941314 - Be more aggressive about GCing and CCing in media mochitests

Status: RESOLVED FIXED

(Core :: Audio/Video, defect)

Description

[Chris Pearce [:cpearce (Not reading bugmail)]]

Attachment: Patch

We can be more aggressive about running the GC and CC in the media mochitests. This should help ensure our decoders clean up their resources, hopefully it will help reduce our random failure rate.

https://tbpl.mozilla.org/?tree=Try&rev=42111df3823b

Comment 1

[Matthew Gregan [:kinetik]]

It looks like you should use SpecialPowers.exactGC for this rather than inventing our own heuristics.

Comment 2

[Paul Adenot (:padenot)]

Comment on attachment 8335619 [details] [diff] [review]
Patch

Review of attachment 8335619 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/media/test/manifest.js
@@ +635,4 @@
>      SpecialPowers.forceGC();
> +    SpecialPowers.forceCC();
> +    SpecialPowers.forceGC();
> +    SpecialPowers.forceCC();

As kinetik says, exactGC is better for this use case, as advertised in the comment at the top:

http://mxr.mozilla.org/mozilla-central/source/testing/specialpowers/content/specialpowersAPI.js#1236

Comment 3

[Chris Pearce [:cpearce (Not reading bugmail)]]

Attachment: Patch v2

Using exactGC() instead.

Comment 4

[Matthew Gregan [:kinetik]]

Comment on attachment 8337439 [details] [diff] [review]
Patch v2

You need to pass a callback as the second argument (first argument (window) can be null) for this to work as it needs to idle the event queue.

With this in place, I wonder if we still need the extra GC in mediaTestCleanup.

Comment 5

[Chris Pearce [:cpearce (Not reading bugmail)]]

Attachment: Patch: Using SpecialPowers.exactGC()

Using SpecialPowers.exactGC. Tests pass locally, try run pending:

https://tbpl.mozilla.org/?tree=Try&rev=216d3988f7ec

Comment 6

[Chris Pearce [:cpearce (Not reading bugmail)]]

Try push: https://tbpl.mozilla.org/?tree=Try&rev=216d3988f7ec

The orange here are known issues.

Had the following failures:
Linux x64 Debug test_unseekable 3 timeouts in 39 runs (Bug 907924)

test_unseekable.html has only ever been logged as timing out 8 times, and only in the past 3 months, so that's a bit suspicious.

MacOSX 10.6 opt: 
* 2 x tests/content/html/content/test/test_bug660959-1.html | application crashed [@ libSystem.B.dylib + 0xd7a] (These crashes are known, we're hanging and forcing a crash to try to get a stack and failing)
* Bug 939633 - Intermittent test_bug419527.xhtml | Test timed out. (known)
* 76176 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/webgl/test_webgl_conformance_test_suite.html | [conformance/glsl/misc/shader-with-glprojectionmatrix.vert.html] Timeout in this test page

MacOSX 10.6 debug:
* 2x Bug 846769 (test_bug495300.html assertion failure) in about 40 runs
 ASSERTION: Start time must be before end time: 'startTime < endTime', file ../../../../content/media/ogg/OggReader.cpp, line 1154

But based on this graph, the memory usage is about the same:
http://people.mozilla.org/~sguo/mochimem/viewer.html?url=http%3A//ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-216d3988f7ec/try-macosx64-debug/try_snowleopard-debug_test-mochitest-1-bm75-tests1-macosx-build1658.txt.gz&url=http%3A//ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-5383aaa0f394/try-macosx64-debug/try_snowleopard-debug_test-mochitest-1-bm76-tests1-macosx-build1629.txt.gz

Here on Win7 debug the difference is more noticable (but is only a few tens of MB, but at least it's consistent):
http://people.mozilla.org/~sguo/mochimem/viewer.html?url=http%3A//ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-216d3988f7ec/try-win32-debug/try_win7-ix-debug_test-mochitest-1-bm73-tests1-windows-build1233.txt.gz&url=http%3A//ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-5383aaa0f394/try-win32-debug/try_win7-ix-debug_test-mochitest-1-bm74-tests1-windows-build908.txt.gz&

Comment 7

[Chris Pearce [:cpearce (Not reading bugmail)]]

Let's see if our orange profile over time changes...

https://hg.mozilla.org/integration/mozilla-inbound/rev/dee18b5520fb

Comment 8

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/dee18b5520fb