You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You send big enough (for your internet channel) message to activate lm buffering.
Loudmouth sends first portion of message and puts remaining part into buffer.
Upon next call to socket writer, loudmouth will try to send all remaining data in one pass - i.e. it will stop only when socket will return error, EAGAIN will not cause it to break the loop and wait for next chance. This is one issue. Situation gets worse, because there's three places, where it should be fixed - plaintext connection (relatively easy), and two ssl engines.
There I am not so sure, but such situation usually results in reconnect due to ping timeout. Maybe loudmouth should take into account buffering state, when estimating channel vitality. Ping requests get appended to the end of buffer, and until that message was sent, they will remain unsent...
P.S. An easy way to get busy enough channel - is to fill it with outgoing torrent traffic.
The text was updated successfully, but these errors were encountered:
Not sure, need to retest the issue. Not right now, though. Probably still valid, but not that important anymore, since it's hard to get really bad internet channel now.
Order of events:
P.S. An easy way to get busy enough channel - is to fill it with outgoing torrent traffic.
The text was updated successfully, but these errors were encountered: