I am attempting to compile gcc 4.4.0 on opensolaris 2009.6
Currently in the box (which is a AMD 64bit machine), I have the gcc 3.4.6 installed.
I unpacked the gcc 4.4.0 tarball.
I set the following env variables:
export CXX=/usr/local/bin/g++
export CC=/usr/local/bin/gcc
Then I ran "configure && make" and this is the error message that I got:
checking for i386-pc-solaris2.11-gcc... /export/home/me/wd/gcc/gcc-4.4.0/host-i386-pc-solaris2.11/gcc/xgcc -B/export/home/me/wd/gcc/gcc-4.4.0/host-i386-pc-solaris2.11/gcc/ -B/usr/local/i386-pc-solaris2.11/bin/ -B/usr/local/i386-pc-solaris2.11/lib/ -isystem /usr/local/i386-pc-solaris2.11/include -isystem /usr/local/i386-pc-solaris2.11/sys-include -m64
checking for suffix of object files... configure: error: in `/export/home/me/wd/gcc/gcc-4.4.0/i386-pc-solaris2.11/amd64/libgcc':
configure: error: cannot compute suffix of object files: cannot compile
See `config.log' for more details.
Anyone has any suggestion as to how to work around this error message?
/Edit:
Content of the config.log is posted here: link text
Normally the GCC build is bootstrapped, i.e. first it uses the system compiler to build GCC C compiler, and then it uses the freshly built compiler to recompile GCC once again (and then even once more time again). The configure line shows that it is not the system compiler but the already-built GCC compiler which is used for configure test there.
Since it fails, the problem is that the freshly-built GCC is somehow "stillborn" here. If config.log will not help you, I'd suggest to ask at gcc-help#gcc.gnu.org.
EDIT: Ah-ha, I think it is the assembler. You are using GNU assembler, but the unsupported options look like they were meant for Sun assembler. This should be solved by adding --with-gnu-as configure option (and then possibly having to specify its path explicitly with --with-as=/usr/gnu/bin/as)
You can also take a look at Solaris-specific GCC build instructions.
There's a readily available build for gcc4, which you can try updating. Its current version is 4.3.3. To get started, install pkg-get from OpenCSW and check out the build from the subversion repository:
svn co https://gar.svn.sourceforge.net/svnroot/gar/csw/mgar/pkg/gcc4/trunk/ gcc4
cd gcc4
gmake package
Related
I struggle with Caffe compilation. Unfortunately I failed to compile it.
Steps I followed:
git clone https://github.com/BVLC/caffe.git
cd caffe
mkdir build
cd build
cmake ..
make all
Running make all fails with the following error message:
[ 2%] Building NVCC (Device) object src/caffe/CMakeFiles/cuda_compile.dir/util/cuda_compile_generated_im2col.cu.o
In file included from /usr/include/cuda_runtime.h:59:0,
from <command-line>:0:
/usr/include/host_config.h:82:2: error: #error -- unsupported GNU version! gcc 4.9 and up are not supported!
#error -- unsupported GNU version! gcc 4.9 and up are not supported!
^
CMake Error at cuda_compile_generated_im2col.cu.o.cmake:207 (message):
Error generating /mydir/caffe/build/src/caffe/CMakeFiles/cuda_compile.dir/util/./cuda_compile_generated_im2col.cu.o
Software version:
OS: Debian.
gcc version: 5.3.1.
nvcc version: 6.5.12.
cat /proc/driver/nvidia/version result:
NVRM version: NVIDIA UNIX x86_64 Kernel Module 352.63 Sat Nov 7 21:25:42 PST 2015
GCC version: gcc version 4.8.5 (Debian 4.8.5-3)
Attempts to solve the problem
1st try
Simple solutions are often best ones, so (as suggested here) I tried to comment out macro checking gcc version from /usr/include/host_config.h (line 82). Unfortunately it doesn't work and compilation fails badly:
1 catastrophic error detected in the compilation of "/tmp/tmpxft_000069c2_00000000-4_im2col.cpp4.ii".
2nd try
I tried to run:
cmake -D CMAKE_CXX_COMPILER=g++-4.8 ..
make
but it fails with exactly the same error message (even though g++-4.8 should be accepted).
3rd try
I've found similar problem (though not related to Caffe) and I tried to solve it as suggested in the accepted answer.
What I did:
I've ran grep -iR "find_package(CUDA" caffe command and found Cuda.cmake file which has find_package(CUDA 5.5 QUIET) in line 225.
I added set(CUDA_HOST_COMPILER /usr/bin/gcc-4.8) to Cuda.cmake, line before line: find_package(CUDA 5.5 QUIET).
I removed everything from build directory and ran cmake and make again - with and without -D CMAKE_CXX_COMPILER=g++-4.8.
Unfortunately result is exactly the same. Caffe probably overwrites it somehow - I didn't figure it out how.
make VERBOSE=1 2>&1 | grep -i compiler-bindir returns nothing.
What's interesting, make VERBOSE=1 prints command that fails, which is:
/usr/bin/nvcc -M -D__CUDACC__ /mydir/caffe/src/caffe/util/im2col.cu -o /mydir/caffe/build/src/caffe/CMakeFiles/cuda_compile.dir/util/cuda_compile_generated_im2col.cu.o.NVCC-depend -ccbin /usr/bin/cc -m64 -DUSE_LMDB -DUSE_LEVELDB -DUSE_OPENCV -DWITH_PYTHON_LAYER -DGTEST_USE_OWN_TR1_TUPLE -Xcompiler ,\"-fPIC\",\"-Wall\",\"-Wno-sign-compare\",\"-Wno-uninitialized\",\"-O3\",\"-DNDEBUG\" -gencode arch=compute_20,code=sm_21 -Xcudafe --diag_suppress=cc_clobber_ignored -Xcudafe --diag_suppress=integer_sign_change -Xcudafe --diag_suppress=useless_using_declaration -Xcudafe --diag_suppress=set_but_not_used -Xcompiler -fPIC -DNVCC -I/usr/include -I/mydir/caffe/src -I/usr/include -I/mydir/caffe/build/include -I/usr/include/hdf5/serial -I/usr/include/opencv -I/usr/include/atlas -I/usr/include/python2.7 -I/usr/lib/python2.7/dist-packages/numpy/core/include -I/mydir/caffe/include -I/mydir/caffe/build
when I add --compiler-bindir /usr/bin/gcc-4.8 flag manually, it prints error:
nvcc fatal : redefinition of argument 'compiler-bindir'
which may be related to this bug report.
Edit: I didn't notice that --compiler-bindir and -ccbin are the same options, and the latter is already set in above command that failed. When I changed -ccbin /usr/bin/cc to -ccbin /usr/bin/gcc-4.8 in above command that failed, it completes successfully. Now I need to find option in Caffe's CMake file that overwrite -ccbin in all subsequent Caffe's CMakes. Looking at cmake/Cuda.cmake:252:list(APPEND CUDA_NVCC_FLAGS ${NVCC_FLAGS_EXTRA} seems to be good way to go.
How can I successfully complete my compilation? Any help is appreciated.
Related SO questions:
host_config.h:unsupported GNU version! gcc versions later than 4.9 are not supported.
CUDA 6.5 complains about not supporting gcc 4.9 - what to do?.
cmake -D CUDA_NVCC_FLAGS="-ccbin gcc-4.8" .. && make causes successful compilation.
Now another problem showed up: linking Google's libgflags or libprotobuf fails probably due to fact that it was compiled with newer gcc version but it's not related to asked question.
My machine runs Ubuntu 15.10, and my default compiler version is gcc 5.2.1 .
Commenting out the #error directive in line 115 of file
/usr/local/cuda-7.5/include/host_config.h
(or whatever the path on your system is) did the trick for me. Caffe compiled fine, all tests ran smoothly.
On the other hand, if one chooses to ignore this and proceed to compile part of the project with one compiler version, part of the project with another (for me it was gcc-4.8 and gcc-5.2.1), linking problems will arise. The linking problems of protobuf and libgflags another answer mentions are not unrelated to this.
