[PATCH 0/3 v2] xrandr: fix misc --gamma bugs

[Andy Ritger]

As I understand it: the gamma-correction tables configured through
XRR[GS]etCrtcGamma store 16:16:16 X Colors.  An X Color has some number
of significant bits, stored in the MSBs of the X Color (reported
in Visual::bits_per_rgb, though the gamma-correction tables are
visual-independent).  The significant bits of the X Color dictate the
size of the gamma-correction table (XRRGetCrtcGammaSize() == 2^sigbits).

PATCH 1/3 corrects xrandr to do gamma *correction* (i.e., 1/gamma)
PATCH 2/3 removes an assumption that XRRGetCrtcGammaSize() == 256.
PATCH 3/3 adjusts gamma-correction computation to produce results in
    the range [0,2^sigbits) << (16 - sigbits).

I found these links helpful:
http://www.poynton.com/PDFs/TIDV/Gamma.pdf (Gamma in computer graphics, page 17)
http://cgit.freedesktop.org/xorg/xserver/tree/hw/xfree86/common/xf86cmap.c:ComputeGamma()

If you compare against the code in xf86cmap.c, note that
CMapScreenPtr::gamma[] is stored in the range [0,2^sigbits), and is
* converted from [0,65536) in ComputeGammaRamp()
* converted to [0,65536) in xf86GetGammaRamp()

In the cover letter for v1 of this patch set, I asked these questions:

> * PATCH 1 changes long standing behavior in xrandr(1).  While the patch 
>   makes the semantics consistent with xgamma(1) and xorg.conf options, 
>   maybe it would be better to keep the existing behavior and just document
>   that behavior in the man page and --help?

If no one else has an opinion, I'm inclined to make the semantics
consistent with xgamma(1) and xorg.conf.

> * hw/xfree86/common/xf86cmap.c:xf86GetGammaRamp() makes up some values 
>   to pack in the low bits of each component of each entry.  Only the MSBs
>   should matter for hardware, but for consistency maybe xrandr(1) should 
>   do something similar to populate the low bits?

If no one else has an opinion, I'm inclined to not bother filling out
the low bits in xrandr(1).

Thanks,
- Andy


Changes since v1:
- in PATCH 2/3: add a comment and sanity check, clarifying why it is
  safe to assume that the size of the gamma ramp is <= 16-bits.

Andy Ritger (3):
  xrandr: use 1/gamma to compute gamma-correction
  xrandr: fix gamma == 1.0 && sigbits != 8
  xrandr: compute gamma-correction in [0,2^sigbits)

 xrandr.c |   63 +++++++++++++++++++++++++++++++++++++++++++++++--------------
 1 files changed, 48 insertions(+), 15 deletions(-)

-- 
1.7.2.5