fix static linkage and make THD statically linked

setup.py

@@ -229,16 +229,6 @@ def run(self):
                       # Python 2.6 requires -fno-strict-aliasing, see
                       # http://legacy.python.org/dev/peps/pep-3123/
                       '-fno-strict-aliasing']
-if os.getenv('PYTORCH_BINARY_BUILD') and platform.system() == 'Linux':
-    print('PYTORCH_BINARY_BUILD found. Static linking libstdc++ on Linux')
-    # get path of libstdc++ and link manually.
-    # for reasons unknown, -static-libstdc++ doesn't fully link some symbols
-    CXXNAME = os.getenv('CXX', 'g++')
-    path = subprocess.check_output([CXXNAME, '-print-file-name=libstdc++.a'])
-    path = path[:-1]
-    if type(path) != str:  # python 3
-        path = path.decode(sys.stdout.encoding)
-    extra_link_args += [path]
 
 cwd = os.path.dirname(os.path.abspath(__file__))
 lib_path = os.path.join(cwd, "torch", "lib")
@@ -265,7 +255,9 @@ def run(self):
 THCUNN_LIB = os.path.join(lib_path, 'libTHCUNN.so.1')
 THPP_LIB = os.path.join(lib_path, 'libTHPP.so.1')
 ATEN_LIB = os.path.join(lib_path, 'libATen.so.1')
-THD_LIB = os.path.join(lib_path, 'libTHD.so.1')
+GLOO_LIB = os.path.join(lib_path, 'libgloo.a')
+GLOO_CUDA_LIB = os.path.join(lib_path, 'libgloo_cuda.a')
+THD_LIB = os.path.join(lib_path, 'libTHD.a')
 NCCL_LIB = os.path.join(lib_path, 'libnccl.so.1')
 if platform.system() == 'Darwin':
     TH_LIB = os.path.join(lib_path, 'libTH.1.dylib')
@@ -276,11 +268,10 @@ def run(self):
     THCUNN_LIB = os.path.join(lib_path, 'libTHCUNN.1.dylib')
     THPP_LIB = os.path.join(lib_path, 'libTHPP.1.dylib')
     ATEN_LIB = os.path.join(lib_path, 'libATen.1.dylib')
-    THD_LIB = os.path.join(lib_path, 'libTHD.1.dylib')
     NCCL_LIB = os.path.join(lib_path, 'libnccl.1.dylib')
 
 if WITH_NCCL and subprocess.call('ldconfig -p | grep libnccl >/dev/null', shell=True) == 0:
-        SYSTEM_NCCL = True
+    SYSTEM_NCCL = True
 
 main_compile_args = ['-D_THP_CORE']
 main_libraries = ['shm']
@@ -343,6 +334,8 @@ def run(self):
         extra_compile_args += ['-DWITH_DISTRIBUTED_MW']
     include_dirs += [tmp_install_path + "/include/THD"]
     main_link_args += [THD_LIB]
+    if platform.system() == 'Linux':
+        main_link_args += [GLOO_LIB]
 
 if WITH_CUDA:
     cuda_lib_dirs = ['lib64', 'lib']
@@ -359,6 +352,8 @@ def run(self):
     extra_compile_args += ['-DCUDA_LIB_PATH=' + cuda_lib_path]
     main_libraries += ['cudart', 'nvToolsExt']
     main_link_args += [THC_LIB, THCS_LIB, THCUNN_LIB]
+    if platform.system() == 'Linux':
+        main_link_args += [GLOO_CUDA_LIB]
     main_sources += [
         "torch/csrc/cuda/Module.cpp",
         "torch/csrc/cuda/Storage.cpp",
@@ -396,6 +391,18 @@ def run(self):
     extra_compile_args += ['-O0', '-g']
     extra_link_args += ['-O0', '-g']
 
+if os.getenv('PYTORCH_BINARY_BUILD') and platform.system() == 'Linux':
+    print('PYTORCH_BINARY_BUILD found. Static linking libstdc++ on Linux')
+    # get path of libstdc++ and link manually.
+    # for reasons unknown, -static-libstdc++ doesn't fully link some symbols
+    CXXNAME = os.getenv('CXX', 'g++')
+    STDCPP_LIB = subprocess.check_output([CXXNAME, '-print-file-name=libstdc++.a'])
+    STDCPP_LIB = STDCPP_LIB[:-1]
+    if type(STDCPP_LIB) != str:  # python 3
+        STDCPP_LIB = STDCPP_LIB.decode(sys.stdout.encoding)
+    main_link_args += [STDCPP_LIB]
+    version_script = os.path.abspath("tools/pytorch.version")
+    extra_link_args += ['-Wl,--version-script=' + version_script]
 
 def make_relative_rpath(path):
     if platform.system() == 'Darwin':

tools/pytorch.version

@@ -0,0 +1,12 @@
+{
+     global:
+         _TH*;
+         TH*;
+         *THP*;
+         *THCP*;
+         PyInit*;
+         init*;
+         state;
+     local:
+         *;
+ };

torch/lib/ATen/CMakeLists.txt

@@ -61,6 +61,10 @@ MESSAGE(STATUS "THNN_LIBRARIES: ${THNN_LIBRARIES}")
 IF ($ENV{TH_BINARY_BUILD})
   MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
   SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
+  IF (UNIX AND NOT APPLE)
+    # hiding statically linked library symbols, this flag is not available for the linker under MACOSX
+    SET(CMAKE_CXX_FLAGS "-Wl,--exclude-libs,libstdc++.a ${CMAKE_CXX_FLAGS}")
+  ENDIF(UNIX AND NOT APPLE)
 ENDIF()
 
