> >Hmm...  Try converting to level 3.
> Same problem.

OK.  This makes it sound to me like there is at least one bug in the
filesystem code...

If mkfs is building bogus filesystems, fsck should have detected
something wrong.  It's possible that mkfs is using odd values for
some of the fs parameters and that's throwing the current kernel
off.

If someone has time, this would be a good thing to track down...
I'd probably start by making a couple of small partitions--one with
mkfs 1.45 on the mac-side (using fsck -c 3 under NetBSD to bring it
up to fslevel 3), then making another with newfs, trying to match it as
closely as possible (in dumpfs output) to the upgraded mkfs fs.  If the
newfs'ed fs shows the same odd behavior as the mkfs-made partition, then
there's probably a bug in the kernel having to do with assumptions being
made about the relationships of certain fs parameters.  In that case,
I might try to narrow down which one(s) are causing problems and look at
ways to fix it.

If the newfs fs just works and the updated mkfs just doesn't, then we
need to find out how to make the mkfs-built partition work, and also how
to make fsck detect the anomalous configuration.

If you (whomever has time and energy to go poking around) find anything
or need any advice or a sounding board, drop me a line.  I know Bill
Studenmund is really busy right now, but he'd probably be willing to
talk about this stuff, too--I know he's been learning a lot about the
higher-level fs code lately...  :-)

-allen