feat: add solver 7702 delegate by igorroncevic · Pull Request #4430 · cowprotocol/services

igorroncevic · 2026-05-22T11:21:34Z

Description

This PR adds the new Solver7702Delegate flow for EIP-7702 parallel settlement submission.

The old forwarder flow required a separately deployed contract address in solver config and separate caller approval handling.
The new delegate makes setup more self-contained:
- the driver deploys the delegate deterministically from the configured submission accounts,
- delegates the solver EOA to it
- and then routes delegated settlement transactions through the solver EOA.

Solver7702Delegate is also much more efficient and secure that its predecessor, due to a minimal attack surface. The contract itself is vendored from https://github.com/cowprotocol/solver-7702-delegate

Changes

Add the vendored Solver7702Delegate artifact and generated Rust bindings.
Wire Solver7702Delegate into contract generation and the contracts facade.
Update EIP-7702 startup setup to:
- validate that submission accounts are signers,
- build delegate init code with approved callers,
- deploy the delegate with CREATE2 when missing,
- delegate the solver EOA to the deployed delegate,
- verify the final EIP-7702 delegation target.
Replace the old forward(target, data) calldata wrapping with the new delegate calldata format targetAddress || targetCalldata.
Add focused tests for delegate address derivation, caller limits, config validation, delegation detection, and delegated transaction rewriting.

How to test

just generate-contracts
just fmt --check
just clippy
just test-unit
just test-e2e parallel_settlement

gemini-code-assist

Code Review

This pull request replaces the CowSettlementForwarder contract with Solver7702Delegate to support EIP-7702 parallel settlement submissions. Key changes include the addition of the Solver7702Delegate artifact and its generated Rust bindings, a refactor of the driver's EIP-7702 setup to use deterministic CREATE2 deployment, and updates to the submission logic to handle the new delegate's fallback mechanism. The PR also adds unit tests for these components and reorganizes existing test files. I have no feedback to provide.

kaze-cow

wow the changes seem very straightforward, nice work! I haven't finished running the actual code locally so give me some time but if the paralell submission test is already working thats a great sign 👍

AryanGodara

looks good mostly, just 2 comments. 1 med, 1 small.
Also replid to the "lets get opinion of backend team" threads 😅

…gate # Conflicts: # contracts/generated/contracts-generated/solver7702delegate/Cargo.toml # contracts/src/main.rs # crates/driver/src/domain/mempools.rs # crates/e2e/tests/e2e/parallel_settlement.rs

jmg-duarte

this is quite big which makes it harder to review, overall it looks ok to me but it's a bit hard to visualize the impact of the change in my head

jmg-duarte · 2026-05-25T15:15:08Z

+                self.max_solutions_to_propose.get() == 1,
+                "solver '{}': max-solutions-to-propose > 1 requires at least one \
+                 submission-account (EIP-7702 parallel submission must be enabled)",


x == 1 & "at least one" are not the same

which one is the correct one?

Its definitely worded wierdly but I think its referring to two different vars. if the var max-solutions-to-propose > 1, then submission-account.length > 0

Good point. The check is about two different fields, but the message was confusing. I changed it to say that max-solutions-to-propose > 1 requires non-empty submission-accounts.

jmg-duarte · 2026-05-25T15:16:01Z

+        anyhow::ensure!(
+            self.submission_accounts
+                .iter()
+                .all(|account| !matches!(account, Account::Address(_))),


same thing but there's no ! almost hiding

Suggested change

.all(|account| !matches!(account, Account::Address(_))),

.any(|account| matches!(account, Account::Address(_))),

jmg-duarte · 2026-05-25T15:20:10Z

+    PrivateKeySigner::from_bytes(&b256!(
+        "59c6995e998f97a5a0044966f0945389dc9e86dae88c7a8412f4603b6b78690d"
+    ))
+    .unwrap()


just in case: leave a note explaining NOT to use this as a production key

its anvil well-known mnemonic signer #1 btw

jmg-duarte · 2026-05-25T15:22:20Z

    std::time::Duration,
 };

