clay-borg/crates
tegwick 48a7bce04d
Some checks failed
ci / check (push) Has been cancelled
CB-WP-0048 T02/T03/T04: Configuration replaces Variant in the state
The three-armed enum could not express a configuration carrying two
modules. It does now: --variant h1 expands to problem_stress.flat_any_open
AND attack_relief.self_soothe_ge4, and scoped_plus_attack_soothe plays.

Legacy names alias forever via serde(alias="variant") plus a
scalar-or-map deserialiser. serde(default) alone would have been a silent
migration bug -- every H1/H2 recording would have come back as baseline.
All 26 scenarios pass unchanged; none pins a state hash.

Three things shipped broken first, all caught by gates rather than by
reading:

1. `profile` was in the state hash. with_config(select("ground-darvo-r0"))
   sets profile=Some("baseline") where setup alone leaves None, so two
   states at THE SAME POINT IN ASPECT SPACE hashed differently and a
   replay bundle stopped reproducing its own initial state. Now
   serde(skip): a hash covers what determines play. This was the open
   judgement from T01 and it did not survive contact with the replay path.

2. A YAML parse inside the event loop. rules() -> resolve() -> catalog()
   re-parsed catalog.yaml per rule check; AM-6 fell to 9,345 events/s
   against a 100,000 target. OnceLock, and aspect validation moved to
   where a configuration is BUILT.

   Then I nearly optimised a phantom: 470k still looked like a 3x
   regression against the "~1.7M on bnt-lap001" reference in the gate's
   own message. Making rules() free measured 491k -- this machine's
   ceiling. Before optimising against a reference, measure the ceiling
   with the suspect code removed.

3. The refusal did not fire on the path a player takes.
   `--module problem_deal.pressure_deck` played a full baseline game and
   reported success, because with_config is a builder and fell back to
   the printed rules -- the silent no-op ADR-0022 exists to refuse. Every
   unit test of resolve() passed. The helper was tested and the driver
   was not, which is CB-WP-0033's finding verbatim. The new test asserts
   refusal BY NAME and that no game was played.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-09 00:54:34 +02:00
..
cb-events CB-WP-0006 T05: K9's assertion, K11's format, and the AM-11 suites 2026-08-01 10:50:52 +02:00
cb-game-runtime CB-WP-0026: collect the rulings -- ten answers that arrived and were never applied 2026-08-05 16:13:37 +02:00
cb-kernel CB-WP-0006 T05: K9's assertion, K11's format, and the AM-11 suites 2026-08-01 10:50:52 +02:00
cb-render-html CB-WP-0048 T02/T03/T04: Configuration replaces Variant in the state 2026-08-09 00:54:34 +02:00