Bug: Seafile-cli AppImage 9.0.16 spawns persistent processes and leaves FUSE mounts after start/stop cycles

Seafile Client
1

Hi Seafile team,

I’m seeing what looks like a process leak plus leftover AppImage mounts when repeatedly starting/stopping the Seafile-cli AppImage. Processes never terminate and FUSE mount points accumulate under /tmp/.mount_*.

Environment

  • Seafile-cli version: 9.0.16 (AppImage)
  • Architecture: x86_64
  • OS: Linux Debian Trixie

Steps to reproduce

Run repeated start/stop cycles:

$ ./Seafile-cli-x86_64-9.0.16.AppImage start
Starting seafile daemon ...
Started: seafile daemon ...
$ ./Seafile-cli-x86_64-9.0.16.AppImage stop
$ ./Seafile-cli-x86_64-9.0.16.AppImage start
Starting seafile daemon ...
Started: seafile daemon ...
$ ./Seafile-cli-x86_64-9.0.16.AppImage stop
$ ./Seafile-cli-x86_64-9.0.16.AppImage start
Starting seafile daemon ...
Started: seafile daemon ...
$ ./Seafile-cli-x86_64-9.0.16.AppImage stop

Observed behavior

Each cycle leaves behind a persistent AppImage process plus a loop process that never exits:

$ ps aux

        ...
...    3463203  ...   Ssl  16:54   0:00 ./Seafile-cli-x86_64-9.0.16.AppImage start
...    3463228  ...   S    16:54   0:00 bash -c      while true; do         sleep 1     done                
...    3463264  ...   Ssl  16:55   0:00 ./Seafile-cli-x86_64-9.0.16.AppImage start
...    3463286  ...   S    16:55   0:00 bash -c      while true; do         sleep 1     done                
...    3463316  ...   Ssl  16:55   0:00 ./Seafile-cli-x86_64-9.0.16.AppImage start
...    3463334  ...   S    16:55   0:00 bash -c      while true; do         sleep 1     done                
...    3463368  ...   Ssl  16:55   0:00 ./Seafile-cli-x86_64-9.0.16.AppImage start
...    3463386  ...   S    16:55   0:00 bash -c      while true; do         sleep 1     done                
...    3463453  ...   S    16:55   0:00 sleep 1
...    3463454  ...   S    16:55   0:00 sleep 1
...    3463456  ...   S    16:55   0:00 sleep 1
        ...

Additionally, FUSE mounts accumulate and remain mounted:

$ mount

      ...
/home/.../Seafile-cli-x86_64-9.0.16.AppImage on /tmp/.mount_Seafil4pqIye type fuse.Seafile-cli-x86_64-9.0.16.AppImage (ro,nosuid,nodev,relatime,user_id=...,group_id=...)
/home/.../Seafile-cli-x86_64-9.0.16.AppImage on /tmp/.mount_SeafilpIGwtz type fuse.Seafile-cli-x86_64-9.0.16.AppImage (ro,nosuid,nodev,relatime,user_id=...,group_id=...)
/home/.../Seafile-cli-x86_64-9.0.16.AppImage on /tmp/.mount_SeafiljX0t11 type fuse.Seafile-cli-x86_64-9.0.16.AppImage (ro,nosuid,nodev,relatime,user_id=...,group_id=...)
/home/.../Seafile-cli-x86_64-9.0.16.AppImage on /tmp/.mount_SeafilGuu45L type fuse.Seafile-cli-x86_64-9.0.16.AppImage (ro,nosuid,nodev,relatime,user_id=...,group_id=...)
      ...

Expected behavior

  • stop should terminate all daemon and helper processes.
  • AppImage mounts should be automatically unmounted when the process exits.
  • Repeated start/stop cycles should not accumulate background loops or FUSE mounts.

Happy to provide additional info and suggestions if needed.

Thanks!

2

Hi,

This issue has been fixed in version 9.0.18.

3

Thanks for addressing this issue.

The solution in 9.0.18, I notice, modifies the infinite cycle that was present in the AppRun script as follows

if [ "$1" = "start" ]; then
    exec bash -c '
    while true; do
        pgrep -x seaf-daemon > /dev/null || break
        sleep 1
    done
    ' &
fi

This approach sort of works, but it fails if there are many users on the system, each running its own seafile daemon.

I suggest that the proper solution would be to remove the infinite loop entirely, and to change the python script app/seaf-cli in such a way that it behaves as the AppImage runtime expects. This change is trivial: suffices to add the argument close_fds=False to the call subprocess.Popen in run_argv.

The code is just one line, anyway I’m submitting a pull request. This solution has been tested on Debian Trixie.

4

Hello @vacaboja ,

Thank you, we will consider this solution.