Code Analysis Six Language Candidate

Six-language source investigation candidate

Draft validated, not Labs: #1263, final source/package head 68f54478ae16d824e43198ac94d6e05746510edf. Actual native stack #1262: JS#1260 → TS#1261 → this candidate, rooted on preserved#1252/93378351d. The cumulative candidate composes original Go#1256/Rust#1258 histories (including frozen#1255), independent Java#1259 and existing README/DSH repair#1242. Python and team/A1/A2 ancestors remain. Seven exact heads are independently asserted as ancestors in the receipt.

These are local Git composition merges, not GitHub PR merges. Original authors, commits, PR ownership and separate Go/Rust stack#1257 are preserved. Maintainers must reconcile combined/original overlap before integrating. No PR approval, formal merge, release or live installation.

One more evidenced source step

Before, the validated language implementations lived in separate artifacts with overlapping shared consumers. After, one source CLI and one regenerated ZIP can investigate Cobra doc_test → doc.GenMan → fillHeader, regex explicit module/use/workspace paths, SLF4J's explicit static class path, Prettier's stable default function, Redux's configured source alias and Click/Jinja's ordinary local imports.

The limits remain visible: Cobra io.Writer.Write has no selected implementation; Rust traits/macros remain opaque or conditional under the captured profile; Redux's barrel stops at src/index.ts and remains a known-target miss. These are not whole-project call graphs or complete change impact. The candidate preserves each language's semantics and freshness checks rather than translating all languages into one guessed dispatch rule.

Fixed samples and finite results

Original source-first manifests and version pins remain in their owning reports, linked from the cumulative replay. This round replays those already exposed samples; it is not a new blind or independently additive evaluation.

Language	Correct / wrong / missed known targets	Preserved limitation
Go	9 /0 /0; zero omitted	5reasonableunknowns. Cobra changed GenMan definition7knownreferences/side. Free functions; methods/interfaces remain unsupported.
Rust	9 /0 /0; zero omitted	2reasonableunknowns. regex/serde_json paired outputs remain partial with actual nonzero exits. Recorded cfg/features/workspace scope; no macro expansion, trait dispatch or full Cargo compilation.
Java	3 /0 /0	Interface/provider selection remains unknown. Explicit bounded class-qualified static paths only.
JavaScript	2 /0 /0	2reasonableunknowns; real Prettier loader1knownreference/side, partial. Factory/callback identity not guessed.
TypeScript	6 /0 /1	1reasonableunknown; known Redux re-export/barrel target remains a miss. Not runtime alias loading.
Python	5 /0 /0; zero omitted	5unknowns. Click/Jinja ordinary local from-imports; member/dynamic/decorated targets unsupported.

Do not aggregate these overlapping selected observations as general precision/recall. Prior independent compiler audits remain attributed to their original heads; they were not relabeled as new successful compilation.

Reproduce and package parity

Run each owning benchmark's public CLI recipe with the candidate source entrypoint and then the extracted ZIP entrypoint. Full commands/exits/configuration/profiles are retained in each *-final-source.json/*-final-zip.json; the report gives all six invocations and dependency/toolchain requirements. Projects are never executed.

python3 benchmarks/code-analysis-languages-combined/verify.py \
  --work /runs --source /candidate --package /extracted --out /receipt.json

Expected replay layout: /runs/combined-<language>-source and -zip, plus both canonical combined-package-one.zip and combined-package-two.zip. 17 raw-facts captures byte-identical;20 saved/paired questions agree after only install/output path normalization. Check commands explicitly use --fail-on never; Rust/JS partial paired queries keep their nonzero exits. Coverage/unknowns are not suppressed to force success.

Final parity and preserved ancestry. Final ZIPsha256 aa74920538583e62d8494ba9244efd031b192b45899f685009547e5a69018f89; two canonical Node22.23.3/zlib1.3.1-e00f703 builds equal,168trackedmodulefilesmatch. Runtime macOSarm64/Node26.12, installedGo1.27.2/Rust1.89.0; no extra toolchain or live installation. No local Windows execution claim.

Separate validation states

• **Local source:**373/373pass,0fail,0skip on both Node26.12 and officialNode22.23.3. The earlier local installed-config failure was a test setup defect: compiler read /private/var, hook compared /var, so no mutation occurred. Controlled probe proves that canonicalizing only the fixture root makes mutation run and the analyzer reject changed input with extract/source-changed. Assertions/producer unchanged.
• **Extracted ZIP:**call/question/paired/native/Python/Java/configuration91/91pass,0skip. Full373tests attempted:372pass/1failure because the self-bootstrap test explicitly requires its own Git baseRevision inside a Git-free archive. This test-environment limitation is recorded separately, not hidden by the focused pass; normal non-Git analysis metadata remains null by design.
• **Other local:**unchanged core19/19, generated/identity/golden checks, README14pass/1Windows-onlyskip and DSHdocs7/7 reused at15c9cf5e3. Existing#1242repair preserved; no new diagram-generation or visual-quality claim.
• Remote final head:CI and DSH integration completed success at68f54478ae16d824e43198ac94d6e05746510edf. Actual analysis log 373/373pass,0skip. Both full regression shards, browser/WebM, ZIP, three-platform packages and Windows jobs passed; deployments skipped. This validates this Draft composition, not Labs integration or successful upstream compilation. Older#1261 remains terminal-failed overall despite353/353analysis; this candidate cannot change that history.
• **Integration/release:**Draft only. None of these candidate capabilities has newly entered Labs. No approval, formal PR merge, release or live installation.

Next: observe final-head remote checks, then continue bounded real source-known misses and wrong targets across supported languages. Go concrete receiver calls and Rust method/trait boundaries merit separate source-first questions; Redux re-export traversal remains a visible known miss.

Go method follow-up

Dependent Draft #1264 adds concrete Go method evidence above this unchanged cumulative head. Its own package, attribution and final-head checks are recorded separately. The original table above remains evidence for #1263, not a rewritten result for the later model.