Digest / November 24, 2025
Codex Issues Digest — November 24, 2025
23 new issues filed ·
13 resolved in openai/codex (UTC day).
New issues (23)
- Undocumented ~7-8 MB size limit on SQLite databases in workspace-write sandbox mode
- handshaking with MCP server failed
- (pty “Device not configured”)
- Tries to use Python when only Python3 available
- /compact doesn't correctly optimize the context (v0.63.0)
- stupid as fuck the 5.1 gpt
- shortcut to copy output to clipboard
- Images not being processed
- uv throws "Permission denied" error inside Codex sandbox on Linux
- SLOW AS HELL
- deleted
- unified_exec leaks tty file descriptors
- Feature Request: Optional “Safe Edit” backup workflow to prevent destructive edits
- GPT-5.1 model returns prematurely with only ~20-30% of tasks completed
- Never Rechecks If any File is Updated
- Invalid prompt: your prompt was flagged as potentially violating our usage policy.
- workspace-write sandbox: creating POSIX semaphores fails with `PermissionError: [Errno 13] Permission denied`
- Pressing shift enter should create a newline in the prompt, but instead just submits it
- Process never ends
- Codex CLI weekly limit % dropping significantly without usage (from 89% to 74% overnight)
- Custom prompts should only be added to context when invoked, not at startup - they occupy context and confuse the model
- Everytime codex tries to use the `apply_patch` tool, it hangs indefinitely.
- I asked codex to do a /review and it started editing files and making changes to the codebase that was never requested
Resolved issues (13)
- stupid as fuck the 5.1 gpt
- config `max_output_tokens` does not work
- GPT-5.1 model returns prematurely with only ~20-30% of tasks completed
- Never Rechecks If any File is Updated
- Pressing shift enter should create a newline in the prompt, but instead just submits it
- Custom prompts should only be added to context when invoked, not at startup - they occupy context and confuse the model
- unified_exec leaks tty file descriptors
- unable to run codex cli with azure openai deployment
- Support `generateContent` (non-streaming) wire\_api for Gemini models in Codex
- Customize context size for oss models?
- Codex with lmstudio Model Provider does not Actually Edit Source Code
- Codex with gpt-oss:120b model appears to forget the task
- qwen2.5-coder is not able to execute any tools