Closed Bug 827372 Opened 8 years ago Closed 8 years ago

Get B2G emulator crashtests running on Cedar


(Testing :: Reftest, defect)

Gonk (Firefox OS)
Not set


(Not tracked)



(Reporter: ahal, Assigned: ahal)




(1 file)

We should get crashtests running on Cedar. Currently they seem to time out somewhere in the reftest harness before starting. I'll investigate.
The problem is that emulator.wait_for_port() is timing out:

The odd part is that if I change 'testing/crashtest/crashtest.list' to 'layout/reftests/reftest.list' everything works fine. It must be something to do with the remote webserver
Didn't notice anything interesting in the logcat either:
I lied, this shows up when running with crashtest.list but not reftest.list

E/GeckoConsole(  298): [JavaScript Warning: "Duplicate resource declaration for 'specialpowers' ignored." {file: "file:///system/b2g/distribution/bundles/" line: 2}]
E/GeckoConsole(  298): [JavaScript Error: "NS_ERROR_ILLEGAL_VALUE: Component returned failure code: 0x80070057 (NS_ERROR_ILLEGAL_VALUE) [nsIXPCComponents_Utils.import]" {file: "jar:file:///system/b2g/omni.ja!/components/HealthReportService.js" line: 10}]
E/GeckoConsole(  298): [JavaScript Error: "SpecialPowersObserverAPI is not defined" {file: "file:///system/b2g/distribution/bundles/" line: 37}]
E/GeckoConsole(  298): While creating services from category 'profile-after-change', could not create service for entry ';1', contract ID ';1'
I/Gecko   (  298): 1357587902227        Marionette      INFO  MarionetteComponent loaded
I/Gecko   (  298): 1357587902251        Marionette      INFO  marionette enabled, loadearly: true
Depends on: 827504
This should be all that's needed on the mozharness side (besides proper configs). Though it probably won't work until bug 827504 is fixed.
Attachment #699378 - Flags: review?(aki)
Attachment #699378 - Flags: review?(aki) → review+
Depends on: 829257
Ok, these are running on Cedar now in three chunks. Chunks 1 and 2 seem to hit bug 818103 consistently on the same test (even though in normal reftests it usually happens on a different test each time). This likely means a different problem with the same end result.

Chunk 3 actually seemed to complete a successful run, but it got marked as a fail by buildbot anyway. Overall these look to be pretty healthy and I might be able to get them green by disabling a few tests. I want to wait for bug 807970 to land first though as that could potentially smooth out a couple of issues.
I think this can be closed, I'll file new a one if future crashtest issues pop up.
Closed: 8 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.