On Thu, 16 Jan 2003, John Klos wrote:

> Yes; the 4000/60 that I have doing binary package bulk builds uses
> softdeps - no problems, and it's been compiling straight for more than a
> month.

Bah! Mine was running for two months before the ST1480N exploded under
the load of swapping.

> OTOH, partitions > 4 gig are iffy...

Ah yes, this still on my todo list of problems to characterize and
document better before submitting a PR.  Those >4GB partitions are evil.
I've got a couple of Seagate 15150N drives which are reported at
precisely 4095MB in size by the kernel at boot time.  Not a single
hiccup on any of them.

I've a few other slightly larger drives that in the range of
4100MB-4200MB in size.  If I build their fs under 1.5.2 and later boot
into 1.6, the boot time fsck under 1.6 sees the filesystems as corrupt
and wipes out the entire thing.  If I build them under 1.6 and boot to
1.5.2, the same thing happens.  If I build them on 1.6, and only boot on
1.6, they seem to mostly work, until they start filling up with data.

I've seen manifestations of this on at least three different drives,
each a different model, between two manufacturers (Quantum and Seagate),
so I don't think it's a fluke.  There's something amuck in the fs code
for >4GB partitions.

I do think >4GB *drive* support is fine, it's just a problem with the
filesystem sizes.

-brian.