fix(p2p): fix EventChannel, add tests

[elenaf9]

Description of change

Refactor Sink and Stream implementation for EventChannel.
The Sink implementation allows to write a message to the channel (blocking or non-blocking).
In case of an non-blocking Sink, the Stream implementation has to be continuously polled. The polling drives the sending of messages once the inner channel has the capacity to receive it (in the SwarmTask, this polling is done in the central loop in Stream::run).
The ChannelSinkConfig allows to configure the channels behaviour:

• in case of ChannelSinkConfig::Block:
  • Sink implementation delegates every call to the inner mpsc::Sender
  • if the inner channel has capacity for the message: the message is send and EventChannel::send returns
  • if the inner channel is full: EventChannel::send blocks until the message can be send
  • Stream implementation is not relevant.
• in case of ChannelSinkConfig::DropLatest | ChannelSinkConfig::BufferLatest:
  • if the inner channel has capacity for the message:
    • start sending the messages
    • polling <EventChannel as Stream> flushes the message
  • if the inner channel is full:
    • if ChannelSinkConfig::BufferLatest
      • Write message to buffer
      • polling <EventChannel as Stream> writes from buffer to inner channel once there is capacity
    • if ChannelSinkConfig::DropLatest: drop message

Type of change

Choose a type of change, and delete any options that are not relevant.

• Bug fix (a non-breaking change which fixes an issue)

How the change has been tested

Added event_channel::test module to test every ChannelSinkConfig.

Change checklist

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[felsweg-iota]

Since an EventChannel<T> can only have the values of ChannelSinkConfig this implementation is Ok, but I would suggest to leave room for extension: Reduce the responsibility of new to one, create new functions like new_blocking(..), new_drop_latest(...), etc...

[elenaf9]

I personally favor having a single new method with an enum, over having a separate new method for each variant. Especially since the rest of the parameters (in this case only capacity) is the same for all variants, hence all 3 methods would have the same signature.

[felsweg-iota]

May I also ask you to describe the bug in a few sentences, that has been fixed? Thank you :)

[elenaf9]

May I also ask you to describe the bug in a few sentences, that has been fixed? Thank you :)

The bug was caused by a fundamental misunderstanding on my side of what Sink::poll_flush does. In the old implementation I assumed that "flushing" should do something like clear all pending requests, hence I actually just cleared the buffer and then called inner.poll_flush. In reality poll_flush is responsible for doing the actual writing of the request and in case of mpsc::Sender it blocks the send operation until the channel has capacity to received the request. Now with this PR it is implemented that poll_flush only blocks in case of ChannelSinkConfig::Block, otherwise it immediately returns and the send-request-once-channel-has-capacity part is done in the Stream implementation

[felsweg-iota]

❤️