Hi,
 
> I pushed some fixes,see https://gerrit.osmocom.org/1530 and
> https://gerrit.osmocom.org/1531  - the remaining failures are all
> LeakSanitizer warnings, and I suspect that while we release all memory
> back to talloc, talloc itself doesn't release it back to glibc
> malloc/free, and hence LeakSanitizer is kicking in.

Just rebased to a new version with both changes applied, and
I can confirm, that the 'gea' test is ok now. Other still fail.

>> Max wrote:
> I wonder why this has not been triggering jenkins build failures?
> Are we using it wrong?

Interesting questions for me too.
Which GCC version is used on buildhost?
Was one compiled manually?


With best regards,
Vadim Yanitskiy.

2017-01-04 17:28 GMT+07:00 Harald Welte <laforge@gnumonks.org>:
On Wed, Jan 04, 2017 at 08:55:45AM +0000, Vadim Yanitskiy wrote:
> I think it's due to --enable-sanitize, because
> when I am trying to compile libosmocore this way:
>
>  ./configure --enable-static --enable-sanitize
>  make check
>
> Some other tests fail (current master version, clang-3.6):
>
>  Regression tests.
>   8: gea                                             FAILED (testsuite.at:51)
>   9: msgfile                                         FAILED (testsuite.at:59)
>  15: lapd                                            FAILED (testsuite.at:96)
>  16: gsm0808                                         FAILED (testsuite.at:102)
>  17: gsm0408                                         FAILED (testsuite.at:108)
>  30: bssgp-fc                                        FAILED (testsuite.at:190)

I pushed some fixes,see https://gerrit.osmocom.org/1530 and
https://gerrit.osmocom.org/1531  - the remaining failures are all
LeakSanitizer warnings, and I suspect that while we release all memory
back to talloc, talloc itself doesn't release it back to glibc
malloc/free, and hence LeakSanitizer is kicking in.

A quick online search didn't seem like anyone else has yet come up with
a solution for this, so I guess we should turn it off.  I remember some
discussion on the list about a year ago where the concensus was to
disable LeakSanitizer by environment variables on the buildhost anyway?
--
- Harald Welte <laforge@gnumonks.org>           http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
                                                  (ETSI EN 300 175-7 Ch. A6)