Guillaume Puthod sez:
> Hello,
>
> We are a software service company specialized in testing.
>
> We are looking for tools on Unix and/or Windows that can generate test
> cases for C++ code.
With all respect, the *Unit test rigs generally dedicate >only< to tests
written by programmers >as< they write the features.
We don't do this:
write code -> spend 6 hours debugging it -> add tests to it
We do this:
write a line of test -> get it to fail -> write enough code to pass ->
test everything -> refactor the design -> test everything
This is a serious workflow which many have found to produce rock-solid code
with a very clean design:
http://c2.com/cgi/wiki?TestDrivenDevelopment
But it's actually much more incremental than one e-mail can express. When
adding an ability to code you generally add just enough test lines to fail.
When they pass, after you try to refactor, you add only one or two more lines
of test, see if these fail, and repeat until done. When you refactor (if you
find a reason to, typically duplication), you break it up into the smallest
steps which can themselves be tested.
So there's simply no room in this loop for a tool that would read static
code, as if it never changes, and writes a batch of tests for them. If you
find a customer that actually thinks they need this ability, they must be
re-educated.
--
Phlip
http://c2.com/cgi/wiki?SheChangeDesignInTheDatabase
-- Now collecting votes for a new group:
news:alt.flame.moderated --
|
