mir/xrandr compatibility lib

[Bryce Harrington]

On Wed, Apr 10, 2013 at 11:12:32AM +1200, Robert Ancell wrote:
> On 09/04/13 07:20, Bryce Harrington wrote:
> > I'm looking into making a wrapper lib for mir that provides a few basic
> > xrandr operations via mir.  I think I understand what's needed but want
> > to get your sanity check before I get too far into it.
> >
> > This mir/randr compatibility library ("mirandr?") will provide the
> > header file X11/extensions/Xrandr.h, but stripped down to only the RANDR
> > 1.0 routines and structure definitions.  That should be adequate for
> > applications and is a reasonably achievable subset (i.e. no multi-head
> > layout operations, properties, or other advanced stuff).  Since
> > libXrandr is in C, this will need to be in C as well.
> >
> > The library will provide mir implementations of these commonly needed
> > RANDR 1.0 calls:
> >
> >   XRRGetScreenInfo()
> >   XRRSetScreenConfig()
> >   XRRQueryVersion()    // Just returns 1.0
> >   XRRQueryExtension()  // Stub; just return true
> >
> > The getter call (retrieving read-only information) is the immediate
> > priority.  The setter routine may be more involved to get right so
> > implementaton will be left to phase 2.
> >
> > There are also a number of accessors and constructor/destructor routines
> > for the various structs.  These are entirely local (they don't make any
> > server calls) so will be reused as-is.
> >
> >
> > The implementation of XRRGetScreenInfo() will essentially just set up a
> > mir connection, grab the display info, and reformat it into the Xrandr
> > structures.  So, borrowing logic from the demo client, it'd look
> > something like:
> >
> >     MirConnection* connection{nullptr};
> >     mir_wait_for(mir_connect(mir_test_socket, __PRETTY_FUNCTION__,
> >                  connection_callback, &connection));
> >     MirDisplayInfo info;
> >     mir_connection_get_display_info(connection, &info);
> >
> >     // Create XRRScreenConfiguration structure
> >     // Copy info.width, info.height to structure
> >
> >     mir_connection_release(connection);
> >     // return structure
> >
> >
> > How's this all sound so far?
> >
> > Bryce
> >
> This sounds good - this is a stepping stone before apps/toolkits become
> Mir aware right?

Right.

> Will this still require a recompile of the application using xrandr or
> is there some linking magic to make this a drop in replacement when
> running under Mir?

Static linking should be straightforward.  The package will provide a
.so version as well, which I'd expect to be swappable with libXrandr.so,
or whatever other linking magic people are doing.

The caveat is that this is only shooting for RANDR level 1.0 - that's
the same minimal RANDR support that -nvidia provided until just
recently.

Apps that need more than RANDR 1.0 probably would be early targets to
make Mir aware.  The RANDR 1.2+ operations are more advanced (e.g. CRTC
fiddling, properties, etc.) and may be too clunky to do within an
adapter.

Grepping XRR in gtk+3.0 it needs a few of the RANDR 1.2 functions, so
perhaps those can either be simulated or stubbed out.  Guess we can
investigate that more when things are further along.

Bryce