GCC 4.5 testing

rhill · Posted: Sat Dec 12, 2009 9:25 pm Post subject: GCC 4.5 testing

GCC 4.5 went into prerelease mode last weekish. I've created the tracker bug for 4.5 porting and have started build testing. As always, anyone filing good bug reports with patches included are welcome to help.

https://bugs.gentoo.org/show_bug.cgi?id=296658

At the moment I'm concerned about things failing with a standard configuration. Ie. if you want to test, please have your system built with safe CFLAGS, use the standard linker (not gold), and leave graphite and LTO off. Also, you need to be running ~arch. It might seem obvious but we always get people that try to run an unreleased compiler on a stable system.

Switching between compiler versions can break C++ programs that use symbol versioning. When switching to an older release from a newer one, if you encounter runtime errors like "libstdc++.so.6: version `GLIBCXX_3.4.9' not found", rebuild the package with the current compiler.

GCC snapshots and svn ebuilds are in the toolchain overlay. The gcc-porting overlay will be used again as a repository for patches not yet applied to the tree.

Bug reports must include patches, and all patches must go upstream.
_________________
by design, by neglect
for a fact or just for effect

me22 · n00b Joined: 30 Jun 2002 Posts: 45 Location: Canada

rhill · Posted: Tue Dec 22, 2009 6:41 am Post subject:

sure, that's fine. we just get always get a number of bug reports from people who use svn builds of toolchain packages on their stable systems and then complain when emerge @world fails. if it's not selected by gcc-config then you don't have to worry.
_________________
by design, by neglect
for a fact or just for effect

me22 · n00b Joined: 30 Jun 2002 Posts: 45 Location: Canada

justinkb · Apprentice Joined: 23 Dec 2008 Posts: 161

Running into this... I guess it's only an annoyance, but still...

SamuliSuominen · Posted: Fri Dec 25, 2009 9:33 pm Post subject:

few · Guru Joined: 03 Mar 2008 Posts: 448

Would it be possible to get a lto useflag for the snapshots? Current gcc-4.5.0_alpha20100107 snapshot fails to compile.
EDIT: 4.5.0_pre999 fails too...

* updating multilib directories to be: ../lib64 ../lib32
cp: cannot stat `/var/tmp/portage/sys-devel/gcc-4.5.0_pre9999/distdir/ecj-4.3.jar': No such file or directory
* ERROR: sys-devel/gcc-4.5.0_pre9999 failed:

Comparing stages 2 and 3
warning: gcc/cc1plus-checksum.o differs
warning: gcc/cc1-checksum.o differs
Bootstrap comparison failure!
gcc/builtins.o differs
make[2]: *** [compare] Error 1
make[2]: Leaving directory `/var/tmp/portage/sys-devel/gcc-4.5.0_alpha20100107/work/build'
make[1]: *** [stage3-bubble] Error 2
make[1]: Leaving directory `/var/tmp/portage/sys-devel/gcc-4.5.0_alpha20100107/work/build'
make: *** [bootstrap-lean] Error 2
* ERROR: sys-devel/gcc-4.5.0_alpha20100107 failed:
* emake failed with bootstrap-lean
*
* Call stack:
* ebuild.sh, line 48: Called src_compile
* environment, line 5040: Called toolchain_src_compile
* environment, line 5600: Called gcc_src_compile
* environment, line 3188: Called gcc_do_make
* environment, line 2978: Called die
* The specific snippet of code:
* emake LDFLAGS="${LDFLAGS}" STAGE1_CFLAGS="${STAGE1_CFLAGS}" LIBPATH="${LIBPATH}" BOOT_CFLAGS="${BOOT_CFLAGS}" ${GCC_MAKE_TARGET} || die "emake failed with ${GCC_MAKE_TARGET}";
*

rhill · Posted: Sat Jan 30, 2010 2:16 am Post subject:

LTO will be enabled if libelf is detected during the build. If you have elfutils installed it should work out of the box.

i know the ldconfig messages are annoying. just delete them if they bother you, they're just pretty-printers for gdb-7.0. i need to sit down and figure out where else they can be stashed that's outside of the library path.

if anyone's had trouble building boost the last week or two, it was fixed today.
_________________
by design, by neglect
for a fact or just for effect

few · Guru Joined: 03 Mar 2008 Posts: 448

rhill · Posted: Thu Feb 04, 2010 3:01 am Post subject:

why is it wrong?

lto is automatically enabled if libelf is found. i guess i could add a USE flag to make it explicit though.
_________________
by design, by neglect
for a fact or just for effect

cruzki123 · Apprentice Joined: 16 May 2008 Posts: 262

I'm having this error whis the two last snapshots:

rhill · Posted: Sat Feb 06, 2010 4:52 am Post subject:

it's an upstream issue. i think it's been fixed now, so wait until the next snapshot or try the svn ebuild.
_________________
by design, by neglect
for a fact or just for effect

cruzki123 · Apprentice Joined: 16 May 2008 Posts: 262

OK thanks.

EDIT: the new snapshop compiles fine. Thanks

few · Guru Joined: 03 Mar 2008 Posts: 448

rhill · Posted: Tue Feb 09, 2010 10:58 pm Post subject:

ah, i was wondering what the min version might be. I'll update the depend next time i touch the snapshots.
_________________
by design, by neglect
for a fact or just for effect

cruzki123 · Apprentice Joined: 16 May 2008 Posts: 262

EDIT: Don't try to compile a kernel with gcc-4.5. Compile fine, but when udev it is trying to do it's magic you obtain a nice kernel panic.

I just found something tricky

if you unistall libelf, gcc doesn't work any more.
I have to gcc-config 4.4.3; emerge elfutils gcc and then try again.
I suppose that at least a use flag shoul be used but I think that elfutils should be in system profile with a very VERY fat warnig.

On the other hand the following packages did not compile in a world rebuild (700+ packages) with graphite enable:
(lto did not work, but with a program I develop works like a charm and it's about 10% faster.)

app-office:
total 8
drwxrwxr-x 7 portage portage 4096 feb 10 01:12 kword-2.1.0
drwxrwxr-x 4 portage portage 4096 feb 10 00:44 openoffice-3.1.1

dev-db:
total 4
drwxrwxr-x 4 portage portage 4096 feb 9 23:56 hsqldb-1.8.0.10

dev-java:
total 140
drwxrwxr-x 4 portage portage 4096 feb 9 22:09 ant-core-1.7.1-r4
drwxrwxr-x 4 portage portage 4096 feb 9 22:34 ant-nodeps-1.7.1
drwxrwxr-x 4 portage portage 4096 feb 9 23:44 bcel-5.2
drwxrwxr-x 4 portage portage 4096 feb 9 23:53 bsh-2.0_beta4-r3
drwxrwxr-x 4 portage portage 4096 feb 10 00:08 dom4j-1.6.1-r6
drwxrwxr-x 7 portage portage 4096 feb 10 00:37 icedtea-6.1.7
drwxrwxr-x 4 portage portage 4096 feb 9 23:45 icu4j-3.8.1-r1
drwxrwxr-x 4 portage portage 4096 feb 9 23:49 iso-relax-20050331-r2
drwxrwxr-x 4 portage portage 4096 feb 9 23:53 javacc-4.2
drwxrwxr-x 4 portage portage 4096 feb 9 23:45 javacup-0.11a_beta20060608
drwxrwxr-x 4 portage portage 4096 feb 10 00:09 jaxen-1.1.1
drwxrwxr-x 4 portage portage 4096 feb 10 00:09 jdom-1.0-r4
drwxrwxr-x 4 portage portage 4096 feb 10 00:10 jdom-jaxen-1.0-r1
drwxrwxr-x 4 portage portage 4096 feb 9 22:34 junit-3.8.2-r1
drwxrwxr-x 4 portage portage 4096 feb 9 22:17 libreadline-java-0.8.0-r2
drwxrwxr-x 4 portage portage 4096 feb 9 23:57 lucene-2.3.2
drwxrwxr-x 4 portage portage 4096 feb 9 22:24 lucene-analyzers-2.3.2
drwxrwxr-x 4 portage portage 4096 feb 10 00:08 msv-20050627-r2
drwxrwxr-x 4 portage portage 4096 feb 9 22:14 relaxng-datatype-1.0-r1
drwxrwxr-x 4 portage portage 4096 feb 9 22:24 rhino-1.5.5-r5
drwxrwxr-x 4 portage portage 4096 feb 10 00:10 rhino-1.6.5
drwxrwxr-x 4 portage portage 4096 feb 9 22:15 saxpath-1.0-r2
drwxrwxr-x 4 portage portage 4096 feb 9 22:46 servletapi-2.4-r5
drwxrwxr-x 4 portage portage 4096 feb 9 22:36 tomcat-servlet-api-4.1.36
drwxrwxr-x 4 portage portage 4096 feb 9 23:58 xalan-2.7.1
drwxrwxr-x 4 portage portage 4096 feb 9 23:51 xalan-serializer-2.7.1
drwxrwxr-x 4 portage portage 4096 feb 9 23:53 xerces-2.9.1
drwxrwxr-x 4 portage portage 4096 feb 9 23:50 xjavac-20041208-r5
drwxrwxr-x 4 portage portage 4096 feb 9 22:13 xml-commons-external-1.3.04
drwxrwxr-x 4 portage portage 4096 feb 9 22:17 xml-commons-resolver-1.2
drwxrwxr-x 4 portage portage 4096 feb 10 00:10 xml-xmlbeans-1.0.4_pre20041217
drwxrwxr-x 4 portage portage 4096 feb 9 23:58 xom-1.0-r7
drwxrwxr-x 4 portage portage 4096 feb 10 00:02 xpp2-2.1.10-r1
drwxrwxr-x 4 portage portage 4096 feb 9 23:46 xpp3-1.1.4c-r1
drwxrwxr-x 4 portage portage 4096 feb 9 23:57 xsdlib-20050627-r2

