[vorbis] Serious GCC/EGCS 2.91 bug found (bites rc2)

[Monty]

> In summary, all the math in GCC seems not to be of high quality.
> I have had many troubles compiling some kind of code (runs good
> using Watcom C for DOS/Win32/OS2) for Linux. I have tested
> three different Linux versions resp. compilers:
> -Red Hat 6.1
> -SuSE 7.0
> -SuSE 7.2 shipped with gcc version 2.95.3!!!
> All they have corrupted my code, even with optimization switched off!

Unless you have identified a specific compiler bug, this does not
indict the compiler.  I've worked in user support, and if I had a dime
for every time someone stormed in complaining that GCC was broken (Sun
CC compiles it just fine! It can't be my code!) and it turned out to
be user error, I'd be a rich man.  I got a few free dinners off of the
obvious bet until my co-volunteers asked me to stop demanding a free
dinner if I was right :-)

The general quality of GCC is roughly that of the compilers around it,
and this is an amazing feat considering that it compiles for virtually
every architecture known to man. 

EGCS was a reimplementation of GCC (the second such rebuild) in order
to make a cleaner optimization framework.  New software has bugs.  I
haven't found any in EGCS for quite a while (2.91 was an early version
and technically was never officially released even though RedHat
decided to ship it).  At this point, EGCS (new gcc) has surpassed the
stability and optimization capability of GCC 2, but with a clean and
elegant optimization framework that will go places.

> Then I kindly suggest perhaps the problem i
> The atanh function seems to be buggy. It gives correct results in
> a stand-alone environment. But it behaves in some kind illegal,
> it corruptes the whole program. By masking atanh by
> #undef atanh

!!! This function is not implemented by GCC!  It is handled by GNU
libc, and *that* has had numerous and recent bug reports filed against
its math behavior.  When making such a strong claim, please take the
minimum effort to identify the real problem!  Even now, we can't tell
if the real problem is GNU libc, but I can say for certain it's not
GCC.

As for mysteriously 'corrupting the program', how about we bet some
beer on it...?

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.