Date: Sat, 12 Dec 92 15:30:43 GMT
From: kuku AT acds DOT physik DOT rwth-aachen DOT de (Christoph Kukulies)
To: djgpp AT sun DOT soe DOT clarkson DOT edu, marwan AT sedal DOT su DOT oz DOT au
Subject: Re:  tanh in libm is buggy?

I'm presently having problems with a ported f77 (f2c) program which 
makes excessive use of trigonometric functions (asin etc.) and 
the guy who wrote it under VMS does not have problems and blames
me for the problems that hist program makes under djgpp libm.
I mean, not severely blame but does not take any effort in looking
into hist code because 'it runs under VMS' 8-O

I'm trying to track it down. I have built a fence around asin
to prevent it from being passed arguments >1.0 smaller -1.0 and now the
code works safer but crashes in a different place. Unfortunately
it behaves like a HeisenBug, it dissapear when you wnat to watch it
with the debugger.

I hope to get it solved soon. These messages about defunct tanh are 
alarming me.

--Chris