Chapter 31

SANDY — Thursday, November 30, 1961

Steve started looping the first failing test case again, and I studied the logic diagrams for a few minutes, looking for a way in which the signal that was associated with the bad connection could affect what was going on now.

I started scoping forward from that signal and, after moving the probe a few times, I said, "Just for fun, let's change this logic card!" and powered down the system. Miguel pulled out the logic card, and said, "Caramba! This looks like something melted here!"

He got a spare card from the cabinet and plugged it in. I turned on the power, and when the system came up, said to Steve, "Let's start at the top, with all of the cases that failed."

He went over to the console, and soon had the programs running. We had made progress; one of the cases that had failed was now running successfully, but there were still a number of failures.

I said to Miguel, "Let's look at that end-op line again!"

He soon had the scope probe set up, and Steve ran all the test cases again. Changing that card got rid of one of the three strange waveforms we had seen on the end-op line yesterday, but it only cleared up the test cases associated with a single byte logic unit control field value.

I said, "OK, let's review what we know about this mess. We had a bad connection that was masking some other failures. We then found a bad component, but that only cleared up one test case. There's still something strange going on here! Steve, what do the cases now failing have in common?"

He looked at his listings for a few minutes. "Right now, the only things I see are the Y branch code and the no-refetch flag."

"Can you give us some additional tests that use those, but do other things with some of the other fields?"

"Sure, but it will take me until lunch time."

"OK, go ahead. We still have a lot of scoping we can do."

Just then, Paul walked in and asked, "What's going on? Steve, I thought you were driving back today!"

Steve said, "That was my plan, but there's already six inches of heavy snow in Maryland, and it's not letting up. Murray told me not to even think of trying to travel today!"

I added, "He's given us a lot of help with test cases and his analysis of what the failing cases have in common. Yesterday we found a bad connection, and a bad component today. Miguel, do you still have that card?"

"Yes, it's over there. I'll get it."

He showed the card to Paul, who studied the logic diagram for a few minutes and then said, "I think you still have another problem, and I suspect that it's something that is putting an unnatural load on this component. If you don't find it, I'm sure this will happen again."

I replied, "I know we have another problem! We still have several of Steve's test cases that are failing. He's about to create some more tests right now, and we should be able to run them after lunch."

"In that case, the most useful thing I can do right now is to get coffee for everyone!"

After we told him how we took our coffee, he went down the hall to the vending machine area and soon returned with six steaming cups.

As Steve continued to scribble code, the rest of us conferred about where we should start to look. I said, "If something is overloading that logic signal, let's try scoping forward from it and see if we find anything. We can split up at lunch time, Steve and I will go as soon as he's finished his coding, and the rest of you can go as soon as we get back. Is that OK with you guys?"

Everyone agreed, and we continued scoping, but we didn't find anything helpful before lunchtime.

STEVE — Thursday, November 30, 1961

As Sandy and I went down to the cafeteria, she told me, "Let's eat down here today. If we take it back upstairs, you'll be punching your programs, and then running them, and won't eat anything until three!"

"Yes, love. You're quite right! Of course, I want to see this resolved as badly as you do!"

"Don't worry! You've given us a lot of help, and I hope we make more progress today, but I'm sure we'll get to the bottom of this mess soon! Besides, this is our last lunch together for several weeks!"

"I know, but remember, it's only a little more than four months, and then we'll be together all the time!"

She gave me such a warm look that it was all I could do not to hug her right then and there in the cafeteria line. We got our food, and found a table in a corner, where I was able at least to briefly squeeze her hand.

I said, "We need to go over this with Mom tonight, but here's what I was thinking about for Christmas. I'll take off Friday the 22nd, and Tuesday the 26th. You did say you'd meet me if I flew up on the Eastern shuttle, didn't you?"

"Of course!"

"I'll probably get one around six on Thursday, then, and I can take a late one back on the 26th. We can go up to my parents on Friday or Saturday, and come back here on Sunday. Do you know what Mary's plans are?"

