Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
Mike Frysinger wrote: On Thursday 10 April 2008, Mike Frysinger wrote: gcc-4.3 seems to be standing up well. since the major gcc-ebuild-specific issues seem to be resolved now, i'll probably do a sweep of bugs to see if there's any patches i'm missing (if you guys know of a bug that should be addressed specifically in the gcc-4.3.0 ebuild, speak up now). then move on to the gcc 4.3 tracker bug (#198121). once this gets below a certain critical mass (i wont know what the critical mass is until it's been de-attained), then we'll be ~arching things. people are recommended to do a quick sweep of the lower hangers (many bugs have simple patches). i'll drop in ~ppc ~amd64 ~x86 as i use those every day. if any other arch is happy now with things, add your ~arch to the commented out list in cvs so i know to include it. the x86 cld revert is in now as well as some more upstream pr fixes. i'll probably let things settle for this week and pending any craziness, move gcc-4.3.0-r1 into ~arch in a week. i'll prob commit some "obvious" gcc-4.3 bugs at the same time. last chance to speak up peeps. -mike Poke. Gimme some gcc 4.3 goodness (besides the fact that I'm using it). But let's slap ~arch with it. -- gentoo-dev@lists.gentoo.org mailing list
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
On Thursday 10 April 2008, Mike Frysinger wrote: > gcc-4.3 seems to be standing up well. since the major gcc-ebuild-specific > issues seem to be resolved now, i'll probably do a sweep of bugs to see if > there's any patches i'm missing (if you guys know of a bug that should be > addressed specifically in the gcc-4.3.0 ebuild, speak up now). > > then move on to the gcc 4.3 tracker bug (#198121). once this gets below a > certain critical mass (i wont know what the critical mass is until it's > been de-attained), then we'll be ~arching things. people are recommended > to do a quick sweep of the lower hangers (many bugs have simple patches). > > i'll drop in ~ppc ~amd64 ~x86 as i use those every day. if any other arch > is happy now with things, add your ~arch to the commented out list in cvs > so i know to include it. the x86 cld revert is in now as well as some more upstream pr fixes. i'll probably let things settle for this week and pending any craziness, move gcc-4.3.0-r1 into ~arch in a week. i'll prob commit some "obvious" gcc-4.3 bugs at the same time. last chance to speak up peeps. -mike signature.asc Description: This is a digitally signed message part.
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
On Friday 11 April 2008, Vaeth wrote: > > So the point is, our current 2.6.24 kernel is safe. > > I can *not* confirm this. Just some days ago, I compiled > hardened-sources-2.6.24 (which uses genpatches-2.6.24-5; > current gentoo-sources uses genpatches-2.6.24-6, but the > difference is obviously not important here [it involves > just an #include for some exotic hardwar]). > > Result: Compiles fine with gcc-4.3 on x86 but dies immediately > at boot (before printing anything) unless acpi=off is used. > (And just to be sure, I disabled every acpi feature except > "general" acpi support - same result). > > It is certainly a compiler problem, because with gcc-4.2 > exactly the same kernel configuration (and even on amd64 > with gcc-4.3 and analogous configuration) has no problem. doesnt sound like a hardened specific problem, nor the string direction problem we've been talking about. the string problem we've been discussing affects userspace *only*. if it fails to reach userspace, your bug is clearly not this bug. -mike signature.asc Description: This is a digitally signed message part.
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
Vaeth wrote: Result: Compiles fine with gcc-4.3 on x86 but dies immediately at boot (before printing anything) unless acpi=off is used. (And just to be sure, I disabled every acpi feature except "general" acpi support - same result). Please file a bug at bugs,gentoo.org, our hardened team surely wants to know about that. Cheers, -jkt -- cd /local/pub && more beer > /dev/mouth signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
> So the point is, our current 2.6.24 kernel is safe. I can *not* confirm this. Just some days ago, I compiled hardened-sources-2.6.24 (which uses genpatches-2.6.24-5; current gentoo-sources uses genpatches-2.6.24-6, but the difference is obviously not important here [it involves just an #include for some exotic hardwar]). Result: Compiles fine with gcc-4.3 on x86 but dies immediately at boot (before printing anything) unless acpi=off is used. (And just to be sure, I disabled every acpi feature except "general" acpi support - same result). It is certainly a compiler problem, because with gcc-4.2 exactly the same kernel configuration (and even on amd64 with gcc-4.3 and analogous configuration) has no problem. -- gentoo-dev@lists.gentoo.org mailing list
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
Mike Frysinger wrote: there is no compile time problem. it's all runtime. i still think carrying the patch until gcc-4.3 goes stable is OK. 1400_prevent-gcc43-optimization-udivdi3.patch fixes compilation issue, another patch (already in some 2.6.24.x, I guess) fixes direction flag (that well-known runtime bug). So the point is, our current 2.6.24 kernel is safe. Cheers, -jkt -- cd /local/pub && more beer > /dev/mouth signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
On Thursday 10 April 2008, Jan Kundrát wrote: > Donnie Berkholz wrote: > > Presuming you're adding the direction-flag patch to 4.3.0 so it doesn't > > break people on a kernel earlier than 2.6.25? > > gentoo-sources-2.6.24-r4 has that patch, at least when looking at the > changelog. Or is it just for compile-time borkage and not for the > direction flag cleaning? there is no compile time problem. it's all runtime. i still think carrying the patch until gcc-4.3 goes stable is OK. -mike signature.asc Description: This is a digitally signed message part.
