Merge branch 'feature/Cpp11' into develop

CMakeLists.txt

@@ -31,6 +31,17 @@ include(ShogunUtils)
 
 SET(CMAKE_INCLUDE_DIRECTORIES_PROJECT_BEFORE ON)
 
+#### set required C++ standard level of the compiler
+set(CMAKE_CXX_STANDARD 11)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
+set(CMAKE_CXX_EXTENSIONS OFF)
+
+### FIXME: remove these flags when the codebase
+# is cleared up
+set(HAVE_CXX11 ON)
+set(HAVE_CXX11_ATOMIC ON)
+set(HAVE_STD_UNORDERED_MAP ON)
+
 ############# minimum library versions ###################
 SET(EIGEN_VERSION_MINIMUM 3.1.2)
 SET(VIENNACL_VERSION_MINIMUM 1.5.0)
@@ -90,13 +101,9 @@ SET(LIBSHOGUN ON CACHE BOOL "Compile shogun library")
 
 IsAnyTrue("${AVAILABLE_INTERFACES}" ANY_INTERFACE_ENABLED)
 IF (${ANY_INTERFACE_ENABLED})
-	IF(RModular)
-		FIND_PACKAGE(SWIG 2.0.5 REQUIRED)
-	ELSE()
-		FIND_PACKAGE(SWIG 2.0.4 REQUIRED)
-	ENDIF()
+	# SWIG3 is the minimum requirement because of C++11 support
+	SET(SWIG_VERSION_MINIMUM 3.0.0)
 
-	SET(SWIG_VERSION_MINIMUM 2.0.4)
 	IF(CSharpModular)
 		# We require SWIG 3.0.7 to support functions with a few SGVector or
 		# SGMatrix arguments. The required SWIG feature is called
@@ -108,11 +115,14 @@ IF (${ANY_INTERFACE_ENABLED})
 		# typemapping created for earlier versions of SWIG.
 		# see: http://www.swig.org/Doc3.0/CSharp.html#CSharp_introduction_swig2_compatibility
 		LIST(APPEND CMAKE_SWIG_FLAGS "-DSWIG2_CSHARP")
-	ELSEIF(RModular)
-		SET(SWIG_VERSION_MINIMUM 2.0.5)
+	ELSEIF(PythonModular)
+		# SWIG was broken for combining -builtin and -modernargs
+		# from v3.0.0 and until 3.0.4.  This bug was fixed in
+		# v3.0.5.  Make CMake emit an error and fail to configure.
+		SET(SWIG_VERSION_MINIMUM 3.0.5)
 	ENDIF()
-	FIND_PACKAGE(SWIG ${SWIG_VERSION_MINIMUM} REQUIRED)
 
+	FIND_PACKAGE(SWIG ${SWIG_VERSION_MINIMUM} REQUIRED)
 	# use our own UseSWIG.cmake in order to be able to enable ccache-swig
 	SET(SWIG_USE_FILE ${CMAKE_SOURCE_DIR}/cmake/UseSWIG.cmake)
 	IF(ENABLE_CCACHE AND CCACHE_SWIG)
@@ -247,41 +257,6 @@ IF(ENABLE_UBSAN)
 	ENDIF()
 ENDIF()
 
-# check for supported c++11 features
-#
-# clang with -std=c++11 and -stdlib=libc++ does not work
-# well with swig generated cxx hence disable c++11 for this case.
-
-# this has been only fixed in swig 2.0.12 or later.
-IF (NOT ((CYGWIN AND ENABLE_TESTING) OR (DARWIN AND COMPILE_MODULAR_INTERFACE
-	AND SWIG_VERSION VERSION_LESS "2.0.12")))
-	INCLUDE(CheckCXX11Features)
-
-	IF (MSVC)
-		IF(MSVC_VERSION VERSION_GREATER "1500")
-			SET(HAVE_CXX11 1)
-		ENDIF()
-	ELSE(MSVC)
-		IF(_HAS_CXX11_FLAG)
-			SET(HAVE_CXX11 1)
-			SET(CMAKE_CXX_FLAGS "${CXX11_COMPILER_FLAGS} ${CMAKE_CXX_FLAGS}")
-			SET(SWIG_CXX_COMPILER_FLAGS "${CXX11_COMPILER_FLAGS} ${SWIG_CXX_COMPILER_FLAGS}")
-		ENDIF()
-
-		IF(_HAS_CXX0X_FLAG)
-			SET(HAVE_CXX0X 1)
-			SET(CMAKE_CXX_FLAGS "${CXX11_COMPILER_FLAGS} ${CMAKE_CXX_FLAGS}")
-			SET(SWIG_CXX_COMPILER_FLAGS "${CXX11_COMPILER_FLAGS} ${SWIG_CXX_COMPILER_FLAGS}")
-		ENDIF()
-	ENDIF()
-ELSEIF(DARWIN AND ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang") AND NOT (CMAKE_CXX_COMPILER_VERSION VERSION_LESS "5.0.0"))
-	# osx clang 5.0.0 or later uses libc++ by default
-	# this is causing problems with source generated by swig version earlier than 3.0.0
-	# force to use libstdc++ for compilation of sources
-	SET(CMAKE_CXX_FLAGS "-stdlib=libstdc++ ${CMAKE_CXX_FLAGS}")
-	SET(SWIG_CXX_COMPILER_FLAGS "-stdlib=libstdc++ ${SWIG_CXX_COMPILER_FLAGS}")
-ENDIF()
-
 # Fix build on Mac OSX 10.10 Yosemite when using mp-gcc-4X.
 # See: https://github.com/shogun-toolbox/shogun/issues/2635
 IF(DARWIN AND (NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang"))

NEWS

@@ -1,9 +1,10 @@
 2016-11-05 Viktor Gal <viktor.gal@shogun-toolbox.org>
 
-	* SHOGUN Release version 5.1.0 (libshogun 17.2, data 0.11, parameter 1)
+	* SHOGUN Release version 6.0.0 (libshogun 18.0, data 0.11, parameter 1)
 
 	* NOTE: Contains major rewrite and clean-up of developer documentation in doc/readme [Heiko Strathmann]	
 	* Add native MS Windows support [Viktor Gal]
+	* Shogun requires the compiler to support C++11 features
 
 	* Features:
 		- LDA now supports 32, 64 and 128 bit floating point numbers [Chris Goldsworthy]