Clang link libc++ statically - c++

On Ubuntu 20.04 I need to use libc++ since the default libstdc++ is lacking required support. In order to not have to ship libc++ I need to statically link against libc++.
I have a static version of libc++ available which has been created locally using the following series of commands (as per https://libcxx.llvm.org/BuildingLibcxx.html):
git clone --single-branch --branch release/15.x --depth 1 https://github.com/llvm/llvm-project.git
mkdir /llvm-project/build
cmake -G "Ninja" -S /llvm-project/runtimes -B /llvm-project/build \
-DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi;libunwind" \
-DLIBCXX_INCLUDE_BENCHMARKS=OFF \
-DLIBCXX_INCLUDE_TESTS=OFF \
-DLIBCXX_ENABLE_SHARED=ON \
-DLIBCXX_ENABLE_STATIC=ON \
-DLIBCXXABI_ENABLE_SHARED=ON \
-DLIBCXXABI_ENABLE_STATIC=ON \
-DLIBCXX_CXX_ABI=libcxxabi \
-DCMAKE_INSTALL_PREFIX=/usr \
-DCMAKE_BUILD_TYPE=Release
ninja -C /llvm-project/build cxx cxx-abi unwind
ninja -C /llvm-project/build install-cxx install-cxxabi install-unwind
I fail to make it achieve the desired results of not of not having any runtime dependency on libc++.
Trying to compile the following simple example:
#include <iostream>
int main(int, const char**) {
std::cout << "Clang " << __clang_version__ << '\n';
#ifdef _LIBCPP_VERSION
std::cout << " Using libc++ " << _LIBCPP_VERSION << '\n';
#endif
return 0;
}
Using clang++ test.cpp -std=c++20 -static-libstdc++ -static-libgcc will result in no undesired runtime dependencies (only linux-vdso, libm, libc, ld-linux), however it is using stdlibc++ which will not work for the actual code.
Using clang++ test.cpp -stdlib=libc++ -std=c++20 -static-libstdc++ -static-libgcc results in lots of linker errors about missing standard stuff.
Using clang++ test.cpp -stdlib=libc++ -std=c++20 -static-libstdc++ -static-libgcc -lc++abi followed by an ldd a.out will show it having a dependency on libc++abi.so and libgcc_s.so.1 which are both undesired, the former more so than the latter.
(Why do I have to explicitly link c++abi anyway, should it not just do it by itself like it does with c++?)
In an attempt to not have the pesky runtime dependency doing clang++ test.cpp -stdlib=libc++ -std=c++20 -static-libstdc++ -static-libgcc -lc++abi -static will result in
ld: error: duplicate symbol: __lll_lock_wait_private
>>> defined at libc-lowlevellock.o:(__lll_lock_wait_private) in archive /lib/x86_64-linux-gnu/libc.a
>>> defined at lowlevellock.c:27
>>> lowlevellock.o:(.text+0x0) in archive /lib/x86_64-linux-gnu/libpthread.a
No idea why it suddenly decides to have pthread in there, as I never specified that and it would not seem to be required given the example code but whatever.
How can one convince clang to actually statically link libc++ (and libc++abi), without errors, so that it will not have non-standard runtime dependencies?

The entire C++ standard library (libc++/libstdc++ etc) all rely on the standard C library and pthreads among others.
C++ standard libs on Linux are just a tiny shell around the infrastructure that the C language provides, the real workhorse.
std::thread is basically just syntax sugar on top of pthreads. This is a hard reality most C++ developers find very hard to stomach.
See examples:
std::threadlibstdc++ libc++
std::chrono libstdc++ libc++
std::filesystem libstdc++ libc++

Related

Automatically select a C++11-compatible g++ version in the Makefile

The problem:
2 version of g++ installed on a computer running Ubuntu 12.04. They are the versions 4.6 and 5.2.
I have to compile a C++11 program using a Makefile. If I use g++ as compiler it calls automatically the version 4.6, it does not support c++11 so the compilation fails. I've followed a tutorial online, so that now if I call g++ it calls automatically the version 5.2 and now it works.
I find this solution not so good, since it works only on my PC. Is there a way to recognize in the Makefile if the default g++ version support C++11 and, in case not, switch to a more recent version?
Thank you!
Is there a way to recognize in the Makefile if the default g++ version support C++11 and, in case not, switch to a more recent version?
You can certainly detect the version of the default compiler available in PATH in your makefile. However, where do you search for another version?
The standard approach is to let the user specify the C compiler through CC and C++ compiler through CXX make variables, e.g.: make CC=/path/to/my/gcc CXX=/path/to/my/g++.
You can always select which gcc to use while invoking make
make CXX=/gcc/path/of/your/choice
otherwise you can detect gcc version using
ifdef CXX
GCC_VERSION = $(shell $(CXX) -dumpversion)
else
GCC_VERSION = $(shell g++ -dumpversion)
endif
in Makefile and while using, you can test if your gcc is >=4.6
ifeq ($(shell expr $(GCC_VERSION) '>=' 4.6), 1)
UPDATE: newer gcc needs -dumpfullversion together (icx is the CC from Intel OneAPI)
$ icx -dumpversion
14.0.0
$ gcc -dumpversion
9
$ icx -dumpfullversion -dumpversion
14.0.0
$ gcc -dumpfullversion -dumpversion
9.3.1
One very simple way is to use conditional statements in your makefile, and go for versions which you know are compatible, and only use the default gcc as a fallback. Here's a basic example:
CXX=g++
ifeq (/usr/bin/g++-4.9,$(wildcard /usr/bin/g++-4.9*))
CXX=g++-4.9
# else if... (a list of known-to-be-ok versions)
endif
The other, more robust method, is to generate your makefile using a script that checks for capabilities using test compilations, kind of like what ./configure usually does. I really don't mean to recommend autotools, though.
The thing to do is build your Makefile to use as many implicit rules as possible. By default compilation uses various environment variables.
The variable $(CXX) is the C++ compiler command and defaults to g++ on Linux systems. So clanging CXX to a different compiler executable will change the compiler for all implicit compile commands.
When you write explicit rules use the same variable that the implicit rules use. So instead of this:
program: program.cpp
g++ -o program program.cpp
Do this:
program: program.cpp
$(CXX) -o program program.cpp
Other variables you should use are:
CPPFLAGS = -Iinclude
CXXFLAGS = -std=c++14 -g3 -O0
Those are for pre-processing flags CPPFLAGS and compiler flags CXXFLAGS and library linking flags LDLIBS.
Using the default environment variables allows the person compiling the project the freedom to control the compilation for their desired environment.
See the GNU make manual
This works for me:
cmake_minimum_required(VERSION 2.8.11)
project(test)
if (${CMAKE_CXX_COMPILER_VERSION} LESS 5.0)
message(FATAL_ERROR "You need a version of gcc > 5.0")
endif (${CMAKE_CXX_COMPILER_VERSION} LESS 5.0)
add_executable(test test.cpp)
You can check in your source code the gcc version and abort compilation if you don't like it. Here is how it works:
/* Test for GCC > 4.6 */
#if !(__GNUC__ > 3 && __GNUC_MINOR__ > 6)
#error gcc above version 4.6 required!
#endif

undefined reference to `shm_open' already with -L /lib -lrt -lpthread

I just want to use the boost library to create a shared memory on an ARM system. It work fine if you want to compile it only under ubuntu. However, when I want to cross compile it with TI's CCSv6 and angstrom toolchain, it keep pushing errors.
Because I do not know how to write a makefile for cross compile, I think using TI their own IDE might be a good choice to avoid further problems.
Here is my code and print out of build console.
#include <boost/interprocess/shared_memory_object.hpp>
#include <boost/interprocess/mapped_region.hpp>
#include <iostream>
using namespace boost::interprocess;
int main()
{
shared_memory_object shdmem{open_or_create, "Boost1", read_write};
shdmem.truncate(1024);
mapped_region region{shdmem, read_write};
}
g++ -std=c++0x -I/usr/include -O0 -g3 -Wall -c -fmessage-length=0 -L /lib -lrt -lpthread -fPIC
The IDE called Code Composer Studio has cross compile settings as below:
Prefix: arm-angstrom-linux-gnueabi-
Path: /usr/local/oecore-x86_64/sysroots/x86_64-angstromsdk-linux/usr/bin/armv5te-angstrom-linux-gnueabi
Build Console:
/usr/include/boost/interprocess/shared_memory_object.hpp:309: undefined reference to shm_open'
/usr/include/boost/interprocess/shared_memory_object.hpp:315: undefined reference toshm_open'
/usr/include/boost/interprocess/shared_memory_object.hpp:327: undefined reference to shm_open'
/usr/include/boost/interprocess/shared_memory_object.hpp:334: undefined reference toshm_open'
collect2: ld returned 1 exit status
make: *** [test] Error 1
undefined reference to shm_open' means it cannot find -lrt for ARM.
In your build command line you need to specify include and library paths to ARM built libraries, not to Ubuntu ones. So -I/usr/include and -L /lib is wrong.
Also you need boost built for ARM, although if you just want to use interprocess library then boost headers should be enough. But you need to copy them into different location because including them from /usr/include includes also other headers specific to Ubuntu.
You can use the cross compiler IDE you mentioned or arm g++ cross compiler which you can install by:
sudo apt-get install g++-arm-linux-gnueabihf. Some headers and libraries for ARM will be installed too.

Need some help figuring out compile-time error: 'Undefined symbols for architecture x86_64: "boost::system::system_category()"' [duplicate]

I'm trying to compile a program on Ubuntu 11.10 that uses the Boost libraries. I have the 1.46-dev Boost libraries from the Ubuntu Repository installed, but I get an error when compiling the program.
undefined reference to boost::system::system_category()
What is it that I do wrong?
The boost library you are using depends on the boost_system library. (Not all of them do.)
Assuming you use gcc, try adding -lboost_system to your compiler command line in order to link against that library.
Linking with a library that defines the missing symbol (-lboost_system) is the obvious solution, but in the particular case of Boost.System, a misfeature in the original design makes it use boost::system::generic_category() and boost::system::system_category() needlessly. Compiling with the flag -DBOOST_SYSTEM_NO_DEPRECATED disables that code and lets a number of programs compile without requiring -lboost_system (that link is of course still needed if you explicitly use some of the library's features).
Starting from Boost 1.66 and this commit, this behavior is now the default, so hopefully fewer and fewer users should need this answer.
As noticed by #AndrewMarshall, an alternative is to define BOOST_ERROR_CODE_HEADER_ONLY which enables a header-only version of the code. This was discouraged by Boost as it can break some functionality. However, since 1.69, header-only seems to have become the default, supposedly making this question obsolete.
Another workaround for those who don't need the entire shebang: use the switch
-DBOOST_ERROR_CODE_HEADER_ONLY.
If you use CMake, it's add_definitions(-DBOOST_ERROR_CODE_HEADER_ONLY).
The above error is a linker error... the linker a program that takes one or more objects generated by a compiler and combines them into a single executable program.
You must add -lboost_system to you linker flags which indicates to the linker that it must look for symbols like boost::system::system_category() in the library libboost_system.so.
If you have main.cpp, either:
g++ main.cpp -o main -lboost_system
OR
g++ -c -o main.o main.cpp
g++ main.o -lboost_system
When using CMAKE and find_package, make sure it is :
find_package(Boost COMPONENTS system ...)
and not
find_package(boost COMPONENTS system ...)
Some people may have lost hours for that ...
I got the same Problem:
g++ -mconsole -Wl,--export-all-symbols -LC:/Programme/CPP-Entwicklung/MinGW-4.5.2/lib -LD:/bfs_ENTW_deb/lib -static-libgcc -static-libstdc++ -LC:/Programme/CPP-Entwicklung/boost_1_47_0/stage/lib \
D:/bfs_ENTW_deb/obj/test/main_filesystem.obj \
-o D:/bfs_ENTW_deb/bin/filesystem.exe -lboost_system-mgw45-mt-1_47 -lboost_filesystem-mgw45-mt-1_47
D:/bfs_ENTW_deb/obj/test/main_filesystem.obj:main_filesystem.cpp:(.text+0x54):
undefined reference to `boost::system::generic_category()
Solution was to use the debug-version of the system-lib:
g++ -mconsole -Wl,--export-all-symbols -LC:/Programme/CPP-Entwicklung/MinGW-4.5.2/lib -LD:/bfs_ENTW_deb/lib -static-libgcc -static-libstdc++ -LC:/Programme/CPP-Entwicklung/boost_1_47_0/stage/lib \
D:/bfs_ENTW_deb/obj/test/main_filesystem.obj \
-o D:/bfs_ENTW_deb/bin/filesystem.exe -lboost_system-mgw45-mt-d-1_47 -lboost_filesystem-mgw45-mt-1_47
But why?
When I had this, problem, the cause was the ordering of the libraries. To fix it, I put libboost_system last:
g++ mingw/timer1.o -o mingw/timer1.exe -L/usr/local/boost_1_61_0/stage/lib \
-lboost_timer-mgw53-mt-1_61 \
-lboost_chrono-mgw53-mt-1_61 \
-lboost_system-mgw53-mt-1_61
This was on mingw with gcc 5.3 and boost 1.61.0 with a simple timer example.
in my case, adding -lboost_system was not enough, it still could not find it in my custom build environment. I had to use the advice at Get rid of "gcc - /usr/bin/ld: warning lib not found" and change my ./configure command to:
./configure CXXFLAGS="-I$HOME/include" LDFLAGS="-L$HOME/lib -Wl,-rpath-link,$HOME/lib" --with-boost-libdir=$HOME/lib --prefix=$HOME
for more details see Boost 1.51 : "error: could not link against boost_thread !"
...and in case you wanted to link your main statically, in your Jamfile add the following to requirements:
<link>static
<library>/boost/system//boost_system
and perhaps also:
<linkflags>-static-libgcc
<linkflags>-static-libstdc++

