"Robert Klemme" <bob.news / gmx.net> wrote in message news:<2hc380Fa9p38U1 / uni-berlin.de>...
> An automated system (supposed it *could* be done) would yield the false
> impression that one would not have to do anything manually any more.  That
> could lead to more bugs because of less attention.

Ah.  By that logic, unit testing is bad, because it gives you the
false impression that your code works just because it passes the
tests.  Therefore, unit testing leads to more bugs.  Real programmers
walk through each line of their code on paper, evaluating each
statement in their head, right?

> > And, no, I won't have to look at 100% of the code to find the 10%.
> > I'll notice the 10% when it fails -- just like I do right now.
> 
> Might be too late then - especially for libs.  IMHO it's more efficient to
> rely on tests.

You will never test everything.  It is a simple fact: there will
always be some case that you simply didn't write a unit test.  This is
the greatest limitation in unit testing: it relies on the programmer
to anticipate errors, and many bugs are unanticipated.

If you had an entirely closed system, it is theoretically possible
that you could write unit tests for all possible cases -- but, then,
you'd have a matrix of all possible answers and you wouldn't need the
program, right?

Unit testing is, perhaps, the most powerful development tool available
to programmers, but it isn't the only tool.  I welcome with enthusiasm
any improvement in the debugging tools available to me.

--- SER