On Mon, 10 Nov 1997, Mike Long wrote:

.. Of course, all that's irrelevant if you're using the i387 assembly
.. version of libm; are you?

Note that I'm seing the problem on both NetBSD/i386 and on NetBSD/mac68k,
both built with the distributed Makefiles.

As I cannot see any significant changes to the libm sources except
for the new way of making RCS Id's (and the inclusion of cdefs.h),
my suspicion is also with the compiler...

To further increase the confusion, I get the following
on NetBSD/pmax:

./perl lib/complex.t | grep -v ^ok
1..1001
not ok 308
# 'cot(acot(z))' expected: '1.2' got: '1' for z = 1.2
not ok 975
# 'acoth $z0' expected: '0.14694666622553-0.2318238045004i' got:
'0.14694666622553-0.270209750135292i' for z = ( 2, 3)
not ok 978
# 'acoth $z0' expected: '0.14694666622553+0.2318238045004i' got:
'0.14694666622553+0.270209750135292i' for z = ( 2,-3)

 uname -a
NetBSD aarhus1.uni-c.dk 1.3_ALPHA NetBSD 1.3_ALPHA (AARHUS1) #31: Mon Nov
10 15:02:27 MET 1997
erik@aarhus1.uni-c.dk:/home/src/sys/arch/pmax/compile/AARHUS1 pmax


- Erik