> I heard a legend of a system, probably not mainframe, on which a
> programmer made the disk error handling routines transient. It worked
> fine until it didn't.
We had a 360/91 with a 2301 drum on selector channel 1. This had
SVCLIB on it...
At some point we got another 2301 and moved a small high usage
local file there (partial parallel jobq). The initialization
code for this "calculated" the number of records per track
by running a channel program like:
CCW read-count
TIC *-8
And at the 2nd index point it got "no record found" leaving the
last (highest) count field in storage. 2301 drums don't
give "no record found" for this, so the channel program just
kept running.
Canceling it failed since it needed a transient area module
and the channel path to SVCLIB was busy...
Looking up the address in the selector channel lights and doing a small
storage change to turn off the chaining bit in the read count CCW freed
things up.
PS: This storage change wasn't typing, lots of switches, one for each bit...
http://www.columbia.edu/cu/computinghistory/img_2631.jpg
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
