Skip to content

Latest commit

 

History

History
190 lines (175 loc) · 18.9 KB

File metadata and controls

190 lines (175 loc) · 18.9 KB

CodeQuality evidence

Configuration and rule contract: CodeQuality. Implementation decision: ADR-033. Working verification: plan.

The owner-selected Prostir EditorConfig was imported byte for byte on 2026-10-01; the final byte comparison passed on 2026-10-02. SHA256: b624db78d435290ba2633a71c901797ff3d66449f76ffd91b4b81724b379b6fb. No imported disabled severity was changed; no authored suppression was added. Twelve rules are source-owned in the CodeQuality slice, with 57 authored TUnit test methods using real compiler, Orleans and ASP.NET metadata. Four new rules enforce file400, aggregate partial type200, executable unit50 and nesting3 limits under the precise metric and bounded exception contract in ADR-033. These authored tests have compiled but have not executed for this source.

Observed gates on 2026-10-02

Gate Actual evidence Result
Current joined working tree on 2026-10-02, based on 9c570f8c33a7a9667507a8e1c0ca68860de3be45 dotnet build KeyLoad.slnx --no-restore --configuration Release Passed all 25 projects with 0 warnings and 0 errors. This is a local, uncommitted source build; no test qualification is implied.
Current joined solution formatter and governance dotnet format KeyLoad.slnx --verify-no-changes --no-restore; node scripts/Features/RepositoryGovernance/verify.mjs; git diff --check; implementation JSON parsing Passed; all 25 project policies and 4 modules preserved. No TUnit, process-recovery, container, RF3, MCP or benchmark workflow ran for this working tree.
Restore dotnet restore KeyLoad.slnx after the embedded library and central CSharp/Common5.0 peer join; strict-embedded-benchmark-paired-restore.log, all25 projects, no new locks Passed development prerequisite; generated Dry compatibility remains pending
Central policy Actual Release MSBuild evaluation of all25 projects, configuration-20261002-25-projects.json;23 ordinary consumers attach the custom analyzer, its own two projects avoid recursion Passed static check, no policy mismatches
Negative policy Build with -p:RunAnalyzers=false fails at EnforceKeyLoadCodeQuality before compilation Expected rejection passed
Analyzer and test project build Release Rebuild of KeyLoad.Analyzers.Tests also builds KeyLoad.Analyzers Passed, zero errors/warnings; not test execution
Initial solution build Release build with all analyzers enabled before prerequisite repairs Failed: 1180 errors, zero warnings; historical source stage
Historical lead solution build Dependency-enabled Release build after the shared Core/contract join Failed: 459 errors, zero warnings; source cut before Query/Artifacts repairs and site restore
Pinned-driver benchmark join Normal dependency-enabled benchmark Release build after Kurrent/Mongo API repairs Failed: 506 analysis/style/XML errors, zero warnings and no type/API-binding errors; historical source cut before final private disposal/type fixes and restored public Kurrent visibility
Historical comparison dependency build Normal dependency-enabled Release build after adapter repairs, before immutable/numeric joins Failed: 16 library diagnostics, zero warnings; nine CA1819, CA1720, CA1710, two CA1032 and three CA2100; the library failure prevented host compilation
Historical prerequisite builds Normal dependency-enabled Release builds of Abstractions, Core, ZoneTree, Client, Query and Artifacts before numeric rules Passed, zero errors/warnings at that source stage; not current numeric or runtime proof
Historical full solution numeric source cut Ordinary enabled Release build, strict-solution-numeric-inventory.log and inventory JSON Failed: 49 errors, zero warnings across ZoneTree6/Core4/Client3/SiteTests36 at that source stage
Historical private storage full solution source cut Ordinary enabled Release build, strict-solution-private-storage-join.log Failed: 54 errors, zero warnings across SiteTests36/CLI13/Query3/Security1/Comparisons1; failed dependencies leave further projects uncompiled
Most recent whole-solution build before embedded/query/storage-test joins Ordinary enabled Release build, strict-solution-host-site-join.log and deduplicated inventory JSON Failed:19 errors, zero warnings across Benchmarks14/CrashHost2/AppHost2/UnitTests1 at that source stage; nonexhaustive historical prerequisite cut, no current whole-solution success
Reviewed numeric prerequisite joins Normal Abstractions/Core/Client/Storage/Security/Query/CLI/Comparisons builds after reviewed private-owner and guard joins Passed, zero errors/warnings with numeric rules enabled; no test execution
