[Debian-ha-maintainers] Bug#803776: libqb FTBFS on linux-amd64 with correct sun_path to bind()

Svante Signell svante.signell at gmail.com
Mon Nov 2 17:39:52 UTC 2015


Source: libqb
Version: 0.17.2.real-4
Severity: normal
Tags: patch
Usertags: linux

Hi,

Currently libqb FTBFS on linux-amd64 for test ipc.test, when applying
the attached patch, fix_ipc_setup_bug.patch. (in fact the unpatched
version FTBFS with gcc-4.9.1-16).

This patch fixes writing of the *.sun_path socket names to become
non-zero for bind() and connect().
The previous race condition is back for the ipc_stress_connections_us
test. Adding the unlink statement solves this problem (unknown how,
since res=-1 due to ENOENT: No such file or directory.)

The patch, rules1.patch removes the named sockets created during the ipc
test to enable successive builds. Otherwise the build fails with:
EADDRINUSE Address already in use (seems like the randomization in
creating these names does not work from build to build)  These removals
can probably be placed somewhere in check_ipc.c, but is placed in
debian/rules for now.

Thanks!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rules1.patch
Type: text/x-patch
Size: 402 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/debian-ha-maintainers/attachments/20151102/4a25e6ee/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fix_ipc_setup_bug.patch
Type: text/x-patch
Size: 1821 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/debian-ha-maintainers/attachments/20151102/4a25e6ee/attachment-0001.bin>


More information about the Debian-ha-maintainers mailing list