When trying to build tensorflow using cmake, I got the following error. I'm not sure whether it's an issue related to tensorflow or grpc. Any idea on what's going on?
This is on Mac OS Sierra.
steps to reproduce:
# clone tensorflow repo
# in tensorflow directory
cd tensorflow/contrib/cmake
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DPYTHON_EXECUTABLE=/usr/local/bin/python3
make tf_tutorials_example_trainer
Errors:
# ...
Scanning dependencies of target tf_tutorials_example_trainer
[100%] Building CXX object CMakeFiles/tf_tutorials_example_trainer.dir/Users/kevenwang/VirtualBoxShared/another_tf/tensorflow/cc/tutorials/example_trainer.cc.o
[100%] Linking CXX executable tf_tutorials_example_trainer
Undefined symbols for architecture x86_64:
"_ares_cancel", referenced from:
on_readable_cb(grpc_exec_ctx*, void*, grpc_error*) in libgrpc_unsecure.a(grpc_ares_ev_driver_posix.cc.o)
on_writable_cb(grpc_exec_ctx*, void*, grpc_error*) in libgrpc_unsecure.a(grpc_ares_ev_driver_posix.cc.o)
"_ares_destroy", referenced from:
grpc_ares_ev_driver_unref(grpc_ares_ev_driver*) in libgrpc_unsecure.a(grpc_ares_ev_driver_posix.cc.o)
...
_grpc_ares_ev_driver_create in libgrpc_unsecure.a(grpc_ares_ev_driver_posix.cc.o)
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[3]: *** [tf_tutorials_example_trainer] Error 1
make[2]: *** [CMakeFiles/tf_tutorials_example_trainer.dir/all] Error 2
make[1]: *** [CMakeFiles/tf_tutorials_example_trainer.dir/rule] Error 2
Maybe you install the grpc lib by Homebrew.
In that case , you need to link the c-ares lib.
e.g.
TARGET_LINK_LIBRARIES(demo "/usr/local/lib/libcares.a" ...)
Related
any ideas for this?
I try to build SDRangel on my M1 Mac but at 77% I get this error ...
-- GIT_EXECUTABLE /opt/local/bin/git
-- GIT LIVE REPO VERSION RETRIEVED
-- X265 RELEASE VERSION 3.5+68-40e37bce9
-- Detected CXX compiler using -O3 optimization level
-- Configuring done
-- Generating done
-- Build files have been written to: /opt/build/sdrangel/build/external/x265/src/x265-build
[ 27%] Performing build step for 'x265'
Consolidate compiler generated dependencies of target common
[ 45%] Built target common
Consolidate compiler generated dependencies of target encoder
[ 65%] Built target encoder
[ 76%] Built target x265-static
[ 77%] Linking CXX shared library libx265.dylib
Undefined symbols for architecture arm64:
"x265_entropyStateBits", referenced from:
_x265_costCoeffNxN_neon in pixel-util.S.o
(maybe you meant: _x265_entropyStateBits)
ld: symbol(s) not found for architecture arm64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[5]: *** [libx265.206.dylib] Error 1
make[4]: *** [CMakeFiles/x265-shared.dir/all] Error 2
make[3]: *** [all] Error 2
make[2]: *** [external/x265/src/x265-stamp/x265-build] Error 2
make[1]: *** [external/CMakeFiles/x265.dir/all] Error 2
make: *** [all] Error 2
I am stuck. Don't know what to try to solve this issue.
I did not find anything anywhere ...
Working on it for most of 2 days, I did find this site that presents a working method of building libx265 stuff on Apple Silicon and works for me on macOS 13.0.1. https://www.osxexperts.net/ Problem is, it only builds the libraries for building it into ffmpeg, no executable binary.
When I tried to get the executable also, by removing -DENABLE_CLI=OFF from the last cmake command, the error comes back. Put it back in, everything completes without error.
Here is the full process that works for me (no executable). It builds a multilibrary that can encode 8, 10, or 12-bit video:
cd x265/source
#### 12 bit
cmake -DCMAKE_INSTALL_PREFIX:PATH=${TARGET} -DHIGH_BIT_DEPTH=ON -DMAIN12=ON -DENABLE_SHARED=NO -DEXPORT_C_API=NO -DENABLE_CLI=OFF .
make
mv libx265.a libx265_main12.a
make clean-generated
rm CMakeCache.txt
#### 10bit
cmake -DCMAKE_INSTALL_PREFIX:PATH=${TARGET} -DHIGH_BIT_DEPTH=ON -DMAIN10=ON -DENABLE_SHARED=NO -DEXPORT_C_API=NO -DENABLE_CLI=OFF .
make clean
make
mv libx265.a libx265_main10.a
make clean-generated && rm CMakeCache.txt
#### 8 bit and all together
# removed -DENABLE_CLI=OFF from last one so we get the x265 binary
cmake -DCMAKE_INSTALL_PREFIX:PATH=${TARGET} -DEXTRA_LIB="x265_main10.a;x265_main12.a" -DEXTRA_LINK_FLAGS=-L. -DLINKED_10BIT=ON -DLINKED_12BIT=ON -DENABLE_SHARED=OFF -DENABLE_CLI=OFF .
make clean
make
mv libx265.a libx265_main.a
# this is Apple's libtool
libtool -static -o libx265.a libx265_main.a libx265_main10.a libx265_main12.a
make install
When I remove the last -DENABLE_CLI=OFF, here's the error:
[100%] Linking CXX executable x265
Undefined symbols for architecture arm64:
"x265_entropyStateBits", referenced from:
_x265_costCoeffNxN_neon in libx265.a(pixel-util.S.o)
_x265_10bit_costCoeffNxN_neon in libx265_main10.a(pixel-util.S.o)
_x265_12bit_costCoeffNxN_neon in libx265_main12.a(pixel-util.S.o)
(maybe you meant: _x265_entropyStateBits)
ld: symbol(s) not found for architecture arm64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[2]: *** [x265] Error 1
make[1]: *** [CMakeFiles/cli.dir/all] Error 2
make: *** [all] Error 2
JimsMBPro:source jim$
EDIT: #Peter Betz, see the issue report. Until the developers correct this error, it will compile if you edit x265/source/common/aarch64/pixel-util.S, line 2410. Add an underscore as follows. Change:
movrel x1, x265_entropyStateBits
to
movrel x1, _x265_entropyStateBits
I am trying to compile from source ypsilon, but couldn't get it installed on Ubunut 22.04 LTS after multiple trial-and-errors.
I followed the official instructions linked above but process fails at cmake --build ., complaining that it couldn't find llvm. Then, I used cmake -DLLVM_DIR=/usr/lib/llvm-14/cmake .... Now, the compilation is OK, but there are many link errors even after trying out -DCMAKE_EXE_LINKER_FLAGS_RELEASE=:
$ cmake -DLLVM_DIR=/usr/lib/llvm-14/cmake -DCMAKE_EXE_LINKER_FLAGS_RELEASE=-L/usr/lib/llvm-14/lib -DCMAKE_BUILD_TYPE=Release .. && cmake --build . |& tee ~/a.txt
-- Build type: Release
-- Found LLVM 14.0.0
-- Using LLVMConfig.cmake in: /usr/lib/llvm-14/cmake
-- Configuring done
-- Generating done
-- Build files have been written to: /home/user/tmp/ypsilon/build
[ 2%] Building CXX object CMakeFiles/ypsilon.dir/src/arith.cpp.o
...
[100%] Linking CXX executable ypsilon
ld.lld: error: undefined symbol: llvm::Type::getInt32Ty(llvm::LLVMContext&)
>>> referenced by ld-temp.o
>>> lto.tmp:(digamma_t::reg_cache_t<8>::copy(llvm::Value*))
>>> referenced by ld-temp.o
>>> lto.tmp:(digamma_t::reg_cache_t<24>::copy(llvm::Value*))
>>> referenced by ld-temp.o
>>> lto.tmp:(digamma_t::reg_cache_t<16>::copy(llvm::Value*))
>>> referenced 96 more times
ld.lld: error: undefined symbol: llvm::ConstantInt::get(llvm::IntegerType*, unsigned long, bool)
...
ld.lld: error: too many errors emitted, stopping now (use -error-limit=0 to see all errors)
clang: error: linker command failed with exit code 1 (use -v to see invocation)
gmake[2]: *** [CMakeFiles/ypsilon.dir/build.make:817: ypsilon] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:164: CMakeFiles/ypsilon.dir/all] Error 2
gmake: *** [Makefile:146: all] Error 2
I am not familiar with cmake or its linking flags on the command line.
Does any one know how to fix the cmake linking issue in the above commands?
