Issue #9079: Added _PyTime_gettimeofday(_PyTime_timeval *tp) to C API
exposed in Python.h. This function is similar to POSIX
gettimeofday(struct timeval *tp), but available on platforms without
gettimeofday().
diff --git a/Python/pytime.c b/Python/pytime.c
new file mode 100644
index 0000000..6fb7695
--- /dev/null
+++ b/Python/pytime.c
@@ -0,0 +1,60 @@
+#include "Python.h"
+
+#ifdef __APPLE__
+#if defined(HAVE_GETTIMEOFDAY) && defined(HAVE_FTIME)
+ /*
+ * _PyTime_gettimeofday falls back to ftime when getttimeofday fails because the latter
+ * might fail on some platforms. This fallback is unwanted on MacOSX because
+ * that makes it impossible to use a binary build on OSX 10.4 on earlier
+ * releases of the OS. Therefore claim we don't support ftime.
+ */
+# undef HAVE_FTIME
+#endif
+#endif
+
+#ifdef HAVE_FTIME
+#include <sys/timeb.h>
+#if !defined(MS_WINDOWS) && !defined(PYOS_OS2)
+extern int ftime(struct timeb *);
+#endif /* MS_WINDOWS */
+#endif /* HAVE_FTIME */
+
+void
+_PyTime_gettimeofday(_PyTime_timeval *tp)
+{
+ /* There are three ways to get the time:
+ (1) gettimeofday() -- resolution in microseconds
+ (2) ftime() -- resolution in milliseconds
+ (3) time() -- resolution in seconds
+ In all cases the return value in a timeval struct.
+ Since on some systems (e.g. SCO ODT 3.0) gettimeofday() may
+ fail, so we fall back on ftime() or time().
+ Note: clock resolution does not imply clock accuracy! */
+#ifdef HAVE_GETTIMEOFDAY
+#ifdef GETTIMEOFDAY_NO_TZ
+ if (gettimeofday(tp) == 0)
+ return;
+#else /* !GETTIMEOFDAY_NO_TZ */
+ if (gettimeofday(tp, (struct timezone *)NULL) == 0)
+ return;
+#endif /* !GETTIMEOFDAY_NO_TZ */
+#endif /* !HAVE_GETTIMEOFDAY */
+#if defined(HAVE_FTIME)
+ {
+ struct timeb t;
+ ftime(&t);
+ tp->tv_sec = t.time;
+ tp->tv_usec = t.millitm * 1000;
+ }
+#else /* !HAVE_FTIME */
+ tp->tv_sec = time(NULL);
+ tp->tv_usec = 0;
+#endif /* !HAVE_FTIME */
+ return;
+}
+
+void
+_PyTime_Init()
+{
+ /* Do nothing. Needed to force linking. */
+}