4.6.0: comments keep their place; renderComposer returns { setValue }

[travisccook]

DroidNet's WCB stored-command value field is becoming a native textarea that stays in step with the row's builder in both directions. That needs two library changes.

Engine

• Comments keep their place. parseWCBValue turns a *** fragment after an empty segment (^^***) into a standalone note step instead of the previous step's label; buildWCBValue writes a note step that follows another step as ^^***text (a note first in the value stays ***text).
• A comment parsed from a single ^ after a delay, a labelled step or another note is marked inline: true and written back with one ^, so A^***c^***d, ;t500^***wait and ***a^***b still rebuild byte for byte.
• A comment directly after a command or raw step (^***) is still that step's label. The WCB skips empty segments and comments alike, so the firmware runs the same commands.
• + Note after a step now writes ^^*** instead of ^*** (which read back as the step's label).

UI

• renderComposer returns { setValue(value) }: a no-op when the value equals the compiled steps; otherwise it re-parses, cancels a step being edited in the add bar, and re-renders the step list, length and safety note, keeping a half-filled add bar. It never calls onChange. Callers that ignore the return value are unaffected.

Tests

• test/comment-steps.test.js: standalone comment parse and build, and the round-trip list (inline comment, comment on its own line, comment first, two comment lines, comment after a delay, a builder note after a command).
• test/ui-set-value.test.js (jsdom): the handle's no-op, re-render, Add bar kept, edit cancelled, no onChange. Adds jest-environment-jsdom@^29.7.0 as a dev dependency.
• The catalog-example and fixture round trips still pass.

npm run validate clean; npm test green.

🤖 Generated with Claude Code

https://claude.ai/code/session_01TpbbHNL9m8Gw9PAjSjiW3C

[github-actions]

Review summary (advisory)

Checks

• npm ci: OK (npm audit reports 5 vulnerabilities in dev dependencies; not blocking).
• npm run validate: PASS. Every board and the cross-file merged catalog show ✓. The only output besides ✓ lines is the non-standard-category warnings, which were already there.
• npm test: PASS. 17 of 17 suites and 498 of 498 tests passed.

Scope: No board files under libraries/boards/ changed, so the board-contribution rules (examples, placeholders, enums, unique ids, confidence) don't apply here. There is no safety-class mislabeling and no storedcommand issue, because no command definitions were touched.

Versioning: libraries/manifest.json:3, package.json and releases.json all move to 4.6.0 and agree with each other. A minor bump fits: the engine's note-serialization output changes and renderComposer gains a return value. No ids, templates or params change. releases.json releasedAt is 2026-10-02.

Engine (src/droidnet-command-library.js)

• It stays DOM-free.
• parseWCBValue and buildWCBValue now treat ^^*** as a standalone note and mark notes parsed from a single ^ with inline: true. Walking through the cases (label, note after a delay, note after a labelled step, note after a note, and a leading note), the parse→build round trip is byte-identical. The tests in test/comment-steps.test.js cover this.
• The only deliberate non-identity is that an empty segment is dropped, except the one in ^^***. The docs and releases.json already say this.
• Minor point: a hand-built {type:'comment'} step without inline that follows a step now serializes as ^^***. That is intended, since it fixes the old ^*** re-parse as a label.

UI (src/droidnet-command-library-ui.js)

• It still goes through buildWCBValue and parseWCBValue.
• cancelEdit() (~line 166) removes the duplicated reset code, and every exit path uses it.
• setValue (~line 388) doesn't call onChange, does nothing when the value equals the compiled one, cancels any edit in progress, and keeps a half-filled add bar. This matches the new docs in docs/INTEGRATION_GUIDE.md.

Other: package.json adds jest-environment-jsdom as a devDependency for the new UI tests, and package-lock.json is updated to match. This is fine, but it is a large lockfile diff.

No blocking issues found. Looks good to merge from a correctness and safety standpoint.