MEMORANDUM

TO:     Bill Campbell
FROM:   Keith Corbett
DATE:   March 18, 1987
SUBJ:   Unix System 5 Patch Tape Testing
COPY:   Ron Genest, Clovis Regis

We have two patch tapes containing fixes to specific Unix problems:

  1) shared memory library [Hakuto needs this fix]
  2) Fortran floating point problems [Bendix needs this fix]

We can't even be sure that these tapes fix the problems as reported.  

I don't have these patches loaded, because we need access to a "vanilla"
system.  This is necessary if we are to have any hope of replicating and
qualifying customer problems.

Therefore, in order to test these and any future patches, I need access
to another Lambda-Plus system.  Obviously I would prefer this to be in
the lab, but having it on our network should suffice.  Unfortunately we
don't have power capacity to add another system, nor do we have space on
the fan-out box to add another system to the network.

What about Ayesha? I don't know who, besides Sandy, uses it anymore.  It
is running an antique alpha version of System 5 (rev 7, not 19).  The
system was kept at that rev by Scott Lander because he had non-standard
software to support the quad video serial port for Uucp connections to
Cambridge.  Presumably this will never again be needed.

My suggestion is that we reload standard software on Ayesha, and use it
to test System 5 patches.  We can restore the users' software (e.g. the
ECO files), and the update should be transparent to anyone using the
system.

For the purposes of configuration testing it might be necessary to add,
temporarily, other peripherals -- Excelan, Iris, and medium-res color
are the options that we have had problems with in the past.

If we can do this soon, we can be ready for testing the complete upgrade
(new boot tapes and /usr files) that Greg McGarry has committed to
prepare.  This update should be ready for alpha testing within 2-3
weeks; after 2 weeks of alpha test we should be ready to distribute it
selectively.

Meanwhile, should we send the patch tapes without any assurance that
they will work?  I believe the risk with Fortran (i.e. to Bendix) is
minimal. 

However, Hakuto doesn't want to upgrade their customers from Version 7
without assurance that their customers will be unable to run their
shared memory applications.  We obviously cannot guarantee this. My
recommendation is that we send Hakuto the patch tape, but tell them it
is up to them to verify that their customers' applications work under
this revision.  [They could test the applications in-house before they
upgrade the customers.]

KMC