EToys Tests thoughts [Was: Re: [V3dot10] eToys in 3.10 and beyond]
johnson at cs.uiuc.edu
Wed Jan 17 22:56:06 UTC 2007
> Thanks for your feedback. I completely agree with step 6 breaking on
> refactoring of message names, it should not be a test. I am now thinking it
> could be useful as a utility (run manually) to support the process of
> separating a module (e.g. etoys) from the image. As I hope any separated
> module (etoys) would be loadable, the utility would compare whether messages
> that existed before separation are still there after reloading the newly
> separated module, does that sound useful or am I missing something..
When I suspect that methods are missing, I usually pull up two images
side by side and compare classes in the browser. If I were looking
for a needle in a haystack (i.e. one method out of thousands) then I'd
write a little script to write out the list from one image and read it
into the other. In any case, I'd do it on an ad hoc basis while
trying to generate the module, and wouldn't expect automated tests for
it. After all, my modularizing tools should just *work*!
It is useful to measure the coverage of tests. Which methods get run
by a test, and which are not run? I know there is a Squeak package to
measure coverage, but I haven't used it and don't remember it
off-hand. But if your tests cover most of the methods in a package,
you don't need to write a test to make sure that the methods are
loaded. The tests will do it automatically.
More information about the V3dot10