The repeat-loop variant of testing in the previous section is an interesting one. It's not, like QuickCheck, a test wholly focused on finding logic bugs. We know that the test will work for most iterations. What's being sought out there is a failure to properly control the indeterminism of the underlying machine. In some sense that variant of test is using a logical check to search for crashes. It is a kind of fuzz test.
Hopper interacts with the filesystem, spawns threads, and manipulates bytes up from files. Each of these activities is subject to failure for want of resources, offsetting errors, or fundamental misunderstandings of the medium. An early version of hopper, for instance, assumed that small atomic ...
