[botandbang]

What and why

sets delete now deletes a published set's record, and its recording's record (computer.localnetwork.dj.recording under the same key, from sets upload), from the Atmosphere account before it removes the set, the set's record first. The publishing permission set grants delete on computer.localnetwork.dj.set and computer.localnetwork.dj.recording (settings and profile stay create and update), PUBLISH_SCOPES matches, and a new repo::delete_record sends a DPoP-bound com.atproto.repo.deleteRecord through the same session, account and scope checks as writes. The set stays when the session may not delete or the server refuses. Signed out, it deletes the set locally and sends nothing. The confirmation line names the records it removes. The mock PDS answers deleteRecord, and a new end-to-end CLI test walks sign-in (a written session), status check, the three publishes, import from the account, unpublish and sign-out against it.

A set made or edited with its drive unplugged now gets MusicBrainz IDs from the tags of the store's copy of each file, found through the libraries' latest snapshots. MusicBrainzIds::read reads a file whose name gives no type by its content.

Also: settings import says "1 setting" instead of "1 settings".

The swap to jacquard-lexicon's SchemaValidator (LOC-85, #187) is not in this PR: 0.12.2 doesn't check string formats, enums, nullable fields, blob limits or constraints inside unions, so it can't replace the current validator without keeping nearly all of it.

Fixes LOC-86
Fixes #188
Refs LOC-82 (the local part: unpublishing and the end-to-end CLI walk; the real-PDS sign-in and publish checks stay open)
Refs #184

How it was tested

On origin/main cc1e7a9 (after #272 and #273):

• cargo fmt --check, and cargo clippy -p locallibrary -p locallibrary-cli --all-targets -- -D warnings.
• cargo nextest run -p locallibrary -p locallibrary-cli filtered on atmosphere, publish, sets, lexicon, musicbrainz, account, bug, mock PDS, recording and upload tests: 417 pass. The new tests passed 3 runs with --retries 0, and deleting_a_published_set_deletes_its_record_and_recording fails with the recording left out of set_unpublish.
• oxfmt --check on the changed lexicon.

Before the move onto cc1e7a9: cargo clippy -p locallibrary-cli --no-default-features -- -D warnings for the account-gated parts, the help, rpc, docs, schema, settings and store tests (498 pass), and each new test failed once with its fix broken.

Not run: the full suite and Windows clippy.

Checklist

• Tests cover the change
• For locallibrary: apps/locallibrary/CHANGELOG.md has a line under Unreleased, if people using it will notice
• The READMEs, docs/ and --help still say what the code does (apps/locallibrary/docs/rpc/ too, for a sidecar change)

[linear-code]

LOC-82

LOC-86

[github-actions]

Coverage Report for apps/web

Status	Category	Percentage	Covered / Total
🟢	Lines	95.12% (🎯 93%) ⬇️ -0.04 pp	7179 / 7547
🟢	Statements	93.72% (🎯 91%) ⬇️ -0.08 pp	10971 / 11706
🟢	Functions	94.13% (🎯 92%) ⬇️ -0.11 pp	2827 / 3003
🟢	Branches	84.14% (🎯 80%) ⬆️ +0.08 pp	5116 / 6080

File Coverage

No changed files found.

Generated in workflow #629 for commit 822eac1 by the Vitest Coverage Report Action

Size

Compared with main. KiB are 1,024 bytes.

	This pull request	main	Change
Client JavaScript	654.4 KiB	657.1 KiB	-2.6 KiB (-0.4%)
Client JavaScript, brotli	218 KiB	218.9 KiB	-986 B (-0.4%)
Client CSS	89.4 KiB	89.5 KiB	-33 B (-0%)
Client CSS, brotli	12.8 KiB	12.9 KiB	-17 B (-0.1%)

[github-actions]

Rust coverage

From run 617. Both are held to the thresholds in .cargo/config.toml.

Tests	Lines	Functions	Regions
All	97.43% (82793/84973)	96.55% (13193/13664)	95.7% (182084/190270)
Hegel property tests alone	95.56% (81196/84973)	94.26% (12880/13664)	93.66% (178209/190270)