updated conditionals for use of time functions in getch.c

[GitMensch]

• prefer clock_gettime over gettimeofday per Posix 2008 deprecation of the later
• use HAVE_... macros + defines instead of own _AVAILABLE macros
• don't define HAVE_CLOCK_GETTIME for older MinGW environments, fixes MinGW: getch.c does not build, no option to work around via Makefile #345

[GitMensch]

Hm, for some reason I've trashed this PR...

But I'd still like the review for the following code, which works on any system I've tested:

/* ftime() is consided obsolete.  But it's all we have for
millisecond precision on older compilers/systems.  We'll
use clock_gettime() or gettimeofday() when available. */

#if defined( _POSIX_C_SOURCE) && (_POSIX_C_SOURCE >= 199309L) \
     && (!defined( __MINGW32__) || defined( CLOCK_REALTIME))
   /* only newer MinGW environments have clock_gettime and those
      have CLOCK_REALTIME as a macro */
   #define HAVE_CLOCK_GETTIME
#elif defined( _DEFAULT_SOURCE) || defined( _BSD_SOURCE) \
     || defined( __FreeBSD__) || defined( __MINGW32__)
   /* POSIX.1-2008 marks gettimeofday() as obsolete,
      recommending the use of clock_gettime instead, so we
      only use that alternative to POSIX_C conditionally */
   #define HAVE_GETTIMEOFDAY
#endif

#if defined( HAVE_CLOCK_GETTIME)
#include <time.h>

long PDC_millisecs( void)
{
    struct timespec t;

    clock_gettime( CLOCK_REALTIME, &t);
    return( t.tv_sec * 1000 + t.tv_nsec / 1000000);
}
#elif defined( HAVE_GETTIMEOFDAY)
#include <sys/time.h>

long PDC_millisecs( void)
{
    struct timeval t;

    gettimeofday( &t, NULL);
    return( t.tv_sec * 1000 + t.tv_usec / 1000);
}
#else    /* neither gettimeofday() or clock_gettime() available */
#include <sys/timeb.h>

long PDC_millisecs( void)
{
    struct timeb t;

    ftime( &t);
    return( (long)t.time * 1000L + (long)t.millitm);
}
#endif

I that's considered OK, then I'm directly checking this in.

[Bill-Gray]

One minor fix : the test on CLOCK_REALTIME is occurring before #including <time.h> and <sys/time.h>. So it'll always fail. And the latter file doesn't exist for Digital Mars or Borland. So before doing that test, I'd put in

#include <time.h>
#ifdef __has_include
   #if __has_include(<sys/time.h>)
       #include <sys/time.h>
   #endif
#endif

(and remove the inclusion of both further down, of course).

With that modification, I can at least confirm that the code compiles without warnings on Digital Mars, Borland 5.5, MinGW on Linux, clang-10, clang-18, and gcc 9.4.

Without that modification, and compiling your snippet separately (outside getch.c), both clangs and gcc failed to recognize that either clock_gettime() or gettimeofday() were available, defaulted to ftime(), and emitted warnings that ftime() is deprecated.

[GitMensch]

So we think that each environment that does has sys/time.h with clock_gettime also provides __has_include?

[Bill-Gray]

I think so, but there's no guarantee that I know of. There may be some platform out there that doesn't support _has_include, does have sys/time.h and CLOCK_REALTIME (but we'd miss it), and where the result would be that we'd percolate down to using either gettimeofday() or ftime(), and we'd get a deprecation warning.

I think it's very unlikely to happen, though. Compilers old enough to lack _has_include are probably also old enough not to deprecate the older functions. And I know for a fact that at least one compiler (Borland) lacks sys/time.h.

[GitMensch]

Let's see what I've broke. I really have to go now :-)