Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.
Hi guys,
So here's my theory:
The client tries to connect and fails, there's still time in the connection window so it registers a task to reconnect:
https://github.com/xrootd/xrootd/blob/c91b1acf4fcf7f39c7b104353bcde1cc6fdd3032/src/XrdCl/XrdClStream.cc#L775-L780
However, note that although we failed to connect and registered a reconnect task we don't disable r/w events from the event loop.
Since we didn't disable the events the socket handler although disconnected is handling events generated by the event loop hence the first stack-trace.
Now at some point the reconnect task gets executed. It seems that in our case it is executed late (out of the connection window) so it triggers a connection error which deadlocks with the event-handling logic in (2).
I think what we need to do is disable events for given socket handler if we schedule a reconnect task. I will try to cook up a patch.
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.
Use REPLY-ALL to reply to list
To unsubscribe from the XROOTD-DEV list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1