Reasons for using regular approvals instead of permit2:
- More simplicity, simpler setup
- Unfortunately we won't be able to "expire" approvals, which is a risk we may be okay with taking - since one advantage of this is that we will be able save on gas by not approving the reactor every single time (is this our intention?)
TODO:
- Do we really need input tokens to pass through our router?
- Handle native ETH (double check how this is transferred in the reactor)
- locked-ether: We do have a withdraw method. Not sure why it doesn't register with slither. Already tried renaming.
- low-level-calls: This low level call is integral to the design of our filler contract.=
Encoding: integration tests are now separate and inside their own test folder
Execution: the final integration test should be inside of the protocol test file now and not in TychoRouterProtocolIntegration.t.sol. For this I had to move the BebopExecutionHarness.t.sol outside of the Bebop test file (because of imports)
Took 24 minutes
# Commit time for manual adjustment:
# Took 2 minutes
Because we don't know the size of hook data, it needs to be at the end of the protocol data. But we also don't know the size of the intermediary swaps. To solve this, we are now ple encoding the intermediary swaps and only then appending the hook data
Took 2 hours 50 minutes
Took 40 seconds
Add hook address to encoded data and then use it in execution
Add execution test for Euler
Took 1 hour 19 minutes
Took 2 hours 40 minutes
Took 3 minutes
Took 2 minutes
- Move encoding integration tests to integration test folder (in the rust project)
- Move protocol integration tests be inside the protocol file. This way we can change the block without any problems and it is easier for integrators
Took 6 minutes
Took 6 minutes
Took 17 minutes
Also added a getForkBlock() on TychoRouterTestSetup.
Moved the router balancer test inside the Balancer file and created a TychoRouterForBalancerV3Test to wrap it
Took 1 hour 10 minutes
Took 21 seconds
Refactor BalancerV3Executor to have an inner _swapCallback method with the real swapping logic. Then we have two external methods:
- handleCallback: called by the router. Here we need to remove the first 68 bytes (4 selector + 32 dataOffset + 32 dataLength)
- swapCallback: called by the Vault directly if we are swapping against the executor directly (no router involved)
Took 1 minute
We don't want to be responsible for holding private keys -> the user is the one that should do this outside of tycho-execution
Done:
- Remove signature from EncodedSolution
- Introduce UserTransferType and pass that everywhere instead of is_permit2_active and token_in_already_in_router
- Remove signing from permit2. Added it to the encoding_utils.rs only
- Mark encode_full_calldata as deprecated
- Backwards compatibility: still accept a signer for the encode_full_calldata case
- Update all tests
Took 2 hours 10 minutes
Took 13 minutes
This way the user is responsible for encoding the Tycho Router method inputs that are used as guardrails in execution.
Interface changes:
- Create EncodedSolution
- StrategyEncoder
- don't need to know have permit2 or token_in_already_in_router as attributes anymore
- encode_strategy returns EncodedSolution now (no method encoding done here now)
- TychoEncoder
- add encode_solution() method. This is the recommended method for users
- needs to have permit2, token_in_already_in_router and router_address as attributes
- permit creation is made in the router now
Also:
- create encoding_utils.rs
- update all tests
Took 2 hours 42 minutes
Took 3 minutes
Took 13 minutes
- In RestrictTransferFrom:
- Compare tokenIn with tokenIn from storage
- Correct docstrings
- Recompute storage slots with new names
- Rename transferFromNeeded to isTransferFromAllowed
- Don't track amount spent but subtract from amount allowed
- In TychoRouter: Rename transferFromNeeded to isTransferFromAllowed
Took 32 minutes
