Fix __owner_group for non-setgid setups

[f-barth]

Summary

• Fix regression from commit 7c07d3a (PR Make it easier to run eth-docker with an app user #2545) where __owner_group always uses the directory's group, causing spurious "Fixing ownership" messages on every start when the directory group differs from the owner's primary group
• Check the setgid bit on the directory: use directory group when set (multi-admin setup), owner's primary group otherwise (standard single-user setup)

Test plan

• On a system without setgid on the eth-docker dir (the common case): verify __owner_group equals the owner's primary group and no spurious "Fixing ownership" messages appear
• On a system with setgid (app-user setup from PR Make it easier to run eth-docker with an app user #2545): verify __owner_group equals the directory's group, preserving multi-admin behavior

Fixes #2570

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Fixes a regression in ethd startup ownership handling by selecting the correct __owner_group depending on whether the project directory has the setgid bit enabled (multi-admin setup) or not (single-user setup), preventing repeated “Fixing ownership” noise.

Changes:

• Detect setgid on the eth-docker directory and use the directory’s group only when setgid is set.
• Fall back to the owner’s primary group when setgid is not set.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

ls -ld . | awk ... is now executed twice (once for __owner, again for __owner_group when setgid is set). Consider capturing the ls -ld . output once and extracting both fields from that single value (or using stat) to reduce duplication and keep owner/group derived from the same source.

[yorickdowne]

I appreciate the thought, Florian, I do. But this does not express the intent. For example, if the directory is user:admin-group and the user is user:user, and there is not setgid, then your code would cause an adjustment of dkg_output to user:user, which is not desirable:

if find .eth/dkg_output \( \! -user "${__owner}" -o \! -group "${__owner_group}" \) -print -quit | grep -q .; then
    if [[ "${__cannot_sudo}" -eq 0 ]]; then
      echo "Fixing ownership of .eth/dkg_output"
      ${__auto_sudo} chown -R "${__owner}:${__owner_group}" .eth/dkg_output
      ${__as_owner} chmod u=rwx,go=rx .eth/dkg_output
    else
      echo "Ownership of .eth/dkg_output should be fixed, but this user can't sudo"
    fi
  fi

I am currently iterating on this multi-user code in #2575 , and just handling the check for whether .env needs to be adjusted better.

Give me a bit of time to test that thoroughly, and for pietjepuk to review it when it's ready. In the meanwhile, please chown -R user:user <eth-docker-dir> and don't do a user:admin-group setup quite yet.

I'll close this PR once I have something that works in a multi-user setup and doesn't break the typical user:user setup