Move to stable Effect 4.0.x (0.2.0)

[tamashi095]

Summary

Refs #10 (issue + numbered spec .github/ISSUE_SPECS/010-move-the-codec-to-stable-effect-4-0-x.md).

Moves @misofm/codec from Effect release candidates (peer 4.0.0-rc.112 || 4.0.0-rc.115) to stable Effect 4.0.x so misofm/app can move to the 4.0.4 train. Version 0.1.1 -> 0.2.0. Not tagged, not published.

• Peer effect is >=4.0.3 <4.1; dev dependency pinned exactly to 4.0.4 (the codec imports only effect, there is no @effect/* package); packed-consumer matrix is Effect 4.0.3 and 4.0.4 (was rc.112/rc.115).
• No source change. The dist/ compiled against 4.0.4 is byte-identical to the 0.1.1 build (all .js, .d.ts and maps), so the public API is identical.
• Adds CHANGELOG.md (shipped in the package) with the behaviour audit and the consumer migration note.

Why: with the app on 4.0.4, engine-web-adapter@0.5.13 (hard dependency effect@4.0.0-rc.112) ran a 4.0.4 codec Layer through its rc.112 runtime and failed with Cannot read properties of undefined (reading 'scheduler'). The adapter port is a separate PR (misofm/engine-web-adapter, branch effect-4.0.4) and depends on this one.

Behaviour audit (Effect rc.112/rc.115 -> 4.0.4)

The codec uses Stream.transformPullBracket, Stream.rechunk, Pull.matchEffect, Cause.done, Scope.addFinalizer, Effect.acquireRelease/scoped, Effect.try/tryPromise, Schema.TaggedError/Defect, Layer.effect, Context.Service. No partition, retry, Schedule, timeout or Worker. A differential script was run against {0.1.1 + rc.115, 0.1.1 + 4.0.4, 0.2.0 + 4.0.4, 0.2.0 + rc.115}; all four produced identical output for (one measured exception below): error message/name/JSON/Cause.pretty/Schema encoding (decode, encode and PCM errors, with and without cause/bigint fields); encoded FLAC SHA-256 + STREAMINFO MD5; decoded PCM SHA-256 + Complete; exit reasons for truncation, source failure (error returned by identity), bad options, bad PCM block; decoder interruption while waiting for input (1 Interrupt, source finalizer once); early take (source finalizer once); Effect.timeout on a stalled source; encode interruption with a pending source (1 Interrupt, no late writes).

Known difference: JSON.stringify of a FlacDecodeError whose cause is a SchemaError is now compact (Effect #8895); Schema encoding and Cause.pretty are unchanged, and the only producer is an ABI-violation path the pinned asset does not take.

Noted Effect changes and why they are unobservable here: Scope.close finalizers uninterruptible (rc.118; the decoder/encoder releases are synchronous), pending interruption kept when an uninterruptible effect fails (4.0.1; no uninterruptible region).

Verification (local, Linux, Node 22.23.2, Bun 1.4.2)

• bun install --frozen-lockfile, bun run check (prettier, tsc incl. benchmark project, 40 pass / 5 native-only skipped / 0 fail): passed
• CODEC_NATIVE_FLAC=<pinned libFLAC 1.5.0 built by scripts/build-native-flac.sh> bun run test:integration: 5 pass
• bun run build:wasm:verify with the pinned Emscripten 6.0.9 at /data/codec-tooling/emsdk: encoder 90da75c7... and decoder 70caf381... reproduced exactly
• bun run pack:check (npm pack, isolated packed consumers on Effect 4.0.3 and 4.0.4: compiled TypeScript consumer, Bun, Node, Chromium/Firefox/WebKit workers, browser bundle has no server imports): passed; tarball SHA-256 9a8cfcd254506617458b41fea315879393e5712c230ba8ecd2f971763ae22071
• npm install --strict-peer-deps effect@{4.0.4,4.0.3} <tarball> + npm ls: valid, one deduped effect; effect@4.0.0-rc.115 is rejected with ERESOLVE
• Old-version crash repro (consumer: effect 4.0.4 + codec 0.1.1 + adapter 0.5.13) reproduces the scheduler TypeError; with codec 0.2.0 + adapter 0.6.0 it succeeds (see adapter PR)

Release (after review/merge; maintainers)

Per .github/RELEASING.md and npm-publish.yml: workflow_dispatch on main with mode=publish, expected_sha=<merge commit>, expected_version=0.2.0; then tag/release v0.2.0 at that commit and record the run URL + registry integrity (as in .github/RELEASE_EVIDENCE/006/registry-release.json). The branch effect-4.0.4-ci-dist exists only so the adapter PR can install this codec from git before publication; delete it after release.

🤖 Generated with Claude Code