restrict keyword used in header breaks some C++ inclusion

[mitchblank]

in xxhash.h:

#if !(defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L))   /* ! C99 */
#  define restrict   /* disable restrict */
#endif

restrict is technically only a keyword in C, not C++ yet this header needs to be includible from either. Most C++ compilers I guess either accept it as an extension or doesn't set __STDC_VERSION__ incorrectly. However, one host in my build farm (a Solaris x86_64 machine with gcc 4.7.3) is getting a compile error when this header is included from C++ code:

include/xxhash.h:233:61: error: expected ',' or '...' before 'dst_state'
include/xxhash.h:234:61: error: expected ',' or '...' before 'dst_state'

According to wikipedia https://en.wikipedia.org/wiki/Restrict

C++ does not have standard support for restrict, but many compilers have equivalents that usually work in both C++ and C, such as the GNU Compiler Collection's and Clang's restrict, and Visual C++'s __restrict and __declspec(restrict).

I'll try fixing it tomorrow, but I'll probably do something like:

#ifdef __cplusplus
#  if defined(__GNUC__) || defined(__clang__)
#    define XXH_RESTRICT __restrict__
#  elif defined (_MSC_VER)
#    define XXH_RESTRICT  __restrict
#  endif
#else
#  if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)   /* C99 */
#    define XXH_RESTRICT restrict
#  endif
#endif
#ifndef XXH_RESTRICT
#  define restrict   /* disable restrict */
#endif

I think defining an xxhash-specific macro like XXH_RESTRICT is less intrusive than potentially defining restrict which might affect how system headers, etc, are interpreted.

[Cyan4973]

Thanks for feedback.

It's also possible to remove restrict which use case within xxhash is extremely limited.
That would reduce complexity.

[Cyan4973]

restrict removed from source in "dev" branch

[mitchblank]

I can confirm my previously failing environment builds after the 386a634 change