dev-lang:
total 4
drwxrwxr-x 7 portage portage 4096 feb 10 00:43 R-2.10.1

dev-libs:
total 8
drwxrwxr-x 7 portage portage 4096 feb 9 22:02 boost-1.41.0-r3
drwxrwxr-x 4 portage portage 4096 feb 9 22:10 soprano-2.3.73

dev-tex:
total 4
drwxrwxr-x 4 portage portage 4096 feb 10 01:03 tex4ht-20090611_p1038-r1

kde-misc:
total 4
drwxrwxr-x 7 portage portage 4096 feb 9 23:44 yawp-0.3.2

media-gfx:
total 4
drwxrwxr-x 7 portage portage 4096 feb 10 00:24 gimp-2.6.8

media-libs:
total 4
drwxrwxr-x 7 portage portage 4096 feb 9 20:54 mesa-7.7-r1

media-plugins:
total 8
drwxrwxr-x 7 portage portage 4096 feb 9 22:08 gst-plugins-ffmpeg-0.10.9
drwxrwxr-x 7 portage portage 4096 feb 9 22:05 gst-plugins-taglib-0.10.17

media-video:
total 8
drwxrwxr-x 7 portage portage 4096 feb 9 22:31 avidemux-2.5.2
drwxrwxr-x 7 portage portage 4096 feb 9 22:11 ffmpeg-0.5_p20373

net-libs:
total 4
drwxrwxr-x 4 portage portage 4096 feb 10 00:20 xulrunner-1.9.2-r1

sci-libs:
total 8
drwxrwxr-x 4 portage portage 4096 feb 9 20:11 blas-reference-20070226
drwxrwxr-x 4 portage portage 4096 feb 9 20:15 lapack-reference-3.2.1

sci-mathematics:
total 8
drwxrwxr-x 7 portage portage 4096 feb 10 01:02 octave-3.2.4
drwxrwxr-x 7 portage portage 4096 feb 10 00:44 rkward-0.5.2

sys-devel:
total 4
drwxrwxr-x 7 portage portage 4096 feb 9 20:24 gcc-4.5.0_alpha20100204

www-client:
total 4
drwxrwxr-x 4 portage portage 4096 feb 10 00:24 mozilla-firefox-3.6-r1

There are some of them that was related to gcc not being compiled with fortran, openmp and gcj. Some other hang in configure o link phase without any reason. I have to kill the prosses. I'm rebuilding gcc and then I will try again.

rtomek · Posted: Wed Feb 10, 2010 8:35 pm Post subject:

well I just built kde 4.4 with the 4.5.0 snapshot from 2 days ago. It all worked!

rhill · Posted: Fri Feb 12, 2010 5:37 am Post subject:

ComaWhite · Tux's lil' helper Joined: 07 Oct 2008 Posts: 125

GCC-4.5 finally hitting RC stage. I wonder if anyone gave up on testing GCC-4.5.0 ;P

rhill · Posted: Mon Apr 05, 2010 1:01 am Post subject:

i haven't had it break anything in months. it's been kinda anticlimactic.

_________________
by design, by neglect
for a fact or just for effect

cruzki123 · Apprentice Joined: 16 May 2008 Posts: 262

Last time I try it broke my system. But I use lto an agresive graphite flags :S

I will tryin a couple of days one more time. is rc1 in toolchain already?

rhill · Posted: Mon Apr 05, 2010 11:30 pm Post subject:

that would probably do it. :wink:

i'd hold off LTO til libtool gets proper support. hopefully graphite works better than it does in 4.4.

rc1 isn't released yet. they're in code freeze right now, they'll branch sometime soon and then do RC's.
_________________
by design, by neglect
for a fact or just for effect

rhill · Posted: Sat Apr 10, 2010 9:39 pm Post subject:

the 4.5 branch has been created and the live ebuild tracks it instead of trunk now. i'm not planning on adding ebuilds for the RCs - just use the live ones, they'll be close enough.
_________________
by design, by neglect
for a fact or just for effect

Etal · Veteran Joined: 15 Jul 2005 Posts: 1932

It's out!

http://gcc.gnu.org/ml/gcc/2010-04/msg00321.html
http://gcc.gnu.org/gcc-4.5/changes.html

Shining Arcanine · Veteran Joined: 24 Sep 2009 Posts: 1110