ThreadRPCServer not finished before exit(0)

[gavinandresen]

I found this debugging a crash-on-exit running the latest code on Windows XP.

Sending bitcoind the RPC 'stop' command, I see the following "Thread exited" messages:

2013-01-22 19:35:01 ThreadRPCServer method=stop
2013-01-22 19:35:03 ThreadSocketHandler exited
2013-01-22 19:35:04 ThreadOpenConnections exited
2013-01-22 19:35:04 ThreadMessageHandler exited
2013-01-22 19:35:04 StopNode()
2013-01-22 19:35:09 Bitcoin exited

There should be a ThreadRPCServer exited message there before 'Bitcoin exited'

I'm 90% sure the crash on Windows is happening in ThreadRPCServer, decrementing vnThreadsRunning[THREAD_RPCLISTENER]. I suspect the vnThreadsRunning destructor is being called before ThreadRPCServer2 returns, because ThreadRPCServer2 is waiting inside io_service.run_one()

The entire threads/fShutdown/StartShutdown()/Shutdown() scheme is fragile and a big mess, and should be cleaned up. Not for the 0.8 release, though...

[Diapolo]

ACK to cleaning that up, I always see that shutdown code and wonder how that is working anyway ^^. Even all that little sleeps() in there or the exit thread... some things look rather weird.

[gavinandresen]

Two more crash-on-exit reports in the rc1 thread:
https://bitcointalk.org/index.php?topic=142151

[sipa]

@gavinandresen Does this still exist after the threads rework?

[gavinandresen]

Closing, fixed by the threads work.