VL_feat Matlab GCC can't compile

Question

I was trying to compile the function c_eoverlap in matlab following the instruction here:

http://www.robots.ox.ac.uk/~vgg/research/affine/evaluation.html#eval_soft

Of course, it wouldn't work out... My Matlab version is 2018a which is pretty new and the gcc -v outputs the following

Also attach this long error log, as I am pretty new to C++, please let me know if there is any other information needed, appreciate for your help!

>> mex repeatability/c_eoverlap.cxx
Building with 'Xcode Clang++'.
Error using mex
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:34:14: error: cannot initialize
a variable of type 'const int *' with an rvalue of type 'const mwSize *' (aka 'const unsigned
long *')
  int const *dims1 = mxGetDimensions(prhs[0]);
             ^       ~~~~~~~~~~~~~~~~~~~~~~~~
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:36:14: error: cannot initialize
a variable of type 'const int *' with an rvalue of type 'const mwSize *' (aka 'const unsigned
long *')
  int const *dims2 = mxGetDimensions(prhs[1]);
             ^       ~~~~~~~~~~~~~~~~~~~~~~~~
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:38:14: error: cannot initialize
a variable of type 'const int *' with an rvalue of type 'const mwSize *' (aka 'const unsigned
long *')
  int const *dims3 = mxGetDimensions(prhs[2]);
             ^       ~~~~~~~~~~~~~~~~~~~~~~~~
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:46:13: error: no matching
function for call to 'mxCreateNumericArray_730'
  plhs[0] = mxCreateNumericArray(2, odim, mxDOUBLE_CLASS, mxREAL);
            ^~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^~~~~~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:1256:1: note: candidate function not
viable: no known conversion from 'int *' to 'const mwSize *' (aka 'const unsigned long *')
for 2nd argument
mxCreateNumericArray(mwSize ndim, const mwSize *dims, mxClassID classid, mxComplexity flag);
^
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:47:13: error: no matching
function for call to 'mxCreateNumericArray_730'
  plhs[1] = mxCreateNumericArray(2, odim, mxDOUBLE_CLASS, mxREAL);
            ^~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^~~~~~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:1256:1: note: candidate function not
viable: no known conversion from 'int *' to 'const mwSize *' (aka 'const unsigned long *')
for 2nd argument
mxCreateNumericArray(mwSize ndim, const mwSize *dims, mxClassID classid, mxComplexity flag);
^
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:48:13: error: no matching
function for call to 'mxCreateNumericArray_730'
  plhs[2] = mxCreateNumericArray(2, odim, mxDOUBLE_CLASS, mxREAL);
            ^~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^~~~~~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:1256:1: note: candidate function not
viable: no known conversion from 'int *' to 'const mwSize *' (aka 'const unsigned long *')
for 2nd argument
mxCreateNumericArray(mwSize ndim, const mwSize *dims, mxClassID classid, mxComplexity flag);
^
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:49:13: error: no matching
function for call to 'mxCreateNumericArray_730'
  plhs[3] = mxCreateNumericArray(2, odim, mxDOUBLE_CLASS, mxREAL);
            ^~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^~~~~~~~~~~~~~~~~~~~~~~~
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:1256:1: note: candidate function not
viable: no known conversion from 'int *' to 'const mwSize *' (aka 'const unsigned long *')
for 2nd argument
mxCreateNumericArray(mwSize ndim, const mwSize *dims, mxClassID classid, mxComplexity flag);
^
/Applications/MATLAB_R2018a.app/extern/include/matrix.h:282:30: note: expanded from macro
'mxCreateNumericArray'
#define mxCreateNumericArray mxCreateNumericArray_730
                             ^
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:198:3: warning: 'delete[]'
applied to a pointer that was allocated with 'new'; did you mean 'delete'?
[-Wmismatched-new-delete]
  delete []tdesc_out;
  ^     ~~~

/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:62:22: note: allocated with
'new' here
  float *tdesc_out = new float[dims2[1]*dims1[1]];
                     ^
/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:199:3: warning: 'delete[]'
applied to a pointer that was allocated with 'new'; did you mean 'delete'?
[-Wmismatched-new-delete]
  delete []tover_out;
  ^     ~~~

/Users/lucius/Desktop/ua/805/a1/repeatability/c_eoverlap.cxx:63:22: note: allocated with
'new' here
  float *tover_out = new float[dims2[1]*dims1[1]];
                     ^
2 warnings and 7 errors generated.

Answer 1

You need to add -compatibleArrayDims to your mex command:

mex -compatibleArrayDims c_eoverlap.cxx 

The code is written for older MATLABs that used 32-bit integers (int) for array sizes. Nowadays these are 64-bits (mwSize). This option forces the interface to work like it did in the older MATLABs. See the mex documentation for details.

There are some scary warnings about using delete[] instead of delete. It's just warnings, so it will compile, but you don't know if the code will do what it is supposed to.