feat(37-13): encounter creation gate — every branch observable#451
Merged
feat(37-13): encounter creation gate — every branch observable#451
Conversation
Spec a to-be-extracted helper `dispatch::apply_confrontation_gate` and
`ConfrontationGateOutcome` enum. Tests cover the full Case A-F matrix:
- A: no current encounter -> Created (encounter.created)
- B: resolved current -> Created
- C: same-type redeclare -> Redeclared (encounter.redeclare_noop)
- D: diff type, beat == 0 -> ReplacedPreBeat (encounter.replaced_pre_beat)
+ actor re-population from snapshot + narrator NPCs
+ old per_actor_state dropped
- E: diff type, beat > 0 -> RejectedMidEncounter (ValidationWarning,
encounter.new_type_rejected_mid_encounter) — state untouched
- F: unknown type -> UnknownType (ValidationWarning,
encounter.creation_failed_unknown_type)
Plus a source-scanning wiring test that verifies dispatch/mod.rs calls
the helper and that the old inline silent-drop branch is gone.
RED: unresolved imports for apply_confrontation_gate +
ConfrontationGateOutcome. Dev extracts gate logic during GREEN.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Extract the inline gate block from dispatch_player_action into a new
dispatch::encounter_gate module with a pure helper that covers all six
cases and emits a distinct WatcherEvent for each:
A: None -> Created (encounter.created)
B: Some(resolved) -> Created (encounter.created)
C: Some(unresolved, same type) -> Redeclared (encounter.redeclare_noop)
D: Some(unresolved, diff, 0) -> ReplacedPreBeat
(encounter.replaced_pre_beat)
E: Some(unresolved, diff, >0) -> RejectedMidEncounter
(encounter.new_type_rejected_mid_encounter, ValidationWarning)
F: def missing -> UnknownType
(encounter.creation_failed_unknown_type, ValidationWarning)
The previous code silently dropped the new confrontation type whenever
an unresolved encounter already existed: no OTEL event, no warning, no
state transition. That was a CLAUDE.md "No Silent Fallbacks" violation
and the root cause for 37-12 (narrator never re-declares because state
never accepts a new type).
Case D (replace-before-any-beat-fired) preserves the invariant that
mid-encounter mechanical state is never clobbered. Case E surfaces
mid-encounter narrator divergence as ValidationWarning so the GM panel
can see the drift without corrupting beat/metric state.
All 9 TEA tests green. 53 lib + 412 integration tests pass. No clippy
warnings. The inline block at the call site shrank from 47 lines to 8.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Addresses all six HIGH findings from Colonel Potter's review plus the mediums on test coverage and doc accuracy. Rule violations - Add #[non_exhaustive] to ConfrontationGateOutcome (Rule #2). The encounter lifecycle is a growing domain; the enum will gain variants. - Fix `let _ = apply_confrontation_gate(...)` at the Case D per-actor leak test (Rule #6). Bind the outcome and assert ReplacedPreBeat so the leak check cannot pass through a silently-wrong variant. Telemetry-lock race - Extract TELEMETRY_LOCK + fresh_subscriber + drain_events into a new `test_support::telemetry` module shared by both 34-11 OTEL tests and 37-13 gate tests. Previously each module had its own lock and could drain each other's events under parallel test execution. Test rigor - Add `assert_source_is_narrator` helper and apply it to every canonical event test (Cases A, C, D, E, F). The production code sets source = "narrator_confrontation" on every branch; no test asserted it until now. - Tighten the wiring test: positive branch scans for `apply_confrontation_gate(` (with the opening paren) so a bare comment mention can't satisfy it; negative branch uses a semantic token pair (`is_none()` + `is_some_and(|e| e.resolved)`) instead of a whitespace- exact multiline match that cargo fmt could silently break. - Rewrite the test-file header: drop RED-phase language, drop the dead `dispatch/mod.rs:1773-1819` line reference, keep it under 20 lines. (CLAUDE.md: don't reference the current task, fix, or callers — those rot.) New regression guards - case_c_same_type_with_beat_zero_still_redeclare: locks match-arm ordering so a future reorder can't promote beat==0 above same-type. - case_f_empty_incoming_type_routes_to_unknown: empty incoming type must route to UnknownType, not silently no-op. - case_f_unknown_type_with_existing_encounter_preserves_state: Case F with a pre-existing encounter must leave the encounter untouched. - case_a_populates_both_player_and_npc_actors: exercises the player actor loop in build_encounter that every other test left untested. Gate code hygiene - Make the Case E match arm's `beat > 0` guard explicit so the code matches the module doc table. Add an unreachable! catch-all covering the compiler-required fallback (guarded arms aren't exhaustive). - Document `apply_confrontation_gate`'s always-emits-one-event contract and the `narrator_npcs` parameter semantics on the function doc. Call site - Bind the gate outcome at dispatch/mod.rs with `let _gate_outcome =` and a comment documenting the intentional discard and the future hook point for surfacing RejectedMidEncounter to the session layer. Tests: 57 lib (up from 53) + 412 integration. Clippy clean, fmt clean. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Addresses the three HIGH and two MEDIUM findings from Colonel Potter's pass-2 review. All are documentation or test-coverage fixes; no logic changes to the gate itself. HIGH: #[non_exhaustive] lying docstring - Rewrote the doc comment on ConfrontationGateOutcome to be honest about intra-crate behavior. The attribute has no compiler bite for pub(crate) consumers inside the defining crate; downstream wildcard arms are a convention, not a compiler guarantee. The new doc says so explicitly. HIGH: Case B missing assert_source_is_narrator - Added the assertion to case_b_resolved_current_encounter_creates_new so every canonical event test verifies source=narrator_confrontation. HIGH: _gate_outcome aspirational comment rot - Dropped the "A future story can match on _gate_outcome to surface RejectedMidEncounter to the session layer" sentence from dispatch/ mod.rs. Replaced with a terse mention that the binding is a named placeholder — same information, no future-tense aspiration, no unowned TODO. MEDIUM: Case F existing-encounter missing actor preservation - Extended case_f_unknown_type_with_existing_encounter_preserves_state to assert resolved flag, actors.len(), first actor name, and per_actor_state are preserved — plus assert_source_is_narrator on the captured event. MEDIUM: Wiring test AND-logic gap - Split the negative-side check from `!(a && b)` to two independent `!a` / `!b` assertions, closing the hole where a partial regression reintroducing just one of the two tokens would slip past. Tests: 57 lib + 412 integration still green. Clippy clean. Fmt clean. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
dispatch_player_actioninto a newdispatch::encounter_gatemodule with a pureapply_confrontation_gatehelper.context-story-37-13.md— every branch emits a distinctWatcherEventon theencounterchannel so the GM panel can see every gate decision.Case matrix
Scope
dispatch/encounter_gate.rs(new, 198 lines),dispatch/mod.rs(call site replacement +pub(crate) modregistration),lib.rs(test module registration).encounter_gate_story_37_13_tests.rs(717 lines, 13 tests covering A-F + wiring + 4 regression guards),test_support.rs(new shared telemetry-lock module),otel_dice_spans_34_11_tests.rs(refactored to use shared lock).apply_beat,find_confrontation_def, beat dispatch, genre packs, narrator prompt.Test plan
cargo clippy -- -D warningscleancargo fmt --checkcleansource = "narrator_confrontation"asserted on every canonical event testdispatch_player_actionand the old inline branch is goneReview history
🤖 Generated with Claude Code