[Dev] Change CXX standard to 17 (#6138) (#6151)

Co-authored-by: Muhammed Fatih BALIN <[email protected]>

CMakeLists.txt

@@ -5,6 +5,9 @@ cmake_minimum_required(VERSION 3.5)
 project(dgl C CXX)
 message(STATUS "Start configuring project ${PROJECT_NAME}")
 
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD_REQUIRED ON)
+
 # cmake utils
 include(cmake/util/Util.cmake)
 include(cmake/util/MshadowUtil.cmake)
@@ -162,8 +165,6 @@ else(USE_CUDA)
   add_library(dgl SHARED ${DGL_SRC})
 endif(USE_CUDA)
 
-set_property(TARGET dgl PROPERTY CXX_STANDARD 14)
-
 # include directories
 target_include_directories(dgl PRIVATE "include")
 target_include_directories(dgl PRIVATE "third_party/dlpack/include")

cmake/modules/CUDA.cmake

@@ -8,7 +8,7 @@ endif()
 ###### Borrowed from MSHADOW project
 
 include(CheckCXXCompilerFlag)
-check_cxx_compiler_flag("-std=c++14"   SUPPORT_CXX14)
+check_cxx_compiler_flag("-std=c++17"   SUPPORT_CXX17)
 
 set(dgl_known_gpu_archs "35" "50" "60" "70")
 set(dgl_cuda_arch_ptx "70")
@@ -248,7 +248,7 @@ macro(dgl_config_cuda out_variable)
   list(APPEND CUDA_NVCC_FLAGS ${NVCC_FLAGS_ARCH})
 
   # 2. flags in third_party/moderngpu
-  list(APPEND CUDA_NVCC_FLAGS "--expt-extended-lambda;-Wno-deprecated-declarations;-std=c++14")
+  list(APPEND CUDA_NVCC_FLAGS "--expt-extended-lambda;-Wno-deprecated-declarations;-std=c++17")
 
   message(STATUS "CUDA_NVCC_FLAGS: ${CUDA_NVCC_FLAGS}")
 

python/setup.py

@@ -145,7 +145,7 @@ def config_cython():
                     library_dirs=library_dirs,
                     libraries=libraries,
                     # Crashes without this flag with GCC 5.3.1
-                    extra_compile_args=["-std=c++14"],
+                    extra_compile_args=["-std=c++17"],
                     language="c++",
                 )
             )