Re: aptitude on alpha (again, sigh)
On Thu, 1 Nov 2001, Daniel Burrows wrote:
> Why does this fix the problem? (ie, is this just working around a bug
> in the code?) The documentation says it has something to do with
> floating-point comparisons being "inexact". Specifically, it mentions
> handling NaN and +-Inf correctly. This seems to be indicating that I
> have a divide-by-zero somewhere!
Underflows and such are also in that category. In this case, I'm not sure
what the offending piece or error really is. Perhaps I'll toss gdb at it
tonight. Usually, such FPEs can be worked around with careful coding,
especially in cases like aptitude where computing NaNs isn't an intended
risk of the software (if this was a math package, I could understand, but
it's not) :-P
C
Reply to: