On Fri, Aug 31, 2012 at 11:49:52AM -0700, Eric Blake wrote:
On 08/29/2012 03:03 AM, Alex Jia wrote:
> From: Alex Jia <Alex Jia ajia(a)redhat.com>
>
> The default flags are inconsistent on both qemuSetSchedulerParameters()
> and qemuGetSchedulerParameters() in libvirt, the qemuGetSchedulerParameters()
> always passes 'VIR_DOMAIN_AFFECT_CURRENT' flag to the
qemuGetSchedulerParametersFlags(),
> it should be a expected behavior, but the qemuSetSchedulerParameters() always
passes
> 'VIR_DOMAIN_AFFECT_LIVE' flag to the qemuSetSchedulerParametersFlags(), if
users use
> default flag=0 or explicitly give a 'VIR_DOMAIN_AFFECT_CURRENT' flag to the
> set_scheduler_parameters() in perl-Sys-Virt, because the flag value is 0, the result
is
> the virDomainSetSchedulerParameters() is called incorrectly.
>
> In fact, it's enough to call the qemuSetSchedulerParametersFlags() regardless of
what flag is
> like 'virsh schedinfo' command.
But doing this makes it harder to talk to older libvirtd that lacked the
new function. I'm not sure whether we want to blindly force the need
for the newer function to exist, or whether we should be reproducing the
same logic for deciding when to try the older API based on the flags
presented by the user.
Yeah, the intent of this conditional code in the Perl bindings is to
be as conservative as possible. So use the old APIs in all scenarios
where they provide the required functionality.
Daniel
--
|:
http://berrange.com -o-
http://www.flickr.com/photos/dberrange/ :|
|:
http://libvirt.org -o-
http://virt-manager.org :|
|:
http://autobuild.org -o-
http://search.cpan.org/~danberr/ :|
|:
http://entangle-photo.org -o-
http://live.gnome.org/gtk-vnc :|