On 10/23/12 10:19, Peter Maydell wrote: > On 23 October 2012 08:03, Gerd Hoffmann <kra...@redhat.com> wrote: >> On 10/22/12 15:51, Peter Maydell wrote: >>> On 22 October 2012 14:28, Gerd Hoffmann <kra...@redhat.com> wrote: >>>> Commit 585f60368f23e6603cf86cfdaeceb89d1169f4b8 appearently breaks the >>>> libvirt feature detection, my guests fail to start with this message: >>>> >>>> error: Failed to start domain fedora-org-virtio >>>> error: internal error qemu does not support SGA >>>> >>>> Running libvirt-0.10.2-2.el6.x86_64 >>> >>> Yeah, this is the commit that means you need an updated libvirt >>> from this point on (and in particular for qemu-1.3). >> >> Well, only problem is that 0.10.2 already is the latest libvirt release. >> Was it really intentional to break things like this? > > This change was postponed to after 1.2 was released to > give libvirt a chance to wean itself off parsing our --help > output.
Yea, I know this has been the plan for a long time and I agree that it is a good move. Only problem is that the switch didn't happen yet. The bits might be landed in libvirt/master, but there is no release with this yet and thus libvirt versions using QOM for feature detection didn't find the way yet into distributions. IMO it is a bit early to stop caring about -help output compatibility in qemu. cheers, Gerd
