| Summary: | sys-libs/glibc-2.9_p20081201-r1: broken toolchain | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Jimmy.Jazz |
| Component: | [OLD] GCC Porting | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
| Status: | RESOLVED NEEDINFO | ||
| Severity: | major | ||
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
ebuild log
sandbox log sys.lst used for the test glibc log same log as 179374 + debug glibc log with an other king of error |
||
|
Description
Jimmy.Jazz
2009-01-18 21:42:23 UTC
Created attachment 178928 [details]
ebuild log
* sys-kernel/linux-headers
Latest version available: 2.6.28-r1
Latest version installed: 2.6.28-r1
Created attachment 179052 [details]
sandbox log
What a mess. I have now the same issue with sandbox ebuild. It won't compile for ABI=x86 again. The toolchain seems broken for amd64.
I finally managed to compile glibc and sandbox after upgrading to sys-apps/portage-2.2_rc23, uninstalled sys-power/pm-utils to avoid the pkg-config bug during a full emerge. Also, this time the distccd server was down during the compilation.
The bug should still be taken in consideration. It shouldn't append and the problem could be related to the modification I have made above. Moreover, the issue is quite similar as when you do a crossdev -t x86_64-pc-linux-uclibc. The compilation won't pass stage 4.
To illustrate the process, I made a little script without any pretention that read a simplified list that is the result of emerge -ep @system output.
#!/bin/sh
trap "exit 2" INT TERM
liste=$1
options=$2
succes=/root/succes.lst
retry=/root/retry.lst
#> ${succes}
#> ${retry}
for i in $(grep -v "^#" < ${liste}); do
echo
echo "proceed ${i}"
emerge -1${options} ${i} || echo ${i} >> ${retry}
echo ${i} >> ${succes}
done
see attachment for the packages list used.
emerge -ep @system > sys.lst
sed -i -e '1,4d;s/\[ebuild ......\] /=/;s/ .*$//;1,1s/^/#/' sys.lst
please be careful about emerge -1. It doesn't stay any packages that request an upgrade. All the packages need to be marked with a R in sys.lst
Created attachment 179183 [details]
sys.lst used for the test
try building with distcc disabled. if that works, clear the ccache of all your machines and try again with distcc enabled. Created attachment 179374 [details]
glibc log
the worst case,
ccache -C
ccache -z
# set | grep -i ccache
CCACHE_DIR=/var/lib/cache/ccache
CCACHE_LOGFILE=/var/log/ccache.log
CCACHE_NOLINK=
CCACHE_PREFIX=
CCACHE_UMASK=002
cat /etc/make.conf
FEATURES="-distcc ccache userpriv sandbox parallel-fetch userfetch fixpackages collision-protect splitdebug"
There is a missing file when CCACHE_PREFIX is null.
No such file or directory
* Checking gcc for __thread support ... [ !! ]
Next log with debug mode activated. Same condition.
Created attachment 179375 [details]
same log as 179374 + debug
Created attachment 179376 [details]
glibc log with an other king of error
Same condition as before but CCACHE_PREFIX has been set
CCACHE_PREFIX="distcc"
# set | grep -i cache
CCACHE_DIR=/var/lib/cache/ccache
CCACHE_LOGFILE=/var/log/ccache.log
CCACHE_NOLINK=
CCACHE_PREFIX=distcc
CCACHE_UMASK=002
DISTCC_DIR=/var/lib/cache/distcc
ccache has been flushed and stat reset
After emerge failed
# ccache -s
cache directory /var/lib/cache/ccache
cache hit 34
cache miss 4273
called for link 58
multiple source files 1
compile failed 30
preprocessor error 26
not a C/C++ file 623
autoconf compile/link 355
unsupported compiler option 361
no input file 165
files in cache 8546
cache size 53.2 Mbytes
max cache size 976.6 Mbytes
The ccache of the distccd server has been reset too. Also there was no activity of distccd daemon when during the time emerge has been run. Controled with distccmon-gnome gui. Just four jobs on localhost were working.
The error is reproducible but not permanent. Also, the type of error depend on the ccache variables value.
genlop glibc
Mon Dec 8 13:04:17 2008 >>> sys-libs/glibc-2.9_p20081201
Sat Jan 10 18:32:00 2009 >>> sys-libs/glibc-2.9_p20081201-r1
Tue Jan 20 14:12:21 2009 >>> sys-libs/glibc-2.9_p20081201-r1
After unsuccessful emerging of the anjuta package, I took a look at ccache log file. It is filed with lines of, argument -x is unsupported When disabling ccache in FEATURES, that package is able to compile again. That is certainly a part of the problem. |
