默认预处理器定义和跨平台编译

Question

I'm trying to figure a "clean way" of using the default preprocessor definitions to determine what parts of my code should be compiled based on the platform and compiler.

My current test setup involves a Windows machine with Visual C++ compiler and a Debian with g++ compiler.

Currently I have something like this:

#if defined (__GNUG__)
    #define ASMMath_EI __attribute__ ((__visibility__("default")))
#elif defined (WIN32)
    #ifdef ASMMath_EXPORTS
        #define ASMMath_EI __declspec(dllexport)
    #else
        #define ASMMath_EI __declspec(dllimport)
    #endif
#endif

extern void ASMMath_EI AsmProblemOne();

And it works, but I figure there might and must be some better definitions I can check for. Or perhaps some more ideal way with CMake? Suggestions?

Answer 1

There is a nice listing of Compiler , Operating System , and Architecture preprocessor names at those links. You could branch for the systems and compilers you care about supporting/detecting. Additionally a lot of this work is already done in boost/config/ (see boost/config/select_compiler_config.hpp as one example for compiler flags) using Boost headers. Not everyone likes including Boost, which is why the first set of links is generic from library specific support.

Answer 2

Answered by the OP in a question edit:

From what I gather the following would be ideal way to implement this:

 // MasterHeader.h #if defined _MSC_VER // Defined by visual studio #define PROJ_TMP_DLL_IMPORT __declspec(dllimport) #define PROJ_TMP_DLL_EXPORT __declspec(dllexport) #else #if __GNUC__ >= 4 // Defined by GNU C Compiler. Also for C++ #define PROJ_TMP_DLL_IMPORT __attribute__ ((visibility ("default"))) #define PROJ_TMP_DLL_EXPORT __attribute__ ((visibility ("default"))) #else #define PROJ_TMP_DLL_IMPORT #define PROJ_TMP_DLL_EXPORT #endif #endif #ifdef PROJ_EXPORTS #define PROJ_API PROJ_TMP_DLL_EXPORT #else #define PROJ_API PROJ_TMP_DLL_IMPORT #endif 

-

 // File.h #include "MasterHeader.h" extern void PROJ_API SomeFunction();