1147243 - Build memory/jemalloc in unified mode

Status: RESOLVED FIXED

(Core :: General, defect)

Comment 1

[(no longer active)]

Attachment: Build memory/jemalloc in unified mode

Comment 2

[Mike Hommey [:glandium]]

Comment on attachment 8582808 [details] [diff] [review]
Build memory/jemalloc in unified mode

Review of attachment 8582808 [details] [diff] [review]:
-----------------------------------------------------------------

::: memory/jemalloc/moz.build
@@ +38,4 @@
>  # Only OSX needs the zone allocation implementation,
>  # but only if replace-malloc is not enabled.
>  if CONFIG['OS_TARGET'] == 'Darwin' and not CONFIG['MOZ_REPLACE_MALLOC']:
>      SOURCES += [

Can you add this one too? (and push to try with ac_add_options --disable-replace-malloc and MOZ_JEMALLOC3=1 in build/mozconfig.common.override to test that it works)

I'd like to know if the various compilers add warnings as a result of this.

Comment 3

[(no longer active)]

See <https://treeherder.mozilla.org/#/jobs?repo=try&revision=01e6bd8d0c98>.  I'm not sure what I need to do to make sure DMD.h is available.  Do you know?

Comment 4

[Mike Hommey [:glandium]]

Don't build with MOZ_DMD=1, it's incompatible with --disable-replace-malloc (should error in configure instead of failing to build later, though)

Comment 5

[(no longer active)]

Well, the reason why I added it was that I was getting the same errors complaining about the lack of DMD.h without it.

Comment 6

[(no longer active)]

Any other suggestions?

Comment 7

[Mike Hommey [:glandium]]

The #include "DMD.h" in nsIMemoryReporter.idl/.h is under a #ifdef MOZ_DMD. So the only way that error happens is if you build with MOZ_DMD.

Comment 8

[(no longer active)]

(In reply to Mike Hommey [:glandium] from comment #7)
> The #include "DMD.h" in nsIMemoryReporter.idl/.h is under a #ifdef MOZ_DMD.
> So the only way that error happens is if you build with MOZ_DMD.

OK then I'm not sure what else to try...  Can we just take the patch as is?

Comment 9

[Mike Hommey [:glandium]]

Not if it breaks --disable-replace-malloc MOZ_JEMALLOC3 builds, which aren't hypothetical: that's what we'd get when jemalloc 3 is enabled by default on aurora.

Comment 10

[(no longer active)]

(In reply to Mike Hommey [:glandium] from comment #9)
> Not if it breaks --disable-replace-malloc MOZ_JEMALLOC3 builds, which aren't
> hypothetical: that's what we'd get when jemalloc 3 is enabled by default on
> aurora.

OK, makes sense.  I pushed again to try with the patch as is and got hit by the same issue.  <https://treeherder.mozilla.org/#/jobs?repo=try&revision=4cd11474f28a>.  I'm giving up.  :/

Comment 11

[Mike Hommey [:glandium]]

That build is not disabling DMD.

Comment 12

[(no longer active)]

(In reply to Mike Hommey [:glandium] from comment #11)
> That build is not disabling DMD.

What's the correct way to disable DMD?  Doesn't --disable-replace-malloc do that?

Comment 13

[Mike Hommey [:glandium]]

(In reply to :Ehsan Akhgari (not reading bugmail, needinfo? me!) from comment #12)
> (In reply to Mike Hommey [:glandium] from comment #11)
> > That build is not disabling DMD.
> 
> What's the correct way to disable DMD?  Doesn't --disable-replace-malloc do
> that?

Maybe it should, but it currently doesn't, and the in-tree mozconfigs explicitly enable it. So hit for that, remove --enable-dmd from the mozconfigs (or add --disable-dmd to mozconfig.common.override, assuming that's included after the other mozconfigs).

Comment 14

[(no longer active)]

Thanks, that finally made things work!

https://treeherder.mozilla.org/#/jobs?repo=try&revision=4ac0783c888e

Comment 15

[(no longer active)]

Attachment: Build memory/jemalloc in unified mode

Comment 16

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/fa8ccc8fb9ed

Comment 17

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

https://hg.mozilla.org/mozilla-central/rev/fa8ccc8fb9ed