On Tue, Feb 04, 2014 at 11:34:39AM -0700, Eric Blake wrote:
On 02/04/2014 11:28 AM, Franky Van Liedekerke wrote:
>> Are you in a position to bisect which libvirt patch introduced the
>> problem? It looks like our first use of qom-list was in 1.1.1, with
>> commit d76a897.
>
> Errr ... I don't really know what you mean with bisect ... but tell me
> what to do, and I'll try my best tomorrow.
'git bisect' is a powerful driver that lets you do a binary search
through git history to find which patch changed a particular behavior.
If you are comfortable using git and building your own libvirt, I can
help you come up with a command line to drive the test for looking for
where this behavior started.
I wonder if we should not add a small section on bugs.html.in about
how git bissect can help locating the bug and the simpler alternatives
like below:
But even if you are not, it would be nice
if you could test a build of libvirt 1.1.0 vs. libvirt 1.1.1; if my
guess above was correct, 1.1.0 will not have the problem. Or, if you
are comfortable using gdb, it might be nice to step through program
execution at the point where qemuMonitorJSONGetObjectListPaths returns
failure because qemu didn't support the qom-list command. And if not,
the fact that you reported it means that one of the regular developers
will hopefully have some time in the near future to further investigate,
based on what we've learned so far. At any rate, we should have a fix
before releasing libvirt 1.2.2.
3 weeks to find the problem and get a fix :-)
Daniel
--
Daniel Veillard | Open Source and Standards, Red Hat
veillard(a)redhat.com | libxml Gnome XML XSLT toolkit
http://xmlsoft.org/
http://veillard.com/ | virtualization library
http://libvirt.org/