I am trying to build GraphChi on OS X Yosemite but get the following error:
fatal error: 'omp.h' file not found
From this question - How to include omp.h in OS X? - I learned that Yosemite uses Clang instead of gcc, which does not include omp.h.
$ which gcc
/usr/bin/gcc
$ gcc -v
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 6.0 (clang-600.0.56) (based on LLVM 3.5svn)
Target: x86_64-apple-darwin14.1.0
Thread model: posix
Next, I installed gcc via Homebrew
$ brew info gcc
gcc: stable 4.9.2 (bottled)
http://gcc.gnu.org
/usr/local/Cellar/gcc/4.9.2_1 (1092 files, 177M)
Built from source with: --without-multilib
and updated $PATH to include the path to the new gcc version
$ echo $PATH
/usr/local/Cellar/gcc/4.9.2_1:usr/local/bin:/opt/local/bin:/opt/local/sbin:/usr/bin:/bin:/usr/sbin:/sbin
however, gcc -v and which gcc still point to the old version, and building GraphChi still doesn't work due to the missing omp.h file
Does anyone know what else I need to do?
Update
locate omp.h returned:
/usr/local/Cellar/apple-gcc42/4.2.1-5666.3/lib/gcc/i686-apple-darwin11/4.2.1/include/omp.h
/usr/local/Cellar/gcc/4.9.2_1/lib/gcc/4.9/gcc/x86_64-apple-darwin14.1.0/4.9.2/include/omp.h
/usr/local/Cellar/gfortran/4.8.2/gfortran/lib/gcc/x86_64-apple-darwin13.0.0/4.8.2/include/omp.h
my ~/.profile:
export PATH=/usr/local/Cellar/gcc/4.9.2_1/lib/gcc/4.9/gcc/x86_64-apple-darwin14.1.0/4.9.2/include:/usr/local/Cellar/gcc/4.9.2_1/bin:usr/local/bin:/opt/local/bin:/opt/local/sbin:$PATH
I solved this with installing gcc with homebrew:
brew install gcc --without-multilib
and then building the source code with
CC=gcc-5 CXX=g++-5 cmake ..
CC=gcc-5 CXX=g++-5 make -j7
Once you have installed gcc-4.9 with homebrew, it will automatically be in your path. To use OpenMP, you just need to make sure you are using the newly installed gcc-4.9, and it will be able to find omp.h.
In the case of GraphChi, you will have to go change line 3 of the Makefile to be gcc-4.9. From there, running make should just work. They describe this in their README, but at least the version they describe is out of date https://github.com/GraphChi/graphchi-cpp#problems-compiling-on-mac.
clang does not support OpenMP yet. Also gcc by default links to Apple's LLVM clang compiler (not the GCC installed from brew).
Instead gcc-4.9 would link to GCC. I think if -fopenmp is specified omp.h is included automatically.
It is possible to manually build a version of clang with OpenMP support, see http://clang-omp.github.io
You shouldn't add the include path to PATH; instead, specify it as CFLAGS, including the -I option. You can export the CFLAGS variable, or set it on the fly.
Depending on how you compile things, you could do
CFLAGS=-I/usr/local/Cellar/gcc/4.9.2_1/lib/gcc/4.9/gcc/x86_64-apple-darwin14.1.0/4.9.2/include/omp.h gcc <whatever>
Of course, in this case you can specify it directly on the gcc command (as -I/usr/local/....), but the CFLAGS variable also works with configure (as configure often won't have an option to specify where it should look for specific include files); probably with make, or even for those installing a Python package: CFLAGS=-I... pip install <some-package>.
Other flags to consider are
CXXFLAGS: C++ specific pre-processor flags
LDFLAGS: linker specific flags (e.g. LDFLAGS=-L/some/path/... for linking with dynamic libraries).
CC: specify the C compiler to use. This is an easy way to avoid the built-in gcc alias for clang on OS X. Just use CC=/usr/local/bin/gcc-4 make or similar.
