On Mon, Feb 19, 2007 at 03:30:25PM +0000, Richard W.M. Jones wrote: > Just for everyone's information. > > http://annexia.org/tmp/libvirt-tls-20070219.patch Can you split out the internal virt/network driver API refactoring to a separate patch so we can apply & test it independantly of the main remote transport. In src/remote_internal.c: + case trans_unix: { + sockname = sockname ? : strdup (LIBVIRTD_UNIX_SOCKET); + + // 108 is hard-coded into the header files as well. +#define UNIX_PATH_MAX 108 + struct sockaddr_un addr; + memset (&addr, 0, sizeof addr); + addr.sun_family = AF_UNIX; + strncpy (addr.sun_path, sockname, UNIX_PATH_MAX); Should really use sizeof(addr.sun_path) - while 108 is hardcoded in the Linux headers, there's no guarentee Solaris hardcodes it to 108 too. In src/remote_internal.h: +//#define LIBVIRTD_UNIX_SOCKET "/var/run/libvirtd/socket" +#define LIBVIRTD_UNIX_SOCKET "/tmp/socket" // Just for testing +#define LIBVIRTD_CONFIGURATION_FILE "/etc/libvirtd.conf" Instead of using /var/run and /etc we should pass in the values of the sysconfdir & localstatedir variables from autoconf. That way, during testing & development people can run autogen.sh --prefix=$HOME/usr and these will automatically point to $HOME/usr/var/run/libvirtd and $HOME/usr/etc/libvirtd.conf Regards, Dan. -- |=- Red Hat, Engineering, Emerging Technologies, Boston. +1 978 392 2496 -=| |=- Perl modules: http://search.cpan.org/~danberr/ -=| |=- Projects: http://freshmeat.net/~danielpb/ -=| |=- GnuPG: 7D3B9505 F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 -=|