> The opther point is that you can't trust your compiler using its maximum
> optimization level. It sometimes produces weired results, and you always
> have to check the accuracy when, as an example, you're using egcs with -O9.
Horsecocky. If EGCS is producing bogus code, that's a bug in your C, in EGCS
or in your headers. I'd wager in the C. In the past nearly six years of
developing Ogg, I've found one gcc bug, no egcs bugs, two Linux libc header
bugs and one libc bug. Each one was exactly that-- A bug, not evil voodoo.
There are plenty of ways to write bad C that will seem OK in -O0, and then
break under optimization because it violates ANSI.
I'm probably arguing semantics (I suspect we actually agree), but I wanted to
pipe up and make the point clear.
Monty
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
