fix: add missing boolean mount flags to fix --flag=value error

[yujunz]

Expand the mountFlags whitelist in both ceMount and eeMount functions to include all boolean flags from the JuiceFS mount command. This fixes the issue where boolean flags like --no-bgjob were incorrectly passed as --no-bgjob=true, causing JuiceFS to fail with "unknown option" error.

Fixes #58

[Dr-Shadow]

Is that AI code ?
It doesn't seem clean to me and that's not reflecting the latest flags on the juicefs mount command.

[yujunz]

Yes, it is generated by claude given the issue as context. It seems getting the flags from document and juicefs mount --help from what I observed.

What is supposed to be the source of truth for latest flags?

[Dr-Shadow]

@yujunz, using AI to generate code for core logic like mount flags without verifying it against the actual source code is problematic.

Even if you used juicefs mount --help as a reference, you can see that some of these flags literally do not exist.

You should refrain from submitting AI-generated guesses. If you want to contribute, you must manually cross-reference the flags with the actual codebase to ensure they are handled correctly.

[yujunz]

cross-reference the flags with the actual codebase

Good point. Please take another look after cross check (by AI 😊)

• EE flag validation: The EE-specific flags (external, internal, gc, dry, flip, no-sync, allow-other, allow-root)
  were not validated against EE source code (not available). These were preserved from the original commit.
• Linux-only flag: update-fstab only works on Linux per the source, but including it is harmless on other platforms
  (JuiceFS will ignore it).

Would you mind contributing to review the change?

[Dr-Shadow]

That's better but honestly I ditched the Docker plugin in favor of a systemd unit service for now since I had other issues.

For example if I want to reuse the same volume 4 times on different services, it mounts the volume 4 times (4 juiceFS processes with each one consuming its own RAM for caching).

I wanted to use --all-squash 1000:1000 as well and it was not supported.

I will open new issues to reflect the different problems I encountered and I would happily test it later.

The big problem with the docker volume plugin in its current state is that it won't translate the mount flags very well and the --flag=value is a very specific case (all-squash is another case).

So it would be probably better to allow another way to pass flags (for example a mountFlags attribute string which would be entirely passed to the mount command).

The problem you will have with the current method is that you will have to maintain the flags on each juiceFS release which can add flag parameters.

Before :

volumes:
  app_data:
    name: app_data
    driver: juicedata/juicefs:1.3.1
    driver_opts:
      name: "app_data"
      metaurl: "rediss://:${REDIS_PASSWORD:?}@${JUICEFS_ENDPOINT:?}:6379/1"
      storage: "minio"
      bucket: "https://${JUICEFS_ENDPOINT:?}/app_data"
      atime-mode: "noatime"
      attr-cache: "1"
      entry-cache: "1"
      dir-entry-cache: "1"
      cache-size: "1024"
      free-space-ratio: "0.1"
      buffer-size: "128"
      no-usage-report: "true"

After :

volumes:
  app_data:
    name: app_data
    driver: juicedata/juicefs:1.3.1
    driver_opts:
      name: "app_data"
      metaurl: "rediss://:${REDIS_PASSWORD:?}@${JUICEFS_ENDPOINT:?}:6379/1"
      storage: "minio"
      bucket: "https://${JUICEFS_ENDPOINT:?}/app_data"
      mountFlags: "--atime-mode noatime --no-bgjob --attr-cache 1 --entry-cache 1 --dir-entry-cache 1 --cache-size 1024 --free-space-ratio 0.1 --buffer-size 128 --no-usage-report"