Re: [PATCH] MIPS: Transform old-style macros to newer "__noreturn" standard.

["Satyam Sharma" <satyam.sharma@gmail.com>]

Hi Peter,

On 5/26/07, H. Peter Anvin <hpa@zytor.com> wrote:
> Robert P. J. Day wrote:
> >> ... and declare functions as:
> >>
> >> __noreturn f();
> >>
> >> ... which is the syntactially sane way of doing it.
> >
> > that may be, but keep in mind that gcc allows attributes to *follow*
> > the parameter list as well, and some people might prefer to do the
> > following:
> >
> >   f() __noreturn;
> >
> > that would fail badly if you defined __noreturn as you suggest.
>
> That's equally moronic that saying that "some people might prefer to
> write 'f() void;'", which is what it's *EXACTLY* equivalent to.  Yes,
> they might "prefer" it, but it's syntactically invalid and the compiler
> won't accept it.  As it shouldn't.
>
> __noreturn here takes the syntactic place of the return type, because
> that's what it IS.

But __attribute__((noreturn)) is simply a _function attribute_. Of course,
it is legal / valid only for functions with return-type void, so it does make
sense to combine both void and __attribute__((noreturn)) in the same
macro like you say. But that's not syntactically necessary. In fact,
grepping through the sources, a lot of people do prefer to place the
attribute _after_ the function declarator.

Anyway, I'm fine either way.

Thanks,
Satyam