v2.2.15 - make check - Conditional jump or move depends on uninitialised value
Hello,
I am currently using Dovecot 2.2.10 on Fedora 16 - 64 bit system
I had made v2.2.10 Fedora 16 rpm file using spec file from ATrpms. http://dl.atrpms.net/all/dovecot.spec
It has been working well from 6 months or so.
Today I tried to make v2.2.15 rpm using same spec file.
But "make check" is giving following error:
fatal_printf_format_fix .............................................. : ok 0 / 190 tests failed ==8834== Conditional jump or move depends on uninitialised value(s) ==8834== at 0x4F7FE99: __strspn_sse42 (in /lib64/libc-2.14.90.so) ==8834== by 0x10FCF1: test_ds_recurse (test-data-stack.c:123) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834== by 0x10FCC9: test_ds_recurse (test-data-stack.c:119) ==8834==
There are more similar errors after that and then rpmbuild aborts.
I already saw that someone reported same error recently: http://dovecot.org/pipermail/dovecot/2014-October/098452.html
But I do not know what is the right solution / patch?
Can I disable "make check" safely? Its a production server so would not take any risk.
Can anyone please tell what should I do?
Thanks in advance.
AMM
On 11/30/2014 05:53 AM, AMM wrote:
__strspn_sse42 (in /lib64/libc-2.14.90.so) Is it possible that you are encountering this issue? https://bugs.kde.org/show_bug.cgi?id=270925 Either way, the error seems to stem from your libc implementation (if it is not the valgrind bug).
If possible, upgrade your valgrind, libc etc.
br, Teemu Huovila
On Monday 01 December 2014 03:41 PM, Teemu Huovila wrote:
On 11/30/2014 05:53 AM, AMM wrote:
__strspn_sse42 (in /lib64/libc-2.14.90.so) Is it possible that you are encountering this issue? https://bugs.kde.org/show_bug.cgi?id=270925 Either way, the error seems to stem from your libc implementation (if it is not the valgrind bug).
If possible, upgrade your valgrind, libc etc.
br, Teemu Huovila
Thank you for reply.
But Dovecot 2.2.10 (and earlier versions) were not throwing this error.
Can I can ignore it by NOT doing "make check"?
OR
is it something serious that if I ignore it can corrupt mailboxes?
Unfortunately for some reasons I can not update libc or valgrind.
Amm
On 12/01/2014 12:41 PM, AMM wrote:
On Monday 01 December 2014 03:41 PM, Teemu Huovila wrote: But Dovecot 2.2.10 (and earlier versions) were not throwing this error.
This test was added in Dovecot version 2.2.14. It is also the only reference to strspn() in the whole project.
Can I can ignore it by NOT doing "make check"? I would say you can safely ignore it, but I can give no guarantee. I have no access to a Fedora 16 system, so I can not verify it, but I would say this is most likely a manifestation of the valgrind bug I linked in my first email.
You could try verification yourself, by using the "Steps to Reproduce" in the linked issue tracker.
br, Teemu Huovila
On Monday 01 December 2014 05:24 PM, Teemu Huovila wrote:
On 12/01/2014 12:41 PM, AMM wrote:
On Monday 01 December 2014 03:41 PM, Teemu Huovila wrote: But Dovecot 2.2.10 (and earlier versions) were not throwing this error. This test was added in Dovecot version 2.2.14. It is also the only reference to strspn() in the whole project. ... You could try verification yourself, by using the "Steps to Reproduce" in the linked issue tracker.
br, Teemu Huovila
Verified it. It is indeed valgrind bug.
Thanks for the help.
Amm.
participants (2)
-
AMM
-
Teemu Huovila