Crashed Related to RTCM Connect Time

[tsawyer]

As RTCM nodes are connected to my hub upwards of 500 hours the connect time display becomes negative. Soon thereafter Asterisk begins to behave erratically and eventually crashes or requires a restart. Disconnecting and reconnecting a RTCM node to prevent exceeding that time prevents related crashes. Guessing this would be a chan_voter issue.

[tsawyer]

Looks like the fix has been discovered for this. The sample code below shows a bug which is caused by an operator precedence error (the typedef to "int"). Apparently this code is all over app_rpt and affects all node types. This is not just a chan_voter issue. This will go a long ways towards fixing a major source of long-term instability in app_rpt.

`
#include <stdio.h>

void main()
{
int hours, minutes, seconds;
long long connecttime, connecttime2;
char conntime[64];
int i;

// ORIGINAL CODE:

   printf ("Original Code:\n");

   connecttime2 = 4294967296L / 2 - 5000L;

   for (i = 0; i < 10; i++) {

           connecttime = connecttime2;

           hours = (int) connecttime/3600000;
           connecttime %= 3600000;
           minutes = (int) connecttime/60000;
           connecttime %= 60000;
           seconds = (int)  connecttime/1000;
           connecttime %= 1000;
           snprintf(conntime, 32, "%02d:%02d:%02d.%d", hours, minutes, seconds, (int) connecttime);
           conntime[32] = 0;

           printf ("time = %s\n", conntime);
           connecttime2 += 1000;
   }

// FIXED CODE:

   printf ("\n\nNew Code:\n");

   connecttime2 = 4294967296L / 2 - 5000L;

   for (i = 0; i < 10; i++) {

           connecttime = connecttime2;

           hours =  connecttime/3600000L;
           connecttime %= 3600000L;
           minutes =  connecttime/60000L;
           connecttime %= 60000L;
           seconds =  connecttime/1000L;
           connecttime %= 1000L;
           snprintf(conntime, 32, "%02d:%02d:%02d.%d", hours, minutes, seconds, (int) connecttime);
           conntime[32] = 0;

           printf ("time = %s\n", conntime);
           connecttime2 += 1000L;
   }

}
/////////////////////////////////////////////////////////////////////////////////////////////////
/////////////////////////////////////////////////////////////////////////////////////////////////
/////////////////////////////////////////////////////////////////////////////////////////////////

[root@alarmpi tmp]# ./a.out
Original Code:
time = 596:31:18.648
time = 596:31:19.648
time = 596:31:20.648
time = 596:31:21.648
time = 596:31:22.648
time = -596:31:23.648
time = -596:31:24.648
time = -596:31:25.648
time = -596:31:26.648
time = -596:31:27.648

New Code:
time = 596:31:18.648
time = 596:31:19.648
time = 596:31:20.648
time = 596:31:21.648
time = 596:31:22.648
time = 596:31:23.648
time = 596:31:24.648
time = 596:31:25.648
time = 596:31:26.648
time = 596:31:27.648
`

[tsawyer]

Over time we've learned that this is issue impacts more than just RTCM the connect time display. Users with other hardware have reported crashes and disconnects occurring around the 600 hour mark. I've seen it with RTCM hardware because that's all I use. But I sense that any two nodes connected to each other on the same server will crash near the 600 hour mark.