The generic, electrolysis and jaegermonkey mozconfigs for linux (didn't look at linux64) are using gcc 4.3.3 for linux in their mozconfigs. As they are (to my understanding) supposed to be mimicking mozilla-central, we should probably have them using the same compilers.

We fixed this for aurora in bug 658167.

The right fix is bug 558180, but we do need parity here in the interim.

Here's the list of mozconfigs that currently *do not* have gcc-4.5 specified, and are not from 2.0 or earlier branches. All of the branches based off m-c should be changing AFAIK, but can someone weigh-in on whether all of the android, maemo, and mobile branches should be using gcc-4.5 too? ./linux/electrolysis/debug/mozconfig ./linux/electrolysis/nightly/mozconfig ./linux/electrolysis/qt/mozconfig ./linux/electrolysis/unittest/mozconfig ./linux/electrolysis/xulrunner/mozconfig ./linux/generic/debug/mozconfig ./linux/generic/nightly/mozconfig ./linux/generic/qt/mozconfig ./linux/jaegermonkey/debug/mozconfig ./linux/jaegermonkey/nightly/mozconfig ./linux/mozilla-aurora/nightly-rpm/mozconfig ./linux/mozilla-aurora/unittest/mozconfig ./linux/mozilla-beta/codecoverage/mozconfig ./linux/mozilla-beta/debug/mozconfig ./linux/mozilla-beta/nightly/mozconfig ./linux/mozilla-beta/nightly-rpm/mozconfig ./linux/mozilla-beta/qt/mozconfig ./linux/mozilla-beta/release/mozconfig ./linux/mozilla-beta/xulrunner/mozconfig ./linux/mozilla-beta/xulrunner-qt/mozconfig ./linux/mozilla-central/nightly-rpm/mozconfig ./linux/mozilla-central/unittest/mozconfig ./linux/ux/nightly-rpm/mozconfig ./linux-android/electrolysis/nightly/mozconfig ./linux-android/generic/nightly/mozconfig ./linux-android/mozilla-aurora/nightly/mozconfig ./linux-android/mozilla-beta/nightly/mozconfig ./linux-android/mozilla-central/nightly/mozconfig ./linux-android/tracemonkey/nightly/mozconfig ./linux-android/try/nightly/mozconfig ./linux-android/ux/nightly/mozconfig ./linux-android-debug/electrolysis/nightly/mozconfig ./linux-android-debug/generic/nightly/mozconfig ./linux-android-debug/mozilla-aurora/nightly/mozconfig ./linux-android-debug/mozilla-beta/nightly/mozconfig ./linux-android-debug/mozilla-central/nightly/mozconfig ./linux-android-debug/tracemonkey/nightly/mozconfig ./linux-android-debug/try/nightly/mozconfig ./linux-android-debug/ux/nightly/mozconfig ./linux-maemo5-gtk/electrolysis/nightly/mozconfig ./linux-maemo5-gtk/generic/nightly/mozconfig ./linux-maemo5-gtk/mozilla-aurora/nightly/mozconfig ./linux-maemo5-gtk/mozilla-beta/nightly/mozconfig ./linux-maemo5-gtk/mozilla-central/nightly/mozconfig ./linux-maemo5-gtk/tracemonkey/nightly/mozconfig ./linux-maemo5-gtk/try/nightly/mozconfig ./linux-maemo5-gtk/ux/nightly/mozconfig ./linux-maemo5-qt/electrolysis/nightly/mozconfig ./linux-maemo5-qt/generic/nightly/mozconfig ./linux-maemo5-qt/mozilla-aurora/nightly/mozconfig ./linux-maemo5-qt/mozilla-beta/nightly/mozconfig ./linux-maemo5-qt/mozilla-central/nightly/mozconfig ./linux-maemo5-qt/tracemonkey/nightly/mozconfig ./linux-maemo5-qt/try/nightly/mozconfig ./linux-maemo5-qt/ux/nightly/mozconfig ./linux-mobile/electrolysis/nightly/mozconfig ./linux-mobile/generic/nightly/mozconfig ./linux-mobile/mozilla-aurora/nightly/mozconfig ./linux-mobile/mozilla-beta/nightly/mozconfig ./linux-mobile/mozilla-central/nightly/mozconfig ./linux-mobile/tracemonkey/nightly/mozconfig ./linux-mobile/try/nightly/mozconfig ./linux-mobile/ux/nightly/mozconfig ./linux64/electrolysis/nightly/mozconfig ./linux64/electrolysis/release/mozconfig ./linux64/generic/debug/mozconfig ./linux64/generic/nightly/mozconfig ./linux64/jaegermonkey/debug/mozconfig ./linux64/jaegermonkey/nightly/mozconfig ./linux64/mozilla-aurora/nightly-rpm/mozconfig ./linux64/mozilla-aurora/unittest/mozconfig ./linux64/mozilla-beta/codecoverage/mozconfig ./linux64/mozilla-beta/debug/mozconfig ./linux64/mozilla-beta/nightly/mozconfig ./linux64/mozilla-beta/nightly-rpm/mozconfig ./linux64/mozilla-beta/release/mozconfig ./linux64/mozilla-beta/xulrunner/mozconfig ./linux64/mozilla-central/nightly-rpm/mozconfig ./linux64/mozilla-central/unittest/mozconfig ./linux64/ux/nightly-rpm/mozconfig ./mobile/linux-i686/mobile-1.1/nightly/mozconfig ./mobile/linux-i686/mobile-1.1/release/mozconfig ./mobile/linux-i686/mobile-browser/nightly/mozconfig ./mobile/linux-i686/mobile-browser/release/mozconfig ./mobile/linux-i686/mobile-e10s/nightly/mozconfig ./mobile/linux-i686/mozilla-aurora/nightly/mozconfig ./mobile/linux-i686/mozilla-beta/nightly/mozconfig ./mobile/linux-i686/mozilla-beta/release/mozconfig

