You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Editing a self created playlist (e.g. changing the name) results that no playlist can't be retrieved anymore. The webui only displays server error 500 and nothing more. The Mobile App displays the same on the playlist page AND the home page. The webui home page works, but displays that no playlist exists (not even the built in ones).
Docker logs of the pinepod and postgres containers are not showing anything when this happens. (Last logs are just from when the containers started)
Only solution i found so far was restoring from backup.
Hey, I wanted to add a little more context based on my experience with how it's happening on my pod. It is definitely true that any edits to a user made smart playlist will result in a 500 error. It seems that the changes are recorded. And it also seems that the system can tolerate this happening a few times before it does give the symptom that the original reporter stated. That is to say it seems that if I edit just one smart playlist and it gets the 500 error it's fine. But if I keep doing it then eventually all my playlists disappear and I have to revert to a known good backup.
I hit this exact bug today and tracked down the root cause via direct DB/log access (Postgres backend). (Investigation and this comment were done by Claude Code on my behalf, not by me directly — flagging that for transparency.)
Editing a user-created playlist through the UI writes NULL into the Playlists.description column (confirmed: my playlist's last_updated timestamp matched the moment the error started). The backend's get_playlists query decodes that column as String rather than Option<String>, so any playlist with a NULL description makes the entireget_playlists response 500 — which is why home page ("no playlists exist") and the playlist page both break at once, exactly as reported above.
Server log (Rust backend, Postgres):
API Error: 500 - Database error: error occurred while decoding column "description": unexpected null; try decoding as an `Option`
Workaround (no data loss, no backup restore needed): set the affected playlist's description back to an empty string instead of NULL, e.g.:
UPDATE"Playlists"SET description =''WHERE description IS NULL;
That immediately restores get_playlists for all users without needing to revert to a backup.
Suggested real fix: either (a) make the edit-playlist endpoint write '' instead of NULL when no description is given, or (b) make the get_playlists query decode description as Option<String> so a NULL doesn't 500 the whole response.
Fixed in #1039 - Episodes missing a title or artwork were breaking the playlist preview. Those are handled now and editing works again. Will be released officially with the next release coming soon.
Activity
djotaku commentedon Aug 7, 2026
Same thing is happening to me. Let me know if there's anywhere I can check in the logs to help
djotaku commentedon Aug 10, 2026
Hey, I wanted to add a little more context based on my experience with how it's happening on my pod. It is definitely true that any edits to a user made smart playlist will result in a 500 error. It seems that the changes are recorded. And it also seems that the system can tolerate this happening a few times before it does give the symptom that the original reporter stated. That is to say it seems that if I edit just one smart playlist and it gets the 500 error it's fine. But if I keep doing it then eventually all my playlists disappear and I have to revert to a known good backup.
kalmatthew commentedon Aug 24, 2026
I hit this exact bug today and tracked down the root cause via direct DB/log access (Postgres backend). (Investigation and this comment were done by Claude Code on my behalf, not by me directly — flagging that for transparency.)
Editing a user-created playlist through the UI writes
NULLinto thePlaylists.descriptioncolumn (confirmed: my playlist'slast_updatedtimestamp matched the moment the error started). The backend'sget_playlistsquery decodes that column asStringrather thanOption<String>, so any playlist with a NULL description makes the entireget_playlistsresponse 500 — which is why home page ("no playlists exist") and the playlist page both break at once, exactly as reported above.Server log (Rust backend, Postgres):
Workaround (no data loss, no backup restore needed): set the affected playlist's
descriptionback to an empty string instead of NULL, e.g.:That immediately restores
get_playlistsfor all users without needing to revert to a backup.Suggested real fix: either (a) make the edit-playlist endpoint write
''instead ofNULLwhen no description is given, or (b) make theget_playlistsquery decodedescriptionasOption<String>so a NULL doesn't 500 the whole response.madeofpendletonwool commentedon Oct 29, 2026
Fixed in #1039 - Episodes missing a title or artwork were breaking the playlist preview. Those are handled now and editing works again. Will be released officially with the next release coming soon.