 IF(NO_CUDA)

torch/lib/THC/CMakeLists.txt

@@ -90,6 +90,10 @@ ENDIF(MAGMA_FOUND)
 IF ($ENV{TH_BINARY_BUILD})
   MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
   SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
+  IF (UNIX AND NOT APPLE)
+    # hiding statically linked library symbols, this flag is not available for the linker under MACOSX
+    SET(CMAKE_CXX_FLAGS "-Wl,--exclude-libs,libstdc++.a ${CMAKE_CXX_FLAGS}")
+  ENDIF(UNIX AND NOT APPLE)
 ENDIF()
 
 IF(APPLE)

torch/lib/THCS/CMakeLists.txt

@@ -42,16 +42,16 @@ endif()
 IF ($ENV{TH_BINARY_BUILD})
   MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
   SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
+  IF (UNIX AND NOT APPLE)
+    # hiding statically linked library symbols, this flag is not available for the linker under MACOSX
+    SET(CMAKE_CXX_FLAGS "-Wl,--exclude-libs,libstdc++.a ${CMAKE_CXX_FLAGS}")
+  ENDIF(UNIX AND NOT APPLE)
+
 ENDIF()
 
 INCLUDE_DIRECTORIES(${CUDA_INCLUDE_DIRS})
 INCLUDE_DIRECTORIES("${CUDA_SDK_ROOT_DIR}/common/inc")
 
-IF ($ENV{TH_BINARY_BUILD})
-  MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
-  SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
-ENDIF()
-
 IF(APPLE)
   IF(${CUDA_VERSION} LESS 6.0)
     # work around for mac os x bug:

torch/lib/THCUNN/CMakeLists.txt

@@ -32,6 +32,10 @@ ENDIF()
 IF ($ENV{TH_BINARY_BUILD})
   MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
   SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
+  IF (UNIX AND NOT APPLE)
+    # hiding statically linked library symbols, this flag is not available for the linker under MACOSX
+    SET(CMAKE_CXX_FLAGS "-Wl,--exclude-libs,libstdc++.a ${CMAKE_CXX_FLAGS}")
+  ENDIF(UNIX AND NOT APPLE)
 ENDIF()
 
 # Detect CUDA architecture and get best NVCC flags

torch/lib/THD/CMakeLists.txt

@@ -83,11 +83,6 @@ ENDIF()
 
 ADD_DEFINITIONS(-D_THD_CORE=1)
 
-IF ($ENV{TH_BINARY_BUILD})
-  MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
-  SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
-ENDIF()
-
 # Can be compiled standalone
 IF(NOT THD_INSTALL_BIN_DIR OR NOT THD_INSTALL_LIB_DIR OR NOT THD_INSTALL_INCLUDE_DIR)
   SET(THD_INSTALL_BIN_DIR "bin" CACHE PATH "THD install binary subdirectory")
@@ -124,14 +119,8 @@ EXCLUDE_DIR(all_cpp ".*/generic/.*\\.cpp$")
 
 INCLUDE_DIRECTORIES(${CMAKE_CURRENT_SOURCE_DIR})
 
-ADD_LIBRARY(THD SHARED ${all_cpp})
-
-IF(THD_SO_VERSION)
-  MESSAGE(STATUS "THD_SO_VERSION: ${THD_SO_VERSION}")
-  SET_TARGET_PROPERTIES(THD PROPERTIES
-    VERSION   ${THD_SO_VERSION}
-    SOVERSION ${THD_SO_VERSION})
-ENDIF(THD_SO_VERSION)
+ADD_LIBRARY(THD STATIC ${all_cpp})
+set_property(TARGET THD PROPERTY POSITION_INDEPENDENT_CODE ON)
 
 TARGET_LINK_LIBRARIES(THD LINK_PRIVATE ${TH_LIBRARIES} ${THPP_LIBRARIES})
 IF(CUDA_FOUND)

torch/lib/THPP/CMakeLists.txt

@@ -63,6 +63,10 @@ ENDIF()
 IF ($ENV{TH_BINARY_BUILD})
   MESSAGE(STATUS "TH_BINARY_BUILD detected. Statically linking libstdc++")
   SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
+  IF (UNIX AND NOT APPLE)
+    # hiding statically linked library symbols, this flag is not available for the linker under MACOSX
+    SET(CMAKE_CXX_FLAGS "-Wl,--exclude-libs,libstdc++.a ${CMAKE_CXX_FLAGS}")
+  ENDIF(UNIX AND NOT APPLE)
 ENDIF()
 
 # Can be compiled standalone

torch/lib/libshm/CMakeLists.txt

@@ -22,6 +22,10 @@ ENDIF ()
 IF ($ENV{PYTORCH_BINARY_BUILD})
   MESSAGE(STATUS "PYTORCH_BINARY_BUILD detected. Statically linking libstdc++")
   SET(CMAKE_CXX_FLAGS "-static-libstdc++ ${CMAKE_CXX_FLAGS}")
+  IF (UNIX AND NOT APPLE)
+    # hiding statically linked library symbols, this flag is not available for the linker under MACOSX
+    SET(CMAKE_CXX_FLAGS "-Wl,--exclude-libs,libstdc++.a ${CMAKE_CXX_FLAGS}")
+  ENDIF(UNIX AND NOT APPLE)
 ENDIF()
 
 ADD_LIBRARY(shm SHARED core.cpp)