Bounded storage metadata production join Normal enabled Storage Release build, strict-storage-bounded-metadata.log SHA256 e8cdfdabeeb8c442095a0d1be61aa2f261adf6bde209b6caccd2c239151e6217 Passed0errors/0warnings; corrected test-source review/registration/allocation/full GitHub qualification pending
Historical comparative host source cut Normal real-dependency host Release build, strict-comparison-host-numeric-owner-join.log Failed: eight ownership/cleanup diagnostics, zero warnings; subsequent preserving repair reviewed
Reviewed Host and SiteTests joins Normal enabled Release builds, strict-comparison-host-cleanup-final.log and strict-site-process-quality-final.log Passed, zero errors/warnings; real cleanup/default/options/lifetime regressions authored, execution pending
Prepared storage mutation source join Normal enabled Release build, strict-storage-prepared-mutations-join.log Passed, zero errors/warnings; one private sorted mutation projection reused for validation/serialization/apply, three real-store regression sources pending CI
Reviewed Host/Site/prepared source formatter Canonical solution formatter with only these joined source paths included, formatter-host-site-prepared-join.log Passed exit0; scoped static proof, complete solution formatter and test execution remain pending
Reviewed embedded scenario library/host Normal enabled Release build, strict-embedded-benchmark-host-final.log Passed0errors/0warnings; actual generated Dry child/TUnit execution pending
Historical UnitTests dependency graph before final lifetime/formatter joins Normal enabled Release build, strict-unit-query-storage-embedded-corrected.log and deduplicated inventory JSON Failed114errors/0warnings at that source cut; post-build snapshots, no runtime qualification
Historical UnitTests dependency build after reviewed lifetime/formatter/alias joins Normal enabled Release build, strict-unit-source-quality-final-join.log SHA256 f7fb4cb72fc94f75ac38018b41bff8ee8ec42bdb0202b78da18fcd4c825ea156 Failed1error/0warnings at that source cut: protected concurrent Core PersistCommandOutcome CA1822. ArtifactTransfer compiled with its original provider CLR type; UnitTests Csc did not run
Latest UnitTests build after pure harness migration Normal enabled Release build, strict-unit-pure-harness-source-join.log SHA256 71e19d7bde4a76ac9fbbc7a83bfeca8eeebfb0600caf7a749abd35e15fa86152 and deduplicated inventory Reached UnitTests Csc, failed3errors/0warnings in protected concurrent BlobRestoreNormalizationTests missing DatabaseEngine. Core's previous CA1822 cleared; all this build's dependencies compiled. UnitTests combined semantic compilation/registration and runtime remain pending
Comparison-only friend metadata Normal enabled Release library build, strict-comparisons-friend-source-join.log Passed0errors/0warnings with the single named ComparisonTests friend; parser/target/genuine runtime joins pending
Reviewed unit formatter join Canonical solution formatter verify with UnitTests include, excluding protected BlobStorage/ClusterRouting/ClientApi and separately pending BenchmarkComparisons/ComparisonHarnessTests; formatter-unit-owned-source-final-join.log Passed exit0 as scoped static evidence; full canonical formatter and actual UnitTests compilation/runtime remain pending
Formatter Canonical solution verify-no-changes Failed on existing source/style findings
Scoped formatter Same command including only the two new analyzer project trees Passed
JSON text source/test formatter Same command including the pooled JSON/shared callers and four new test files Passed after removing two unused imports; no runtime test execution
Latest governance node scripts/Features/RepositoryGovernance/verify.mjs after the embedded library joined Passed: root policy preserved,25 projects, four modules, local policies and empty installed-skill catalog
Whitespace / import review git diff --check, byte comparison and lead joined-diff review Passed static checks
New analyzer tests and full product suites Exact delivered-SHA GitHub Actions qualification Pending; these new tests have not executed
Numeric complexity Four source-owned rules configured and compiled; real compiler fixtures and complete graph qualification required Configured, full gate pending; no weakened thresholds or new exceptions
Numeric coverage Compatible collector, container export, strict raw-report verifier and real baseline Not configured; cannot claim passing

Generated TUnit registration sources confirm the new methods were discovered during compilation. That is source-generation evidence, not a regression result. No tests, recovery qualifications or load benchmarks ran locally.

The initial solution errors came from Abstractions and ServiceDefaults: 1078 CS1591 missing public XML comments, 49 CA1819 array properties, 22 KLD0001 literal machine keys, 21 IDE0011 missing braces, three CA1032 exception-constructor findings, three CA1062 validation findings, two CA1307 string comparisons and one each of CA1716/CA1724 public naming findings. Their failed dependency graph leaves 18 further projects uncompiled; this is not a complete source-debt inventory. ServiceDefaults and contract XML prerequisites have since been repaired. The latest lead solution build at the contract-documentation stage had 50 CS1591, 49 CA1819, one CA1062 and one CA1716 error in Abstractions. Storage contract XML and argument guards were then authored; a normal worker dependency build reached the remaining 49 CA1819 and one CA1716 diagnostic. ADR-041 now governs the explicit read-only CLR contract/consumer migration with stable wire bytes and strict null/default rejection. Abstractions and its joined Core, Query, ZoneTree and Client consumers compiled with their real dependencies before the numeric-rule stage. The subsequent solution build reached 72 Replication, 361 Comparisons, 20 Artifacts, five Query and one missing site-assets error. Query and Artifacts were then repaired and rebuilt cleanly; restore supplied the site assets. Replication and comparison source are being integrated with their concurrent owners. These counts identify that exact earlier build, not a fresh remaining-error total or runtime result.

An earlier comparison-library source cut contained 62 CS1591, 16 CA1062, 13 KLD0001, nine CA1819, six CA2234, three CA2100, two each CA1032/CA2002/KLD0024/IDE0290 and one each CA1720/CA1710/KLD0022/CA1068/CA1849/IDE0019. Shared corpus XML and the preserving sole-host/library split are source-joined. Adapter XML/private helpers, shared guards and source review proceed under AC-CQ-007; public collection/enum/ exception and SQL designs now have separate accepted implementation contracts. Counts are from that exact build, not a fresh complete solution inventory. The IDE0011 join retained all 22,210 non-brace tokens in its 28-file cut; this does not establish compliance with remaining file/type/function limits or runtime correctness.

Compiler SARIF 2.1 reports contain real rule IDs, messages and source locations. The initial report snapshot included 88 informational style findings and 33 CA1822 records suppressed by TUnit's standard instance-test-method suppressor; those suppressed records do not count as build errors. Per-project paths avoid collisions. CI is configured to upload diagnostics even when builds fail and to qualify analyzer rules independently of the product graph.

Local ignored evidence: artifacts/code-quality/report.md, summary.json, configuration.json, solution-build.log, analyzer-build.log, formatter.log, disabled-analyzers-rejected.log, configuration-20261002-24-projects.json and per-project Release/net10.0/diagnostics.sarif. New joined build/inventory logs are retained under /private/tmp/keyload-performance-baseline/; the complete solution source cut is strict-solution-after-core-join.log and its inventory JSON. The later document-image Core join is strict-core-document-image-join.log. The pooled JSON/replay Core join is strict-core-pooled-json-join.log; the historical host dependency cut after braces is strict-comparison-host-after-braces.log with its inventory. After preserving U/V/X repairs the actual normal dependency cut strict-comparison-preserving-join.log contains 41 errors, zero warnings. After reviewed Y HTTP/native guard and ownership packets, strict-comparison-adapter-join.log contains 16 errors, zero warnings: nine CA1819, one CA1720, one CA1710, two CA1032 and three CA2100. That library failure prevented host compilation at that source stage; new argument/ownership cases have not executed. The exact before-source cut is preserved in ignored artifacts. ADR-044 now accepts the separate immutable harness/naming stage with valid JSON/configuration/wire values preserved; source or malformed-input fixtures do not qualify remote cleanup, numeric coverage or performance. The later numeric rules and both analyzer project graphs compile in strict-analyzer-numeric-tests-build.log. The historical full numeric cut is strict-solution-numeric-inventory.log and its JSON inventory. Subsequent Core and Client joins passed in strict-core-numeric-guard-join.log and strict-client-numeric-predicate-reviewed.log, respectively. The later solution cut strict-solution-private-storage-join.log contains54errors/0warnings and is not exhaustive. The reviewed joins are0errors/0warnings in strict-storage-private-owner-final-review.log, strict-comparison-numeric-readiness-final.log, strict-security-redaction-guard-join.log, strict-cli-feature-owner-final.log and strict-query-feature-owner-final.log. The reviewed Host cleanup and SiteTests joins are also0errors/0warnings in strict-comparison-host-cleanup-final.log and strict-site-process-quality-final.log. Site JSON metadata owns copied options, and all seven actual generated DTO creators are parameterless with unchanged initializers. The real Node/options/default/null and static-listener regressions are source-authored only. The older UnitTests dependency graph emitted225errors/ 0warnings in strict-unit-host-site-prerequisites.log. The latest recorded joined build is114errors/0warnings in strict-unit-query-storage-embedded-corrected.log, with its matching deduplicated inventory; no finding belongs to the reviewed query/storage/prepared/embedded scopes. The earlier whole-solution build, strict-solution-host-site-join.log, stopped at19 errors/0warnings. Its verified inventory is strict-solution-host-site-inventory.json; source hashes are explicitly post-build shared-checkout snapshots, not an exact delivered SHA. That historical prerequisite cut predates the clean enabled embedded-host build and latest114-finding UnitTests cut. New source work requires a fresh integrated build and retains all old raw evidence. These test migration/style findings need accepted bounded joins, with fake comparison cases replaced by real caller proof. Full graph, source-owned rule execution, complete formatter and exact-SHA CI remain pending. Coverage research is complete; compatible TUnit already resolves extension18.11.2 with MTP2.4.1, but explicit collection, container export and authentic numeric baseline remain pending. Reports are build artifacts, not committed source or a fabricated green baseline.

Final reviewed source packets preserve the UTF132-method scope, CDE77original+4new, all43 B methods/10arguments and original allocation windows. B3's throwing wrapper retains every original error and final actual-worker awaits. Two exact method groups and the five-file literal/token-identical whitespace packet were reviewed; the two resulting long methods received a separate preserving coverage-helper extraction (live type190/method47, adapter type58/method46 by source audit). CQ018's six existing ordered-content assertions preserve all7methods, exact bytes, WAL/positions/reopen and real ManagedCode transfer. Every packet has full hashes/ diff/source maps and strongest plus lead review; none is runtime/test qualification. The earlier lifetime build stopped at ArtifactTransfer CS0029 before UnitTests compilation; the newer source-quality build clears that alias defect but stops at the concurrently edited protected Core prerequisite. Neither one-error cut is an exhaustive current solution inventory. The previous broad formatter's112 findings included protected BlobStorage; that raw static cut is preserved and is not this stage's write scope or a compiler result. Accepted source joins and their open combined gates are in quality plan.

Canonical CI baseline

Dispatched after the task plan: 36930711012, source SHA 9c570f8c33a7a9667507a8e1c0ca68860de3be45, completed successfully. Every required baseline job succeeded:

Downloaded run artifacts to artifacts/code-quality/baseline-ci/; the comparison suite includes smoke, json-1k-c8 and json-16k-c4 JSON/Markdown/CSV results. Run metadata is retained in artifacts/code-quality/baseline-run.json. This run covers existing committed main, not the dirty checkout or new analyzers.

The failed build/formatter rows above are source-specific historical snapshots. Commit 3559225a5f918160e46e32c9a812c3f71790e382 is pushed to protected main. Exact-SHA run 36988949282 failed: the analyzer suite passed 84/88 tests, comparison passed 2/4, and Docker/Aspire RF3 passed 3/23. Governance failed on Ubuntu, macOS and Windows; UnitTests and child-process recovery did not run because the verification jobs stopped at governance. The known span/cohesion, resource naming/image-scope, RF3 invalid-read, and shared-cluster cleanup repairs are being joined. This run does not qualify the candidate; a new full exact-SHA workflow is required. ADR-033 remains Accepted and numeric coverage is still unconfigured.