I had never solved this problem, just hoped that I would be able to 
re-compile df with soft-float, and the problem would go away. But it 
got worse:

My 128MB /home partition, with two brand-new users and nothing in their 
directories except .profile, etc., gave me the file system full warning 
when I added a third user. That struck a bell, so I dug in the archives 
and found that someone had warned me previously that the file system 
built by mkfs was funky. So I rebooted to single-user, umount-ed, 
newfs-ed, and mount-ed the /tmp partition, and copied /home off to /tmp 
to do the same to /home. That went well. So I figured I should do this 
with all my partitions.

When I tried copying /var to /tmp, the SCSI drive went into a 
time-out-and-repeat loop. So, after reboot and fsck and getting a 
repeat, I tried shutting down and taking my SCSI MO off line. Got the 
same result with only the one SCSI drive. So I decided to try 
reformatting and starting from scratch.

This time I used the install kernel and set up to install from John 
Klos's ftp site (referred to in Bruce Oneel's pages). Setting up the 
file system goes well, the kernel installs, but on the next tarball 
(base, right?), the SCSI timed out again. Same time-out and repeat loop.

Any suggestions? In particular, is a full 68040 known to help the SCSI 
subsystem on systems where this occurs?

On 2004.4.1, at 09:40  PM, Joel Rees wrote:

> No segfaults! Yeah! I even tested it compiling a four line helloworld 
> with some floating point math decorations, and it doesn't complain. 
> (Okay, that's not much of a test, but it does look promising.)
>
> I just finished unpacking the tarball for pkgsrc into /usr. No source 
> loaded yet. All the non-X stuff from Bruce's softfloat tarballs is 
> unpacked into the tree.
>
> my /usr partition is 2.6 G.
>
> fsck -pf on the unmounted partition gives me
>
>     84203 files, 334772 used, 2076374 free (2078 frags, 259287 blocks, 
> 0.1% fragmentation)
>
> df reports 3806548 1K blocks, 1730174 used, 1695719 available, 50% 
> capacity. (It was reporting 43% capacity before I unpacked pkgsrc, but 
> I didn't write those numbers down.)
>
> Perhaps df missed the softfloat flag for the compile? Or is it 
> something else?
>
> No segfaults! Yeah! I'm going to get busy and get apache running on 
> this thing!
>
>
-- Joel Rees

I had a manager at Radio Shack many years ago
who told me that opinions were like armpits.
Everybody has one and they all smell but your own.

But everybody has more than one,
and the more you cover them up, the worse they smell.