Closed
Bug 1015178
Opened 7 years ago
Closed 7 years ago
B2G reftests don't killAndGetStack on timeout
Categories
(Testing :: Reftest, defect)
Testing
Reftest
Tracking
(Not tracked)
RESOLVED
FIXED
mozilla32
People
(Reporter: ahal, Assigned: ahal)
References
Details
Attachments
(1 file, 2 obsolete files)
2.21 KB,
patch
|
ahal
:
review+
|
Details | Diff | Splinter Review |
We implemented kill and get stack for B2G in mozrunner, but reftests still use b2gautomation.py. Ideally we'd move them off b2gautomation.py and onto mozrunner, but the simple thing to do here is to just re-implement it. http://mxr.mozilla.org/mozilla-central/source/testing/mozbase/mozrunner/mozrunner/remote.py#246
Assignee | ||
Comment 1•7 years ago
|
||
Should be all that's needed. Try run with the patch from bug 966646 to help reproduce timeouts: https://tbpl.mozilla.org/?tree=Try&rev=cfee277e40ae
Assignee: nobody → ahalberstadt
Status: NEW → ASSIGNED
Assignee | ||
Comment 2•7 years ago
|
||
So that somewhat worked, after the timeout the process was killed and checkForCrashes invoked.. but minidump_stackwalk failed with "header signature mismatch": https://tbpl.mozilla.org/php/getParsedLog.php?id=40287378&tree=Try&full=1#error1 Ted, what does it mean when this happens?
Flags: needinfo?(ted)
Comment 3•7 years ago
|
||
We chatted on IRC, the current working theory is that maybe we're pulling the dump from the device too soon and it's not finished writing yet. ahal is going to try putting a sleep in there to test that theory.
Flags: needinfo?(ted)
Assignee | ||
Comment 4•7 years ago
|
||
Try run with sleep: https://tbpl.mozilla.org/?tree=Try&rev=d90894750c8e If this works, I'll modify it to poll the b2g process before committing.
Assignee | ||
Comment 5•7 years ago
|
||
I think that worked: https://tbpl.mozilla.org/php/getParsedLog.php?id=40481364&tree=Try&full=1#error0 I'll get a patch up shortly.
Assignee | ||
Comment 6•7 years ago
|
||
Updated patch to wait for b2g process to stop existing. New try run just to be sure it still works: https://tbpl.mozilla.org/?tree=Try&rev=50e86a011121
Attachment #8427754 -
Attachment is obsolete: true
Attachment #8429567 -
Flags: review?(wlachance)
Comment 7•7 years ago
|
||
Comment on attachment 8429567 [details] [diff] [review] b2g_kill_and_get_stack Review of attachment 8429567 [details] [diff] [review]: ----------------------------------------------------------------- Looks ok to me, one minor suggestion. ::: build/mobile/b2gautomation.py @@ +155,5 @@ > self.lastTestSeen, int(timeout)) > + self._devicemanager.killProcess('/system/b2g/b2g', sig=signal.SIGABRT) > + > + timeout = 10 # seconds > + for i in range(0, timeout): I think the marionette approach is somewhat more intuitive: starttime = datetime.datetime.now() while datetime.datetime.now() - starttime < datetime.timedelta(seconds=timeout):
Attachment #8429567 -
Flags: review?(wlachance) → review+
Comment 8•7 years ago
|
||
(In reply to Andrew Halberstadt [:ahal] from comment #5) > I think that worked: > https://tbpl.mozilla.org/php/getParsedLog. > php?id=40481364&tree=Try&full=1#error0 Yeah, that looks good. Good call!
Assignee | ||
Comment 9•7 years ago
|
||
Updated patch with suggestion. Carry r+ forward.
Attachment #8429567 -
Attachment is obsolete: true
Attachment #8430019 -
Flags: review+
Assignee | ||
Comment 10•7 years ago
|
||
https://hg.mozilla.org/integration/b2g-inbound/rev/33a2a8aaed24
Comment 11•7 years ago
|
||
https://hg.mozilla.org/mozilla-central/rev/33a2a8aaed24
Status: ASSIGNED → RESOLVED
Closed: 7 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla32
You need to log in
before you can comment on or make changes to this bug.
Description
•