Client keeps updating tip after Shutdown requested #3966

1. 
   laanwj commented at 11:00 am on March 27, 2014: member

   I’ve shut down my client (on testnet) and it keeps running, printing messages like this:

   02014-03-27 10:57:57 Pre-allocating up to position 0x1000000 in blk00006.dat
   12014-03-27 10:58:00 Pre-allocating up to position 0x100000 in rev00006.dat
   22014-03-27 10:58:00 UpdateTip: new best=00000000c92aeb1fbb19c781dea3f28248a8cb0d54c3776e207f1021455b76a6  height=208103  log2_work=58.297563  tx=1325101  date=2014-03-26 13:29:00 progress=0.999795
   

   My guess is that some newly introduced loop is ignoring interruptions.

2. laanwj added the label Bug on Mar 27, 2014
3. laanwj added the label Priority Low on Mar 27, 2014
4. 
   laanwj commented at 11:02 am on March 27, 2014: member

   After a minute it finally really quit:

   02014-03-27 11:01:33 UpdateTip: new best=000000000c540716cb66222d5ed90de88bbddbe812b83d15d7527c835b2a5e14  height=208174  log2_work=58.298491  tx=1333404  date=2014-03-27 11:21:16 progress=1.000003
   12014-03-27 11:01:33 ProcessBlock: ACCEPTED
   22014-03-27 11:01:33 msghand thread interrupt
   32014-03-27 11:01:33 Shutdown : In progress...
   42014-03-27 11:01:33 StopNode()
   52014-03-27 11:01:34 Shutdown : done
   62014-03-27 11:01:34 Shutdown finished
   72014-03-27 11:01:34 Shutdown result: 1
   82014-03-27 11:01:34 Stopping thread
   92014-03-27 11:01:34 Stopped thread
   

5. laanwj renamed this:
   Client keeps updating tip after Shutdown
   Client keeps updating tip after Shutdown requested
   on Mar 31, 2014
6. 
   laanwj commented at 4:48 pm on April 1, 2014: member

   My guess is that it was in this loop: https://github.com/bitcoin/bitcoin/blob/master/src/main.cpp#L2086 Introduced in 75f51f2a63e0ebe34ab290c2b7141dd240b98c3b

   Seemingly, disconnecting and connecting the new blocks can take quite a while. A boost::interruption_point() every so many blocks would solve this issue.

   Just need a way to reproduce this…

7. 
   sipa commented at 5:12 pm on April 1, 2014: member

   If a large batch of orphans was received, and they just got connected (by receiving their missing parent link), you can end up with a very long (minutes…) lock on cs_main while that batch is being connected. I believe I mentioned that in the PR :)
8. 
   rebroad commented at 5:08 am on April 6, 2014: contributor

   I’ll have a go at fixing this…
9. 
   laanwj commented at 11:51 am on September 8, 2014: member

   This hasn’t been a problem for a long time.
10. laanwj closed this on Sep 8, 2014

    ---

11. DrahtBot locked this on Sep 8, 2021

Contributors
laanwj sipa rebroad

Labels
Bug