src/cpu/testers/rubytest/Check.cc (Diff revision 2)

You don't need to do %lld with cprintf/warn/dprintf.  %d works just fine.

src/cpu/testers/rubytest/Check.cc (Diff revision 2)

If you're going to panic, why bother with the warn?  I know that you're mirroring what is already there, but it doesn't really make sense to me.  It would probably be nice to make the error messages a little more verbose, but then again, perhaps that should just be a different changeset.

You have a lot of fatals that I think should be panics.  Ask Steve about that.  Also this diff didn't address my comments to your previous diff.

I second Nate: there's fatal() and panic() also take cprintf-style arguments, so there's no point in doing a warn() first followed by fatal() or panic(); the whole message should just be passed to fatal()/panic().  I also don't understand why you kept the ASSERT(0) lines; seems like they should be panics too.

See http://m5sim.org/wiki/index.php/Coding_Style#Fatal_vs._panic for the difference between fatal() and panic() (if you haven't already).  It's a little hard to tell from just a few lines of context which one is correct, but basically anything that's a programming error rather than a user/configuration error should be panic(), so I'm guessing that some of your fatals should be panics based on that.  Also, since the old ERROR_MSG macro called abort(), it's more similar to panic(), so I'd say by default you should use panic unless you know that something reflects a user/configuration error.

This last comment is a little unrelated, but I noticed that several of the ERROR_MSG() calls are followed by return statements even though the program terminates.  panic() and fatal() are marked as "noreturn" so you should probably be able to eliminate those pointless return statements without incurring any compiler error messages.

Assuming everything compiles and runs, it looks good.

Looks good to me, thanks for the changes!