[self-interest] UnitTests

David Goehrig dave at nexttolast.com
Tue Jan 11 20:53:04 UTC 2011


On Jan 10, 2011, at 5:45 PM, Randy <randy.smith at gmail.com> wrote:

> Creating an otherwise low-resistance language with the property that you can't create anything without a test would offer an interesting check of his hypothesis.  Is there such a language? If not, is it even possible? (Who tests the tests?)

Funny you should say this, because I know some developers who have been trying to do just this with their PHP tool chain. Their version control will not even check out code if the unit tests don't pass. They have tried to enforce 100% code coverage too, which has the obvious result being that none of the tests actually catch any actual bugs because touching a line is deemed testing it. 

This has not improved code quality, test quality, or reduced frequency of production failures. It has meant that the code base is now 2x as large and work often stops when someone manages to check in code that breaks someone else's unit tests. 

This does not mean that such a tool are 100% wasteful as they have prevented a lot of buggy code from entering production. If only by the virtue of preventing most development from occurring in the first place. 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.selflanguage.org/pipermail/self-interest/attachments/20110111/7f6f8f09/attachment.html>


More information about the Self-interest mailing list