-
Notifications
You must be signed in to change notification settings - Fork 2.2k
Segmentation fault during sync #5633
Comments
It looked like |
@gumbo: which commit was your Aleth syncd to? Does it include the host gc work? |
@halfalicious I think it was 932fdd9 so it doesn't |
Still seeing the exception during mainnet sync though I'm not hitting the segfault (maybe because I have multiple connected peers?):
|
Hit the segfault on mainnet sync:
|
@gumb0 I was curious about this so I started by investigating why the disconnect exceptions are occurring...I've figured that part out, I still need to investigate how this can result in an AV though. The disconnect exceptions occur because a peer can successfully reconnect while there's still an outstanding
aleth/libethereum/EthereumCapability.cpp Lines 611 to 619 in ebf792d
A possible fix could be to add another function to |
Maybe would be nicer if the peer reconnects during this 2 seconds delay to immediately cleanup the old peer state and create new session (I guess this is how it worked before we introduced the 2 sec delay). But I don't know how complicated that would be. |
@gumb0 Is the only way the AV could have happened is if the |
@gumb0 Could perhaps |
This was fixed by PR #5834 |
This crash happened to me while running Ropsten sync with Release build under gdb on Linux.
The log end with
The stack of all threads
The text was updated successfully, but these errors were encountered: