Handling of own messages

[neekolas]

Summary

• Adds support for matching intents to incoming messages in the receive flow
• Updates the intent state depending on whether the message matches the group epoch
• Adds a sync function to encapsulate publishing, receiving, and post-commit flows
• Refactors message processing to branch out to the new flow if an intent is found

Notes

There are a few error cases that are hard to handle

1. Retryable vs non-retryable errors. When to abort
2. What happens if an intent can never be fully fulfilled? We probably need an error intent status for intents that are impossible (trying to add a member that doesn't exist, trying to add a member already present in the group)
3. Application messages from the wrong epoch. These may actually be fine depending on the max_past_epochs config We now don't reject application messages from the wrong epoch. There will be some follow-up work to try and catch application messages that are from outside the max_past_epochs window.

TODO

• Add support for writing a transcript of members added/removed
• Better mapping between commits and payloads. Should include a reference to the staged commit hash or something to make sure that when merge_pending_commit gets called it is actually being called on the correct commit
• More tests around failure modes and recovery

[neekolas]

I added this line to make my test case handling conflicting messages produce the right errors

[richardhuaaa]

To be explicit, what is the reason we can fire-and-forget messages but we can't do the same for other intents?

I'm guessing because OpenMLS only allows us to have one pending commit at a time? Which is not an issue for messages

[neekolas]

That's right. Especially with max_past_epochs we should be able to assume that sending a message will "just work" in almost all cases and not need to be re-queued. There are some edge cases where your local state is way out of sync with the network where it would need a retry, but they should be rare enough I think the trade-off is worth it.