Reason:
- The non-bulk notice regression test should exercise the real resolve-by-newest call path.
Changes:
- Resolve a mocked conflict through `_anyResolveConflictByNewest` and assert that its success notice remains visible.
Reason:
- Bulk newest resolution should not display one success notice per file, while non-bulk resolution must retain its existing notice.
Changes:
- Pass an explicit notice flag only through the bulk resolution path.
- Cover bulk suppression, non-bulk notices, and ten-file progress updates.
Bulk newest-revision conflict resolution created one notice per resolved file
and updated progress for nine out of every ten files.
Keep per-file bulk resolution success messages in the log without displaying
notices, update notice-level progress once every ten files, and cover both
behaviours with unit tests.
Consume commonlib's delivery lifecycle, classify finite replication entry points, and document why the five-minute inactivity fuse is only a leak safety valve. Record the remote-activity counter correction and preserve continuous replication's unbounded live channel.
Reason:
- A client that remains open during a remote database rebuild can retain the previous Security Seed and upload documents encrypted with the wrong key.
Changes:
- Fetch the remote Security Seed at every replication preflight instead of reusing the process cache.
- Add a regression test and an unreleased change note for issue #1018.
Address the maintainer review on #949: determine the platform through the
plugin's own service layer (services.API.isMobile()) rather than Obsidian's
Platform API directly, matching the existing call in ObsidianLiveSyncSettingTab.
Applies to both PR-introduced sites: the runtime guard (ModuleObsidianEvents)
and the settings-pane toggle (PaneSyncSettings).
The TFile import becomes type-only so deps.ts is no longer pulled at runtime;
the unit test drives the platform through the services.API.isMobile() mock.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replication is suspended when the Obsidian window becomes hidden (document.hidden),
so LiveSync and Periodic stop syncing while minimised until the window is focused.
Add keepReplicationActiveInBackground (default off, desktop only). When enabled, the
window-visibility handler no longer suspends on hide, so replication keeps running while
minimised. Becoming visible forces a teardown before reopen (LiveSync only) so a stalled,
half-open channel is always replaced.
Includes the setting definition (src/lib submodule), a desktop-only toggle in the Sync
pane shown for LiveSync and Periodic, a docs/settings.md entry, and unit tests for the
visibility handler.
- Database fetching (a.k.a. Reset Synchronisation on This Device) on the initialisation now supports streaming and is faster (CouchDB only)
- The database fetching process has been streamlined, and database operations are now suspended until it has been completed
- The initial synchronisation process has been simplified, making it easier to synchronise files with the remote server
- We can select the remote database to fetch from during the initialisation, when there are multiple remote databases configured (e.g. multiple CouchDBs or S3 remotes)
This commit may contains behavioural changes.
- Fix for the issue with corrupted log displays
- Wrap the activeDocument
- Reduced potential type errors and strengthened certain checks
- Made error handling more robust (by rewriting the error class)
vorotamoroz