v0.4.1

• Requires rasuvaeff/understudy ^0.9 || ^0.10. The engine's 0.10 closes six
  defects of the 1.0 review, the widest of them being that an optional
  parameter no longer has to be spelled in a specification: when(fn () => $repository->find(1)) now matches a contract whose remaining parameters
  carry defaults, and an unspelled position is rendered as … in reports so
  it is never mistaken for a written any(). Nothing this adapter does changes
  — the arity suppression it carries is about Arg::rest(), which still
  covers the contract's required parameters and only those.
• Drops the ^0.8 term, as 0.4.0 said it would in the release that
  follows the engine's.
• Moves the Pest fixture's path-repository pins with the release, as this
  package's own rule requires: the engine to 0.10.0, this package to
  0.4.1, and the fixture's requirement to ^0.4. A path repository is
  canonical, so a stale pin does not fall back to Packagist — left alone,
  ^0.8 would have kept the README's recipe resolvable and never once run
  it against 0.10.

v0.4.0

A minor rather than a patch: the trait's PHPUnit lifecycle hooks are all
@internal. Their protected visibility is required for PHPUnit discovery,
not an invitation to override them. A test class that owns post-conditions
must use the documented trait alias to compose them explicitly.

v0.3.0

A minor rather than a patch: two methods of the trait become @internal, and
roave/backward-compatibility-check reads that as the contract narrowing it is
— which Composer's caret already treats as breaking on 0.x.

• Allows rasuvaeff/understudy ^0.8 || ^0.9. A bridge, not a widening for
  its own sake: the engine's 0.9 is the 1.0 candidate — every contract decision
  of the 1.0 review lands there, and 1.0 follows once it has been driven by real
  packages — and a project taking it must be able to keep this adapter without a
  window in which composer require silently installs the 0.8 engine beside
  it. The ^0.8 term is dropped in the release that follows the engine's.
• The two hook methods of the trait, understudyPrepareContext() and
  understudyResetContext(), are @internal. They are protected because
  PHPUnit discovers hooks by attribute and a trait cannot hide them from the
  class using it, not as override points: a subclass that overrode one would
  silently lose the guard or the reset. understudyStrictStubs() stays the
  one documented override, and assertPostConditions() stays public API
  because the README's explicit-composition recipe calls it through an alias.
• Both READMEs stop dating lean() to «understudy 0.4+»: with a floor of
  ^0.8 every engine this adapter installs beside has it. llms.txt carries
  the family's # rasuvaeff/… heading.

v0.2.0

• Requires rasuvaeff/understudy ^0.8, and requires it as a single term. The
  accumulating union it carried (^0.4 || ^0.5 || …) had to be widened by hand
  on every core release, and a package that misses one becomes uninstallable
  beside its own engine.
• #[DoesNotPerformAssertions] works again. The trait counted an assertion
  unconditionally, so a test marked with that attribute reported "performed 1
  assertion" and went risky — on exactly the tests the attribute is written for
  ("this should simply not throw"), in a class that inherits the trait from a
  project-wide base class along with everything else. A test that created no
  double is now left alone entirely.
• The context guard no longer blames a test that never ran. A double
  created in setUpBeforeClass() fills the context before #[Before], and the
  message sent the reader looking for an earlier test that skipped cleanup, an
  unused trait and a swallowed assertPostConditions() — none of which was the
  case. It now names setUpBeforeClass() first.
• The Pest example runs. The second snippet of that section called when()
  without importing it — Call to undefined function when() — in a section
  whose whole subject is which functions to import under which names. Both
  READMEs are fixed; the package's own Pest fixture had it right all along.
• Both READMEs and llms.txt stop pointing at Understudy::strict($double) as
  the per-double form of strict stubs. It is strict dispatch — "fail on any
  call no expectation matched" — and says nothing about a stub that was
  configured and never called; the per-double equivalent is
  when(…)->times(n).
• Both READMEs and llms.txt say that verification runs before your
  teardown here and after it under understudy-testo, so a test whose
  expectation is fulfilled by teardown itself fails in one runner and passes in
  the other.

