Bug 128109 – Turn GenGC on

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Mark Hahnenberg]

GenGC looks like it's almost ready for primetime. We'll start by enabling it on X86-64.

[Mark Hahnenberg]

Created attachment 223008 [details]
Patch

[Oliver Hunt]

Comment on attachment 223008 [details]
Patch

\o/

[Build Bot]

Comment on attachment 223008 [details]
Patch

Attachment 223008 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/5653330361581568

New failing tests:
http/tests/security/contentSecurityPolicy/source-list-parsing-paths-01.html

[Build Bot]

Created attachment 223020 [details]
Archive of layout-test-results from webkit-ews-11 for mac-mountainlion-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: webkit-ews-11  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.5

[Geoffrey Garen]

Please file follow-up bugs for known performance issues. (I believe you mentioned a SunSpider regression to me in person.)

[Mark Hahnenberg]

(In reply to comment #5)
> Please file follow-up bugs for known performance issues. (I believe you mentioned a SunSpider regression to me in person.)

I filed bug 128129 for the sunspider regression.

[WebKit Commit Bot]

Comment on attachment 223008 [details]
Patch

Clearing flags on attachment: 223008

Committed r163330: <http://trac.webkit.org/changeset/163330>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Andreas Kling]

Looks like 3% regression on Membuster.

[Andreas Kling]

Filed bug 128160 and bug 128161 to track Membuster and DYEB regressions.

[Antti Koivisto]

PLT and html5 loading test agree this was ~3% performance regression.

[Csaba Osztrogonác]

(In reply to comment #7)
> (From update of attachment 223008 [details])
> Clearing flags on attachment: 223008
> 
> Committed r163330: <http://trac.webkit.org/changeset/163330>

It seems the GenGC has a serious bug, because from this change
there are many flaky failures on the Apple Mac bots:
- http://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK1%20%28Tests%29?numbuilds=100
- http://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK2%20%28Tests%29?numbuilds=100
- http://build.webkit.org/builders/Apple%20MountainLion%20Release%20WK1%20%28Tests%29?numbuilds=100

Apple Mac EWS bots are broken too because of this regression.

Could you fix it as soon as possible please?

[Mark Hahnenberg]

(In reply to comment #12)
> (In reply to comment #7)
> > (From update of attachment 223008 [details] [details])
> > Clearing flags on attachment: 223008
> > 
> > Committed r163330: <http://trac.webkit.org/changeset/163330>
> 
> It seems the GenGC has a serious bug, because from this change
> there are many flaky failures on the Apple Mac bots:
> - http://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK1%20%28Tests%29?numbuilds=100
> - http://build.webkit.org/builders/Apple%20Mavericks%20Release%20WK2%20%28Tests%29?numbuilds=100
> - http://build.webkit.org/builders/Apple%20MountainLion%20Release%20WK1%20%28Tests%29?numbuilds=100
> 
> Apple Mac EWS bots are broken too because of this regression.
> 
> Could you fix it as soon as possible please?

My guess is that these flaky tests depend on some object being finalized before the test times out. Technically this is a bad assumption because garbage collection isn't designed to run finalizers before a hard deadline. OTOH, it worked before so we should be able to make it work again. I'm working on a simple patch to fix this.