CXX: specify the C++ compiler to use.
According to this post (https://github.com/mxcl/homebrew/pull/2953), the flag "--with-mpi" should enable boost_mpi build support for the related homebrew formula, so I am trying to install boost via homebrew like this:
brew install boost --with-mpi
However, the actual boost mpi library is not being build and can not be found.
There is currently some work being done around this, according to: https://github.com/mxcl/homebrew/pull/15689
In summary, I can currently build boost, but it seems the "--with-mpi" flag is being ignored. Could someone please check, if I should be able to build boost (with mpi support) on Mac OS X Mountain Lion (10.8)?
The (verbose) output generates these lines:
MPI auto-detection failed: unknown wrapper compiler mpic++
Please report this error to the Boost mailing list: http://www.boost.org
You will need to manually configure MPI support.
warning: skipping optional Message Passing Interface (MPI) library.
note: to enable MPI support, add "using mpi ;" to user-config.jam.
note: to suppress this message, pass "--without-mpi" to bjam.
note: otherwise, you can safely ignore this message.
Not sure how exactly I can fix this and get the mpi stuff to be build - any ideas?
Just in case this helps anyone else along the line, here's how I fixed this. The main error is MPI auto-detection failed: unknown wrapper compiler mpic++, any typing mpic++ at the command line verified that it was not working properly for me. I used brew to install open-mpi, but the same error was showing in the verbose output for installing boost. A run of brew doctor showed that openmpi was not linked properly, so I fixed those errors and reran brew -v install boost --with-mpi --without-single and it finally built and installed all of the libraries without a problem
To anyone that comes across this, the package migrated to boost-python and boost-mpi separate from boost. Use brew install boost-mpi
Just get it worked on OSX 10.11.5. I've tried brew, but with no luck.
Suppose you already have gcc installed. Here are what I've done:
1. Find and disable (but do not remove) clang
clang alway cause headaches. There would be a lot of warnings when building Boost.
which clang, which should give you /usr/bin/clang
Rename it: sudo mv clang clang_mac_remove, also for clang++: sudo mv clang++ clang++_mac_remove. You can change the names back if you need them in future.
2. Install OpenMPI
If you already installed using brew, uninstall first. Becasue it would have used clang as the compiler wrapper by default. You need to change the wrapper to gcc.
Download the package.
Specify the wrapper compiler to gcc and g++:
./configure CC=gcc CXX=g++ F77=ifort FC=ifort --prefix=/usr/local
Below may take a long time.
make all
sudo make install
Reference: https://wiki.helsinki.fi/display/HUGG/Open+MPI+install+on+Mac+OS+X
3. Install Boost MPI
Download the package.
Run ./bootstrap.sh (can open it first and specify the toolset to gcc, otherwise, the default option is darwin for mac).
Add using mpi ; in project-config.jam file. Then ./b2 —with-mpi will only build the mpi library.
Then, all built libraries can be found in the folder ~/Downloads/boost_1_61_0/stage/lib.
Copy or move them to /usr/local/lib or any other commonly used library path.
Reference: http://www.boost.org/doc/libs/1_61_0/doc/html/mpi/getting_started.html
4. Compile with Boost MPI
LIBRARY DIR = -L/usr/local/lib
INCLUDE = -I/usr/local/include/
LINKER = -lboost_mpi -lboost_serialization
e.g.
mpic++ -std=c++11 -I/usr/local/include/ -c boost_test.cpp -L/usr/local/lib -lboost_mpi -lboost_serialization
Good luck!
