[SIDESWAP] Fix USDt to LBTC atomic swap - dealer's suggested recv_amount

[coelhogonzalo]

closes #66 for more details check the ticket

[coelhogonzalo]

I contacted the Blockstream team and asked if they could update the python binding to expose a Python-callable TxOut.unblind_with_ephemeral_sk(ephemeral_sk, address_blinding_pk)

Will update this PR when they reply.

[coelhogonzalo]

I contacted the Blockstream team and asked if they could update the python binding to expose a Python-callable TxOut.unblind_with_ephemeral_sk(ephemeral_sk, address_blinding_pk)

Will update this PR when they reply.

I fixed it by using wallycore directly.

[andycreed0x]

Good fix! Thanks for the PR

[andycreed0x]

You can make this testnet-compatible with a small change. As-is, this would break if you ever want to test swaps on testnet. From Claude's analysis:

addr_family is always hardcoded to "lq", but testnet addresses use tlq. So confidential_addr_segwit_to_ec_public_key(tlq1..., "lq") will throw a raw ValueError before ever reaching the PSET.

# In _verify_pset, derive and pass addr_family:
addr_family = "tlq" if network == "testnet" else "lq"

[andycreed0x]

If recv_amount < fee_tolerance_sats, min_recv_delta goes negative, meaning a dealer could deliver 0 of the receive asset and still pass verification.

# Clamp to zero so the tolerance can't exceed the agreed amount
min_recv_delta = max(0, recv_amount - fee_tolerance_sats)

[andycreed0x]

The whole function is mocked out in every swap manager test.
(Something I always ask myself when reviewing: "is this test actually exercising the real functions and hitting the DB, or is everything mocked?"... might be useful for building tests.)

According to Claude:
what's missing is a test that doesn't patch this function and feeds it a real PSET ideally one honest case and one with a tampered asset.

[marinate305]

Conflict in pyproject.toml appeared after PR #97 merged — likely a version bump conflict. Please rebase onto develop to resolve before merging.