The spy that counted the wrong calls

The trap people carry over from Mockery most often, and the one that leaves a green suite rather than a confusing failure.

The test

$repository = Understudy::for(BookRepository::class);
$expected = new Book('Dune');

expect(fn () => $repository->save($expected));

// The subject also saves something the test never asked about.
$repository->save($expected);
$repository->save(new Book('Neuromancer'));

Understudy::verifyAll();

This passes.

What it reports

Adding one line — Understudy::nothingElse($repository) — turns it into:

verifyAll() alone: passed

Understudy `BookRepository` received 1 call(s) nothing accounted for:
    save(Book#1 {title: 'Neuromancer'})

Why the first check passed

expect() counts the calls matching its own arguments. It claimed save($expected), it got exactly one, and it is satisfied. It never said anything about the rest of the traffic — nor should it, because a test that stubs three collaborators and exercises two of them is not asserting on the third.

A Mockery spy that counted every call to save did make that wider claim. The migration has to say it out loud.

The fix

expect(fn () => $repository->save($expected));
Understudy::nothingElse($repository);        // ← the wider claim

Or both in one:

Understudy::allVerified($repository);

nothingElse() takes any number of doubles, so one line closes out a whole test — and it names every offender rather than stopping at the first.

What counts as accounted for

An expect() that matched, or a successful verify(). A when() stub accounts for nothing: it is permission, not a description of what happened. See Has everything been described?.