jared/cc-os

Watch

Fork 0

Pipeline cannot converge when a red suite-check needs a test-side edit #544

New issue

Open

opened 2026-09-09 11:57:29 +00:00 by jared · 2 comments

jared commented

2026-09-09 11:57:29 +00:00

Owner

Context

During #535 implementation (os-sdlc test-path resolution), the suite-check gate went red because a pre-existing test encoded the old contract the ticket was replacing. The os-sdlc map has no route for "the failing test itself must change", so the repair loop routed to programmer-repair, which cannot edit test files.

Observed

The test OsSdlc::Runner::GateCommandsTest#test_lint_present_and_test_lint_nil_for_a_production_only_diff failed because it asserted the old return-a-missing-path behavior. The map routed suite-check/fail → programmer-repair, which lacks test-editing authority. Three repair rounds burned the bound; the run ended bound_exhausted:suite-check/fail. The fix took one test-fixture edit by hand.

Reproduce

n/a: This is a map-routing design gap, not reproducible without a full pipeline run whose suite failure stems from a test file outside the ticket's authored scope.

Expected

When suite-check/fail occurs in a test file the ticket did not author, the pipeline should route the failure to a test-repair node (or test-writer), with the failing test named; or the escalate verdict should report "test-side conflict" so a human decides in one turn instead of three.

Illustration

# Candidate 1: escalate verdict signals test-side conflict
class SuiteCheckGate
  def verdict
    return "escalate:test-side-conflict" if failure_in_unowned_test?
  end
end

# Candidate 2: map routes suite-check/fail → test-writer when failure is in test file
# map.routes.suite-check.fail routes to test-writer_repair, passes failing_test: name

Origin

Trigger: /os-sdlc:implement #535; suite-check red on round 3 of repair loop
Improvised this session: fixture edit by hand to match new contract
Chain: test-route-gap ← suite-check-map-design ← DESIGN (ADR-0160, map-contents-as-configuration)
Root candidate: this ticket
Where: plugins/os-sdlc/lib/os_sdlc/dispatch/maps.yaml
Session: 1a3b7fd0-0319-465b-8414-0ab70560de3d (origin), d233a2f7-977a-419d-99ea-011b6e1e66f8 (this session)
Transcript: https://claude.ai/code/session_01QqjuRkVBpZyEdtkWywpU2Q

Skeptic verdict: VERDICT: CREATE — Origin chain resolves to a root candidate (this ticket, ADR-0160 map-contents design) with observed pipeline failure (bound_exhausted on #535) and a hand improvisation only as the workaround, not the cause — genuine map-routing gap, not speculative.

## Context During #535 implementation (os-sdlc test-path resolution), the suite-check gate went red because a pre-existing test encoded the old contract the ticket was replacing. The os-sdlc map has no route for "the failing test itself must change", so the repair loop routed to programmer-repair, which cannot edit test files. ## Observed The test `OsSdlc::Runner::GateCommandsTest#test_lint_present_and_test_lint_nil_for_a_production_only_diff` failed because it asserted the old return-a-missing-path behavior. The map routed suite-check/fail → programmer-repair, which lacks test-editing authority. Three repair rounds burned the bound; the run ended `bound_exhausted:suite-check/fail`. The fix took one test-fixture edit by hand. ## Reproduce n/a: This is a map-routing design gap, not reproducible without a full pipeline run whose suite failure stems from a test file outside the ticket's authored scope. ## Expected When suite-check/fail occurs in a test file the ticket did not author, the pipeline should route the failure to a test-repair node (or test-writer), with the failing test named; or the escalate verdict should report "test-side conflict" so a human decides in one turn instead of three. ## Illustration ```ruby # Candidate 1: escalate verdict signals test-side conflict class SuiteCheckGate def verdict return "escalate:test-side-conflict" if failure_in_unowned_test? end end # Candidate 2: map routes suite-check/fail → test-writer when failure is in test file # map.routes.suite-check.fail routes to test-writer_repair, passes failing_test: name ``` ## Origin - Trigger: /os-sdlc:implement #535; suite-check red on round 3 of repair loop - Improvised this session: fixture edit by hand to match new contract - Chain: test-route-gap ← suite-check-map-design ← DESIGN (ADR-0160, map-contents-as-configuration) - Root candidate: this ticket - Where: plugins/os-sdlc/lib/os_sdlc/dispatch/maps.yaml - Session: 1a3b7fd0-0319-465b-8414-0ab70560de3d (origin), d233a2f7-977a-419d-99ea-011b6e1e66f8 (this session) - Transcript: https://claude.ai/code/session_01QqjuRkVBpZyEdtkWywpU2Q --- **Skeptic verdict:** VERDICT: CREATE — Origin chain resolves to a root candidate (this ticket, ADR-0160 map-contents design) with observed pipeline failure (bound_exhausted on #535) and a hand improvisation only as the workaround, not the cause — genuine map-routing gap, not speculative.

jared added the

plugin/os-sdlc

filed-by/agent

labels

2026-09-09 11:57:29 +00:00

jared referenced this issue

2026-09-09 11:57:37 +00:00

Diff-test gate loops when a lib file has only scenario-split tests and no mirrored _test.rb #535

jared referenced this issue

2026-09-09 14:29:34 +00:00

#534 A: Project: target keyword required; JSON round-trip #537

jared referenced this issue

2026-09-09 15:06:58 +00:00

#534 B: Schema 21: project_config column; bind at open; Implementation#project #538

jared referenced this issue

2026-09-09 18:07:17 +00:00

#534 E: BriefFacts asks, never queries #541

jared referenced this issue

2026-09-09 18:53:35 +00:00

#534 F: Handoff facts by convention #542

jared referenced this issue

2026-09-09 18:54:56 +00:00

Brief reads every fact from the Implementation aggregate; project config binds at open (was: brief ignores target) #534

jared added the

category/owner-ruling

label

2026-10-10 13:32:09 +00:00

jared commented

2026-10-18 18:58:40 +00:00

Author

Owner

Resolution

Done: Mechanism: a test-side red is routed to programmer-repair, which cannot edit tests. Occurrences: #665 "os-sdlc map: route a red diff-test caused by a test-file change to a role that may edit tests"; #709 "poodr-ticket-implementation map: a red diff-test caused by a test-file defect after the green phase has no route to test-writer".

Evidence: #665 adds ticket #641 impl 18 (Required work named a test-file conversion, two repair rounds changed nothing; asks for an ADR-level map edge). #709 adds #632 impl 35 (refactorer gate red from the test-writer's process-global stubs under parallelize_me!; three programmer-repair rounds 382-384, then bound_exhausted); it also asks for a programmer to declare 'test defect' to short-circuit the loop.

Follow-ups: Ruling on 544 should cover the post-green refactorer case and a test-defect declaration, per #709

## Resolution **Done:** Mechanism: a test-side red is routed to programmer-repair, which cannot edit tests. Occurrences: #665 "os-sdlc map: route a red diff-test caused by a test-file change to a role that may edit tests"; #709 "poodr-ticket-implementation map: a red diff-test caused by a test-file defect after the green phase has no route to test-writer". **Evidence:** #665 adds ticket #641 impl 18 (Required work named a test-file conversion, two repair rounds changed nothing; asks for an ADR-level map edge). #709 adds #632 impl 35 (refactorer gate red from the test-writer's process-global stubs under parallelize_me!; three programmer-repair rounds 382-384, then bound_exhausted); it also asks for a programmer to declare 'test defect' to short-circuit the loop. **Follow-ups:** Ruling on 544 should cover the post-green refactorer case and a test-defect declaration, per #709

jared referenced this issue

2026-10-18 18:59:19 +00:00

os-sdlc map: route a red diff-test caused by a test-file change to a role that may edit tests #665

jared referenced this issue

2026-10-18 18:59:20 +00:00

poodr-ticket-implementation map: a red diff-test caused by a test-file defect after the green phase has no route to test-writer; three programmer-repair rounds are spent for nothing #709

jared commented

2026-11-06 20:54:03 +00:00

Author

Owner

Occurrence: Route a suite failure in a new test back to test-writer
Seen: poodr-ticket-implementation map, suite-check edge to programmer-repair
Run: implementation 246 (#1614), dispatches 2351 and 2352, session 9d8c2e15-00c1-4dec-b177-e749fa05ec67
Adds: A suite-wide guard case. UseCaseTagCoverageTest failed on a new test file. The test-writer gate runs only the new tests, so it did not see the guard, and the programmer escalated.

Occurrence: Route a suite failure in a new test back to test-writer Seen: poodr-ticket-implementation map, suite-check edge to programmer-repair Run: implementation 246 (#1614), dispatches 2351 and 2352, session 9d8c2e15-00c1-4dec-b177-e749fa05ec67 Adds: A suite-wide guard case. UseCaseTagCoverageTest failed on a new test file. The test-writer gate runs only the new tests, so it did not see the guard, and the programmer escalated.