Coral

Appearance

Release compliance checklists ​

Historical release-compliance evidence retained for maintainer reference.

ReleaseState
0.2.4Published and verified
0.2.2Closed without publication

0.2.4 published release ​

  • [x] Root, six workspace manifests, internal pins, lockfile, and MCP server report 0.2.4 atomically; GraphIR remains 0.2.0/schema 1.0.0.
  • [x] The Slice 60 design narrows R-039 to a compatibility shared engine, conservative opt-in profiles, an editor-only worker, and focused performance gates; independent design review accepted with no medium/high findings.
  • [x] Shared ELK and 500-node drag/zoom budgets pass initial exact-toolchain calibration on the attributable controlled release machine with 20 samples, zero browser retries, and recorded raw evidence. GitHub-hosted shared runners are explicitly non-authoritative for hard timing budgets.
  • [x] Final committed candidate reruns both performance gates on that controlled machine with exact SHA/tree, commands, environment, raw samples, and zero retries recorded.
  • [x] Exact Node 24.19.0/npm 12.0.2 clean install, lint, typecheck, builds, all unit/coverage/browser/API/consumer/native/package/doc/diff gates pass.
  • [x] Exact npm audit --audit-level=high exits zero with three moderate transitive findings and no high/critical finding; no breaking force-fix was applied. CI and registry-consumer audit evidence remain separately required.
  • [x] Guarded OIDC commit/version checks, contiguous partial-publication recovery, structural npm provenance validation, registry-only verification, and adversarial closure tests pass independent review and full verification.
  • [x] Release-branch CI run 34562701499 passes the then-defined ten jobs on exact candidate cb1440e44f924e9b7a6a53bb34e521952dc0ee8a: maintained Linux/macOS/Windows, Windows full verify/consumer smoke, pristine Chromium 59/59, historical hosted performance, API compatibility, React 18/19, coverage, audit, packages and docs.
  • [x] NVDA/Windows and VoiceOver/Safari are explicitly experimental and untested for 0.2.4; their retained optional script is non-blocking and no automated result is represented as screen-reader evidence.
  • [x] Final candidate push run 34598566896, PR run 34599676223, and merged main run 34600831120 pass all nine deterministic jobs; independently audited PR/main Chromium runs are 59/59 without retry or flaky annotation.
  • [x] OIDC publication, registry-only 0.2.4 verification, tag, GitHub release and issue closure complete in their real environments.

The 0.2.4 release is complete. Controlled-machine timing evidence and external CI evidence remain distinct; neither substitutes for the other.

0.2.3 published baseline ​

  • [x] @graph-ir/core@0.2.0 is published, public, and installable with the expected registry integrity.
  • [x] A clean checkout runs npm ci on Node 24.19.0/npm 12.0.2 using only the public registry dependency.
  • [x] npm run verify succeeds on Node 24.19.0/npm 12.0.2: 1,700 tests pass and one intentionally skipped MCP transport test remains documented.
  • [x] Tree-sitter generation and 27 corpus tests succeed on the local release platform.
  • [x] Vite output contains the lazy Whisper worker and required WASM assets.
  • [x] npm audit --audit-level=high reports no high/critical advisory on the final lockfile.
  • [x] Each public package passes npm pack --dry-run.
  • [x] Public blank-consumer ESM/type smoke passes for every public package on Linux x64 using public @graph-ir/core@0.2.0.
  • [x] Native install smoke passes on Linux x64.
  • [x] Registry-backed clean install, all 1,700 tests, blank-consumer checks, and native install smoke pass on Windows x64 in the maintained on-host VM.
  • [x] Native install smoke passes on macOS 14 arm64 in CI.
  • [x] A clean-source Windows x64 VM run passes toolchain validation, lint, typecheck, production builds, 1,700 unit tests with one intentional skip, and all 27 Tree-sitter corpus cases using public @graph-ir/core@0.2.0.
  • [x] All 17 Chromium tests pass, including downloaded content, raster decode, conversion-loss reporting, and recovery geometry.
  • [x] Coverage gates pass for viz and MCP.
  • [x] Requirements, architecture, roadmap, public docs, and package docs use the unpublished-0.2.2 / first-supported-0.2.3 boundary.
  • [x] The five 0.2.2 release-review defects retain permanent regression tests.
  • [x] PDF, Node 26, and TypeScript 7 are assigned to 0.3.0.
  • [x] npm confirms coreyt100 as owner of the coral-viz organization; all scoped Coral packages use @coral-viz/*.
  • [x] Revalidate the renamed Coral artifacts with the clean install, full unit, coverage, browser, documentation, audit, pack, consumer, and native Linux gates.
  • [x] Publish and registry-verify all five renamed Coral artifacts. A blank registry-only consumer passes runtime, type, executable, dependency, and native-parser checks with zero audit findings.

0.2.2 remains closed without publication. This checklist governs the first supported/public release of its repaired and extended feature set.