High Unicode in tags is garbled on save

[slycurgus]

I'm encountering an issue with high-range Unicode characters in song tags as well as in artist/album folder names.

Problem Description: UTF-32 characters are being garbled (re-encoded as a series of �) by Subsonic. I can input them into the text boxes on the tag edit page but on saving and returning to the track list they're � again. The same happens when reading the artist/album folder names from disk - on initial import the artist/album are garbled.
Troubleshooting Steps: Ensured the characters are correct on disk (folder paths) and in the tags. Checked that the user subsonic runs as has Unicode locales set. Not sure what else I can do.
Playback Application and version: Web interface
Subsonic Version: 6.1.3 (build e408c9) – October 31, 2017
Server Version: jetty-6.1.x, java 1.8.0_181, HSQLDB, Linux (97.7 MB / 191.0 MB)
Hardware Platform: Debian Linux (Debian 9.5, "Stretch")
Java Memory Limit: 200MB (in /etc/default/subsonic)
Problem Filename: Unable to reproduce it here - I get a mysqli error if I try to submit the post. You can see the artist, album, and track names at the Bandcamp link below. The filename format is

Code: Select all

/var/music/Music/<artist>/<album>/<track number> <track name>.mp3

Output from ffmpeg -i: Tells me ffmpeg is not found (though playback is fine, it's just display that's the problem)
Last ten log file lines: Nothing relevant appears in the logs. The last 10 lines are (part of) a successful media library scan.

The album in question is available for viewing on Bandcamp (with the title and artist name visible for comparison, depending on your OS/browser font support) at https://lapfoxtrax.com/album/--13.

Very interestingly, the artist name displays almost perfectly on the DSub Android app. So it appears to be stored correctly and just displayed improperly on the web interface. If I reload the tag edit page on web and re-save the tags, however, DSub also displays �s.

I've worked around the issue for now by renaming the artist/album folders with "near enough" representations that are lower in Unicode, but would prefer to be able to have it faithful to the original characters if possible.

[slycurgus]

Some additional information: it seems the "playing" pane at the bottom of the web interface displays the information more or less correctly (it gets one character wrong that DSub gets right, and gets right one character that DSub gets wrong). So it seems it's just the main view that's incorrect; see top-right vs bottom-left of the image below:



Track names seem fine, it's just the artist name that struggles. Album title also seems to be problematic (though as mentioned I've replaced that with clean ASCII as a workaround).

I wonder that it's part of the code that reads directory names from the filesystem? It seems it would be better if the artist/album were taken from the tags instead of from disk....