There are two tests in this challenge. Both deal with the absolute difference between two random positive integer parameters.
Test 1 ("difference must not be zero") only succeeds if
- the first parameter is less than 10
- or the difference is not zero.
The smallest falsified sample is
[10, 10]
Test 2 ("difference must not be small") only succeeds if
- the first parameter is less than 10
- or the difference is not between 1 and 4.
The smallest falsified sample is
[10, 6].
Test 3 ("difference must not be one") only succeeds if
- the first parameter is less than 10
- or the difference is not exactly 1.
The smallest falsified sample is
[10, 9].
Shrinking is a challenge in these examples because it requires keeping up a dependency between two distinct parameters. Additionally, it can be a challenge to find a first failing sample when generation of integers is naively done uniformly across the realm of positive integers.
Test 3 seems the most difficult one to shrink because shrinking parameters individually will never lead to a smaller and falsifying sample. This is also the hardest to find a falsifying sample in the first place.