A simple matter of…

It’s been quite a struggle to get virt-manager working (particularly using “qemu+ssh” transport as I do). Firstly there was some braindead hard-coding of ssh ports (to 22) in both virt-manager and libvirt (given that it’s the default WTF did they hardcode “-p 22″ – they could have just skipped it!). Secondly there was some _really_ braindead hard-coding of “-l root” too! I mean, who runs boxes with remote root login enabled? [first bug http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=513605] raised (and fixed).

Now I’ve fixed the second part – which is that libvirt0 runs “nc -U /var/run/libvirt/libvirt-sock” on the remote end. Minor lart for using “nc” rather than “netcat” (which avoids a bit of namespace pollution). The main problem is that “-U” doesn’t work in “netcat-traditional” package, only “netcat-openbsd”. TBH since libvirtd has already created this socket, I’m not exactly sure what nc is doing (looks like it’s sending to the file as a Unix domain socket – I wonder how that differs from “cat”? I presume libvirt0 is using it bi-directionally. Ah, that’s _exactly_ the difference) I’ve only ever used netcat over TCP before. In any case, I’ve raised [http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=516482 another bug ] to cover this.