RE: [AD] 4.0.2 release

["Vincent Penquerc'h" <Vincent.Penquerch@xxxxxxxxxx>]

Title: RE: [AD] 4.0.2 release

> The problem is a mix of both: the range of the hue (which is
> a float) is
> 0->360, but we must tolerate -epsilon for 0, because of the
> precision errors
> introduced by the hue blender (likely due to -ffast-math with
> gcc).

I'm a bit stupid for answering something I don't know about,
sorry. I'd have simply capped the intermediate values and
continued. Anyway, how about switching to fixed ? This would
prevent any problems with --ffast-maths, wouldn't it ?
I'll stop speaking about stuff I don't know now ;)

--
Vincent Penquerc'h