Comments on: Fight starter #66: The Exercise Suite http://caines.ca/blog/programming/fight-starter-66-the-exercise-suite/#utm_source=feed&utm_medium=feed&utm_campaign=feed Shell-Shocked Ramblings from the Trenches of Software Development Tue, 03 Jan 2012 16:51:44 -0600 http://wordpress.org/?v=2.8.6 hourly 1 By: Cainus http://caines.ca/blog/programming/fight-starter-66-the-exercise-suite/comment-page-1/#comment-427 Cainus Tue, 27 Jul 2010 04:21:23 +0000 http://caines.ca/blog/?p=266#comment-427 I like that question: "How many lines of code does it take to break a program?". That really shines a light on the mindset that's stopping people from unit-testing more doesn't it? I like that question: “How many lines of code does it take to break a program?”. That really shines a light on the mindset that’s stopping people from unit-testing more doesn’t it?

]]> By: Adam Purkiss http://caines.ca/blog/programming/fight-starter-66-the-exercise-suite/comment-page-1/#comment-426 Adam Purkiss Mon, 26 Jul 2010 21:39:23 +0000 http://caines.ca/blog/?p=266#comment-426 I agree 100% ;) Does 100% coverage have any value beyond the proportion of code tested with meaningful asserts validating results? I would argue that the burden of proof is not there, but lies with those who contend it doesn’t have any additional value. I would also add that for type safe languages or strict typing in general, there is another easily avoided error that 100% coverage handles for you: (where arrays start at 0, for example) Integer i[1]; I[1] = 86; Of course this is easy to see here with literals for element indicators. But in large code bases with loops and variables pointing to the n’th element of arrays, having tests that simply execute through code will easily weed out these issues. Pointing to a missing element in an array is a runtime bomb for even the strongest typed languages, and has absolutely nothing to do with validating return values with asserts. ~ How many lines of code does it take to break a program? ~ I agree 100% ;) Does 100% coverage have any value beyond the proportion of code tested with meaningful asserts validating results? I would argue that the burden of proof is not there, but lies with those who contend it doesn’t have any additional value. I would also add that for type safe languages or strict typing in general, there is another easily avoided error that 100% coverage handles for you: (where arrays start at 0, for example)

Integer i[1];
I[1] = 86;

Of course this is easy to see here with literals for element indicators. But in large code bases with loops and variables pointing to the n’th element of arrays, having tests that simply execute through code will easily weed out these issues. Pointing to a missing element in an array is a runtime bomb for even the strongest typed languages, and has absolutely nothing to do with validating return values with asserts.

~ How many lines of code does it take to break a program? ~

]]>