fix(instrument): name stored-snapshot reconstruction errors honestly - #179
fix(instrument): name stored-snapshot reconstruction errors honestly#179cursor[bot] wants to merge 5 commits into
Conversation
After process restart, load the stored locale, digest, item set, and publication state before starting a new session. A Published snapshot may start sessions on that exact form. Missing identity is absent. Duplicate stored item versions fail closed. Exact persist replay stays Duplicate. Publication evidence is not invented on reload. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
load_instrument_release rebuilds Draft, Review, Published, Suspended, and Retired rows. A corrupt row is a stored-snapshot failure, not a published-only failure. Review reconstruction still cannot publish without rebound evidence. Co-authored-by: Seongho Bae <seonghobae@users.noreply.github.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Review (#179 8712307)
Prefer this head over #171 for instrument-release reload. InconsistentEvidence is now a stored-snapshot reconstruction failure: the Display text names a corrupt durable row, not a published-only adapter. from_persisted_snapshot plus load_instrument_release reconstruct any stored lifecycle, omit event history and bound evidence, and keep exact persist replay as Duplicate. The Review-without-rebound-evidence test matches the documented reactivation rule.
No blocking defect on this honesty slice. Domain tests cover Draft/Review/Suspended/Retired session eligibility and the Review Publish fail-closed path. PostgreSQL tests cover Published reload, Draft reload, missing identity, blank/numeric identity, isolation, and duplicate stored item versions.
Residual buyer gap, not a reason to rewrite this head: after restart a worker can load one known release_ref but cannot enumerate which stored Published forms may start sessions. #165 is in-process HTTP only and does not read PostgreSQL. #164 remains the session-start landing (load_published_instrument_release). Do not merge this PR in parallel with #164; rebase after #164. Do not fold HTTP or session start into this honesty slice. Do not merge #171.
This automation cannot approve or merge. After exact-head Runtime CI, coverage, and an independent last-push review, call load_instrument_release with the release reference a session or catalog already holds. If reconstruction is Published, start sessions on that exact locale, digest, and item set. If it is Draft, Review, Suspended, or Retired, do not start sessions. If reconstruction fails, treat the row as corrupt stored evidence and do not start sessions.
Sent by Cursor Automation: Fix Issues
2 participants
Superseded
This persisted instrument-release reconstruction slice is fully contained in #193. Do not merge this predecessor.
Fresh exact ancestry evidence immediately before closure:
87123079430065b88cd8c34e8850c83d07de0e464ee1c01e0918030befcf3b13c37b7b0079cf6587#193 retains generic stored-release reload and honest reconstruction errors and adds a fail-closed catalog of only startable Published stored releases, ordered deterministically. #193 remains Draft and subject to unchanged exact-head CI/security/review gates. Closing this PR does not promote successor behavior to protected-main truth.