Open Bug 1389319 Opened 2 years ago Updated Last year

Intermittent bugs/1313772.xhtml == bugs/1313772.xhtml | image comparison, max difference: 14, number of differing pixels: 43

Categories

(Core :: Layout, defect, P5)

defect

Tracking

()

REOPENED

People

(Reporter: intermittent-bug-filer, Unassigned, NeedInfo)

References

Details

(Keywords: intermittent-failure, leave-open, Whiteboard: [stockwell unknown])

Attachments

(1 file)

Resolving intermittent Stylo test failure bugs that haven't had any recent activity.
Status: NEW → RESOLVED
Closed: 2 years ago
Resolution: --- → WORKSFORME
This has 32 fails in the last 7 days. All of them happening on windows7-32. 

An example of the most recent log: https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-central&job_id=145161249&lineNumber=1585
Flags: needinfo?(bugs)
Whiteboard: [stockwell needswork]
To Neerja for a look. This needs some fuzzy-if() on Windows too.
Flags: needinfo?(bugs) → needinfo?(npancholi)
Assignee: nobody → npancholi
Flags: needinfo?(npancholi)
This started occurring again on December 6th.
Since then, there have been 56 total failures.

https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/KxHra6z_TJqfeUWO9NH6zQ/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1

Occurrences by platforms:
- 46 on Windows 7
- 4 on Linux x64
- 4 on windows10-64
- 1 on OS X 10.10
- 1 on linux64-stylo-disabled

Occurrences by build type:
- 28 on opt
- 23 on pgo 
- 5 on debug

Here is a link from a recent relevant log file:
https://treeherder.mozilla.org/logviewer.html#?repo=mozilla-inbound&job_id=150217208&lineNumber=4175

And a link to the reftest-analyzer:
https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/e8fZ-6FcSw-Yxi3PhjWuCA/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1
Flags: needinfo?(npancholi)
(In reply to Tiberius Oros[:tiberius_oros] from comment #11)

Added a patch for this so cancelling needinfo.
Flags: needinfo?(npancholi)
Comment on attachment 8942342 [details]
Bug 1389319 - Add fuzzy annotation for 1313772.xhtml for non-Android platforms.

https://reviewboard.mozilla.org/r/212624/#review218338

This failure actually stopped happening, the day after comment 11.  In fact, it's only failed a couple times (all at once) in the time since this bug was filed:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1389319&startday=2017-08-09&endday=2018-01-12&tree=all

So: that feels to me like the December spike (Comment 11) was caused by an actual bug which was later addressed (elsewhere), and we probably don't need to (& shouldn't) paper over it given that it's not causing trouble right now.
Attachment #8942342 - Flags: review?(dholbert) → review-
Having said that -- it looks like this is still kind of a problem, since we've gotten one recent report on Jan 10th:
 https://treeherder.mozilla.org/logviewer.html#?repo=autoland&job_id=155396795&lineNumber=8754
...which happened to be on Linux (not Windows), but still had max-difference 14 number of differing pixels 43.

The reftest-analyzer screenshot looks harmless (it looks like a subtle color rounding error on the edge of a shape inside of a scaled-up image), so it seems annotation-worthy to me -- but a Windows-specific annotation probably isn't what we want, given that the only recent report was on Linux.

So: I'd r+ a version of the attached patch with a simple "fuzzy(...)" annotation rather than "fuzzy-if(...)".  Or, I'd also be fine leaving this un-annotated (given the 1 failure per month frequency we're seeing right now) -- and if it gets bad again, then we'll notice & maybe be able to figure out why.
(In reply to Daniel Holbert [:dholbert] from comment #16)
I would say that since the unconditional fuzzy condition would have a much larger threshold compared to the 43 pixels we see here (because it already has fuzzy-if(Android,27,874) in the reftest.list), maybe it would make sense to not add this and wait to see if it becomes worse like dholbert suggested.
I don't think we'd have to merge the general annotation with the Android-annotation... We could, for example, probably do:
  fuzzy-if(Android,[big numbers]) fuzzy-if(!Android,[smaller numbers])

But it's also fine to leave things as they are, too.
(In reply to Daniel Holbert [:dholbert] from comment #18)

Ok, I guess if the number of pixels says the same then it makes sense to fuzzy-if it. Updated the patch :)
Comment on attachment 8942342 [details]
Bug 1389319 - Add fuzzy annotation for 1313772.xhtml for non-Android platforms.

https://reviewboard.mozilla.org/r/212624/#review218350

r=me with commit message brought up to date:

::: commit-message-4db16:1
(Diff revision 2)
> +Bug 1389319 - Add fuzzy annotation for 1313772.xhtml for Windows. r?dholbert

s/for Windows/for non-Android platforms/

(Per the stats in comment 12, it looks like we've had reports of this pretty much everywhere, so I think it's accurate enough to say this tweak is aiming to target all non-Android platforms.)
Attachment #8942342 - Flags: review?(dholbert) → review+
Pushed by npancholi@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/19bda1338749
Add fuzzy annotation for 1313772.xhtml for non-Android platforms. r=dholbert
Assignee: npancholi → nobody
The leave-open keyword is there and there is no activity for 6 months.
:svoisen, maybe it's time to close this bug?
Flags: needinfo?(svoisen)
You need to log in before you can comment on or make changes to this bug.