The cost of a bug [modified]

[DaveAuld]

keyresearcher wrote:

The most expensive bug was obviously the $3 billion bug in the mars lander(rover?) software.

What about the Ariane 5 rocket, $7 Billion in development cost + $500 Million in rocket and payload, and *poof* gone in an instant; http://www.ima.umn.edu/~arnold/disasters/ariane.html[^] That would make it a costly bug?

Dave Find Me On: Web|Facebook|Twitter|LinkedIn

---

Folding Stats: Team CodeProject

[Lost User]

Rage wrote:

So what does DD's car look like

I've deduced it doesn't look like a Bus.

[Lost User]

I bet it was aa engineered 'mistake'.

[hairy_hats]

Rage wrote:

So what does DD's car look like ?

I think this[^] is it.

[Sascha Atrops]

That was the first bug I thought of. Fortunately until now no nuclear power plant was destroyed by a software bug. I could imagine that this would seriously beat the record.

[BrainiacV]

There were two instances when I was doing computer controlled conveyor systems...costs unknown. One system had a bug that was extremely situation sensitive. The end result would be a carton counted twice and a carton disappearing from the accounting. The client had insisted that the tracking information be written to the floppy disk as well as the hard drive. They feared hard drive failure was given as the rationale for that specification. The computer controlled the tracking (my code) and would raise a flag to the PLC (Programmable Logic Controller) that controlled the carton release. The PLC code was to raise a confirming flag when it did so, my program would drop the flag to reset and the PLC was to drop it's flag in response and to show that it was ready. Turned out the PLC programmer took the cheap way out, in that he didn't have any real logic controlling the raising and dropping of his flag. He simply passed the status of the first photoeye as the carton went past. Furthermore, he just used my flag to energize the carton release. As long as my flag was up, cartons would be released. Conveyor systems we had worked on over two years had used this logic, neither of us realizing we were talking at cross purposes. Operationally, it seemed to function as specified, the PC reacting in microseconds compared to the PLC reacting in milliseconds, my flag went up, his went up, mine went down, his went down. Quick aside here, when I was teaching programming, I'd use this as an example of how something as simple as two binary flags could be misinterpreted. Meanwhile back at the plot, it turned out that if a carton was being released AND a carton was verified as sorted or not was being written to the HDD and the floppy drive, the MS-DOS that we were running on top of would seize control and not release until the floppy data was written. During this time the PC would not see the PLC signal drop, kept the release flag high, releasing another box. Carton tracking did not start until the PLC flag dropped. (or in this case, was seen to drop). So the first carton would not be tracked. The label information from the first carton would be read from the barcode scanner buffer and assigned to second carton, and then to prevent any FIFO errors, the scanner buffer would be flushed, losing the label from the second carton. So the second carton would go to where the first one was supposed to go, the first carton, not actively tracked would come around for a second pass, be read again and tracked to its intend

[DaveAuld]

Highly unlikely a software bug could cause a nuclear power plant failure. The design of the systems with both high availability and multiple redundancy and layers of independent protection, backed up by mechanical fail safe mechanisms, means it is highly unlikely it ever could occur.

Dave Find Me On: Web|Facebook|Twitter|LinkedIn

---

Folding Stats: Team CodeProject

[Matt McGuire]

The software I develop is industrial control and automation for cold storage of food, it controls the pressures, temperatures, atmospheric gases of around 140 facilities mostly in the north west of the US, each facility has around 20 to 30 rooms, each room containing 1/2 to 2 million $ in food product, plus must record all control data points in an encrypted format. A lot of these rooms fall under 40 day export; in which the temperature must not reach 32.0f or above (this will throw them out of the program, and have to start over) but the temperature can not get below 30.0f do to freezing conditions. to make the food sleep so they last longer the oxygen in the room averages from 0.5% to 1.5% any lower will kill the fruit, too high the food spoils (19.5% normal atmo), and the CO2 must purged to stay well below 2.0% to prevent scalding. Machine rooms must maintain steady suction and discharge pressure by controlling condenser and compressor equipment. I’m not even getting into the safety controls and backup systems. :omg: So no stress here. Bugs in the interface are forgivable, but not on the control side. BTW I am the only developer here, but we are looking to expand; any takers? long term employment: 11+ years for me.

[FrankLaPiana]

One of the Mars landers had a fairly famous bug - one portion of the code in metric, the other portion in English units. Costs? Must have been in tens of millions. Another case of medical diagnostic machine that had buggy software (CAT scanner? or something similar?) The bug killed three people. "Cost" was in millions.

[FrankLaPiana]

Accounting / Order software - didn't have the logic for tax calculations in CA. Stubbed the code with a "todo comment". Forgot about it... time for QA. California too busy to send us test data. Six months later, the company realized that it had "forgotten" to collect $450,000 in sales tax. Financial trading software. One simple routine to convert an alphanumeric string to an integer. Tested by programmers, QA and customers for over 6 months. ONE customer sends the strings with fractions, "10.0". All other customers send the strings as integers, "10". The new version goes live, and that one customer has immediate problems (10.0 was being converted to 100). After 3 hours of waffling and hand-ringing, management decides to shut down the entire system rather than just the one customer. System was restarted, but all customers automatically switched to other firm's trading systems. By the time customers started using our software again... dunno, must have been millions of dollars in lost revenue. Shit happens, and then you get on with life.

[Lost User]

Worked on an application that was used to control a concrete-factory. An inverted boolean caused the app to dump about four cubic metres of concrete on the floor (as opposed to dumping it in a mixer) No idea what it costed. It had to be cleaned up before we could continue with anything else.

Bastard Programmer from Hell :suss:

[Member 4194593]

I think it was a medical device which could be used as both an EBT (Electron Beam Tomography) and as an XRAY device, they would drop a metal target in front of the electron beam to produce the XRAYs when XRAYs were desired (needed a stronger electron beam to produce the XRAYs than for EBT). It was software controlled, but if an error was made in selecting what was supposed to be used, and if the technician hit backup to correct the mistake, the machine would lift the target (in the process of backing up), but somehow forgot to drop it back down as it was forward spaced to continue the input specification. The resulting beam burned holes in people. Nasty painful death. I think it was a Thorac xxx, but can't seem to find any reference to it on Google (GE may have had a hand in suppressing that info). I do not remember where I read this (many tears ago), maybe Scientific American. Dave.

[James W Taylor]

Remember Chernobyl. No matter the quality of the design, it doesn't work if switched off.

[Member 4194593]

Found it. Google "electron beam accidents" for more info. Around 1986/1987, the "Therac-25". Dave.

[R Erasmus]

I would guess, Y2K bug standing at +-$308 billion

"Program testing can be used to show the presence of bugs, but never to show their absence." << please vote!! >>

[Sascha Atrops]

I am a computer scientist. I do programming for 25 years now, so I saw a lot of code. I know that working computer programs save a lot of lives i.e. in airbags or emergency assistants. I accept programs which take over control in a moment when a human cannot react anymore. In this case the assistant is an additional chance. But I do not trust computer programs, so I prefer systems that cannot arrive in situations of uncontrollable risks.

[DaveAuld]

That wasn't a bug, that was bad original design.

Dave Find Me On: Web|Facebook|Twitter|LinkedIn

---

Folding Stats: Team CodeProject