CMake build options BUILD_STATIC_LIBRARY, BUILD_SHARED_LIBRARY, and B…

[axelsommerfeldt]

…UILD_DEVEL_PACKAGE added (default=ON)

These three new CMake build options control what kind of libraries will be build and what contents will be put into the resulting package:

• BUILD_STATIC_LIBRARY controls if a static restbed library should be build
• BUILD_SHARED_LIBRARY controls if a shared restbed library should be build
• BUILD_DEVEL_PACKAGE controls if a development package should be build, i.e. C++ header files included

The default value for all three options is ON to provide backward-compatibility with older restbed versions which do not offer these build options yet.

[axelsommerfeldt]

How does the addition of CMake options (which do not cause any change unless set to OFF) could affect a regression test?

96/116 Test  #96: yield_callback_leads_to_service_crash_regression_test_suite ................Exit code 0xc0000409
***Exception:   0.10 sec

It seems that either the "yield_callback_leads_to_service_crash" bug isn't really fixed or something else here is fishy.

BTW: The result of the tests on Debian 11:

96/116 Test  #96: yield_callback_leads_to_service_crash_regression_test_suite ................   Passed    0.01 sec
...
100% tests passed, 0 tests failed out of 116

Total Test time (real) =  23.69 sec

[madebr]

Only MSVC uses file names that conflict between shared and static libraries.
Also, the .dll does not really need the -shared suffix.
So set the ARCHIVE_OUTPUT_NAME property additionally.

Suggested change

	if ( WIN32 )
	# Workaround to avoid name clash of static lib and dynamic import lib under windows.
	set_target_properties( ${SHARED_LIBRARY_NAME} PROPERTIES OUTPUT_NAME "${PROJECT_NAME}-shared" )
	if ( MSVC )
	# Workaround to avoid name clash of static lib and dynamic import lib under windows.
	set_target_properties( ${SHARED_LIBRARY_NAME} PROPERTIES OUTPUT_NAME ${PROJECT_NAME} ARCHIVE_OUTPUT_NAME "${PROJECT_NAME}-shared" )

This results in the following files:

restbed-shared.lib # import library of restbed.dll (shared library)
restbed.dll        # shared library
restbed.lib        # static library

[axelsommerfeldt]

Thanks for your feedback. But I'm not sure what the proposed change of yours has to do with my Pull Request. I just have introduced three boolean switches (which default to ON) and have surrounded already existing CMake code with if ( ) ... endif() (and have indented the lines within it), that's all. It's a very trivial change which should have no impact except offering the possibility to set one (or more) of these new switches to OFF. Therefore I have no idea why one of the tests fails when run on MS-Windows.

Regarding the MS-Windows build: I can neither comment on this nor check it since I'm using Linux only and have no experience with CMake and MS-Windows. But I don't see the correlation with my changes, and furthermore it's a test execution that fails under MS-Windows, not the build, at least that is my understanding.

[madebr]

You need to install restbed.dll too.

Suggested change

	install( TARGETS ${SHARED_LIBRARY_NAME} LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR} ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR} COMPONENT library )
	install( TARGETS ${SHARED_LIBRARY_NAME} LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR} ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR} COMPONENT library RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR} COMPONENT runtime)

[axelsommerfeldt]

What does this proposed change has to do with my Pull Request? I just have indented this already existing line because I have surrounded it with if( )..endif().

If there is something to improve for the MS-Windows build, I suggest making a separate Pull Request.