🧵 Improve synchronization of connection_state transitions#494
Merged
Conversation
If `disconnected?` returns true, the connection state is most likely already `logout`. But, just in case, we'll ensure it's set every time.
Net::IMAP#disconnect will still attempt to enter the logout state first, but it won't wait for the lock until after the socket has been shutdown.
The reciever thread should close the connection before it finishes, and `@sock.shutdown` should've been enough to trigger that. But this ensures the socket is closed right away, without needing to wait on whatever the receiver thread might be in the middle of doing.
It should always be safe to call `@sock.close` without external synchronization. And synchronizing here with could get stuck waiting e.g. for a response_handler looping in the receiver thread.
This should work both for TCPSocket and for OpenSSL::SSL::SSLSocket. Both will respond to `#to_io` with the TCPSocket.
Eventually the states will store small bits of information about each state on the state objects, e.g: which mailbox was selected or what caused the logout. So for `logout`, the first attempt wins and shouldn't be overwritten. For other states, like `selected`, the last attempt _should_ win. On the other hand, state transitions should be reworked so that: * state transitioning commands cannot run simultaneously * all state transitions are effected _inside_ the receiver thread
This pushes `ensure state_logout!` from the receiver's `Thread.start` down into `#receive_responses`. As a side-effect, this also pulls the state change inside the receiver thread's exception handling. But that's fine: it fits better inside `receive_responses`, and `state_logout!` really should never raise an exception anyway.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Handles a few more connection_state synchronization issues I saw while working on #493.
While attempting to get the tests running under JRuby,
#disconnectwould sometimes deadlock. So, although I haven't seen that problem elsewhere and JRuby is still failing tests with another threading issue, this did get most tests passing for me (but not in CI).