Windows desktop 26.707.9981.0 hangs after loading serialport.node — missing napi_module_register
Summary
The ChatGPT Codex Windows desktop app renders and briefly accepts input, then stops accepting clicks about 5–7 seconds after launch. Reinstalling the Microsoft Store package and the official web installer does not help.
Environment
- Windows 11 25H2, build 26200.8875
- Failing package:
OpenAI.Codex_26.707.9981.0_x64__2p2nqsd0c76g0 - Working package: official signed offline build
OpenAI.Codex_26.707.3351.0_x64__2p2nqsd0c76g0 - Failing Electron/Chromium runtime:
150.0.7871.115
Reproduction
- Launch package version 26.707.9981.0.
- The window renders and may accept input briefly.
- After roughly 5–7 seconds, the UI stops accepting clicks and Windows records Application Error 1000 with exception code
0xc06d007f.
Debugger finding
Live CDB/WinDbg debugging catches a delay-load ERROR_PROC_NOT_FOUND while loading:
@worklouder/device-kit-oai -> @worklouder/wl-device-kit -> serialport -> @serialport/bindings-cpp/build/Release/serialport.node
The delay-load record identifies:
- Requested DLL:
node.exe - Missing export:
napi_module_register dwLastError:0x7f(ERROR_PROC_NOT_FOUND)- The current module handle resolves to
ChatGPT.exe, where the requested export is absent.
The main process imports the Codex Micro service dynamically, but the renderer mounts the Codex Micro bridge and calls getState() on normal startup even when no Codex Micro hardware is present.
Attempts that did not help
- App reload and Windows restart
- Clearing only disposable Chromium UI state
- Disabling GPU and GPU compositing
- Disabling optional browser/computer-use features
- Microsoft Store repair/reinstall
- Full uninstall followed by the official website installer
Confirmed workaround
Install OpenAI's official signed offline MSIX version 26.707.3351.0. It remains responsive beyond the failure window, records no Application Error 1000 event, loads existing conversations, and accepts real navigation clicks.
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