I am trying to set up a completely recent and up to date gcc environment... basically gcc 4.6.3 and binutils 2.22. Not the fairly old gcc 4.2.1 (llvm) that comes with Xcode 4.2.1. My problem is not with compiling (for now, at least). I can compile binutils, gcc, and both at the same time perfectly. However each time I try to compile I find something like this appear at the end of configure's output
checking where to find the target ar... just compiled
checking where to find the target as... host tool
checking where to find the target cc... just compiled
checking where to find the target c++... just compiled
checking where to find the target c++ for libstdc++... just compiled
checking where to find the target dlltool... just compiled
checking where to find the target gcc... just compiled
checking where to find the target gcj... just compiled
checking where to find the target gfortran... just compiled
checking where to find the target gccgo... host tool
checking where to find the target ld... host tool
checking where to find the target lipo... host tool
checking where to find the target nm... just compiled
checking where to find the target objdump... just compiled
checking where to find the target ranlib... just compiled
checking where to find the target strip... just compiled
checking where to find the target windres... just compiled
checking where to find the target windmc... just compiled
I made a symlinks for ld, opcodes, bfd, gas, gprof, and binutils in the gcc source folder to get this output... (if that helps)
As you can see I can get a "just compiled" version of most software, by I just can't figure out how to compile as and ld. I have read the README file in the ld folder (which in the the binutils package) and it says to execute:
make all-ld
I do and being that I already "made" the source, it tells me nothing is left to do...
The only thing I can think of now is to temporally modify my path to not include ld. This would be pretty easy...
aka a quick check with
whereis ld
anyway immediately after thinking of this I relized it would be stupid being now I would have nothing to link gcc (or even a new version of ld (if I got to work at all))
So, in conclusion, I am trying to compile a new copy of ld and as...
The facts:
I am running Mac OS X Lion.
I am trying to compile gcc 4.6.3 using the default gcc that comes with Mac OS X Lion (gcc 4.2.1).
The configure scripts seem to think I do not want to compile a new linker and assembler.
I have successfully built a working gcc and binutils but without a new linker and assembler.
I have tried to compile binutils and gcc with the --with-ld=yes and --with-gold=no option, they didn't seem to do anything.
I built gmp v5.0.4, mpfr v3.1.0, and mpc v0.9 using gcc v4.2.1, installing them in the same prefix as used in the gcc and binutils configure.
For each configure I am explicitly stating the locations of gmp, mpfr, and mpc.
I built gmp, mpfr, and mpc separately from gcc (doing so, resulted in some build errors).
gcc, binutils, gmp, mpfr, and mpc are all build in a separate directory which is not a subdirectory of the source directory (as recommended by the gcc installation instructions).
I am compiling gcc with the following options:
../gcc-4.6.3/configure --prefix=$PREFIX --with-gmp=$PREFIX --with-mpfr=$PREFIX --with-mpc=$PREFIX --enable-checking=release --program-suffix=-4.6.3 --enable-ld=yes --enable-gold=no
I mainly followed this tutorial.
Because I have no other compiler on my computer I am forced to use the mac os x default llvm compiler (what I described earlier as gcc 4.2.1) which I read here doesn't work, however I don't have gcc-4.2 on my computer.
Compiling gcc-4.7.0 (the very recent snapshot) using the new version of gcc I just compiled (gcc-4.6.3) results in compiler errors (I set the environment variables CC and CXX).
I have tried compiling gcc first, than binutils, and the other way around, eventually I decided building gcc with binutils was the way to go. The output above you was with both compiled at once
I read this little post on gold, and tried adding --enable-gold and --enable-ld. They changed nothing
The order in which I did things:
Compile gmp.
Install it.
Compile mpfr.
Install it.
Compile mpc.
Install it.
Make symlinks to the main gcc src directory for ld, gas, binutils, opcodes, gprof, bfd.
Configure gcc.
Find that "status report."
Think of another way to possibly build ld and as.
Fail at that.
Go on Stack Overflow.
as "fazo" said, I am not really telling anyone much about how I compiled binutils... so here it is...
I tried first compiling gcc, than building binutils using gcc, that compiled fine, but no ld, as, or even gold. Then I tried using the llvm compiler to build binutils then gcc. It also worked fine, but still no ld, as, or gold. Finally, I read here that I could make symlinks to some items in binutils and gcc would detect and build them... I did. and gcc found most of them (as stated before I put symlinks in the gcc directory for ld, opcodes, bfd, gas, gprof, and binutils) My actual problem here is telling the configure script to compile a new assembler and linker. Also, what boggles me, is in the binutils README folder it says that a mere ./configure make make install will build and install all the tools, I am trying to figure out why this is not the case.
