Re: GCC 2.95 and floating point precision

[craig@jcb-sc.com]

>The author is working on fixing it, but this really isn't a great way to
>promote gcc2.95. I know all you people put an incredible amount of work into
>gcc2.95 and for most things it excels, but unless I'm much mistaken it could
>have been left in with a warning that it might generate bad code.
>-fno-strict-aliasing was for the purpose of backward compatibility and it is
>essential. Something of the same style for clobber would have really
>simplified issues.

Excuse me, but...*all* compilers might generate bad code.  If you know of
any counterexamples, please use those, and, let us know.

Also, I have not been watching this thread *real* closely, but, my
impression is, you have not yet identified any bug in gcc pertaining
to your code.

Therefore, it's quite likely the bug is in your code, and, as is so often
the case, you're complaining about gcc exposing it by optimizing it
sufficiently to expose the bug.

If Toon wants to try to debug your code for you, great, but, unless I've
missed where you've already done this, please read the FAQ item on how
to report bugs and track this bug down.  Then, you'll probably be able to
fix your code -- or, if you happen to find a compiler bug, you'll have
a more useful bug report to submit.

"My program doesn't behave as I expect with your new compiler, but it
behaved okay with the old" (or variations thereof) are almost always
statements made by people who have buggy programs, but haven't figured
that out yet.

(My guess is that you're hitting the truncating-spills problem.)

        tq vm, (burley)