chore: merge main into dev-2.3.0, keeping the 2.3.0 work intact

[dianakocsis]

TL;DR for the reviewer

This merge changes no contract code. The diff is 34 files: 24 deploy-address records and 10 deploy scripts, all from #516 and #517, each byte-identical to origin/main. If you see anything under contracts/, test/ or snapshots/ in this diff, something is wrong — there should be nothing.

That is the entire review. The rest of this description explains why a merge this boring needed a hand-written resolution.

The problem

main's aebfa39 (#514) reverted #491 (nested V4_SWAP) and #497 (bounded command input decoding). dev-2.3.0 forked at fb25ff0, before that revert, so it still contains both.

That matters because #502 depends on #491. #502 is +10 lines / −0 in Dispatcher.sol — a guard that sits inside the if (poolManager.isUnlocked()) block #491 created:

if (poolManager.isUnlocked()) {              // <- #491
    if (!NestedUnlock.isPermitted())         // <- #502
        revert NestedExecutionNotPermitted();
    ...
}

Remove #491 and there is no branch for #502 to guard. They are one unit: #491 added the feature and carried audit finding M-01; #502 fixes M-01 by replacing chain-state inference with an explicit opt-in. main took the other remedy for M-01 — delete the feature outright — which is correct for main, since main tracks the deployed 2.1.2 router. 2.3.0 takes the keep-and-gate path.

Why a plain merge is unsafe

A default merge re-applies the revert. Nine files conflict:

contracts/base/Dispatcher.sol
lib/v4-periphery
snapshots/UniversalRouterTest.json
test/foundry-tests/V4SwapWithinUnlock.t.sol          (modify/delete)
test/integration-tests/gas-tests/__snapshots__/*.snap  (5 files)

Those are the visible half. The dangerous half is silent: git cleanly auto-merges contracts/libraries/Commands.sol and contracts/modules/Payments.sol, deleting command 0x0f (UNWRAP_WETH_EXACT / unwrapWETH9Exact) with no conflict marker. Resolving only the nine conflicts and committing would ship that loss invisibly.

Worse, resolving the nine "sensibly" — taking main's side, since main is the release branch — yields #491: 0, #502: 0, #497: 0. All three gone, including the audited work.

Resolution applied

keep dev-2.3.0:  contracts/  test/  snapshots/  README.md  lib/v4-periphery
take main:       deploy-addresses/  script/

Deliberately not git revert aebfa39 — that would restore #491 and #497 but not #502, which was authored on top of them.

Verification

check	result
contracts/, test/, snapshots/, README.md, submodule pin vs dev-2.3.0 (8b8595a)	byte-identical
incoming files	34, all deploy-addresses/ (24) or script/ (10)
each incoming file vs origin/main (b67f125)	byte-identical
#491 nested branch / #502 guard / #497 0x0f / #497 checkInputLength	present (1 / 1 / 1 / 17 sites)
RouterParameters.sol (the struct the new scripts construct)	identical on main and dev-2.3.0, so the scripts compile against these contracts

Merge parents: 8b8595a (dev-2.3.0) and b67f125 (main, including #517).

Follow-ups

lib/v4-periphery stays pinned at 0cc6e164. The bump to the restored-#584 periphery stack is a separate commit once Uniswap/v4-periphery#604 lands.

Once this merges, #512's diff stops understating the change. GitHub was diffing #512 against merge-base fb25ff0, which predates the revert, so it rendered as 16 files / +687 −71. What would actually land on main is 59 files / +1854 −239. Anyone auditing #512 before now was reviewing roughly a third of it.

This also makes dev-2.3.0 a true superset of main, so future main→dev merges are conflict-free.

Moving dev-2.3.0 makes #507 stale; the restack (#507 → #508 → #509, with gas snapshots regenerated at each level) follows.