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
Donnie Berkholz wrote: Presuming you're adding the direction-flag patch to 4.3.0 so it doesn't break people on a kernel earlier than 2.6.25? gentoo-sources-2.6.24-r4 has that patch, at least when looking at the changelog. Or is it just for compile-time borkage and not for the direction flag cleaning? Cheers, -jkt -- cd /local/pub && more beer > /dev/mouth signature.asc Description: OpenPGP digital signature
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
On Thursday 10 April 2008, Mike Frysinger wrote: > Also, you'll have to provide a URL to said change. i havent seen a > patch for it in my random driftings on the interweb. > -mike I was just researching the issue, so had this handy: http://gcc.gnu.org/ml/gcc-patches/2008-03/msg00417.html -- /PA -- gentoo-dev@lists.gentoo.org mailing list
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
On Thursday 10 April 2008, Donnie Berkholz wrote: > On 02:57 Thu 10 Apr , Mike Frysinger wrote: > > gcc-4.3 seems to be standing up well. since the major > > gcc-ebuild-specific issues seem to be resolved now, i'll probably do a > > sweep of bugs to see if there's any patches i'm missing (if you guys know > > of a bug that should be addressed specifically in the gcc-4.3.0 ebuild, > > speak up now). > > Presuming you're adding the direction-flag patch to 4.3.0 so it doesn't > break people on a kernel earlier than 2.6.25? I didn't see it in a quick > glance at the patch tarball. i had no plans to revert the behavior in question. i could be persuaded to carry such a patch though until 2.6.25 goes stable and gcc-4.3 goes stable. presumably the time frame of both of those should be "long enough". also, you'll have to provide a URL to said change. i havent seen a patch for it in my random driftings on the interweb. -mike signature.asc Description: This is a digitally signed message part.
Re: [gentoo-dev] gcc-4.2 / gcc-4.3 plans
On 02:57 Thu 10 Apr , Mike Frysinger wrote: > gcc-4.3 seems to be standing up well. since the major gcc-ebuild-specific > issues seem to be resolved now, i'll probably do a sweep of bugs to see if > there's any patches i'm missing (if you guys know of a bug that should be > addressed specifically in the gcc-4.3.0 ebuild, speak up now). Presuming you're adding the direction-flag patch to 4.3.0 so it doesn't break people on a kernel earlier than 2.6.25? I didn't see it in a quick glance at the patch tarball. Thanks, Donnie -- gentoo-dev@lists.gentoo.org mailing list
[gentoo-dev] gcc-4.2 / gcc-4.3 plans
i dont plan on suggesting any gcc-4.2 version for stable. arches of course are free to determine if gcc-4.2 works better for them than gcc-4.1 and thus move to stable. gcc-4.3 seems to be standing up well. since the major gcc-ebuild-specific issues seem to be resolved now, i'll probably do a sweep of bugs to see if there's any patches i'm missing (if you guys know of a bug that should be addressed specifically in the gcc-4.3.0 ebuild, speak up now). then move on to the gcc 4.3 tracker bug (#198121). once this gets below a certain critical mass (i wont know what the critical mass is until it's been de-attained), then we'll be ~arching things. people are recommended to do a quick sweep of the lower hangers (many bugs have simple patches). i'll drop in ~ppc ~amd64 ~x86 as i use those every day. if any other arch is happy now with things, add your ~arch to the commented out list in cvs so i know to include it. -mike signature.asc Description: This is a digitally signed message part.