1347654 - mochitest-browser-chrome jobs "leak" if run multiple times in automation

Status: RESOLVED DUPLICATE of bug 1391694

(Testing :: Mochitest, defect)

Description

[William Lachance (:wlach)]

In testing bug 1343327, it seems like running mochitest-browser-chrome tests repeatedly (i.e. using the `repeat` or `run until fail` options) causes reports of "leaks" in the harness. Examples:

e10s browser_ext_tabs_reload_bypass_cache.js

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f9f1ae807d384f42912547621443e46db90b00f7&group_state=expanded&selectedJob=83857318

e10s browser_async_remove_tab.js:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f9f1ae807d384f42912547621443e46db90b00f7&group_state=expanded&selectedJob=83758884

I have confirmed that just running them once does not exhibit this behaviour. Also plain mochitests run just fine (have not tried other variants yet).

Ted, sorry to bug you, but I'm not sure who to ask: have you seen anything like this before? Any ideas on what might be happening? If you're not sure, feel free to just cancel the NI and I'll try to investigate further myself.

Comment 1

[(not currently active) Ted Mielczarek]

This isn't something I've seen before, sorry. The only thing I could think is that maybe the repeat option doesn't properly clean up the test the same way a normal run does?

Comment 2

[William Lachance (:wlach)]

(In reply to Ted Mielczarek [:ted.mielczarek] from comment #1)
> This isn't something I've seen before, sorry. The only thing I could think
> is that maybe the repeat option doesn't properly clean up the test the same
> way a normal run does?

Yeah, I suspected the same when I first saw this, though I couldn't find any smoking guns when I looked at the code. Hopefully someone will be able to figure this out eventually...