View previous topic :: View next topic |
Author |
Message |
satanskin Guru
Joined: 25 Apr 2005 Posts: 353
|
Posted: Thu Nov 10, 2005 7:59 pm Post subject: |
|
|
So what is the difference between the 2.3.6 release and the HEAD branch? I don't know what the HEAD branch is, sorry. |
|
Back to top |
|
|
Redeeman l33t
Joined: 25 Sep 2003 Posts: 958 Location: Portugal
|
Posted: Thu Nov 10, 2005 8:43 pm Post subject: |
|
|
HEAD is the very latest development code |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Thu Nov 10, 2005 8:53 pm Post subject: |
|
|
Redeeman wrote: | satanskin wrote: | nxsty wrote: | Btw, I'm woking on a updated overlay. It will include a new snapshot from HEAD from today (like the 2.3.5.2005XXXX snapshots in portage) with updated patches, updated fedora add ons and gcc4 ssp support. |
Why not just use the official 2.3.6 release? Just came out today! woot! |
his overlay is based on that, but 2.3.6 still needs the libm patch to be real good |
The libm patch is still included and used. Check the patching step when you emerge it, it should be applied.
Glibc 2.3.6 is in portage now. I'll sync my ebuild with it soon. Now I don't have to exclude a lot of upstream fixed patches anymore. |
|
Back to top |
|
|
mbar Veteran
Joined: 19 Jan 2005 Posts: 1990 Location: Poland
|
Posted: Thu Nov 10, 2005 9:09 pm Post subject: |
|
|
And what about this glibc-2.3.6 ebuild being broken? "Portage sucks." it says |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Thu Nov 10, 2005 9:52 pm Post subject: |
|
|
mbar wrote: | And what about this glibc-2.3.6 ebuild being broken? "Portage sucks." it says |
I just noticed that when syncing my ebuild. I don't know, I never had any problems updating to 2.3.6 and I've had no problem reports in this thread about it. Im merging it now to see what happens. Perhaps it's only the devel versions of portage that's affected? |
|
Back to top |
|
|
Tiger683 Veteran
Joined: 08 Jan 2005 Posts: 1347 Location: Heffner's House
|
Posted: Thu Nov 10, 2005 9:57 pm Post subject: |
|
|
it emerges perfectly....with gcc-4.1 too _________________ Retired gentoo user |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Thu Nov 10, 2005 10:58 pm Post subject: |
|
|
Ok the new overlay is online, compile and run tested on amd64 and x86. Ping me if you have any problems emerging it! |
|
Back to top |
|
|
Redeeman l33t
Joined: 25 Sep 2003 Posts: 958 Location: Portugal
|
Posted: Thu Nov 10, 2005 11:51 pm Post subject: |
|
|
nxsty wrote: | Redeeman wrote: | satanskin wrote: | nxsty wrote: | Btw, I'm woking on a updated overlay. It will include a new snapshot from HEAD from today (like the 2.3.5.2005XXXX snapshots in portage) with updated patches, updated fedora add ons and gcc4 ssp support. |
Why not just use the official 2.3.6 release? Just came out today! woot! |
his overlay is based on that, but 2.3.6 still needs the libm patch to be real good |
The libm patch is still included and used. Check the patching step when you emerge it, it should be applied.
Glibc 2.3.6 is in portage now. I'll sync my ebuild with it soon. Now I don't have to exclude a lot of upstream fixed patches anymore. |
i did not mean to say that you dont apply it, the intention of that post was to tell satanskin that 2.3.6 still neded libm |
|
Back to top |
|
|
Kuroki Kaze n00b
Joined: 16 Aug 2005 Posts: 19 Location: Net-Sphere
|
Posted: Sun Nov 13, 2005 6:23 pm Post subject: |
|
|
Warning! your patchset added to crontab!
=)
simple update script:
Code: |
#!/bin/bash
W='/usr/bin/wget'
T='/bin/tar'
CWD=$PWD
GLIBCOVERDIR="/home/ftp/Linux/Gentoo/glibc-amd64"
source /etc/make.conf
cd $GLIBCOVERDIR
$W -N http://snigel.no-ip.com/~nxsty/linux/glibc-overlay.tar.bz2
$T -xvjf glibc-overlay.tar.bz2 -C $PORTDIR_OVERLAY/sys-libs
find $PORTDIR_OVERLAY/sys-libs/glibc/ -iname \*ebuild -exec ebuild '{}' digest \;
cd $CWD
|
_________________ Blame! |
|
Back to top |
|
|
Kuroki Kaze n00b
Joined: 16 Aug 2005 Posts: 19 Location: Net-Sphere
|
Posted: Sun Nov 13, 2005 7:20 pm Post subject: |
|
|
hmmm... can you add this in ebuild file:
Code: |
IUSE="nls pic build nptl nptlonly erandom hardened userlocales multilib selinux
glibc-compat20 glibc-omitfp linuxthreads-tls profile nomalloccheck custom_cflags"
setup_flags() {
....
if ! use custom_cflags ; then
strip-flags
strip-unsupported-flags
filter-flags -m32 -m64 -mabi=*
fi
....
if ! use custom_cflags ; then
filter-flags -O?
append-flags -O2
fi
}
|
_________________ Blame! |
|
Back to top |
|
|
enderandrew l33t
Joined: 25 Oct 2005 Posts: 731
|
Posted: Sun Nov 13, 2005 7:28 pm Post subject: |
|
|
I thought this patch was predominately for AMD64. Should I put this on non-AMD64 systems? _________________ Nihilism makes me smile. |
|
Back to top |
|
|
Gnufsh Guru
Joined: 28 Dec 2002 Posts: 400 Location: Portland, OR
|
Posted: Sun Nov 13, 2005 7:57 pm Post subject: |
|
|
THe libm and sting patches might give problems. I don't know if glibc as a fallback for that or not. I don't really think it's a good idea. |
|
Back to top |
|
|
Kuroki Kaze n00b
Joined: 16 Aug 2005 Posts: 19 Location: Net-Sphere
|
Posted: Sun Nov 13, 2005 8:07 pm Post subject: |
|
|
Gnufsh wrote: | THe libm and sting patches might give problems. I don't know if glibc as a fallback for that or not. I don't really think it's a good idea. |
thats only for crazy geeks =) _________________ Blame! |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Sun Nov 13, 2005 10:05 pm Post subject: |
|
|
Kuroki Kaze wrote: | hmmm... can you add this in ebuild file: |
I don't think that's such a good idea. Using nonstandard optimizations for glibc will proably just cause problems. Use the glibc-omitfp USE-flag if you want higher compiler optimizations. |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Sun Nov 13, 2005 10:08 pm Post subject: |
|
|
enderandrew wrote: | I thought this patch was predominately for AMD64. Should I put this on non-AMD64 systems? |
You can, the x86_64 parts isn't built on other arches. I use it on other x86 systems too. Allthough most fixes I had besides the x86_64 patches is included in glibc 2.3.6 so there isn't much different from the 2.3.6 ebuild in portage on x86. |
|
Back to top |
|
|
Kuroki Kaze n00b
Joined: 16 Aug 2005 Posts: 19 Location: Net-Sphere
|
Posted: Sun Nov 13, 2005 10:37 pm Post subject: |
|
|
CFLAGS="-O3 -fforce-addr -fforce-mem -march=k8 -pipe -fomit-frame-pointer" - compiled!
so it's my choise... _________________ Blame! |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Mon Nov 14, 2005 9:10 am Post subject: |
|
|
Kuroki Kaze wrote: | CFLAGS="-O3 -fforce-addr -fforce-mem -march=k8 -pipe -fomit-frame-pointer" - compiled!
so it's my choise... |
Yea, but people have a tendency to try to push it as far as it goes. If I added a custom-cflags USE a lot of people would try to compile it with -ffast-math, -funroll-loops etc. which would be pointless and only break their systems. Your optimizations are actually the same as turned on by glibc-omitfp except the -fforce-addr btw. |
|
Back to top |
|
|
enderandrew l33t
Joined: 25 Oct 2005 Posts: 731
|
Posted: Mon Nov 14, 2005 9:28 am Post subject: |
|
|
Isn't -fforce-addr somewhat redundant?
