On Mon, Jun 20, 2011 at 10:34:42AM -0700, Jason Gerecke wrote:
> On Sun, Jun 19, 2011 at 9:01 PM, Peter Hutterer
> <peter.hutte...@who-t.net> wrote:
> > The NVIDIA binary driver doesn't support RandR 1.2. Test for the NV-CONTROL
> > extension and assume the binary driver is running when this extension is
> > detected. Then use Xinerama to query the screen offset and set the
> > transformation matrix accordingly.
> >
> > User commands:
> > xsetwacom set "device name" MapToOutput HEAD-0
> > xsetwacom set "device name" MapToOutput HEAD-1
> >
> > Signed-off-by: Peter Hutterer <peter.hutte...@who-t.net>
> > ---
> > configure.ac | 2 +-
> > man/xsetwacom.man | 3 +-
> > tools/xsetwacom.c | 64
> > ++++++++++++++++++++++++++++++++++++++++++++++++++++-
> > 3 files changed, 66 insertions(+), 3 deletions(-)
> >
> > diff --git a/configure.ac b/configure.ac
> > index 408e75f..c151e54 100644
> > --- a/configure.ac
> > +++ b/configure.ac
> > @@ -53,7 +53,7 @@ AC_CHECK_LIB([m], [rint])
> > PKG_CHECK_MODULES(XORG, [xorg-server >= 1.7.0] xproto xext kbproto
> > inputproto randrproto)
> >
> > # Obtain compiler/linker options for the xsetwacom tool
> > -PKG_CHECK_MODULES(X11, x11 xi xrandr)
> > +PKG_CHECK_MODULES(X11, x11 xi xrandr xinerama)
> >
> > # Obtain compiler/linker options for libudev used by ISDV4 code
> > PKG_CHECK_MODULES(UDEV, libudev)
> > diff --git a/man/xsetwacom.man b/man/xsetwacom.man
> > index 3a5fa40..564e1d0 100644
> > --- a/man/xsetwacom.man
> > +++ b/man/xsetwacom.man
> > @@ -126,7 +126,8 @@ outputs may be obtained with the xrandr tool. The
> > output mapping
> > configuration is a onetime setting and does not track output
> > reconfigurations; the command needs to be re-run whenever the output
> > configuration changes. When used with tablet rotation, the tablet must be
> > -rotated before it is mapped to the new screen.
> > +rotated before it is mapped to the new screen. When running the NVIDIA
> > +binary driver, the output names are "HEAD-0" and "HEAD-1".
> > This parameter is write-only and cannot be queried.
>
> See comment below on expanding syntax to "HEAD-N" (where N can be any
> digit from 0 to 9). If you remove the NV-CONTROL restriction, you may
> also want to mention the syntax works for any driver exposing info
> through Xinerama.
>
> > .TP
> > \fBMode\fR Absolute|Relative
> > diff --git a/tools/xsetwacom.c b/tools/xsetwacom.c
> > index 1a313df..94c9db2 100644
> > --- a/tools/xsetwacom.c
> > +++ b/tools/xsetwacom.c
> > @@ -37,6 +37,7 @@
> > #include <X11/Xatom.h>
> > #include <X11/extensions/XInput.h>
> > #include <X11/extensions/Xrandr.h>
> > +#include <X11/extensions/Xinerama.h>
> > #include <X11/XKBlib.h>
> >
> > #define TRACE(...) \
> > @@ -2058,8 +2059,64 @@ static void set_output_xrandr(Display *dpy, XDevice
> > *dev, param_t *param, int ar
> > XRRFreeScreenResources(res);
> > }
> >
> > +/**
> > + * Adjust the transformation matrix based on the twinview settings. This
> > + * would better be done with libXNVCtrl but until they learn to package it
> > + * properly, rely on Xinerama. Besides, libXNVCtrl isn't available on RHEL,
> > + * so we'd have to do it through Xinerama there anyway.
