Anything left to do in CMake ?

[Alexander Neundorf]

On Sunday 11 March 2012, Yury G. Kudryashov wrote:
> Alexander Neundorf wrote:
> > On Sunday 11 March 2012, Yury G. Kudryashov wrote:
> >> Alexander Neundorf wrote:
> >> > Hi,
> >> > 
> >> > I think things are looking quite good, and we got most of the things
> >> > we wanted to get into cmake in.
> >> 
> >> I think that we need dependent Export sets support.
> >> 
> >> Example: strigi.
> >> The project consists of several repos. You can either build these repos
> >> one by one, or download all using "meta" repo, and build the whole
> >> package.
> >> 
> >> I think that each component (libstreamanalyzer etc.) should install its
> >> own Export file. But if you build the whole project, you get a library
> >> in export B that depends on a library in export A.
> > 
> > Maybe it's not a good idea to offer two different ways how strigi is
> > built. I would see this as an inconvenience because somebody tries to do
> > something unsupported.
> 
> Why is it bad? They use splitted repos for development, then release a
> combined tarball.

I did not mean to say it's bad.
It's just that they are doing something which is not really supported, so they 
get issues (...and it was their choice to do something unsupported).

> Another example: a program (say, digikam) can either use system library
> (say, libkexiv2) or compile own copy stored in subdirectory. If digikam
> would like to EXPORT one of the libraries linked to libkexiv2, it will face
> the same problem.

Isn't it good enough if they export their own libkexiv2 then too ?

Anyway, it would be nice if you could work on this for cmake 2.8.9.
No objections at all :-)

Alex