At 1:43 AM -0800 12/12/97, Christopher R. Bowman wrote:
>On Thu, 11 Dec 1997, Henry B. Hotz wrote:

I should have clarified that the following comments about MACE-5 refer to a
completely different disk partition and the binary sets are from a
corresponding snapshot.

>>The last few days (since I got back from vacation) I've been seeing a *lot*
>>of errors on the screen from the MACE-5 kernel which I reported before and
>>think indicate collisions.  MACE-5 now frequently dies after the starting
>>network boot message and the traceback suggests to me that it is running
>>out of memory allocating socket buffers.  If this is still relevant let me
>>know what info you want.
>
>Yes MACE5 is giving me lots of collision errors, but seems to work despite
>them as I have ftp'd all the BETA source and bin tarballs including the
>xsrc which has got to be close to 113 Megs of data with out other problems
>or crashes.  I am recompileing a kernel now on my 660av to see if this will
>go away before I look throught the sources.

DH tells me that MACE-5 is compiled with some debugging flag turned on and
so the errors I see usually would not get printed.  Collisions *are* a
normal event on a CSMA/CD network.  I mentioned it because there was a
change in their frequency coupled with a change in the machine's stability
while booting.

I am currently running with MACE-5 and the -beta binaries and see none of
the three serious problems I reported, just the syslog nattering.  I
haven't stressed the configuration much though.  If building a custom
kernel makes a difference, please let me know

Signature failed Preliminary Design Review.
Feasibility of a new signature is currently being evaluated.
h.b.hotz@jpl.nasa.gov, or hbhotz@oxy.edu