Re: [glob2-devel] gradient improvement

[Simon Schuler]

> As you can see, this is more than unlikely anything to happen here.
> The "worst" gradient computation is GT_FORBIDDEN. But as all values
> are 255, it's fine. This reach big numbers because the forbidden
> gradiens starts with all places but the not-forbidden as sources,
> minus the place one can't go to, like stone, and water for
> non-swimming gradients.
> 
> So, first, I think you agree, but this really don't worth to worry
> about this. We should put a comment into the code anyway.
> 
> If you need ordered listedAddr[] for furthe optimisation, just don't
> process the GradientType==GT_UNDEFINED. Let the simple gradient
> computation handle it.
> 
> Is that ok with you Kai and Simon ?
> 
Yep, no problem for me... I was wrong to call it a bug without having
a closer look at it, possible bug would have been better. It really
seems unlikely to cause trouble as long as the users of the function
know the risk. But a little comment does really avoid confusion...