Re: [PATCH 12/12] use a shell function for AC_TYPE_INTx_T

[Eric Blake]

Paolo Bonzini <bonzini <at> gnu.org> writes:

> 
> 2008-10-27  Paolo Bonzini  <bonzini <at> gnu.org>
> 
>       * lib/autoconf/types.m4 (_AC_TYPE_INT_BODY, _AC_TYPE_UNSIGNED_INT_BODY):
>       New.
>       (_AC_TYPE_INT, _AC_TYPE_UNSIGNED_INT): Define and use a shell function.

Let's fix this up for another round of review, as well.

> +m4_define([_AC_TYPE_INT_BODY],
> +[  AS_LINENO_PUSH([$[]1])
> +  AC_CACHE_CHECK([for int$[]2_t], [$[]3],
> +    [AS_VAR_SET([$[]3], [no])
> +     for ac_type in 'int$[]2_t' 'int' 'long int' \

Should that be "int$[]2_t", rather than '', since we want shell expansion of $2 
to take place in creating $ac_type?

> +# _AC_TYPE_INT(NBITS)
> +# -------------------
> +AC_DEFUN([_AC_TYPE_INT],

A comment describing this m4 macro would be nice. (And several other macros 
touched by this patch).

> +[AC_REQUIRE_SHELL_FN([ac_func_c_find_intX_t],
> +  [AS_FUNCTION_DESCRIBE([ac_func_c_find_intX_t], [LINENO BITS],
> +    [Finds a signed integer type with width BITS, setting a cache variable
> +     accordingly.])],
> +    [$0_BODY])]dnl

Is this really C specific?  Can C++ or Objective C use it?  In other words, do 
we want _c_ in the macro name?

> +  AC_CACHE_CHECK([for uint$[]2_t], $[]3,
> +    [AS_VAR_SET([$[]3], [no])
> +     for ac_type in 'uint$[]2_t' 'unsigned int' 'unsigned long int' \

Same question on '' vs. "".

>      m4_bmatch([$1], [^\(8\|32\|64\)$],

Unrelated to the patch, but what happened to 16?

-- 
Eric Blake