fix(stage-state): v0.18.2b — symmetric agent_role via in-class resolve_role seam

QA REJECT-P1 remediation (seq 29). start()/done()/is_done()/get_output() were asymmetric: start() passed the mapped role but done()/is_done()/ get_output() passed '' via the default, addressing a different unique key (run_id, stage, agent_role, item_key). On every normal run this orphaned the start() row in 'running'; on resume is_done() always returned false → duplicate-post vector. Fix: add resolve_role(stage, agent_role) private static seam in Run_Stage_State. When the caller passes '' the role is derived from Stage_Display_Map::default_agent_role(stage) before any DB operation. Non-empty roles (Phase-2 fan-out) pass through unchanged. All four write/lookup methods (start, done, fail, get) call resolve_role. is_done() and get_output() go through get() so they benefit too. Mid-run-upgrade compat (v0.18.1 → v0.18.2): get() falls back to a legacy role='' row on miss so a run started under v0.18.1 (all rows have agent_role='') can still resume under v0.18.2. One extra indexed query, only on miss, only for known stages. Tests: two new RunStageStateTest cases — test_role_tagged_start_and_done_are_symmetric: start+done with 'editor', assert is_done('editor')=true and is_done('')=false. test_legacy_empty_role_row_is_found_on_fallback: resolve to 'editor', primary miss, legacy role='' hit → is_done=true. ARCHITECTURE.md §22: agent_role column and #22 Phase 1 description updated — no longer 'Phase-2 only / '' default'; now documents Phase 1 population from Stage_Display_Map. Agent-Session: prab-v0182-maintenance-20260612
2026-06-12 02:01:01 +00:00
parent 3424af137e
commit 3e257c1855
4 changed files with 163 additions and 91 deletions
--- a/ARCHITECTURE.md
+++ b/ARCHITECTURE.md
@@ -536,7 +536,7 @@ dropped on uninstall.
 | id          | BIGINT UNSIGNED | Auto-increment PK                                  |
 | run_id      | VARCHAR(36)     | Pipeline run UUID                                  |
 | stage       | VARCHAR(50)     | Stage name (see stage vocabulary)                  |
-| agent_role  | VARCHAR(50)     | Fan-out dimension for Phase-2 quorum ('' default)  |
+| agent_role  | VARCHAR(50)     | Fan-out dimension. Populated in Phase 1 from `Stage_Display_Map::default_agent_role()` (e.g. 'writer', 'editor', 'publisher'). Phase-2 quorum passes multiple roles for the same stage. |
 | item_key    | VARCHAR(64)     | Scopes article-level stages within multi-article runs ('' for run-level) |
 | status      | VARCHAR(20)     | pending / running / done / failed / halted         |
 | attempt     | SMALLINT UNSIGNED | Re-entry counter                                 |
@@ -831,8 +831,9 @@ multi-step publish paths. Four subsystems:
   keyed by `_prautoblogger_run_id` + `_prautoblogger_idea_hash` (check-before-insert) so a
   retried run cannot duplicate a post. `PRAutoBlogger_Run_Reaper` (same daily cron as #19)
   marks runs stuck `running` > 2× expected stage wall-clock as `failed`; such runs render
-   as **"incomplete"** in audit queries. Quorum logic itself is Phase 2 — the schema just
-   reserves its dimensions.
+   as **"incomplete"** in audit queries. Quorum logic (multiple roles per stage) itself is Phase 2. In Phase 1,
+   `agent_role` is populated on every row from `Stage_Display_Map::default_agent_role()`
+   so the run timeline is self-describing and the idempotency key is fully populated.

