419648 - tryserver win32 builds fail since bug 415928 was landed

Status: RESOLVED DUPLICATE of bug 415928

(Release Engineering :: General, defect)

Description

[[e-mail] (:bhearsum)]

Ted suspects that because the try servers do not --enable-jemalloc that non-jemalloc win32 builds are broken. I'm trying a build with --enable-jemalloc to test that theory.

Either way, this has brought up an important issue: the try server uses different mozconfig files than the nightlies. I think we should make it checkout the nightly mozconfig files and use those. There's no reason I can think of to have it use it's own.

Comment 1

[[e-mail] (:bhearsum)]

Nick points out that there's some differences between these configs. For one, RUN_AUTOCONF_LOCALLY=1 is in the tryserver configs -- this is pretty important. Something like that _could_ be set as an env variable, but I don't think that's the right way to go about this.

Maybe we just have to deal with keeping these configs in sync.

Comment 3

[[e-mail] (:bhearsum)]

Adding --enable-jemalloc to the tryserver config didn't fix the problem. There's a bunch of other differences between the two -- I'm going to try a build using the nightly mozconfig next.

Comment 4

[[e-mail] (:bhearsum)]

This build didn't work either.

Gavin landed a fix in bug 415928 that Ted thinks will fix this. There's a build spinning right now to test it.

Comment 5

[[e-mail] (:bhearsum)]

Latest win32 build built fine.

Comment 6

[:Gavin Sharp [email: [e-mail]]]

(In reply to comment #1)
> Nick points out that there's some differences between these configs. For one,
> RUN_AUTOCONF_LOCALLY=1 is in the tryserver configs -- this is pretty important.

> Maybe we just have to deal with keeping these configs in sync.

The tryserver mozconfigs are likely to always contain a superset of the settings in the nightly config though, right? Could you just checkout+include the nightly mozconfig in the tryserver config?

Comment 7

[[e-mail] (:bhearsum)]

Yeah, that's a good idea. I totally forgot we could do that.