Re: Copying sc/sd/sw into unittest libs (Re: namespace / typing thrash ...)

On Wed, Apr 18, 2012 at 09:33:51PM +0200, Lubos Lunak wrote:
>  This:
>
> On Monday 16 of April 2012, Markus Mohrhard wrote:
> > 2012/4/16 Lubos Lunak <l.lunak@suse.cz>:
> > > On Monday 16 of April 2012, Michael Meeks wrote:
> > > >       Oh - and finally (Lubos) I pushed an item to the ESC agenda to
> > > > discuss whether we should be exposing tons of classes and their symbols
> > > > in the product, just to make unit tests work :-)
> > >
> > >  I assume this is about 69d46dd7a6adfffd71da055bb65108c80d27395f .
> ...
> > I was really annoyed by the fact that is was changed without at least
> > asking and noticing the people who are affected by this change. There
> > were good reasons to have the old behaviour and I spend some ours
> > searching for a bug because I had to export a method for ucalc. IMHO
> > such basic things should not be changed without noticing and
> > discussing with the people who are directly working in that area.

If I got Markus right, his problem was not that he had to export a method, but
that somebody changed unittests from static linking the library it tests to
dynamic linking. I wholeheartly agree: A unittest should be allowed to see the
internals of what it tests -- esp. as the "unit" is something way smaller that
one of our (huge) libraries.

If you care about the size of the build output(*), make the "unit" that the test
tests able to be standalone, so that only that subset needs to be statically
linked into the unittest. Thats a big harder, but something that would benefit
the codebase as a whole (decoupling FTW etc.).

Best,

Bjoern


(*) And those who are doing unconditioned debug build are just asking for huge
build dirs -- nobody forces you to do so.