Summary

In the logs for linux platform tests, there is a section where the test-linux.sh script is forced to sleep for 15 seconds.

Example from linux64:

[task 2019-04-29T16:56:11.387Z] if $NEED_WINDOW_MANAGER; then
[task 2019-04-29T16:56:11.387Z]     # This is read by xsession to select the window manager
[task 2019-04-29T16:56:11.388Z]     echo DESKTOP_SESSION=ubuntu > $HOME/.xsessionrc
[task 2019-04-29T16:56:11.389Z] 
[task 2019-04-29T16:56:11.389Z]     # note that doing anything with this display before running Xsession will cause sadness (like,
[task 2019-04-29T16:56:11.389Z]     # crashes in compiz). Make sure that X has enough time to start
[task 2019-04-29T16:56:11.389Z]     sleep 15
[task 2019-04-29T16:56:11.389Z]     # DISPLAY has already been set above
[task 2019-04-29T16:56:11.390Z]     # XXX: it would be ideal to add a semaphore logic to make sure that the
[task 2019-04-29T16:56:11.390Z]     # window manager is ready
[task 2019-04-29T16:56:11.390Z]     /etc/X11/Xsession 2>&1 &

Scope

Investigate if feasible to:

• reduce the sleep duration, or;
• eliminate entirely

Test Plan

Ensure the changes do not have adverse impact by running all linux jobs.

Baseline:
https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&resultStatus=pending%2Crunning%2Csuperseded%2Cusercancel%2Cretry%2Csuccess%2Ctestfailed%2Cbusted%2Cexception&classifiedState=unclassified&tier=1%2C2%2C3&group_state=expanded&revision=e8aebe488b2f2e567940577de25013d00e818f7c&selectedJob=243355619

elimination of sleep 15 clause

https://treeherder.mozilla.org/#/jobs?repo=try&group_state=expanded&revision=ac823a45a223844ead72f64b811f8a0112b69d8d&searchStr=linux&selectedJob=244588779

Current results are preliminary (not enough tests run), however it appears that eliminating the sleep 15 clause from test-linux.sh script has no adverse impact.

A sample test from each suite for every Linux platform has been run, and so far all results are green.

conclusion

At this point I believe it is safe to say that it is possible to eliminate this extraneous sleep to gain 15s back per chunk for all Linux platforms. This will add up to a significant time savings.

I believe the same script is used for all/many Android tests.

(In reply to Geoff Brown [:gbrown] from comment #2)

I believe the same script is used for all/many Android tests.

I'll be sure to check Android, just to ensure compatibility if I press ahead with the change.

https://hg.mozilla.org/mozilla-central/rev/0e9e744accd3

Backed out changeset 0e9e744accd3 (Bug 1548924) for causing leaks

Backout link: https://hg.mozilla.org/integration/autoland/rev/8bde97e9f598f2bea1c0182de7d0b7f52126c933

Push with failures: https://treeherder.mozilla.org/#/jobs?repo=autoland&resultStatus=success%2Ctestfailed%2Cbusted%2Cexception&searchStr=linux%2Cx64%2Casan%2Cmochitests%2Ctest-linux64-asan%2Fopt-mochitest-browser-chrome-e10s-3%2Cm%28bc3%29&revision=0e9e744accd3ed7b7b0ff2436904ed78a92bf9e2

Log link: https://treeherder.mozilla.org/logviewer.html#?job_id=245234416&repo=autoland

(In reply to Mihai Alexandru Michis [:malexandru] from comment #7)

Backed out changeset 0e9e744accd3 (Bug 1548924) for causing leaks

Backout link: https://hg.mozilla.org/integration/autoland/rev/8bde97e9f598f2bea1c0182de7d0b7f52126c933

Push with failures: https://treeherder.mozilla.org/#/jobs?repo=autoland&resultStatus=success%2Ctestfailed%2Cbusted%2Cexception&searchStr=linux%2Cx64%2Casan%2Cmochitests%2Ctest-linux64-asan%2Fopt-mochitest-browser-chrome-e10s-3%2Cm%28bc3%29&revision=0e9e744accd3ed7b7b0ff2436904ed78a92bf9e2

Log link: https://treeherder.mozilla.org/logviewer.html#?job_id=245234416&repo=autoland

From what I see, leaks for linux64-asan/opt-mochitest-browser-chrome-3 started much earlier than my push in autoland: https://treeherder.mozilla.org/#/jobs?repo=autoland&resultStatus=success%2Ctestfailed%2Cbusted%2Cexception&searchStr=linux64-asan%2Fopt-mochitest-browser-chrome-e10s-3&revision=f9833a7d765abb8eba66b95cfc183845db8e1b32

The above push is where I see the first instance of the leaks.

I will be pushing additional try runs with my changes as well as updated mozilla-central with no changes; I expect the leaks to be present in my baseline push, since it doesn't appear that my changes are related to the leaks in my opinion.

I relanded this

https://hg.mozilla.org/mozilla-central/rev/65a693623cee

(In reply to Mihai Alexandru Michis [:malexandru] from comment #10)

I relanded this

Thanks, if it causes the leaks to occur again then I'm more than happy to take another look.