View previous topic :: View next topic |
Author |
Message |
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Thu Oct 25, 2007 5:10 pm Post subject: |
|
|
Oh no, I have tried quiete a few and there are lots of problems, for example ffmpeg gives an ICE and mplayer simply don't compil...
It's still really experimental |
|
Back to top |
|
|
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Sat Oct 27, 2007 12:59 pm Post subject: |
|
|
Just emerged gcc-4.3.0_alpha20071026 and emerge -va1 ffmpeg mplayer still produce ICE and don't compil... |
|
Back to top |
|
|
Dottout l33t
Joined: 07 Mar 2006 Posts: 882
|
Posted: Sun Oct 28, 2007 6:52 pm Post subject: |
|
|
with gcc-4.3.0-alpha20071026 glibmm now compiles fine |
|
Back to top |
|
|
unK l33t
Joined: 06 Feb 2007 Posts: 769
|
Posted: Mon Oct 29, 2007 8:36 pm Post subject: |
|
|
Gregoire wrote: | Just emerged gcc-4.3.0_alpha20071026 and emerge -va1 ffmpeg mplayer still produce ICE and don't compil... |
With gcc-4.3.0_alpha20071005 & 1012 I compiled my whole world, no one package failed. ffmpeg and mplayer both compiled succesfully also.
It seems gcc devs finally fixed -ftree-loop-linear flag (see this bug)! It doesn't cause ICE with xorg-server and ghostscript-gpl compilation anymore (I tried this package 'cause eremini was complaining about this a few pages before). I'm going to check this flag out. :)
PS. It's nothing superb about this flag, I just like its name ;p |
|
Back to top |
|
|
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Mon Oct 29, 2007 9:57 pm Post subject: |
|
|
unK wrote: | Gregoire wrote: | Just emerged gcc-4.3.0_alpha20071026 and emerge -va1 ffmpeg mplayer still produce ICE and don't compil... |
With gcc-4.3.0_alpha20071005 & 1012 I compiled my whole world, no one package failed. ffmpeg and mplayer both compiled succesfully also.
It seems gcc devs finally fixed -ftree-loop-linear flag (see this bug)! It doesn't cause ICE with xorg-server and ghostscript-gpl compilation anymore (I tried this package 'cause eremini was complaining about this a few pages before). I'm going to check this flag out.
PS. It's nothing superb about this flag, I just like its name ;p |
Is that under x86_64 ? Which flags ? |
|
Back to top |
|
|
zxy Veteran
Joined: 06 Jan 2006 Posts: 1160 Location: in bed in front of the computer
|
|
Back to top |
|
|
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Tue Oct 30, 2007 8:25 am Post subject: |
|
|
There still some things missing in last ebuild to compil under x86_64 :
Code: |
--- gcc-4.3.0_alpha20071026.ebuild 2007-10-30 08:00:56.000000000 +0100
+++ gcc-4.3.0_alpha20071026.ebuild-greg 2007-10-27 14:26:00.000000000 +0200
@@ -4,6 +4,7 @@
ETYPE="gcc-compiler"
GCC_FILESDIR=${PORTDIR}/sys-devel/gcc/files
+GCC_MAKE_TARGET="bootstrap"
inherit toolchain
@@ -51,6 +52,7 @@
# Fix cross-compiling
epatch "${GCC_FILESDIR}"/4.1.0/gcc-4.1.0-cross-compile.patch
+ epatch /usr/portage/local/layman/toolchain/sys-devel/gcc/files/multilib
}
pkg_postinst() {
|
With multilib posted a few posts ago :
Code: |
diff -urN fixincludes.orig/fixincl.x fixincludes/fixincl.x
--- fixincludes.orig/fixincl.x 2007-07-26 00:51:47.000000000 +0100
+++ fixincludes/fixincl.x 2007-08-08 00:39:26.000000000 +0100
@@ -6349,7 +6349,7 @@
* content bypass pattern - skip fix if pattern found
*/
tSCC zStdio_Stdarg_HBypass0[] =
- "include.*(stdarg.h|machine/ansi.h)";
+ "include.*(stdarg.h|machine/ansi.h)|gentoo-multilib";
#define STDIO_STDARG_H_TEST_CT 1
static tTestDesc aStdio_Stdarg_HTests[] = {
diff -urN fixincludes.orig/inclhack.def fixincludes/inclhack.def
--- fixincludes.orig/inclhack.def 2007-07-26 00:51:47.000000000 +0100
+++ fixincludes/inclhack.def 2007-08-08 00:18:06.129542000 +0100
@@ -3457,12 +3457,12 @@
/*
* Arrange for stdio.h to use stdarg.h to define __gnuc_va_list.
