Re: [Qemu-devel] [PATCH 7/8] VNC Acceptance test: check protocol version

["Daniel P. Berrangé" <berrange@redhat.com>]

On Fri, Jun 07, 2019 at 02:12:07PM -0400, Cleber Rosa wrote:
> On Fri, Jun 07, 2019 at 06:29:15PM +0100, Daniel P. Berrangé wrote:
> > On Fri, Jun 07, 2019 at 11:22:22AM -0400, Cleber Rosa wrote:
> > > This goes a bit further than the other tests, and does a basic (read
> > > only) interaction with the VNC protocol.
> > > 
> > > This is not a enough to perform a handshake, but enough to make sure
> > > that the socket is somewhat operational and that the expected initial
> > > step of the handshake is performed by the server and that the version
> > > matches.
> > 
> > The GTK-VNC project provides a low level library libgvnc that can
> > be used to talk the RFB protocol in a fairly fine grained manner.
> > This is built using GObject, so is accessible from Python thanks
> > to GObject Introspection.
> >
> 
> Interesting.
> 
> > We could use libgvnc for exercising the VNC server instead of writing
> > custom RFB code. For your simple test just sending/receiving the
> > version it won't save much, but if we ever want to test TLS or
> > SASL integration, it would save alot of work dealing wth the auth
> > handshake and subsequent encryption needs.
> >
> 
> Absolutely.
> 
> > The main limitation it would have is that it would only work well
> > for sending "well formed" RFB protocol messages. If we ever want to
> > send intentionally evil/bad RFB data to check QEMU's VNC server
> > security hardening it would be harder.
> >
> 
> Right.  Still, there's a lot that can be done until we eventually
> exaust all possibilities and look into sending bad messages.
> 
> > As the maintainer of GTK-VNC though, I would be open to adding more
> > APIs to the low level gvnc library to facilitate QEMU's testing
> > needs if we want.
> >
> 
> I personally need to get acquainted with the currently available APIs
> first, but it looks like you alread have ideas for extensions that
> would come in handy.
> 
> Also, the one concern I have is how to deploy the library and Python
> bindings so that we can host those more advanced tests and still allow
> for a "make check-acceptance"-like experience.  What I mean is, I
> expect the Python bindings to be easily installed by pip, by I'd be
> (positively) surprised if the libgvnc would also have such an easy
> bootstrap.
> 
> Any ideas on this?

IMHO we shouldn't try to install anything from pip. It should all be
available from distro repos already. In fact if you have "virt-install"
or "virt-manager" installed, you should already have the required
python bits.

On RHEL/Fedora you need python3-gobject to get the GObject introspection
support and then all you need is the native gvnc library. There is no
python library for gvnc because GObject introspection makes this
redundant - the python API is dynamically created at runtime using
FFI to access the native library APIs.


Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|