Opus finding
approxOut precision loss via Number() on 18-decimal wei amounts produces wrong amountOutMinimum (potential silent loss or revert)
highbughigh
- scripts/reposition.ts:313-322
- scripts/reposition.ts:246-248
swapAmountIn is in wei (1e18 scaling). Number(swapAmountIn) loses precision for any amount above ~9e15 (Number.MAX_SAFE_INTEGER = 2^53 ≈ 9.007e15), i.e. above ~0.009 ETH or ~0.009 DIEM. The conversion goes through float and back to BigInt, which (a) silently drops precision and (b) compounds with rounding of diemPerWeth. amountOutMin is then derived from this approximate value with only 3% slippage — meaning the slippage buffer may be smaller than the float quantization error. For larger LPs this either reverts the swap (overly aggressive min) or under-protects against MEV/price impact (if approxOut overshoots).
Recommendation
Compute approxOut in BigInt: derive a fixed-point price from sqrtPriceX96 (e.g., price1Per0Q96 = sqrtPriceX96^2 / 2^96) and do (amountIn * price) >> 96 in BigInt. Apply pool fee and slippage in BigInt only.