tests: add solver delegate coverage

[igorroncevic]

Summary

This PR adds focused test coverage for Solver7702Delegate.

What changed

• Added unit tests for fallback behavior:
  • short calldata, including empty calldata, simply receives ETH and returns no data
  • approved callers forward packed bytes20(target) || payload calldata
  • unauthorized callers revert
  • target return data and revert data bubble back to the caller
• Added gas snapshots for the success paths that matter, using cleaner last-call snapshots for delegate fallback gas.
• Added mainnet fork replay tests for real historical CoW transaction batches.
• Added support for user-supplied historical transaction hashes through COW_HISTORICAL_TX_HASHES.
• Fork replay now fetches each original transaction target and calldata from RPC, rolls the fork to that transaction, compares a direct replay against the delegated replay, and uses vm.etch to simulate the EIP-7702 delegation.
• Removed settlement-specific mock calldata construction and settlement test interfaces from the fork tests.
• Reduced fork RPC config to ETH_MAINNET_RPC_URL; despite the name, it can point to another network when replaying tx hashes from that network.
• Added README notes for replaying custom transactions with a user-provided RPC URL and comma-separated tx hash list.
• Added test-only Solhint exceptions and checked gas snapshots in CI.

How to test custom transactions

ETH_MAINNET_RPC_URL=<your_rpc_url> \
COW_HISTORICAL_TX_HASHES=0xabc...,0xdef... \
just test --match-test test_fork_historicalTransaction_directVsDelegated_userSuppliedTxHashes

The supplied tx hashes must exist on the RPC's network, and the RPC must support the historical state needed by vm.rollFork(txHash).

Test plan

Run:

just all

[gemini-code-assist]

Code Review

This pull request introduces a comprehensive test suite for the Solver7702Delegate contract, including unit, integration, and fork tests, along with necessary mocks and gas snapshots. It also adds documentation to the fallback function regarding the expected calldata format. Feedback focuses on correcting the invalid Solidity pragma version 0.8.34 to a stable release and improving fuzz test coverage by allowing empty payloads instead of substituting them with 0x00.

[Copilot]

Pull request overview

Adds a comprehensive Foundry test suite around Solver7702Delegate’s EIP-7702 delegated fallback forwarding behavior, including unit, integration (mock GPv2), and fork-based regression tests.

Changes:

• Added unit/integration/fork tests plus supporting mocks and utilities for constructing settlement-like calldata and revert/return-data scenarios.
• Added gas snapshot JSONs for the new test suites.
• Updated tooling/config (test solhint overrides, coverage exclusions to ignore lib/, Foundry remappings, .env.example, and OpenZeppelin submodule pin).

Reviewed changes

Copilot reviewed 25 out of 26 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
test/utils/SettlementUtils.sol	Utility library to build mock/real GPv2 settlement calldata for integration/fork tests.
test/utils/MathUtils.sol	Small test-only math helper library.
test/Solver7702Delegate.t.sol	Unit tests covering constructor behavior and fallback forwarding/revert bubbling.
test/Solver7702Delegate.integration.t.sol	Integration tests using mocked GPv2 settlement/authenticator.
test/Solver7702Delegate.fork.t.sol	Fork tests against real GPv2 settlement across networks plus historical calldata replay.
test/mocks/targets/RevertingTarget.sol	Target contract producing standard Solidity revert shapes (custom error/string/panic/empty).
test/mocks/targets/RawRevertTarget.sol	Target contract that reverts with arbitrary raw bytes.
test/mocks/targets/PayableFallbackTarget.sol	Payable fallback/receive target used to assert exact forwarding semantics.
test/mocks/targets/NonpayableTarget.sol	Nonpayable target used to assert value-forwarding failure behavior.
test/mocks/MockSettlement.sol	Simple payable settlement-like target for delegate integration tests.
test/mocks/MockGPv2Settlement.sol	Minimal GPv2 settlement mock that checks solver allowlist and executes interactions.
test/mocks/MockGPv2Authenticator.sol	Minimal solver allowlist mock used by integration tests.
test/dependencies/settlement/IGPv2Settlement.sol	Test-local GPv2 settlement interface/types used by mocks and calldata builders.
test/dependencies/settlement/IGPv2Authenticator.sol	Test-local GPv2 authenticator interface used by mocks.
test/BaseTest.t.sol	Shared test harness: approved callers, packed calldata helper, delegation attachment helper.
test/.solhint.json	Test-folder solhint overrides to accommodate patterns used in the new tests.
src/Solver7702Delegate.sol	Added calldata-format documentation to the fallback function.
snapshots/Solver7702DelegateTest.json	Gas snapshot baselines for unit tests.
snapshots/Solver7702DelegateIntegrationTest.json	Gas snapshot baselines for integration tests.
snapshots/Solver7702DelegateForkTest.json	Gas snapshot baselines for fork tests.
Justfile	Adjusted test/coverage commands (including excluding lib/ from coverage).
foundry.toml	Added OpenZeppelin remapping and formatting adjustment.
foundry.lock	Added OpenZeppelin submodule pin.
.gitmodules	Added OpenZeppelin contracts submodule.
.env.example	Added RPC URL environment variables for fork testing.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kaze-cow]

Good start! The tests could use more foundry provided utilities and there were some unusual patterns that I couldn't figure out.

[kaze-cow]

if practical it would be good to only disable next line rathre than disabling the whole file. Otherwise it is easy to miss out on legitimate checks (or understand why they are needed to be disabled in the first place)

[fedgiac]

I agree it isn't good to do this kind of global ignore for the aforementioned reasons.
I'd argue in the other direction though: most of these should be ignored for all tests, we should even add ignores to the template.

• avoid-low-level-calls, this is the only one I'd possibly keep, it's nice for readability if we use proper calls and it's ok to add in-line exceptions when needed.
• gas-small-strings, this is for gas efficiency and tests shouldn't be developed with gas efficiency in mind.
• max-states-count, I think it's fine if tests that need it have a lot of state, so it's fine to remove them globally. If you think this helps keeping the tests simple, we can keep it however.

[igorroncevic]

low level calls are particularly useful when calling a delegate contract because there are no signatures to be called there, so we simply pack the data in the expected format (targetAddress + calldata) and call the solver directly.

[fedgiac]

Very quick review on my part, I like how tests are structured but there are a few things I don't understand.

[fedgiac]

I agree it isn't good to do this kind of global ignore for the aforementioned reasons.
I'd argue in the other direction though: most of these should be ignored for all tests, we should even add ignores to the template.

• avoid-low-level-calls, this is the only one I'd possibly keep, it's nice for readability if we use proper calls and it's ok to add in-line exceptions when needed.
• gas-small-strings, this is for gas efficiency and tests shouldn't be developed with gas efficiency in mind.
• max-states-count, I think it's fine if tests that need it have a lot of state, so it's fine to remove them globally. If you think this helps keeping the tests simple, we can keep it however.

[igorroncevic]

The tests are significantly simpler and have had a major rework, so some reviews might be outdated.

[Copilot]

Pull request overview

Copilot reviewed 13 out of 14 changed files in this pull request and generated 4 comments.

[kaze-cow]

I was able to try out the new fork tests! super nice, ergonomic. only minor things left.

I noticed that the gas snapshots measuring how much gas was used for each different approved caller appears to have been removed. I actually liked this breakdown--any chance we can bring it back?

worth mentioning that rollFork unfortunately does take a lot of time because it has to replay all the transactions prior to the target one to get a perfect simulation. But its a 1st time thing only because anvil caches previous calls to disk.

[kaze-cow]

all of my remaining comments were addressed. We can get rid of BASE_RPC_URL by using a different transaction (discussed in private call), so once that is resolved I am happy to merge.