Port-vax archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Thoughts about a floating point exception fix



Hi.

Sounds to me like the packages' problem.  Do their authors simply not
care about portability to non-IEEE floating point?

The problems are definitely with the packages. When I tried to talk with the python folks, for instance, they objected (well, a few of them) purely on the ideological ground that VAX is irrelevant, regardless of whether their assumptions are correct or whether embedded environments might have non-IEEE software floating point. The perl folks fixed things upstream.

What if we made sure all the soft-float stuff is in libm and compiled
certain tools such as awk and selected pkgsrc packages to use
software floating point?

Do you mean software VAX FP or software IEEE FP?   The latter sounds
more consistent with the rest of your mail, but also sounds completely
perverse for port-vax, not to mention, as you say, causing no end of
trouble if less than the whole system is built that way.

I was thinking of VAX FP, but with software routines which would check for data that would trigger exceptions.

I think this would be a good way to fix some of these longstanding
issues.

I disagree.  It wouldn't fix anything; it would merely paper over some
of the worst of the portability bugs in the software in question.

I used the word "fix" without qualifying - you're right that it's not actually fixing as much as it's providing better behavior. If, for example, we had a "fixed" awk which doesn't dump core, pkgsrc would work a lot better, even if awk were slower.

I think the software should be fixed.  If we (FWVO "we") don't have the
human cycles for that, I think the broken tests should be disabled when
building for VAX (or other non-IEEE FP) - or simply don't use the
(broken) software in question.

For commonly used software, definitely.

John


Home | Main Index | Thread Index | Old Index