* On 4BSD-derived systems, stdio.h defers to machine/ansi.h; that's
- * OK too.
+ * OK too. Gentoo multilib uses a wrapper for stdio.h; skip it.
*/
fix = {
hackname = stdio_stdarg_h;
files = stdio.h;
- bypass = "include.*(stdarg\.h|machine/ansi\.h)";
+ bypass = "include.*(stdarg\.h|machine/ansi\.h)|gentoo-multilib";
/*
* On Solaris 10, this fix is unncessary; <stdio.h> includes
* <iso/stdio_iso.h>, which includes <sys/va_list.h>.
|
Would it be possible to include those in toolchain ? |
|
Back to top |
|
|
Thargor Apprentice
Joined: 23 Dec 2005 Posts: 239 Location: Bamberg/Germany
|
Posted: Tue Oct 30, 2007 10:44 am Post subject: |
|
|
zxy wrote: | http://gcc.gnu.org/ml/gcc/2007-10/msg00441.html
So 4.3 is slowly finishing stage3. |
For people like me which do not know that much about gcc development:
What follows after stage 3? RC-versions?
And what does slushy mean? If I remember correctly, gcc-4.2.1 was marked slushy. I know the normal meaning of the word, but it doesn't fit that good in this place...
Thanks in advantage _________________ Ideas are bulletproof |
|
Back to top |
|
|
zxy Veteran
Joined: 06 Jan 2006 Posts: 1160 Location: in bed in front of the computer
|
Posted: Tue Oct 30, 2007 6:54 pm Post subject: |
|
|
Thargor wrote: | zxy wrote: | http://gcc.gnu.org/ml/gcc/2007-10/msg00441.html
So 4.3 is slowly finishing stage3. |
For people like me which do not know that much about gcc development:
What follows after stage 3? RC-versions?
And what does slushy mean? If I remember correctly, gcc-4.2.1 was marked slushy. I know the normal meaning of the word, but it doesn't fit that good in this place...
Thanks in advantage |
But after reading this http://gcc.gnu.org/ml/gcc/2007-10/msg00462.html htread i'm not so sure it's comming quickly. _________________ Nature does not hurry, yet everything is accomplished.
Lao Tzu |
|
Back to top |
|
|
yamakawa Guru
Joined: 28 Jul 2003 Posts: 340
|
Posted: Tue Oct 30, 2007 7:23 pm Post subject: |
|
|
Genewb wrote: |
The multilib patch is still needed:
|
Thank you for reminding me of my old post...I have poor memory.
Thargor wrote: |
What follows after stage 3? RC-versions?
|
AFAIK here are GCC development plans
http://gcc.gnu.org/develop.html
http://gcc.gnu.org/wiki/GCC_4.3_Release_Planning
I made svn ebuild for 4.3.0. did not closely examine depend/rdepend etc, but so far it works fine for me.
As for using svn gcc, emerge -e system failed at cpio and tar, both of which can be fixed by adding -std=gnu89 CFLAG and gpm, which is discussed here and fixed by a patch.
Code: |
GCC_MAKE_TARGET="bootstrap"
ETYPE="gcc-compiler"
GCC_FILESDIR=${PORTDIR}/sys-devel/gcc/files
LOCAL_GCC_FILESDIR=/usr/portage/local/myown/sys-devel/gcc/files
ESVN_REPO_URI="svn://gcc.gnu.org/svn/gcc/trunk"
inherit toolchain subversion
DESCRIPTION="The GNU Compiler Collection. Includes C/C++, java compilers, pie+ssp extensions, Haj Ten Brugge runtime bounds checking"
SRC_URI=""
LICENSE="GPL-2 LGPL-2.1"
KEYWORDS="-*"
RDEPEND=">=sys-libs/zlib-1.1.4
|| ( >=sys-devel/gcc-config-1.3.12-r4 app-admin/eselect-compiler )
virtual/libiconv
>=dev-libs/gmp-4.2.1
>=dev-libs/mpfr-2.2.0_p10
!build? (
gcj? (
gtk? (
|| ( ( x11-libs/libXt x11-libs/libX11 x11-libs/libXtst x11-proto/xproto x11-proto/xextproto ) virtual/x11 )
>=x11-libs/gtk+-2.2
x11-libs/pango
)
>=media-libs/libart_lgpl-2.1
)
>=sys-libs/ncurses-5.2-r2
nls? ( sys-devel/gettext )
)"
DEPEND="${RDEPEND}
>=sys-apps/texinfo-4.2-r4
>=sys-devel/bison-1.875
>=${CATEGORY}/binutils-2.16.1"
PDEPEND="|| ( sys-devel/gcc-config app-admin/eselect-compiler )"
if [[ ${CATEGORY} != cross-* ]] ; then
PDEPEND="${PDEPEND} elibc_glibc? ( >=sys-libs/glibc-2.3.6 )"
fi
svn_pkg_setup() {
if [[ -z ${I_PROMISE_TO_SUPPLY_PATCHES_WITH_BUGS} ]] ; then
die "Please \`export I_PROMISE_TO_SUPPLY_PATCHES_WITH_BUGS=1\` or define it in your make.conf if you want to use this ebuild. This is to try and cut down on people filing bugs for a compiler we do not currently support."
fi
}
S=${WORKDIR}
src_unpack() {
subversion_src_unpack
use vanilla && return 0
cd "${S}"
# Fix cross-compiling
epatch "${GCC_FILESDIR}"/4.1.0/gcc-4.1.0-cross-compile.patch
epatch "${LOCAL_GCC_FILESDIR}"/gcc-4.3.0-multilib.patch
}
pkg_postinst() {
toolchain_pkg_postinst
einfo "This gcc-4 ebuild is provided for your convenience, and the use"
einfo "of this compiler is not supported by the Gentoo Developers."
einfo "Please file bugs related to gcc-4 with upstream developers."
einfo "Compiler bugs should be filed at http://gcc.gnu.org/bugzilla/"
}
|
|
|
Back to top |
|
|
mbar Veteran
Joined: 19 Jan 2005 Posts: 1990 Location: Poland
|
Posted: Tue Oct 30, 2007 7:32 pm Post subject: |
|
|
thanks for the ebuild |
|
Back to top |
|
|
unK l33t
Joined: 06 Feb 2007 Posts: 769
|
Posted: Tue Oct 30, 2007 7:58 pm Post subject: |
|
|
Gregoire wrote: | unK wrote: | Gregoire wrote: | Just emerged gcc-4.3.0_alpha20071026 and emerge -va1 ffmpeg mplayer still produce ICE and don't compil... |
With gcc-4.3.0_alpha20071005 & 1012 I compiled my whole world, no one package failed. ffmpeg and mplayer both compiled succesfully also.
It seems gcc devs finally fixed -ftree-loop-linear flag (see this bug)! It doesn't cause ICE with xorg-server and ghostscript-gpl compilation anymore (I tried this package 'cause eremini was complaining about this a few pages before). I'm going to check this flag out.
PS. It's nothing superb about this flag, I just like its name ;p |
Is that under x86_64 ? Which flags ? |
No, it's x86. Flags are simple "-O2 -march=native -pipe -fomit-frame-pointer". |
|
Back to top |
|
|
Gregoire Apprentice
Joined: 15 Apr 2006 Posts: 292
|
Posted: Tue Oct 30, 2007 8:15 pm Post subject: |
|
|
unK wrote: |
No, it's x86. Flags are simple "-O2 -march=native -pipe -fomit-frame-pointer". |
OK, then there no so many improvement in using the gcc in respect to code speed, as there is with new -march=core2.
But I am still interested if you are able to build a recent svn ffmpeg (like the one from overlay berkano) with your gcc-4.3 ?
