[Bug c/118005] Diagnostic for mixing noipa and inline mentions nonline rather noipa

[alx at kernel dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118005

--- Comment #2 from Alejandro Colomar <alx at kernel dot org> ---
(In reply to Andrew Pinski from comment #1)
> So noipa implies noinline.

Hmmm, I thought it could be considered something like
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117853>, where you're showing an
internal transformation.  Although in this case it doesn't hurt, since users of
noipa actually know that it implies noinline, so I guess it's okay-ish.

> The diagnostic is correct and is wanted.
> 
> On the question of using -Wattribute or it's own flag, I think we don't want
> to do that. Since the fix is to remove inline keywords.

Do you mean that I should never write a function is both inline and noipa?
Please allow me to disagree.  My actual code is something like the following:


// <foo.h>:
#if (FOO_INTERNAL)
#define foo_inline  inline
#else
#define foo_inline  inline [[gnu::noipa]]
#endif

foo_inline int foo(void);

inline int
foo(void)
{
    return 42;
}


// foo.c
#define FOO_INTERNAL
#include "foo.h"

extern inline int foo(void);