mp3 files with special Char in Filenames or Folder names are not displayed in UI

[Fabien-35235]

This may be a duplicate of #5100.
However, no exceptions were found in log files.

Description

Files with special characters (éàèç...) in filename or directory name do not appear on the embedded UMS UI (port 9002), and are not shown on any DNLA player that connect to UMS.

This is not linked with bad encoding of the filenames: everything is UTF-8 from the start

The filesystem with mp3 files is mounted from a Linux distribution, e.g. Ubuntu

-Dfile.encoding=UTF-8 is used by the startup script

Example

/musique/French/song1.mp3 is shown
/musique/French/à-chanter.mp3 is not shown
/musique/Français/song1.mp3 is not shown (no folder, no file)

No file with accentuated letters is shown.
All files with regular letters are shown and played.
Mode, owner, ACL of files/directories are similar
Changing 1 letter in file/directory enables access , e.g. Français -> Francais

Wild guess

UMS does not support UTF-8 Non Latin-1 characters in the file names, maybe because these are not correctly managed in the underlying database

NB: airsonic-advanced, in the same situation, (same host, similar Docker/Compose file, same mounted directories) provides a correct access to the files, so this is not, definitely, a file access issue.

Setup 1

Using the standard Docker image with a Docker/compose file:

services:
  universalmediaserver:
    image: universalmediaserver/ums
    container_name: UMS
    network_mode: host
    environment:
      - UMS_PROFILE=/profile 
      - PUID=1000
      - PGID=1000
      - TZ=Europe/Paris
    volumes:
      - /opt/my-omv/UMS:/profile
      - /srv/dev-disk-by-uuid-96c47f8d-03d0-40a4-99e4-de9cce7679f1/Music:/musique
    restart: 'unless-stopped'

Setup 2:

Making your own Docker, from standard Ubuntu:

services:
  ums:
    container_name: UMS2
    network_mode: host
    build:
      context: .
      dockerfile_inline: |
        FROM ubuntu:latest
        RUN apt-get update
        RUN apt-get install nano
        RUN apt-get install -y default-jre 
        RUN apt-get install -y default-jre-headless
        RUN apt-get install -y mediainfo dcraw vlc mplayer mencoder
        # RUN apt-get install -y lib32z1 lib32ncurses5 libbz2-1.0:i386 libstdc++6:i386
        RUN apt-get install -y p7zip
        RUN apt-get install -y wget
        RUN wget https://github.com/UniversalMediaServer/UniversalMediaServer/releases/download/15.3.0/UMS-Linux-15.3.0-x86_64.tgz
        RUN tar xzvf UMS-Linux-15.3.0-x86_64.tgz
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Europe/Paris
    volumes:
      - /srv/dev-disk-by-uuid-96c47f8d-03d0-40a4-99e4-de9cce7679f1/Video:/video:ro
      - /srv/dev-disk-by-uuid-96c47f8d-03d0-40a4-99e4-de9cce7679f1/Music:/musique
    command: /ums-15.3.0/UMS.sh
    restart: 'unless-stopped'

Extract from the logs:

INFO 12:46:20.937 [main] Build: https://github.com/UniversalMediaServer/UniversalMediaServer/commit/33887b553c47b4b4a624b62a2aa565dc3f4fd1a2 (2025-12-11)
INFO 12:46:20.937 [main] Working directory: /ums-15.3.0
...
INFO 12:46:21.178 [System Information Logger] System information:
OS: GNU/Linux Ubuntu 24.04.3 LTS (Noble Numbat) build 6.12.57+deb12-amd64 64-bit
JVM: OpenJDK 64-Bit Server VM 17.0.17 (64-bit) by BellSoft
JVM args: -Xmx1280M -Xss2048k -Dfile.encoding=UTF-8 -Djava.net.preferIPv4Stack=true -Djna.nosys=true
Language: English (United States)
Encoding: UTF-8 | resolved UTF-8 | native ANSI_X3.4-1968 | jnu ANSI_X3.4-1968 | stdout ANSI_X3.4-1968 | stderr ANSI_X3.4-1968
CPU: Intel(R) N100 with 4 cores (unknown)
Physical Memory: 15.4 GiB
Free Memory: 11.9 GiB
Maximum JVM Memory: 1.3 GiB

[SubJunk]

I named a mp3 file à-chanter song.mp3 and it appeared on my UMS interface. We really need the full trace logs like @Mik-S-UMS said in his comment #5100 (comment)

[Eskell106]

The reason for this behavior is incorrect jnu encoding (ANSI_X3.4-1968 in your example). sun.jnu.encoding charset is used for file paths (check https://openjdk.org/jeps/400), so setting file.encoding isn't enough. In JDK 17 you can't set jnu encoding manually, as it's taken from default system charset. You need to set locales in your container explicitly (both your build and official UMS images rely on ubuntu images, so configuration can be done the same way). Just add the following variables to environment section:
LANG=C.UTF-8
LC_ALL=C.UTF-8

When you see that jnu encoding had changed to UTF-8 in startup log, everyting should work as expected.

This also fixes same problems with other non-ascii file paths.

[SubJunk]

Thanks for your comment