Codex VSCode extension on Windows stalls and crashes on v. 0.4.51

Resolved 💬 26 comments Opened Dec 13, 2025 by ysmartin Closed Dec 16, 2025

💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

What version of Codex is running?

0.4.51

What subscription do you have?

Plus

Which model were you using?

GPT-5.1-Codex-Max

What platform is your computer?

Microsoft Windows NT 10.0.22631.0 x64

What issue are you seeing?

This error appears on the Codex chat view.

An error has occurred
Codex crashed with the following error:

Codex process errored: Codex process is not available
thread 'tokio-runtime-worker' panicked at core\src\util.rs:21:9:
Function call output is missing for call id: call_Mx1hEL51XPh0uQKthJIcbenN
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
Some things to try:

Check your config.toml for invalid settings
Check your settings to disable running in WSL if you are seeing compatibility issues
Try downloading a different version of the extension
Click reload to restart the Codex extension, or visit our documentation for additional help.

Open Config.toml
Reload

What steps can reproduce the bug?

Open Codex chat Window. Click on a previous conversation. The chat view opens, it keeps trying to load the conversation, and after a while, the error above appears.

What is the expected behavior?

_No response_

Additional information

Same results on a portable installation of VSCode.
The error disappears on WSL mode (but the list of chats in chat history is different).

View original on GitHub ↗

26 Comments

etraut-openai contributor · 7 months ago

Thanks for reporting the problem. I suspect it has something to do with one or more of your saved session files.

It would be really useful to have a full stack crawl. You can get that by setting the RUST_BACKTRACE environment variable to 1 before launching your IDE. Then trigger the crash.

We may also be able to get some additional information from your log files. Trigger the crash, then reload the extension and create a new conversation. Submit a dummy prompt like "Hi". This will allocate a session (thread) ID that we can use for tracking purposes. Then use the /feedback slash command to upload your logs and paste the thread ID here. I can then look at your logs to see if there's a stack crawl. I don't know if the crash will be logged and flushed prior to the process being terminated.

MRCHEERS · 7 months ago

me too

hezhexi2002 · 7 months ago

same on macos

oreomilktea · 7 months ago

I am having the same issue on macOS, since the chatGPT 5.2 update

etraut-openai contributor · 7 months ago

Please see my note above if you'd like to help us find and fix this problem.

hatmatter · 7 months ago

This is on Linux. Here's the vscode-server log with rust backtrace set to full for the Codex process when it crashes:

2025-12-14 14:07:10.513 [error] [CodexMcpConnection] CLI: thread 'tokio-runtime-worker' panicked at core/src/util.rs:21:9:
Custom tool call output is missing for call id: call_6rgLnmTUA09GmY45w0ajdmqd
stack backtrace:
2025-12-14 14:07:10.513 [error] [CodexMcpConnection] CLI: 0:     0x7fcf8a449e42 - <unknown>
   1:     0x7fcf89c60ae3 - <unknown>
   2:     0x7fcf8a4493af - <unknown>
   3:     0x7fcf8a449ca3 - <unknown>
   4:     0x7fcf8a449938 - <unknown>
   5:     0x7fcf8a449212 - <unknown>
   6:     0x7fcf8a48c4d8 - <unknown>
   7:     0x7fcf8a48c439 - <unknown>
   8:     0x7fcf8a48fe3c - <unknown>
   9:     0x7fcf8830e03f - <unknown>
  10:     0x7fcf882ea417 - <unknown>
  11:     0x7fcf890c9044 - <unknown>
  12:     0x7fcf890c7ff5 - <unknown>
2025-12-14 14:07:10.514 [error] [CodexMcpConnection] CLI: 13:     0x7fcf8841c667 - <unknown>
  14:     0x7fcf884044ef - <unknown>
  15:     0x7fcf883cc6ea - <unknown>
  16:     0x7fcf8855cc99 - <unknown>
  17:     0x7fcf885760df - <unknown>
  18:     0x7fcf885445bb - <unknown>
  19:     0x7fcf887daba3 - <unknown>
  20:     0x7fcf88775770 - <unknown>
  21:     0x7fcf8a4dd789 - <unknown>
  22:     0x7fcf8a4dbcb2 - <unknown>
  23:     0x7fcf8a4e378b - <unknown>
  24:     0x7fcf8a4bf7ab - <unknown>
  25:     0x7fcf8a4c5cfa - <unknown>
  26:
2025-12-14 14:07:10.514 [error] [CodexMcpConnection] CLI: 0x7fcf8a49025f - <unknown>

joce · 7 months ago

Same, Windows on WSL2.

