Fixed podcast sync
Build and Publish / Build and Publish Docker Image (push) Successful in 2m8s

This commit is contained in:
ab committed 2026-11-22 17:11:37 +02:00
1 parent 441316f78f
commit 319f0bcba7
21 files changed
+1067 -276

No files matched your search

+14 -2
View File
@@ -31,8 +31,20 @@ its publications, clear saved audio, or delete the podcast for every user. Audio
retention is disabled by default. When enabled, its day count is measured from the
download time; the periodic job removes old audio while keeping episode descriptions
and listening progress. Removed audio is downloaded again only on an explicit save
or playback request. Podcast personal state is not replicated to trusted Frid devices
yet; device synchronization continues to advertise protocol v3.
or playback request.
Connected devices use Frid device-sync v4 to exchange podcast subscriptions, share
decisions, listening progress and listening-session records. Adding or removing a
subscription on one connected device updates the others; unsubscribe and share-off
registers survive offline devices reconnecting. Older v3 clients keep music sync and
receive no unsupported podcast operations or snapshot sections.
The main player search also finds accessible podcasts by words in their titles,
descriptions and authors. All query words must match; a private subscription is visible
only to its owner. Podcast search and the admin artist/release/track search use the
explicit ICU `und` collation `furumusic_unicode`, so case matching does not depend on
the database's initial locale (including databases created with the `C` locale).
PostgreSQL must include ICU support to apply migration `m_0055_podcast_sync_unicode_search`.
## Access and public sharing
+6
View File
@@ -62,10 +62,16 @@ retention and administrator actions:
```bash
node --test tests/podcasts.test.cjs
FURU_PODCAST_TEST_DATABASE_URL=postgresql://... cargo test podcasts -- --include-ignored
FURU_PODCAST_TEST_DATABASE_URL=postgresql://... cargo test admin_unicode_search -- --include-ignored
```
Use a disposable database. The PostgreSQL test creates and drops a unique schema;
its audio files use a unique temporary directory. Normal podcast tests need no network.
The podcast fixture includes real Frid v4 subscription operations, snapshot exchange,
offline unsubscribe replay, legacy-peer filtering, and private Unicode search. The
admin fixture exercises list counts, all three catalogue lists and the track picker
with Cyrillic and Greek searches. Use a PostgreSQL build with ICU support; running
these fixtures on a database created with locale `C` covers the locale regression.
The access regression suite uses a disposable migrated database and a local OIDC
provider with signed tokens and PKCE: