Codex Desktop Plugins page renders broken on Intel Mac

Open 💬 3 comments Opened Apr 29, 2026 by liangxin001

<img width="1261" height="806" alt="Image" src="https://github.com/user-attachments/assets/5d675dd8-ea68-4ff6-b6fa-898f9cd95324" />

Environment

  • Codex Desktop: 26.422.62136 (2180)
  • Codex CLI: codex-cli 0.125.0
  • macOS: 15.4.1 (24E263)
  • Architecture: x86_64 / Intel Mac
  • App executable: Mach-O 64-bit x86_64

Summary

The Codex Desktop Plugins page renders incorrectly on Intel Mac. Text and plugin cards are clipped/misaligned, with large blank areas and partial letters visible. This happens on the latest Codex Desktop version available to me.

Steps to Reproduce

  1. Launch Codex Desktop on Intel Mac.
  2. Open the Plugins page.
  3. Observe the plugin list/cards area.

Expected Behavior

The Plugins page should render plugin cards and text normally.

Actual Behavior

The Plugins page layout is corrupted: text/card layers appear horizontally displaced, partial letters are visible, and cards/rows are clipped or misaligned.

Troubleshooting Already Tried

The issue still reproduces after:

  • Launching with --disable-gpu.
  • Clearing/renaming GPU-related caches: GPUCache, DawnGraphiteCache, DawnWebGPUCache.
  • Clearing/renaming frontend caches: Cache, Code Cache, blob_storage.
  • Launching with stronger rendering flags:

--disable-gpu --disable-gpu-compositing --disable-gpu-rasterization --disable-accelerated-2d-canvas --disable-features=SkiaGraphite,UseSkiaRenderer,CanvasOopRasterization,Vulkan,Metal,WebGPU,UseDawn.

  • Launching with a clean user data dir:

--user-data-dir=/private/tmp/codex-clean-profile.

Relevant Log Notes

Logs are under:
~/Library/Logs/com.openai.codex/2026/04/29/

Notable lines observed:

  • unsupported feature enablement workspace_dependencies
  • plugin/list requests taking several seconds
  • primary runtime initially failed to fetch, later installed

Screenshot

A screenshot is available and shows the corrupted Plugins page layout.

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