Is fastcall broken?

[Angus]

Andrew Haley-5 wrote:
> 
> Angus writes:
>  > 
>  > I have gcc 4.0.2 and 4.1.2 and in the one particular case if I use the
>  > fastcall attribute valgrind reports a bunch of errors, and I get a
>  > segmentation fault, for which fastcall is key. But I believe it is more
> than
>  > just key, I believe the fastcall attribute to be the cause, since when
> I run
>  > valgrind on my app w/out fastcall, or in a 64-bit environment, valgrind
>  > doesn't report anything wrong.
>  >    I'm using the fastcall attribute in 3 places. One is this
> stand-alone
>  > function, and the other two are pure virtual methods of a class. Oddly
>  > enough, the pure virtuals aren't giving me any trouble at all (since
> I'm
>  > very caseful to declare the overrides with like attributes). In all
> cases
>  > the full attributes declaration is: __attribute__((const, nothrow,
>  > fastcall)).
>  >    Could I be using fastcall wrong, or is it just a dangerous attribute
> to
>  > work with?
> 
> I wouldn't say it was dangerous, exactly.  I suspect there is a
> problem soemewhere with your declarations, but until you give us a
> test case we won't know.
> 

    I figured out what the problem was. I forgot to include the header file
with the definition, so obviously there was no way for the function to know
that it was supposed to use the fastcall convention, but all the callers
did. It's very unlike me to omit the inclusion of the header, even though
for functions it is almost never necessary. But the one time it was
necessary, I didn't do it. Thanks.

   BTW, in my opinion it is dangerous. Usually one can rely on compile or
link errors to catch mismatched function characteristics, but with
attributes there is no such checking. So even if you aren't doing something
*really* dangerous, like working with virtual methods, you might do what I
did, and you'll never know about it until you notice you've mismatched your
attributes. So if you ask me, attributes like this one should be used
sparingly, and with much caution.
-- 
View this message in context: http://www.nabble.com/Is-fastcall-broken--tf3995991.html#a11359697
Sent from the gcc - Help mailing list archive at Nabble.com.