909128 - Thunderbird - Per-pixel scroll views don't redraw correctly.

Status: RESOLVED FIXED

(Core :: Web Painting, defect)

Description

[Josiah Bruner [:jsbruner]]

Can't be positive this is a layout issue, but starting here. Could be a widget->Cocoa bug, but unlikely.

Thunderbird Daily 26.0a1 (2013-08-25), now has major issues when scrolling. Parts of messages move, others don't. However, this definitely looks like an issue with redrawing.

My best guess is that something caused the redraw to not be called often enough, but what and why is unknown to me and beyond my skill-set.

I'll upload a video of the glitch shortly.

Comment 1

[Josiah Bruner [:jsbruner]]

Attachment: RenderingGlitch.mov

Screencast of the bug.

Comment 2

[Josiah Bruner [:jsbruner]]

Hmm, loading the video may prove difficult. So adding a dropbox link as well. See URL.

Comment 3

[Josiah Bruner [:jsbruner]]

Also note that reaching the top or bottom of a view correctly redraws everything.

Comment 4

[Josiah Bruner [:jsbruner]]

The current UX Nighty's add-on manager looks like it has a similar issue. FYI.

Getting a regression window now...

Comment 6

[Josiah Bruner [:jsbruner]]

Last good: Daily 26.0a1 (2013-08-23)
First bad: Daily 26.0a1 (2013-08-24)

Changesets in between the 23rd and the 24th:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=fb2318875cd4&tochange=01576441bdc6

Also according to the duped bug this happens on Windows as well. Save to say this is an all platforms issue then.

Comment 7

[Josiah Bruner [:jsbruner]]

Actually the other bug probably shouldn't have been duped. The issues are almost certainly related but not necessarily the same.

So keeping this bug specifically about the scrolling issues on OS X...

Comment 8

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Matt, any insight on the regression range here?  (You landed a bunch of things in that range, though none pops out as obviously related.)

Comment 9

[Josiah Bruner [:jsbruner]]

FYI, toggling layers.acceleration.disabled fixes things. I'm wondering if bug 909128 is the cause...

Comment 10

[Josiah Bruner [:jsbruner]]

Umm... Bug 907926 I mean.

Comment 11

[Matt Woodrow (:mattwoodrow)]

Attachment: Fix CopySurface

Yes, definitely a regression from bug 907926.

Comment 12

[helkav]

(In reply to Josiah Bruner [:JosiahOne] from comment #9)
> FYI, toggling layers.acceleration.disabled fixes things. I'm wondering if
> bug 909128 is the cause...

I tried toggling this as well : switch from true to false. It seems to resolve the rendering issues

I have no idea what this means/does . what are the disadvantages (if any of keeping it set to false?)

Comment 13

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 795216 [details] [diff] [review]
Fix CopySurface

Review of attachment 795216 [details] [diff] [review]:
-----------------------------------------------------------------

A unit test would be nice

Comment 14

[Matt Woodrow (:mattwoodrow)]

Looks like Moz2D has a unit test for this, plus we have reftests that hit this path.

We just don't run BasicLayers reftests for mac, or any of the Moz2D tests.

https://hg.mozilla.org/integration/mozilla-inbound/rev/6efe5d76831e

Comment 15

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/6efe5d76831e