Release validation

A release must validate the language, native runtime, installed packages, editor, and documentation together. This page describes the review requirements; the release process supplies the publication commands and platform jobs.

Review source changes against their intended behavior and contributor standards. Resolve findings with concrete regressions where behavior is affected. A review summary names its scope and remaining limitations; it does not claim correctness for code or platforms that were not examined.

Native redistribution review checks each target’s actual dependency closure, retained notices and corresponding source inputs. A notice inventory copied from another platform is insufficient. The Linux C++ runtime recipe backports the upstream aligned-allocation overflow fix and rejects a missing or changed required dynamic export, including its version, binding, visibility and data size, or a newer libc requirement. Its real native regression checks oversized and ordinary aligned allocations, threads, once-only initialization and asynchronous errors. The exact receipt belongs to the sealed artifact being qualified.

The release gates include type checking, language and runtime regressions, LLVM backend parity, sanitizer circuits, clean installed-CLI/native-library consumers, package and extension checks, executable documentation examples, and documentation drift/site checks. Preserve the exact candidate identity and failed reports when a gate rejects a release. Run the repaired candidate through the required gates again.

August 1.0 supports the three hosts in the platform matrix. The tagged preparation run retains the exact-source producer, consumer and editor results; the release retains the reviewed compiler packs, npm archives, VSIX and checksums. npm publication verifies all four registry archives. The public-install check checks ordinary downloads after publication.

Performance and safety gyms describe their measured scope. HTTP conformance, identity-provider hardening and unsupported targets retain the known limits and roadmap.

The retained qualification records include host-specific conformance, performance, lifecycle and gym reports, plus the exact installed CLI and editor receipts. The tagged consumer jobs used the qualified local compiler archive before publication; the later public-install report separately checks its public download. All results describe finite checks on the recorded hosts.

Preview evidence

The 0.23.0 release retains the compiler packs, npm archives, VSIX and checksums produced from c47626dc75f3822cc78391b1e7697f87022ae2b9. Release preparation passed all three producer and clean-consumer jobs. npm publication verified all four public archives. The public Linux ARM64 report records installation from empty caches, real PostgreSQL and HTTP operations, native cleanup, locked offline reuse and relocated execution. Its runner and fixtures are retained under experiments/native-ingestion-release. This evidence does not replace application acceptance tests.