Bring the Anchor v1 copies in line with v2 where the Anchor version allows

[mikemaccana]

Summary

CONTRIBUTING.md now says the anchor-v1/ copies differ from their v2 counterparts only where the Anchor version forces it (#156). This PR applies that rule to one known case, the interest-bearing helper, and to everything a scan of all v1/v2 pairs turned up.

Changes

Interest-bearing (first commit). v2 split check_mint_data into check_rate_authority, which takes the parsed InterestBearingConfig. v2 needed that because of its exclusive borrows (44f7c35); only the parsing route is forced. The v1 copy now has the same helper, name and signature, and its handlers parse the extension with anchor-spl's get_mint_extension_data.

Drift scan fixes (second commit).

• Transfer-hook counter and account-data-as-seed (a real bug). The v1 copies still had the bug v2 fixed in 14847e8. The hook added one to the stored count, logged it, and dropped it, because counter_account wasn't mut. The count read one after every transfer. This ports the fix:
  • the mut constraint and the write-back
  • a new CounterOverflow error for the checked add, which previously reused AmountTooBig
  • a test assertion that reads the stored count after the hooked transfer
• Account field visibility. Fields that v2 declares pub are now pub in the v1 copies of account-data, checking-accounts, program-derived-addresses, rent, transfer-sol and nft-operations. The other v1 examples already did this.
• Order book. OrderTreeRoot derives Default as in v2, replacing a hand-written impl that clippy flagged as derivable.

How the scan worked

1. For every anchor/ + anchor-v1/ pair, I compared the sets of function names, #[error_code] variants, pub fields, constants and seed strings in the program source, plus the test names.
2. I listed every non-merge commit since July that changed a v2 program's source without touching its v1 copy. Then I checked the behavior-changing ones against v1: vault-strategy's allocation, MAX_ASSETS and ApprovedAsset, lending's index names, and nft-minter's update_authority_is_signer, plus the handler-adding commits. All of these were already in v1 or expressed through v1's own API.

The remaining differences are all forced by the Anchor version; the root changelog lists them:

• v2's transfer-hook entrypoint fallback
• a local session-token reader replacing the v1-only session-keys crate
• the last_restart_slot syscall wrapper (v1 does the same check through the sysvar)
• the wincode span in rent
• zero-copy padding
• the Pyth account traits
• the betting market's borrow-release helper
• v2's borrow-model fixes (44f7c35, eff9f03, 16d30e4)

Verification

• Every touched v1 copy and its tests compile with no errors.
• Clippy warning counts match main in each copy. The order book goes from 10 to 9, because the flagged impl is gone.
• Not run locally: the LiteSVM suites, which need the SBF build. This PR's Anchor v1 CI job runs them.

🤖 Generated with Claude Code

https://claude.ai/code/session_01F8Wo5zNLqZ2NQqSjGD7WpK

---

Generated by Claude Code