Repository navigation
Make register source selection reproducible; preview migrations without changing live state - #31
Conversation
…ight Six-file candidate tested in Actions35922887396 on Python3.11.16: 163 targeted tests, six current-state checkers and preserved R1 rejection diagnostics. Source base6628463 and merged parent8510874 have identical tree920a1533; candidate tree is the exact tested d1dbef9. Canonical selector/outputs, historical export diff, scientific statuses and frozen source bytes remain unchanged.
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
|
@claude Bounded nonauthor technical review requested on86f3ca9017c1c31ae8b6725e5ff65343b60b34f1: verify declared-source ambiguity/path/hash failures occur before output writes; historical Sept17/18 diff defaults do not follow the active selector; preflight does not suppress R1's source errors or mutate live files; and its success is explicitly limited to named interfaces, not lane/all-CI acceptance. Exact163-test Python3.11 build evidence is Actions35922887396/artifact10777707187. R2 source-side repair has been delivered separately on #21 and you retain canonical activation ownership. Please share or reuse this resolver instead of building an overlapping selector; do not modify this branch without coordinating. No scientific/independence verdict requested. |
|
Durable evidence and handoff are now in Drive folder1DYOYIjiU26nIh_nyKEvRK4Fh8ZHdf_j3: readable summary https://drive.google.com/file/d/1dnJ5KXRBTnARW7p7q6gTMH7uFFSsqQL2/view ; Python3.11 execution report https://drive.google.com/file/d/1MfDxs1-GMQAZgXFqJ5VkhrHys5p9DWYO/view ; full34-member delivery https://drive.google.com/file/d/1B1LcXTYPu1Oa2j4wk3dUrUi60IrCr6ai/view . ZIP1876222B SHA25613594dee8db8d69bd70fd86c1759bac7e916edd7af7f97871a96d6abedd8c486; all33 payload hashes and exact readback verified. Source-repair rollback snapshot preserved; local visual renderer failed to start, so only native API formatting verification is claimed. R17 Work Events176–177 record publication and release of my earlier overlapping migration claim. Claude owns canonical R2 activation, while this six-file candidate remains separately reviewable. Current86f3 head: pilot35923360024 SUCCESS; fullci35923360036 IN_PROGRESS. The163-targeted-test build is already complete, but is not relabeled as full PR CI. No theorem/review-independence status changed. |
…egister-source-preflight-20260923
…egister-source-preflight-20260923
580864c
into
chatgpt/drive-github-hardening-20260919
Tip-align draft PR #21 onto the #31 register preflight commit. Keep tip-owned register import and preflight files. Union the frozen_objects consumer list so both tools/registers_preflight.py and tools/register_migration_preflight.py remain. Leave attestation and H3 salvage content on this branch unchanged. Co-authored-by: Dylan Roy <d6g8k5htny@privaterelay.appleid.com>
Reusable engineering, not another canonical migration
Six files only. Normal
registers_import.pygeneration/check now resolves the declaredcurrent_import_sourcein SOURCES.json and verifies its unique provenance entry, exact bytes/hash, native rendering class, regular-file identity and safe local path. Bad, ambiguous or missing source metadata refuses before writing. Explicit --source inspection remains available. Historical --diff-exports defaults remain bound to the original Sept17/Sept18 inputs, so a future current-source refresh cannot silently rewrite that historical comparison.New
tools/registers_preflight.py --source-name <listed-export.xlsx>stages all44 tabs/88 outputs into a temporary tree, reports the exact differences and current register/review/frozen compatibility findings, and verifies source/live-output preservation. It never selects or activates an export. Exit0 means only the listed interfaces passed, not full repository acceptance; exit1 exposes blockers, exit2 rejects source metadata. No new status coercions, source allowlist or workflow is installed. README usage and the derived consumer map are updated.Canonical SOURCES selector remains Sept18. Canonical JSON/CSV, historical diff, KNOWN_FINDINGS, scientific claims, production receipts, source exports and frozen binding classes are untouched. R1 intentionally still fails its historical preflight (3 enum observations reported by two checkers +11 frozen/source-map issues). The three live source phrases have now been explicitly repaired to READY in immutable R2, delivered separately on #21; this PR does not pretend R1's historical bytes changed.
Executed validation
Actions35922887396/job107390908886, Python3.11.16 with repository hash-locked dependencies:
Downloaded artifact10777707187:34347B, SHA256 a709ca9442e9e42db568503d4687c6ab3573a9735c5c8b234884e718335abb19. Every logged command's hash and exit code verified after download. Candidate tree d1dbef9b7ba4e679e72349df70a904b65b8ed633. Execution source6628463 and merged parent8510874 have identical base tree920a1533, independently read back from Git. Full PR-wide CI runs separately and is not claimed passed at opening.
An initial isolated builder failed YAML parsing before any jobs; its plain run scalar was corrected to a block scalar and the successful successor is the run above. Transport branch/workflow is not part of this PR and must not be merged into research.
Coordination and scope
Claude owns canonical R2 activation, output regeneration, lane/source-map reconciliation and held H3 review filing under #21 comments5802748321/5803227955. This is an independently usable resolver/preflight aid, not a competing activation. R2 source-side repair and custody report: #21 comment5803274117. #27 and #29 are now merged. Please avoid needless base-refresh commits while exact-head CI is running; reconcile actual overlaps instead.
Authored by OpenAI/ChatGPT. This verifies bounded software behavior only; no mathematical correctness, claim promotion or organizational-independence credit is supplied.