Queue repair 2.1 - Windows x64

Repair author: IveSan83 (Eugene Bondarenko)

Source: https://github.com/IveSan83/codex-vscode-queue-repair

This release supports Windows x64 / openai.chatgpt 26.930.31428.

Download codex-vscode-queue-repair-2.1.0.zip, extract it into a new folder, close VS Code and run START.cmd. Requirements: stable VS Code with code in PATH, Node.js 22+ (x64), npm and Python 3.11+.

The installer downloads the pinned original extension from its official distribution URL, verifies its checksum, runs all 76 registered tests in both Node and VS Code's Electron/Node runtime, builds and checks the finished VSIX, and installs it. To build and verify without installing, run setup.ps1 -VerifyOnly.

The repair adds a separate atomic revisioned journal, concurrent-writer protection, durable retired IDs, dead-owner recovery, positive delivery reconciliation and canonical follower state. It also normalizes successful void JSON bridge responses. The official codex-cli 0.160.0 runtime is preserved.

The included stress test runs 180 two-window add/remove cycles per runtime with competing writes. Packaging checks every original archive entry and the real relative module dependency chain from the finished VSIX.

The ZIP contains this project's repair source, tests and scripts. It contains no OpenAI VSIX, user settings, private paths, queue contents or credentials. MIT license applies to the independent repair; see NOTICE.md for scope.

Automated tests cover storage, coordinator and package behavior using controlled transport. Live UI and production model delivery still need user verification. Unknown delivery remains paused for review; no exactly-once external action guarantee or model generation speed claim is made.

Related upstream storage report: microsoft/vscode#339391 . Composer draft/focus problems reported separately are outside this repair's verified scope.