The final command comes out to be:
../binutils-2.22/configure --prefix=$PREFIX --with-gmp=$PREFIX --with-mpfr=$PREFIX --with-mpc=$PREFIX --program-suffix=-4.6.3 --enable-ld --with-gnu-as --with-gnu-ld
yet I still get no ld and no gas.
Any help is greatly appreciated! (also if you want some info I haven't supplied, just ask)
thanks... :)
Afaik with standard builds (configure;make), the results are named differently. gas-new and gld-new or so.
Note that OS X doesn't normally use binutils but Apple's own Mach-O linker (cctools). Even if you succeeded, it might not lead to a workable solution
Binutils does not support an assembler or linker on OS X, see the binutils configure script:
case "${target}" in
*-*-chorusos)
;;
aarch64-*-darwin*)
noconfigdirs="$noconfigdirs ld gas gdb gprof"
noconfigdirs="$noconfigdirs sim target-rda"
;;
arm-*-darwin*)
noconfigdirs="$noconfigdirs ld gas gdb gprof"
noconfigdirs="$noconfigdirs sim target-rda"
;;
powerpc-*-darwin*)
noconfigdirs="$noconfigdirs ld gas gdb gprof"
noconfigdirs="$noconfigdirs sim target-rda"
;;
i[3456789]86-*-darwin*)
noconfigdirs="$noconfigdirs ld gprof"
noconfigdirs="$noconfigdirs sim target-rda"
;;
x86_64-*-darwin[912]*)
noconfigdirs="$noconfigdirs ld gas gprof"
noconfigdirs="$noconfigdirs sim target-rda"
;;
*-*-darwin*)
noconfigdirs="$noconfigdirs ld gas gdb gprof"
noconfigdirs="$noconfigdirs sim target-rda"
;;
For any darwin (i.e. OS X) target, this adds ld and gas to the noconfigdirs list. That means those directories will be skipped during the build.
To build GCC on OS X you need to use the native OS X assembler and linker (which are part of the "Xcode Command Line Tools" package, from the App Store).
I realized, that it was basically a mac issue. After installing and getting Ubuntu it built the linker and assembler just fine... still having a bit of trouble compiling gcc, but that is because of Debian's new multi architecture support... but this is another issue... Thanks for your help Marco van de Voort
my custom built gcc 4.6.0, installed in my home directory, on ubuntu 10.04, links the system libstdc++ instead of the custom built one, most of the time (as evidenced by ldd). to be more puzzling, using this newly built gcc, custom compiled openmpi libraries are linked correctly, and this is the only software i have compiled that behaves ok. does anybody have any explanation for this, or a workaround?
thanks
Isn't there an option to statically link the libstdc into the gcc when you configure it? --disable-shared if I understand how it works correctly. Worst case make another compile of gcc with that switch and see if you run into the issue.
I don't know why this isn't detailed more clearly on the GCC website for end-users. The GCC FAQ clearly states this is a common problem wrt libstdc++. Environment variables are troublesome. Wrapping the linker, nobody knows how to do that. Editing /etc/ld.so.conf isn't an option. Adding -Wl,-rpath everywhere, come on. The easiest solution is the specs file. For a typical 64-bit x86 Linux system, go into your custom gcc installation, in dirname `g++ -print-libgcc-file-name`and then run g++ -dumpspecs > specs. Edit that file, find the *link_command: section. After %(link_libgcc) add -rpath /home/user/bin/gcc-9/lib64 (of course use your own path). Or add the same rpath to end of *link: section. Alternatively, configure gcc with --with-specs='%{!static:%x{-rpath=/home/user/bin/gcc9/lib64} %x{-enable-new-dtags}}' . Enjoy your own C++ compiler that generates binaries that just work.
See also:
GCC specs file: how to get the installation path
Linking g++ 4.8 to libstdc++
How to configure libstdc++ with GCC 4.8?