Subject: Re: +Infinity D_Float on VAX?
To: Anders Magnusson , Brian D Chase <bdc@world.std.com>
From: I. Souvatzis <ignatios@cs.uni-bonn.de>
List: port-vax
Date: 03/29/1999 13:32:13
On Mon, Mar 29, 1999 at 01:29:33PM +0200, Anders Magnusson wrote:
> > Finally got the problems resolved with Povray running under NetBSD/vax
> > over the weekend.  A friend of mine trudged through the povray code and
> > found the problem to be in a test against the HUGE_VAL define from
> > <math.h> which gets its value from the __infinity symbol in libc.a or more
> > specifically from /usr/src/lib/libc/arch/vax/gen/infinity.c
> > 
> > I've checked all the other architectures listed under
> > /usr/src/lib/libc/arch, and it looks like just about all of them use
> > 
> >   const char __infinity[] = { 0, 0, 0, 0, 0, 0, (char)0xf0, 0x7f };
> > 
> This is completely wrong on VAX (as also stated in the comment in
> the C source file). The VAX FP format does not have infinity; it uses
> floating point traps instead. I don't remember exactly how/what this
> should be instead, someone that like floating point specifications
> should fix this :-)

well, I guess setting __infinity to the value HUGE_VAL should have is ok
on the VAX.	

	-is