Once bug 923500 has rolled out, we can clean up remoteautomation.py and make it use mozcrash::check_for_java_exception() instead.
Created attachment 813604 [details] [diff] [review] Patch v1 Make remoteautomation.py use the new mozcrash check_for_java_exception(). Will obviously need to wait until bug 923500 rolled out etc.
Attachment #813604 - Flags: review?(jmaher)
Comment on attachment 813604 [details] [diff] [review] Patch v1 Review of attachment 813604 [details] [diff] [review]: ----------------------------------------------------------------- please verify on try server for the tegras sys path. ::: build/mobile/remoteautomation.py @@ +10,5 @@ > import subprocess > > from automation import Automation > from devicemanager import NetworkTools, DMError > +import mozcrash are there dependencies here? We run on the tegras which isn't mozharness, so we need to ensure all dependencies of mozcrash are available, more specifically the modules are in the sys.path (this requires sys.path hacking, but is probably already done at the runner script)
Attachment #813604 - Flags: review?(jmaher) → review+
(In reply to Joel Maher (:jmaher) from comment #2) > > +import mozcrash > > are there dependencies here? We run on the tegras which isn't mozharness, > so we need to ensure all dependencies of mozcrash are available, more > specifically the modules are in the sys.path (this requires sys.path > hacking, but is probably already done at the runner script) automation.py.in already uses mozcrash on Android, and appends the necessary directories to the path here: https://hg.mozilla.org/integration/mozilla-inbound/file/bf856e026871/build/automation.py.in#l38 ...which unless I'm mistaken, permanently changes the path for all subsequent imports in all files. But yeah, will verify on Try (you don't need to worry about a sheriff not using Try... would be kind of hypocritical not to... ;-))
With intentional-crash patch applied: remote: https://tbpl.mozilla.org/?tree=Try&rev=a29796f74d73
(In reply to Ed Morley [:edmorley UTC+1] from comment #4) > With intentional-crash patch applied: > remote: https://tbpl.mozilla.org/?tree=Try&rev=a29796f74d73 Ok, few issues: 1) On the pandas, the releng pypi mirror is missing mozcrash-0.9, I'll file a bug to have it added. 2) In mozharness for pandas we should specify a minimum version of mozcrash, so we fail early in cases like this. Guessing we do this somewhere here: https://hg.mozilla.org/build/mozharness/file/55fa038de460/scripts/android_panda.py#l133 https://hg.mozilla.org/build/mozharness/file/55fa038de460/scripts/android_panda_talos.py#l88 3) (But not an issue with this patch in particular) The panda mochitest-1 run had an uncaught exception whilst trying to process the java-exception, and yet the exception didn't turn the run orange. Joel, presume I'm just good to file a releng bug for #1? Happy for me to do #2 (ie have a >= for mozcrash-0.9)?
More specifically: > 1) On the pandas, the releng pypi mirror is missing mozcrash-0.9 ... > 2) In mozharness for pandas we should specify a minimum version of mozcrash 16:40:31 INFO - Installing mozcrash into virtualenv /builds/panda-0854/test/build/venv 16:40:31 INFO - Running command: ['/builds/panda-0854/test/build/venv/bin/pip', 'install', '--download-cache', '/builds/panda-0854/test/build/venv/cache', '--no-index', '--find-links', 'http://pypi.pvt.build.mozilla.org/pub', '--find-links', 'http://pypi.pub.build.mozilla.org/pub', 'mozcrash'] in /builds/panda-0854/test/build 16:40:31 INFO - Copy/paste: /builds/panda-0854/test/build/venv/bin/pip install --download-cache /builds/panda-0854/test/build/venv/cache --no-index --find-links http://pypi.pvt.build.mozilla.org/pub --find-links http://pypi.pub.build.mozilla.org/pub mozcrash 16:40:33 INFO - Ignoring indexes: http://pypi.python.org/simple/ 16:40:33 INFO - Downloading/unpacking mozcrash 16:40:33 INFO - Downloading mozcrash-0.5.tar.gz 16:40:33 INFO - Storing download in cache at ./venv/cache/http%3A%2F%2Freleng-puppet2.build.scl1.mozilla.com%2Fpython%2Fpackages%2Fmozcrash-0.5.tar.gz > 3) (But not an issue with this patch in particular) The panda mochitest-1 > run had an uncaught exception whilst trying to process the java-exception 16:57:15 INFO - INFO | zombiecheck | Reading PID log: /tmp/tmp1PoLeMpidlog 16:57:16 INFO - Traceback (most recent call last): 16:57:16 INFO - File "/builds/panda-0854/test/build/tests/mochitest/runtests.py", line 1005, in runTests 16:57:16 INFO - onLaunch=onLaunch 16:57:16 INFO - File "/builds/panda-0854/test/build/tests/mochitest/runtestsremote.py", line 542, in runApp 16:57:16 INFO - return self._automation.runApp(*args, **kwargs) 16:57:16 INFO - File "/builds/panda-0854/test/build/tests/mochitest/automation.py", line 899, in runApp 16:57:16 INFO - crashed = self.checkForCrashes(os.path.join(profileDir, "minidumps"), symbolsPath) 16:57:16 INFO - File "/builds/panda-0854/test/build/tests/mochitest/remoteautomation.py", line 118, in checkForCrashes 16:57:16 INFO - javaException = mozcrash.check_for_java_exception(logcat) 16:57:16 INFO - AttributeError: 'module' object has no attribute 'check_for_java_exception' 16:57:16 INFO - /data/anr/traces.txt not found 16:57:16 INFO - Automation Error: Received unexpected exception while running application 16:57:16 INFO - WARNING | leakcheck | refcount logging is off, so leaks can't be detected! 16:57:16 INFO - runtests.py | Running tests: end.
yes, we will need releng bugs for both of those issues. Once those are updated, we can work on rolling this out, I would think this would be a 1 day turnaround. please cc, kmoir, aki, armen, callek on the bug(s) for releng.
(In reply to Ed Morley [:edmorley UTC+1] from comment #6) > 2) In mozharness for pandas we should specify a minimum version of mozcrash In fact, I think this will be made redundant by bug 921596.
All seemed good on the try retriggers now that the releng pypi store has mozcrash-0.9, so landed: https://hg.mozilla.org/integration/mozilla-inbound/rev/7652896ed115
Status: ASSIGNED → RESOLVED
Last Resolved: 5 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla27
You need to log in before you can comment on or make changes to this bug.