[Lcms-user] Patches for floating point curve segments

[Paul M. <mi...@av...>]

I've been testing some floating point monitor profiles, and I think I have found a couple of bugs in the floating point sampled curve support in the latest lcms

cmsBuildTabulatedCurveFloat builds a curve which always evaluates to 0 for a value of 0.  I have changed it to use a constant segment which matches the first sample  for values <= 0, and also added a similar segment matching the last sample for values > 1.0 (the ICC spec says that the first and last segments in a segmented curve should be function segments).

EvalSegmentedFN gives the wrong result when evaluating a sampled segment where x1 - x0 is not 1.0

I have also added tests to testcms.c to verify the changes.

Suggested patches attached....

(apologies if this turns out to be a duplicate: I sent it to the list request address by mistake)

- Paul