Last released ffmpeg is simply much too incomplete to play h.264 stream more or less well |
|
Back to top |
|
|
unK l33t
Joined: 06 Feb 2007 Posts: 769
|
Posted: Tue Oct 30, 2007 9:00 pm Post subject: |
|
|
Gregoire wrote: | But I am still interested if you are able to build a recent svn ffmpeg (like the one from overlay berkano) with your gcc-4.3 ? |
Compiled successfully, even with "-O2 -march=native -pipe -fomit-frame-pointer -funroll-loops -ftree-loop-linear". But unfortunately gcc-4.3.0 is not able to compile kdelibs with these flags (well, even with sane set, it seems some dependencies got broken because of these insane cflags ) so I'm going to try this set with gcc-4.2.2, I'm very curious about "how -funroll-loops, -ftree-loop-linear and -ftracer may damage my system" ; ) |
|
Back to top |
|
|
treffer Apprentice
Joined: 14 Dec 2004 Posts: 150
|
Posted: Thu Nov 01, 2007 4:27 pm Post subject: |
|
|
gcc 4.3 with gcj useflag requires a zip/jar implementation, so you will need zip or any other package that provides zip/fastjar/.... (e.g. older gcc with gcj)
Compiling gcc 4.3, try 2... _________________ root@localhost# whois POEM-RIPE55-SONG
root@localhost# : ( ) { : | : & } ; : |
|
Back to top |
|
|
likewhoa l33t
Joined: 04 Oct 2006 Posts: 778 Location: Brooklyn, New York
|
Posted: Sun Nov 04, 2007 5:50 am Post subject: |
|
|
anyone running into this
Quote: |
ld -m elf_i386 -m elf_i386 -o .tmp_vmlinux1 -T arch/i386/kernel/vmlinux.lds
arch/i386/kernel/head.o arch/i386/kernel/init_task.o init/built-in.o
--start-group usr/built-in.o arch/i386/kernel/built-in.o
arch/i386/mm/built-in.o arch/i386/mach-default/built-in.o
arch/i386/crypto/built-in.o kernel/built-in.o mm/built-in.o fs/built-in.o
ipc/built-in.o security/built-in.o crypto/built-in.o block/built-in.o
lib/lib.a arch/i386/lib/lib.a lib/built-in.o arch/i386/lib/built-in.o
drivers/built-in.o sound/built-in.o arch/i386/pci/built-in.o
arch/i386/power/built-in.o net/built-in.o --end-group
kernel/built-in.o: In function `getnstimeofday':
(.text+0x1eba5): undefined reference to `__udivdi3'
kernel/built-in.o: In function `do_gettimeofday':
(.text+0x1ecca): undefined reference to `__udivdi3'
kernel/built-in.o: In function `update_wall_time':
(.text+0x1f0f4): undefined reference to `__udivdi3'
make: *** [.tmp_vmlinux1] Error 1
|
also see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32494
This on x86. |
|
Back to top |
|
|
nbvcxz Guru
Joined: 02 Sep 2005 Posts: 379 Location: Kraków / PL
|
Posted: Sun Nov 04, 2007 8:10 am Post subject: |
|
|
just look into https://forums.gentoo.org/viewtopic-t-608387.html there is gcc-4.3.0-compilation-fix.patch included _________________ nBVCXz
zen-kernel (bfq compcache) | /tmp -> tmpfs | ext4 | zsh | xfce | schedtool |
|
Back to top |
|
|
likewhoa l33t
Joined: 04 Oct 2006 Posts: 778 Location: Brooklyn, New York
|
Posted: Sun Nov 04, 2007 1:14 pm Post subject: |
|
|
thanks buddy i don't know how i forgot that setting. |
|
Back to top |
|
|
b33fc0d3 Retired Dev
Joined: 22 Jun 2006 Posts: 128
|
Posted: Wed Nov 07, 2007 3:22 am Post subject: Working CFLAGS |
|
|
On x86_64, i've been using these flags succesfully for weeks (if not months). A fully compiled system with glibc-2.7 compiled with them too.
Code: | CFLAGS="-march=athlon64 -O3"
CFLAGS="${CFLAGS} -ftree-vectorizer-verbose=3"
CFLAGS="${CFLAGS} -frecord-gcc-switches"
CXXFLAGS="-march=athlon64 -O3 -ftree-vectorizer-verbose=3 -fpermissive"
LDFLAGS="-Wl,-O1,--enable-new-dtags,--sort-common,--hash-style=gnu,--as-needed" |
|
|
Back to top |
|
|
baigsabeeh Guru
Joined: 28 Sep 2005 Posts: 520 Location: North Brunswick, NJ
|
Posted: Thu Nov 08, 2007 1:03 am Post subject: |
|
|
What's --enable-new-dtags do? _________________ BSD > SysV > Linux
BSD FTW! |
|
Back to top |
|
|
FireBurn Apprentice
Joined: 19 Sep 2004 Posts: 170 Location: Edinburgh, UK
|
Posted: Thu Nov 08, 2007 1:55 am Post subject: |
|
|
It's enabled by default anyway so it doesn't need to be specified in your LDFLAGS |
|
Back to top |
|
|
likewhoa l33t
Joined: 04 Oct 2006 Posts: 778 Location: Brooklyn, New York
|
Posted: Fri Nov 09, 2007 7:24 am Post subject: |
|
|
i ran into some app-text/recode-3.6-r2 problems and fixed them with this small patch.