 ### #20: PHPUnit test infrastructure + WordPress-Core PHPCS compliance (v0.10.1)

--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -21,12 +21,17 @@ Four confirmed-live maintenance fixes (CTO thread `2026-06-pipeline-v2-phase1` s
  HTTP status parsed from the exception message, and the first 200 chars of the error body —
  so a 27-day outage like the seq-22 llm_research 404 incident is diagnosable after the fact.
  No behaviour change; logging only.
- **`run_stages.agent_role` back-write (P3)**: `Pipeline_Runner`, `Article_Worker`, and
-  `Content_Generator` were calling `Run_Stage_State::start()` with `agent_role = ''`. All four
-  stage-start call sites now pass `Stage_Display_Map::default_agent_role( $stage )` so the run
-  timeline is self-describing. Role is written at INSERT only (the `start()` call); the
-  `ON DUPLICATE KEY UPDATE` path does not include `agent_role`, preserving the
-  `run_id+stage+agent_role+item_key` idempotency-key semantics on resume.
+- **`run_stages.agent_role` symmetric write (P3)**: `Pipeline_Runner`, `Article_Worker`, and
+  `Content_Generator` were calling `start()` with the mapped role but leaving `done()`,
+  `is_done()`, and `get_output()` with `agent_role = ''`. With `agent_role` in the UNIQUE key,
+  the mismatch addressed different rows — stranding `start()` rows in `running` forever and
+  breaking idempotent resume (duplicate-post vector). Fixed via an in-class seam in
+  `Run_Stage_State::resolve_role()`: when the caller passes `''`, the role is derived from
+  `Stage_Display_Map::default_agent_role($stage)` before any DB query, ensuring every method
+  addresses the same row. Explicit non-empty roles (Phase-2 fan-out) are passed through
+  unchanged. `get()` falls back to `role=''` on miss for mid-run-upgrade compat with v0.18.1
+  rows. Two new `RunStageStateTest` cases lock the symmetric-role and fallback contracts.
+  ARCHITECTURE.md §22 updated to document Phase 1 role population.
 - **Runware float deprecation (P3)**: `normalize_seed()` in `class-runware-image-support.php`
  applied the modulo operator to a float (`microtime(true) * 1000`), triggering
  `Deprecated: Implicit conversion from float … to int` on every image generation. Fixed
--- a/includes/core/class-run-stage-state.php
+++ b/includes/core/class-run-stage-state.php
@@ -6,40 +6,33 @@ declare(strict_types=1);
 *
 * Per-run per-stage state machine over `wp_prautoblogger_run_stages`.
 *
- * What: One row per (run_id, stage, agent_role, item_key) holding
- *       pending|running|done|failed|halted plus the stage's output
- *       snapshot (meta_json). This is what makes resume idempotent on the
- *       chained-cron architecture: a `done` stage is NEVER re-run or
- *       re-charged — re-entry reuses its stored output; `done` is sticky
- *       (an upsert cannot demote it). agent_role is the Phase-2 fan-out
- *       dimension (quorum members; the quorum logic itself is Phase 2);
- *       item_key scopes article-level stages because one run_id spans all
- *       N articles of a batch run ('' for run-level stages). Self-healing:
- *       a missing table degrades every method to a no-op/false.
- * Who triggers it: Pipeline_Runner (run-level research/analysis),
- *       Article_Worker (per-article generate/review/publish),
- *       Content_Generator (per-LLM-stage outline/draft/polish),
- *       Run_Reaper (stuck-stage sweep + payload pruning).
- * Dependencies: WordPress $wpdb, wp_json_encode.
+ * One row per (run_id, stage, agent_role, item_key) holding
+ * pending|running|done|failed|halted plus the output snapshot (meta_json).
+ * Done stages are sticky and never re-charged on resume. Self-healing:
+ * a missing table degrades every method to a no-op/false.
 *
- * @see core/class-run-state.php   — Run-level lifecycle + ledger row.
- * @see core/class-run-reaper.php  — Sweeps stages stuck in 'running'.
- * @see ARCHITECTURE.md #21        — Idempotency-key design.
+ * v0.18.2: role is resolved from Stage_Display_Map when caller passes ''
+ * so start/done/is_done/get_output are always symmetric. Explicit non-empty
+ * roles (Phase-2 fan-out) are honoured as-is. get() falls back to role=''
+ * on miss for mid-run-upgrade compat with v0.18.1 rows.
+ *
+ * @see core/class-run-state.php  — Run-level lifecycle + ledger row.
+ * @see core/class-run-reaper.php — Sweeps stages stuck in 'running'.
+ * @see ARCHITECTURE.md #22       — Pipeline v2 Phase 1 substrate.
 */
 class PRAutoBlogger_Run_Stage_State {

 	/** @var bool|null Per-request "run_stages table exists" probe result. */
 	private static ?bool $table_ok = null;

-	/** Fully-qualified run_stages table name. */
+	/** @return string Fully-qualified run_stages table name. */
 	public static function table_name(): string {
 		global $wpdb;
 		return $wpdb->prefix . 'prautoblogger_run_stages';
 	}

 	/**
-	 * Whether the run_stages table exists and is queryable. Cached per
-	 * request; never throws.
+	 * Whether the run_stages table exists and is queryable. Cached per request.
 	 *
 	 * @return bool
 	 */
@@ -49,7 +42,7 @@ class PRAutoBlogger_Run_Stage_State {
 		}
 		global $wpdb;
 		if ( null === $wpdb ) {
-			return false; // Not cached: $wpdb may appear later in boot.
+			return false;
 		}
 		$table = self::table_name();
 		// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery,WordPress.DB.DirectDatabaseQuery.NoCaching
@@ -59,18 +52,17 @@ class PRAutoBlogger_Run_Stage_State {
 	}

 	/**
-	 * Stable hash identifying one article idea within a run (post-creation
-	 * idempotency + stage item scoping both key off this).
+	 * Stable 16-hex hash for one article idea within a run.
 	 *
 	 * @param PRAutoBlogger_Article_Idea $idea The idea.
-	 * @return string 16-hex-char hash.
+	 * @return string
 	 */
 	public static function idea_hash( PRAutoBlogger_Article_Idea $idea ): string {
 		return substr( md5( $idea->get_suggested_title() . '|' . $idea->get_topic() ), 0, 16 );
 	}

 	/**
-	 * Stage item_key for an article idea ('' is reserved for run-level stages).
+	 * Stage item_key for an article idea ('' reserved for run-level stages).
 	 *
 	 * @param PRAutoBlogger_Article_Idea $idea The idea.
 	 * @return string e.g. 'idea:1a2b3c4d5e6f7a8b'.
@@ -80,22 +72,18 @@ class PRAutoBlogger_Run_Stage_State {
 	}

 	/**
-	 * Mark a stage as entered (running). Upsert: first entry inserts the
-	 * row; re-entry bumps `attempt` — but a `done` stage is sticky and is
-	 * neither demoted nor re-counted.
-	 *
-	 * Side effects: one INSERT … ON DUPLICATE KEY UPDATE.
+	 * Mark a stage as running. Sticky: done stages are never demoted.
 	 *
 	 * @param string $run_id     Run UUID.
 	 * @param string $stage      Stage name (see Stage_Display_Map).
-	 * @param string $agent_role Fan-out dimension ('' default in Phase 1).
+	 * @param string $agent_role Fan-out dimension ('' = resolve from stage map).
 	 * @param string $item_key   Article scope ('' for run-level stages).
-	 * @return void
 	 */
 	public static function start( string $run_id, string $stage, string $agent_role = '', string $item_key = '' ): void {
 		if ( '' === $run_id || ! self::is_available() ) {
 			return;
 		}
+		$agent_role = self::resolve_role( $stage, $agent_role );
 		global $wpdb;
 		$table = self::table_name();
 		$now   = current_time( 'mysql' );
@@ -109,33 +97,26 @@ class PRAutoBlogger_Run_Stage_State {
 					attempt = IF(status = 'done', attempt, attempt + 1),
 					status = IF(status = 'done', 'done', 'running'),
 					updated_at = VALUES(updated_at)",
-				$run_id,
-				$stage,
-				$agent_role,
-				$item_key,
-				$now,
-				$now
+				$run_id, $stage, $agent_role, $item_key, $now, $now
 			)
 		);
 	}

 	/**
-	 * Mark a stage done, persisting its output for resume-without-recharge.
-	 *
-	 * Side effects: one INSERT … ON DUPLICATE KEY UPDATE.
+	 * Mark a stage done and persist its output for resume-without-recharge.
 	 *
 	 * @param string      $run_id     Run UUID.
 	 * @param string      $stage      Stage name.
-	 * @param string      $agent_role Fan-out dimension.
+	 * @param string      $agent_role Fan-out dimension ('' = resolve from stage map).
 	 * @param string      $item_key   Article scope.
-	 * @param string|null $output     Stage output to snapshot (pruned after R days).
-	 * @param float       $cost_usd   Cost attributed to the stage.
-	 * @return void
+	 * @param string|null $output     Stage output snapshot (pruned after R days).
+	 * @param float       $cost_usd   Cost attributed to this stage.
 	 */
 	public static function done( string $run_id, string $stage, string $agent_role = '', string $item_key = '', ?string $output = null, float $cost_usd = 0.0 ): void {
 		if ( '' === $run_id || ! self::is_available() ) {
 			return;
 		}
+		$agent_role = self::resolve_role( $stage, $agent_role );
 		global $wpdb;
 		$table = self::table_name();
 		$now   = current_time( 'mysql' );
@@ -152,32 +133,24 @@ class PRAutoBlogger_Run_Stage_State {
 					meta_json = VALUES(meta_json),
 					updated_at = VALUES(updated_at),
 					finished_at = VALUES(finished_at)",
-				$run_id,
-				$stage,
-				$agent_role,
-				$item_key,
-				$cost_usd,
-				$meta,
-				$now,
-				$now,
-				$now
+				$run_id, $stage, $agent_role, $item_key, $cost_usd, $meta, $now, $now, $now
 			)
 		);
 	}

 	/**
-	 * Mark a stage failed (does not demote a `done` stage).
+	 * Mark a stage failed (does not demote a done stage).
 	 *
 	 * @param string $run_id     Run UUID.
 	 * @param string $stage      Stage name.
-	 * @param string $agent_role Fan-out dimension.
+	 * @param string $agent_role Fan-out dimension ('' = resolve from stage map).
 	 * @param string $item_key   Article scope.
-	 * @return void
 	 */
 	public static function fail( string $run_id, string $stage, string $agent_role = '', string $item_key = '' ): void {
 		if ( '' === $run_id || ! self::is_available() ) {
 			return;
 		}
+		$agent_role = self::resolve_role( $stage, $agent_role );
 		global $wpdb;
 		$table = self::table_name();
 		$now   = current_time( 'mysql' );
@@ -186,43 +159,53 @@ class PRAutoBlogger_Run_Stage_State {
 			$wpdb->prepare(
 				"UPDATE {$table} SET status = 'failed', updated_at = %s, finished_at = %s
 				WHERE run_id = %s AND stage = %s AND agent_role = %s AND item_key = %s AND status != 'done'",
-				$now,
-				$now,
-				$run_id,
-				$stage,
-				$agent_role,
-				$item_key
+				$now, $now, $run_id, $stage, $agent_role, $item_key
 			)
 		);
 	}

