[mythtv] MBE segfaults when a slave backend goes away

[Robert Kulagowski]

I'm troubleshooting some of the slowdowns that I'm seeing with "The 
Nanny".  Anyway, during a scheduling run of TN, I had two programs 
recording on the MBE / mysql machine / NFS server and one on a slave 
backend.

Here's what the MBE looked like:

<snip>
The Nanny - "The Rosie Show"        45 1045  23 23:00-23:30  1 14 14  A 
0 14   0
The Nanny - "Freida Needa Man"      45 1045  23 23:30-00:00  1 13 13  A 
0 13   0
---  print list end  ---
2004-04-15 18:24:14 Scheduled 311 items in 58.7951 seconds.
2004-04-15 18:29:24 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:25 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:26 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:27 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:28 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:29 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:29 428     QUERY_REMOTEENCODER 
16[]:[]RECORD_PENDING[]:[]3...
2004-04-15 18:29:30 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:31 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:32 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:33 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:34 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:35 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:36 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:37 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:38 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:39 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:40 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:41 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:42 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:43 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:44 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:45 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:46 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:48 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:49 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:50 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:51 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:52 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:53 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:54 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:55 36      QUERY_REMOTEENCODER 16[]:[]GET_STATE
2004-04-15 18:29:55 Slave backend: frontend has left the building
Mutex destroy failure: Device or resource busy
Segmentation fault

On the frontend, I had:
2004-04-15 18:24:09 IOBOUND - blocking in ThreadedFileWriter::Write()
2004-04-15 18:24:14 IOBOUND - blocking in ThreadedFileWriter::Write()
2004-04-15 18:24:14 IOBOUND - blocking in ThreadedFileWriter::Write()
2004-04-15 18:24:14 IOBOUND - blocking in ThreadedFileWriter::Write()
2004-04-15 18:29:24 1       0
2004-04-15 18:29:25 1       0
2004-04-15 18:29:26 1       0
2004-04-15 18:29:27 1       0
2004-04-15 18:29:28 1       0
2004-04-15 18:29:29 1       0
2004-04-15 18:29:29 2       OK
2004-04-15 18:29:30 1       0
2004-04-15 18:29:31 1       0
2004-04-15 18:29:32 1       0
2004-04-15 18:29:33 1       0
2004-04-15 18:29:34 1       0
2004-04-15 18:29:35 1       0
2004-04-15 18:29:36 1       0
2004-04-15 18:29:37 1       0
2004-04-15 18:29:38 1       0
2004-04-15 18:29:39 1       0
2004-04-15 18:29:40 1       0
2004-04-15 18:29:41 1       0
2004-04-15 18:29:42 1       0
2004-04-15 18:29:43 1       0
2004-04-15 18:29:44 1       0
2004-04-15 18:29:45 1       0
2004-04-15 18:29:46 1       0
2004-04-15 18:29:48 1       0
2004-04-15 18:29:49 1       0
2004-04-15 18:29:50 1       0
2004-04-15 18:29:51 1       0
2004-04-15 18:29:52 1       0
2004-04-15 18:29:53 1       0
2004-04-15 18:29:54 1       0
At this point I hit CTRL-C on the slave backend, and got the subsequent 
segfault on the MBE.

I'm running the MBE in gdb now, just in case it happens again.