Regex Boost library linking in release mode warns "duplicate section has different size" when using mingw-w64 toolchain

When linking my project in the release mode I am getting the following warning:
myProject-libs/release/libboost_regex-mt-s-1.50.0.a(cpp_regex_traits.o): duplicate section `.data$_ZZN5boost16cpp_regex_traitsIcE21get_catalog_name_instEvE6s_name[boost::cpp_regex_traits<char>::get_catalog_name_inst()::s_name]' has different size
I suspect that the cause could be that the boost library is compiled with different options than I use for my project, but I don't know how to find the difference (boost didn't output these options during the build).
In order to compile the boost for win32 on Ubuntu 12.04 I used this procedure:
tar jxf boost_1_50_0.tar.bz2
cd boost_1_50_0
./bootstrap.sh
echo "using gcc : 4.6 : i686-w64-mingw32-g++ : <rc>i686-w64-mingw32-windres <archiver>i686-w64-mingw32-ar ;" > user-config.jam
./bjam toolset=gcc target-os=windows --address-model=32 variant=release threading=multi threadapi=win32 link=static runtime-link=static --prefix=/opt/boost_1_50_0-release-static-windows-32 --user-config=user-config.jam -j 10 --without-mpi --without-python -sNO_BZIP2=1 -sNO_ZLIB=1 --layout=tagged install
In order to compile files in my project I use something like
i686-w64-mingw32-g++ -march=corei7 -mfpmath=sse -m32 -Wall -fmessage-length=0 -I"/opt/boost_1_50_0-release-static-windows-32/include" -std=c++0x -O3 -g0 -DNDEBUG -I"myProject/src/cpp" -c -o myProject/build/release/src/cpp/myproject.o myproject/src/cpp/myproject.cpp
The tests I have indicate that the regexps run fine but still I would like to resolve the warning.
EDIT
I found that additional options to the boost compiler can be added using a cxxflags= argument of bjam.
Example:
bjam cxxflags='-fPIC' ....
Maybe making sure to pass the same arguments as I do to the project could resolve the problem (particularly the arguments related to optimizations as suggested in the linked question).
Your compilers were compiled with different options :) Compiling the library on Linux and the program on Windows result in a situation where there is an identically named .data segment, but they aren't the same size. That could theoretically be interesting, inasmuch as a data segment is writable, but in practice, it shouldn't matter. Unless there is evidence to suggest this causes a problem of which I'm not aware, you can safely suppress that warning; I don't know how you'd make it go away, though.
I recently encountered this problem (i.e. linker warning "duplicate section has different size") when trying to compile boost for Windows using mingw.
The issue I had was that I compiled my application with -std=c++14 but when compiling boost I didn't specifically provide a dialect flag (which defaulted to -std=c++98 for g++ 5.3.0). Adding the dialect flag -std=c++14 when compiling boost solved the problem for me. See this answer for an explaination on how to set cxxflags when compiling boost.
I believe my solution might have worked for you (your application was compiled with -std=c++0x but boost was not provided any dialect flag). Although I am 6 years too late, I'll leave my answer here for others who happen to stumble-upon this issue.