Ran with RUST_BACKTRACE=1 as @etraut-openai suggested, and was subsequently told by the extension's error message to run with RUST_BACKTRACE=full, which I did.

Got this in the log:

2025-12-14 20:15:14.930 [warning] [warn] is production release true warning false
2025-12-14 20:15:18.399 [error] Error fetching vscode://codex/git-current-branch: TypeError: Cannot destructure property 'gitRoot' of 'undefined' as it is undefined.
2025-12-14 20:15:40.671 [error] Error fetching vscode://codex/git-current-branch: TypeError: Cannot destructure property 'gitRoot' of 'undefined' as it is undefined.
2025-12-14 20:15:40.785 [error] [CodexMcpConnection] CLI: thread 'tokio-runtime-worker' panicked at core/src/util.rs:21:9:
Function call output is missing for call id: call_jYm97nThdA5SaJd7gg2ywS6C
stack backtrace:
2025-12-14 20:15:40.785 [error] [CodexMcpConnection] CLI: 0:     0x74854ce49e42 - <unknown>
   1:     0x74854c660ae3 - <unknown>
   2:     0x74854ce493af - <unknown>
   3:     0x74854ce49ca3 - <unknown>
   4:     0x74854ce49938 - <unknown>
   5:     0x74854ce49212 - <unknown>
   6:     0x74854ce8c4d8 - <unknown>
   7:     0x74854ce8c439 - <unknown>
   8:     0x74854ce8fe3c - <unknown>
   9:     0x74854ad0e03f - <unknown>
  10:     0x74854acea417 - <unknown>
  11:     0x74854bac90cc - <unknown>
  12:
2025-12-14 20:15:40.786 [error] [CodexMcpConnection] CLI: 0x74854bac7ff5 - <unknown>
  13:     0x74854ae1c667 - <unknown>
  14:     0x74854ae044ef - <unknown>
  15:     0x74854adcc6ea - <unknown>
  16:     0x74854af5cc99 - <unknown>
  17:     0x74854af760df - <unknown>
  18:     0x74854af445bb - <unknown>
  19:     0x74854b1daba3 - <unknown>
  20:     0x74854b175770 - <unknown>
  21:     0x74854cedd789 - <unknown>
  22:     0x74854cedbcb2 - <unknown>
  23:     0x74854cee378b - <unknown>
  24:     0x74854cebf7ab - <unknown>
  25:     0x
2025-12-14 20:15:40.786 [error] [CodexMcpConnection] CLI: 74854cec5cfa - <unknown>
  26:     0x74854ce9025f - <unknown>
2025-12-14 20:16:27.216 [warning] [CodexMcpConnection] Attempted to send app-server message but stdin is destroyed: {"id":"CodexWebviewProvider.webview:c16f9bc3-f975-4839-ab5d-a306984994e9","method":"thread/list","params":{"limit":20,"cursor":null,"modelProviders":null}}
2025-12-14 20:16:27.216 [error] [CodexMcpConnection] Codex process fatal error: Codex process errored: Codex process is not available, signal=null, exitCode=null
2025-12-14 20:16:27.216 [error] [CodexWebviewProvider] Fatal error: Codex process errored: Codex process is not available
2025-12-14 20:16:27.222 [warning] [CodexMcpConnection] Attempted to send app-server message but stdin is destroyed: {"id":"CodexWebviewProvider.webview:b20edd3a-e6e1-448a-bb48-569841866d65","method":"thread/list","params":{"limit":20,"cursor":null,"modelProviders":null}}
2025-12-14 20:16:27.222 [error] [CodexMcpConnection] Codex process fatal error: Codex process errored: Codex process is not available, signal=null, exitCode=null
2025-12-14 20:16:27.222 [error] [CodexWebviewProvider] Fatal error: Codex process errored: Codex process is not available

As asked, I started a new discussion and called /feedback. The feedback thread ID is: 019b1f96-87f0-7971-add1-e0a777c2b430.

etraut-openai contributor · 7 months ago

Thanks for posting that. Unfortunately, it's not as useful as I had hoped. The production build of the CLI doesn't include symbols for stack frames. If it were the debug build, we'd see a more useful stack crawl.

joce · 7 months ago

Can't you repro on your end? It's a 100% repro here.

etraut-openai contributor · 7 months ago

No, I'm not able to repro the problem. What steps are you taking to repro the problem? Does it occur as soon as the extension loads? Or does it occur only when loading previous local conversations? Is it only with certain conversations — for example, those created with an older model or older version of the extension?

joce · 7 months ago

Repro steps:

Open VSCode (Codex is loaded in the right side window, and appears to be working).
Press the "Task history" button (no crash there). The list of past convos appears.
Select one of the convo (any).
The plugin appears to be stalled (no visual update) for a while (more than 60 seconds).
Crash.

I'm trying to reload convos from today, I do not recall changing the model since they were created, but... maybe?

My current setting is 5.2 medium reasoning.

The other setting I might have been using was 5.1-codex-max with medium reasoning as well. I tried to reload with that setting and I still get a crash.

Attempting to reload the simple "Hi!" test convo from earlier works. So I guess it might require a longer convo? 🤷‍♂️

I'm guessing you're probably trying on a the OpenAI network or VPN? Not sure if trying on an external network might have any effect for you?

hatmatter · 7 months ago

It is only triggered when clicking into a previous session. For me it is only one particular session that triggers it, but it is with 100% repro for that one session.

joce · 7 months ago

oh! You're right @hatmatter! It's only the one convo that crashes!

joce · 7 months ago

@etraut-openai could it help if I shared the session?

etraut-openai contributor · 7 months ago

Yes, that might help. The problem is that if the crash occurs when you select the session, you won't be able to share it using our /feedback mechanism. Here's an idea... if you install the latest codex CLI, can you resume that session (codex resume and then select the session from the list)? Or does the CLI also crash when selecting that session? If it does crash, that's a valuable clue. If it doesn't crash, you could then use the /feedback slash command to upload the session.

joce · 7 months ago

@etraut-openai All right. It loads just fine in the CLI. I did as instructed. Here's the thread ID: 019b1e0a-8f2f-7b13-8de0-413593f6499e.

Logging out for the night. will check back in the morning.

Melonan · 7 months ago

same issue, however, while I use codex resume , it shows empty, I find the session record in ~/.codex/sessions
whose path and info as below, it seems contain the thread ID:

~\.codex\sessions\2025\12\11>ls
rollout-2025-12-11T22-17-46-019b1090-8ad2-7411-9252-7f337abc688b.jsonl

etraut-openai contributor · 7 months ago

@joce, thanks for uploading that. I'm able to repro the problem with that session.

@Melonan, codex resume filters based on your current working directory. If you want it to show all sessions, you need to use codex resume --all.

kixponent · 7 months ago

thank you guys for fix this error. I think this is due to version error and I downgraded it, and fixed it. 4.51 is occuring, but 4.5 is stable

Melonan · 7 months ago

@joce, thanks for uploading that. I'm able to repro the problem with that session. @Melonan, codex resume filters based on your current working directory. If you want it to show all sessions, you need to use codex resume --all.

I've upload the session : 019b1090-8ad2-7411-9252-7f337abc688b , hope it helps

etraut-openai contributor · 7 months ago

Thanks for your help in finding the repro. I've identified the underlying cause of the crash. It occurs when restoring a session that contains a tool call that was interrupted (canceled).

sr-tamim · 7 months ago

Facing same issue...

Tri-Miquant · 7 months ago

Same issue also, it's appear when I set thinking = 'high' with 'codex-max'
Now I cannot do anything with Codex

PS: I'm using Codex Extension with Window 11

ysmartin · 7 months ago

Thanks for reporting the problem. I suspect it has something to do with one or more of your saved session files. It would be really useful to have a full stack crawl. You can get that by setting the RUST_BACKTRACE environment variable to 1 before launching your IDE. Then trigger the crash. We may also be able to get some additional information from your log files. Trigger the crash, then reload the extension and create a new conversation. Submit a dummy prompt like "Hi". This will allocate a session (thread) ID that we can use for tracking purposes. Then use the /feedback slash command to upload your logs and paste the thread ID here. I can then look at your logs to see if there's a stack crawl. I don't know if the crash will be logged and flushed prior to the process being terminated.

After other responses, I understand that the RUST_BACKTRACE part doesn't apply anymore.
I've followed the process suggested, but I wasn't able to create a new conversation either. Thus, what I've done is triggering the crash, then reloading and opening another conversation that didn't trigger the crash (created with the previous version after I submitted this issue). The thread id is 019b2168-4a37-7872-9596-a461c31763f4

gpeal contributor · 7 months ago

A fix has been rolled out in 0.5.52 pre-release. We will do a production release (0.4.42) tomorrow. Sorry for the inconvenience here.

joce · 7 months ago

A fix has been rolled out in 0.5.52 pre-release. We will do a production release (0.4.42) tomorrow. Sorry for the inconvenience here.

@gpeal @etraut-openai No worries and thanks for the very quick turnaround! 😄