scm_t_subr warnings

scm_t_subr warnings

[Julian Graham]

Hey Guilers,

Andy and Ludo and I were discussing this on IRC and it was suggested
that we move the question to the mailing list. I'm trying to compile
some code -- using `gcc -pedantic' -- that invokes `scm_c_make_gsubr',
and I'm getting the following warning:

  warning: ISO C forbids passing argument 5 of ‘scm_c_make_gsubr’
between function pointer and ‘void *’ [-pedantic]
  /usr/local/include/guile/2.0/libguile/gsubr.h:63:13: note: expected
‘scm_t_subr’ but argument is of type ‘struct scm_unused_struct *
(*)(struct scm_unused_struct *, struct scm_unused_struct *)’

I was confused, because I was sure that Guile defines scm_t_subr as
`SCM (*) ()', meaning that an `scm_t_subr' is of unspecified arity.
And I was right, but only at libguile compilation time. From __scm.h:

  #ifdef BUILDING_LIBGUILE
  typedef SCM (* scm_t_subr) ();
  #else
  typedef void *scm_t_subr;
  #endif

Thus the warning: ISO C lets you cast any kind of pointer to `void *'
-- except for a function pointer. Ludovic suggested that this bit of
preprocessor magic exists to support C++, in which the `()' style of
function prototyping is equivalent to `(void)'. But that leaves people
like me who want to be, well, pedantic, in a tough spot. Is there
anything we can do about this? One thing I was thinking was that we
could support the C++ case (and others) more explicitly. E.g.:

  #ifdef __cplusplus
  typedef void *scm_t_subr
  #else
  typedef SCM (* scm_t_subr) ();
  #endif

What do you think?


Regards,
Julian

Re: scm_t_subr warnings

[Ludovic Courtès]

Hello,

Julian Graham <joolean@gmail.com> skribis:

> Andy and Ludo and I were discussing this on IRC and it was suggested
> that we move the question to the mailing list. I'm trying to compile
> some code -- using `gcc -pedantic' -- that invokes `scm_c_make_gsubr',
> and I'm getting the following warning:
>
>   warning: ISO C forbids passing argument 5 of ‘scm_c_make_gsubr’
> between function pointer and ‘void *’ [-pedantic]
>   /usr/local/include/guile/2.0/libguile/gsubr.h:63:13: note: expected
> ‘scm_t_subr’ but argument is of type ‘struct scm_unused_struct *
> (*)(struct scm_unused_struct *, struct scm_unused_struct *)’
>
> I was confused, because I was sure that Guile defines scm_t_subr as
> `SCM (*) ()', meaning that an `scm_t_subr' is of unspecified arity.
> And I was right, but only at libguile compilation time. From __scm.h:
>
>   #ifdef BUILDING_LIBGUILE
>   typedef SCM (* scm_t_subr) ();
>   #else
>   typedef void *scm_t_subr;
>   #endif

The rationale is given above in gsubr.h:

  /* The type of subrs, i.e., Scheme procedures implemented in C.  Empty
     function declarators are used internally for pointers to functions of
     any arity.  However, these are equivalent to `(void)' in C++, are
     obsolescent as of C99, and trigger `strict-prototypes' GCC warnings
     (bug #23681).  */

(This refers to <http://savannah.gnu.org/bugs/?23681>.)

It is still an obsolescent feature as of C11.

Would it work for you to just cast the argument to scm_t_subr?  This is
what snarf.h (and guile-snarf) does.

> Thus the warning: ISO C lets you cast any kind of pointer to `void *'
> -- except for a function pointer.

[...]

>   #ifdef __cplusplus
>   typedef void *scm_t_subr
>   #else
>   typedef SCM (* scm_t_subr) ();
>   #endif
>
> What do you think?

I think it’s not just C++, but also future C standards and current
‘-Wstrict-prototypes’.

We can test for C++ or C > 11, but there’s no way do test for
-Wstrict-prototypes or -pedantic.  There’s #pragma GCC diagnostic, but
it cannot be used here.

So my current recommendation would be to add an explicit cast to
scm_t_subr, unless someone has a better idea.

Thanks,
Ludo’.