[Libvirt-cim] [PATCH 1 of 2] Initialize libvirt early so that we avoid the backend register race

[Kaitlin Rupert]

Dan Smith wrote:
> # HG changeset patch
> # User Dan Smith <danms at us.ibm.com>
> # Date 1195154883 28800
> # Node ID ecc46606e90388217b94916c1eab2ceeb051064a
> # Parent  670294145fcc1a786594f5110e2248cd4ec0b7ed
> Initialize libvirt early so that we avoid the backend register race
>
> This fixes the errant behavior we see with parallel calls to
> connect_by_classname().  This behavior was certainly possible with lv_connect(),
> but I imagine we were getting lucky most of the time.
>
> Signed-off-by: Dan Smith <danms at us.ibm.com>
>   

I was also seeing the same issue after applying the 
connect_by_classname().  I'm using Pegasus with 
forceProviderProcesses=true.  I would see the issue the first call after 
restarting Pegasus. 

After applying this patch, I haven't seen the error.  I've tried 
restarting Pegasus about 10 times, no error yet. 

-- 
Kaitlin Rupert
IBM Linux Technology Center
karupert at us.ibm.com