> > In short I refer to algoritms that tries to be _too_ "clever" as
> > heuristics while some uses it for normal "clever" algoritms as well.
>
> I think that's the point. Your definition is rather non-standard.
OK, well, it isn't really import for the issue at hand.
> > That said I remember reading somewhere that there are _some_
> > optimizations that GNU C employs at higher (like -O6)
>
> There's nothing like -O6 in GCC. -O3 is the maximum.
Perhaps not, but it was not I that wrote the topic, that as you
see includes -O6.
> > optimization level that tries to be _too_ clever and thereby
> > causing it to generate wrong code for some strange cases.
>
> I'm not aware of any such optimization that breaks standards-compliant
> code. That would be a bug. There are, however, optimizations
> that break
> non-standards-compliant code, where the user tried to be too clever.
Sure who is at fault is determined primarily determinte by the standard.
However what I and Ove was discussioning was
1. Whether the standard really is supposed to be interpreted that way
2. Whether the standard makes sense.
> If you still think that GCC has a bug, please report this!
> Or, probably
> better, ask at [EMAIL PROTECTED] whether the behavior you are seeing is
> intended or really a bug. If you think the documentation is not clear
> enough, a patch would be welcome as well!
At the very least the documentation should be clarified, since using
__attribute__((unused)) hides a warning that might cause a real nasty
suprise for the user that adds it because he interpreted the standard
the way I do, which is also IMHO the sensible interpretation of
__attribute__((unused)) without looking at the documentation.
