feat(buy-sol) PR 1: route native SOL buys to solvers as wSOL

[squadgazzz]

Description

Program v0.4.1 lets an order buy native SOL. The intent puts the System Program ID where the buy mint goes. Solvers only swap into token accounts, so a native SOL buy passed through as-is gives them a buy token they can't route.

So the driver hands solvers a wSOL buy instead, with the taker's wSOL ATA as the destination. The output can't go to the wSOL buffer PDA, because FinalizeSettle pays a native buy out of the state PDA and the program can't unwrap the buffer into it. Unwrapping the wSOL and funding the state PDA at settle time is BE-330.

Native SOL buys still can't reach /solve, since the orderbook rejects them at creation and the autopilot cut drops orders whose payout account isn't a token account. So only quotes change for now.

Changes

• Native SOL buys go to solvers with wSOL as buyMint and the taker's wSOL ATA as buyDestination
• /quote rejects selling wSOL for native SOL with QuoteSameTokens, the same rule EVM applies to WETH and ETH in is_same_buy_and_sell_token

How to test

New unit and API tests.

Related issues

BE-329

[linear-code]

BE-329

[claude]

Claude finished @squadgazzz's task in 3m 8s —— View job

---

Review

• Gather context (diff, related code)
• Review changes
• Post findings

I traced the full path and this change is clean and well-scoped. No blocking issues found — I'm not posting inline comments because I didn't find anything that should lead to a change.

What I verified

• The wSOL substitution lives only in the solver-facing DTO (Order::new) and quote flow. The settle path (domain/settlement.rs) builds intents from the untouched domain order, so the native-SOL→wSOL mapping never leaks into on-chain settlement — consistent with the PR scoping this to quotes/solve only.
• The quote path goes quote → compute_solutions → solver.solve → Auction::new → Order::new(taker = self.pubkey(), …), so a native-SOL buy is presented to the engine as a wSOL buy and priced by the wSOL mint. The new native_sol_buy_quote_is_priced_as_wsol test exercises exactly this.
• All Order::new call sites are updated for the new taker parameter; buys_native_sol() compares the buy mint against ENCODED_NATIVE_SOL_TRANSFER, and native_mint/associated_token_address are used correctly (wSOL is a classic SPL token, so the token-2022 caveat on associated_token_address doesn't apply here).
• /quote rejecting sell wSOL → buy native SOL as QuoteSameTokens matches the stated EVM parity. The wrap direction isn't guarded, but native SOL can't be a sell token, so that's moot rather than a gap.

Non-blocking notes (already covered by the PR / BE-330, no action needed)

• The native-SOL buy destination is the taker's (solver's) wSOL ATA rather than the order owner's account. That's fine for quotes (never settled) and the PR calls out that native buys can't reach /solve today (orderbook + autopilot cut). Worth keeping front-of-mind for BE-330 so the settle work routes the payout to the user rather than inheriting a solver-owned destination.

Overall: LGTM. Good comments and test coverage.
· branch solana-driver/be-329-native-sol-buys-as-wsol

[tilacog]

I foresee issues when creating setup accounts (solver/user ATAs and buffer PDAs) w/ this new native token denomination ENCODED_NATIVE_SOL_TRANSFER.

We can merge as is, and I'll provide a follow up PR.
I see #4986 already addresses these concerns.