🤖 Generated with Claude Code

AI-Generated Description

TL;DR for the reviewer

This merge changes no contract code. The diff is 34 files: 24 deploy-address records and 10 deploy scripts, all from #516 and #517, each byte-identical to origin/main. If you see anything under contracts/, test/ or snapshots/ in this diff, something is wrong — there should be nothing.
That is the entire review. The rest of this description explains why a merge this boring needed a hand-written resolution.

The problem

main's aebfa39 (#514) reverted #491 (nested V4_SWAP) and #497 (bounded command input decoding). dev-2.3.0 forked at fb25ff0, before that revert, so it still contains both.
That matters because #502 depends on #491. #502 is +10 lines / −0 in Dispatcher.sol — a guard that sits inside the if (poolManager.isUnlocked()) block #491 created:

if (poolManager.isUnlocked()) {              // <- #491
    if (!NestedUnlock.isPermitted())         // <- #502
        revert NestedExecutionNotPermitted();
    ...
}

Remove #491 and there is no branch for #502 to guard. They are one unit: #491 added the feature and carried audit finding M-01; #502 fixes M-01 by replacing chain-state inference with an explicit opt-in. main took the other remedy for M-01 — delete the feature outright — which is correct for main, since main tracks the deployed 2.1.2 router. 2.3.0 takes the keep-and-gate path.

Why a plain merge is unsafe

A default merge re-applies the revert. Nine files conflict:

contracts/base/Dispatcher.sol
lib/v4-periphery
snapshots/UniversalRouterTest.json
test/foundry-tests/V4SwapWithinUnlock.t.sol          (modify/delete)
test/integration-tests/gas-tests/__snapshots__/*.snap  (5 files)

Those are the visible half. The dangerous half is silent: git cleanly auto-merges contracts/libraries/Commands.sol and contracts/modules/Payments.sol, deleting command 0x0f (UNWRAP_WETH_EXACT / unwrapWETH9Exact) with no conflict marker. Resolving only the nine conflicts and committing would ship that loss invisibly.
Worse, resolving the nine "sensibly" — taking main's side, since main is the release branch — yields #491: 0, #502: 0, #497: 0. All three gone, including the audited work.

Resolution applied

keep dev-2.3.0:  contracts/  test/  snapshots/  README.md  lib/v4-periphery
take main:       deploy-addresses/  script/

Deliberately not git revert aebfa39 — that would restore #491 and #497 but not #502, which was authored on top of them.

Changes

• deploy-addresses/ (24 files): Record UniversalRouterV2_1_2 addresses across all networks, add new chain files (arc.json, megaeth.json, robinhood.json)
• script/deployParameters/ (10 files): Add deploy scripts for Arc, HyperEVM, MegaETH, Robinhood; update BaseSepolia, Ink, Unichain, UnichainSepolia, Worldchain, Zora with live addresses

Verification

check	result
contracts/, test/, snapshots/, README.md, submodule pin vs dev-2.3.0 (8b8595a)	byte-identical
incoming files	34, all deploy-addresses/ (24) or script/ (10)
each incoming file vs origin/main (b67f125)	byte-identical
#491 nested branch / #502 guard / #497 0x0f / #497 checkInputLength	present (1 / 1 / 1 / 17 sites)
RouterParameters.sol (the struct the new scripts construct)	identical on main and dev-2.3.0, so the scripts compile against these contracts
Merge parents: 8b8595a (dev-2.3.0) and b67f125 (main, including #517).

Follow-ups

lib/v4-periphery stays pinned at 0cc6e164. The bump to the restored-#584 periphery stack is a separate commit once Uniswap/v4-periphery#604 lands.
Once this merges, #512's diff stops understating the change. GitHub was diffing #512 against merge-base fb25ff0, which predates the revert, so it rendered as 16 files / +687 −71. What would actually land on main is 59 files / +1854 −239. Anyone auditing #512 before now was reviewing roughly a third of it.
This also makes dev-2.3.0 a true superset of main, so future main→dev merges are conflict-free.
Moving dev-2.3.0 makes #507 stale; the restack (#507 → #508 → #509, with gas snapshots regenerated at each level) follows.