Pairing details and delete-versus-edit, 2026-09-24

The composed #141 and #178 changes passed make check with the repository's pinned Node 26.8.2/npm 11.19.1 and Rust 1.98.0: 144 core tests, 369 server tests, two CLI tests, 977 plugin tests, 70 dashboard tests and 767 contract tests. Rust line coverage was 94.68%. The source and history secret scans reported no leaks. The core suite retains its existing one ignored test.

Pairing details passed 17 focused plugin mutation probes (M510–M526) and five server mutations for envelope size, nonce shape, base64 decoding, claim storage and creator polling. Delete-versus-edit passed ten focused probes (M650–M657 and recut M252/M253). Each compiled and was killed by its corresponding behavioral tests. These focused runs do not substitute for the final complete mutation matrix, which is still pending composition of the remaining release changes.

The server keeps the claimant vault details sealed. The creator decrypts and validates them before presenting approval; older claims without details stay compatible. Tests cover a closed modal during decryption and invalid payloads without consuming a claim or enrolling a device.

The delete-versus-edit regression includes two independently syncing clients, a startup upload racing settlement, preserved deletion history, replay, and a later edit. See the research and decision record.

Native pairing

Two isolated profiles of unmodified desktop Obsidian 1.13.7 paired through the native interface. The approving device named the claimant's disposable vault and its 215 notes before approval. The candidate bundle SHA-256 was 301a1d13f38cc799254d5f6871cac8e65d6d5462bf3c37f408dfe0571d6dc544. After approval both vaults held the same 217 files with identical contents.

The screenshot crops the actual approval dialog after the pairing-code display was closed; it contains no code or credential. The quickstart uses this capture to explain what the approving person should check. The later settings refresh fix is covered by modal-close callback regressions.

Native delete-versus-edit

On candidate bundle 7557e642f146659c8dfe3cf0e974f405d8ca904d6364898102ddaf5a92d3efed, the second desktop's sync was disabled through Community plugins. Its native editor appended a sentence while the first desktop deleted the same disposable note through Delete current file. The first desktop had published a tombstone before the second reconnected.

Re-enabling sync restored the edited note on both devices, with the same SHA-256 and no conflict copy. Eleven observations over 300.1 seconds found unchanged file records, unchanged contents and no new copy. A separate same-line offline conflict ended with one copy and both sentences preserved; its real screenshot illustrates the user-facing outcome. No repeating notice was observed. The automated server test, rather than a native screenshot, proves the single current head and retained tombstone.

Current-candidate phone acceptance remains separate from these desktop checks.