On Thu, Sep 20, 2007 at 08:06:22PM +0100, Richard W.M. Jones wrote:
Daniel P. Berrange wrote:
On Thu, Sep 20, 2007 at 01:10:28PM +0100, Richard W.M. Jones wrote:
Daniel P. Berrange wrote:
The use of '-o PasswordAuthentication=no' isn't sufficient because even with pulic key / gssapi auth schemes it may prompt for passwords and/or passphrases.
I think we need to use '-T -o BatchMode=yes', and regardless of the no_tty flag I think it could be wise to add '-e none'. Yes, that's right. Try this patch.
There's something not quite right going on. It correctly prevents the prompting of passwords, but the '?' seems to be passed through to the remote end. So the server side tries to open 'qemu:///system?' and fails.
Now technically '?' is a valid URI character and so qemu_driver.c should handle it, by parsing the URI properly, rather than STREQ(qemu:///system), but I think the client side should avoid appending the '?' too if there is no parameter to pass across.
The URI is reconstructed using xmlSaveUri from libxml2, and as far as I can see that always appends a '?' character.
The logic of remote_internal.c has got rather convoluted since qemu support was added so I can't see why this doesn't hit us in the current code, but it certainly hits us now. Only way I can see around this is to write proper URI parsing into qemu_driver.c ...
Ok, lets commit the no_tty patch then & i'll look at fixing the QEMU driver to parse things properly. 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 -=|