[cmake-developers] linked-usage-cleanup regressed automoc

[Alexander Neundorf]

On Thursday 14 March 2013, Brad King wrote:
> On 03/13/2013 05:36 PM, Alexander Neundorf wrote:
> > On Wednesday 13 March 2013, Stephen Kelly wrote:
> >> I think Alex' objection is only related to thinking that the case of a
> >> header-only-library-with-automoc-generated-cxx-file should be an error.
> > 
> > Yes, looks like it.
> > If this is really a feature (and not a bug), then this is probably
> > needed. I still think it's kind of strange.
> > I mean, this turns a header-only library, and for those I would expect
> > that it was a deliberate decision to make it header-only, into a normal
> > link library.
> 
> There is no such thing as a header-only add_library.  It would have no
> link language or objects and CMake would not generate.  The point is to
> support a header-only Qt library where the moc part is the only source
> of C++ code.

Yes.
 
> > Also, at least in theory, it could be a C (and not C++) header-only
> > library, which would then turn into C++.
> 
> Don't set the AUTOMOC property on the target if you don't want it.

Yes, there's no problem with this.
It's the global
set(CMAKE_AUTOMOC ON)
which may have unexpected effects.

> > Really, who is using this ?
> 
> If no one was using it I wouldn't have started this thread and
> this discussion wouldn't be happening (yet).  It was in one of Kitware's
> other projects but I don't remember which one.  Since then they've
> stopped using this though and moved the header into another library that
> also has real sources.
> 
> However, since it worked in 2.8.10 others could be using it so it will
> take a post-2.8.11 policy to drop it.

Well, if this is indeed used, I don't feel like removing it :-)

Would you object a check that if a target has at least one C or Fortran source 
file, but no C++ source file, it should skipped for automoc ?

Alex