On Thu, Dec 20, 2012 at 12:45:33PM +0100, Gerd Hoffmann wrote: > >>>> I doubt I manage to finish (and test!) it before xmas. > > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > > > >>> Okay, the QMP interface seems sane to me (from libvirt POV). However, > >>> what about other chardev types like pipe and vc? And I guess pty can be > >>> covered by tty, right? > > > > I think that is the missing part. > > Exactly. > > /me wades through the socket code (unix+tcp) right now, which needs some > refactoring to make it fly. > > >> From libvirt's POV, I think the most important chardev types are pty, > >> unix and tcp. The other types are pretty rarely used AFAICT. > > pty looks like another non-trivial challenge. How does libvirt gather > the pty device today? IIRC there is some stderr parsing? Or was it > info chardev? With QMP we probably want switch to a more sane model > here ...
Yes, these days we use info chardev, or query-chardev as appropriate for the monitor mode. 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 :|