The steps I tried are as follows:
sudo apt install clang lld
git clone https://github.com/fujita-y/ypsilon.git
cd ypsilon/ ; mkdir -p build ; cd build
cmake -DLLVM_DIR=/usr/lib/llvm-14/cmake -DCMAKE_EXE_LINKER_FLAGS_RELEASE=-L/usr/lib/llvm-14/lib -DCMAKE_BUILD_TYPE=Release .. && cmake --build .
I'm trying to build LUAGLM (from https://github.com/gottfriedleibniz/lua) to bind GLM library to give access in Lua using cmake. When i build, the building completes ok but the linker does not giving error:
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[2]: *** [glm.so] Error 1
make[1]: *** [CMakeFiles/glm.dir/all] Error 2
make: *** [all] Error 2
Above these lines there is a very long list starting with:
[ 10%] Building CXX object CMakeFiles/liblua.dir/onelua.c.o
[ 20%] Linking CXX shared library liblua54.dylib
[ 20%] Built target liblua
[ 30%] Building CXX object CMakeFiles/liblua_static.dir/onelua.c.o
[ 40%] Linking CXX static library liblua54_static.a
[ 40%] Built target liblua_static
[ 50%] Building C object CMakeFiles/lua.dir/lua.c.o
[ 60%] Linking CXX executable lua
[ 60%] Built target lua
[ 70%] Building CXX object CMakeFiles/luac.dir/onelua.c.o
[ 80%] Linking CXX executable luac
[ 80%] Built target luac
[ 90%] Building CXX object CMakeFiles/glm.dir/libs/glm_binding/lglmlib.cpp.o
[100%] Linking CXX shared module glm.so
Undefined symbols for architecture x86_64:
"glm_pushmat(lua_State*, glmMatrix const&)", referenced from:
glm_mat_add(lua_State*) in lglmlib.cpp.o
glm_mat_sub(lua_State*) in lglmlib.cpp.o
glm_mat_mul(lua_State*) in lglmlib.cpp.o
glm_mat_negate(lua_State*) in lglmlib.cpp.o
glm_mix(lua_State*) in lglmlib.cpp.o
glm_saturation(lua_State*) in lglmlib.cpp.o
glm_orthonormalize(lua_State*) in lglmlib.cpp.o
...
I'm on MacOS Intel (i386-apple-darwin11.3.0)
Not sure where to go next, do I need to edit the CMakeList somehow?
Any help much appreciated
EDIT:
The cmake error log contains the following :
Compiling the C compiler identification source file "CMakeCCompilerId.c" failed. Compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/cc Build flags: Id flags:
The output was: 1 ld: library not found for -lSystem clang: error: linker command failed with exit code 1 (use -v to see invocation)
Compiling the CXX compiler identification source file "CMakeCXXCompilerId.cpp" failed. Compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/c++ Build flags: Id flags:
The output was: 1 ld: library not found for -lc++ clang: error: linker command failed with exit code 1 (use -v to see invocation)
This is actually a bug in the CMakeList.txt because the library-list for the glm.so module does not contain the liblua.
If you apply the following patch
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 8cf8a1a6..6eda2a17 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
## -940,7 +940,7 ## ENDIF()
ADD_LIBRARY(glm MODULE ${SRC_LIBGLM})
TARGET_INCLUDE_DIRECTORIES(glm PRIVATE ${INCLUDE_DIRECTORIES})
-TARGET_LINK_LIBRARIES(glm PRIVATE ${LIBS})
+TARGET_LINK_LIBRARIES(glm PRIVATE ${LIBS} liblua_static)
IF( LUA_BUILD_AS_DLL )
TARGET_LINK_LIBRARIES(glm PUBLIC ${interpretor_target})
TARGET_COMPILE_DEFINITIONS(glm PRIVATE LUA_BUILD_AS_DLL)
to the freshly cloned repository and then do (inside the cloned repository):
git submodule update --init
mkdir build
cd build
cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release -DONE_LUA=ON ..
make
you should find the files glm.so, lua and luac in your build directory. At least that worked on my Intel MacBook.
Using liblua_static in the CMakeList.txt for the library glm makes sure, that the resulting glm.so does not need the liblua dynamic library but contains the required code statically. That avoids problems when loading glm.so and the liblua library not being in the library path.
I opened an issue in the GitHub repository for it. So hopefully it will get fixed for everybody soon.
