Re: Review: Refactored baker code to allow for more advanced lut output

Jeremy Selan <jeremy...@...>

In case the nan/inf bug is holding you up - I've fixed the crash and
pushed it to a local branch:

I'm holding off though in pushing this to the main trunk pending a bit
more work.

I'm in the process of adding explicit unit tests to all image
processing operators, to guarantee that we have uniform nan/inf
handling through the entire library. I hope to get it all wrapped up
next week. (FYI - my intent is to preserve nan/inf value to the
greatest extend possible). I.e., if you run a 'nan' though a 1d lut,
it preserves the nan value. If you run a nan though a 3d lut (in any
input channel), the output will be nan in all 3 output color channels.
I believe this best preserves image integrity and obeys the design
principle of 'least surprise'.

-- Jeremy

Join to automatically receive all group messages.