957088 - GenerationalGC: Frequent test timeouts running mochitest browser chrome in debug builds

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jon Coppeard (:jonco)]

There are now frequent test timeouts when running the mochitest bc tests, for example:

https://tbpl.mozilla.org/php/getParsedLog.php?id=32526876&tree=Try

These are annoying because they can hide real test failures.

This is caused by the time spent in minor GC.  I've found that taking out the hashtable checking assertions fixes the issue, but I don't want to do this for all builds as I think these are really important to keep.

I've experimented with only running the checks when necessary (i.e. when the relevant tables had entries added) but this didn't totally fix the issue.

My current thought is to also add a configure flag to enable these assertions and only turn them on for shell builds.

Comment 1

[Steve Fink [:sfink] [:s:]]

The M-bc tests have also been bumping up against memory limits, especially with the Address Sanitizer builds, so there is some pressure to split them up already. I don't know the state of that, though. One of bug 931002, bug 927295, bug 939036, or bug 931340 ought to have more info.

Comment 2

[Steve Fink [:sfink] [:s:]]

The above bugs are useless. It's bug 819963, which looks like it'll happen before too long, though it's a bit of a mess.

Comment 3

[Jon Coppeard (:jonco)]

(In reply to Steve Fink [:sfink] from comment #2)

Unfortunately it's not just that the whole test suite is timing out - individual tests are timing out too.  So we need to do something more that just split the suite up.

Comment 4

[Jon Coppeard (:jonco)]

Attachment: zeal-mode-hashtable-checks

As suggested by Terrence, we can add a new zeal mode to enable these checks.  That way the fuzzers can use them without them slowing down out testing.

Comment 5

[Terrence Cole [:terrence]]

Comment on attachment 8360458 [details] [diff] [review]
zeal-mode-hashtable-checks

Review of attachment 8360458 [details] [diff] [review]:
-----------------------------------------------------------------

\o/ r=me

Comment 6

[Jon Coppeard (:jonco)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/dc286d6f6ef8

Comment 7

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/dc286d6f6ef8

Comment 8

[Jon Coppeard (:jonco)]

Oh, I broke windows builds (GGC only) with my use of "and" for "&&" in #if directives.

Comment 9

[Jon Coppeard (:jonco)]

Attachment: fix-windows-build-errors

Here's a patch to fix that.

Comment 10

[Terrence Cole [:terrence]]

Comment on attachment 8361805 [details] [diff] [review]
fix-windows-build-errors

Review of attachment 8361805 [details] [diff] [review]:
-----------------------------------------------------------------

r=me

Comment 11

[Jon Coppeard (:jonco)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/9a84eebfcd41

Comment 12

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/9a84eebfcd41