GCC 3.4.4 is released!

Message

wing · Post by **wing** » Fri May 20, 2005 3:21 am

Yup, released yesterday. Well, I'm gonna see if I can bootstrap this one (with FSF Binutils-2.16 and glibc-2.3.5), I'm compiling it now, then I'm gonna do an emerge -e world.

Anyways to those who want to know, here you will find all bugs fixed by gcc-3.4.4, yup, it's just another bugfix release!

Here it is:
ftp://ftp.gnu.org/gnu/gcc/gcc-3.4.4/

though you may feel kind enough to check out the GCC mirrors list.

Have fun with the release!

nxsty · Post by **nxsty** » Fri May 20, 2005 3:08 pm

Finally! I hope it gets into the tree soon. This release contains a lot of regression fixes backported from 4.0 so it's probably very stable. This also ends the need to use the semi broken 20050110 snapshot for the updated symbol visibility patches.

G2k · Post by **G2k** » Fri May 20, 2005 4:38 pm

wing wrote:Yup, released yesterday. Well, I'm gonna see if I can bootstrap this one (with FSF Binutils-2.16 and glibc-2.3.5), I'm compiling it now, then I'm gonna do an emerge -e world.

why the hell would you emerge -e world? it's not worth it...

wing · Post by **wing** » Fri May 20, 2005 5:51 pm

G2k wrote:
wing wrote:Yup, released yesterday. Well, I'm gonna see if I can bootstrap this one (with FSF Binutils-2.16 and glibc-2.3.5), I'm compiling it now, then I'm gonna do an emerge -e world.
why the hell would you emerge -e world? it's not worth it...

If I did multiple emerge -e world on gcc-4.0.1 and glibc-2.3.6 snapshots, I can certainly do them on this one. 'sides, my machine is fast enough to only take around a day or so to do ~500 packages.

Rainmaker · Post by **Rainmaker** » Fri May 20, 2005 10:39 pm

what kind of machine is that??

emerge -eDp world | genlop --pretend

...

Estimated update time: 2 days, 14 hours, 53 minutes.

emerge -eDp world | wc -l
884

This is minus +-10 lines for the "normal" output, so about 850 packages.

G2k · Post by **G2k** » Fri May 20, 2005 10:51 pm

you won't really notice any considerable changes if you're the average desktop user...you'd have to be one of the google guys or something to notice speed increases and whatnot.

Arainach · Post by **Arainach** » Sat May 21, 2005 2:22 am

Change compared to what? gcc 3.4.3-r1 over 3.3.4 feels like a hardware upgrade. Switching from plain 3.4.3-r1 to 3.4.3-r1 built with 3.4.3-r1 reduced my compile times by an additional 10-25% on huge ebuilds (like the toolkit). As far as an emerge -e world.... I'll probably do it over a weekend when I have a lot of excess time. That shouldn't have a HUGE increase, but anything helps.

bollucks · Post by **bollucks** » Sat May 21, 2005 5:35 am

Arainach wrote:Switching from plain 3.4.3-r1 to 3.4.3-r1 built with 3.4.3-r1 reduced my compile times by an additional 10-25% on huge ebuilds (like the toolkit).

I don't claim to understand the ebuild system which may stuff this process up, but when gcc is built it rebuilds itself with itself and that is the final binary... ie gcc 3.4.3 when built with any other gcc ends up building itself again with gcc 3.4.3..

rhill · Post by **rhill** » Sat May 21, 2005 7:34 am

Arainach wrote:Switching from plain 3.4.3-r1 to 3.4.3-r1 built with 3.4.3-r1 reduced my compile times by an additional 10-25% on huge ebuilds (like the toolkit).

are you sure that's not ccache? that's a ridiculous amount of improvement.

I don't claim to understand the ebuild system which may stuff this process up, but when gcc is built it rebuilds itself with itself and that is the final binary... ie gcc 3.4.3 when built with any other gcc ends up building itself again with gcc 3.4.3..

gcc builds in three stages. it builds itself with the system compiler, then uses that compiler to build itself again with profile-generation, then uses that profile information to build the final compiler (make profiledbootstrap). the currently popular idea is that in the first part, the new compiler built with the old won't be as "advanced" as if it was the new being built with the new, so the general philosophy is to build it twice. whether this has any affect or not i don't know, but if it does it's probably pretty small. i used to believe it but now i'm actually not that convinced it's any better - especially these days where later versions are proving to be worse performers.

