Davyd Norris writes:
> Just supped the most recent source and started a full build from a
> 1.2BETA box, hoping for improvements in the NCR and com drivers. Built a
> kernel and installed it and it came up looking great, then gave me an
> NCR fatal error (command 9 I think) but kept going all the way to login.
> I rebooted and got through ok the next time, and started to build gcc.

Ah, at this point you really, really had to force an fsck and I
suspect you didn't....

> The machine wedged hard about half way through the build, and when I
> rebooted it came up with a ton of errors in my /usr partition (the
> largest one BTW), necessitating a restore of named, dig, zoneinfo, ed
> and a few others.

Not that surprised. You probably had a partially corrupted file system
after the first incident.

Here is the best I can tell you.

1) If you get the machine to boot without spewing error codes, your
   NCR driver/controller is probably (now) fine.
2) If you ever get your machine to spew error codes, be very sure to
   fsck the thing for real at the next reboot.
3) If you were spewing error codes but have ceased to do so, thank the
   gods and cross your fingers.
4) The NCR driver is crap. Sometimes we manage to make it less crappy
   for a while, but frankly I rue the day I bought an NCR controller.

Perry