Code: |
--- src/recodext.h 2001-01-04 09:36:54.000000000 -0500
+++ src/recodext.h 2007-11-08 21:11:24.000000000 -0500
@@ -218,7 +218,7 @@
enum recode_symbol_type type : 3;
/* Non zero if this one should be ignored. */
- bool ignore : 2;
+ bool ignore : 1;
};
struct recode_surface_list
|
source: http://lists.opensuse.org/opensuse-commit/2007-10/msg00854.html |
|
Back to top |
|
|
rhill Retired Dev
Joined: 22 Oct 2004 Posts: 1629 Location: sk.ca
|
Posted: Sun Nov 11, 2007 2:45 pm Post subject: Re: Working CFLAGS |
|
|
b33fc0d3 wrote: |
CFLAGS="${CFLAGS} -ftree-vectorizer-verbose=3"
|
this doesn't actually do anything unless -ftree-vectorize is enabled. now that http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32893 is fixed on the trunk, vectorization isn't as unstable as it has been in previous releases. that said, we still don't support it. _________________ by design, by neglect
for a fact or just for effect |
|
Back to top |
|
|
rhill Retired Dev
Joined: 22 Oct 2004 Posts: 1629 Location: sk.ca
|
Posted: Sun Nov 11, 2007 11:12 pm Post subject: |
|
|
here's a patch for gmp which should fix the multiple declaration errors when including gmp.h
Code: | diff -Naur gmp-4.2.1-orig/gmp-h.in gmp-4.2.1/gmp-h.in
--- gmp-4.2.1-orig/gmp-h.in 2006-04-11 12:10:15.000000000 -0600
+++ gmp-4.2.1/gmp-h.in 2007-11-11 16:57:26.000000000 -0600
@@ -24,6 +24,7 @@
#if defined (__cplusplus)
#include <iosfwd> /* for std::istream, std::ostream, std::string */
+#include <cstdio> /* for std::FILE */
#endif
@@ -418,9 +419,15 @@
/* gcc has __inline__ in all modes, including strict ansi. Give a prototype
for an inline too, so as to correctly specify "dllimport" on windows, in
- case the function is called rather than inlined. */
+ case the function is called rather than inlined.
+ GCC 4.3 and above with -std=c99 or -std=gnu99 implements ISO C99
+ inline semantics, unless -fgnu89-inline is used. */
#ifdef __GNUC__
+#ifdef __GNUC_STDC_INLINE__
+#define __GMP_EXTERN_INLINE extern __inline__ __attribute__ ((__gnu_inline__))
+#else
#define __GMP_EXTERN_INLINE extern __inline__
+#endif
#define __GMP_INLINE_PROTOTYPES 1
#endif
diff -Naur gmp-4.2.1-orig/tests/cxx/t-locale.cc gmp-4.2.1/tests/cxx/t-locale.cc
--- gmp-4.2.1-orig/tests/cxx/t-locale.cc 2006-03-14 09:57:54.000000000 -0600
+++ gmp-4.2.1/tests/cxx/t-locale.cc 2007-11-11 16:55:36.000000000 -0600
@@ -20,6 +20,7 @@
MA 02110-1301, USA. */
#include <clocale>
+#include <cstdlib>
#include <iostream>
#include "gmp.h" |
https://bugs.gentoo.org/show_bug.cgi?id=198830 _________________ by design, by neglect
for a fact or just for effect |
|
Back to top |
|
|
b33fc0d3 Retired Dev
Joined: 22 Jun 2006 Posts: 128
|
Posted: Mon Nov 12, 2007 3:47 am Post subject: Re: Working CFLAGS |
|
|
dirtyepic wrote: | this doesn't actually do anything unless -ftree-vectorize is enabled. |
note that i'm using -O3. |
|
Back to top |
|
|
|