3 ms·

"50,000 freshly written and barely tested lines of code"... it sounds like they're doing it wrong.

notimetorelax 14y ago
Considering their success so far I wouldn't be so sure.
jaggederest 14y ago
In my opinion, code is never fully tested unless it's produced under extreme clean room conditions (i.e. autopilots, nuclear plant software, space shuttle softwware), or has been used in production for a fair amount of time. Even if your code has full path coverage, you're almost certainly going to interact with an unpredictable outside world that can cause code to fail in novel and interesting ways.
- DanBC 14y ago
  > Even if your code has full path coverage, you're almost certainly going to interact with an unpredictable outside world that can cause code to fail in novel and interesting ways. Yes. (http://www.f-22raptor.com/news_view.php?nid=267 http://www.f-22raptor.com/news_view.php?nid=267) > Don Shepperd told CNN Television that the onboard navigation, communications and fuel systems crashed as the planes crossed the International Date Line. > The problem seems to have arisen not from the time change, but from the change in longitude from W179.99 degrees to E180 which occurs on the International Date Line. > The USAF refused to specify the cause of the issue saying only that the aircraft "experienced a software problem involving the navigation system en route from Hickam to Kadena". The Wikipedia article List of Software Bugs has some other scary examples of billion dollar systems destroyed, or people killed, by bugs. (http://en.wikipedia.org/wiki/List_of_software_bugs http://en.wikipedia.org/wiki/List_of_software_bugs)
- scottbartell 14y ago
  "barely tested" seems to imply that they... barely... have written any tests at all. While I can understand that even full coverage doesn't guarantee anything, it sure can help. I'm not advocating full test converge, I'm just saying that if I wrote 50k lines of code, I really wouldn't want to be able to say that I have "barely tested it".

Considering their success so far I wouldn't be so sure.

In my opinion, code is never fully tested unless it's produced under extreme clean room conditions (i.e. autopilots, nuclear plant software, space shuttle softwware), or has been used in production for a fair amount of time. Even if your code has full path coverage, you're almost certainly going to interact with an unpredictable outside world that can cause code to fail in novel and interesting ways.

DanBC 14y ago
> Even if your code has full path coverage, you're almost certainly going to interact with an unpredictable outside world that can cause code to fail in novel and interesting ways. Yes. (http://www.f-22raptor.com/news_view.php?nid=267 http://www.f-22raptor.com/news_view.php?nid=267) > Don Shepperd told CNN Television that the onboard navigation, communications and fuel systems crashed as the planes crossed the International Date Line. > The problem seems to have arisen not from the time change, but from the change in longitude from W179.99 degrees to E180 which occurs on the International Date Line. > The USAF refused to specify the cause of the issue saying only that the aircraft "experienced a software problem involving the navigation system en route from Hickam to Kadena". The Wikipedia article List of Software Bugs has some other scary examples of billion dollar systems destroyed, or people killed, by bugs. (http://en.wikipedia.org/wiki/List_of_software_bugs http://en.wikipedia.org/wiki/List_of_software_bugs)
scottbartell 14y ago
"barely tested" seems to imply that they... barely... have written any tests at all. While I can understand that even full coverage doesn't guarantee anything, it sure can help. I'm not advocating full test converge, I'm just saying that if I wrote 50k lines of code, I really wouldn't want to be able to say that I have "barely tested it".