We're linking gtest-libxul on demand by design. This is great because we're not increasing the turn around time for everyone just in case they might happen to run some test. Sadly this is tripping our timeout for 'make check'. Can we get it increased to the same timeout as our build step? I don't have data on how frequent timeouts are during the 'make check' phase but I hope this number is low so the impact of doing this should be low as well. My ideal solution is to have a separate build phase where we build&prepare our tests but this is a big project sadly. It would be a huge win for developers turn around times. But unless we can tackle this problem increasing the timeout is what I'm requesting here.

jmaher, do you know who can drive this? I imagine it's a simple change but I don't know who to ask.

Managed to track this down - think I have the right buildbot step/timeout value...

Ben is out for a week so here's a driveby instead. Does the 10 minute value come from looking at jobs ? The linking of libxul for the main app takes much longer than 10 minutes on windows, where compile step has a two hour timeout.

from looking at the patch :edmorley posted, it looks as though the check command has a 5 minute limit.

(In reply to Nick Thomas [:nthomas] from comment #3) > Ben is out for a week so here's a driveby instead. Does the 10 minute value > come from looking at jobs ? The linking of libxul for the main app takes > much longer than 10 minutes on windows, where compile step has a two hour > timeout. It's currently 5 mins without output (rather than total job length) aiui. We're intermittently hitting the current 5 min timeout in bug 886079 - the 10 mins I'm adjusting it to here is a bit of a guess.

Comment on attachment 785019 [details] [diff] [review] Make timeout 10 mins rather than 5 Review of attachment 785019 [details] [diff] [review]: ----------------------------------------------------------------- stamp. do we need a similar thing in mozharness, or are we not running make check through it?

(In reply to Ben Hearsum [:bhearsum] from comment #6) > do we need a similar thing in mozharness, or are we not running make check > through it? I don't believe we support it yet (both through inspecting the mozharness repo and build logs for platforms where we do use mozharness for the other steps).

In production.

Should we close this bug then?

Yup :-)