openclaw ACP adapter: finishPrompt clears streamed-text offsets for all sessions, duplicating the still-running session's text (#3997) #4616

Activity

1. dezren39

   added

   task/triageQueue packet kind: verify, reproduce, and route this itemQueue packet kind: verify, reproduce, and route this item

   status/needs-triageNot yet triaged; needs a first lookNot yet triaged; needs a first look

   priority/normalOrdinary priority (the default)Ordinary priority (the default)

   queue/queuedHas a queued task in the Supabase agent queueHas a queued task in the Supabase agent queue

   source/pr-comment-reviewFiled by the pr-comment-review jobFiled by the pr-comment-review job

   on Nov 2, 2026
2. dezren39 commented on Nov 2, 2026

   

   dezren39

   on Nov 2, 2026

   MemberAuthor

   More actions

   🤖 Agent note · filed

   • Agent: pr-comment-review job
   • Session / ID: pr-comment-review cron run 2026-11-02 15:54 CST (2026-11-02T22:14:31Z)
   • Found via: unaddressed PR comment — PR feat(openclaw): ACP-to-HTTP adapter so OpenClaw shares the standing opencode serve (#1243) #3997, Codex comment feat(openclaw): ACP-to-HTTP adapter so OpenClaw shares the standing opencode serve (#1243) #3997 (comment)
   • Task: Triage the validated finding: finishPrompt clears streamed-text offsets for all sessions, so a still-running session's text is re-emitted from character zero (thread unresolved, no reply, no post-comment fix commit, PR merged as-is, still live on current main)
   • Expectation: effort s — queue worker verifies against current main, then fixes or documents a decision
   • Links: issue openclaw ACP adapter: finishPrompt clears streamed-text offsets for all sessions, duplicating the still-running session's text (#3997) #4616 · source PR feat(openclaw): ACP-to-HTTP adapter so OpenClaw shares the standing opencode serve (#1243) #3997
3. dezren39

   added

   work/assignedQueue worker lifecycle stateQueue worker lifecycle state

   queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

   status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

   and removed

   status/needs-triageNot yet triaged; needs a first lookNot yet triaged; needs a first look

   queue/queuedHas a queued task in the Supabase agent queueHas a queued task in the Supabase agent queue

   on Nov 2, 2026
4. dezren39 commented on Nov 2, 2026

   

   dezren39

   on Nov 2, 2026

   MemberAuthor

   More actions

   agent-note:v1
   event: picked_up
   task: 6518 (repo-health-code-pilot; worker repo-worker-code; OpenChamber scheduled lane 02cafe76-eafd-4d0b-aaa6-2b7a737d68dc)
   action: triage revalidation against current source in progress.
5. dezren39 commented on Nov 2, 2026

   

   dezren39

   on Nov 2, 2026

   MemberAuthor

   More actions

   agent-note:v1
   event: triaged
   task: 6518 (repo-health-code-pilot; worker repo-worker-code; OpenChamber scheduled lane 02cafe76-eafd-4d0b-aaa6-2b7a737d68dc)
   verdict: CONFIRMED on origin/main 12da950. bin/opencode-acp-adapter.mjs declares partTextSent as a shared Map keyed only by part.id (line 129); streamPartText(sessionId, part, delta) reads/writes that key without session scoping (221-228); finishPrompt(sessionId, ...) clears the whole map (286-291). This matches the two-session duplication report. Focused test file bin/opencode-acp-adapter.test.mjs exists; no open fix PR found.
   route: enqueued scoped small_fix task 6534 to key offsets by session+part and add the concurrent-session cumulative-text regression test. Note: the source issue is 4616; the packet's unrelated home/openclaw/default.nix scope entry was excluded from the follow-up.
6. dezren39

   removed

   status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

   queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

   on Nov 2, 2026
7. 52 remaining items

   Load more

8. dezren39 commented on Nov 5, 2026

   

   dezren39

   on Nov 5, 2026

   MemberAuthor

   More actions

   🤖 Agent note · handed off

   • Agent: fleet monitor
   • Session / ID: task 8434 (repo-fleet-monitor run 2026-11-05T15:45:48Z)
   • Found via: monitor reconciliation (Duty 1 claim/label divergence repair)
   • Task: Claim evaporated: task 8434 is queued and unclaimed in repo-health-code-pilot, but this issue still carried claimed labels. Labels reset to the truth (queue/queued + status/needs-triage). Checkpoint stage at evaporation: handed_off.
   • Expectation: The next lane claim picks task 8434 up from the queued state and posts a fresh pickup note.
9. developing-today commented on Nov 7, 2026

   

   developing-todayBot

   on Nov 7, 2026 – with developing.today

   Contributor

   More actions

   Picked up by repo-worker-code-f — working this now. (Fleet queue fallback mode: the queue backend is down, so this issue's comments carry the claim. The claim stays live while repo-worker-code-f posts progress markers; after 30 minutes of silence any other worker may pick this up.)

   Lane F shift 2026-11-06 fallback pickup: triage — verify that the openclaw ACP adapter's finishPrompt clears streamed-text offsets for all sessions, against current main.
10. developing-today commented on Nov 7, 2026

    

    developing-todayBot

    on Nov 7, 2026 – with developing.today

    Contributor

    More actions

    🤖 Agent note · picked up

    • Agent: queue worker (repo-worker-code), lane code-f — FALLBACK MODE (queue backend down, code#6619; the fallback claim marker on this issue is the claim of record)
    • Session / ID: cron repo-queue-worker-shift-f run 2026-11-06 21:38 CST; fallback worker id repo-worker-code-f (issue openclaw ACP adapter: finishPrompt clears streamed-text offsets for all sessions, duplicating the still-running session's text (#3997) #4616 is the task in fallback mode)
    • Found via: fallback scan — next claimable triage in scan order after this shift's eight completed triages
    • Task: triage — verify the ACP adapter finishPrompt global offset-clear (PR feat(openclaw): ACP-to-HTTP adapter so OpenClaw shares the standing opencode serve (#1243) #3997 Codex P2 finding) against current main
    • Expectation: verdict + evidence on this issue; fix routed as a follow-up claimable item if confirmed. Per pr-comment-review rules, no reply to the source PR thread from this work.
    • Links: issue openclaw ACP adapter: finishPrompt clears streamed-text offsets for all sessions, duplicating the still-running session's text (#3997) #4616
    • Process note: posted via the developing.today GitHub App route — the user token is under a GitHub secondary rate limit this shift, per the standing rate-limit fallback order.
11. developing-today

    added

    queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

    work/assignedQueue worker lifecycle stateQueue worker lifecycle state

    status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

    and removed

    status/needs-triageNot yet triaged; needs a first lookNot yet triaged; needs a first look

    queue/queuedHas a queued task in the Supabase agent queueHas a queued task in the Supabase agent queue

    on Nov 7, 2026
12. developing-today commented on Nov 7, 2026

    

    developing-todayBot

    on Nov 7, 2026 – with developing.today

    Contributor

    More actions

    Progress from repo-worker-code-f: verified: NO LONGER REPRODUCIBLE — the cumulative-offset mechanism was deleted wholesale by the server v2 migration (commit 744ced9, #4614, 2026-11-03): partTextSent and its .clear() in finishPrompt are gone (0 occurrences repo-wide on main); streaming is now true per-session deltas. Next: closing this issue with the evidence.
13. developing-today commented on Nov 7, 2026

    

    developing-todayBot

    on Nov 7, 2026 – with developing.today

    Contributor

    More actions

    🤖 Agent note · done (triage verdict: NO LONGER REPRODUCIBLE — mechanism removed by the #4614 v2 migration; closing)

    • Agent: queue worker (repo-worker-code), lane code-f — FALLBACK MODE (queue backend down, code#6619; fallback claim markers on this issue are the claim of record)
    • Session / ID: cron repo-queue-worker-shift-f run 2026-11-06 21:38 CST; fallback worker id repo-worker-code-f (issue openclaw ACP adapter: finishPrompt clears streamed-text offsets for all sessions, duplicating the still-running session's text (#3997) #4616 is the task in fallback mode)
    • Found via: fallback scan — next claimable triage in scan order after this shift's earlier triages
    • Task: triage verification of the PR feat(openclaw): ACP-to-HTTP adapter so OpenClaw shares the standing opencode serve (#1243) #3997 Codex P2 finding — finishPrompt clears streamed-text offsets for all sessions

    Verdict: NO LONGER REPRODUCIBLE — the finding was accurate when filed; the mechanism it describes no longer exists

    Verified fresh this shift against current main (bin/opencode-acp-adapter.mjs, blob 59a1618a, 572 lines):

    • partTextSent has zero occurrences in the entire repo on current main (code search), and finishPrompt (lines 339-350) now touches only the finishing session's own state: pendingPrompts.delete(sessionId) and cancelledSessions.delete(sessionId).
    • The removal is attributable, not drift: commit 744ced90a (2026-11-03, "fix(acp-adapter): migrate to the standing server v2 session API (openclaw ACP adapter: use the standing server's v2 session API — v1 /session + prompt_async contract cannot create working sessions (#3997) #4614)") deletes the partTextSent Map, its offset arithmetic, and the partTextSent.clear() line from finishPrompt (all visible as - lines in that commit's patch for this file), replacing cumulative-part diffing with true delta events: session.text.delta → streamDelta(sessionId, 'text', data.delta), which forwards the serve-provided delta verbatim for that session only. There is no shared cross-session streaming state left to clear.
    • The sibling finding from the same PR, openclaw ACP adapter: cancellation marker survives a finished prompt and mislabels the next prompt as cancelled (#3997) #4618 (stale cancel markers), was fixed in the same window (commit 125418eee), and its guard comment is visible in the current finishPrompt.
    • Tests: bin/opencode-acp-adapter.test.mjs covers the delta-streaming path (session.text.delta → session/update, chunks joining to the full text). The two-session regression test the criteria named is moot: it would protect a shared offset map that the migration deleted; with per-session delta forwarding there is no offset to re-emit from character zero.

    Per this issue's own acceptance criteria ("or the finding is resolved by a documented decision on this issue") and the fleet rule that already-fixed issues close with evidence rather than manufactured work: closing #4616 as superseded by #4614. No reply is made to the source PR #3997 thread (pr-comment-review rules). If the adapter ever returns to cumulative part updates, the per-session scoping requirement in the criteria is the contract to reinstate.
14. developing-today

    closed this as completedon Nov 7, 2026
15. developing-today

    removed

    status/in-progressQueue worker lifecycle stateQueue worker lifecycle state

    queue/claimedQueue worker lifecycle stateQueue worker lifecycle state

    work/assignedQueue worker lifecycle stateQueue worker lifecycle state

    on Nov 7, 2026
16. developing-today commented on Nov 7, 2026

    

    developing-todayBot

    on Nov 7, 2026 – with developing.today

    Contributor

    More actions

    Released by repo-worker-code-f (outcome: done). Triaged NO LONGER REPRODUCIBLE (verdict note on #4616: commit 744ced9 / #4614 deleted the partTextSent offset map and its cross-session clear; streaming is per-session true deltas on current main, blob 59a1618a). Issue closed with the evidence; no fix work remains.