Usually it'll be something like -lYourLibrary not found, if -lSystem for example is not found, try adding it to your CMakeLists.
Are you still having the issue?
I'm interested in statically linking PROJ to a library created with pybind11.
but, I get error by cmake.
Why am I getting this error and how can I fix it?
My CMakeLists.txt is this.
cmake_minimum_required(VERSION 3.4)
project(myearth LANGUAGES CXX)
add_subdirectory(pybind11)
pybind11_add_module(myearth MyEarth.cpp)
include_directories(${CMAKE_SOURCE_DIR}/PROJ/include)
add_library(proj4 STATIC IMPORTED)
set_target_properties(proj4 PROPERTIES IMPORTED_LOCATION ${CMAKE_SOURCE_DIR}/PROJ/lib/libproj.a)
target_link_libraries(myearth PRIVATE proj4)
CMake Error.
I think fPIC option is for shared library. so, libproj.a is not necessary this option because this is static library.
(venv) ~/Projects/LinkLibTest/MySample/build$ make
[ 50%] Building CXX object CMakeFiles/myearth.dir/MyEarth.cpp.o
[100%] Linking CXX shared module myearth.cpython-36m-x86_64-linux-gnu.so
/usr/bin/ld: ../PROJ/lib/libproj.a(proj_4D_api.o): relocation R_X86_64_PC32 against symbol `pj_errno' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status
make[2]: *** [CMakeFiles/myearth.dir/build.make:98: myearth.cpython-36m-x86_64-linux-gnu.so] Error 1
make[1]: *** [CMakeFiles/Makefile2:100: CMakeFiles/myearth.dir/all] Error 2
make: *** [Makefile:91: all] Error 2
(venv) ~/Projects/LinkLibTest/MySample/build$
I created libproj.a by the following command.
$ ./configure --prefix=/output --disable-shared
$ sudo make
$ sudo make install
Linux Mint 20.3 64bit
CMake v3.22.2
PROJ v5.2.0 (https://proj.org/)
Python v3.6.8
Thank you.
Shared libraries must have position independent code, but your version of PROJ was built without it; you will have to rebuild PROJ with PIC.
$ ./configure --prefix=/output --with-pic
I am able to do this without CMake using a handwritten Makefile, like so:
g++ $(CXSCINC) -c -fPIC cellComplex_extern.cpp -o cellComplex_extern.o
g++ $(CXSCINC) -shared -Wl -o cellComplex_lib.so cellComplex_extern.o $(CXSCLIB) -lcxsc
This gets me shared library cellComplex_lib.so, which then gets picked up by ctypes as a dynamically linked library (lib = ctypes.cdll.LoadLibrary('./cellComplex_lib.so') for later use.
My project has moved to CMake as a build system and I am looking to emulate the functionality of my Makefile above.
So far I have discovered the add_library() command for CMakeLists.txt, but the link to the CXSC library is never made (as evidenced by horrible complaining when I run make after cmake.
How can I tell CMake to build cellComplex_lib with the third-party library CXSC?
-- non-working CMakeLists.txt --
add_library(include/python/cellComplex_extern OBJECT
include/python/cellComplex_extern.cpp ${all_headers})
add_library(include/python/cellComplex_lib SHARED
include/python/cellComplex_extern)
target_link_libraries(include/python/cellComplex_lib ${CXSC_LIB_DIR}/libcxsc.a)
Result of running cmake followed by make:
.
.
.
[ 75%] Built target include/python/cellComplex_extern
Linking CXX shared library libinclude/python/cellComplex_lib.dylib
ld: can't open output file for writing: libinclude/python/cellComplex_lib.dylib, errno=2 for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[2]: *** [libinclude/python/cellComplex_lib.dylib] Error 1
make[1]: *** [CMakeFiles/include/python/cellComplex_lib.dir/all] Error 2
make: *** [all] Error 2
I think you need to use target_link_libraries
target_link_libraries(include/python/cellComplex_lib ${CXSLIB})
This is what I use during Win32 development:
link_directories(${LIB_ROOT_DIR}/lib ${LIB_ROOT_DIR2}/lib/morelibs)
add_library(MyDll1 SHARED File1.cpp File2.cpp)
add_library(MyDll2 SHARED File3.cpp File4.cpp)
add_dependencies(MyDll2 Dll1)
target_link_libraries(MyDll2 Dll1 some.lib another.lib)
Here you specify that Dll2 requires Dll1 and two other external lib's.