RandR version 1.2 revisited

[Xavier Bestel]

Le mercredi 13 septembre 2006 à 10:43 -0700, Keith Packard a écrit :
> > - There should be a way for an app to specify it will sync to the vblank
> > of a particular CRTC (for uncomposited cases I guess).
> 
> Yes, that's what the sync extension is for if it ever gets hooked up.
> I'm not sure what RandR could do about this.

The problem is in selecting which vblank to sync to. AFAICS the only way
is for the app to look at RandR's map of the CRTCs, deduce which one
matches the most with its own window coordinates, and then ask to be
synchronized with this one. The trouble is in correlating Sync's and
RandR's Output id.

> > - There should be a user-selectable "Driving Output" which will set the
> > main subpixel order, main vblank, XVidMode target, etc. for applications
> > unaware of RandR1.2. That may solve point 10.1.
> 
> I'm not sure we need to make this selectable; it's purely a migration
> issue, and few (if any) applications actually listen to this part of
> RandR in any case. I suggest we just fix a simple policy (like sub-pixel
> order always follows the first output in the list driven by the CRTC,
> (which is user-selectable, btw)

That would have helped for the gamma setting: the "gamma setter" could
choose a particular output to drive, and then change gamma/whatever. No
need to touch the one-head-only parts of the protocol.
I admit it's a bit circonvoluted.

	Xav