On 03/27/2013 09:32 AM, Stephen Kelly wrote: > http://thread.gmane.org/gmane.comp.programming.tools.cmake.user/35578/focus=35666 > > All I know is that it doesn't seem to cause a build error for me when I > export a target with LINK_PUBLIC Qt5::Core LINK_PRIVATE Qt5::Script and then > have the caller only find Qt5Core and not Qt5Script. > > Would that be because libQt5Script.so.5 is in the same dir as libQt5Core.so.5
Yes, so the linker can find it there. > and the Qt5::Script entry in LINK_DEPENDENT_LIBRARIES is ignored? The Qt5::Script entry in IMPORTED_LINK_DEPENDENT_LIBRARIES_<CONFIG> will be skipped by cmComputeLinkInformation::AddSharedDepItem if it is not an available target name because it is not a full path. When generating the Qt5 imported targets if you know dependencies will be in the same directory then you can just leave them out of the IMPORTED_LINK_DEPENDENT_LIBRARIES_<CONFIG> property altogether. > Also, when building the executable, the link line contains the full path to > Qt5Core, and -Wl,-rpath,/path/to/qt5prefix/lib. Maybe that's what makes it > work for me either? The -rpath option will not help at link time, only -rpath-link. The above explains why it works though. > At any rate, changing this to be a fatal error should at least come with a > policy, no? Yes. However, I'm also fine with just leaving out implementation dependencies from the feature proposed in this thread because the conditions under which there will be a link-time error due to the missing target are quite obscure. -Brad -- Powered by www.kitware.com Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ Follow this link to subscribe/unsubscribe: http://public.kitware.com/cgi-bin/mailman/listinfo/cmake-developers
