On 10/13/07, Robert Burrell Donkin <[EMAIL PROTECTED]> wrote:

> ...but keith's proposal is definitely worthwhile: it's very easy to add
> new features without ensuring that existing ones work. perhaps 0.2
> tasks could be created along the lines suggested...?...

Sure, I didn't mean to dismiss Keith's proposal! Sorry if that's how I
came across.

I think adding more and more precise test cases is key in avoiding
regressions and making sure thinks continue to work as expected.  And
if things have to change, the diffs in the test cases show exactly
what happened.

-Bertrand

Reply via email to