>>>>> "Barton" == Barton Willis <willisb@unk.edu> writes:
Barton> (1) There is a bug in sign
Barton> (%i1) sign((sqrt(3)-2)*x);
Barton> (%o1) NEG
Barton> This should evaluate to PNZ.
Yep. Haven't tracked it down yet, but it's easily confused when an
arg is negative.
I think your hack of (not (mminusp* (neg y))) is a decent workaround
until the sign bug is fixed.
It doesn't help that the sign routines use lots of global speical
variables to return results so tracing doesn't include everything you
need.
Ray