Re: On white-box tests

[Mike Gran]

On Wed, 2009-08-19 at 15:53 +0200, Ludovic Courtès wrote:
> Mike Gran <address@hidden> writes:
> 
> > On Wed, 2009-08-19 at 10:38 +0200, Ludovic Courtès wrote:

> I just wanted to hear what you and others thought about the issue,
> because I think unit tests are a crucial part of software development.

OK.  To say something slightly more cogent.  I think when a current
phase of development centers around modifying low-level code, it is
useful to have a set of low-level tests.  If those tests fail, it
reminds the developer that s/he has modified something upon which other
routines rely.

I wrote the string-internals tests to indicate to me when I'd done
something that had unexpected side-effects.  They are intentionally
white-box; they are intentionally reflexive.

There is a danger that those tests, should they remain, could be seen as
indicating software requirements, which they do not.  The software
requirement specification for Scheme (RnRS) is high level and leaves
much of the implementation detail unspecified.

I think it is a good idea to leave them in, probably with comments that
express that they test the implementation, not the specification.   I
also think that it is a good idea to segregate them from tests that
exercise the actual software requirements.

But, I can see an equally valid argument for stripping them out once
strings are no longer in flux, for example at release 2.0, assuming it
is bug free ;-) or perhaps 2.1.

Thanks,

Mike




> 
> Thanks,
> Ludo'.
> 
> 
>