Skip to content

fix: drop unreachable action=defer from champion-hold ratification --help - #14322

Merged
rjwalters merged 1 commit into
mainfrom
feature/issue-14271
Nov 19, 2026
Merged

rjwalters merged 1 commit into
mainfrom
feature/issue-14271

Conversation

@loom-fleet-dispatch

@loom-fleet-dispatch loom-fleet-dispatch Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

Summary

The Ratification doc comment still advertised action=defer even though the #13932 fail-closed gate in ratification_line exits 3 on an unreadable head before decide can ever produce Action::Defer on this verb's stdout. This narrows the action= enumeration to the values the verb can actually emit, mirroring what #14253 did for verdict=.

Changes

  • loom-daemon/src/cli/champion_hold_cmd.rs: removed |defer from the action= list in the Ratification doc comment (help text only; no behaviour change).

Acceptance Criteria Verification

Criterion Status Verification
Ratification doc comment lists action=<none|hold|stands|release|clear> ✅ git diff shows exactly that enumeration; live ./target/debug/loom-daemon champion-hold ratification --help prints action=<none|hold|stands|release|clear> with 0 defer occurrences
Action::Defer, decide, role-runner lane, tick table unchanged ✅ Diff is 1 file / 1 line (the action= line only); git grep Action::Defer still finds the variant untouched in ratification.rs and role_runner/ratification_lane.rs
cargo check -p loom-daemon passes ✅ Finished dev profile in 2m 11s (run with RUSTC_WRAPPER= — the host sccache server has a stale temp dir, /tmp/loom-nt/.../sccachetNoLMC, which fails every cold compile; likely why prior builder passes on this issue died unclassified. Not fixed here — host-tool issue, out of scope)

Test Plan

  • RUSTC_WRAPPER= cargo check -p loom-daemon -j 2 — green.
  • RUSTC_WRAPPER= cargo build -p loom-daemon -j 2 then loom-daemon champion-hold ratification --help — no defer, enumeration reads none|hold|stands|release|clear.
  • grep -rn 'action=<' loom-daemon/src — no other stale enumeration (single hit, this line).
  • grep -rn 'clear|defer\|defer>' defaults .loom .claude — no stale doc-side enumeration; tick table derives defer from exit code as the issue notes.

TDD: no — help-text-only change, no executing behaviour to test first (per builder-pr.md § "Test-First Discipline", docs/config-only diffs omit the line's test leg).

Note: PR #14253 (open, loom:pr) narrows the adjacent verdict= line of the same comment and is not yet on origin/main (verified git show origin/main:... still lists unreadable). This PR branches from main and does not touch verdict=, so merging both leaves at most a trivial adjacent-line conflict, per the curation sequencing note. That conflict is expected, not an oversight.

Closes #14271

loom dashboard

@loom-fleet-dispatch loom-fleet-dispatch Bot added the loom:review-requested PR ready for Judge to review. Applied by: Builder when opening PR. label Nov 18, 2026
@loom-fleet-dispatch

loom-fleet-dispatch Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor Author

PR #14322's loom:reviewing claim is held by the judge of sweep step-judge-14322-1795011894 on host host-a6032f0c. This comment is a PR lease record (#9401): its liveness signal is its own forge-assigned updated_at, renewed while the judge works and released when it is done. Another judge that finds this lease live stands down before touching a worktree. See defaults/docs/lease-record.md § "PR-phase leases".

@loom-fleet-dispatch loom-fleet-dispatch Bot added the loom:reviewing Judge is reviewing this PR. Applied by: Judge only. Stale after LOOM_STALE_REVIEWING_MINUTES (30m). label Nov 18, 2026
@loom-fleet-dispatch

Copy link
Copy Markdown
Contributor Author

✅ Approved!

A one-line doc-comment narrowing in loom-daemon/src/cli/champion_hold_cmd.rs —
action=<none|hold|stands|release|clear|defer> → action=<none|hold|stands|release|clear>.
Scope is exactly what #14271 asked for: 1 file, 1 line, no behaviour change.

Spec reconciliation

spec=9a40400dbf94617e comments=6979480977@2026-11-18T13:09:53Z,6979659624@2026-11-18T13:16:28Z,6979673857@2026-11-18T13:16:12Z,6979686272@2026-11-18T13:17:44Z,6979752853@2026-11-18T13:19:26Z,6979769804@2026-11-18T13:20:09Z,6980531570@2026-11-18T14:11:04Z
baseline: none - no earlier spec baseline available; reviewed against current requirements

Reviewed against #14271's current body (Observed section + Curator Enhancement
acceptance criteria) and its trusted comments. No material difference between
the current requirements and this PR's scope or implementation.
The Champion
promotion comment (2026-11-18T13:16:12Z) re-verified the same premise against
origin/main and reached the same conclusion; its sequencing note about #14253
is honoured (see below).

Premise re-verified independently (not taken from the PR body)

The claim is that action=defer can never appear on this verb's stdout. Verified
by reading the code at 3d5a12d0, not by trusting the description:

  • ratification_line (champion_hold_cmd.rs:286-288) fails closed on
    !unreadable_candidates(&files, &cfg).is_empty() → fail("head content") →
    exit 3, before scan/decide run.
  • unreadable_candidates (ratification.rs:195-201) selects on
    status != "removed" && is_candidate(f, cfg) && f.head_content.is_none() —
    byte-identical to the predicate that puts a file into scan's unreadable
    vector (ratification.rs:171-178). scan returns Verdict::Unreadable only
    when that vector is non-empty, so past the gate the verdict is strictly
    Clear or Hold.
  • decide (ratification.rs:359-375) emits Action::Defer from exactly one
    arm, (Verdict::Unreadable(_), Episode::None | Episode::Released { .. }),
    which is therefore unreachable from ratification_line.

Consumer contract is unaffected — the Champion tick derives defer from the
exit code, never from parsing the enumeration:
defaults/.claude/commands/loom/champion-ratification-hold.md:82 →
case "$RT_RC" in 0) ;; 2) RT_ACTION=none ;; *) RT_ACTION=defer ;; esac, with
line 83 defaulting an empty RT_ACTION to defer. The tick table's defer
row (line 69) stays reachable. Correctly left untouched, as #14271 specified.

Out-of-scope boundaries held: git grep confirms Action::Defer,
decide, role_runner/ratification_lane.rs and the tick table are unchanged
by this diff (the diff is the single action= line).

Class sweep

Searched for other stale enumerations of the same shape, as a pattern not a
one-off:

$ grep -rn 'action=<' --include=*.rs loom-daemon/src
loom-daemon/src/cli/champion_hold_cmd.rs:61:    /// unreadable> action=<none|hold|stands|release|clear> pr=<N>

$ grep -rn 'action=.*defer' defaults .loom .claude --include=*.md
(no output)

Single source of truth, now correct. No other instance in the repo.

Test Execution

Test plan from PR description:

  1. RUSTC_WRAPPER= cargo check -p loom-daemon -j 2 — ⏭️ Skipped: CI is the test
    gate (Host-side test burden to a bare minimum: CI is the test gate, Builders don't run the suite (operator direction 2026-10-13) #11690) and is green at this exact head — Build loom-daemon (debug, shared), Rust Linting & Build Check, macOS Compile Check and all 5
    Rust Unit Tests shards pass on 3d5a12d0.
  2. cargo build + loom-daemon champion-hold ratification --help shows no
    defer — ⚠️ Reported by the builder, not re-derived here. I did not run
    the live binary (a cold loom-daemon build on this shared 8-core worker is
    ~2+ min and the host's sccache is reportedly broken). What I did verify is
    that the help text has exactly one source — the clap doc comment at
    champion_hold_cmd.rs:58-62 — and that it now reads
    action=<none|hold|stands|release|clear> with zero defer occurrences.
  3. grep -rn 'action=<' loom-daemon/src — ✅ Executed: single hit, output above.
  4. No stale doc-side enumeration in defaults/.loom/.claude — ✅ Executed:
    no matches, output above.

Live-verification triggers (A) and (B) both checked and neither fires: the
diff drives no browser/scraper/DOM parsing, and #14271's body contains none of
the silent-omission fragments (scanned; no matches). No circular-fixture
exposure — there is no fixture, and no executing behaviour changed.

TDD: no — help-text-only change — accepted as plausible and advisory: a
doc-comment-only diff has no executing behaviour to drive a failing test from.

Non-blocking follow-up filed

Filed #14332: Action::Defer turns out to be unreachable in production
altogether, not just on this verb's stdout. decide is its only producer and
ratification_line is decide's only caller, so the gate above makes the
variant live only in its own unit test (ratification/tests.rs:263). #14271's
note that Defer stays "live in the role-runner lane" does not survive reading
role_runner/ratification_lane.rs:204-222, which matches scan's Verdict
directly and carries its own MatchState::Unreadable — it never touches
Action. That is out of this PR's scope (#14271 explicitly said leave the
variant alone, and this PR correctly did), so it is tracked separately rather
than held against the diff.

Merge sequencing

#14253 (open, loom:pr) narrows the adjacent verdict= line of this same doc
comment and has not merged — confirmed live: gh pr view 14253 reports
state=OPEN, mergedAt=null. This PR's base therefore still carries
verdict=<clear|hold|unreadable>, mergeStateStatus is CLEAN, and the PR
body flags the expected trivial adjacent-line conflict if #14253 lands first.
That disclosure is accurate and is not an oversight.

loom dashboard

@loom-fleet-dispatch loom-fleet-dispatch Bot added loom:pr PR approved by Judge, ready for Champion auto-merge. Applied by: Judge. and removed loom:review-requested PR ready for Judge to review. Applied by: Builder when opening PR. loom:reviewing Judge is reviewing this PR. Applied by: Judge only. Stale after LOOM_STALE_REVIEWING_MINUTES (30m). labels Nov 18, 2026
@loom-fleet-dispatch

loom-fleet-dispatch Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor Author

PR #14322's (no label: the lease is the claim) claim is held by the champion of sweep step-champion_merge-14322-1795017936 on host host-e1d4c843. This comment is a PR lease record (#9401): its liveness signal is its own forge-assigned updated_at, renewed while the champion works and released when it is done. Another champion that finds this lease live stands down before touching a worktree. See defaults/docs/lease-record.md § "PR-phase leases".

@loom-fleet-dispatch

Copy link
Copy Markdown
Contributor Author

Champion: Holding for Human Merge

Judge-approved and passing the mechanical safety criteria, but not
auto-merging:

  • Blast radius: touches loom-daemon/src/cli/champion_hold_cmd.rs, the champion-hold CLI (sticky-hold logic and its janitor, the champion_hold* entry of the kept set, "the merge gate itself"). The change is a one-line help-text edit with no behaviour change, but the gate cannot vouch for a change to itself, and kept-set paths are never self-released.

Next steps — this hold stays in force until one of these happens; Champion re-reading the same diff will not clear it:

  • A human releases it and merges in one audited step: ./.loom/scripts/merge-pr.sh 14322 --release-hold
  • Or applies loom:auto-merge-ok to override the hold; Champion merges on the next tick
  • Or comments here to clear it explicitly, starting the comment with the instruction: clear the hold — <why>
  • Or the situation actually changes: a new push that narrows the concern, or a deeper Judge re-review

Whichever path releases it, Champion will post a comment naming what changed before it merges.

Keeping loom:pr. This PR stays in the queue and is re-checked each tick against the release conditions above.


Automated by Champion role

loom dashboard

@loom-fleet-dispatch loom-fleet-dispatch Bot added the loom:blocked Waiting on something named (another change, a capability); released once it resolves label Nov 18, 2026
@rjwalters

Copy link
Copy Markdown
Member

clear the hold — operator ruling (rjwalters, 2026-11-19, attended /fleet:decide round 20261119-r17, q49): release and merge. Recorded by the attended agent session.

The kept-set hold is right in general (the gate cannot vouch for itself); this diff at 3d5a12d0459c was read line by line and is text only: help text or prompt wording, no behaviour change.

@rjwalters

Copy link
Copy Markdown
Member

Champion hold released by operator @rjwalters released the human-merge hold on this PR and is merging it at approved head 3d5a12d0459cf306d70ae7504333b1d6cdfe38d4.

merge-pr.sh --release-hold verified: labels were only loom:pr + loom:blocked, the Champion hold-state head and the Judge-approved head both equal 3d5a12d0459cf306d70ae7504333b1d6cdfe38d4.

loom dashboard

@rjwalters rjwalters removed the loom:blocked Waiting on something named (another change, a capability); released once it resolves label Nov 19, 2026
@rjwalters
rjwalters merged commit 7c9cd30 into main Nov 19, 2026
24 checks passed
@rjwalters
rjwalters deleted the feature/issue-14271 branch November 19, 2026 22:49
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

loom:pr PR approved by Judge, ready for Champion auto-merge. Applied by: Judge.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

champion-hold ratification: --help still advertises action=defer, which the verb can no longer emit

1 participant