[Prev][Next][Index][Thread]

Re: Floating Point Exceptions in Executor



>>>>> "Alexander" == Alexander Lehmann <alexlehm@rbg.informatik.th-darmstadt.de> writes:

    Alexander> Apparently, it depends on the CPU whether FPEs are
    Alexander> caught or not I just compiled a small test program on
    Alexander> m68k and hppa on hpux and on Solaris/sparc.  On the
    Alexander> motorola CPU, the program caused a core dump on the
    Alexander> other one it run returns a Inf value.  I have no idea
    Alexander> how this is handled on Macs, since the program I tested
    Alexander> apparently runs on real Macs, I would guess that it
    Alexander> just ignores the exception. If this is the case, then
    Alexander> Executor should do to.

SANE allows programs to set what happens on divide-by-zero.
Executor's SANE implementation doesn't properly handle signaling
exceptions, but it *does* work properly if the program requests that
division by zero not cause an exception and instead generate +/-Inf.

I would guess that the division by zero really isn't supposed to
happen (and doesn't happen when the program is run on a real Mac).

-Mat


References: