Removed code that emits a warning on float16, if desired that should be a separate PR

The float16 warnings still seem to show up in the tests?

Codecov Report

Merging #11595 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #11595      +/-   ##
==========================================
+ Coverage    85.7%    85.7%   +<.01%     
==========================================
  Files         327      327              
  Lines       81985    81987       +2     
==========================================
+ Hits        70264    70266       +2     
  Misses      11721    11721

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 19522e2...b63687b. Read the comment docs.

What was the problem with float16?

float16 never emit a npy_set_floatstatus_invalid in np.maximum, np.minimum even if Nan appears, unlike other floats. In the original PR I added npy_set_floatstatus_invalid to the HALF_{maximum, minimum} loops, but this goes beyond the original issue and changes current behavior, so I removed it in a subsequent changeset

@mattip Sounds like float16 should raise a warning just to be consistent, what is the argument against that?

I thought the argument was: "lets not go there for a minor release"? then again, it might have been, lets put in the quick bug fix before the release happens at the time.

I definitely agree that we should enable the warnings for float16, but I would say that now that 15.0 is out, it is simplest to just wait for the next release for the float16 part (I assume the float32 is partially a regression here), it is not a huge deal and has been buggy forever.

See PR #11660 which extends this for 1.16 (no backport of that PR needed)

Thanks Matti.