16 Mar
2017
16 Mar
'17
9:40 p.m.
William Pechter:
VMS source fiche was very common of sites owned by large corporations.
Their IT staff used it to research bugs... and as sample code for
writing their own drivers etc...
=====
Indeed, I used the VMS source microfiche to learn how to
handle various sorts of errors (machine checks, memory
errors) better in UNIX. Stock VAX systems at the time
just crashed on any error, but it turned out that many of
them admitted recovery: some errors were transient,
others could be ridden over by disabling some piece of the
hardware.
This led to an amusing event on the VAX-11/750 that at the
time handled e-mail as uucp node research!. (Its internal
name on our datakit node was grigg.) People noticed that
the system was running slowly. I checked and discovered
that the CPU itself seemed to be a bit slower. Then I
checked logs and discovered that a week earlier, there had
been a cache error; my new recovery code had turned off
the failing half of the cache, logged the error, and forged
ahead.
At the next convenient time, we took the system down and ran
DEC's standalone diagnostics. (Contrary to the rude stories
one hears, those diags were in fact pretty thorough.) The
problem didn't show up, so we booted grigg back up again,
secure in the knowledge that if the problem was persistent,
my code would let us know without crashing. (I don't think
it ever showed up again.)
We also learned to pay more attention to console messages!
Norman Wilson
Toronto ON