[cairo] [RFC] Add device_extents api

[Jeff Muizelaar]

On Tue, Feb 03, 2009 at 05:13:55PM -0500, Behdad Esfahbod wrote:
> Jeff Muizelaar wrote:
> > On Tue, Feb 03, 2009 at 02:46:20PM -0500, Behdad Esfahbod wrote:
> >> Jeff Muizelaar wrote:
> >>> On Tue, Feb 03, 2009 at 12:25:48PM -0500, Behdad Esfahbod wrote:
> >>>> Jeff,
> >>>>
> >>>> The current implementation is broken indeed, and we plan to fix it. 
> >>> How do we plan on doing this?  I don't know if we can just change the
> >>> current implementation as that will likely break all the applications
> >>> that currently use it. For example, if, like moonlight, you switch to an
> >>> identity matrix before getting the extents and expect to get the device
> >>> extents we can't just start returning the user space extents because
> >>> they'll be too small.
> >> Huh?  If matrix is identity, the broken and ideal extents are the same.
> >> Either I'm missing something huge or you're talking about something else.
> >> This problem can be fixed and doesn't need new API.
> > 
> > Yep, when the matrix is identity they're the same, but what should you
> > get if the matrix isn't identity?
> 
> The tight bounding box, in user space.  Currently what we return isn't tight,
> and that's incorrect.

Yeah, I guess that would work. It's going to mean multiplying all of the
device space points back into userspace though.

-Jeff