[PATCH] support: Print timestamps in timeout handler

[Gabriel F. T. Gomes]

On Tue, 13 Nov 2018, Florian Weimer wrote:

>This is sometimes useful to determine if a test truly hang, or if it
>was making progress (logging information to standard output) and was
>just slow to complete.
>
>2018-11-13  Florian Weimer  <fweimer@redhat.com>
>
>	support: Print timestamps in timeout handler.
>	* support/support_test_main.c (print_timestamp): New function.
>	(signal_handler): Use it to print the termination time and the
>	time of the last write to standard output.

Looks good to me.

>+static void
>+print_timestamp (const char *what, struct timeval tv)
>+{
>+  struct tm tm;
>+  if (gmtime_r (&tv.tv_sec, &tm) == NULL)
>+    printf ("%s: %lld.%06d\n",
>+            what, (long long int) tv.tv_sec, (int) tv.tv_usec);
                    ~~~~~~~~~~~~~~~            ~~~~~
OK.

>+  /* Do this first to avoid further interference from the
>+     subprocess.  */
>+  struct timeval now;
>+  bool now_available = gettimeofday (&now, NULL) == 0;
>+  struct stat64 st;
>+  bool st_available = fstat64 (STDOUT_FILENO, &st) == 0 && st.st_mtime != 0;

OK.

>+  if (now_available)
>+    print_timestamp ("Termination time", now);
>+  if (st_available)
>+    print_timestamp ("Last write to standard output",
>+                     (struct timeval) { st.st_mtim.tv_sec,
>+                         st.st_mtim.tv_nsec / 1000 });

OK.

The st_mtim member of stat64 is only available in XOPEN2K8 mode, which is
fine for the test cases.

PS: I was reading about these types in the manual, but there's no mention
to st_mtim, only to st_mtime, which is provided when *not* in XOPEN2K8.
Does this deserve mention in the manual?  (not a problem for this patch).