Closed Bug 1339062 Opened 3 years ago Closed 3 years ago
.0 == http://10 .2 .2:8854/tests/layout/reftests/webcomponents/input-transition-1 .html .0 | image comparison, max difference: 229, number of differing pixels: 23 .2 .2:8854/tests/layout/reftests/webcomponents/input-transition-1-ref .html
Filed by: cbook [at] mozilla.com https://treeherder.mozilla.org/logviewer.html#?job_id=76793742&repo=mozilla-inbound https://queue.taskcluster.net/v1/task/cocOx579RPqEmthXA_Qnjw/runs/0/artifacts/public/logs/live_backing.log https://hg.mozilla.org/mozilla-central/raw-file/tip/layout/tools/reftest/reftest-analyzer.xhtml#logurl=https://queue.taskcluster.net/v1/task/cocOx579RPqEmthXA_Qnjw/runs/0/artifacts/public/logs/live_backing.log&only_show_unexpected=1
I'll try to narrow the regression range: https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&filter-searchStr=android%20reftest&tochange=2f9d2be88b3854db220c9f1696a40140a360c9a2&fromchange=739f1c4bee0df80fcf014af10a6c83eaccc80211
:jet - This Android-only reftest failure is very frequent. Can you find someone to look at it?
(In reply to Geoff Brown [:gbrown] from comment #3) > I'll try to narrow the regression range: > https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&filter- > searchStr=android%20reftest&tochange=2f9d2be88b3854db220c9f1696a40140a360c9a2 > &fromchange=739f1c4bee0df80fcf014af10a6c83eaccc80211 The earliest I could find was https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=2f9d2be88b3854db220c9f1696a40140a360c9a2&filter-searchStr=android+reftest, but the failure was so infrequent on that changeset, I would say it is unclear which changeset is responsible.
:jet, any chance we can get some attention to this bug? :ni myself to follow up on Tuesday if needed.
Maybe similar to bug 1342240, though this test doesn't seem to be the first test in the run. Is it the first test in the run with a checked checkbox in it?
See Also: → 1342240
Seems like a decent chance this is a regression from either bug 605985 or bug 1338293.
The test is doing "el.checked = true" in the callback. Since the check-mark is an image on Android it seems there is a race between loading it and the 'transitionend' event that removes the 'class' attribute, which triggers the reftest screenshot. http://searchfox.org/mozilla-central/source/layout/reftests/webcomponents/input-transition-1.html I think we just need to wait for it to load somehow. It's a bit mysterious why this turns up now though... Does radio/checkbox elements have additional style sheets on Android besides forms.css?
thanks :dbaron and :mats for chiming in here. :snorp, do you know the answer to :mats question in comment 11?
Flags: needinfo?(jmaher) → needinfo?(snorp)
I'm not an expert on these bits, but there is some css for forms in https://dxr.mozilla.org/mozilla-central/source/mobile/android/themes/core/content.css
This has been failing at 50+ times/week for the last few weeks, I want to disable this test, ni? myself to write a patch for that later today.
Waiting 1 second before taking the reftest snapshot seems to cure it. It's a bit ugly, but think I'd prefer that over disabling the test. https://treeherder.mozilla.org/#/jobs?repo=try&revision=c9a3fd396a48f2a6dbd1d7dec337308f8b75839c
Attachment #8847113 - Flags: feedback?(jmaher)
that seems reasonable- if we do that, we should consider leaving this bug open or filing a follow up to investigate the issues.
Comment on attachment 8847113 [details] [diff] [review] wallpaper Review of attachment 8847113 [details] [diff] [review]: ----------------------------------------------------------------- this is really hacky, but it will help ensure we are avoiding timing issues.
Attachment #8847113 - Flags: feedback?(jmaher) → feedback+
Pushed by email@example.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/efbc8ac7bca1 Wallpaper an orange failure by delaying the snapshot for 1s. r=me
Whiteboard: [stockwell needswork] → [stockwell fixed]
You need to log in before you can comment on or make changes to this bug.