cast_to_largest_integral_type() strips 64-bits values down to 32-bits

[jeremfg]

When compiling on Windows, we use MinGW32, which compiles a 32-bit executable.
As such, the type "size_t" is 32-bit wide even thought the largest type supported (uintmax_t) is 64-bit wide.

So whenever we call "will_return" or "assert_int_equal" and similar macros, we cannot pass a 64-bit value as it gets stripped down to 32-bit by the first cast to size_t. I've copied the culprit macro below, notice the double casting happening there.

// Perform an unsigned cast to uintmax_t.
#define cast_to_largest_integral_type(value) \
    ((uintmax_t)((size_t)(value)))

We've thus far circumvented the problem by creating a local "cmockery.h" file with the following content:

#include <setjmp.h>
#include <stdarg.h>
#include <stddef.h>
#include <stdint.h>
#include_next "cmockery.h"

/* Needed because original function "will_return" does a call to cast_to_largest_integral_type(),
 * which strips the upper 32-bits of a 64-bit integer. This happens because there is an intermediate
 * cast using size_t
 */
#define will_return64(function, value)          \
    _will_return(#function, __FILE__, __LINE__, \
                 value, 1)

/* Needed because original function "assert_int_equal" does some call to cast_to_largest_integral_type(),
 * which strips the upper 32-bits of a 64-bit integer. This happens because there is an intermediate
 * cast using size_t
 */
#define assert_int_equal64(a, b) \
    _assert_int_equal((uintmax_t) a, (uintmax_t) b, __FILE__, __LINE__)