Makefile：将 C 代码转换为 mex 代码（链接错误）答案

【问题标题】：Makefile: Converting C-code to mex code (Linking error)Makefile：将 C 代码转换为 mex 代码（链接错误）
【发布时间】：2015-04-13 14:56:50
【问题描述】：

我有一个使用 makefile 的C-code which works fine。现在，我正在尝试 convert it to mex 文件，以便我可以从 Matlab 运行它。我也是using makefile approach。但是，mex 的 makefile 给了我错误。

这是我想要与 mex 文件一起编译的 tsnnls lib。

C项目组织：

tsnnls_test_DKU.c
Include_4_TSNNLS.c
Include_4_TSNNLS.h

"Include_4_TSNNLS.*" 文件有函数TestingLibraries() 调用第三部分库；虽然我试图保持“tsnnls_test_DKU.c”非常简单：

原始代码：tsnnls_test_DKU.c

int TestingLibraries() ;

int main( int argc, char* argv[] ) 
{

    int k = TestingLibraries() ;
    return(1);
}

现在，代码已更改：

更改代码：tsnnls_test_DKU.c:

#include "mex.h"
#include <math.h>

int TestingLibraries() ;

void mexFunction (int nlhs, mxArray *plhs[],  int nrhs, const mxArray *prhs[]) {

    int k = TestingLibraries() ;

}

原始制作文件（有效

CXX       = gcc
FLAGS     = 
CFLAGS    = "-m64"   ## "-m32"
INCLUDE_TSNNLS  = -I/home/dkumar/libtsnnls-2.3.3 -I/home/dkumar/libtsnnls-2.3.3/tsnnls  -I/home/dkumar/libtsnnls-2.3.3/tsnnls/taucs_basic

#################### ALL TSNNLS lib related ##########################
## this is where all object file (*.o) generated by tsnnls library makefile are located.
tsnnl_PATH = /home/dkumar/libtsnnls-2.3.3/tsnnls/

# Here is a simple Make Macro.
OBJS_tsnnls1        = tsnnls_test_DKU.o
OBJS_ADD        = Include_4_TSNNLS.o

# Here is a Make Macro that uses the backslash to extend to multiple lines.
OBJS_tsnnls0 =  libtsnnls_la-taucs_malloc.o libtsnnls_la-taucs_ccs_order.o \
    libtsnnls_la-taucs_ccs_ops.o libtsnnls_la-taucs_vec_base.o \
    libtsnnls_la-taucs_complex.o libtsnnls_la-colamd.o \
    libtsnnls_la-amdbar.o libtsnnls_la-amdexa.o \
    libtsnnls_la-amdtru.o libtsnnls_la-genmmd.o \
    libtsnnls_la-taucs_timer.o libtsnnls_la-taucs_sn_llt.o \
    libtsnnls_la-taucs_ccs_base.o libtsnnls_la-tlsqr.o \
    libtsnnls_la-tsnnls.o libtsnnls_la-lsqr.o

 ## adding "$(OBJS_PATH)" to each word in "$(OBJS)"
# which in our case is basically to add the same folder in front of all "*.o" object files.
OBJS_TEMP = $(addprefix $(tsnnl_PATH), $(OBJS_tsnnls0)) 

# OBJS_LOC is in current working directory,
OBJS_tsnnlsALL = $(OBJS_TEMP)   $(OBJS_tsnnls1)

# Libraries for tsnnls
STLIB_tsnnls    = /usr/local/lib/taucs_full/lib/linux/libtaucs.a 
LIBS_tsnnls     = -largtable2 -llapack -lblas -lquadmath -lm

########################################################################
TARGET  = tsnnls_test_DKU
REBUILDABLES = $(OBJS_tsnnls1) $(TARGET) 
LIBS      =    $(LIBS_tsnnls)  $(STLIB_tsnnls)

INCLUDE = $(INCLUDE_TSNNLS)

all : $(TARGET)
    echo All done

clean : 
    rm -f $(REBUILDABLES)   
    echo Clean done

$(OBJS_ADD):    Include_4_TSNNLS.c
    gcc -c -o Include_4_TSNNLS.o Include_4_TSNNLS.c

