GCC 4.3.0 testing!

Message

DualCpuUser · Post by **DualCpuUser** » Mon Dec 18, 2006 10:16 am

nbvcxz wrote:Has anybody succeeded in building system with 4.3.0 - mine always fails at glibc / btw any important news in 20061216 except of ppc bugfix?

Yeah my whole system, minus firefox and a cpl of others, but all of qt/kde/beryl works.

nbvcxz · Post by **nbvcxz** » Mon Dec 18, 2006 11:06 am

@Ron (DualCpuUser): could You show your emerge --info?

Gentree · Post by **Gentree** » Mon Dec 18, 2006 11:37 am

20061216:

a few failures includeing glib. All seem to be basically the same thing. Looks like more strict systax requirements.

Code: Select all

timezone fails
zic.o:zic.c:(.text+0x3c0): first defined here
scheck.o: In function `__sigismember':
scheck.c:(.text+0x440): multiple definition of `__sigismember'

gzip -1.3.9 fails
/usr/include/sys/types.h:90: error: two or more data types in declaration speci$

checking for socklen_t... no
checking for socklen_t equivalent... configure: error: Cannot find a type to us$

!!! Please attach the following file when filing a report to bugs.gentoo.org:
!!! /usr/portage/tmp/portage/sys-apps/coreutils-6.2/work/coreutils-6.2/config.l$

!!! ERROR: sys-apps/coreutils-6.2 failed.


i686-pc-linux-gnu-gcc -std=gnu99 -I. -I. -march=athlon-xp -O2 -pipe -fomit-fram$
{standard input}: Assembler messages:
{standard input}:1478: Error: symbol `fstatat64' is already defined
{standard input}:1496: Error: symbol `fstat64' is already defined

!!! ERROR: net-misc/rsync-2.6.9-r1 failed.

/s_signbit.o
../sysdeps/ieee754/dbl-64/s_signbit.c:27: error: redefinition of '__signbit'
../sysdeps/i386/fpu/bits/mathinline.h:135: error: previous definition of '__sig$
make[2]: *** [/usr/portage/tmp/portage/sys-libs/glibc-2.5/work/build-default-i6$
make[2]: *** Waiting for unfinished jobs....

R=\"/var/empty\" -DSSH_RAND_HELPER=\"/usr/lib/misc/ssh-rand-helper\" -DHAVE_CON$
{standard input}: Assembler messages:
{standard input}:1095: Error: symbol `fstatat64' is already defined
{standard input}:1113: Error: symbol `fstat64' is already defined

!!! ERROR: net-misc/openssh-4.5_p1 failed.

weedy · Post by **weedy** » Mon Dec 18, 2006 2:43 pm

Code: Select all

/var/tmp/portage/sys-libs/glibc-2.5.90.20061214/work/build-default-i686-pc-linux-gnu-nptl/math/s_isinff.o
../sysdeps/ieee754/dbl-64/s_signbit.c:27: error: redefinition of '__signbit'
../sysdeps/i386/fpu/bits/mathinline.h:135: error: previous definition of '__signbit' was here
make[2]: *** [/var/tmp/portage/sys-libs/glibc-2.5.90.20061214/work/build-default-i686-pc-linux-gnu-nptl/math/s_signbit.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory `/var/tmp/portage/sys-libs/glibc-2.5.90.20061214/work/glibc-2.5/math'
make[1]: *** [math/subdir_lib] Error 2
make[1]: Leaving directory `/var/tmp/portage/sys-libs/glibc-2.5.90.20061214/work/glibc-2.5'
make: *** [all] Error 2

no problem with 4.1.1 :/

nbvcxz · Post by **nbvcxz** » Mon Dec 18, 2006 4:48 pm

@weedy: this seems to be the same error I got trying to emerge glibc with 4.3.0

Gentree · Post by **Gentree** » Mon Dec 18, 2006 5:28 pm

redefinition of '__signbit'

this post explains the cause and indicates a patch exists if anyone is interested/impatient:

http://www.nabble.com/Error-when-re-com ... 42237.html

DualCpuUser · Post by **DualCpuUser** » Mon Dec 18, 2006 9:43 pm

This is where using my method is a pita.

I update my svn sources daily, export the tree, FIX*** WANT_.... in gcc/c-decl.c to ZERO, also I fix BASE version to 4.3.98, so as not to have any directories collide.

Code: Select all

/* Set this to 1 if you want the standard ISO C99 semantics of 'inline'
   when you specify -std=c99 or -std=gnu99, and to 0 if you want
   behaviour compatible with the nonstandard semantics implemented by
   GCC 2.95 through 4.2.  */
#define WANT_C99_INLINE_SEMANTICS 0

then you tar.bz2 up the directory and stuff it in your distfiles, then I re-digest it turn on test in features then test it. if its the same tests results or less I install it.

neuron · Post by **neuron** » Mon Dec 18, 2006 11:52 pm

anyone done any performance comparisons between stable and 4.3?

The new vectorizer work looks interesting, only thing I really need more performance in is wine though, really hoping for a noticable performance increase

Joaquin24 · Post by **Joaquin24** » Tue Dec 19, 2006 12:10 am

soo, the time has come to switch to gcc-4.3 thnxs for the good news!!!, don't know why gcc-4.1 is still shouting out about "moved libs", when they have never been moved...

edited:
gcc-4.3 didn't compile here
AMD +1700
256MB + 512 SWAP

The new vectorizer work looks interesting,

Has the 4.1 already been patched, with this new fixes from 4.3 ?

this is my make.conf and ...

Code: Select all

CFLAGS="-march=athlon-xp -O2 -pipe -ftree-vectorize"
CHOST="i686-pc-linux-gnu"
CXXFLAGS="${CFLAGS} -fvisibility-inlines-hidden"
MAKEOPTS=""
ACCEPT_KEYWORDS=""
#USE="  X alsa arts avi berkdb bitmap-fonts cairo cdr cli crypt cups dbus dlloa$USE="  X alsa arts avi berkdb bitmap-fonts cairo cdr cli crypt cups dbus dlload$FEATURES=""
INPUT_DEVICES="keyboard mouse"
VIDEO_CARDS="nv nvidia vesa"
PORTAGE_NICENESS=8
LINGUAS="es"
PORTDIR_OVERLAY="/usr/local/portage /usr/local/portage/xeffects/trunk /usr/loca$PORTAGE_ELOG_CLASSES="log error warn"
PORTAGE_ELOG_MAILURI="------------------------"
ALSA_CARDS="via82xx"

I am re-emerging all gnome, (had mixed versions 2.16 & 2.14), no problem at all, till now..

FireBurn · Post by **FireBurn** » Tue Dec 19, 2006 8:22 pm

Could someone please make a simple overlay with a working tool chain that can compile a glibc binutils and gcc and then recompile it

Thanks

Mike

neuron · Post by **neuron** » Tue Dec 19, 2006 8:34 pm

Joaquin24 wrote:
The new vectorizer work looks interesting,
Has the 4.1 already been patched, with this new fixes from 4.3 ?

4.1 has old stuff, 4.3 has far more work on it. That said, you probably dont want to enable it globally, it has known conflict with some stuff.

I'm semi considering patching glibc's ebuild to enable it personally though, dunno how much gain I'll get though

Gentree · Post by **Gentree** » Tue Dec 19, 2006 10:02 pm

Joaquin24 wrote:soo, the time has come to switch to gcc-4.3 thnxs for the good news!!!, don't know why gcc-4.1 is still shouting out about "moved libs", when they have never been moved...

edited:
gcc-4.3 didn't compile here
AMD +1700
256MB + 512 SWAP

The new vectorizer work looks interesting,
Has the 4.1 already been patched, with this new fixes from 4.3 ?

this is my make.conf and ...
Code: Select all
CFLAGS="-march=athlon-xp -O2 -pipe -ftree-vectorize"
CHOST="i686-pc-linux-gnu"
CXXFLAGS="${CFLAGS} -fvisibility-inlines-hidden"
MAKEOPTS=""
ACCEPT_KEYWORDS=""
#USE="  X alsa arts avi berkdb bitmap-fonts cairo cdr cli crypt cups dbus dlloa$USE="  X alsa arts avi berkdb bitmap-fonts cairo cdr cli crypt cups dbus dlload$FEATURES=""
INPUT_DEVICES="keyboard mouse"
VIDEO_CARDS="nv nvidia vesa"
PORTAGE_NICENESS=8
LINGUAS="es"
PORTDIR_OVERLAY="/usr/local/portage /usr/local/portage/xeffects/trunk /usr/loca$PORTAGE_ELOG_CLASSES="log error warn"
PORTAGE_ELOG_MAILURI="------------------------"
ALSA_CARDS="via82xx"
                       
I am re-emerging all gnome, (had mixed versions 2.16 & 2.14), no problem at all, till now..

"AMD +1700" is that an athlon or an athlon-xp ? If it's an earlier athlon you should be using that in CFLAGS.

Also dont forget this is alpha release still. Dont dive straight in with new optimisations like tree-vectorize, just get the compiler build as a first step.

BTW if you want more speed get youself a decent CPU

I recently got a athon-xp mobile 1600+ off ebay for 20 euros and it's quite happy running at 2.2GHz

You're likely to get an order of magnitude more from o/c-ing an athlon then you will out of compiler tweeks.

Nothing to stop you doing both of course.

I'm semi considering patching glibc's ebuild to enable it personally though, dunno how much gain I'll get though

If you can locate a patch to get it to compile glibc pls post a link. I'd like to have a stable toolchain before I start recompiling world.

baigsabeeh · Post by **baigsabeeh** » Wed Dec 27, 2006 4:50 am

I might go for GCC 4.3.0 and recompile my system with it. It's been pretty slow lately anyway, couldn't hurt.

neuron · Post by **neuron** » Wed Dec 27, 2006 12:29 pm

baigsabeeh wrote:I might go for GCC 4.3.0 and recompile my system with it. It's been pretty slow lately anyway, couldn't hurt.

ohhh yes it can

, good if your bored though

, personally I'm testing that only in chroot.

DualCpuUser · Post by **DualCpuUser** » Wed Dec 27, 2006 12:32 pm

I was thinking about doing a chroot too, but what's the destructions on setting it up? Basically just another partition like gentoo install??

Well as it sits if anything get hosed, I boot into another partition install and chroot into my main one to fix things.

Firefox won't build, haven't cared enough to figure out why.

neuron · Post by **neuron** » Wed Dec 27, 2006 12:58 pm

DualCpuUser wrote:I was thinking about doing a chroot too, but what's the destructions on setting it up? Basically just another partition like gentoo install??

Well as it sits if anything get hosed, I boot into another partition install and chroot into my main one to fix things.

Firefox won't build, haven't cared enough to figure out why.

no need for a seperate partition, just unpack a stage* file in a directory and chroot right into it.

zxy · Post by **zxy** » Wed Dec 27, 2006 8:06 pm

First you'll have to get gcc 4.3 compiled at all. I had no luck with any of the alpha releases.
(amd64: glibc 2.5-r3, binutils 2.17.50.0.8, gcc 4.1.1-r3)

baigsabeeh · Post by **baigsabeeh** » Wed Dec 27, 2006 9:36 pm

Can't you keep multiple compilers. If so, then I'd build both GCC 4.2.0 and 4.3.0 in an instant and test both of them while keeping GCC 4.1.1-r3. It would be the best solution, so if anything goes kaput then I can rebuild on the fly.

drwook · Post by **drwook** » Wed Dec 27, 2006 9:43 pm

baigsabeeh wrote:Can't you keep multiple compilers. If so, then I'd build both GCC 4.2.0 and 4.3.0 in an instant and test both of them while keeping GCC 4.1.1-r3. It would be the best solution, so if anything goes kaput then I can rebuild on the fly.

Sure you can. I'd always advise a backup though...

baigsabeeh · Post by **baigsabeeh** » Wed Dec 27, 2006 10:01 pm

drwook wrote:
baigsabeeh wrote:Can't you keep multiple compilers. If so, then I'd build both GCC 4.2.0 and 4.3.0 in an instant and test both of them while keeping GCC 4.1.1-r3. It would be the best solution, so if anything goes kaput then I can rebuild on the fly.
Sure you can. I'd always advise a backup though...

Of course, I have a Stage4 of my system ready from Nov. 20, but I'll make a new one just for this.

baigsabeeh · Post by **baigsabeeh** » Wed Dec 27, 2006 10:14 pm

This is beautiful. I can't even get 1 minute into the GCC 4.2.0 install before it craps out when it says something about file ends being corrupted. anyway, I'm going ahead with GCC 4.3.0

baigsabeeh · Post by **baigsabeeh** » Wed Dec 27, 2006 11:37 pm

I can't even get GCC 4.3.0 to compile. I'll post the error when I try again, but right now GCC 4.2.0 is compiling and then I'm going to move up to Glibc 2.5.90, see how that works out.

zxy · Post by **zxy** » Wed Dec 27, 2006 11:47 pm

Maybe compiling gcc-4.3.0 with gcc-4.2.0 would work. I didn't try it yet, but I plan that for tomorrow.

baigsabeeh · Post by **baigsabeeh** » Wed Dec 27, 2006 11:53 pm

That could be the case. Here is my emerge --info. You guys think there is something that might be screwing with it? I did an emerge -eav system earlier today and it worked fine.

Code: Select all

sbmegsy@manchester ~ $ emerge --info
Portage 2.1.2_rc4-r1 (default-linux/amd64/2006.1, gcc-4.1.1, glibc-2.5-r0, 2.6.19-gentoo-r2 x86_64)
=================================================================
System uname: 2.6.19-gentoo-r2 x86_64 AMD Athlon(tm) 64 X2 Dual Core Processor 3800+
Gentoo Base System version 1.12.8
Last Sync: Wed, 27 Dec 2006 16:20:01 +0000
dev-java/java-config: 1.3.7, 2.0.31
dev-lang/python:     2.4.4
dev-python/pycrypto: 2.0.1-r5
sys-apps/sandbox:    1.2.18.1
sys-devel/autoconf:  2.13, 2.61
sys-devel/automake:  1.4_p6, 1.5, 1.6.3, 1.7.9-r1, 1.8.5-r3, 1.9.6-r2, 1.10
sys-devel/binutils:  2.17.50.0.8
sys-devel/gcc-config: 1.3.14
sys-devel/libtool:   1.5.22
virtual/os-headers:  2.6.17-r2
ACCEPT_KEYWORDS="amd64 ~amd64"
AUTOCLEAN="yes"
CBUILD="x86_64-pc-linux-gnu"
CFLAGS="-march=athlon64 -O3 -pipe -ftree-vectorize -fomit-frame-pointer -fprefetch-loop-arrays -frename-registers"
CHOST="x86_64-pc-linux-gnu"
CONFIG_PROTECT="/etc /usr/kde/3.5/env /usr/kde/3.5/share/config /usr/kde/3.5/shutdown /usr/lib64/mozilla/defaults/pref /usr/share/X11/xkb /usr/share/config"
CONFIG_PROTECT_MASK="/etc/env.d /etc/env.d/java/ /etc/gconf /etc/java-config/vms/ /etc/revdep-rebuild /etc/terminfo"
CXXFLAGS="-march=athlon64 -O3 -pipe -ftree-vectorize -fomit-frame-pointer -fprefetch-loop-arrays -frename-registers"
DISTDIR="/usr/portage/distfiles"
FEATURES="autoconfig distlocks metadata-transfer sandbox sfperms strict"
LDFLAGS="-Wl,-O1 -Wl,--sort-common -s -Wl,--hash-style=both"
MAKEOPTS="-j3"
PKGDIR="/usr/portage/packages"
PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress --force --whole-file --delete --delete-after --stats --timeout=180 --exclude=/distfiles --exclude=/local --exclude=/packages"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY="/usr/local/portage /usr/local/overlays/toolchain_overlay /usr/portage/local/layman/xeffects /usr/local/custom_kernels /usr/portage/local/layman/xeffects /usr/portage/local/layman/pok3d"
SYNC="rsync://rsync.namerica.gentoo.org/gentoo-portage"
Unset:  CTARGET, EMERGE_DEFAULT_OPTS, INSTALL_MASK, LANG, LC_ALL, LINGUAS, PORTAGE_RSYNC_EXTRA_OPTS

sbmegsy@manchester ~ $

DualCpuUser · Post by **DualCpuUser** » Wed Dec 27, 2006 11:55 pm

If you can't get one fo the alpha's of 4.3 to build, then something else it wrong. I do my own thing with an overlay ebuild and change my to 4.3.98 in BASE.Ver so the directories are unquie. I also make sure multislot is in my package.use file for gcc, so you can keep your gcc-4.1.1