On Fri, Jun 08, 2012 at 10:17:05AM +0200, Andreas Färber wrote:
Am 08.06.2012 09:44, schrieb Anthony Liguori:
On 06/08/2012 03:11 PM, Andreas Färber wrote: Since qdev->id is NULL 90% of the time, I don't think a user can realistically rely on it. I don't think changing the type of the data in the error is going to be a problem.
Doesn't libvirt ignore the contents of an error object?
I'm out of my field there, those questions are for Luiz and the libvirt guys to answer. (Context is ongoing DeviceState -> Object transition on qom-next branch, properties being moved to Object and what info to include in Error objects then)
Libvirt will look at two fields in the a JSON 'error' reply. We will pass 'desc' field through to the libvirt code - we treat it as an opaque value. We will look at the string value in the 'class' field to check for certain types of error - eg we strcmp against things like MigrationExpected, DeviceNotActive, CommandNotFound, KVMMissingCap, DeviceInUse, etc. All other fields are ignored. Regards, 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 :|