P1: Apply Python semantic rename atomically with rollback

[tomdps]

Parent

Tracking parent #249; epic #243. This is one vertical delivery slice. Do not close or bypass the parent tracker from this child.

Goal

Apply only a reviewed semantic preview after exact validation binding, path/version/hash rechecks, deterministic commit, final verification, and proven rollback.

Dependency contract

• Hypothetical (validate pre-write) evaluation must be comprehensive — graph-backed checks are silently skipped #197
• P0 — Harden Python type checking and configuration semantics #245
• fix(python): make ruff and pytest toolchain status honest #211
• P1: Preview Python semantic rename as checksummed WorkspaceEdit #265 (P1: Preview Python semantic rename as checksummed WorkspaceEdit)

Required behavior

1. Only opcore edit apply of an exact serialized A plan may apply Python semantic edits; one-shot rename --apply remains refused. Recompute planHash/planId, repo/context/input/baseline/preimage fingerprints, and reject any tamper or drift before staging.
2. Extend normalization/apply only to a safe single regular-file RenameFile returned by semantic rename. Reject Create/Delete/folders, overwrite/ignore flags, malformed order, unsupported metadata, destination/source/case/cycle collisions, and unreviewed operations.
3. Construct one exact Hypothetical (validate pre-write) evaluation must be comprehensive — graph-backed checks are silently skipped #197 after-state and require mandatory Python syntax, canonical imports, and P0 — Harden Python type checking and configuration semantics #245 authoritative type gate plus only explicitly requested fix(python): make ruff and pytest toolchain status honest #211 Ruff/pytest gates. Verify overall and every mandatory/requested run is exact passed and validation recomputed after-state hash equals preview; disabled/unavailable/invalid/timeout/findings/skipped/not-applicable/provider/infrastructure/malformed states refuse without writes.
4. Before/after validation and immediately precommit, canonicalize and lstat-walk source/destination/ancestors; refuse symlink, hardlink, special/generated/private/ignored/cache/read-only/unwritable/out-of-root/case/identity/inode swaps. Support verified UTF-8/BOM/line endings/final newline and measured mode/uid/gid only; refuse unsupported encoding/metadata rather than overclaim.
5. Snapshot reviewed content/metadata, stage all writes, recheck precommit identities, commit deterministically, reread final checksums/metadata against afterStateHash, and clean temps. Validation/stage refusal reports not_started/unchanged. Commit/postverify failure restores reverse order and verifies every path/metadata/directory; incomplete restore is rollback_failed with uncertain paths.
6. Guarantee all-or-nothing only for handled in-process failures with verified rollback; make no crash/power-loss durability or ASP host apply claim.

Acceptance criteria

• Multi-file preview/check/apply, unsaved overlay disk/service preimages, plan/id/field tamper, document/disk/context/unrelated-universe drift, safe RenameFile ordering, destination/source/unsupported resource cells are covered.
• Generated/ignored/read-only/out-of-root/symlink/ancestor+inode swap/hardlink/special/case/encoding and UTF-8 metadata preservation cases refuse or verify exactly.
• Every Hypothetical (validate pre-write) evaluation must be comprehensive — graph-backed checks are silently skipped #197/P0 — Harden Python type checking and configuration semantics #245/fix(python): make ruff and pytest toolchain status honest #211 validation state refuses before write unless exact passed; injected stage/commit/postverify/rollback/cleanup failures prove unchanged or verified restoration and explicit uncertainty on rollback failure.
• Packed preview then apply reports validation/tool/snapshot/final/rollback evidence; stale/tampered/failure cases are nonzero applied:false with no ambient fallback.

Observable outcomes

• Clean success carries positive execution/provenance evidence; zero diagnostics alone is never proof.
• Findings and every unavailable/invalid/timeout/crash/resource/stale/unsupported state are machine-readable and fail or degrade exactly as the capability contract says.
• No implicit fallback, partial result, or skipped execution is reported as a pass.
• Target source, configuration, lockfiles, environments, and caches remain unchanged except where this issue explicitly owns a validation-gated atomic edit; every temporary resource/process is cleaned.

Non-goals

• module/package moves
• ASP apply authority
• tool install/checker selection
• power-loss journal
• arbitrary binary encodings
• readiness claims

Verification

Run focused behavior first, then the configured repository proof only after the acceptance matrix works:

• npm run build
• node --test tests/edit-python-rename-preview.test.mjs tests/edit-python-rename-apply.test.mjs tests/edit-atomic-writer.test.mjs tests/edit-atomic-writer-mixed.test.mjs tests/edit-validation.test.mjs tests/validation-overlays.test.mjs tests/validation-runner.test.mjs tests/contracts.test.mjs tests/installed-bins.test.mjs
• npm run lint
• npm run pack:check
• bash ./scripts/ci/run-local-ci-equivalent.sh
• Attach exact real-tool/artifact versions, argv/config/cwd/source, normalized result examples, before/after cleanliness evidence, and packed-install proof where the slice changes public package behavior.

Execution rule

Implement only after a just-in-time review against current dev. Keep exactly one implementation run active. PR base is dev; require green checks, clean scoped diff, merged PR, closed child, parent checklist update, focused reproductions, and opcore-ci proof before starting the next child. Do not claim Python readiness/parity, ASP authority, old-tool replacement, security/SAST, all-stack support, automatic fixes, or a blended score.