Re: [Dri-devel] fix for logic blend/color ops (radeon / r200)

[Roland S. <rsc...@hi...>]

Ian Romanick wrote:
> Roland Scheidegger wrote:
> 
>> Here's a small patch which makes color / blend logic ops work with 
>> r200/radeon (untested on radeon). It's actually a copy&paste job from 
>> the mesa sources, mesa has the comment
>>    /* This is needed to support 1.1's RGB logic ops AND
>>     * 1.0's blending logicops.
>>     */
>> wherever it tries to figure out if logic ops are actually enabled. 
>> Looks like a trap in the OGL specification ;-), maybe other drivers 
>> might be affected?
>>
>> Patch fixes samples/blendeq and samples/blendxor (on my rv250), I hope 
>> it doesn't break anything.
> 
> 
> Good catch.
> 
> [snip]
> 
>> -   if ( ctx->Color.ColorLogicOpEnabled ) {
>> +   if ( ctx->Color.ColorLogicOpEnabled || (ctx->Color.BlendEnabled && 
>> mode == GL_LOGIC_OP)) {
> 
> 
> There's derrived state that Mesa calculates just for this case.  The 
> test should be 'if ( ctx->Color._LogicOpEnabled )'.

Yes, saw that derived state. For some reason I can't figure out now, I 
thought it might not be a good idea to use Mesa derived state directly. 
Ah well, this patch corrects this (even simpler, and should also save 1 
or 2 clock cycles ;-)).

Roland