[mikemaccana]

Depends on quicknode/solana-program-examples#152. Merge that first. This PR names two tests that exist only on that branch, test_donation_cannot_round_a_later_deposit_to_zero and test_inflating_liquidity_through_own_trades_does_not_pay, and describes the behavior it ships.

Two first-deposit defenses in the companion repository did not work the way the book said:

• Token swap: it minted later deposits against the bare LP supply while withdrawals divided by supply plus the floor.
• Perpetual futures: it divided by the bare supply both ways, so the withheld 1,000 shares locked nothing. A provider who was also the pool's only trader could pay funding into liquidity and take part of the next deposit. Against the old program the attacker spent about 1,000 USDC and withdrew about 1,500.

The companion PR counts the minimum in every share divisor, and these chapters follow it.

• Constant-Product AMM:
  • Alice's deposit and withdrawal now name the same divisor.
  • The inflation-attack bullet now gives the attack's price: a donation of about 101 times the victim's deposit, with 100/101 of it lost to the floor. It also explains why the floor must be in both divisors and names the test that runs the attack.
  • The rules list and the What You Learnt list are updated to match.
• Perpetual Futures:
  • Carol's paragraph now says both handlers count the withheld shares.
  • The failure mode explains that pricing shares against liquidity stops a vault donation but not a provider's own funding payments, and gives the test's numbers.
• The Adversary: the "Ignore donations" defense now says a recorded pool size stops donations only while nothing else the attacker controls can raise it, with the perpetual futures pool as the example.

Rebased onto current main, where the Adversary section and Carol's paragraph had been rewritten. main's text is kept, with these additions merged into it. npm run lint, node scripts/check-figure-geometry.mjs and npm run build-pdf all pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_01HagqhpPNqcywT37qqgnFey