Re: gawk 3.1.2g problems with NaNs and substr etc.

[Paul Eggert]

Aharon Robbins <address@hidden> writes:

> I resent your labelling this a "bug".

Sorry.  The word wasn't meant to be pejorative.  It just popped out of
my keyboard, I guess because I was sending email to bug-gnu-utils.

I've been finding 64-bit shortfalls in software for many years and I
agree with you that there are a lot more where these came from.  At
first, this was more of a hobby, but as time goes on these problems
will gradually become more important.  I think GNU "sort", for
example, is now regularly used with buffers larger than 4 GiB on some
platforms, and it wouldn't tolerate problems like this in its
buffer-handling code.

> I would be really surprised to learn of *any* contemporary program that
> doesn't have similar design shortfalls.

Well like I said I make a bit of a hobby of it.  If anyone can find
such shortfalls in GNU "sort", diffutils, tar, Bison, patch, etc.,
please let me know.


PS.  In 1995 I wrote to bug-emacs that a new Emacs feature was "dead
on arrival" on my platform due to some configuration problem or
another.  It took RMS a couple of days to forgive me for using the
phrase.  What can I say?  No insult was intended there, either: I was
just reporting a problem.  You can see my original bug report, if you
like, and see whether you think I was out of line there, too:
<http://groups.google.com/groups?selm=9506232051.AA00394%40light.twinsun.com>