On 1/16/07, Greg Lindstrom <[EMAIL PROTECTED]> wrote:
> <snip>
>
> > > Not to mention that I doubt you have time to do 100% coverage of
> > > the test cases that you have come up with (scriptable or otherwise,
> > > unit or otherwise) before making a commit to the trunk. I'd expect
> > > that your coverage percentage is different based on what type of
> > > change you're committing, and I'd expect that you have different
> > > percentages for when you are preparing an official release.
> >
> >         Be realistic. Just because there isn't 100% coverage and 100%
> > thoroughness and 100% exactness, etc., doesn't mean that things don't
> > get tested sufficiently to determine if they work. There are
> > certainly edge cases that slip by, and these are fixed as soon as
> > they are noted.
>
>
> That's how it's suppose to work.  If some pathological case gets through, we
> write a new test to catch it, then fix the code and verify it with *ALL* of
> the previous tests (plus the new test).  Over time, we have developed a test
> suite with hundreds of tests cases.  Sounds like you have it right, Ed.
> --greg

He originally stated that he expected everyone to run every possible
combination of inputs before committing a change but later clarified
this and  described a process with a limited set of tests that had to
pass before something is committed to the trunk. His first statement
was unreasonable. I objected. The clarification was reasonable.

-- 
sheila

_______________________________________________
Post Messages to: [email protected]
Subscription Maintenance: http://leafe.com/mailman/listinfo/dabo-users

Reply via email to