feat: add Machima DEX adapter for KyberSwap onchain execution

[0xvsr]

Why did we need it?

On-chain Solidity adapter enabling KyberSwap's router to execute swaps through Machima DEX on Base. Machima is a Uniswap V3 fork with a custom tax layer and authorization system.

The adapter:

• Receives tokenIn from KyberSwap's router
• Approves and calls MachimaAggregatorRouter.swap()
• Handles amountUnused for partial fills (XMA sells that hit the price floor)
• Returns unused tokens to the caller

Deployed contracts (Base):

Contract	Address
MachimaAggregatorRouter	0x566250347E1401615B3e043918fc290B98448578

Companion PR: MachimaLabs/kyberswap-dex-lib (off-chain liquidity source)

Related Issue

N/A — new integration.

Release Note

• New adapter: MachimaAdapter
• Targets MachimaAggregatorRouter which wraps the protocol's MachimaSwapAdapter with a standard aggregator interface
• No special deployment requirements — the adapter is stateless and uses the router address passed via calldata

How Has This Been Tested?

• Foundry fork tests against Base mainnet (MachimaAdapter.t.sol)
  • Buy path (WETH → XMA): verifies output > 0 and token delivery
  • Sell path (XMA → WETH): verifies output > 0 and token delivery
  • Invalid pair: verifies revert
  • Expired deadline: verifies revert
• forge build clean

Screenshots (if appropriate):

N/A

[minhtr09]

Hi @0xvsr, please allow edits by maintainers.

[0xvsr]

Thanks @minhtr09 — enabled 'allow edits by maintainers', and fixed the failing CI (it was forge fmt --check; reformatted the adapter files to the repo's fmt config). Should be green now.

[minhtr09]

Hi @0xvsr, could you check your tests as well? I saw some failing tests.

[0xvsr]

@minhtr09 Fixed the failing tests — two things were wrong:

1. The fork test pinned an outdated router deployment. Updated to AggregatorRouter v1.1.0 (0xa25D1158B7Cf373DC3787793A52933dB0A0CaD89, verified on Basescan). Same swap() signature; notable improvement for this adapter: on partial fills at the XMA sell floor, v1.1.0 refunds the unconsumed input to the recipient (the adapter), which is exactly what the amountUnused return value expects.
2. test_sellXmaForWeth assumed sells always fill, but the XMA pool has an on-chain sell price floor — when the live price is pinned at the floor (as it is at some blocks), every sell reverts with SwapFailed(). The test now accepts that specific revert as valid (asserting the selector), and asserts the normal fill path whenever the floor isn't binding at the forked block.

All 6 tests pass locally against a live Base fork. Note the suite forks https://mainnet.base.org, so CI results can vary with public RPC rate limits — happy to gate the fork tests behind an env var if you'd prefer that for CI stability.

[0xvsr]

Hey @minhtr09 - thanks for the reviews so far, really appreciate the back-and-forth. Just checking in - the test fixes from my last push (3aeae05) address the failing tests you flagged:

• Router updated to v1.1.0 (verified on Basescan)
• XMA sell test now correctly handles the floor-pinned state
• All 6 tests pass on a live Base fork

Happy to gate the fork tests behind an env var for CI stability if that's preferred, just let me know.

Is there anything else blocking approval? The companion off-chain PR (kyberswap-dex-lib#1483) is also ready, so once both land KyberSwap can start routing through Machima pools immediately.

[minhtr09]

Hi @0xvsr, the CI still failed. Please make sure you use Foundry 1.7.1 and run forge fmt. Also, for the RPC, please read it from the env variable RPC_8453

[0xvsr]

Hey @minhtr09 — thanks for the quick feedback, really appreciate how responsive you've been throughout this.

Both issues fixed in 34b1e6d:

1. RPC: Tests now read from RPC_8453 env variable instead of hardcoding
2. Formatting: Ran forge fmt with Foundry 1.7.1 — reformatted MachimaAdapter.sol

All 6 tests pass locally on a live Base fork. CI should be green now. Let me know if anything else is needed!