Skip to content

Commit fc4e490

Browse files
committed
Register source-bound native operation tests for functional coverage
1 parent 5ded69d commit fc4e490

3 files changed

Lines changed: 2074 additions & 362 deletions

File tree

‎docs/ADR/ADR-033-code-quality.md‎

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -455,3 +455,12 @@ Root stages: update the accepted contract and matching closed count/name; exerci
455455
## KL079 authenticated strict execution and completed binder retirement
456456

457457
TASK-KL079-STRICT-ORIGINAL-EXECUTION-008 and TASK-CQ-KL079-STRICT-SOURCE-ADMISSION-082 freeze the exact original run38046690429/attempt1/sourcec67d4dac normal/scalar provider archives, identical48 native instances and private canonical binding proof before implementation. Root replaces only the final KL079 census object with its exact strict proposal, preserves all12 prior objects, removes its census-only admission, and retires the completed one-shot binder together with the shared three-name source registry and its independent native literal oracle. Unknown source names remain rejected. Actual current-image preparation/tamper/refusal/restoration/healthy native flows, exact-source Linux strict48 normal/scalar execution, full source/DLL/PDB/UID/process/resource settlement and full functional coverage remain required; this appendix is not an Implemented or qualified status. Existing189 Query paths,50 ordinary slots, exclusive heavy ownership, timeouts, authorization, atomic cuts, original results, RF3 topology, collector thresholds and source completeness remain unchanged. Root joins/builds/tests/pushes while feature owners continue disjoint work. No runtime compatibility/migration or fallback path is introduced; rollout and rollback keep the strict task/source-admission/oracle coherent and preserve all immutable original evidence.
458+
459+
460+
### TASK-CQ-CURRENT-WHOLE-CLASSIFICATION-044
461+
462+
REQ-CQ-039..045 / AC-CQ-039..045, ADR-033/117: the finite whole122/corrected18/Recovery47/Telemetry5/SampleRange5 intake is classified through actual current-image native execution identities. The original Main TUnit report executed 386 instances (385 passed, one retained Unix socket path failure); the same-image short-path rerun passed that exact failed instance. Its original native report supplies UID, namespace, class, method, display name and source span for 81 new instances across 21 reviewed whole-operation owners and 20 existing instances across four changed owners. The separately executed original PartitionRecordInventoryTests full literal family/page operation passed and supplies the fifth changed owner's one native instance. Admission rebinds only these 21 existing instances and adds only these 81 genuinely observed new instances. All original inventory rows, classifications and group assignments remain retained; new classes have one deterministic group assignment.
463+
464+
The canonical inventory parser, actual 102-instance UID/source-span admission and all five exact disjoint group selectors passed. The declared inventory union is 3464 (3041 functional and 423 required-non-contributor), not a claim of a complete native CLI census. Original reports and independent current PE/PDB/compiled-source evidence are immutable; the before/after Unit observation is identical for all 1922 declared compile sources, native metadata and compiled receipt. LocalRf3ImageContextSnapshotTests and NativeTUnitSelectionTests remain required-non-contributor. Product contributors remain the separate reviewed subset; new Recovery/Integration owners retain their complete mandatory process/RF3 gates.
465+
466+
The independently owned unfiltered 1800-second CLI discovery ended with exit 137 and retained settlement/process failures, with producer exit, stdout/stderr readers and disposal joined. It produced no native list and is not a passing census. Authentic selected-execution identity enrollment does not substitute for required complete CI census or coverage. Preserve all thresholds 80/70/90, no-decrease, full native source/PE/PDB binding, strict unknown-instance rejection, required controls, public RF3, recovery and Linux qualification. This JSON/document-only checkpoint neither closes a product task nor promotes local evidence into delivered Linux acceptance. Rollback retains the complete inventory/docs checkpoint together and preserves every original report.

‎docs/Features/CodeQuality.md‎

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2226,3 +2226,12 @@ The owner2026-10-11 normal-only SIMD decision also retires the removed ScalarUni
22262226
The 2026-10-11 current-source checkpoint passed the canonical full Release build with zero warnings/errors and no drift across 7344 source inputs. Fresh native discovery observed 3383 Unit and 394 Integration instances; the strict Unit census matched the reviewed inventory after independent PE/PDB/source verification. The complete 22 operations in the eight original classes containing the previous nine failures passed without skips, source/image/entry drift, or disabled intrinsics. Native TUnit admission remained 50; actual overlap for this selected batch was 16. Changed-file SDK format checks passed after all 62 original whitespace diagnostics were repaired through reviewed native patches.
22272227

22282228
This checkpoint is local development evidence. The original failed runs remain separate; the later focused startup successes do not establish their cause. The final native Roslynk diagnostics invocation failed with an incomplete Streamable HTTP reply, so no clean native diagnostic result is claimed. Required Linux, full functional coverage, recovery, public RF3 and task-acceptance gates remain mandatory; no additional KL task is closed by this checkpoint.
2229+
2230+
2231+
### TASK-CQ-CURRENT-WHOLE-CLASSIFICATION-044
2232+
2233+
REQ-CQ-039..045 / AC-CQ-039..045, ADR-033/117: the finite whole122/corrected18/Recovery47/Telemetry5/SampleRange5 intake is classified through actual current-image native execution identities. The original Main TUnit report executed 386 instances (385 passed, one retained Unix socket path failure); the same-image short-path rerun passed that exact failed instance. Its original native report supplies UID, namespace, class, method, display name and source span for 81 new instances across 21 reviewed whole-operation owners and 20 existing instances across four changed owners. The separately executed original PartitionRecordInventoryTests full literal family/page operation passed and supplies the fifth changed owner's one native instance. Admission rebinds only these 21 existing instances and adds only these 81 genuinely observed new instances. All original inventory rows, classifications and group assignments remain retained; new classes have one deterministic group assignment.
2234+
2235+
The canonical inventory parser, actual 102-instance UID/source-span admission and all five exact disjoint group selectors passed. The declared inventory union is 3464 (3041 functional and 423 required-non-contributor), not a claim of a complete native CLI census. Original reports and independent current PE/PDB/compiled-source evidence are immutable; the before/after Unit observation is identical for all 1922 declared compile sources, native metadata and compiled receipt. LocalRf3ImageContextSnapshotTests and NativeTUnitSelectionTests remain required-non-contributor. Product contributors remain the separate reviewed subset; new Recovery/Integration owners retain their complete mandatory process/RF3 gates.
2236+
2237+
The independently owned unfiltered 1800-second CLI discovery ended with exit 137 and retained settlement/process failures, with producer exit, stdout/stderr readers and disposal joined. It produced no native list and is not a passing census. Authentic selected-execution identity enrollment does not substitute for required complete CI census or coverage. Preserve all thresholds 80/70/90, no-decrease, full native source/PE/PDB binding, strict unknown-instance rejection, required controls, public RF3, recovery and Linux qualification. This JSON/document-only checkpoint neither closes a product task nor promotes local evidence into delivered Linux acceptance. Rollback retains the complete inventory/docs checkpoint together and preserves every original report.

0 commit comments

Comments
 (0)