Luca Olivetti wrote:
I still couldn't find the cause of color bleeding: both the color manager (apart from the error fixed by the patch) and the spu encoder seems ok to me.
I found it and now it's gone :-) Apply the attached patch, additional to the previous one. Apart from the name change of a couple of variables, it only removes the bogus optimization that avoided calculating the index if the color hadn't changed (but since it might be in a different region/section it could well have a different index even if it's the same color).
Bye