set_mode_major vs RandR transforms

[Keith Packard]

On Tue, 2009-06-09 at 16:13 +0200, Matthias Hopf wrote:
> On Jun 08, 09 22:27:03 -0700, Keith Packard wrote:
> >  1) set_mode_major no longer sees the current state
> >     of the hardware. That was already partially true -- the
> 
> The driver can track the state itself if it needs to.
> I don't see any issues here.

Yes, I agree, although it seems sub-optimal in some ways.

> This combination bothers me a little. We don't use set_mode_major in
> radeonhd yet, so there it is a no-brainer, but how do you implement a
> driver that wants to support both old style (1.6) and new style (1.7)?

You continue to set the mode/x/y/rotation values in the crtc and do not
set the transform or transformPresent fields. Old X servers would never
have set transformPresent or the transform, so you will not get
transforms with that version. New X servers will set it correctly before
calling your function.

> You have to be able to determine whether you should set transformPresent
> to FALSE or not - best on runtime, if the ABI doesn't change.

Never setting it is correct -- it isn't modified otherwise in the
existing code.

Thanks for making me check this again; I was uneasy about this case as
well, hence the long message. Given what I've found, it looks like we
can get as correct a behaviour as possible with the above plan.

-- 
keith.packard at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.x.org/archives/xorg-devel/attachments/20090609/24419cf8/attachment.pgp