[Spice-devel] Further thoughts on the write polling in the xf86_spice_xql driver

[Alon Levy]

On Mon, Jun 04, 2012 at 08:42:17AM -0500, Jeremy White wrote:
> >> However, while the X server main loop has a facility that lets us add fds to it's
> >> select read mask, it has no such facility to let us add fds to it's *write* select mask.
> > 
> > Right. This is the problem. I think the right solution is to patch the X
> > server.
> 
> Okay; I unimaginatively decided that would be Hard.  Wouldn't that break
> the ABI, and therefore be rejected unilaterally?  I'm willing to try to
> craft a patch and submit it; but maybe I just need encouragement <grin>.

I started looking at this, talked to Dave Airlie and he suggested to ask
on xorg-devel either Adam Jackson or Keith Packard about one of two
suggested solutions I fielded, expand BlockHandler with a pReadmask or
add a third callback (in addition to BlockHandler, WakeupHandler)
BlockWriteHandler. The later won't break the internal Xorg / module ABI,
but is a little ugly IMO. The former breaks it.

> 
> Cheers,
> 
> Jeremy
> _______________________________________________
> Spice-devel mailing list
> Spice-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/spice-devel