Closed Bug 963103 Opened 11 years ago Closed 10 years ago

Intermittent test_bug369306.html | The focus should not have been changed with URL=data:text/html,<script>blur();</script>

Categories

(Core :: DOM: Events, defect)

Product:
Core
Component:
DOM: Events
Platform:
ARM
Gonk (Firefox OS)
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED WORKSFORME

People

(Reporter: cbook, Unassigned)

References

(
URL
)

Details

(Keywords: intermittent-failure)

b2g_emulator_vm b2g-inbound opt test mochitest-5 on 2014-01-23 04:01:11 PST for push 20fcdbb2d5cc slave: tst-linux64-spot-015 https://tbpl.mozilla.org/php/getParsedLog.php?id=33454070&tree=B2g-Inbound 10189 ERROR TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/bugs/test_bug369306.html | The focus should not have been changed with URL=data:text/html,<script>blur();</script>
OS: Linux → Gonk (Firefox OS)
Hardware: x86 → ARM
Summary: Intermittent TEST-UNEXPECTED-FAIL | /tests/dom/tests/mochitest/bugs/test_bug369306.html | The focus should not have been changed with URL=data:text/html,<script>blur();</script> → Intermittent test_bug369306.html | The focus should not have been changed with URL=data:text/html,<script>blur();</script>
It's not a b2g only fail. I can repo this bug on Linux 64 as well
By running this command ./mach mochitest dom/tests/mochitest/bugs/test_bug369306.html Repro rate is around 50% on my desktop
So, a couple of things I've noticed while trying to figure out why this is permaorange with the patches from bug 991037: 1. The mochitest window is chrome (on m-c, all platforms apparently), meaning that the dom.disable_window_flip pref should have no effect at all, and ought to cause the test to always fail (since chrome windows are allowed to mess around with focus of other windows all they want, see http://dxr.mozilla.org/mozilla-central/source/dom/base/nsGlobalWindow.cpp#6576 and http://dxr.mozilla.org/mozilla-central/source/dom/base/nsGlobalWindow.cpp#7492). 2. Leaving the dom.disable_window_flip pref set to its default value still allows the test to pass most of the time. 3. Explicitly setting the pref to false also allows the test to pass most of the time. For these reasons, I think the test is simply broken, and needs to be disabled and fixed.
Continuing to experiment; the stuff the test does to attempt to cause focus to change never even reaches the code in nsGloblaWindow that does things like check the pref. This is the only reason this test passes; it really should always fail. For some reason, occasionally the action is executed, causing the orange.
Blocks: 991037
Ok, more experimentation; it looks like we aren't bypassing the pref like I thought we were, but when the pref is flipped, the Focus/Blur just fail later on. Not sure what is going on here...
Since we're just going to disable the PeerConnection tests on b2g emulator, this no longer blocks.
No longer blocks: 991037
Inactive; closing (see bug 1180138).
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.