"No, but Mom will know. She said that she would stay over at Walt's again if Mary's going to be here. However, she may come for New Years instead of Christmas, depending on how things work out with the kids and their father. I'll plan on taking that Friday and Tuesday off as well."

I finished off my salad, and we took our apples back upstairs to the test cell. Sandy looked at Miguel, but he shook his head, indicating that there had been no progress while we were at lunch. He and Aaron then went down to lunch.

I punched up my new programs, and ran them through the assemblers. Just as I finished with that, Sarah walked in. I told her what I had been doing, and how I had modified my Stretch control program to make it easy to select the tests to run and how to run them. I ran the new tests, noting down which ones had failed.

I exclaimed, "This is crazy! All the tests use the no-refetch flag and the Y branch code. I can't make sense yet which other fields are involved and which ones aren't."

Sarah said, "Just out of curiosity, how long has it been since you ran the standard diagnostics?"

Sandy said, "They ran perfectly a week before Steve found the first problems. We haven't tried them since."

"Try them now!"

Sandy loaded the diagnostic tape, and set it to run through all the Reaper tests. There were five tests that failed.

Miguel and Aaron had returned from lunch, and after a fruitless hour of scoping, Sandy said, "OK, let's review what's happened once more. On Monday we found the problem with the fast memory, but we're all pretty confident that isn't related to this problem. We noticed an ugly waveform on the end-op line, but didn't track it down to anything. Tuesday we identified Y branch code 9 and the no-refetch flag as being involved in all of the failures, and Steve told us that the instruction was executing one time more than it should have. I can't help but think that the end-op signal has something to do with that! He also showed us that the byte logic unit control field was involved, and we identified which code points were involved Wednesday morning. Today, we found a bad connection, but when we fixed it, some tests that hadn't failed before started failing."

Miguel said, "Yes, and right after that, I found that bad component."

"Remember that Paul said that was probably caused by something overloading the circuit?"

"Yes, but what would that be? I traced the signal from that circuit in the logic diagrams, and didn't see anything that could do it!"

She paused, and then exclaimed, "Wait a minute! Maybe the design is OK, but there's a wiring error!"

"Nothing failed until the beginning of this week, isn't that what you said?"

"I have it! We need to review the ECs that were applied in the past two weeks, particularly those that affect that area of the machine! We need to check each wire!"

The phone rang, and Aaron answered it. "It's Paul. The fast memory group wants the machine, and they want it for the rest of the day, unless you want to wait three weeks for the fix."

Sandy said, "Tell him OK. We can use the time to review the ECs and then start tracing wires in the morning!" She saw the expression on my face, and added, "I'm sorry about this, I know you wanted to be involved in getting to the bottom of this bug! I'll just say again, without your help this week, we wouldn't be anywhere near as close as we are now!"

I replied, "I understand, I guess, and the fast memory is really important to us. You'll keep me posted, won't you?"

"Of course! I hope that I have good news for you tomorrow night by the time you get home!"

Sandy took a big notebook out of a cabinet, and said, "This is the EC history for the system. We're going to start here, and see which changes are the most likely to have caused the problem. Steve, I'm not sure you can help us much with this."

Sarah said, "In that case, let's go over this test case control program again. I like how you've set it up, but it looks like specifying the various options is a little bit cumbersome, and I've been thinking about a different approach."

We found a table, and were soon absorbed in studying my code. After an hour, she pulled out a large blank piece of paper, and sketched out a scheme for specifying the options that would be easy to use for the most common cases, with more interactive input for the unusual cases.

I said, "That looks good. I wish I had a chance to see it work!"

"You'll be back in January for the acceptance test, won't you?"

"Yes, assuming that they find the problem that's been plaguing us all week!"

"Sandy is a brilliant engineer, and I'm sure that she'll have it completely resolved in well under a week. She was right; without your insight and help, they wouldn't be nearly as close now as they are!"