stdint.h与C ++不兼容？

Question

I tried to compile a source code package, and found version-specific issues.

When I do that on my computer, everything goes well; but when I compile it on another computer, it produces a lot of claims that INT32_MAX is not defined. Both computer runs a Debian system, and the difference is my computer uses Testing repo and has gcc 4.9, and the other computer uses Stable repo., which has slightly older gcc 4.7.

Then I have a detailed look inside /usr/include/stdint.h . By surprise, on the computer claiming the undefined macros, all int range macros are defined inside a non-C++ condition:

/**
 * I forget the exact definition,
 * but it looks like this:
 */
#ifndef __cplusplus ......
......
# define INT32_MIN XXX
# define INT32_MAX XXX
......
#endif

As a result, the package won't see these standard range macros, as it is a C++ project, and uses g++ to compile.

Why the stdint.h of gcc 4.7 is designed like this? Is that means gcc 4.7 don't want me to use those integer ranges with C++, and gcc 4.9 allows it?

And the most important: how should I work around this?

Answer 1

In C++ you're recommended to use std::numeric_limits #include <limits> :

Usage example from cplusplus.com

// numeric_limits example
#include <iostream>     // std::cout
#include <limits>       // std::numeric_limits

int main () {
  std::cout << std::boolalpha;
  std::cout << "Minimum value for int: " << std::numeric_limits<int>::min() << '\n';
  std::cout << "Maximum value for int: " << std::numeric_limits<int>::max() << '\n';
  std::cout << "int is signed: " << std::numeric_limits<int>::is_signed << '\n';
  std::cout << "Non-sign bits in int: " << std::numeric_limits<int>::digits << '\n';
  std::cout << "int has infinity: " << std::numeric_limits<int>::has_infinity << '\n';
  return 0;
}

Answer 2

Include climits or limits.h that contains sizes of integral types. This header defines constants with the limits of fundamental integral types for the specific system and compiler implementation used. The limits for fundamental floating-point types are defined in cfloat ( earlier float.h). The limits for width-specific integral types and other typedef types are defined in cstdint (earlier stdint.h).

You should include cstdint instead of stdint.h