G2k · Post by **G2k** » Sat May 21, 2005 7:39 am

Arainach wrote:gcc 3.4.3-r1 over 3.3.4 feels like a hardware upgrade.

...uh...sure it does

rhill · Post by **rhill** » Sat May 21, 2005 8:44 am

actually it does.

StringCheesian · Post by **StringCheesian** » Sun May 22, 2005 8:29 pm

Does 3.4.4 fix any bugs that the 3.4.3 ebuild doesn't already apply patches for?

There's a Gentoo bugzilla entry for 3.4.4, but it seems to have been ignored... What could be taking so long?

nxsty · Post by **nxsty** » Sun May 22, 2005 9:08 pm

StringCheesian wrote:Does 3.4.4 fix any bugs that the 3.4.3 ebuild doesn't already apply patches for?

There's a Gentoo bugzilla entry for 3.4.4, but it seems to have been ignored... What could be taking so long?

Yes a lot, check the list on gcc.gnu.org. But it also deprecates several patches.

wing · Post by **wing** » Sun May 22, 2005 11:19 pm

these are the only packages that failed

Code: Select all

media-libs/faad2-2.0-r6
media-libs/faac-1.24
media-plugins/gst-plugins-faac-0.8.8

So there you are! Only 3 out of 500 packages from a gnome-centric system.

johntramp · Post by **johntramp** » Mon May 23, 2005 8:15 pm

lol I am still using 3.3.5. Maybe it is time to upgrade.

nxsty · Post by **nxsty** » Tue May 24, 2005 7:53 am

And now it's finally in portage!

I'm going to wait untill it hits unstable before I upgrade.

wing · Post by **wing** » Wed May 25, 2005 3:49 pm

hooray! I wondered where -fvisibility-inlines-hidden went!

johntramp · Post by **johntramp** » Wed May 25, 2005 8:17 pm

nxsty wrote: I'm going to wait untill it hits unstable before I upgrade.

eh'?

rhill · Post by **rhill** » Thu May 26, 2005 4:26 am

it's hard-masked

wing · Post by **wing** » Thu May 26, 2005 5:51 am

dirtyepic wrote:it's hard-masked

Yeah, and this also presented a problem to me when I tried to emerge the official gcc-3.4.4 ebuild. I couldn't figure out a filter for package.mask that would block the gcc-4.0 snapshots, so I just copied it to my overlay.

StringCheesian · Post by **StringCheesian** » Fri May 27, 2005 8:50 am

Adding this to /etc/portage/package.keywords will allow 3.4.4 but not 4.x.x:

Code: Select all

# Unmask every gcc version less than 4:
<sys-devel/gcc-4 -*

I don't understand why they mask some packages with KEYWORDS="-*" in the ebuild and others using package.mask... it can be confusing when trying to unmask something

An entire ~x86 system compiled from stage 1 (including KDE 3.4.0) with gcc 3.4.4 (using the ebuild in portage) is working fine here, BTW. klineakconfig segfaults, but I don't know if that's gcc's fault. Other than that no problems whatsoever. I don't see why it's masked...

didl · Post by **didl** » Fri May 27, 2005 10:38 am

StringCheesian wrote: I don't see why it's masked...

I do

Just because it works on your box, with your version of glibc, binutils, ....
doesn't mean it is ready for general use. There are quite a lot of things that
need testing which takes time. Thorough testing of gcc is extremely crucial
for a source based distro like Gentoo, hence I'd gladly wait a few more weeks
rather than having to resurrect an unusable box lateron.

StringCheesian · Post by **StringCheesian** » Fri May 27, 2005 1:18 pm

didl wrote:Just because it works on your box, with your version of glibc, binutils, ....
doesn't mean it is ready for general use.

Oh yeah. But isn't that what unstable is for? Stuff that's thought to work but unproven?

Maedhros · Post by **Maedhros** » Fri May 27, 2005 1:31 pm

StringCheesian wrote:Oh yeah. But isn't that what unstable is for? Stuff that's thought to work but unproven?

Nope, it's for ebuilds that are thought to work, but are unproven.

johntramp · Post by **johntramp** » Fri May 27, 2005 7:46 pm

Maedhros wrote:
StringCheesian wrote:Oh yeah. But isn't that what unstable is for? Stuff that's thought to work but unproven?
Nope, it's for ebuilds that are thought to work, but are unproven.

wow I didn't know that. And so hard mask is for unstable apps?

GCC 3.4.4 is released!

GCC 3.4.4 is released!

Re: GCC 3.4.4 is released!

Re: GCC 3.4.4 is released!