815071 - Update jemalloc 3 to (mostly) 3.2

Status: RESOLVED FIXED

(Core :: Memory Allocator, defect)

Description

[Mike Hommey [:glandium]]

I'd like to update jemalloc3 to commit 9906660eb7365abb54e4495407ffddb1069ef654, which only differs from the 3.2.0 release with http://www.canonware.com/pipermail/jemalloc-discuss/2012-November/000517.html and another patch to the jemalloc profiling feature that we don't enable, so it's essentially a no-op patch.

Comment 1

[Mike Hommey [:glandium]]

Attachment: Update jemalloc3 to commit 9906660

Comment 2

[Mike Hommey [:glandium]]

Comment on attachment 685073 [details] [diff] [review]
Update jemalloc3 to commit 9906660

There's a bug in arenas_extend_ctl that makes it fail to build on windows, and build something very wrong on other platforms.

Comment 3

[Mike Hommey [:glandium]]

Attachment: Update jemalloc3 to commit 6eb84fb

This updates to commit 6eb84fbe315add1e1d4f8deedc25d260fff3ae97, but I want http://www.canonware.com/pipermail/jemalloc-discuss/2012-November/000527.html for bug 804303, and it's not committed upstream yet.

Comment 4

[Mike Hommey [:glandium]]

Comment on attachment 687667 [details] [diff] [review]
Update jemalloc3 to commit 6eb84fb

That can land without http://www.canonware.com/pipermail/jemalloc-discuss/2012-November/000527.html for bug 804303 for now. The only impact of not having that patch is there's a possibility of a dead-lock on fork() when building with --enable-replace-malloc *and* running with the libreplace_jemalloc.dylib lib on mac. And that patch actually introduces another more important problem with libreplace_jemalloc.dylib, so all in all, it's not that bad not to land with it.

Comment 5

[Mike Hommey [:glandium]]

http://hg.mozilla.org/integration/mozilla-inbound/rev/e6f7e0e1257b

Comment 6

[(no longer active)]

https://hg.mozilla.org/mozilla-central/rev/e6f7e0e1257b

Comment 7

[Mike Hommey [:glandium]]

Comment on attachment 687667 [details] [diff] [review]
Update jemalloc3 to commit 6eb84fb

[Approval Request Comment]
User impact if declined: required for bug 804303
Testing completed (on m-c, etc.): baked on m-i/m-c for a few days
Risk to taking this patch (and alternatives if risky): NPOTB
String or UUID changes made by this patch: none

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/e04eec5f7c6d

This didn't apply cleanly to beta, so I didn't push it there.

Comment 9

[Mike Hommey [:glandium]]

The base jemalloc version was a different one, but updating only requires changing upstream.info and running update.sh.
http://hg.mozilla.org/releases/mozilla-b2g18/rev/fc084e346eb5

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Good to know for the future, thanks!