RE: Problems when compiling Wine 20001202 with gcc -O6

[Patrik Stridvall]

> On Tue, 5 Dec 2000, Patrik Stridvall wrote:
> > So if optimization can't analyze what inline assembler does it
> > shouldn't do any optimizations that depends on what the assembler
> > code does.
> 
> Can you please provide a minimal example? (I missed the original mail,
> but will try to have a look and have it fixed, if it really is a bug.)

Look in wine/debugger/windbg.spec.c:

There is a row
        static void init(void) __attribute__((unused));
followed by a use of "init" in assembler code which
GNU C doesn't realize exists and therefore optimizes
the implementation away if -O6 is present.

> > That said IIRC the documentation for -O6 says it uses heuristics
> > which sort of means that generated code might not be correct
> > even if gcc is bug free. In other words don't use -O6.
> 
> Your reading of the documentation is not correct.
> 
> Plus, I'd really be interested to see where you found -O6 
> being described.
> GCC does not have such an option (but automatically reduces 
> it to -O3).

I said IIRC (If I Recall Correctly). That is I think
I have read it somewhere, persumably in the documentation.

If you say it isn't there I probably read it somewhere else
or simple didn't recall it correctly.