# Final linking
$(TARGET) : $(OBJS_tsnnlsALL)  $(OBJS_ADD) $(LIBS)
    $(CXX) -g -o $@ $(INCLUDE) $(CFLAGS) $^

更改的 Makefile 无法运行 请注意，我已使用 gcc 编译器将链接部分恢复为原始部分；但是，我想在这里使用 mex。

我根据自己的理解更改了一些行，并使用了标签“# mex”。

此外，Include_4_TSNNLS.o 正在成功创建。此外，之前的“include”问题已通过硬连线所有路径得到解决。

  MEXSUFFIX  = mexa64                       # mex
  MATLABHOME = /usr/local/MATLAB/R2011b             # mex
  MEX        = /usr/local/MATLAB/R2011b/bin/mex         # mex

  MEXFLAGS  = -cxx CC='$(CXX)' CXX='$(CXX)' LD='$(CXX)'     # mex

CXX       = gcc
FLAGS     = 
CFLAGS    = -fPIC -pthread -DMX_COMPAT_32 -DMATLAB_MEX_FILE -m64   ## "-m32"
INCLUDE_TSNNLS  = -I/usr/local/MATLAB/R2011b/extern/include  -I/home/dkumar/libtsnnls-2.3.3 -I/home/dkumar/libtsnnls-2.3.3/tsnnls  -I/home/dkumar/libtsnnls-2.3.3/tsnnls/taucs_basic

#################### ALL TSNNLS lib related ##########################
## this is where all object file (*.o) generated by tsnnls library makefile are located.
tsnnl_PATH = /home/dkumar/libtsnnls-2.3.3/tsnnls/

# Here is a simple Make Macro.
OBJS_tsnnls1        = tsnnls_test_DKU.o
OBJS_ADD        = Include_4_TSNNLS.o

# Here is a Make Macro that uses the backslash to extend to multiple lines.
OBJS_tsnnls0 =  libtsnnls_la-taucs_malloc.o libtsnnls_la-taucs_ccs_order.o \
    libtsnnls_la-taucs_ccs_ops.o libtsnnls_la-taucs_vec_base.o \
    libtsnnls_la-taucs_complex.o libtsnnls_la-colamd.o \
    libtsnnls_la-amdbar.o libtsnnls_la-amdexa.o \
    libtsnnls_la-amdtru.o libtsnnls_la-genmmd.o \
    libtsnnls_la-taucs_timer.o libtsnnls_la-taucs_sn_llt.o \
    libtsnnls_la-taucs_ccs_base.o libtsnnls_la-tlsqr.o \
    libtsnnls_la-tsnnls.o libtsnnls_la-lsqr.o

 ## adding "$(OBJS_PATH)" to each word in "$(OBJS)"
# which in our case is basically to add the same folder in front of all "*.o" object files.
OBJS_TEMP = $(addprefix $(tsnnl_PATH), $(OBJS_tsnnls0)) 

# OBJS_LOC is in current working directory,
OBJS_tsnnlsALL = $(OBJS_TEMP)   $(OBJS_tsnnls1)

# Libraries for tsnnls
##STLIB_tsnnls  = /usr/local/lib/taucs_full/lib/linux/libtaucs.a 
STLIB_tsnnls    = /usr/local/lib/libtaucs.a 
#LIBS_tsnnls    = -largtable2 -llapack -lblas -lquadmath -lm
LIBS_tsnnls     = -largtable2 -llapack -lblas -lm
########################################################################
## TARGET   = tsnnls_test_DKU

TARGET_WO_EXTN =  tsnnls_test_DKU               # mex
TARGET      = TARGET_WO_EXTN

