[libvirt] libvirt-0.9.3 week freeze, RC1 version to test

[Matthias Bolte]

2011/6/27 Laine Stump <laine at laine.org>:
> On 06/27/2011 07:32 AM, Matthias Bolte wrote:
>>
>> 2011/6/27 simon<simonjin at linux.vnet.ibm.com>:
>>>
>>> 于 2011-6-27 13:01, Daniel Veillard 写道:
>>>
>>>   So we are entering the week freeze for the new version 0.9.3
>>> I have made an rc1 tarball available at:
>>>    ftp://libvirt.org/libvirt/libvirt-0.9.3-rc1.tar.gz
>>> along with rpms for those interested
>>> We pushed a lot of patches in the last few days, we know there is at
>>> least some cleanup issues, but I tried it locally and basic operations
>>> seems to work fine,
>>>
>>>   please give it a try too and report issues,
>>>
>>>    thanks !
>>>
>>> Daniel
>>>
>>> On fc14.x86_64
>>> TEST: networkxml2argvtest  FAIL
>>>
>>> Download  libvirt-0.9.3-rc1.tar.gz
>>>
>>> ./configure&&  make&&  make check
>>>
>>> It report 1 of 52 tests failed
>>>
>>> TEST: networkxml2argvtest
>>>       !.!!!!!                                  7   FAIL
>>> FAIL: networkxml2argvtest
>>
>> This a known problem when you configure libvirt with localstatedir
>> other than /var, what happens when you just run configure without
>> arguments.
>
> It looks like that's what he did, so maybe this is a different problem (or
> possibly his configure defaults to /usr/local/var?).
>
> If the default localstatedir isn't /var, try doing "./autogen.sh --system"
> instead of ./configure. If that doesn't cure the problem, try running the
> test with debug logging turned on to see exactly what error is reported:
>
>  LIBVIRT_LOG_OUTPUTS=3:stderr VIR_TEST_DEBUG=2 ./networkxml2argvtest
>

5 of 6 failing tests are fixed by this patch I just pushed:

  https://www.redhat.com/archives/libvir-list/2011-June/msg01358.html

The last one is related to the fact that
networkBuildDhcpDaemonCommandLine writes dnsmasq config files to disk
relative to DNSMASQ_STATE_DIR and DNSMASQ_STATE_DIR is configure
--localstatedir dependent.

You also noticed this in your review:

  https://www.redhat.com/archives/libvir-list/2011-June/msg01284.html

As Dan suggested on IRC this should be refactored and will then also
fix the last test failure.

<danpb> yeah this is bad code structure IMHO
<danpb> the call to networkSaveDnsmasqHostsfile() should be removed
from networkBuildDhcpDaemonCommandLine / networkBuildDnsmasqArgv
<danpb> whatever calls networkBuildDhcpDaemonCommandLine  should be
made responsible for actually creating the hosts file
<danpb> the networkBuildDnsmasqArgv method should solely create the
CLI args, and assume someone else has taken care of actualyl creating
the file on disk

-- 
Matthias Bolte
http://photron.blogspot.com