[MetaCRS] [Proj] [osgeo4w-dev] Using cmake of osgeo projects?

[Charles Karney]

On 07/25/2013 08:01 PM, Greg Troxel wrote:
>
> Charles Karney <charles.karney at sri.com> writes:
>
>> On 06/29/2013 09:05 PM, Mateusz Loskot wrote:
>>> I don't know what's typical, regarding developer's reaction on CMake adoption
>>> proposal, but as a big fan of CMake, I wouldn't paint it in pink colours only.
>>>
>>> Thus, I understand, for example, Frank's experience may be very different.
>>>
>>> Finally, if a software has a build system that is usable and works on
>>> all supported
>>> platforms, I also understand a team may go for lean approach: don't change it.
>>>
>>
>> I basically agree.  However, let's not forget those users who need, for
>> one reason or another, to compile these packages themselves.
>
> I find it odd to talk about how one can't compile packages oneself when
> they use autoconf.
>
> Is this really about accommodating windows without using cygwin?  I have
> found that autoconf/automake/libtool is really quite workable.

Windows support is the important issue for some users.  CMakes's other
big advantage relative to autoconf is that it provides a more systematic
way of discovering dependent packages.

I agree that, as a user of a package, autoconf on non-Windows systems
works well.  However, for the developer of a package, autoconf seems
like an ugly hacked-together mess, which I would not recommend to anyone
starting on a new package.  Nevertheless, I concede that the issue gets
more complicated for large packages which already have autoconf support.

> Does cmake support
>
>     building in an objdir, with a read-only srcdir
>
>     cross-compiling
>
>     the equivalent of 'make dist' and 'make distcheck'
>
>     building on a system not previously known to the build system, as
>     long as feature tests are doable

The answer to these is yes.  In particular, building outside the source
tree is the standard procedure with cmake.