Desktop mozilla-beta needs to wait until we merge there in 6 weeks time.

Bug 660616 adds ./linux/jaegermonkey/qt/mozconfig

Would it be better to move the default gcc to gcc-4.5 and updating the mozconfigs for those stragglers that still need to use gcc-4.3 instead?

Being explicit can be helpful for people looking at the mozconfigs without having to jump in the logs but we can let others weigh in. aki, jhford can we switch to gcc-4.5 for mobile?

Maemo uses scratchbox so it uses a different set of compilers. I'm not entirely sure for Android -- we're cross compiling though, so I'd guess we're not using the standard compiler. Linux mobile desktop can most likely switch without issues, assuming it builds+runs afterwards.

JaegerMonkey's lucky, it can see that it's busted. It's tickling a gcc-4.3 bug on Linux64 which makes one test fail, and it would be wonderful if it could see that it actually is a gcc-4.3 bug in a way other than by merging to TraceMonkey (which is sort of scheduled to happen this weekish) and only then finding that it's still broken on 4.5 (though not on any developer's 4.5). Of course, it would also be wonderful to see whether or not it's busted on Linux PGO, which is on everywhere it's headed, and off where it is, too.

Comment on attachment 538180 [details] [diff] [review] unbust JaegerMonkey http://hg.mozilla.org/build/buildbot-configs/rev/5b5eecf69b70

I'm explicitly ignoring the following from the above list: * anything maemo * anything android * anything mozilla-beta * anything under mobile/ (linux-mobile/ *is* updated so mobile desktop builds will use the new compiler)

Comment on attachment 538262 [details] [diff] [review] Use gcc 4.5 on all m-c derived branches http://hg.mozilla.org/build/buildbot-configs/rev/2aade2ef4c0f

This made it into production today -- is there anything left ?

(In reply to comment #13) > This made it into production today -- is there anything left ? Don't think so, but I'll field dev complaints (if any).

(In reply to comment #14) > Don't think so, but I'll field dev complaints (if any). So far, they haven't figured out where to complain, but the combination of gcc-4.5 and not doing PGO causes bug 654858. At least build-system (the misplaced bug 604621 comment 16) and services-central (the misstarred bug 604621 comment 17 and the retriggerings in http://tbpl.mozilla.org/?tree=Services-Central&rev=ca188fd87727) have discovered that they're broken, and it's only a matter of time before they figure out how. Since mozilla-central does PGO on Linux, we probably want to be doing it on project branches anyway, at least until the plan to mostly do non-PGO and only occasionally do PGO comes up against this problem and has to fix it to move forward.

Silly tbpl.m.o, thinking things happen within 12 hours, make that the retriggerings in http://dev.philringnalda.com/tbpl/?tree=Services-Central&rev=ca188fd87727 that you can't see in tbpl.m.o.

It's just copy-paste, how badly could I mess that up? :) I used "1.9 vs" as my marker for things that needed to change, so if anyone didn't copy-paste that, I would have missed them.

(In reply to comment #15) > So far, they haven't figured out where to complain, but the combination of > gcc-4.5 and not doing PGO causes bug 654858. Bug 655003 is one possible place to complain. It reverts to -Os when PGO is not enabled. But I assume doing the same build on all branches is sensible.

Comment on attachment 538770 [details] [diff] [review] pgo for all http://hg.mozilla.org/build/buildbot-configs/rev/c8acaca6ba1e

Is needs‑reconfig? a proxy for needs-merge-to-production? here ?

(In reply to comment #20) > Is needs‑reconfig? a proxy for needs-merge-to-production? here ? Probably. Changing project branches to do PGO may raise some eyebrows, but it is just bringing it in line with m-c.

(In reply to comment #20) > Is needs‑reconfig? a proxy for needs-merge-to-production? here ? ...and Aki's merge has accomplished this now.

(In reply to comment #21) > Probably. Changing project branches to do PGO may raise some eyebrows, but > it is just bringing it in line with m-c. That'd be balanced by lowering the eyebrows of the branch owners who have had perma-orange Linux reftests for a while. Thanks for the fix!

Armen, is there anything here that still needs to be tracked for Firefox 6?

(In reply to comment #24) > Armen, is there anything here that still needs to be tracked for Firefox 6? All good. We just had to make sure it got fixed since we wanted all project branches to be synchronized with mozilla-central (see bug 658167).