RE: [PATCH] OMAP3630: DSS2: Enable Pre-Multiplied Alpha Support

[Tomi Valkeinen <tomi.valkeinen@nokia.com>]

On Wed, 2010-11-03 at 08:57 +0100, ext Taneja, Archit wrote:
> Hi,
> 
> linux-omap-owner@vger.kernel.org wrote:
> > Alpha Support
> > 
> 
> [snip]
> 
> >> 
> >> +static void _dispc_set_pre_mult_alpha(enum omap_plane plane, bool +enable) {
> >> +	if (!dss_has_feature(FEAT_PREMUL_ALPHA))
> >> +		return;
> >> +
> >> +	BUG_ON(!dss_has_feature(FEAT_GLOBAL_ALPHA_VID1) &&
> >> +		plane == OMAP_DSS_VIDEO1);
> > 
> > What is the rationale for having the function return, if
> > FEAT_PREMUL_ALPHA is not supported, but BUG if plane is video1 and
> > GLOBAL_ALPHA_VID1 is not supported?
> 
> Premultiplied alpha is available on omap36xx and above, but on 36xx
> since global alpha itself isn't supported for video1 then writing to
> the pre multiplied alpha bit is incorrect.
> 
> It could have been possible to make a new feature like FEAT_PRE_MULT_VID1
> but I think its redundant as FEAT_GLOBAL_ALPHA_VID1 is enough to determine
> if we should set the pre multiplied alpha bit for video1 plane or not.

I was referring to the first check using return, and the second using
BUG(). If nobody is supposed to call that function when

!dss_has_feature(FEAT_GLOBAL_ALPHA_VID1) && plane == OMAP_DSS_VIDEO1)

then why is it ok to call that function when 

!dss_has_feature(FEAT_PREMUL_ALPHA)

Shouldn't they both be either returns or BUGs?

  Tomi