v0.1.7

• Documentation review fixes. Both READMEs gained the missing Security
  section; the API table no longer says «reset-in-finally» (the reset has
  always lived in #[After]). The trait's docblock now shows the explicit
  composition in the order the README mandates — the user's post-conditions
  first, verification second. The stale engine-constraint comment in
  examples/readme-usage.php caught up with ^0.7. AGENTS.md no longer links
  to the retired _plans/UNDERSTUDY-PLAN.md.

v0.1.6

• Allow rasuvaeff/understudy ^0.7. Widened rather than raised.
• The #[Before] guard is checked through real PHPUnit. It was covered only
  by a unit test calling runGuard() directly, which proves the condition and
  nothing about what a user sees — the guard fires from #[Before], so the
  failure lands on the BLAMELESS class, the one that ran after the leak, and
  that attribution is the whole diagnostic. A fixture now leaks a context from
  a class without the trait and pins the report, the attribution and the exit
  code.
• The Pest recipe runs in CI, and it was broken. The test that executes it
  skips itself unless the fixture project is installed, which was every run of
  every job — so the recipe had never actually been checked, and the fixture
  still stubbed find(7) with when() and claimed the same call with
  expect(): two registrations of one call, which the engine has refused with
  ConflictingExpectation since 0.3. The README's Pest section has been right
  all along; the fixture that was supposed to prove it had drifted, in the
  third place this same mistake has now been found. The job has a checkout of
  its own because the fixture's path repositories are written for the monorepo
  layout — this package and the engine side by side — which is deliberate:
  the recipe runs against the working tree of both.
• infection/infection moves to ^0.35, matching the monorepo templates.
  Re-measured: 12 of 12 mutants killed, unchanged.
• AGENTS.md said the unit tests avoid PHPUnit's assertion counter; they use
  it, by its new name. What actually guards the rename is the PHPUnit ^11.5
  matrix job, and the note says so now.

v0.1.5

• Allow rasuvaeff/understudy ^0.6. Widened rather than raised: the adapter
  works against every 0.x line it has ever supported, and consumers on the
  older ones should not be cut off from it.
• The shipped trait's docblock taught the usage this package documents as
  wrong. It stubbed find(7) with when() and claimed the same call with
  expect() — two registrations of one call, which the engine refuses — and
  it armed the expectation below the action, where it counts zero. llms.txt
  carried the same snippet. Both now show the corrected form the README and
  examples/readme-usage.php have used since 0.1.0, and both name the two
  engine rules that decide it. Fixes #13.
• DocumentedUsageTest pins the three copies of the Usage snippet to each
  other. The original mistake survived two releases because nothing ran the
  snippet; the fix reached the README and the runnable example and stopped
  there, at the two copies nothing executes — which a consumer reads in their
  IDE and an assistant reads as llms.txt.

v0.1.4

• The Requirements section of both READMEs said rasuvaeff/understudy
^0.1 || ^0.2 || ^0.3 while composer.json has allowed ^0.4 since 0.1.3,
  and the usage examples already use 0.4 idioms.
• Allow rasuvaeff/understudy ^0.5. The core release narrows what a
  closing scope() verifies and refuses two impossible matcher
  configurations — both changes to the consumer's own test code, neither
  reaching this adapter, which needs no code change.

v0.1.3

• Allow rasuvaeff/understudy ^0.4: Arg::rest(), Arg::captor(),
  Understudy::delegate(), Understudy::lean() and rendered property hooks
  are all additive — the adapter needs no code change.

• README (EN+RU): documented that the #[After] reset runs after
  tearDown() while the call log still retains returned values, the Windows
  "Directory not empty" failure that surfaces it, and the two remedies —
  Understudy::lean() (understudy 0.4+) and Understudy::scope()
  (rasuvaeff/understudy#63).

v0.1.2

• Allow rasuvaeff/understudy ^0.3 (the engine refuses colliding same-call
  when()/expect() registrations with ConflictingExpectation from 0.3.0;
  nothing in this adapter changes behaviour).