fix(redis-lock): URL parsing fix + lock domain single-flight (PR-1, Issue #704)

[jlin53882]

PR-1: URL parsing fix + lock domain single-flight

What Changed

• New: src/redis-lock.ts — parseRedisUrl() + determineLockDomain() single-flight

  • parseRedisUrl(): handles legacy URLs without scheme (e.g. "localhost:6379" → "redis://localhost:6379")
  • Has scheme URLs (redis://, rediss://, redis://?tls=true, etc.): passed directly to ioredis unchanged
  • Old implementation used .replace('redis://', '') which broke rediss:// URLs (removed "re" leaving "s://...") and TLS query strings
  • determineLockDomain(): single-flight decision — all concurrent callers share the same Promise, so the whole process decides redis vs file lock only once
  • If redis-lock.js import fails, defaults to 'file'

• New: test/redis-url-parsing.test.mjs — 7 test cases (all passing)

  • Legacy without scheme: "localhost:6379" → "redis://localhost:6379"
  • Standard redis:// + auth: preserved unchanged
  • TLS rediss://: preserved unchanged (was broken before)
  • TLS + Auth: preserved unchanged
  • Query string redis://host:6379?tls=true: preserved unchanged (was broken before)
  • IPv4 IP:port: scheme added
  • IPv6 [::1]:port: scheme added

• Modified: src/store.ts — removed dead code getLockDomain() (was not called by anyone)

• Modified: scripts/verify-ci-test-manifest.mjs — registered redis-url-parsing.test.mjs

• Modified: package.json — added ioredis ^5.10.1 devDependency (module-level import requires it)

PR Split Context

This PR is Phase 1 of 3 for Issue #704 (Redis distributed lock).

Phase	Content	Status
PR-1 (this PR)	URL parsing + single-flight	done
PR-2	RedisLockManager full implementation	pending
PR-3	Integration tests + CI	pending

Verification

node test/redis-url-parsing.test.mjs
# 7/7 passing

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e993f781cd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Release probe Redis connection before returning

When determineLockDomain() detects a healthy Redis instance, it returns 'redis' but drops the manager without calling disconnect(). In that path, the underlying ioredis socket/timers stay alive for the lifetime of the process, which can leak a connection and keep short-lived CLI/test processes from exiting cleanly once this function is used. If this function is only meant to decide lock domain, the health-check client should be explicitly closed (or retained and reused).

Useful? React with 👍 / 👎.

[jlin53882]

✅ P2 已修復

determineLockDomain() 在確認 Redis 健康後，現在會主動呼叫 manager.disconnect() 釋放 probe client，避免 connection leak 和 process 無法乾淨退出的問題。

Fix commit： 7472e3f

if (manager && await manager.isHealthy()) {
  _lockDomainDecision = 'redis';
  await manager.disconnect(); // 釋放 probe client，避免 connection leak
  return 'redis';
}

PR #739 現有兩個 P1/P2 維護者問題均已解決。

[rwmjhb]

PR #739 Review: fix(redis-lock): URL parsing fix + lock domain single-flight (PR-1, Issue #704)

Verdict: RESOLVE-CONFLICTS-FIRST | Author: jlin53882 | Merge state: DIRTY

Pipeline short-circuited at the conflict gate after R0 verification. Deep review deferred until the branch rebases cleanly onto the base.

Problem Statement (R1)

The PR attempts to add Redis-aware locking support for memory writes, including safer Redis URL handling for legacy host:port inputs, rediss:// URLs, TLS query strings, and a single-flight decision between Redis and file locking. This targets high-concurrency deployments where local file locks may not coordinate writes safely across processes or hosts.

Why This Stopped Here

GitHub reports mergeable=CONFLICTING (merge_state_status=DIRTY) for this PR. Reviewing the diff now would:

1. Give feedback against a branch the author must rewrite anyway
2. Produce findings that may be invalidated by the conflict resolution
3. Waste review cycles on code that cannot be merged as-is

Recommended Action

Author should:

1. Rebase this branch onto the latest base (or merge the base into this branch)
2. Resolve all merge conflicts
3. Push the rebased branch — the re-review will be picked up automatically

---

Reviewed at 2026-05-09T12:53:40Z | R0+R1 gate | Conflict gate

[jlin53882]

@rwmjhb PR #739 改成用 PR #781 接續PR 修復