Ordinary source8 compatibility evidence
Local-only evidence packet for selected ordinary source8 compatibility. Final source 7b06b8ea1b7ef902ee0b3d0ec2971a9a7a1f1acd; baseline edf096cd33350865ecfc3c65ec266a6c0d3c4000. This is neither a runnable release nor confirmation of remote publication.
evidence.zip:1,220,828bytes,128members, SHA256 879ec5a2198898a8a6d792d163710ca1d6ca83f8e7773ee093dcc93456c77378. manifest.json pins every member's byte count/hash and records its original local file or immutable Git-diff source. ZIP CRC/member readback passed during creation. The independent final packet audit passes with no unresolved finding.
Result and exact source boundary
Core execution at 0c07f53ab517626f536f79dbaa2885a587bf1db7 passes298/298, including all16 former schema-refusal failures and six new tests. Only two explicitly recorded existing test names change. Focused filter passes7: six new tests plus one renamed existing control; it is not seven added tests.
Six selected integration targets at0c07 pass60tests: source_project33, source_identity_updates3, source_nested_updates12, source_files8, review_regressions checked_save_3 and source7_contract selected migration1. Final r3_migration at7b06 passes5, totaling65 selected integration passes. Each target's nonzero actual list and runtime inventory/artifact agree. The checked-save functions reach source then JSON. Windows cfg-excluded Unix cases are not claimed.
Final7b06 differs from0c07 only in the r3_migration test oracle and its implementation note. Among560 declared inputs only that test changes; other559 are byte-identical, including production, core library tests and CLI sources. Final migration suite, strict core all-target Clippy and format were rerun. Unaffected core/six integration/CLI evidence is reused honestly from0c07, not presented as full rerun at finalHEAD. Both Clippy runs exit0. Format exits1 with exactly inherited raw hash c1c6b12861c8117ce6ffc6d6742ebf92618aafb10160ec6f99d370ca72ebab2d; it did not pass.
Actual non-test CLI built at0c07 passes check ENTRY --json and rename-alias ENTRY team rescue_team --json. Explicit-ID three-file Unicode/CRLF fixture, full expected preview texts, exact two changed-file reports and complete unchanged disk inventory/bytes are checked. The independent complete-Project public-API test provides the model oracle; CLI JSON has no Project field. No --write invocation was used. CLI executable22,081,536bytes SHA256 46b2d1324aaff0ba5a95345d667063aa36e4362281bd6b6f0fdcce6e0c98cb81; core executable43,324,928bytes SHA256 b2261dd9168eaa582d478d4989610b37a5c13cec3f42ec93b5c651c2f3c190e9. Executable bytes are not duplicated in this evidence ZIP.
Preserved failures
- Initial33b86 focused5pass/1fail: new fixture assertion incorrectly assumed existing example traces. Explicit authored trace coverage replaces the assumption.
- Intermediate520 focused6pass, list298 and full294pass/4fail: four old private tests still expected ordinary8 header refusal. Narrow version assertions reconcile with selected compatibility while preserving substantive typed/model/numeric checks; two exact names are updated.
- 0c07 r3_migration4pass/1fail: expected migrated JSON omitted current required empty
definitions:{}. Final oracle adds exactly that empty map after asserting original absence/decoded emptiness; production remains unchanged. - First independent checker exits1 due to its own Windows text-decoding corruption of the intended Unicode comment literal. CHECKER-NOTE.md, executed failed checker and observed traceback preserve this evaluator defect. Traceback is transcribed from the actual exec tool result, not separately redirected raw stderr. Correction changes only that literal to explicit
\u96e8; no source, fixture or receipt changes. Corrected checker completes successfully.
Historical failed binaries are recorded by their original hashes; they are not falsely compared to later overwritten files. All initial logs, receipts, declared-input manifests and exact launcher copies remain in the archive.
Navigation and reproduction
Read RUNTIME-REVIEW.md, SOURCE-REVIEW.md, TEST-REVIEW.md, LAUNCHER-REVIEW.md and VERIFICATION-PLAN.md. ORDINARY-RUNTIME-READBACK.json pins the independently reconciled records; inspect_runtime.py is the data-only checker. Its original checkout invocation was:
python -I -B .local/ordinary-source8-review-20260925/inspect_runtime.py 7b06b8ea1b7ef902ee0b3d0ec2971a9a7a1f1acd 7b06b8e 6
The checker requires original local layout, immutable Git objects and current artifacts and refuses to overwrite an existing result. It is not a portable test runner. Raw source-bound launchers record exact Cargo argv/environment; no archived code needs execution to inspect this packet. Archive contains28Cargo gate records,2CLI receipts with stdout/stderr,3original CLI input files, exact launchers, baselineEDF full/list/fmt anchored to prior immutable ZIP 2f5e0c5257eba11908b60e02b76d529fcd53f11edd057c859d3cb1b34f41d7c4, four baseline-to-checkpoint Git patches, implementation note, toolchain evidence, contracts/design and review records. The10 readable review/checker mirrors use -text attributes to preserve raw bytes.
Selected contract and adoption map/critique remain authoritative. This result admits supported ordinary records at schema8; it does not enable public typed source, complete typed formatting, adoption/history integration, universal allocation guarantees or depth128 qualification. Preserve DEPTH_128_VALIDATED=false, the documented full-pipeline small-stack limitation and user-deferred Linux work.
Packet narrative correction
Independent audit found that the first CHECKER-NOTE conflated the executed checker's153-byte corrupted literal with a161-byte secondary diagnostic probe that used Windows default decoding. The corrected note distinguishes both observations and the actual149-byte fixture. The first packet candidate is preserved locally; only the CHECKER-NOTE member changed in this replacement packet. All raw application evidence, both checker versions and successful readback remain byte-identical. See CHECKER-NOTE for the exact byte account.