> For system libraries like this it is often better to just link them in > with -l. The compiler should be able to find them if it is installed > correctly.
MinGW is installed in standard fashion. Nothing unusual there. No environment variables, like LIB etc., are set that might cause problems. > if(MINGW) > target_link_libraries(foo uuid ole32 shell32) > endif(MINGW) > There is a good chance CMake will get it wrong. So, the > correct CMake solution for system/compiler libraries is to just use the > name of the library. > Something like this: > if(MINGW) > target_link_libraries(foo uuid ole32 shell32) > endif(MINGW) To me that seems counter to the whole reason to use CMake in the first place. I could have just written Makefile to do that. This is where I'm coming from. I wrote a XML parser to deal with Atmel configuration files on Linux. I wanted it to be cross platform. I knew that the autotools would be a problem to make a Windows version, so I thought I'd give CMake a try, even bought your book. Everything worked just great on Linux, no problems at all. Quickly had a working program. However as chronicled in a couple of threads here I've found using CMake with MinGW in conjunction with wxWidgets to be a nightmare. I've spent days trying to get things to link, and still don't have a working program. I know many of the issue have been self inflicted as I learn CMake. I expected everything to 'just work' the way it did under Linux. Apparently my expectation was to optimistic, after all this is Windows. Perhaps if someone could point me to a working example of MinGW (Not MSYS)/wxWidgets/CMake would help? Miguel has been especially patient with me, both on and off the list here, in getting my CMake/MinGW/wxWidgets issues sorted out, and I want to publicly thank him for all of his help. _______________________________________________ CMake mailing list CMake@cmake.org http://www.cmake.org/mailman/listinfo/cmake