From: Werner S. <sm...@ia...> - 2006-11-22 21:39:51
|
Hi Alan, sorry for the late reply. > > I believe the -rpath mechanism worked under our ABS for > build-tree tests under Cygwin with shared libraries. So we may be up > against a bug in our PLplot CBS, a bug in the CMake generator that Werner > used on Cygwin, or a bug in CMake itself for the Cygwin platform. > > So, Werner, what generator did you use for your Cygwin experiment? Hmm, I didn't specify a generator explicitly. Just let Cmake decide, it choose UNIX Makefiles normally. > If experiments with a variety of generators are unsuccessful, The only other generator available is KDevelop3, so I think there is not much choice here. then to help > sort out what is really going on I suggest you implement a simple > "hello-world" style CMake example with the shared library built in one > subdirectory and an example built in another subdirectory that calls that > library with the appropriate cmake commands to set up ctest for that > example. > > (One of the nice advantages of CMake is it usually only takes a few minutes > to set up simple test cases like this.) > > If you can get that simple test case to work on Cygwin with any > generator, but that same generator does not work for our CBS, then > obviously we need to > improve our CBS. If the simple example does not work under any > circumstances, then it should be straightforward to make it part of the bug > report to the CMake developers so they can easily replicate the problem for > themselves. Yes, I was just thinking about doing some tests here - but cygwin hasn't updated cmake to 2.4.4 so far, so I think this has to wait a little. Regards, Werner |