1482887 - Remote automation should kill already running instances rather than dying

Status: RESOLVED FIXED

(Testing :: General, enhancement)

Description

[Bob Clary [:bc] (inactive)]

adb_android.py's launch_application has an option to fail if an instance is already running which can be useful however remoteautomation.py does not attempt to recover the situation. I think we should have RProcess kill any running instance prior to attempting to launch.

Comment 1

[Bob Clary [:bc] (inactive)]

Attachment: RProcess-stop-application.patch

try coming soon.

Comment 2

[Bob Clary [:bc] (inactive)]

https://treeherder.mozilla.org/#/jobs?repo=try&author=[e-mail]&filter-tier=1&filter-tier=2&filter-tier=3&group_state=expanded

Comment 3

[Geoff Brown [:gbrown]]

Comment on attachment 8999615 [details] [diff] [review]
RProcess-stop-application.patch

Review of attachment 8999615 [details] [diff] [review]:
-----------------------------------------------------------------

Traditionally, we stop the test app when the harness is starting. For example,

https://dxr.mozilla.org/mozilla-central/rev/4e56a2f51ad739ca52046723448f3129a58f1666/testing/mochitest/runtestsremote.py#95

...and also, we try to make sure that the app is stopped after a test run completes:

https://dxr.mozilla.org/mozilla-central/rev/4e56a2f51ad739ca52046723448f3129a58f1666/build/mobile/remoteautomation.py#90-92

But I think there are some holes in that logic: Like on a mochitest run with multiple browser sessions, if the browser goes to the background, if get_top_activity fails, etc. So, I like this!

::: build/mobile/remoteautomation.py
@@ +241,5 @@
>              self.stdoutlen = 0
>              self.utilityPath = None
>  
> +            if app and self.device.process_exist(app):
> +                print("remoteautomation.py %s is already running. Stopping...")

If we could do without this message, I think we could avoid the process_exist() call -- it looks like calling stop_application is okay whether the process is running or not. But the message might be useful -- alert us to unexpected behavior, etc. -- so probably best to keep it as it is.

Comment 4

[Bob Clary [:bc] (inactive)]

Thanks. I ran into this a bunch this weekend trying to get mach working for android hardware testing. Maybe we can chase it down.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/3f7bfeff4b2f
Remote automation should kill already running instances rather than dying, r=gbrown.

Comment 6

[Dorel Luca [:dluca]]

https://hg.mozilla.org/mozilla-central/rev/3f7bfeff4b2f