[ovs-discuss] [openvswitch 2.6.0] testsuite: 7 8 111 113 119 121 125 381 761 1859 1860 1861 1862 1863 1864 1868 1870 2247 failed

Ben Pfaff blp at ovn.org
Fri Sep 30 17:26:04 UTC 2016


On Fri, Sep 30, 2016 at 03:07:17AM +0100, Stuart Cardall wrote:
> The patches improved the failures to 11:
> 
> _testsuite: 7 8 761 1859 1860 1861 1862 1863 1864 1868 1870 failed:_
> 
> 7: vsctl-bashcomp - basic verification             FAILED
> (completion.at:352)
> 8: vsctl-bashcomp - argument completion            FAILED
> (completion.at:454)
> 761: tunnel - Geneve option present                  FAILED (tunnel.at:639)
> 1859: insert monitored table, insert excluded table   FAILED
> (ovsdb-server.at:1072)
> 1860: insert monitored table, update excluded table   FAILED
> (ovsdb-server.at:1072)
> 1861: update monitored table, insert excluded table   FAILED
> (ovsdb-server.at:1072)
> 1862: update monitored table, update excluded table   FAILED
> (ovsdb-server.at:1072)
> 1863: delete monitored table, insert excluded table   FAILED
> (ovsdb-server.at:1072)
> 1864: delete monitored table, update excluded table   FAILED
> (ovsdb-server.at:1072)
> 1868: ovsdb-server/set-sync-exclude-tables            FAILED
> (ovsdb-server.at:1151)
> 1870: ovsdb-server/disconnect-active-server           FAILED
> (ovsdb-server.at:1253)

The failures for the ovsdb-server tests are because of the POSIX
noncompliant version of "diff" that Alpine includes by default.  Install
a POSIX-compliant diff (e.g. "apk add diffutils") to fix the problem.

The failure of test 761 seems to be a memory error.  Running under
valgrind, with "make check-valgrind TESTSUITEFLAGS=761", I get a useless
backtrace though:

    ==29841== Invalid free() / delete / delete[] / realloc()
    ==29841==    at 0x4C92A6E: free (vg_replace_malloc.c:530)
    ==29841==    by 0x405717F: ??? (in /lib/ld-musl-x86_64.so.1)
    ==29841==  Address 0x4e9a180 is in a rw- mapped file /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so segment

What does it take to get useful valgrind results on Alpine?

Thanks,

Ben.



More information about the discuss mailing list