Rick Fochtman wrote: > Consider chewing gum and bailing wire?? :-)
An old war story is always fun, ne c'est pas? I've told this one before here, but it is right up this little side alley Ed Jaffee started. This one involves a missing hammer. I had standalone time on a 360/75 one weekend, from 10 PM Saturday to 6 AM Sunday. Very unusual, but I needed every minute. I had made several passes (and taken several standalone dumps -- on a fast 1403-N1 [1,100 lpm] printer) when it started red-lighting, but only when I ran my new code. Of course a program could not (or should not) cause a machine to halt completely, but I'd already been around the block enough times to know that if you did just the wrong thing, you could elicit some unwanted behavior, and it might be days or even weeks before it could all get resolved and management would realize that the finger of blame should have been pointing at IBM all along. I knew what I had changed in my code, so after a couple of false starts, I had branched around a subroutine that was using some very simple floating point instructions to accumulate some data so as not to have to save and restore any of R0-R15; I found if I didn't call this code then no red light resulted. I figured I could step thru the subroutine instruction by instruction, to find the problem that I had assumed was a bad core way up at the very high end of LCS where I had hidden my data. That quickly got tedious so I formatted the restart PSW at X'000' to start up again after the call to the subroutine, so I didn't have to re-IPL each time. I'd just hit RESET and then RESTART. Back then, boys and girls, there were no special control registers or machine state that needed preserving: RESET would reset virtually everything; RESTART would simply load a new PSW from X'0' and off you could go again, with OS/360 and your program still running. Any pending I/O was toast, however, but there was none going on at that point, anyway. It took a while, but I finally determined that the box was red lighting on an otherwise ordinary AW instruction, and no LCS was involved. Even if I moved the AW to some other location or even if I executed it, the box still red lighted. So it was time to call the CE. By then I had used and mostly wasted 4 hours. The CE that showed up took only 15 minutes to arrive, because he lived nearby. I guess that's why they gave him the call. But he was our CE's 2nd level manager and area specialist, so I was surprised. I started to explain the problem, but he just said "show me." So I re-IPLed and let it red-light. He looked at the lights and just said "Oh! Floating." He then opened the front frame, pulled out the gate, opened a card cover and ran his hands -- ever so gently -- over the ends of the cards inserted into the backplane there. I guessed he was feeling for a hot card or something sophisticated like that. He found what he was looking for, grinned, walked over to his tool briefcase, extracted a HUGE screwdriver, took it by the business end (not the handle), and proceeded to beat the living daylights out of the end of a couple of cards on which his fingers had stopped. I mean he BEAT those cards like he was chopping wood. I was speechless. He stopped, put his screwdriver up, closed his case, put the gate back and closed the cover, put his suit jacket back on (yes, at this time, if an IBM employee showed up at a customer site, they were in a dark suit, white shirt, and appropriate tie, even if it was 2 AM on a Sunday morning), and said, as he was walking away, not even facing me, "try it now," and left before I even had a chance to pick my jaw up off the floor. Of course, the box (but not my code) worked fine after that. I found him in the CE room later that week and, in awe, asked him about it. He didn't seem to think he had done anything slick. He was actually embarrassed that it took him so long to remember on which cards the floating logic was located (the 360/75 was not a microcoded machine -- it was the first machine to implement the S/360 instruction set completely in hardware). I asked him how he knew it was a floating point instruction that was causing the problem, and he simply said "I glanced at the lights." I then asked him about literally beating the cards. He said the connectors got corroded sometimes, but the simplest way to clean them was to reseat them by just pushing them a little bit further into their slot on the backplane. He said if you pulled one out, you risked something else going wrong, so it was just easier to take a hammer to the card cage. He didn't have a hammer in his tool briefcase, so he used the biggest thing he had: the giant screwdriver normally used for frame shipping screws. -- WB ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
