be1b76dddd
* test(#4699): add failing-first coverage for skipping complete phases in next_phase * fix(#4699): skip already-complete phases in the next_phase cascade Both next-phase scans selected the numerically lowest phase above N without consulting completion state, so completing a reopened phase persisted an already-[x] phase as STATE.md current_phase while roadmap.analyze correctly named the outstanding one (issue repro: completing 2 with phases 1 and 3 already [x] returned next_phase 03). The cascade collects the complete phase numbers from the roadmap checkboxes (milestone-scoped, comparePhaseNum-deduped) and skips them in both the disk scan and the roadmap scan; a [x] checkbox row and its heading sibling both name a phase that is never next. Heading-only and checkbox-less roadmaps behave exactly as before. * test(#4699): align the negative-control expectation with the disk spelling * test(#4699): pin the STATE.md persistence and the all-later-complete tail corner Review findings: the regression never asserted STATE.md current_phase (the issue's actual harm), and the all-later-phases-[x] corner (is_last_phase true, next_phase null) was unpinned. A changeset fragment is included. * docs(#4699): backfill changeset PR number * test(#4700): add failing-first coverage for the queued headless mine * fix(#4700): queue the headless mempalace mine and surface skipped captures The capture's mine ran in the foreground with no lock handling: MemPalace wraps every mine in a per-palace lock, so any concurrent writer (two phases finishing a stage at once, a git-hook refresh mining the same palace) made it exit 1 (MineAlreadyRunning) and the onError: skip step silently dropped the capture — unlost for CONTEXT/PLAN/SUMMARY files that can be re-filed, unrecoverable for execute:wave:post problem-fix pairs. The mine now queues via --daemon --background (MemPalace #2029: the daemon holds a job refused the lock and runs it when the holder exits), and the report step gains the queued and skipped outcomes per #4700's requirement that a skipped capture never stay silent. Option 2 (write_routing.cli) is unreleased at MemPalace 3.9.0; option 3 (retry) re-enters the same lock race — both declined in the PR body. * fix(#4700): queue the wave:post problems fragment's headless mine too The issue names the execute:wave:post problem-fix pair as the unrecoverable loss (no source file to re-file later); the capture-problems fragment's headless mine ran foreground like the capture capability's did. Same fix: --daemon --background, with the lock-deferral rationale inline. * docs(#4700): backfill changeset PR number * fix(#4682): register the stale-reverification part in the capability registry The new steps/ part is a shipped workflow file; gen-capability-registry --check requires it in the committed registry. --------- Co-authored-by: sim <sim@local>
2.5 KiB
2.5 KiB
Capture problems → fixes (MemPalace)
Gate first. Read .planning/config.json. If mempalace.enabled is not true, or mempalace.capture_artifacts is false, skip this entire section and let the wave complete unchanged. (This contribution is only injected when the capability is enabled; the capture_artifacts check lets you turn capture off without disabling the rest of the capability.)
Otherwise — after verifying this wave, persist any confirmed problem→fix pairs into the palace so they are recalled in future phases. This is best-effort; if MemPalace is unreachable, skip silently — capture never fails a wave.
For each confirmed bug/issue resolved in this wave:
- Resolve the wing (
mempalace.wing, elseproject_code, else project dir) and targetroom: problems. - Dedupe first. Call
mempalace_check_duplicate(interactive) before filing so re-runs don't create duplicate drawers. - File the drawer verbatim. Store the problem statement and its fix as a drawer in
room: problems— interactive:mempalace_add_drawer; headless: stage the artifact under theproblems/folder and runmempalace mine --daemon --background(no--roomflag; the background queue defers the write when another writer holds the palace lock, #4700 — see CLI reference; room assignment is viadetect_room()folder-path match per the mining guide; use the same staging pattern documented ingsd-mempalace-captureStep 3). Include provenance (source_file, phase id). - Mirror the KG fact when
mempalace.mirror_kgis on: add(<bug>, fixed_by, <fix>)withvalid_from= the phase date viamempalace_kg_add. - Mode awareness (
mempalace.memory_mode). Underaugmentthe fact is an additive mirror alongside.planning/graphs/. Underkg_backend/replacethe palace is the authoritative store for the fact; GSD still writes.planning/graphs/through its normal graphify, so an unreachable palace never loses it.
Captures are idempotent: deterministic drawer IDs + check_duplicate mean re-running the wave re-files the same content without duplication. On any error, skip and let the wave complete normally.