Compute Magic Amounts failure

[gakonst]

Running forge t --optimize-runs 800 -vvvv --fork-block-number 13806816 while I was testing in foundry-rs/foundry#821, I got a failing test case for the test_computeMagicAmounts test, courtesy of our recently updated fuzzer logic (foundry-rs/foundry#735), which otherwise passes with dapptools:

Failed tests:
[FAIL. Counterexample: calldata=0x1e0eabea0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000c9, args=[0, 0, 201]] test_computeMagicAmounts(uint256,uint256,uint24) (runs: 31, μ: 2428, ~: 1417)

  function test_computeMagicAmountsRegression() public {
        uint256 inventory0 = 0;
        uint256 inventory1 = 0;
        uint24 halfWidth = 201;
        if (int24(halfWidth) < blend.MIN_WIDTH() / 2) return;
        if (int24(halfWidth) > blend.MAX_WIDTH() / 2) return;

        (uint256 amount0, uint256 amount1) = blend.computeMagicAmounts(inventory0, inventory1, int24(halfWidth));

        assertLt(amount0, inventory0);
        assertLt(amount1, inventory1);
    }

[adamegyed]

Thanks for pointing this out @gakonst! After investigating, we've found the issue to be specifically from the zero inventory case. To address this, we'll adjust the assertions to be <= rather than just <, which is a more accurate requirement for all cases in the test.