get_time_in_ns: switch to CLOCK_MONOTONIC

(previously TIME_UTC corresponed with CLOCK_REALTIME)

The function is intended to measure intervals rather to wall-clock time,
so that CLOCK_MONOTONIC is more eligible - if there is a clock jump
caused by clock_settime, it may cause problems. Even until now, this
was not much likely, anyways, because this would happen just when an
administrator sets the time with clock_settime. NTP adjusts the time
gradually and never jumps back (see ntp_adjtime()) and it also adjusts
both clocks at the same time, anyways (the clock is actually one,
REALTIME is adding an fixed offset). Not sure about Windows (where the
clock_getime API is provided by MinGW, not the system iteslf).

Author: MartinPulec

src/tv.c

@@ -250,14 +250,14 @@ uint32_t get_std_video_local_mediatime(void)
 }
 
 /**
- * @returns nanoseconds since implementation defined epoche (timespec_get def)
+ * @returns nanoseconds since some time point (CLOCK_MONOTONIC)
  */
 time_ns_t
 get_time_in_ns()
 {
         struct timespec ts = { 0, 0 };
-        if (timespec_get(&ts, TIME_UTC) == 0) {
-                perror("timespec_get");
+        if (clock_gettime(CLOCK_MONOTONIC, &ts) != 0) {
+                perror("get_time_in_ns: clock_gettime");
         }
         return (ts.tv_sec * NS_IN_SEC) + ts.tv_nsec;
 }