fix(locks): make cron admission atomic

[sjinks]

Summary

• replace request-local cache counters with an atomic options-table lock state
• assign each lock row a random generation and prevent stale cleanup from releasing a recovered or recreated lock
• release all event action locks through the generation-aware path and remove inactive lock rows

Testing

• PHP syntax checks
• PHPCS
• git diff --check
• targeted PHPUnit is blocked locally because /tmp/wordpress-tests-lib and its MariaDB-backed WordPress test database are unavailable

Refs: https://linear.app/a8c/issue/PLTFRM-2759

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

[Copilot]

Pull request overview

This PR refactors Cron Control’s concurrency locking to use an atomic, database-backed lock state (stored in the options table) and introduces generation-aware release logic to prevent stale workers from accidentally releasing newer/recovered locks.

Changes:

• Replaced object-cache-based lock counters with an atomic INSERT ... ON DUPLICATE KEY UPDATE lock state stored in wp_options.
• Added generation tracking per admitted lock acquisition and made free_lock() generation-aware (including deleting inactive lock rows).
• Added unit tests covering admission limits, stale recovery, and old-generation cleanup behavior.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
includes/class-lock.php	Implements atomic options-table lock state + generation-aware acquisition/release and parsing helpers.
includes/class-events.php	Routes event action lock release through Lock::free_lock() consistently.
tests/unit-tests/test-lock.php	Adds unit tests validating concurrency limits and generation-safe cleanup semantics.

Suppressed comments (2)

includes/class-lock.php:152

• random_int() can throw, and this code path currently doesn’t handle that. A failure here would prevent priming locks and may cascade into cron admission failures. Consider the same try/catch + wp_rand() fallback used in check_lock().

 $wpdb->query(
$wpdb->prepare(
"INSERT IGNORE INTO `$wpdb->options` (`option_name`, `option_value`, `autoload`) VALUES (%s, %s, 'no')",
self::get_key( $lock ),
self::build_lock_state( 0, time(), random_int( 1, \PHP_INT_MAX ) )
)

includes/class-lock.php:192

• random_int() can throw and isn’t caught here, which could make reset_lock() fatally error (potentially leaving locks stuck). Consider catching \Throwable and falling back to wp_rand() for generation selection.

 $now = time();
$generation = random_int( 1, \PHP_INT_MAX );
$result = $wpdb->query(

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

Suppressed comments (1)

Previously missed (1) — in code that hasn't changed since the last review.

tests/unit-tests/test-lock.php:37

• Despite the test name, these calls are sequential, so the previous read-then-increment implementation would also pass this test. The PR's core regression is simultaneous admissions observing the same prior count; add a test using separate database connections/processes synchronized to contend on the same lock and assert that only one acquisition succeeds.

 $this->assertTrue( Lock::check_lock( $lock, 1 ) );
$this->assertFalse( Lock::check_lock( $lock, 1 ) );