856763 - Build and deploy gcc 4.5.4 on all linux/linux64 build slaves

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P2)

Description

[Ryan VanderMeulen [:RyanVM]]

We currently use gcc 4.5.2 as our default compiler for linux/linux64 builds. However, this is causing is frequent pain in bug 820796 as it leads to burned nightlies and PGO builds. In dev-platform, Mike suggested that we try upgrading to gcc 4.5.4 to see if the situation improves since gcc 4.7 still has some issues that prevent us from switching over to it.

Note that this bug is not currently about switching over to version 4.5.4, only that it be made available for testing.

Comment 1

[Ryan VanderMeulen [:RyanVM]]

As much as I love seeing hours of builder time get wasted every day due to bug 820796, can we please up the priority on this to see if it helps?

Comment 2

[Rail Aliiev [:rail]]

Sure, I can generate a package, but it would be great to have a person who can validate the binaries. maybe glandium? ;)

Comment 3

[Ryan VanderMeulen [:RyanVM]]

I think he already volunteered himself in the dev.platform thread about this :)

Comment 4

[Rail Aliiev [:rail]]

Attachment: gcc 4.5.4 spec file

Renames, cleanup and DRY.

Comment 5

[Rail Aliiev [:rail]]

Attachment: gcc 4.5.4 spec file

Oops, s/0moz4/0moz1/

Comment 6

[Rail Aliiev [:rail]]

Attachment: configs

Once I have the packages ready I'll test if they don't conflict.

Comment 7

[Justin Wood (:Callek)]

Comment on attachment 733707 [details] [diff] [review]
configs

Review of attachment 733707 [details] [diff] [review]:
-----------------------------------------------------------------

Only comment here is, since we now know 4.7 is broken why don't we just remove it while here? (not blocking)

Comment 8

[Justin Wood (:Callek)]

Comment on attachment 733704 [details] [diff] [review]
gcc 4.5.4 spec file

Review of attachment 733704 [details] [diff] [review]:
-----------------------------------------------------------------

As long as you test for no conflicts, r=me

Before switch we'll definitely need assurances that any compat changes (due to gcc switch) are understood and signed off on, and figure out if we can (and want to) limit updates based on that :-)

Comment 9

[Mike Hommey [:glandium]]

(In reply to Justin Wood (:Callek) from comment #7)
> Comment on attachment 733707 [details] [diff] [review]
> configs
> 
> Review of attachment 733707 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> Only comment here is, since we now know 4.7 is broken why don't we just
> remove it while here? (not blocking)

That would instantly break dxr and asan. And it's still useful.

Comment 10

[Justin Wood (:Callek)]

(In reply to Mike Hommey [:glandium] from comment #9)
> (In reply to Justin Wood (:Callek) from comment #7)
> > Only comment here is, since we now know 4.7 is broken why don't we just
> > remove it while here? (not blocking)
> 
> That would instantly break dxr and asan. And it's still useful.

Ahhh, was under the impression it was only there for try, so no worries then.

Comment 11

[Rail Aliiev [:rail]]

Comment on attachment 733704 [details] [diff] [review]
gcc 4.5.4 spec file

http://hg.mozilla.org/build/rpm-sources/rev/d5aafe2f63d4

Comment 12

[Rail Aliiev [:rail]]

Comment on attachment 733707 [details] [diff] [review]
configs

http://hg.mozilla.org/build/buildbot-configs/rev/78b3db20dd24

all 3 gcc packages live together very friendly :)

Comment 13

[Rail Aliiev [:rail]]

Will be available on try after next reconfig. /tools/gcc-4.5.4-0moz1 is the new prefix.

Comment 14

[[e-mail] (:bhearsum)]

in production

Comment 15

[Rail Aliiev [:rail]]

Al done here. Feel free to reopen if we need to change something in the paackage.