[sane-devel] Patches for SANE 1.0.9 be/fe compiled on IRIX

[Andrea Suatoni]

Henning Meier-Geinitz wrote:

> > Using the SGI X server in 24 bit (SGI boxes are big endian), the red and blue
> > channels are swapped without the change. You may enclose the changes I've
> > applied using #ifdef __sgi (this will be valid only on SGI machines, using
> > either GCC or the SGI compiler).
> 
> Ok, but I'd rather try to understand what's up here. Basically, you
> are chnaging the rules for the order of colors in memory with your
> patch. While it was rgb for big-endian and bgr for little-endian your
> patch swaps it to make it the other way round.

It's not always like the way you describe. In fact, my SGI O2 is currently
running the following visual (among the 29 supported):

screen #0:
  dimensions:    1280x1024 pixels (350x280 millimeters)
  resolution:    93x93 dots per inch
  depths (5):    1, 8, 12, 15, 24
  root window id:    0x40
  depth of root window:    24 planes
  number of colormaps:    minimum 1, maximum 11
  default colormap:    0x3d
  default number of colormap cells:    256
  preallocated pixels:    black 0, white 16777215
  options:    backing-store NO, save-unders NO
  largest cursor:    32x32
  current input event mask:    0x72003c
    ButtonPressMask          ButtonReleaseMask        EnterWindowMask          
    LeaveWindowMask          StructureNotifyMask      SubstructureRedirectMask 
    FocusChangeMask          PropertyChangeMask       
  number of visuals:    29
  default visual id:  0x32
<snip>
  visual:
    visual id:    0x32
    class:    TrueColor
    depth:    24 planes
    available colormap entries:    256 per subfield
    red, green, blue masks:    0xff, 0xff00, 0xff0000
    significant bits in color specification:    8 bits

Of course, you have also the following, but it's not the default one when
running the X server in TrueColor:

  visual:
    visual id:    0x33
    class:    TrueColor
    depth:    24 planes
    available colormap entries:    256 per subfield
    red, green, blue masks:    0xff0000, 0xff00, 0xff
    significant bits in color specification:    8 bits

Other SGI boxes may vary, depending on the XDM settings chosen by the sys
admin, and on the type of graphic subsystem installed.

On IRIX, all the visuals are usable at the same time. It may be possible,
eventually, to choose a visual (PseudoColor or TrueColor) with matches also
the RGB mask (instead of the default BGR mask). The old Maelstrom X11 porting,
for example, did exactly this.

> So it looks like (at least on your system) the color order is bgr
> despite it's big-endian. Could you test all the cases (24/4, 24/3, and
> 32)?
> 
> Does 32 exist at all?

No, it doesn't on mine. But other SGI boxes could have them (the Fuel, for
example, has also a 30 bitplanes visual, because it can have 10 bits of
resolution for color channel). I would suggest to not rely at all on the
default visual, but try to select the closest one among the availables. My fix
for xcam was, as said, just a quick hack.

Andrea