Trilinos - c++: error trying to exec 'cc1plus': execvp: No such file or directory

Question

I want to write a simple program which uses Trilinos. After many problems with configuration script I've managed to write it, so it launches without any problem. The problem is, when I try to compile my own code ( both Makefile and my test program I've rewritten according to the original examples ), it throws the error:

c++: error trying to exec '/usr/lib/gcc/x86_64-linux-gnu/4.9/cc1plus': execv: Argument list too long

So I tried to launch my Makefile with env -i make to make enough space for "the arguments". It worked, but now it throws another error:

c++: error trying to exec 'cc1plus': execvp: No such file or directory
Makefile:62: recipe for target 'test.o' failed
make: *** [test.o] Error 1

---

My attempts:

I've tried to find any solution to this problem, but I've found nothing which worked so far.

I've checked if my GCC and G++ are the same version and they are:

gcc (Ubuntu 4.9.1-16ubuntu6) 4.9.1
Copyright (C) 2014 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

g++ (Ubuntu 4.9.1-16ubuntu6) 4.9.1
Copyright (C) 2014 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Then I've tried to reinstall G++:

sudo aptitude reinstall g++-4.9

Still, no success.

Finally I've checked, if my g++ command points to the correct destination:

martin@martin-Aspire-E1-531:~$ which g++
/usr/bin/g++

/usr/bin/g++ is a link to the file g++-4.9, which is a correct file to launch the compiler.

So, I don't have any idea what should I do know to compile it correctly. Do you?

---

My files

.do-configure(file for configuring Trilinos)

#!/bin/bash

# Set this to the root of your Trilinos source directory.
TRILINOS_SOURCE_PATH=~/Stazene/trilinos-11.12.1-Source

#
# You can invoke this shell script with additional command-line
# arguments.  They will be passed directly to CMake.
#
EXTRA_ARGS=$@

#
# Each invocation of CMake caches the values of build options in a
# CMakeCache.txt file.  If you run CMake again without deleting the
# CMakeCache.txt file, CMake won't notice any build options that have
# changed, because it found their original values in the cache file.
# Deleting the CMakeCache.txt file before invoking CMake will insure
# that CMake learns about any build options you may have changed.
# Experience will teach you when you may omit this step.
#
rm -f CMakeCache.txt

#
# Enable all primary stable Trilinos packages.
#
cmake \
  -D CMAKE_INSTALL_PREFIX:FILEPATH="~/.trilinos" \
  -D CMAKE_BUILD_TYPE:STRING=DEBUG \
  -D CMAKE_C_COMPILER=/usr/bin/mpicc \
  -D CMAKE_CXX_COMPILER=/usr/bin/mpicxx \
  -D CMAKE_Fortran_COMPILER=/usr/bin/mpif90 \
  -D TPL_ENABLE_MPI:BOOL=ON \
  -D MPI_BASE_DIR:FILEPATH="/usr/include/mpi" \
  -D MPI_EXEC:FILEPATH="/usr/include/mpiexec" \
  -D Boost_INCLUDE_DIRS="/usr/include/boost" \
  -D Trilinos_ENABLE_Fortran:BOOL=OFF \
  -D Trilinos_ENABLE_SECONDARY_TESTED_CODE=ON \
  -D Trilinos_ENABLE_TESTS:BOOL=ON \
  -D Trilinos_ENABLE_Epetra=ON \
  -D Trilinos_ENABLE_ALL_PACKAGES:BOOL=ON \
$EXTRA_ARGS \
$TRILINOS_SOURCE_PATH

Makefile

# CMAKE File for "MyApp" application building against an installed Trilinos

#This file is an adaptation of the CMakeLists.txt file that was converted from 
#the buildAgainstTrilinos example. This Makefile was designed to be used in a
#flat directory structure. If you would like to run this example you will need
#put this file and src_file.cpp, src_file.hpp, main_file.cpp from
#buildAgainstTrilinos into a new directory. You will then need to set the
#environment variable MYAPP_TRILINOS_DIR to point to your base installation of
#Trilinos. Note that this example assumes that the installation of Trilinos that
#you point to has Epetra enabled.

# Get Trilinos as one entity
include ~/.trilinos/Makefile.export.Trilinos

# Make sure to use same compilers and flags as Trilinos
CXX_WITH_ECHO=$(Trilinos_CXX_COMPILER)
CXX_WITHOUT_ECHO=@$(Trilinos_CXX_COMPILER)

CXX=$(CXX_WITHOUT_ECHO)
CC=$(Trilinos_C_COMPILER)
FORT=$(Trilinos_Fortran_COMPILER)

CXX_FLAGS=$(Trilinos_CXX_COMPILER_FLAGS) $(USER_CXX_FLAGS)
C_FLAGS=$(Trilinos_C_COMPILER_FLAGS) $(USERC_FLAGS)
FORT_FLAGS=$(Trilinos_Fortran_COMPILER_FLAGS) $(USER_FORT_FLAGS)

INCLUDE_DIRS=$(Trilinos_INCLUDE_DIRS) $(Trilinos_TPL_INCLUDE_DIRS)
LIBRARY_DIRS=$(Trilinos_LIBRARY_DIRS) $(Trilinos_TPL_LIBRARY_DIRS)
LIBRARIES=$(Trilinos_LIBRARIES) $(Trilinos_TPL_LIBRARIES)

LINK_FLAGS=$(Trilinos_EXTRA_LD_FLAGS)

#just assuming that epetra is turned on.
DEFINES=-DMYAPP_EPETRA


default: build_all

# Echo trilinos build info just for fun
info:
    @echo "\nFound Trilinos!  Here are the details: "
    @echo "   Trilinos_VERSION = $(Trilinos_VERSION)"
    @echo "   Trilinos_PACKAGE_LIST = $(Trilinos_PACKAGE_LIST)"
    @echo "   Trilinos_LIBRARIES = $(Trilinos_LIBRARIES)"
    @echo "   Trilinos_INCLUDE_DIRS = $(Trilinos_INCLUDE_DIRS)"
    @echo "   Trilinos_LIBRARY_DIRS = $(Trilinos_LIBRARY_DIRS)"
    @echo "   Trilinos_TPL_LIST = $(Trilinos_TPL_LIST)"
    @echo "   Trilinos_TPL_INCLUDE_DIRS = $(Trilinos_TPL_INCLUDE_DIRS)"
    @echo "   Trilinos_TPL_LIBRARIES = $(Trilinos_TPL_LIBRARIES)"
    @echo "   Trilinos_TPL_LIBRARY_DIRS = $(Trilinos_TPL_LIBRARY_DIRS)"
    @echo "   Trilinos_BUILD_SHARED_LIBS = $(Trilinos_BUILD_SHARED_LIBS)"
    @echo "End of Trilinos details\n"

build_all: test.out
    @echo "CXX_FLAGS: $(CXX_FLAGS)"

# build the 
test.out: test.o
    $(CXX) $(CXX_FLAGS) test.cpp -o test.out $(LINK_FLAGS) $(INCLUDE_DIRS) $(DEFINES) $(LIBRARY_DIRS) $(LIBRARIES)

test.o:
    $(CXX) -c $(CXX_FLAGS) $(INCLUDE_DIRS) $(DEFINES) test.cpp

.PHONY: clean
clean:
    rm -f *.o *.a *.exe *.out

.CPP file

//
// This example includes conditional MPI initialization, getting an
// Epetra communicator wrapper, and printing out Epetra version
// information.
//

// This defines useful macros like HAVE_MPI, which is defined if and
// only if Epetra was built with MPI enabled.
#include <Epetra_config.h>

#ifdef HAVE_MPI
    // Your code is an existing MPI code, so it presumably includes mpi.h directly.
    #  include <mpi.h>
    // Epetra's wrapper for MPI_Comm.  This header file only exists if
    // Epetra was built with MPI enabled.
    #  include <Epetra_MpiComm.h>
#else
    #  include <Epetra_SerialComm.h>
#endif // HAVE_MPI

#include <Epetra_Version.h>


//
// ... Your other include files go here ...
//

// Do something with the given communicator.  In this case, we just
// print Epetra's version to the given output stream, on Process 0.
void exampleRoutine (const Epetra_Comm& comm, std::ostream& out) {
    if (comm.MyPID () == 0) {
    // On (MPI) Process 0, print out the Epetra software version.
      out << Epetra_Version () << std::endl << std::endl;
    }
}

int main (int argc, char *argv[]) {
    // These "using" declarations make the code more concise, in that
    // you don't have to write the namespace along with the class or
    // object name.  This is especially helpful with commonly used
    // things like std::endl.
    using std::cout;
    using std::endl;

    #ifdef HAVE_MPI
        // Start up MPI, if using MPI.  Trilinos doesn't have to be built
        // with MPI; it's called a "serial" build if you build without MPI.
        //
        // It's bad form to ignore the error codes returned by MPI
        // functions, but we do so here for brevity.
        (void) MPI_Init (&argc, &argv);
        cout << "MPI ok" << endl;

        // Wrap MPI_COMM_WORLD in an Epetra communicator wrapper.
        // Epetra_MpiComm is a subclass of Epetra_Comm, so you may use it
        // wherever an Epetra_Comm is required.
        Epetra_MpiComm comm (MPI_COMM_WORLD);
    #else
        // Make a "serial" (non-MPI) communicator.  It doesn't actually
        // "communicate," because it only has one process, whose rank is
        // always 0.  Epetra_SerialComm is a subclass of Epetra_Comm, so you
        // may use it wherever an Epetra_Comm is required.
        Epetra_SerialComm comm;
        cout << "Serial" << endl;
    #endif

    // Epetra_Comm has methods that wrap basic MPI functionality.
    // MyPID() is equivalent to MPI_Comm_rank, and NumProc() to
    // MPI_Comm_size.
    //
    // With a "serial" communicator, the rank is always 0, and the
    // number of processes is always 1.
    const int myRank = comm.MyPID ();
    const int numProcs = comm.NumProc ();

    cout << "My PID is:" << myRank << endl;

    if (myRank == 0) {
        cout << "Total number of processes: " << numProcs << endl;
    }

    // Do something with the new Epetra communicator.
    exampleRoutine (comm, cout);

    // This tells the Trilinos test framework that the test passed.
    if (comm.MyPID () == 0) {
        cout << "End Result: TEST PASSED" << endl;
    }

    #ifdef HAVE_MPI
        // Since you called MPI_Init, you are responsible for calling
        // MPI_Finalize after you are done using MPI.
        (void) MPI_Finalize ();
    #endif // HAVE_MPI

    return 0;
}

Answer 1

Finally I've found out that the problem was env -i command - it removed even the g++ path, so the makefile must be launched without it.

Answer 2

Is GCC_EXEC_PREFIX set in your environment? Check out https://stackoverflow.com/a/15262886 where there was a very similar issue, resolved by not setting GCC_EXEC_PREFIX in the environment.