Re: Coverage for libltdl/slist.c and fallout

[Ralf Wildenhues]

Hi Gary,

* Gary V. Vaughan wrote on Sun, Dec 06, 2009 at 05:55:25PM CET:
> On 1 Dec 2009, at 06:39, Ralf Wildenhues wrote:
> > So, there is two choices: remove the API, or add test coverage.  Which
> > alternative do you prefer?
> 
> I don't plan to do either.  But, of course, I'd much rather you didn't remove
> half of the functionality from slist.[ch] just because I don't have time to
> write unit tests.

I'm fine with whoever writes unit tests.  I would even write them, but
for that I'd need to understand how it's supposed to work.  There is no
documentation, no examples, no other code that uses slist this way that
I can look at.

BTW, so far I haven't removed anything that wouldn't be fixable with a
cast or two.

More generally, I am really convinced that libltdl quality is the way it
is only because authors never really cared to ensure their code really
does what it was supposed to do.  If we continue to treat testing and
coverage as an afterthought, there is little reason to believe that is
going to change.  So yes, I pretty much think that all code that isn't
exercised by the testsuite but could be, does not belong in the tree.

If writing tests is too hard for some reason, we need to make it easier
to do so.  That would then be probably be even more important than any
new feature.

Cheers,
Ralf