I think we check if they are the same as before, not if they are not null, so we may print null there. I wouldn't comment only because it's a debug print.

If the option might be null, lets use .getOrElse("_"). We replace the unset-value with a _. Or use none if you prefer a word.

I wouldn't comment only because it's a debug print.

IMHO, that doesn't matter. The code must be read and understood nevertheless.

If the option might be null, lets use .getOrElse("_"). We replace the unset-value with a _. Or use none if you prefer a word.

Consider we are already printing null for the other argument:

Java version changed (Some(/opt/homebrew/opt/openjdk@17) -> null), re-starting server

IMHO, I would fallback to the real used ones on the rhs, e.g. BuildInfo.millVersion and system.

I changed it to use 0.0.0 for the unknown last Mill version and <system> for the unspecified JVM version. @lolgab WDYT?

Maybe <unknown> instead of 0.0.0?

Thank you! @lefou