Skip to content
Source
Sign inSign up

[Bug]: wrong thread ownership duplicate message #33608

Send to my agent
Closedwarpdesign(GitHub profile, opens in a new tab) opened this issue 2 comments

Before submitting

  • I searched existing issues and did not find a duplicate.
  • I included enough detail to reproduce or investigate the problem.

Area

apps/desktop

Steps to reproduce

  1. start working in a thread
  2. after a few minutes, the owning thread message appears

Expected behavior

No interruption.

Actual behavior

After a few minutes, I get the message:

Ownership coordination: thread 01a1ab54-5a65-7c31-8c43-c1dadb9785ea owns this shared checkout now. You resumed after previous interrupt and changed hud/scene files. STOP edits, builds, launches and goal continuation now; preserve your results. No more provider continuation: this primary will review/preserve changes and complete user fixes. I will interrupt you to prevent concurrent writes.

The thing is that the thread 01a1ab54-5a65-7c31-8c43-c1dadb9785ea does not exit, so I don't know why I'm getting this message.

Note that this started happening right after updating to 0.1.8: it never happened before.

Impact

Major degradation or frequent failure

Version or commit

Version 0.1.8 (bf469bb19669)

Environment

macOS 15.5

Logs or stack traces

Screenshots, recordings, or supporting files

Image Image

Workaround

Tell the agent to keep working and ignore this message.

  1. warpdesign(GitHub profile, opens in a new tab) added labelsneeds-triage

  2. juliusmarmingecommented

    T3 Code doesn't generate this message. Another agent wrote it and sent it to your thread.

    • The header on the message (screenshot 1) names the sending thread, Port Lands of Lore to Amiga 1200. Its agent sent the text to your thread as a steer with the t3_thread_send tool. Open thread under the message takes you to it. The "ownership coordination" wording is that agent's own. It isn't in T3 Code's source, either on main or in your build bf469bb196.
    • 01a1ab54-5a65-7c31-8c43-c1dadb9785ea is a Codex session ID, not a T3 thread ID. Codex puts its own session ID into every shell it runs as CODEX_THREAD_ID, a UUIDv7 that starts 01a1… at current dates. T3 puts the T3 thread ID next to it as T3CODE_THREAD_ID. The sending agent quoted the wrong one, so thread_not_found was the correct answer.

    I checked this on main (ee59ab22a8) with a fresh dev server and a Codex thread. The agent's shell had CODEX_THREAD_ID=01a1c2aa-fc43-7e72-… and T3CODE_THREAD_ID=thread:project:…. t3_thread_read on the first returned thread_not_found; on the second it returned the thread.

    To stop the interruptions, tell the agent in "Port Lands of Lore to Amiga 1200" not to message or interrupt other threads. You can also give each thread its own worktree so agents don't try to coordinate a shared checkout.

    Closing this as working as intended. If a steer arrives with no sending thread shown above it, or Open thread goes to a thread you didn't start, please reopen with that screenshot.

  3. juliusmarminge closed this

  4. @juliusmarminge the sending thread (Port Lands of Lore to Amiga 1200) is the thread I'm working on, that's the thing...

    Image

Sign in to join this conversation.