 	/**
-	 * Fetch one stage row.
+	 * Fetch one stage row. Falls back to role='' on miss for mid-run-upgrade
+	 * compat with v0.18.1 rows (one extra indexed query, only on miss).
 	 *
 	 * @param string $run_id     Run UUID.
 	 * @param string $stage      Stage name.
-	 * @param string $agent_role Fan-out dimension.
+	 * @param string $agent_role Fan-out dimension ('' = resolve from stage map).
 	 * @param string $item_key   Article scope.
-	 * @return array<string, mixed>|null Row or null.
+	 * @return array<string, mixed>|null
 	 */
 	public static function get( string $run_id, string $stage, string $agent_role = '', string $item_key = '' ): ?array {
 		if ( '' === $run_id || ! self::is_available() ) {
 			return null;
 		}
+		$agent_role = self::resolve_role( $stage, $agent_role );
 		global $wpdb;
 		$table = self::table_name();
 		// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery,WordPress.DB.DirectDatabaseQuery.NoCaching
 		$row = $wpdb->get_row(
 			$wpdb->prepare(
 				"SELECT * FROM {$table} WHERE run_id = %s AND stage = %s AND agent_role = %s AND item_key = %s",
-				$run_id,
-				$stage,
-				$agent_role,
-				$item_key
+				$run_id, $stage, $agent_role, $item_key
 			),
 			ARRAY_A
 		);
-		return is_array( $row ) ? $row : null;
+		if ( is_array( $row ) ) {
+			return $row;
+		}
+		// v0.18.1 upgrade compat: rows started before this fix carry role=''.
+		// One extra indexed query, only on miss, only for known stages.
+		if ( '' !== $agent_role ) {
+			// phpcs:ignore WordPress.DB.DirectDatabaseQuery.DirectQuery,WordPress.DB.DirectDatabaseQuery.NoCaching
+			$legacy = $wpdb->get_row(
+				$wpdb->prepare(
+					"SELECT * FROM {$table} WHERE run_id = %s AND stage = %s AND agent_role = '' AND item_key = %s",
+					$run_id, $stage, $item_key
+				),
+				ARRAY_A
+			);
+			return is_array( $legacy ) ? $legacy : null;
+		}
+		return null;
 	}

 	/**
@@ -230,7 +213,7 @@ class PRAutoBlogger_Run_Stage_State {
 	 *
 	 * @param string $run_id     Run UUID.
 	 * @param string $stage      Stage name.
-	 * @param string $agent_role Fan-out dimension.
+	 * @param string $agent_role Fan-out dimension ('' = resolve from stage map).
 	 * @param string $item_key   Article scope.
 	 * @return bool
 	 */
@@ -244,7 +227,7 @@ class PRAutoBlogger_Run_Stage_State {
 	 *
 	 * @param string $run_id     Run UUID.
 	 * @param string $stage      Stage name.
-	 * @param string $agent_role Fan-out dimension.
+	 * @param string $agent_role Fan-out dimension ('' = resolve from stage map).
 	 * @param string $item_key   Article scope.
 	 * @return string|null
 	 */
@@ -258,13 +241,10 @@ class PRAutoBlogger_Run_Stage_State {
 	}

 	/**
-	 * Fail every still-open (pending/running) stage of one item — used by
-	 * the worker's catch-all so a crashed article leaves no stage stuck
-	 * in 'running' until the reaper. Done stages are untouched.
+	 * Fail every still-open stage of one item. Done stages are untouched.
 	 *
 	 * @param string $run_id   Run UUID.
 	 * @param string $item_key Stage item key ('' = run-level stages).
-	 * @return void
 	 */
 	public static function fail_open_for_item( string $run_id, string $item_key ): void {
 		if ( '' === $run_id || ! self::is_available() ) {
@@ -278,10 +258,7 @@ class PRAutoBlogger_Run_Stage_State {
 			$wpdb->prepare(
 				"UPDATE {$table} SET status = 'failed', updated_at = %s, finished_at = %s
 				WHERE run_id = %s AND item_key = %s AND status IN ('pending','running')",
-				$now,
-				$now,
-				$run_id,
-				$item_key
+				$now, $now, $run_id, $item_key
 			)
 		);
 	}
@@ -290,4 +267,21 @@ class PRAutoBlogger_Run_Stage_State {
 	public static function flush_cache(): void {
 		self::$table_ok = null;
 	}
+
+	/**
+	 * Resolve the effective agent role for a stage.
+	 *
+	 * '' → derive from Stage_Display_Map (unknown stages map to '').
+	 * Non-empty → returned as-is (Phase-2 explicit fan-out role).
+	 *
+	 * @param string $stage      Stage name.
+	 * @param string $agent_role Caller-supplied role ('' = auto-resolve).
+	 * @return string Effective agent role.
+	 */
+	private static function resolve_role( string $stage, string $agent_role ): string {
+		if ( '' !== $agent_role ) {
+			return $agent_role;
+		}
+		return (string) PRAutoBlogger_Stage_Display_Map::default_agent_role( $stage );
+	}
 }
--- a/tests/unit/Core/RunStageStateTest.php
+++ b/tests/unit/Core/RunStageStateTest.php
@@ -137,4 +137,76 @@ class RunStageStateTest extends BaseTestCase {
 		$this->assertFalse( \PRAutoBlogger_Run_Stage_State::is_done( 'run-1', 'draft' ) );
 		$this->assertNull( \PRAutoBlogger_Run_Stage_State::get_output( 'run-1', 'draft' ) );
 	}
+
+	/**
+	 * start(role) → done(role) → is_done(role) must be true; is_done('') must
+	 * be false — this catches the v0.18.2 regression where done() passed ''
+	 * while start() passed a real role, so done() addressed a different row.
+	 *
+	 * In this test the wpdb mock records inserts via query() and simulates
+	 * the SELECT via get_row():
+	 *   is_done('review','editor') → get_row finds the done row → true.
+	 *   is_done('review','')      → resolve_role→'editor'; primary: miss;
+	 *                               legacy fallback (role=''): miss → false.
+	 */
+	public function test_role_tagged_start_and_done_are_symmetric(): void {
+		$get_row_results = array(
+			array( 'status' => 'done', 'meta_json' => null ), // primary hit for 'editor'.
+			null,   // primary miss for legacy check (role='editor' resolved, not '').
+			null,   // legacy fallback (role='') also not found.
+		);
+		$call_index      = 0;
+		$this->wpdb->method( 'get_var' )->willReturn( 'wp_prautoblogger_run_stages' );
+		$this->wpdb->method( 'query' )->willReturn( 1 );
+		$this->wpdb->method( 'get_row' )->willReturnCallback(
+			static function () use ( &$get_row_results, &$call_index ) {
+				$result = $get_row_results[ $call_index ] ?? null;
+				++$call_index;
+				return $result;
+			}
+		);
+
+		\PRAutoBlogger_Run_Stage_State::start( 'run-2', 'review', 'editor', 'idea:xyz' );
+		\PRAutoBlogger_Run_Stage_State::done( 'run-2', 'review', 'editor', 'idea:xyz' );
+
+		// Explicit role finds the done row.
+		$this->assertTrue(
+			\PRAutoBlogger_Run_Stage_State::is_done( 'run-2', 'review', 'editor', 'idea:xyz' ),
+			'is_done with explicit role should be true'
+		);
+		// '' resolves to 'editor'; primary misses; legacy '' fallback also misses.
+		$this->assertFalse(
+			\PRAutoBlogger_Run_Stage_State::is_done( 'run-2', 'review', '', 'idea:xyz' ),
+			'is_done with legacy empty role should be false (no legacy row exists)'
+		);
+	}
+
+	/**
+	 * Mid-run-upgrade resume: a row created by v0.18.1 (agent_role='') must
+	 * still be found by is_done('') under v0.18.2 via the legacy fallback.
+	 * resolve_role('review','') → 'editor'; primary query misses; legacy
+	 * fallback (role='') hits → is_done returns true.
+	 */
+	public function test_legacy_empty_role_row_is_found_on_fallback(): void {
+		$call_index      = 0;
+		$get_row_results = array(
+			null,   // primary query (role='editor'): miss.
+			array( 'status' => 'done', 'meta_json' => null ), // legacy fallback (role=''): hit.
+		);
+		$this->wpdb->method( 'get_var' )->willReturn( 'wp_prautoblogger_run_stages' );
+		$this->wpdb->method( 'get_row' )->willReturnCallback(
+			static function () use ( &$get_row_results, &$call_index ) {
+				$result = $get_row_results[ $call_index ] ?? null;
+				++$call_index;
+				return $result;
+			}
+		);
+
+		// is_done('') resolves to 'editor', misses, falls back to role='',
+		// finds the v0.18.1 row → true.
+		$this->assertTrue(
+			\PRAutoBlogger_Run_Stage_State::is_done( 'run-3', 'review', '', 'idea:abc' ),
+			'v0.18.1 empty-role row must be found via legacy fallback'
+		);
+	}
 }