Re[2]: [Gnash-commit] gnash ChangeLog backend/render_handler_agg_style.h

[Udo Giacomozzi]

Wednesday, May 28, 2008, 3:31:43 PM, you wrote:
>> +    m_need_premultiply(false)

zl> Shouldn't the span color always need to be premultiplied?

In theory, yes, but...


>> +      if (trans_color.m_a < 255)
>> +        m_need_premultiply=true;

zl> This is an just optimization to avoid some unnecessary
zl> 'premultiplication' when alpha is 255, right?

Correct. For fully opaque gradients there is no need to do any
premultiplication. BTW, someone on the AGG list proposed a different
solution - trying that now...


zl> BTW, there is also an 'm_force_premultiply' in agg_style_bitmap, which
zl> seems to be always 'false'. Do we really need that, too?

Yes, it was used earlier and seems to be unnecessary. However I let
it there in case we come accross similar bugs again.

Udo