comment 4•5 years ago
|
||
With fix reverted: https://treeherder.mozilla.org/#/jobs?repo=try&group_state=expanded&revision=571325faba86f392dd1933a930295ece5c030fa1 shows test failing
With fix applied: https://treeherder.mozilla.org/#/jobs?repo=try&group_state=expanded&revision=585a99a48b4531439f3e7f766f4cf0abfe1538e3 shows test passing
|
|
Assignee | |
Comment 5•5 years ago
|
||
|
||
Comment 7•5 years ago
|
||
| bugherder | ||
Description
•