--- Begin Message ---
|
Subject: |
coreutils-8.30: test sort-u-FMR fails after rebuild |
|
Date: |
Tue, 28 Aug 2018 20:42:25 +0200 |
|
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:21.0) Gecko/20100101 Thunderbird/52.9.1 |
Hi,
after updating Gentoo's toolchain to GCC 8.2.0 and an entire world rebuild
test "misc/sort-u-FMR" is failing:
> + require_valgrind_
> + valgrind --error-exitcode=1 true
> + echo 0
> + printf '%0900d\n' 1
> + valgrind --error-exitcode=1 sort --p=1 -S32b -u in
> ==46265== Memcheck, a memory error detector
> ==46265== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
> ==46265== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
> ==46265== Command: sort --p=1 -S32b -u in
> ==46265==
> ==46265== Conditional jump or move depends on uninitialised value(s)
> ==46265== at 0x49249CE: ??? (in /lib64/libc-2.27.so)
> ==46265== by 0x48F4B31: ??? (in /lib64/libc-2.27.so)
> ==46265== by 0x7FFFFFFFFFFFF: ???
> ==46265==
> ==46265== Conditional jump or move depends on uninitialised value(s)
> ==46265== at 0x491EE81: ??? (in /lib64/libc-2.27.so)
> ==46265== by 0x1100D6: ??? (in
> /var/tmp/portage/sys-apps/coreutils-8.30/work/coreutils-8.30/src/sort)
> ==46265== by 0x2F: ???
> ==46265== by 0x402BCC2: ???
> ==46265== by 0x30: ???
> ==46265== by 0x5: ???
> ==46265== by 0x402A36F: ???
> ==46265== by 0x1FFEFF471F: ???
> ==46265== by 0xA00000000000009: ???
> ==46265== by 0x1FFEFF55A6: ???
> ==46265== by 0x402B84F: ???
> ==46265== by 0x402A7DF: ???
> ==46265==
> ==46265==
> ==46265== HEAP SUMMARY:
> ==46265== in use at exit: 0 bytes in 0 blocks
> ==46265== total heap usage: 0 allocs, 0 frees, 0 bytes allocated
> ==46265==
> ==46265== All heap blocks were freed -- no leaks are possible
> ==46265==
> ==46265== For counts of detected and suppressed errors, rerun with: -v
> ==46265== Use --track-origins=yes to see where uninitialised values come from
> ==46265== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 0 from 0)
Not sure what's causing the problem. On a system which was built using
gcc-7.3.0 test is passing if I only switch GCC to v8.2.0 and rebuild.
However, I was able to reproduce the test failure on 3 physical systems
after complete rebuild.
Complete logs are available at
https://gist.github.com/Whissi/dde2960c1339a73d66c09613864488a5
--
Regards,
Thomas Deutschmann / Gentoo Linux Developer
C4DD 695F A713 8F24 2AA1 5638 5849 7EE5 1D5D 74A5
--- End Message ---
--- Begin Message ---
|
Subject: |
Re: coreutils-8.30: test sort-u-FMR fails after rebuild |
|
Date: |
Thu, 30 Aug 2018 15:29:07 +0200 |
|
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:21.0) Gecko/20100101 Thunderbird/52.9.1 |
Update:
This was caused by valgrind which was unable to read read-only PT_LOAD
segments and .rodata. In the end this wasn't triggered by glibc-2.27 but
by binutils which now defaults to "ld -z" (I didn't notice before
because bintuils was updated many weeks before so this was actually also
my first complete rebuild using the new binutils).
There's already a valgrind patch:
https://sourceware.org/git/gitweb.cgi?p=valgrind.git;a=commit;h=64aa729bfae71561505a40c12755bd6b55bb3061
And a follow-up patch:
https://sourceware.org/git/gitweb.cgi?p=valgrind.git;a=commit;h=e752326cc050803c3bcfde1f8606bead66ff9642
This bug report can be closed, thanks.
--
Regards,
Thomas Deutschmann / Gentoo Linux Developer
C4DD 695F A713 8F24 2AA1 5638 5849 7EE5 1D5D 74A5
--- End Message ---