Closed Bug 1205648 Opened 9 years ago Closed 8 years ago

wpt focus pseudo selector test unstable with e10s

Categories

(Core :: DOM: Core & HTML, defect)

defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla48
Tracking Status
e10s + ---
firefox48 --- fixed

People

(Reporter: jgraham, Assigned: jgraham)

References

(Blocks 1 open bug)

Details

testing/web-platform/meta/html/semantics/selectors/pseudo-classes/focus.html.ini

Disabled for now
Still fails on e10s debug runs. Is this something you'd have cycles to look at, Boris?

https://treeherder.mozilla.org/logviewer.html#?job_id=18236939&repo=try#L11543

14:51:00     INFO - TEST-UNEXPECTED-FAIL | /html/semantics/selectors/pseudo-classes/focus.html | input3 has the attribute autofocus - assert_array_equals: lengths differ, expected 1 got 0
14:51:00     INFO - testSelector/<@http://web-platform.test:8000/html/semantics/selectors/pseudo-classes/utils.js:12:5
14:51:00     INFO - Test.prototype.step@http://web-platform.test:8000/resources/testharness.js:1382:20
14:51:00     INFO - test@http://web-platform.test:8000/resources/testharness.js:496:9
14:51:00     INFO - testSelector@http://web-platform.test:8000/html/semantics/selectors/pseudo-classes/utils.js:10:3
14:51:00     INFO - @http://web-platform.test:8000/html/semantics/selectors/pseudo-classes/focus.html:23:5
14:51:00     INFO - TEST-PASS | /html/semantics/selectors/pseudo-classes/focus.html | input1 has the focus 
14:51:00     INFO - TEST-PASS | /html/semantics/selectors/pseudo-classes/focus.html | tabindex attribute makes the element focusable 
14:51:00     INFO - TEST-PASS | /html/semantics/selectors/pseudo-classes/focus.html | editable elements are focusable 
14:51:00     INFO - TEST-PASS | /html/semantics/selectors/pseudo-classes/focus.html | ':focus' matches focussed body with tabindex 
14:51:00     INFO - TEST-PASS | /html/semantics/selectors/pseudo-classes/focus.html | ':focus' doesn't match focused elements in iframe
Component: web-platform-tests → DOM
Flags: needinfo?(bzbarsky)
Product: Testing → Core
I'm confused.  The failure listed in comment 1 (ignoring for the moment that it's now moved to esting/web-platform/tests/html/semantics/selectors/pseudo-classes/focus-autofocus.html and made non-racy in <https://github.com/w3c/web-platform-tests/commit/2d402a62a0b602ab5076cf57ef4bda967afb8290>) is not specific to e10s, right?  It presumably failed across the board.

I think we should just try removing testing/web-platform/meta/html/semantics/selectors/pseudo-classes/focus.html.ini and doing another try push; I expect it to pass.

And the "instability" mentioned in comment 0 was presumably because of the bogus raciness of the test, fwiw.  But it's hard to tell, since comment 0 doesn't say what exactly was unstable.
Flags: needinfo?(bzbarsky) → needinfo?(ryanvm)
Comment 1 included a link to the Try push doing exactly what you proposed. It included https://hg.mozilla.org/try/rev/2687e9ac8dd2 in the push.
Flags: needinfo?(ryanvm)
To be clear, that Try push had e10s forced on via https://hg.mozilla.org/try/rev/c7f7e72c32e4, so it was in fact an e10s test run (even though Treeherder shows it as non-e10s). It's a hack since Buildbot doesn't know about e10s versions of those jobs on Linux debug and enabling them for Try "officially" has other consequences.
> Comment 1 included a link to the Try push doing exactly what you proposed.

Yes, but it didn't include the actual test fix I'm referring to, which landed as part of http://hg.mozilla.org/mozilla-central/rev/6ab07b5c4c3e on March 19, two days after your try push...
Flags: needinfo?(ryanvm)
So, it seems we need to do a wpt update to pick up that fix and Try again? I can start that.
Oh, sorry, I misread the comment and thought you meant it had landed upstream on the 19th.
INFO - TEST-OK | /html/semantics/selectors/pseudo-classes/focus.html | took 2045ms

Across the board \m/
Assignee: nobody → james
Depends on: 1257799
https://hg.mozilla.org/mozilla-central/rev/9abade266139
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla48
Component: DOM → DOM: Core & HTML
You need to log in before you can comment on or make changes to this bug.