+const CREATE2_DEPLOYER: Address = address!("4e59b44847b379578588920cA78FbF26c0B4956C");


you can use the value from the driver

jmg-duarte · 2026-05-25T15:22:42Z

+}
+
+fn solver_delegate_address(callers: [Address; 2]) -> Address {
+    let mut approved_callers = [Address::ZERO; 5];


use MAX_ALLOWED_CALLERS (or what was the name of the var)

jmg-duarte · 2026-05-25T15:23:49Z

+/// Ensure EIP-7702 delegate deployment and solver delegation are set up for all
+/// solvers with parallel submission accounts. Called once at driver startup.


the error descriptions disappeared

jmg-duarte · 2026-05-25T15:24:17Z

-        let forwarder = config.forwarder_contract.ok_or_else(|| {
-            anyhow::anyhow!(
-                "solver {}: submission_accounts configured but forwarder_contract missing",
-                config.name
-            )
-        })?;


thought that this had been removed already

This check is now moved out of EIP-7702 setup and into config validation, where it belongs.

jmg-duarte · 2026-05-25T15:32:25Z

+fn delegation_status(code: &[u8]) -> DelegationStatus {
+    if code.is_empty() {
+        DelegationStatus::Empty
+    } else if code.len() == 23 && code.starts_with(&DELEGATION_PREFIX) {


23 is a magic number

jmg-duarte · 2026-05-25T15:32:36Z

+    Empty,
+    DelegatedTo(Address),
+    OtherCode,


docs for the variants

jmg-duarte · 2026-05-25T15:33:50Z

+    OtherCode,
+}
+
+fn delegation_status(code: &[u8]) -> DelegationStatus {


take it or leave it: maybe?

Suggested change

fn delegation_status(code: &[u8]) -> DelegationStatus {

impl DelegationStatus {

fn from_code(code: &[u8]) -> Self {

...

}

}

squadgazzz · 2026-05-25T20:15:36Z

-    // Check delegation status.
-    let code = provider.get_code_at(solver_address).await?;
-    let needs_delegation = !is_delegated_to(&code, forwarder);
+    let mut approved_callers = [Address::ZERO; MAX_APPROVED_CALLERS];


Worth rejecting duplicates and Address::ZERO here. Both change (or alias with the padding for) the CREATE2 target with no signal to the operator. A typo in the TOML would silently produce a working-but-wrong delegate.

anyhow::ensure!( callers.iter().all(|a| *a != Address::ZERO), "submission accounts cannot include the zero address" ); let mut seen = std::collections::HashSet::with_capacity(callers.len()); anyhow::ensure!( callers.iter().all(|a| seen.insert(*a)), "submission accounts must be unique" );

Also worth a comment noting that reordering submission accounts in the TOML changes the delegate address. The test asserts it, but operators won't read the test.

Agreed, added these checks

squadgazzz · 2026-05-25T20:15:36Z

+        .github(
+            "Solver7702Delegate",
+            // TODO(post-audit): bump to the audited commit hash or release tag.
+            "cowprotocol/solver-7702-delegate/acde4b0cd452d208c2ea0f0fbbf83e3698decbf8/out/\


The vendor URL is at commit acde4b0c..., but the committed JSON's sourceCommit is 4273853b.... Re-running the vendor command would pull a different artifact and overwrite the checked-in one. Please pin them to the same commit (the post-audit TODO can update both at once).

Thanks, I did not realize out folder needed to be exposed for this to work, so now that it's exposed I've pinned the commit to the latest one.

squadgazzz · 2026-05-25T20:15:36Z

+        "CREATE2 deployer {CREATE2_DEPLOYER:?} has unexpected code",
+    );
+
+    let tx_sender = config.account.address();


This recovery path pays gas from the solver EOA. If the operator only funds submission accounts (reasonable setup), startup fails with a confusing OOF error. No auth is needed in DeployOnly mode, so a submission account could send the CREATE2 deploy instead. Non-blocking.

Good point. In DeployOnly mode the deploy tx can be sent from a submission account, so I changed it to use the first submission account when available.

squadgazzz · 2026-05-25T20:15:36Z

 }

+impl Config {
+    fn validate(&self) -> Result<()> {


The submission-accounts-must-be-signers check lives in Config::validate, but the main-account-must-be-signer check is over in eip7702::setup. Both are the same kind of EIP-7702 precondition. Moving the main-account check into validate would catch misconfig at construction. Minor.

