On Wed, May 17, 2006 at 08:59:57PM -0700, Ben Pfaff wrote: > Jason Stover <[EMAIL PROTECTED]> writes: > > > On Wed, May 17, 2006 at 08:50:26PM -0700, Ben Pfaff wrote: > >> John Darrington <[EMAIL PROTECTED]> writes: > >> > >> > I'm not sure exactly what these functions do, but are you advocating > >> > we should ignore floating point errors? It's ok to turn off FPE's if > >> > we can ensure that the result of a floating point error will be NaN, > >> > but if not, then I would rather have PSPP crash, than to carry on and > >> > give erroneous output. > >> > >> Yes, the result should be a NaN, and PSPP should present that to > >> the user as a "system-missing" value. In my opinion, at least. > >> > >> That is the effect of disabling exceptions, or it should be. > > > > Using the -mieee did cause PSPP to produce a NaN in the output. > > Well, that's good news at least. > > Does NetBSD have feholdexcept() or __setfpucw()?
No. I'm not sure what it has for FPE's. man -k float doesn't show much. -Jason _______________________________________________ pspp-dev mailing list [email protected] http://lists.gnu.org/mailman/listinfo/pspp-dev