#TARGET =  $(TARGET_WO_EXTN).$(MEXSUFFIX)           # mex

REBUILDABLES = $(OBJS_tsnnls1) $(TARGET) *.mexa64  
LIBS      =    $(LIBS_tsnnls)  $(STLIB_tsnnls)
INCLUDE = $(INCLUDE_TSNNLS)

all : $(TARGET)
    echo All done

clean : 
    rm -f $(REBUILDABLES)   
    echo Clean done

$(OBJS_ADD):    Include_4_TSNNLS.c
    gcc -c -o $(INCLUDE) Include_4_TSNNLS.o Include_4_TSNNLS.c

# CHANGED FROM HERE     # mex
tsnnls_test_DKU.o: tsnnls_test_DKU.c  Include_4_TSNNLS.c $(OBJS_tsnnlsALL)  $(OBJS_ADD)
    $(CXX) $(CFLAGS) $(INCLUDE)  -c $^

# Final linking
$(TARGET): tsnnls_test_DKU.o     $(OBJS_tsnnlsALL)  $(OBJS_ADD)   $(LIBS) 
    $(MEX) $(MEXFLAGS)  -output $(TARGET_WO_EXTN) $^      -largeArrayDims

我收到以下链接错误：

dkumar@kumar-Aspire-E1-510 ~/CPP_ExampleCodes_DKU/Using_tsnnls_DKU_copy_2_MEX $ make
make: Circular tsnnls_test_DKU.o <- tsnnls_test_DKU.o dependency dropped.
gcc -fPIC -pthread -DMX_COMPAT_32 -DMATLAB_MEX_FILE    -I/usr/local/MATLAB/R2011b/extern/include  -I/home/dkumar/libtsnnls-2.3.3 -I/home/dkumar/libtsnnls-2.3.3/tsnnls  -I/home/dkumar/libtsnnls-2.3.3/tsnnls/taucs_basic -c tsnnls_test_DKU.c Include_4_TSNNLS.c /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_malloc.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_order.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_ops.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_vec_base.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_complex.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-colamd.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdbar.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdexa.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdtru.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-genmmd.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_timer.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_sn_llt.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_base.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-tlsqr.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-tsnnls.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-lsqr.o Include_4_TSNNLS.o
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_malloc.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_order.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_ops.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_vec_base.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_complex.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-colamd.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdbar.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdexa.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdtru.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-genmmd.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_timer.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_sn_llt.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_base.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-tlsqr.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-tsnnls.o: linker input file unused because linking not done
gcc: warning: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-lsqr.o: linker input file unused because linking not done
gcc: warning: Include_4_TSNNLS.o: linker input file unused because linking not done
/usr/local/MATLAB/R2011b/bin/mex             -cxx CC='gcc' CXX='gcc' LD='gcc'         -output tsnnls_test_DKU                tsnnls_test_DKU.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_malloc.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_order.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_ops.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_vec_base.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_complex.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-colamd.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdbar.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdexa.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-amdtru.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-genmmd.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_timer.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_sn_llt.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_base.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-tlsqr.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-tsnnls.o /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-lsqr.o Include_4_TSNNLS.o /usr/lib/libargtable2.so /usr/lib/liblapack.so /usr/lib/libblas.so /usr/lib/x86_64-linux-gnu/libm.so /usr/local/lib/libtaucs.a      -largeArrayDims

Warning: You are using gcc version "4.8.2-19ubuntu1)".  The version
         currently supported with MEX is "4.3.4".
         For a list of currently supported compilers see: 
         http://www.mathworks.com/support/compilers/current_release/

/usr/bin/ld: /home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_order.o: relocation R_X86_64_32 against `.rodata' can not be used when making a shared object; recompile with -fPIC
/home/dkumar/libtsnnls-2.3.3/tsnnls/libtsnnls_la-taucs_ccs_order.o: error adding symbols: Bad value
collect2: error: ld returned 1 exit status

    mex: link of ' "tsnnls_test_DKU.mexa64"' failed.

make: *** [tsnnls_test_DKU] Error 1

有人可以帮我提供链接选项吗？

【问题讨论】：

使用 -I(matlab_dir)/extern/include 将 makefile 指向标头
@user3528438 我认为我已经使用了 include : INCLUDE = -I$(MATLABHOME)/extern/include $(INCLUDE_TSNNLS) 然后在制作 TARGET 时包含它。但是，我可能在编译器选项中传递了错误，无法弄清楚。
在代码的底部，您使用“gcc -c -o...”进行编译并使用“$(MEX) -g -o ....”进行链接。由于您的 $(INCLUDE) 在 $(MEX) 之后，因此它被视为链接器选项而不是编译器选项。尝试将 $(INCLUDE) 和 $(CFLAGS) 移近 gcc。
@user3528438 我试过关注；但是，仍然得到同样的错误：$(OBJS_ADD): Include_4_TSNNLS.c gcc -c -o Include_4_TSNNLS.o Include_4_TSNNLS.c $(CFLAGS) $(INCLUDE) 然后$(TARGET) : $(OBJS_tsnnlsALL) $(OBJS_ADD) $(LIBS) $(MEX) -g -o $@ $^
我对 mex 不太了解。但是，当然在规则 $(OBJS_ADD): Include_4_TSNNLS.c 中，您已经包含了 mex 包含库。您包括在链接阶段。有人可能会给你精确的命令。

标签： c matlab makefile mex

【解决方案1】：

问题在于 $(OBJS_tsnnls0) 中列出的所有目标文件（或者至少是 libtsnnls_la-taucs_ccs_order.o）都在没有 -fPIC gcc 编译器选项的情况下进行了编译。 gcc 手册页说 -fPIC

生成与位置无关的代码 (PIC) 适用于共享库

请注意，MEX 文件是共享库。因此，所有目标代码都完美地链接到一个独立的可执行文件（不需要 PIC），但它只是没有被编译为链接到 MEX 文件（或任何共享库）。如果你有原始源文件，你应该能够通过运行

使用正确的开关重新编译它们

$(MEX) -c -o 文件名.o 文件名.c

对于每个源文件。

【讨论】：

该库似乎使用了一些自动工具（libtool）。在 $(OBJS_tsnnls0) 中 configure 之后使用 makefile 时，我可以看到如下命令：libtool: compile: gcc -DHAVE_CONFIG_H -I. -I.. -Wall -MT libtsnnls_la-taucs_ccs_order.lo -MD -MP -MF .deps/libtsnnls_lataucs_ccs_order.Tpo -c taucs_basic/taucs_ccs_order.c -fPIC -DPIC -o .libs/libtsnnls_la-taucs_ccs_order.o 。它似乎已经使用-fPIC。我应该将此命令更改为什么？
是的，我有下载的库和所有源代码：jasoncantarella.com/wordpress/software/tsnnls。它的 makefile 正在使用makefile.am 创建。但是，鉴于我不擅长 C，我无法弄清楚 makefile.am->makefile 和 change the flag or compiler 的方式。你能自己试试这个库并帮我解决这个问题吗？为此，我真的会为你欠下我的生命。我很乐意特别thank you in my paper or put you as my co-author。这将解决医学成像中的一个突出问题；因此，非常重要

【解决方案2】：

您对tsnnls_test_DKU.o 具有循环依赖关系，因为OBJS_tsnnlsAll 扩展为包含OBJS_tsnnls1，它在Makefile 的第一行中定义为tsnnls_test_DKU.o。

您也没有构建OBJS_tsnnlsAll 的规则。您只是将所有目标文件的名称传递给编译器，由于 -c 标志而忽略它们。

您应该从tsnnls_test_DKU.o 规则中删除$(OBJS_tsnnlsALL)，并从最终目标中删除tsnnls_test_DKU.o，因此最后两条规则如下所示：

# CHANGED FROM HERE     # mex
tsnnls_test_DKU.o: tsnnls_test_DKU.c  Include_4_TSNNLS.c $(OBJS_ADD)
    $(CXX) $(CFLAGS) $(INCLUDE)  -c $^

# Final linking
$(TARGET): $(OBJS_tsnnlsALL)  $(OBJS_ADD)   $(LIBS) 
    $(MEX) $(MEXFLAGS)  -output $(TARGET_WO_EXTN) $^      -largeArrayDims

【讨论】：