[MetaCRS] Using cmake of osgeo projects?

[Charles Karney]

Sorry for hitting multiple mailing lists...

I'm currently going through the process of compiling various packages
for Windows (proj, libtiff/geotiff, gdal, libLAS) and hitting the
various (but typical) bumps...

In my group at SRI, we've more-or-less completely switched to cmake for
configuring and building our software.   The benefits are:

* a single set of configuration files for Linux/Windows;
* makes the process by which dependent packages are discovered
   systematic;
* makes the process of installing software systematic;
* thus making the package you're building easy for other packages to
   discover;
* running tests can be integrated into the build.

While cmake has its warts, it is much less arcane and easier to
use than autoconf scripts used for Unix and Linux systems.  It has
been a godsend when building software which depends on several third
party packages, several of our own packages, and where the selection of
package versions, compiler, and 32-bit vs 64-bit, may depend on the
particular project.

libLAS already uses cmake.  But it appears to be the exception in the
realm of geospatial software.  Is there any interest in adopting cmake
for other packages?  Incidentally, the choice is not either/or...  For
example, GeographicLib supports Windows solution files, vanilla
makefiles, autoconf, and cmake.  (And I know from this experience that
the cmake configuration is much the easiest to maintain.)

Unfortunately, I'm not in a position to undertake this work myself.  But
if people have questions about cmake, I may be able to help.

   --Charles
-- 
Charles Karney <charles.karney at sri.com>
SRI International, Princeton, NJ 08543-5300

Tel: +1 609 734 2312
Fax: +1 609 734 2662