Open Bug 1341310 Opened 7 years ago Updated 2 years ago

Intermittent forms/legend/1339287.html == forms/legend/1339287-ref.html | image comparison, max difference: 255, number of differing pixels: 14562

Categories

(Core :: Layout, defect)

Product:
Core
Component:
Layout
Type:
defect

Tracking

()

People

(Reporter: intermittent-bug-filer, Unassigned)

References

Details

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

Attachments

(1 file)

stack
8.21 KB, text/plain
Details 
Hmm, that's odd.  Either the fix was insufficient, or the <legend>
was reflowed with a too small available space or something.

FTR, the above job is:
Linux x64 debug Reftests executed by TaskCluster with e10s test-linux64/debug-reftest-e10s-5 tc-R-e10s(R5)

So this is actually reproducible on my machine...  in 1 of 100 runs.
I'll try to capture it in rr somehow...
Assignee: nobody → mats
Attached file stack — 
The fix in bug 1339287 is working correctly, but we later get a RebuildAllStyleData
request and then we have lost the information that we should ignore mDisplay on
the parent context...
I think this is related to what Boris reported in bug 1340622.
Assignee: mats → nobody
Depends on: 1340622
this is failing primarily on android debug e10s, almost at the threshold where we want to see a fix, I am going to mark this as needswork so  Ican keep an eye on it.
Whiteboard: [stockwell needswork]
oh, picking up a bit more :(  I suspect March 8th as the time window as per:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1341310&startday=2017-03-01&endday=2017-03-23&tree=all

doing some backfilling and retriggers on autoland where the first android instance showed up:
https://treeherder.mozilla.org/#/jobs?repo=autoland&filter-searchStr=reftest-28&tochange=9e21c2dc1b0da1c450462a3f5458d6c6d547fd3a&fromchange=902c21221272f2ff6c86bb83e11a4310d4dace53

as a note, we only run this test every 5th push, so the failure rate is much higher on android

ni myself to follow up on the retriggers
Flags: needinfo?(jmaher)
and the root cause is bug 1343078:
https://treeherder.mozilla.org/#/jobs?repo=autoland&filter-searchStr=reftest-28&tochange=74bd66f2490e43cb9dfc9a8dbc4a0f01a2629126&fromchange=fd66e733b07562475d910ed0928f35cb7f700759

:bz, can you take a look at this an see if you can get a fix for this or help find someone who could help out with this?
Blocks: 1343078
Flags: needinfo?(jmaher) → needinfo?(bzbarsky)
That regression range and regressing bug is interesting, because neither the test nor the reference has any placeholders.  And while https://hg.mozilla.org/integration/autoland/rev/44750ea79b9f could affect some other things, they aren't present in this testcase either, afaict.

In any case, this intermittent quite predates the fix for bug 1343078...  Is that just being blamed for this becoming more common, or something else?

In any case, comment 3 is right.  We should fix bug 1340622.
Flags: needinfo?(jmaher)
Flags: needinfo?(cam)
Flags: needinfo?(bzbarsky)
the only difference is this started on Android after we identified that issue, maybe it is a slight rendering or timing quirk and now the same root cause shows up on Android, lets work on fixing bug 1240622 and see where this ends up.
Flags: needinfo?(jmaher)
this has reduced in frequency- appears the android failures are less, but the other platforms are similar.
Whiteboard: [stockwell needswork] → [stockwell unknown]
21 failures, android is the key here which started around 4 weeks ago:
https://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=1341310

I don't see an urgency to investigate this, but this is one of those intermittent bugs we should keep an eye on- possibly look at why android is causing problems.
The switch to Stylo, which happens to fix bug 1340622, probably resolved this.
Flags: needinfo?(cam)
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: