GCC 3.4

Message

nepenthe · Post by **nepenthe** » Tue Apr 13, 2004 4:50 pm

Just change the line in the GCC 3.4 ebuild you already have from...

SNAPSHOT="2004-03-27"

to

Code: Select all

SNAPSHOT="2004-04-07

then redigest the ebuild

Lv · Post by Lv » Thu Apr 15, 2004 10:22 am

the pre-release is in a new directory, so that wont work.
EDIT: if you want the pre-release that is. the snapshot is still in the snapshot directory, but the pre-release is in the prerelease-3.4.0-20040406 directory...

I have a slightly amd64-specific ebuild for testing purposes in my devspace. This is where I insert the disclaimer that it is completely and totally unsupported and that bugs shouldn't be filed in bugzilla, even if the bug reporting URL has been updated to say so....

http://dev.gentoo.org/~lv/gcc-3.4.0-200 ... erlay.tbz2

I'd be interested in knowing the rate of success on other archs... especially x86 and ppc. if you have a portage overlay, just untar this inside that directory. if you dont feel like setting up a portage overlay for some reason, you can untar it inside /usr/portage, but it'll be erased next time you emerge sync.

good luck and enjoy, I guess.

Spawn of Lovechild · Post by **Spawn of Lovechild** » Thu Apr 15, 2004 10:32 am

With a littlte bit of ln -s charm I made the ebuild from this thread work on the prerelease and I had good results for x86 - everything I tried compiled, however I didn't try big things like glibc.

ett_gramse_nap · Post by **ett_gramse_nap** » Thu Apr 15, 2004 11:22 am

When would the pre-release show up in portage?

() · Post by () » Thu Apr 15, 2004 12:01 pm

Does GCC 3.4 include the new parser structure, rewritten from Bison to recursive descent, and precompiled headers? I'm hoping this will speed up C++ compilation

Markus_T · Post by **Markus_T** » Thu Apr 15, 2004 2:03 pm

() wrote:Does GCC 3.4 include the new parser structure, rewritten from Bison to recursive descent, and precompiled headers? I'm hoping this will speed up C++ compilation

Yes, look here for further information:
http://gcc.gnu.org/gcc-3.4/changes.html

() · Post by () » Thu Apr 15, 2004 2:05 pm

Ah, very nice. I remember using #pragma once from my Visual Studio days, beats the traditional preprocessor guard.

Lv · Post by Lv » Thu Apr 15, 2004 2:30 pm

ett_gramse_nap wrote:When would the pre-release show up in portage?

no idea. if it were entirely up to me, it would only be installable on amd64, as we have a definate need for it. i mean... gcc 3.4 JUST added an -march option for our arch. a lot of amd64-specific improvements and fixes are included in the new release... enough that re-compiling with gcc 3.4 makes it feel like you just bought a new computer.

EDIT - "only installable on amd64", because a number of things still break and i'm not fond of the idea that i would be breaking things

oh yeah, if you want a quick and easy way to recompile as much of your system as possible with gcc 3.4 without having the whole process stop as soon as a package fails to compile (there /will/ be a nice number of packages that fail to compile), run the following:

emerge world -ep | grep ^.ebuild | awk '{ print $4 }' | sed 's/^/\=/' | xargs -n1 emerge

Master_Of_Disaster · Post by **Master_Of_Disaster** » Fri Apr 16, 2004 4:34 am

Code: Select all

gcc abi-note.S -c  -I../include -I. -I/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu -I.. -I../libio -I../nptl -I/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere -I../sysdeps/x86_64/elf -I../nptl/sysdeps/unix/sysv/linux/x86_64 -I../nptl/sysdeps/unix/sysv/linux -I../nptl/sysdeps/pthread -I../sysdeps/pthread -I../nptl/sysdeps/unix/sysv -I../nptl/sysdeps/unix -I../nptl/sysdeps/x86_64 -I../sysdeps/unix/sysv/linux/x86_64 -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps/unix/sysv -I../sysdeps/unix/x86_64 -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/x86_64/fpu -I../sysdeps/x86_64 -I../sysdeps/wordsize-64 -I../sysdeps/ieee754/ldbl-96 -I../sysdeps/ieee754/dbl-64 -I../sysdeps/ieee754/flt-32 -I../sysdeps/ieee754 -I../sysdeps/generic/elf -I../sysdeps/generic -nostdinc -isystem /usr/lib/gcc/x86_64-pc-linux-gnu/3.4.0/include -isystem //usr/include -D_LIBC_REENTRANT -D_LIBC_REENTRANT -include ../include/libc-symbols.h       -DHAVE_INITFINI -DASSEMBLER  -I/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu/.  -Wa,--noexecstack  -o /var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu/abi-note.o -MD -MP -MF /var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu/abi-note.o.dt
gcc init.c -c -std=gnu99 -O2 -Wall -Winline -Wstrict-prototypes -Wwrite-strings -freorder-blocks -march=k8 -pipe      -I../include -I. -I/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu -I.. -I../libio -I../nptl -I/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere -I../sysdeps/x86_64/elf -I../nptl/sysdeps/unix/sysv/linux/x86_64 -I../nptl/sysdeps/unix/sysv/linux -I../nptl/sysdeps/pthread -I../sysdeps/pthread -I../nptl/sysdeps/unix/sysv -I../nptl/sysdeps/unix -I../nptl/sysdeps/x86_64 -I../sysdeps/unix/sysv/linux/x86_64 -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu -I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet -I../sysdeps/unix/sysv -I../sysdeps/unix/x86_64 -I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/x86_64/fpu -I../sysdeps/x86_64 -I../sysdeps/wordsize-64 -I../sysdeps/ieee754/ldbl-96 -I../sysdeps/ieee754/dbl-64 -I../sysdeps/ieee754/flt-32 -I../sysdeps/ieee754 -I../sysdeps/generic/elf -I../sysdeps/generic -nostdinc -isystem /usr/lib/gcc/x86_64-pc-linux-gnu/3.4.0/include -isystem //usr/include -D_LIBC_REENTRANT -D_LIBC_REENTRANT -include ../include/libc-symbols.h       -DHAVE_INITFINI -o /var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu/init.o -MD -MP -MF /var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu/init.o.dt
abi-note.S: Assembler messages:
abi-note.S:66: Warning: rest of line ignored; first ignored character is `0'
abi-note.S:67: Warning: rest of line ignored; first ignored character is `2'
abi-note.S:66: Error: can't resolve `.note.ABI-tag' {.note.ABI-tag section} - `' {*UND* section}
abi-note.S:67: Error: can't resolve `.note.ABI-tag' {.note.ABI-tag section} - `' {*UND* section}
make[2]: *** [/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/buildhere/csu/abi-note.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory `/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2/csu'
make[1]: *** [csu/subdir_lib] Error 2
make[1]: Leaving directory `/var/tmp/portage/glibc-2.3.3_pre20040207/work/glibc-2.3.2'
make: *** [all] Error 2

!!! ERROR: sys-libs/glibc-2.3.3_pre20040207 failed.
!!! Function src_compile, Line 531, Exitcode 2
!!! (no error message)

This occurs with all (!) snapshots >= 2004-04-07 (USE="-multilib -java" on amd64). Has somebody got an idea how to get over this?
Furthermore, gcc refuses to find -lgcc_s though it's directory /usr/lib/gcc/x86_64-pc-linux-gnu/3.4.0/../lib64 is in /etc/ld.so.conf. If I specify it via -L, it works, but otherwise not...

deepspace9 · Post by **deepspace9** » Fri Apr 16, 2004 6:39 am

This occurs with all (!) snapshots >= 2004-04-07 (USE="-multilib -java" on amd64). Has somebody got an idea how to get over this?

I saw this with the prerelease too. Used an earlyer version, and that worked fine

Furthermore, gcc refuses to find -lgcc_s though it's directory /usr/lib/gcc/x86_64-pc-linux-gnu/3.4.0/../lib64 is in /etc/ld.so.conf. If I specify it via -L, it works, but otherwise not...

This is a symbolic link problem. search the ebuild for "-lgcc_s" and you will find the needed links. Somehow these links are not made...

Spawn of Lovechild · Post by **Spawn of Lovechild** » Fri Apr 16, 2004 6:56 am

2004-04-14 is out btw.

ett_gramse_nap · Post by **ett_gramse_nap** » Fri Apr 16, 2004 9:18 am

Spawn of Lovechild wrote:2004-04-14 is out btw.

Then, maybe it's time for me stop lurking around in this thread and get into some action...

Are there any common pitfalls one should be aware of when starting to use 3.4? Or is it just to emerge it and then start emerging anything else?

Master_Of_Disaster · Post by **Master_Of_Disaster** » Fri Apr 16, 2004 9:29 am

Is this problem with glibc I mentioned above Gentoo specific? Or should this be reported as a bug in gcc? Perhaps building a newer glib would help?

NeighborhoodGullwings · Fri Apr 16, 2004 9:43 pm

ett_gramse_nap wrote:
Spawn of Lovechild wrote:2004-04-14 is out btw.
Then, maybe it's time for me stop lurking around in this thread and get into some action...

Are there any common pitfalls one should be aware of when starting to use 3.4? Or is it just to emerge it and then start emerging anything else?

No you simply emerge it, and some things will compile with it and others won't. Fortunately, there's gcc-config to switch between different versions of gcc.

ett_gramse_nap · Post by **ett_gramse_nap** » Fri Apr 16, 2004 9:48 pm

NeighborhoodGullwings wrote:
ett_gramse_nap wrote:
Spawn of Lovechild wrote:2004-04-14 is out btw.
Then, maybe it's time for me stop lurking around in this thread and get into some action...

Are there any common pitfalls one should be aware of when starting to use 3.4? Or is it just to emerge it and then start emerging anything else?
No you simply emerge it, and some things will compile with it and others won't. Fortunately, there's gcc-config to switch between different versions of gcc.

Aha, didn't know that! Thanks!

Master_Of_Disaster · Post by **Master_Of_Disaster** » Sat Apr 17, 2004 1:28 pm

Ignore my whining about compile errors, it turns out that compiling binutils with too agressive CFLAGS caused it. After compiling binutils with -O2 everything worked fine.

Spawn of Lovechild · Post by **Spawn of Lovechild** » Sat Apr 17, 2004 3:04 pm

Master_Of_Disaster wrote:Ignore my whining about compile errors, it turns out that compiling binutils with too agressive CFLAGS caused it. After compiling binutils with -O2 everything worked fine.

maybe that will teach ya.

Lv · Post by Lv » Sat Apr 17, 2004 3:16 pm

new updated overlay at http://dev.gentoo.org/~lv/gcc-3.4.0-200 ... erlay.tbz2

besides being a new pre-release, it should fix the libgcc_s.so problem people are seeing on amd64

can someone on x86 let me know where this ebuild installs your libgcc_s.so? on amd64 it gets installed in /usr/lib/gcc-lib/x86_64-pc-linux-gnu/lib64/, however i doubt x86 would have a lib64. perhaps it's installed into lib? or the versioned 3.4.0 directory where it should be?

Lv · Post by Lv » Sat Apr 17, 2004 3:21 pm

ett_gramse_nap wrote:
Spawn of Lovechild wrote:2004-04-14 is out btw.
Then, maybe it's time for me stop lurking around in this thread and get into some action...

Are there any common pitfalls one should be aware of when starting to use 3.4? Or is it just to emerge it and then start emerging anything else?

dont install binutils with anything in CFLAGS other than -O2 or it will break. once gcc 3.4 is in the tree, i'll see about adding a pkg_setup() to the binutils ebuild to change CFLAGS to -O2 for you if you're using gcc 3.4.

anything else that has a problem just simply refuses to compile. this includes gcj, so add nogcj to your USE.

darkless · Post by **darkless** » Sat Apr 17, 2004 3:45 pm

I found and applied a patch that fixes the glibc compilation error regarding "fixup" that shows up when using a recent GCC-3.4 snapshot.

Grab the ebuild here:
http://www.sundebo.dk/dark-sources/glib ... 207.ebuild

And the patch here:
http://www.sundebo.dk/dark-sources/glib ... ixup.patch

Put the patch in the sys-libs/glibc/files/2.3.3/ folder and digest the ebuild.

MikeP · Post by **MikeP** » Sat Apr 17, 2004 5:49 pm

well i have a question now, precompiled headers etc improve speed, and quite some people use it already, has anyone done some benchmarks then to compare the speed of 3.3 and 3.4 in compiling different things?

NeighborhoodGullwings · Sat Apr 17, 2004 6:16 pm

Thanks darkless for the fixup patch, I'm trying it out now.

Is anyone else still having problems compiling a 2.4 kernel (vanilla 2.4.26 at the moment, though I've also tried 2.4.25 and 2.4.24). Recent gcc 3.4 snapshots won't work for me with the kernel. It doesn't matter if I use -fno-unit-at-a-time or not either.

Here's the errors I get:

Code: Select all

gcc -D__KERNEL__ -I/usr/src/linux-2.4.26/include -Wall -Wstrict-prototypes -Wno-trigraphs -O2 -fno-strict-aliasing -fno-common -fomit-frame-pointer -pipe -mpreferred-stack-boundary=2 -march=athlon   -nostdinc -iwithprefix include -DKBUILD_BASENAME=sched  -fno-omit-frame-pointer -c -o sched.o sched.c
sched.c:213: error: conflicting types for 'reschedule_idle'
sched.c:210: error: previous declaration of 'reschedule_idle' was here
sched.c:213: error: conflicting types for 'reschedule_idle'
sched.c:210: error: previous declaration of 'reschedule_idle' was here
sched.c:371: error: conflicting types for 'wake_up_process'
/usr/src/linux-2.4.26/include/linux/sched.h:603: error: previous declaration of 'wake_up_process' was here
sched.c:371: error: conflicting types for 'wake_up_process'
/usr/src/linux-2.4.26/include/linux/sched.h:603: error: previous declaration of 'wake_up_process' was here
sched.c:409: error: conflicting types for 'schedule_timeout'
/usr/src/linux-2.4.26/include/linux/sched.h:148: error: previous declaration of 'schedule_timeout' was here
sched.c:409: error: conflicting types for 'schedule_timeout'
/usr/src/linux-2.4.26/include/linux/sched.h:148: error: previous declaration of 'schedule_timeout' was here
sched.c:739: error: conflicting types for '__wake_up'
/usr/src/linux-2.4.26/include/linux/sched.h:595: error: previous declaration of '__wake_up' was here
sched.c:739: error: conflicting types for '__wake_up'
/usr/src/linux-2.4.26/include/linux/sched.h:595: error: previous declaration of '__wake_up' was here
sched.c:749: error: conflicting types for '__wake_up_sync'
/usr/src/linux-2.4.26/include/linux/sched.h:596: error: previous declaration of '__wake_up_sync' was here
sched.c:749: error: conflicting types for '__wake_up_sync'
/usr/src/linux-2.4.26/include/linux/sched.h:596: error: previous declaration of '__wake_up_sync' was here
sched.c:759: error: conflicting types for 'complete'
/usr/src/linux-2.4.26/include/linux/completion.h:31: error: previous declaration of 'complete' was here
sched.c:759: error: conflicting types for 'complete'
/usr/src/linux-2.4.26/include/linux/completion.h:31: error: previous declaration of 'complete' was here
sched.c:769: error: conflicting types for 'wait_for_completion'
/usr/src/linux-2.4.26/include/linux/completion.h:30: error: previous declaration of 'wait_for_completion' was here
sched.c:769: error: conflicting types for 'wait_for_completion'
/usr/src/linux-2.4.26/include/linux/completion.h:30: error: previous declaration of 'wait_for_completion' was here
sched.c:804: error: conflicting types for 'interruptible_sleep_on'
/usr/src/linux-2.4.26/include/linux/sched.h:600: error: previous declaration of 'interruptible_sleep_on' was here
sched.c:804: error: conflicting types for 'interruptible_sleep_on'
/usr/src/linux-2.4.26/include/linux/sched.h:600: error: previous declaration of 'interruptible_sleep_on' was here
sched.c:815: error: conflicting types for 'interruptible_sleep_on_timeout'
/usr/src/linux-2.4.26/include/linux/sched.h:601: error: previous declaration of 'interruptible_sleep_on_timeout' was here
sched.c:815: error: conflicting types for 'interruptible_sleep_on_timeout'
/usr/src/linux-2.4.26/include/linux/sched.h:601: error: previous declaration of 'interruptible_sleep_on_timeout' was here
sched.c:828: error: conflicting types for 'sleep_on'
/usr/src/linux-2.4.26/include/linux/sched.h:597: error: previous declaration of 'sleep_on' was here
sched.c:828: error: conflicting types for 'sleep_on'
/usr/src/linux-2.4.26/include/linux/sched.h:597: error: previous declaration of 'sleep_on' was here
sched.c:839: error: conflicting types for 'sleep_on_timeout'
/usr/src/linux-2.4.26/include/linux/sched.h:598: error: previous declaration of 'sleep_on_timeout' was here
sched.c:839: error: conflicting types for 'sleep_on_timeout'
/usr/src/linux-2.4.26/include/linux/sched.h:598: error: previous declaration of 'sleep_on_timeout' was here
sched.c:210: warning: 'reschedule_idle' declared `static' but never defined
make[2]: *** [sched.o] Error 1
make[2]: Leaving directory `/usr/src/linux-2.4.26/kernel'
make[1]: *** [first_rule] Error 2
make[1]: Leaving directory `/usr/src/linux-2.4.26/kernel'
make: *** [_dir_kernel] Error 2

I have tried fixing this myself but it is beyond me, the errors just keep cropping up.

NeighborhoodGullwings · Sat Apr 17, 2004 7:54 pm

The fixup patch worked perfectly darkless. Now if I could get the kernel to work again....

MikeP · Post by **MikeP** » Sat Apr 17, 2004 8:52 pm

switch to 2.6

NeighborhoodGullwings · Sat Apr 17, 2004 10:38 pm

MikeP wrote:switch to 2.6

That isn't a very helpful answer, although that's what I was expecting anyway. I don't want to switch to 2.6 for other various reasons beyond the scope of this thread.