[teamleaderleo]

A background-task option added in #33347 reached the task spec without being forwarded through the stable-start helper, so acpmux failed to compile with an undefined result_line. Forward the option from ordinary background starts and explicitly keep it off for native workflow machine tasks.

Validation at d746e624a89b1059cca122b7f9ab8f9718d56e7e: fleet run 41678018220 passed formatting, 3 focused library tests, 13 background-task integration tests and strict acpmux library clippy. The existing integration tests cover output, completion notices and task lifetime; compilation is the regression oracle for the missing argument.

The automatic clippy job’s remaining failure is the existing panic ratchet. Its compiler JSON artifact (13251567529, job 126063559512) has no deny-lint diagnostic in either changed file, and none on this repair’s added lines. Strict library clippy passed in the focused run.

Leo-ask: ask-6ef2624788

“go do all those/help out where you can + examine how you can improve yourself and your available options and capabilities and speed”

🤖 Generated with Codex

Agent session: agent session

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review d746e62: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr34053 -f ref=lane/bg-result-line-forward -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4168460606701 (2e9aa60ffa)
• cmux NEXT nightly 1.0.0-nightly.4168717571501 (4ca639b0ca)
• cmux NEXT nightly 1.0.0-nightly.4168960686802 (6076a8e983)
• cmux NEXT nightly 1.0.0-nightly.4169064870802 (3bfe74f2e0)

[teamleaderleo]

Review (claude-opus-5-5 medium, coordinator): SHIP at d746e62

This fixes the base compile break E0425 (result_line not found at hub/bg_tasks.rs:160). It happened because #33347 added result_line to bg_start and bg_run::Spec after #33779 had split bg_start_with_id out. The fix:

• bg_start_with_id now takes result_line: bool as its last parameter.
• bg_start passes its own result_line through.
• The workflow machine task passes false, since workflow children report through their durable path.

Spec { .., result_line } now binds. I read the function at the head. The diff is identical, line for line, to the 3-line fix pushed separately on fix/base-red-acpmux-result-line (b19fc01), which has no PR.

Follow-ups: none.

[lawrencecchen]

Review (independent-subagent): SHIP at d746e62

Findings (read at the exact head; I did not build on the review machine):

1. Callers. git grep at the head shows two callers of bg_start_with_id. Both pass the right value:
   • hub/bg_tasks.rs bg_start passes its own result_line argument. Its only caller is server/bg_tasks.rs, which reads params.resultLine (default false). The option is not silently dropped.
   • hub/workflow_activity.rs (the durable machine task with a fixed id) passes false. This is correct: that path came from workflow: execute durable local agent and machine activities #33779, before resultLine existed. workflow_bg_wait decides success from state and exitCode only, so false keeps its old behavior. The stable-id replay check compares only command and cwd, so a replayed record is not affected.
   • No other code builds a bg_run::Spec, and nothing else calls hub.bg_start directly.
2. acpmux: run_wait waits for an Actions run or a PR's checks and wakes the chat once #33347 behavior. This still works. run_wait (acpmux_mcp_bg.rs:80) sends "resultLine": true. That value now reaches Spec.result_line, then Record.result_line, then bg_view sets lastLine, and bg_run::outcome returns the last line as the summary once the task ends.
3. Compile. Green at this head: cargo check (x86_64-pc-windows-gnu), which checks acpmux, and these two modules are not cfg-gated. acpmux tests (lib, hub_integration) and cmux-next fast tier (cargo) pass too. The cargo clippy --all-targets step in clippy (changed cmux-tui packages) passes. The 8-argument function is allowed by the crate-level #![allow(clippy::too_many_arguments)].
4. Clippy job red, base finding. The only red step in that job is the Rust panic ratchet: acpmux indexing_slicing 714 > 686, plus counts in many unrelated crates. All named hits are in files this PR does not touch (acpmux_mcp.rs, adopt_live.rs, cli/*, ...). The 3-line diff adds no indexing or panics. The same ratchet is red on unrelated PRs (for example lane/native-workflow-remote-cache-guard, run 41677307458). This PR does not cause it.
5. Rest of the merge. feat(acpmux): durable native workflow CI machines [ask-6ef2624788] #33982 did not change bg_tasks.rs. The result_line field is wired the same way in Spec, Record, the runner (bg_run.rs:551) and bg_view. I found no other break between the two PRs. Correction to the cause in the brief: the missing parameter comes from acpmux: run_wait waits for an Actions run or a PR's checks and wakes the chat once #33347 meeting workflow: execute durable local agent and machine activities #33779's bg_start_with_id, not feat(acpmux): durable native workflow CI machines [ask-6ef2624788] #33982.

Note: feat-cmux-next (f07e8d9) still has the compile red. The duplicate fix #34054 is closed, so this PR is the live fix.