defines, frameworks now also generate CMakeConfigDeps targets

[AbrilRBS]

Changelog: Fix: defines and frameworks now also generate CMakeConfigDeps targets.
Docs: Omit

This comes as a fix for the opengl recipe in Macos, which has this package_info():

        self.cpp_info.set_property("cmake_file_name", "opengl_system")

        self.cpp_info.bindirs = []
        self.cpp_info.includedirs = []
        self.cpp_info.libdirs = []
        if self.settings.os == "Macos":
            self.cpp_info.defines.append("GL_SILENCE_DEPRECATION=1")
            self.cpp_info.frameworks.append("OpenGL")
        elif self.settings.os == "Windows":
            self.cpp_info.system_libs = ["opengl32"]
        elif self.settings.os in ["Linux", "FreeBSD", "SunOS"]:
            pkg_config = PkgConfig(self, 'gl')
            pkg_config.fill_cpp_info(self.cpp_info, is_system=self.settings.os != "FreeBSD")

which did not create a target for opengl, and then having this as a dependency made it error out

[franramirez688]

@AbrilRBS The PR looks good, but I think it could be the moment to refactor that a little bit?

• The function name _get_cmake_lib sounds confusing to me. Would it not be better to have something like _get_cmake_target?
• The first check is getting a bit convoluted and fragile:

if info.exe or not (info.package_framework or info.frameworks or info.includedirs or info.libs
                            or info.system_libs or info.defines):

What I'm wondering, if we only return empty when if info.exe. Would it not be more robust to return an empty CMake target instead? Would it be a problem to return an empty one?

    def _get_cmake_lib(self, info, components, pkg_folder, pkg_folder_var, comp_name=None):
        if info.exe:
            return

I'm asking because if we're checking the defines field, what about cflags, cxxflags, etc.?

[memsharded]

It was called get_cmake_lib because there is another method called def _get_exes(self, that takes care of creating the executable targets, while this is creating the library targets.

I don't exactly recall why the check is there to not create targets if those are empty, but this is incubating, so it is good to try to remove and leave the if info.exe: return only and see how it goes

[memsharded]

@franramirez688 lets merge this as-is because the change is fully clear, and have the refactor in another PR?