ACP (with Codex CLI) in Windows WSL2 hangs with message "failed to get old checkpoint" caused by "not implemented yet"

Reproduction steps

I don't have a stable reproduction, but this happens to me quite often.

Open Zed in Windows WSL2.
Use the Codex CLI external agent (though I don't think it is relevant to only this specific ACP agent).
My configuration is "Default" + "gpt-5.4" + "High".
Send some messages.

Current vs. Expected behavior

Happens quite randomly (but quite often for me), it just hang there in the generation step forever. And when I open the logs I see the following:

2026-03-22T20:07:17-04:00 ERROR [crates/acp_thread/src/acp_thread.rs:1889] failed to get old checkpoint

Caused by:
    not implemented yet

I'm not sure if it really is "not implemented yet", or it is that I landed at a theoretically unreachable place? If it really is "not implemented yet", I think it should at least not silently hang.

Zed version and system specs

Zed: v0.228.0+stable.203.8421009ef8a022df1196d54bb42fd94366ec0988 (Zed)
OS: Windows 10.0.26200
Memory: 31.5 GiB
Architecture: x86_64
GPU: Intel(R) Arc(TM) Graphics || Intel Corporation || 32.0.101.8332

Attach Zed log file

The relevant piece of log is already mentioned above.

Relevant Zed settings

settings.json

{
  "agent_servers": {
    "codex-acp": {
      "default_config_options": {
        "reasoning_effort": "high",
        "model": "gpt-5.4",
        "mode": "auto"
      },
      "type": "registry"
    }
  },
  // ...
}

Relevant Keymap

Nothing relevant.

(for AI issues) Model provider details

No response

If you are using WSL on Windows, what flavor of Linux are you using?

Ubuntu

Unfortunately this just seems like a hard to pin down bug - I tried for several minutes and I couldn't pin anything down on Zed stable:

Do you have any inkling on what could be triggering this?

The "hanging" happens quite randomly for me... But the log that I showed, I see it almost everytime, even if I just send a hello.

zed-52153-1.mp4

And you see the hang really happens randomly, for instance this one (hello clearly does not take 5min to response, unless it's codex-side issue?)

Thanks for the response. Quick questions:

Where are you (roughly) based? Shanghai?
Do you face these same response delays when using the Codex CLI standalone or other LLM apps or is it only via the Zed integration?

Currently I'm in Cambridge, MA, US. I don't use Codex CLI very often, so I haven't really experienced delay with it. Maybe the above 5-min delay is not a good example; I have previously experienced hours of hanging (and I eventually interrupted it).

I think it might be hard to reproduce the hanging issue. But are you able to reproduce the log message that I saw? That one is kinda happens to me on every message.

Thanks.

I think the logs might be an issue if you're on an older version of Zed - we've moved over to newer ACPs.

Are you getting the message on latest stable Zed v0.230.1?

Yes, the screen recording I uploaded yesterday was from the latest stable:

Zed: v0.230.1+stable.210.bf3a14350a4fa4b67747afe2e50128fb8fea4fa3 (Zed) 
OS: Windows 10.0.26200
Memory: 31.5 GiB
Architecture: x86_64
GPU: Intel(R) Arc(TM) Graphics || Intel Corporation || 32.0.101.8332

I am seeing somewhat similar error when using Opencode ACP (with Github Copilot GPT5.4 High)

2026-04-23T13:13:18+03:00 ERROR [crates/acp_thread/src/acp_thread.rs:2228] failed to get old checkpoint

Caused by:
    No such file or directory (os error 2)

Not able to reproduce it consistently. Just sometimes my message does no go through and errors with the above error message.

opencode --version
1.14.19

Zed 0.233.7 
Commit: 2f844c3cdb1380f992a892f173eb45b6bcaad24c
Version: 0.233.7+stable.244.2f844c3cdb1380f992a892f173eb45b6bcaad24c

I am on macos.

Zed development moves fast and a significant number of bugs become outdated. If you can reproduce this bug on the latest stable Zed, please let us know by leaving a comment with the Zed version, it helps us focus on the right issues. If the bug doesn't appear for you anymore, feel free to close the issue yourself; otherwise, the bot will close it in a couple of weeks. But even after it's closed by the bot, you can leave a comment with the version where the bug is reproducible and we'll reopen the issue. (This bot will only ask about this issue once) Thanks!

This issue was closed due to inactivity. If you're still experiencing this problem, please leave a comment with your Zed version so that we can reopen the issue.

Navigation Menu

Repository navigation

ACP (with Codex CLI) in Windows WSL2 hangs with message "failed to get old checkpoint" caused by "not implemented yet" #52153

Reproduction steps

Current vs. Expected behavior

Zed version and system specs

Attach Zed log file

Relevant Zed settings

Relevant Keymap

(for AI issues) Model provider details

If you are using WSL on Windows, what flavor of Linux are you using?

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

ACP (with Codex CLI) in Windows WSL2 hangs with message "failed to get old checkpoint" caused by "not implemented yet" #52153

Description

Reproduction steps

Current vs. Expected behavior

Zed version and system specs

Attach Zed log file

Relevant Zed settings

Relevant Keymap

(for AI issues) Model provider details

If you are using WSL on Windows, what flavor of Linux are you using?

Activity

miguelraz commented on Apr 5, 2026

Charlie-XIAO commented on Apr 5, 2026

Charlie-XIAO commented on Apr 5, 2026

miguelraz commented on Apr 5, 2026

Charlie-XIAO commented on Apr 5, 2026

miguelraz commented on Apr 5, 2026

Charlie-XIAO commented on Apr 5, 2026

esamattis commented on Apr 23, 2026

github-actions commented on Jul 24, 2026

github-actions commented on Aug 21, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions