1412297 - Some Android test crashes are caused by aborts, but no reason is logged

Status: RESOLVED FIXED

(Firefox for Android Graveyard :: Testing, defect)

Description

[Geoff Brown [:gbrown]]

If a test harness job hangs or runs for too long, the harness may abort the job, causing a crash. In these cases, the harness logs the reason for the abort, takes a screenshot, kills Firefox, and collects and reports the crash.

Recently, I see tests failing exactly like this, except the cause of the abort is not logged, and not apparent.

For instance, bug 1412223: A screenshot was taken, the crash was caused by SIGABRT, but there is no sign of a hang or a timeout.

Comment 1

[Geoff Brown [:gbrown]]

Logs show "mozdevice ERROR" a few seconds earlier.

Comment 2

[Geoff Brown [:gbrown]]

Attachment: retry getTopActivity, explain kill

A minor tweak to how we wait for the remote browser: I'm hoping the retry will make an infrequent failure practically non-existent. Since failures are typically 60 second timeouts, I am reluctant to retry more than once.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=9364bc567c340ae9ddd680efca67b3b4dc6490db

Comment 4

[Joel Maher ( :jmaher ) (UTC -8)]

Comment on attachment 8923911 [details] [diff] [review]
retry getTopActivity, explain kill

Review of attachment 8923911 [details] [diff] [review]:
-----------------------------------------------------------------

this looks fairly safe

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/0341d179a4ba
Retry devicemanager getTopActivity once; r=jmaher

Comment 6

[Natalia Csoregi [:nataliaCs]]

https://hg.mozilla.org/mozilla-central/rev/0341d179a4ba