Agreed. I moved the main-account-must-be-signer check into Config::validate, next to the submission account signer validation.

kaze-cow

my feedback was addressed or superseded by other feedbacks. in other news, I was able to test with the playground and verify the wholistic change is working as expected, so looks good to me 👍

# Description This PR removes the old `CowSettlementForwarder` implementation and its config surface. The forwarder is being replaced by `Solver7702Delegate` in the follow-up PR. Removing the old contract, artifact, bindings, and explicit `forwarder-contract` config first keeps the stack easier to review: this PR is only the cleanup of the old path, while cowprotocol#4430 introduces the new delegate behavior. # Changes * Remove `CowSettlementForwarder.sol` and its artifact. * Remove generated `cowsettlementforwarder` bindings and facade wiring. * Remove `forwarder_contract` / `forwarder-contract` from driver config loading and solver config. * Remove e2e setup fields and test code that deployed or passed a forwarder address. * Keep the replacement `Solver7702Delegate` implementation for the follow-up PR. ## How to test This PR is intended to be validated together with cowprotocol#4430, which adds the replacement implementation. 1. `just generate-contracts` 2. `just fmt --check` 3. `just clippy` 4. `just test-unit`

igorroncevic · 2026-05-26T10:53:00Z

Thinking that just generate-contracts might not be handling Foundry artifacts properly.

It says that deployedBytecode isn't present, but it's clearly available in the out/Solver7702Delegate.sol/Solver7702Delegate.json but under deployedBytecode.object.

squadgazzz

LGTM

igorroncevic requested a review from a team as a code owner May 22, 2026 11:21

gemini-code-assist Bot reviewed May 22, 2026

View reviewed changes

igorroncevic added 2 commits May 22, 2026 13:30

feat: remove old settlement forwarder

e94152c

feat: add solver 7702 delegate

fdcb176

igorroncevic force-pushed the feat/solver-7702-delegate branch from 8af659d to fdcb176 Compare May 22, 2026 11:32

igorroncevic force-pushed the feat/remove_old_forwarder_contract branch from 4dc8037 to e94152c Compare May 22, 2026 11:32

igorroncevic mentioned this pull request May 22, 2026

feat: remove old settlement forwarder #4429

Merged

kaze-cow requested changes May 22, 2026

View reviewed changes

kaze-cow reviewed May 22, 2026

View reviewed changes

Comment thread contracts/src/vendor.rs Outdated

igorroncevic added 4 commits May 22, 2026 16:07

fix: CI

cb275c3

fix CI

c3ed392

fix toml formatting

c207776

address pr feedback

f11cd9d

AryanGodara reviewed May 25, 2026

View reviewed changes

Comment thread crates/driver/src/infra/solver/eip7702.rs Outdated

Comment thread crates/driver/src/infra/solver/eip7702.rs

Base automatically changed from feat/remove_old_forwarder_contract to main May 25, 2026 14:56

igorroncevic added 3 commits May 25, 2026 16:59

pr feedback

79e3a46

Merge remote-tracking branch 'origin/main' into feat/solver-7702-dele…

a7c24df

…gate # Conflicts: # contracts/generated/contracts-generated/solver7702delegate/Cargo.toml # contracts/src/main.rs # crates/driver/src/domain/mempools.rs # crates/e2e/tests/e2e/parallel_settlement.rs

fix ci toml formatting

afef2b7

jmg-duarte reviewed May 25, 2026

View reviewed changes

squadgazzz reviewed May 25, 2026

View reviewed changes

kaze-cow approved these changes May 26, 2026

View reviewed changes

pr feedback

87285fd

fix generated contracts

b19a4b3

squadgazzz approved these changes May 26, 2026

View reviewed changes

	.all(\|account\| !matches!(account, Account::Address(_))),
	.any(\|account\| matches!(account, Account::Address(_))),

		/// Ensure EIP-7702 delegate deployment and solver delegation are set up for all
		/// solvers with parallel submission accounts. Called once at driver startup.

-fn delegation_status(code: &[u8]) -> DelegationStatus {
+impl DelegationStatus {
+	fn from_code(code: &[u8]) -> Self {
+	...
+	}
+}

Conversation

igorroncevic commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Changes

How to test

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

kaze-cow left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AryanGodara left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

jmg-duarte left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

igorroncevic commented May 22, 2026 •

edited

Loading