> > + */
> > +static void set_output_twinview(Display *dpy, XDevice *dev, param_t
> > *param, int argc, char **argv)
>
> I'd prefer if this were named e.g. 'set_output_xinerama' since there
> is nothing inherently TwinView-related in it.
point taken, will amend
> > +{
> > + char *output_name;
> > + int event, error;
> > + XineramaScreenInfo *screens;
> > + int nscreens;
> > + int head;
> > +
> > + if (!XineramaQueryExtension(dpy, &event, &error))
> > + {
> > + fprintf(stderr, "Unable to set screen mapping. Xinerama
> > extension not found\n");
> > + return;
> > + }
> > +
> > + output_name = argv[0];
> > + if (strncmp(output_name, "HEAD-", 5) != 0)
> > + {
> > + fprintf(stderr, "Please specify the output name as HEAD-0
> > or HEAD-1\n");
> > + return;
> > + }
> > +
> > + head = output_name[5] - '0';
> > + if (head != 0 && head != 1)
> > + {
> > + fprintf(stderr, "Please specify the output name as HEAD-0
> > or HEAD-1\n");
> > + return;
> > + }
> > +
> > + screens = XineramaQueryScreens(dpy, &nscreens);
> > +
> > + if (nscreens == 0)
> > + {
> > + fprintf(stderr, "Xinerama failed to query screens.\n");
> > + goto out;
> > + } else if (nscreens != 2)
> > + {
> > + fprintf(stderr, "Found %d screens, but this command
> > requires exactly 2.\n", nscreens);
> > + goto out;
> > + }
>
> My code uses convert_value_from_user to allow an arbitrary (positive)
> integer as a head number. That's definitely overkill, but supporting
> up to 10 heads (still probably overkill) would require only a trivial
> change to this code and work with the rarer tri- and quad-head setups.
it'd be simple enough to do a head > nscreens here anyway, so I can easily
amend this, pluse the convert_value_from_user.
> > +
> > + _set_matrix(dpy, dev,
> > + screens[head].x_org, screens[head].y_org,
> > + screens[head].width, screens[head].height);
> > +
> > +out:
> > + XFree(screens);
> > +}
> > +
> > static void set_output(Display *dpy, XDevice *dev, param_t *param, int
> > argc, char **argv)
> > {
> > + int opcode, event, error;
> > +
> > if (argc != param->arg_count)
> > {
> > fprintf(stderr, "'%s' requires exactly %d value(s).\n",
> > param->name,
> > @@ -2067,7 +2124,12 @@ static void set_output(Display *dpy, XDevice *dev,
> > param_t *param, int argc, cha
> > return;
> > }
> >
> > - set_output_xrandr(dpy, dev, param, argc, argv);
> > + if (XQueryExtension(dpy, "NV-CONTROL", &opcode, &event, &error))
> > + {
> > + TRACE("Nvidia binary driver detected.\n");
> > + set_output_twinview(dpy, dev, param, argc, argv);
> > + } else
> > + set_output_xrandr(dpy, dev, param, argc, argv);
> > }
>
> Why make the Xinerama codepath dependent on the NV-CONTROL extension?
> Its a reasonable fallback in a number of circumstances beyond the user
> being stuck with the binary nVidia driver. Why not try RandR first,
> and if it fails (version too old, output not found, etc) give Xinerama
> a shot?
do we have any meaningful drivers that don't support RandR 1.2 but support
Xinerama? except the nvidia binary driver of course?
either way, simple enough change. I'll send out new patches tomorrow, I
can't test them at all today.
thanks.
Cheers,
Peter
------------------------------------------------------------------------------
EditLive Enterprise is the world's most technically advanced content
authoring tool. Experience the power of Track Changes, Inline Image
Editing and ensure content is compliant with Accessibility Checking.
http://p.sf.net/sfu/ephox-dev2dev
_______________________________________________
Linuxwacom-devel mailing list
Linuxwacom-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxwacom-devel
