Closed Bug 1252266 Opened 4 years ago Closed 4 years ago

[e10s] Various APZ reftests fail on OSX

Categories

(Core :: Panning and Zooming, defect)

Unspecified
macOS
defect
Not set

Tracking

()

RESOLVED FIXED
Tracking Status
e10s + ---
firefox47 --- fixed

People

(Reporter: RyanVM, Assigned: RyanVM)

References

(Blocks 1 open bug)

Details

(Whiteboard: [gfx-noted])

Looks like some fail consistently across the board, a couple fail intermittently (like position-fixed-in-scroll-container.html), and some only fail on 10.6. Looking at the manifest, I suspect some of these tests had issues on Windows as well and got disabled for similar reasons.

10.6:  https://treeherder.mozilla.org/logviewer.html#?job_id=17373159&repo=try
10.10: https://treeherder.mozilla.org/logviewer.html#?job_id=17372405&repo=try
All the 10.10 failures look like they just need to be fuzzed. On 10.6 it looks like the scrollbar positions aren't getting updated, possibly because we're assuming overlay scrollbars somewhere for OS X. Not really sure, I'll dig into that. I have a 10.6 machine that I can use to try and repro locally, maybe.
Also seeing "layout/reftests/invalidation/scroll-inactive-layers.html | failed reftest-no-paint" on OSX 10.10 only.
Keywords: leave-open
Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/11fa0cbdf5a7 because your reordering caused bg-fixed-child-mask.html and group-opacity-surface-size-1.html to start failing on Linux and Win8, at least, so far.
http://mxr.mozilla.org/mozilla-central/source/layout/tools/reftest/README.txt#55 being the oddly-placed explanation that last-matching wins, so if you want to actually have a skip-if work, you have to put it after everything else, not before.
(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #1)
> On 10.6 it
> looks like the scrollbar positions aren't getting updated, possibly because
> we're assuming overlay scrollbars somewhere for OS X.

I don't think this is right. I did a new try push with more logging [1] and everything looks normal in the layer tree. The scrollthumb is layerized as expected, there is a layer with the right vscrollbar tag, and it has a shadow transform that is what I would expect. So I'm not really sure what's going on here.

[1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=d8cd72edf11c&group_state=expanded&selectedJob=17436762
Did another try push with some more logging [1], the 10.6 output seems very similar to the 10.10 output, so I have no clue as to what's going wrong. I'm not even sure what the next step here would be. Maybe running it again with texture dumping enabled, but even that I wouldn't expect to show any differences because if there is a problem it's likely in the compositing.

[1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=7bca9fbdd9e9&group_state=expanded
(In reply to Ryan VanderMeulen [:RyanVM] from comment #2)
> Also seeing "layout/reftests/invalidation/scroll-inactive-layers.html |
> failed reftest-no-paint" on OSX 10.10 only.

I've spun this off to bug 1253690.
The remaining reftest failures are all 10.6 which we're not going to support in 46 onwards. So I'm going to close this bug.
Assignee: nobody → ryanvm
Status: NEW → RESOLVED
Closed: 4 years ago
Keywords: leave-open
Resolution: --- → FIXED
Duplicate of this bug: 1244813
You need to log in before you can comment on or make changes to this bug.