i386-*bsd, was Re: FreeBSD 3.x

[Marc Espie <espie at quatramaran dot ens dot fr>]

In article <16954.928489480@upchuck.cygnus.com> you write:
>
>I need someone with an x86 background that uses freebsd-3.x to debug an
>interesting problem.

Give me the exact failure mode you get on FreeBSD.
The  same test does fail on OpenBSD.  I think it might be for the same
reasons.

>I've got a few tests (ex ieee/rbug.c) which are failing on a freebsd-3.0
>system I have access to.
>
>I found this to be a little odd since the other x86 ports do not seem to
>have this problem.

You sure ? 

>What's even more interesting is I can take a static binary that fails on
>the freebsd box and run it without any problems on my linux box.

>There's definitely something interesting happening with this test :-)  I
>want/need to know is whether or not there's a compiler bug causing the
>problem (of course).  But I don't have a good enough background on the
>x86 to debug it -- especially since it uses floating point and you can't
>access the FP registers with the debugger.


could it be that:
/* This goes away when the math-emulator is fixed */
#undef TARGET_DEFAULT
#define TARGET_DEFAULT \
  (MASK_80387 | MASK_IEEE_FP | MASK_FLOAT_RETURNS | MASK_NO_FANCY_MATH_387)

might be at fault ?

This is about the only thing the three BSD share that might explain this 
failing on our platforms.

It could also be that the same test also fails on netbsd, but since the
netbsd crowd doesn't seem to believe in sending test results, I have no
way to check...