The test do not work without acks as we stop the event loop as soon
as a message is considered as "sent" when actually the event loop
and swarm may not have yet sent the message.
The ack allow to avoid this issue as the message was considered "sent"
only once the other party sent a response. However, the ack brings
other issue so a review needs to be done to select the appropriate
solution.
If dialing Bob fails Alice waits for the acknowledgement of the transfer proof indefinitely.
The timout prevents her execution from hanging.
Added a ToDo to re-visit the ack receivers. They don't add value at the moment and should be removed.
Alice was attempting to create a new event loop using the same listen addr as the old one which was still running. This commit aborts the event loop before creating a new one.
Upgrade bitcoin harness dependency to latest commit
Upgrade backoff to fix failing tests. The previous version of backoff had a broken version of the retry function. Upgraded to a newer comit which fixes this problem.
Upgrade hyper to 0.14 as the 0.13 was bringing in tokio 0.2.24
Upgraded bitcoin harness to version that uses tokio 1.0 and reqwest 0.11
Upgrade reqwest to 0.11. Reqwest 0.11 uses tokio 1.0
Upgrade libp2p to 0.34 in preparation for tokio 1.0 upgrade
- Properly check the timelocks before trying to redeem
- Distinguish different failure scenarios and reactions to it.
- if we fail to construct the redeem transaction: wait for cancel.
- if we fail to publish the redeem transaction: wait for cancel but let the user know that restarting the application will result in retrying to publish the tx.
- if we succeed to publish the tx but then fail when waiting for finality, print error to the user (secreat already leaked, the user has to check manually if the tx was included)
This is not really a factory as a factory design pattern is about
producing several instances.
In the current usage, we are only interested in one swap instance. Once
the swap instance is created, the factory becomes useless. Hence, it is
more of a builder pattern.
- Introduce Test abstraction instead of tow harnesses, move test specific data into Test
- Change the abstraction from actors to swap, because we are creating swaps, not actors
- rename actor::swap to run, because we are running a swap
This makes the redeem assertion reusable for all tests with a redeem scenario.
Since the factory was not a clean factory before and is now doing even more it was renamed to harness.
This includes the introduction of the --data-dir parameter instead of the --database.
Both the seed file and the database are stored in the data-dir, the database in sub-folder `database`.
Created network, storage and protocol modules. Organised
files into the modules where the belong.
xmr_btc crate moved into isolated modulein swap crate.
Remove the xmr_btc module and integrate into swap crate.
Consolidate message related code
Reorganise imports
Remove unused parent Message enum
Remove unused parent State enum
Remove unused dependencies from Cargo.toml
There are no refund timelock, only a cancellation timelock and punish
timelock.
Refund can be done as soon as the cancellation transaction is published.
As Bob is dialing Alice, we now ensure that we are connected to Alice
at each step that needs communication.
If we are not connected, we proceed with dialing.
In an attempt to improve libp2p usage, we also add known address of
Alice first and only use peer_id to dial.
This ensures that we use the expected peer id.
The usage of the peer id is incorrect as we do not even check it when
dialing. For now, we can ignore it.
We can then re-introduce it and use it properly at a later stage.
Reworked Alice XmrLocked state transition handler to handle the
scenario when Alice received the encsig but Bob refunds.
Previously Alice was trying to redeem after receiving the encsig
without checking if t1 had elapsed.
Previously state0 had to be set after creating Alice's behaviour.
With the event loop we no longer has access to the swarm so
set_state0() has to be called indirectly through a channel. This
means it is difficult to guarantee state0 is being set due to the
asynchronous nature of channels. This was solved by initialising
Alice with state0.
Previously the libp2p swarm had to be manually polled within the
protocol execution code to execute actions such as sending a
message. The swarm is now wrapped in SwarmDriver which polls the
swarm in a seperate task
Tracing should be initialized by test and the `_guard` kept alive within the test.
Re-using this code in different tests does not really have any additional value.
Instead of specifying what messages we want to include, I went for a filter that excludes noise.
That way we get more useful logging.
Use reusable test init functions for happy path test
Extract tracing setup to reusable function
Move test initialization to seperate functions
Increase stack size in CI
Fix monero max finality time
Force Bob swarm polling to send message 2
Run Bob state to xmr_locked in punish test to force the sending of
message2. Previously Bob state was run until btc_locked. Although
this was the right thing to do, message2 was not being sent as the
swarm was not polled in btc_locked. Alice punish test passes.
Add info logging to executor
This introduces a lot of duplication between the binary and the
library, but it's okay because this module should only be a temporary
measure until we allow recovery to be handled by the original state
machine.
Also, fix a bug in `xmr_btc::alice::action_generator` caused by the
incorrect assumption that Alice's ability to punish Bob could be
determined before the cancel transaction hits the blockchain.
