4.1 KiB
Ordinary default omission: core verification recovered
The ordinary source formatter and lossless updater at 7bb366f66a8c6f8b6f66fc9b04c9e3b39f84b78e now pass the complete core test suite on Windows. This closes the compilation and core runtime gaps identified in the reviewed follow-up. It does not establish consumer, native GUI, source8 integration, Linux, or release acceptance.
Evidence
| Closed gate | Result | Scope and UTC interval |
|---|---|---|
production-followup-1 |
180 passed, 0 failed, 0 ignored | Core library and eight selected integration targets; 11:32:40–11:36:22, 221.4 seconds |
production-core-2 |
511 passed, 0 failed, 1 ignored | Complete oa-core Cargo test invocation, including doc-tests; 11:36:51–11:40:23, 211.7 seconds |
The selected run is a subset of the full run, not an additional 180 distinct cases. The sole ignored test is the existing opt-in inheritance_chain_validation_scaling_probe; it was not executed by these gates. Both commands exited successfully and recorded 613 unchanged inputs at the same published code commit. Neither stream reports a compiler warning or error.
The private module's 22 default-policy/parser tests passed. The repaired External/MixedLegacy language expectations and guarded invalid mutations passed. The corrected port-anchor reset test reached and passed its previously unexecuted re-add checks. All three former authored-default erasure witnesses passed as ordinary enabled tests. The borrowed insertion spans therefore have compilation and affected core runtime evidence in addition to the earlier independent static review; no memory or timing improvement is inferred from whole-suite duration.
Both captures have input-manifest SHA-256 b9a161be8aca264a6dcf4e7df5cbb7a4f9d999411f1da97430ec0eb3a9929566 and source ZIP SHA-256 79dfd5167779094f30f0ee7bdcd678935f7471b294fa2193cfcfd4a5a843459c. The portable evidence archive retains the original inputs, raw streams, start/end receipts and capture helper. Its read-only verifier checks membership, hashes, source contents, command outcomes and raw totals. Executable hashes are historical execution receipts; the archive does not distribute binaries.
Reproduction
Environment: native Windows, Rust/Cargo 1.98.1, stable-x86_64-pc-windows-gnullvm, LLVM-MinGW 20260826 UCRT, locked offline dependencies, CARGO_INCREMENTAL=0. The full run used CARGO_BUILD_JOBS=1; the selected run did not set that environment variable. Use the established Windows toolchain environment setup documented by the project.
cargo test --locked --offline -p oa-core --lib --test language --test port_anchors --test source_project --test default_omission_fidelity --test default_omission_output --test default_omission_updates --test default_omission_preservation --test default_omission_admission --message-format=json -- --nocapture
cargo test --locked --offline -p oa-core --no-fail-fast --message-format=json -- --nocapture
python harness/experiments/default-followup-20260913/verify.py
The original disk-exhaustion failure remains preserved in the earlier production archive. These successful runs follow manual disk cleanup; they do not rewrite that failed result.
Acceptance boundary
Strict production Clippy and actual desktop, CLI and server consumer gates remain necessary. Static review mapped those consumers to the shared formatter/updater and identified a useful HTTP witness: an authored empty port-anchor field and comment must survive an unrelated move and restart, while canonical export omits the default. A separately reviewed test draft is outside these captured execution inputs; it is not covered by the core pass.
The source8 ownership partition, integration decision and live work remain in Gitea issues 7 and 11. Linux checks remain user-paused. Existing native evidence for the earlier diagnostic contribution does not validate this new formatter in a GUI workflow. No release or integration completion is claimed here.