On Sat, 16 Jul 2011 17:51:50 -0400, Jonathan M Davis <jmdavisp...@gmx.com>
wrote:
I think that the problem is a combination of NaN and real. If I change
the
first three tests (which all test using init) to use is, then all of the
tests
pass until real is tested with T.max, and it's doesn't survive being
swapped
twice). So, maybe something special needs to be done for NaN, and the
issue
with real may involve issues with padding. I don't know.
float is float changed for 2.054. == always has made two nans not be
compared equal. 'is' was changed, but I think the change is not good
enough for your unit tests.
See bug http://d.puremagic.com/issues/show_bug.cgi?id=3632
-Steve
