Filed by: philringnalda [at] gmail.com https://treeherder.mozilla.org/logviewer.html#?job_id=2994975&repo=autoland https://queue.taskcluster.net/v1/task/AWIoNulCTkqiN_b-NYnhYQ/runs/0/artifacts/public%2Flogs%2Flive_backing.log

This looks like infrastructure failure. 22:05:28 INFO - File "/home/worker/workspace/build/venv/local/lib/python2.7/site-packages/mozdevice/devicemanagerADB.py", line 229, in pushFile 22:05:28 INFO - (localname, destname)) 22:05:28 INFO - DMError: Attempted to push a file (/tmp/tmps6bQBD) to a directory (/storage/sdcard/tests/xpc/p/mozinfo.json)! 22:05:28 INFO - Traceback (most recent call last): 22:05:28 INFO - File "/home/worker/workspace/build/tests/xpcshell/remotexpcshelltests.py", line 639, in <module> 22:05:28 INFO - main() 22:05:28 INFO - File "/home/worker/workspace/build/tests/xpcshell/remotexpcshelltests.py", line 634, in main 22:05:28 INFO - **vars(options)): 22:05:28 INFO - File "/home/worker/workspace/build/tests/xpcshell/runxpcshelltests.py", line 1448, in runTests 22:05:28 INFO - raise exceptions[0] 22:05:28 INFO - mozdevice.devicemanager.DMError: Attempted to push a file (/tmp/tmps6bQBD) to a directory (/storage/sdcard/tests/xpc/p/mozinfo.json)! 22:05:28 ERROR - Return code: 1 22:05:28 ERROR - No tests run or test summary not found > Error is raised by mozdevice/devicemanagerADB.py#228 > if self.dirExists(destname): > raise DMError("Attempted to push a file (%s) to a directory (%s)!" % > (localname, destname)) It seems .../p/mozinfo.json is a directory, so there might be another related bug somewhere.

This doesn't ring any bells, but I also haven't dealt with the Android test harnesses in a while. Maybe gbrown has a clue?

Yes, this failure is unrelated to the test - it is an infra problem. The thing to look for is that mozharness cannot communicate with the emulator after the test: WARNING - Unable to communicate with emulator via adb

Near as I can tell from auditing the last 300 pushes, Android xpcshell tests have four states, then: * green * timed out * retried on buildbot * invalid, that's not an actual failure, that's supposed to have been retried Enough, I'm hiding the taskcluster ones wherever they run. I don't actually see much of any point in running the buildbot ones either.

We've improved detection of this condition -- it shouldn't happen now.