Darren New <dne### [at] sanrrcom> wrote:
> Did you do this with Hopido?  What parts could you test?

  Interactive graphical programs are quite a lot harder to test automatically
than libraries which take some input and produce some well-defined output as
a result.

  With a library (like a function parser) automatic testing is relatively
easy: The testing program simply uses the library, feeds it some input and
checks what the library returns to see if it's correct (or, in the case of
the function parser, within acceptable limits, as we are talking about
floating point numbers after all, so bit-by-bit equivalence between what
the compiler produces and what the library calculates cannot be expected).

  With an interactive graphical program such automatic testing is not that
easy. You can't simply write a program which calls some functions and checks
what those functions return. You would have to programmatically emulate user
interaction, and somehow check that the program does the right thing.
There's a million things which could theoretically go wrong, and most of
them are things which are very hard if not impossible to check
programmatically (eg. visual glitches, wrong timings, etc.)

  There's also the question of deadlines. Writing a fully automated test
for an interactive graphical program could at worst take even more time
than implementing the program itself, and that's often just not an option,
especially in a commercial project. The cost-benefit of such testing scheme
would be dubious.

-- 
                                                          - Warp

Post a reply to this message