Re: error: Support the compiler's control flow analysis better (was: copy-file: Silence gcc warnings)

[Pádraig Brady]

On 28/05/2023 13:50, Pádraig Brady wrote:
> On 28/05/2023 13:31, Pádraig Brady wrote:
> > On 27/05/2023 21:53, Bruno Haible wrote:
> > > +# ifndef _GL_NO_INLINE_ERROR
> > > +#  undef error
> > > +#  define error(status, ...) \
> > > +     ((rpl_error)(0, __VA_ARGS__), (status) ? exit (status) : (void)0)
> > > +# endif
> >
> > We might need to cast STATUS to bool to avoid the
> > following failure from coreutils CI:
> >
> > In file included from src/die.h:22,
> >                     from src/chroot.c:27:
> > src/chroot.c: In function 'main':
> > src/chroot.c:362:25: error: '?:'using integer constants in boolean context 
> > [-Werror=int-in-bool-context]
> >      362 |         error (warn ? 0 : EXIT_CANCELED, 0, "%s", (err));
> > ./lib/error.h:422:33: note: in definition of macro 'error'
> >      422 |      ((error)(0, __VA_ARGS__), (status) ? exit (status) : (void)0)
> >          |                                 ^~~~~~
>
> Actually casting with (bool), or using !! does NOT help here.
>
> It looks to be due to the nested use of ?: that's triggering the issue.
>
> $ gcc --version
> gcc (Debian 10.2.1-6) 10.2.1 20210110

To avoid this one can use `(status) != 0`.

There still is a gotcha (hit in dd.c in coreutils)
where if you define an error macro yourself
you get a macro redefinition error,
but that's something we can/should handle on the coreutils side anyway.

cheers,
Pádraig