pb-notify won't reconnect between network disturbances #1
Assignees
Labels
Comments
tsani
added a commit
that referenced
this issue
Apr 30, 2017
pb-notify now dispatches notifications when pushes are received. See #2 although we still don't implement full notification mirroring. This is done by watching the event stream for push-type tickles. When one is received, the list of pushes since the last batch has been processed is requested. Each is formatted appropriately is dispatched as a libnotify notification. Architecturally, this is done with a separate thread with an IORef holding the timestamp of the last processed push. Initially, this variable is filled with the timestamp of the latest push with activity in the account. This thread is synchronized with the event-loop thread via a channel. When a tickle is received on the event-loop thread, it sends a message over the chan notifying the HTTP thread that the list of pushes needs to be updated. The connectivity problems -- see #1 -- are partially resolved. Connection to the websocket is wrapped with exception-handling code so that it can be retried every five seconds. HTTP requests are also wrapped with retry logic until they succeed. Waiting for messages on the event stream is wrapped with a timeout of 35 seconds. (Since pushbullet sends a nop every 30 seconds.) If the timeout is hit, then the event-loop dies and the websocket connection is restarted. There are still some connectivity problems though. For instance, if the network is down and pb-notify is launched, then we get repeated `getAddrInfo: name or service not found` errors, as we expect, but if the network subsequently comes up, these errors continue rather than a connection successfully being established. I'm not sure why this is happening, but I suspect it has something to do with a networking library being used under the hood, as I've noticed similar behaviour in glirc2. Sometimes, doing a /reconnect on it doesn't work despite the network being up! Similarly, I get repeated DNS resolution errors there as well when the error occurs.
|
I believe that this issue is broader than just tpb, as glirc2 appears affected by it too. It might have something to do with the Haskell RTS or the underlying network library incorrectly doing some kind of caching. The specific error that I get in glirc2 is a |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
If pb-notify is running during a network outage, then when the network comes back up, rather than crash or reconnect, it just becomes unresponsive. The process remains alive, but no more notifications are delivered.
The text was updated successfully, but these errors were encountered: