[Bug] omp has finished running but the left side is still running

[orca-discord-issues]

Steven via Discord

omp has finished running but the left side is still running

[OrcaWin]

Leaving this open — the Discord report is a screenshot of OMP finished while something on the left is still running, which is not enough to close as a duplicate.

Likely related (not confirmed the same root cause):

• [Bug]: OMP agent status stuck at "working" in sidebar after turn completes — panes whose ORCA_PANE_KEY is a reminted "$$<hash>:L$$" never reach the hook pipeline #14018 — OMP sidebar stays working after the turn completes, on panes whose ORCA_PANE_KEY is a reminted $$<hash>:L$$ token
• [Bug]: Split-pane agents remain jointly Running after one finishes #12228 — split-pane agents remain jointly Running after one finishes

[AmethystLiang]

thanks for reporting. looking into it

[nwparker]

Current-source audit: the reminted-pane-key failure mode identified in related #14018 is now fixed by merged #15714, which routes $$…:L$$ OMP hook keys to their canonical pane identity and includes regression coverage for the completed-turn/sidebar settlement path.

This report's screenshot does not identify whether it is that same reminted-key path or the separate split-pane status issue tracked in #12228. I am leaving #15557 open pending a reproduction/log showing the pane key and final OMP hook/status event so we do not close a distinct bug prematurely.

[nwparker]

Current audit: merged #17209 and #15714 cover OMP agent-end settlement and pane-key reminting. Open PR #14702 addresses the remaining dense/nested/parked split-pane title-to-leaf attribution path, with local and real SSH runtime proof and all recorded CI checks green.

That PR explicitly reproduces a separate sparse-ID case (split → close pane → split again) that it does not fix. I am keeping this issue open until that sparse mapping is either fixed with a focused regression or the reporter confirms their case is covered by the merged paths.

[nwparker]

Current-main fix prepared as #21671. I reproduced the remaining sparse-pane-ID case (split → close → split) and added PTY/layout-based attribution. Focused title/leaf tests 33/33, related runtime/title tests 62/62, web typecheck, changed-file quality, and hooks pass. PR includes ELI5 and full proof; CI is running.

[nwparker]

PR #21671 is merged (5cc89b6). It fixes the remaining sparse-pane-ID attribution case after split → close → split. Focused title/runtime tests, related suites, web typecheck, quality, full CI, package, Windows package, and verify passed.

[nwparker]

Fixed by #21671.

[orca-discord-issues]

Shadorain via Discord

I get this also, good work guys havent seen the issue again!

[orca-discord-issues]

FireSly via Discord

I also get this. Happy to test this is you need more data.