[cairo] [PATCH]: pixman: New functions for converting format to/from masks

[Soeren Sandmann]

Carl Worth <cworth at redhat.com> writes:

> Here is a small (two commit) patch series that adds this little bit of
> new API to pixman:
> 
> 	typedef struct _pixman_format_masks {
> 	    int bpp;
> 	    unsigned long alpha_mask;
> 	    unsigned long red_mask;
> 	    unsigned long green_mask;
> 	    unsigned long blue_mask;
> 	} pixman_format_masks_t;
> 
> 	/* Format conversion */
> 	PIXMAN_EXPORT
> 	pixman_bool_t   pixman_format_code_to_masks (pixman_format_code_t code,
> 	                                             pixman_format_masks_t *masks);
> 
> 	PIXMAN_EXPORT
> 	pixman_bool_t   pixman_format_masks_to_code (pixman_format_masks_t *masks,
> 	                                             pixman_format_code_t *code_ret);
> 
> See the patch itself for complete documentation of these functions.
> 
> This functionality currently lives within cairo itself, but really
> needs to live in pixman. As can be seen in the patch series, the cairo
> versions are already very incomplete with respect to what formats
> pixman provides.

The main issue I see with this is that a pixman format can't really be
fully described by that struct. For example formats with 16 bits per
channel or floating point formats will not be describable.

What does describe a pixman format is the format code itself,
though. 

The functions above do not really produce anything information that a
pixman user doesn't already have. Because of this, and because the
main use of these functions would be to convert to and from XImage
formats, I don't see it as all that bad if they live in cairo.

As far as I can tell, the pixman_format_code_to_masks() in your patch
is already as general as the pixman version could be, given that only
32 bit or less fixpoint formats are possible for XImage.

The _from_masks() function could be written with
PIXMAN_FORMAT(t,bpp,a,r,g,b) macro to construct a format by
itself.

The issue with that is that there is no guarantee that the format
would be supported by pixman, so we would need a couple of new
functions:

        pixman_format_supported_destination ()
        pixman_format_supported_source()

Both functions would simply return TRUE for all supported formats,
except that the destination one would return FALSE for YUV formats.

(Also, I'd prefer that the code goes into pixman-utils.c instead of
the new pixman-format.c).


Soren