It seems to be the same thing as --fforce-mem, which is already enabled at -O2. So he is repeating it twice basically, right? _________________ Nihilism makes me smile. |
|
Back to top |
|
|
Kuroki Kaze n00b
Joined: 16 Aug 2005 Posts: 19 Location: Net-Sphere
|
Posted: Mon Nov 14, 2005 12:46 pm Post subject: |
|
|
enderandrew wrote: | Isn't -fforce-addr somewhat redundant?
It seems to be the same thing as --fforce-mem, which is already enabled at -O2. So he is repeating it twice basically, right? |
yeah.... how can i miss it?...
hmmm... lets try somthing else =) _________________ Blame! |
|
Back to top |
|
|
energyman76b Advocate
Joined: 26 Mar 2003 Posts: 2048 Location: Germany
|
Posted: Mon Nov 14, 2005 2:57 pm Post subject: |
|
|
Hi,
and fomit-frame-pointer is pointless on amd64. _________________ Study finds stunning lack of racial, gender, and economic diversity among middle-class white males
I identify as a dirty penismensch. |
|
Back to top |
|
|
Redeeman l33t
Joined: 25 Sep 2003 Posts: 958 Location: Portugal
|
Posted: Mon Nov 14, 2005 3:04 pm Post subject: |
|
|
energyman76b wrote: | Hi,
and fomit-frame-pointer is pointless on amd64. |
why |
|
Back to top |
|
|
loftwyr l33t
Joined: 29 Dec 2004 Posts: 970 Location: 43°38'23.62"N 79°27'8.60"W
|
Posted: Mon Nov 14, 2005 4:06 pm Post subject: |
|
|
You should probably update the intro to this thread to say 2.3.6 and show the benchmark comparison to 2.3.6 just so people don't get confused when they install it. _________________ My emerge --info
Have you run revdep-rebuild lately? It's in gentoolkit and it's worth a shot if things don't work well.
Celebrating 5 years of Gentoo-ing. |
|
Back to top |
|
|
loftwyr l33t
Joined: 29 Dec 2004 Posts: 970 Location: 43°38'23.62"N 79°27'8.60"W
|
Posted: Mon Nov 14, 2005 4:07 pm Post subject: |
|
|
Redeeman wrote: | energyman76b wrote: | Hi,
and fomit-frame-pointer is pointless on amd64. |
why |
gcc on x86-64 removes the frame pointer by default. _________________ My emerge --info
Have you run revdep-rebuild lately? It's in gentoolkit and it's worth a shot if things don't work well.
Celebrating 5 years of Gentoo-ing. |
|
Back to top |
|
|
Kuroki Kaze n00b
Joined: 16 Aug 2005 Posts: 19 Location: Net-Sphere
|
Posted: Mon Nov 14, 2005 4:34 pm Post subject: |
|
|
loftwyr wrote: | Redeeman wrote: | energyman76b wrote: | Hi,
and fomit-frame-pointer is pointless on amd64. |
why |
gcc on x86-64 removes the frame pointer by default. |
#man gcc
....
-O also turns on -fomit-frame-pointer on machines where doing so does not interfere with debugging.
.... _________________ Blame! |
|
Back to top |
|
|
nxsty Veteran
Joined: 23 Jun 2004 Posts: 1556 Location: .se
|
Posted: Tue Nov 15, 2005 12:07 pm Post subject: |
|
|
Glibc 2.3.6-r1 was added today so I've updated my overlay to use it. The changes are only interesting if you're using binutils 2.16.9* and/or gcc 4.1. |
|
Back to top |
|
|
|