undefined reference to boost::system::system_category() when compiling

I'm trying to compile a program on Ubuntu 11.10 that uses the Boost libraries. I have the 1.46-dev Boost libraries from the Ubuntu Repository installed, but I get an error when compiling the program.
undefined reference to boost::system::system_category()
What is it that I do wrong?
The boost library you are using depends on the boost_system library. (Not all of them do.)
Assuming you use gcc, try adding -lboost_system to your compiler command line in order to link against that library.
Linking with a library that defines the missing symbol (-lboost_system) is the obvious solution, but in the particular case of Boost.System, a misfeature in the original design makes it use boost::system::generic_category() and boost::system::system_category() needlessly. Compiling with the flag -DBOOST_SYSTEM_NO_DEPRECATED disables that code and lets a number of programs compile without requiring -lboost_system (that link is of course still needed if you explicitly use some of the library's features).
Starting from Boost 1.66 and this commit, this behavior is now the default, so hopefully fewer and fewer users should need this answer.
As noticed by #AndrewMarshall, an alternative is to define BOOST_ERROR_CODE_HEADER_ONLY which enables a header-only version of the code. This was discouraged by Boost as it can break some functionality. However, since 1.69, header-only seems to have become the default, supposedly making this question obsolete.
Another workaround for those who don't need the entire shebang: use the switch
-DBOOST_ERROR_CODE_HEADER_ONLY.
If you use CMake, it's add_definitions(-DBOOST_ERROR_CODE_HEADER_ONLY).
The above error is a linker error... the linker a program that takes one or more objects generated by a compiler and combines them into a single executable program.
You must add -lboost_system to you linker flags which indicates to the linker that it must look for symbols like boost::system::system_category() in the library libboost_system.so.
If you have main.cpp, either:
g++ main.cpp -o main -lboost_system
OR
g++ -c -o main.o main.cpp
g++ main.o -lboost_system
When using CMAKE and find_package, make sure it is :
find_package(Boost COMPONENTS system ...)
and not
find_package(boost COMPONENTS system ...)
Some people may have lost hours for that ...
I got the same Problem:
g++ -mconsole -Wl,--export-all-symbols -LC:/Programme/CPP-Entwicklung/MinGW-4.5.2/lib -LD:/bfs_ENTW_deb/lib -static-libgcc -static-libstdc++ -LC:/Programme/CPP-Entwicklung/boost_1_47_0/stage/lib \
D:/bfs_ENTW_deb/obj/test/main_filesystem.obj \
-o D:/bfs_ENTW_deb/bin/filesystem.exe -lboost_system-mgw45-mt-1_47 -lboost_filesystem-mgw45-mt-1_47
D:/bfs_ENTW_deb/obj/test/main_filesystem.obj:main_filesystem.cpp:(.text+0x54):
undefined reference to `boost::system::generic_category()
Solution was to use the debug-version of the system-lib:
g++ -mconsole -Wl,--export-all-symbols -LC:/Programme/CPP-Entwicklung/MinGW-4.5.2/lib -LD:/bfs_ENTW_deb/lib -static-libgcc -static-libstdc++ -LC:/Programme/CPP-Entwicklung/boost_1_47_0/stage/lib \
D:/bfs_ENTW_deb/obj/test/main_filesystem.obj \
-o D:/bfs_ENTW_deb/bin/filesystem.exe -lboost_system-mgw45-mt-d-1_47 -lboost_filesystem-mgw45-mt-1_47
But why?
When I had this, problem, the cause was the ordering of the libraries. To fix it, I put libboost_system last:
g++ mingw/timer1.o -o mingw/timer1.exe -L/usr/local/boost_1_61_0/stage/lib \
-lboost_timer-mgw53-mt-1_61 \
-lboost_chrono-mgw53-mt-1_61 \
-lboost_system-mgw53-mt-1_61
This was on mingw with gcc 5.3 and boost 1.61.0 with a simple timer example.
in my case, adding -lboost_system was not enough, it still could not find it in my custom build environment. I had to use the advice at Get rid of "gcc - /usr/bin/ld: warning lib not found" and change my ./configure command to:
./configure CXXFLAGS="-I$HOME/include" LDFLAGS="-L$HOME/lib -Wl,-rpath-link,$HOME/lib" --with-boost-libdir=$HOME/lib --prefix=$HOME
for more details see Boost 1.51 : "error: could not link against boost_thread !"
...and in case you wanted to link your main statically, in your Jamfile add the following to requirements:
<link>static
<library>/boost/system//boost_system
and perhaps also:
<linkflags>-static-libgcc
<linkflags>-static-libstdc++