Closed Bug 1380971 Opened 7 years ago Closed 7 years ago

Intermittent bugs/1375315-11.html == bugs/1375315-11-ref.html | image comparison, max difference: 255, number of differing pixels: 48

Categories

(Core :: Graphics: WebRender, defect, P1)

defect

Tracking

()

RESOLVED FIXED
mozilla57

People

(Reporter: intermittent-bug-filer, Assigned: gbrown)

References

Details

(Keywords: intermittent-failure, Whiteboard: [wr-mvp] [stockwell disabled])

Attachments

(1 obsolete file)

Text in the test says "bullets should be green"; in recent failed cases -- all on linux64-qr -- the bullets are red.

This test has failed infrequently for most of its history; failures became much more frequent today.
Blocks: 1375315
If it's all qr, I'd say it's probably a Quantum Render bug, not a stylo bug per se.
Failures are frequent enough now that they need attention as soon as possible.

All failures are linux64-qr. I doubt we have a way to skip a test on -qr only and I'm not sure who might look at this from a Quantum Render perspective. Checking in with :bz as test author.
Flags: needinfo?(bzbarsky)
Whiteboard: [stockwell needswork]
As test author, there should be no magic here.  We should just be painting green circles.

What is the "linux64-qr" platform?  Is that webrender?  If so, that sounds like there's just a webrender bug that needs to be fixed.

Looking at a failing run the test in fact just fails: the bullets that should be green are red instead.  This suggests that the bug is an invalidation issue.

Assuming we're talking about webrender here, Milan, who can look into this?
Flags: needinfo?(bzbarsky) → needinfo?(milan)
> I doubt we have a way to skip a test on -qr only

Oh, and if we don't, we need to fix that ASAP.  It's a non-shipping configuration, so far, and disabling tests just because they fail in that configuration, losing coverage in our actual shipping configurations, is not really ok.
Assignee: nobody → jmuizelaar
We can probably just disable this test with webrender turned on.
Flags: needinfo?(milan)
(In reply to Jeff Muizelaar [:jrmuizel] from comment #16)
> We can probably just disable this test with webrender turned on.

That sounds promising. Hopefully I'm missing something, but I don't see how to do that. I filed bug 1398223 in case there is a generic concern.
(In reply to Geoff Brown [:gbrown] from comment #17)
> That sounds promising. Hopefully I'm missing something, but I don't see how
> to do that. I filed bug 1398223 in case there is a generic concern.

If you add skip-if(webrender) to the reftest line that should do it.
Component: Layout → Graphics: WebRender
Attached patch skip on webrender (obsolete) — Splinter Review
Failures are still very frequent; let's skip-if(webrender), at least temporarily.
Attachment #8907171 - Flags: review?(jmuizelaar)
See Also: → 1382486
Comment on attachment 8907171 [details] [diff] [review]
skip on webrender

Too many failures! Going ahead with random-if, as in bug 1382486.
Attachment #8907171 - Attachment is obsolete: true
Attachment #8907171 - Flags: review?(jmuizelaar)
Pushed by gbrown@mozilla.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/cacc4579fb63
Mark reftest 1375315-11.html as random on webrender for frequent intermittent failures; r=me,test-only
Keywords: leave-open
Whiteboard: [stockwell needswork] → [stockwell disabled]
(In reply to OrangeFactor Robot from comment #31)
> For more details, see:
> https://brasstacks.mozilla.com/orangefactor/
> ?display=Bug&bugid=1380971&startday=2017-09-25&endday=2017-10-01&tree=all

There are two failures listed here, both are misclassifications. I'm going to close this bug as fixed, since we fixed it by marking the test random-if. I have a spreadsheet tracking all the webrender-specific failures so there's no need to keep this bug open.
Assignee: jmuizelaar → gbrown
Status: NEW → RESOLVED
Closed: 7 years ago
Keywords: leave-open
Resolution: --- → FIXED
Target Milestone: --- → mozilla57
Priority: -- → P1
Whiteboard: [stockwell disabled] → [wr-mvp] [stockwell disabled]
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: