Idyntree: Build devel branch fails on MacOS when USES_MATLAB is ON

[lrapetti]

At first it failed with the same error experienced in https://github.com/robotology/idyntree/issues/630, fixed by recompiling a clean build.

However, after turning IDYNTREE_USES_MATLAB ON, it was failing with the following message:

[ 99%] Building CXX object bindings/matlab/CMakeFiles/iDynTreeMEX.dir/autogenerated/iDynTreeMATLAB_wrap.cxx.o
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:4549:30: error: non-constant-expression cannot be narrowed from type 'int' to 'mwSize' (aka 'unsigned long') in initializer list [-Wc++11-narrowing]
        const mwSize dims[]={nMatrices,nRows,nCols};
                             ^~~~~~~~~
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:4549:30: note: insert an explicit cast to silence this issue
        const mwSize dims[]={nMatrices,nRows,nCols};
                             ^~~~~~~~~
                             static_cast<mwSize>( )
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:4549:40: error: non-constant-expression cannot be narrowed from type 'int' to 'mwSize' (aka 'unsigned long') in initializer list [-Wc++11-narrowing]
        const mwSize dims[]={nMatrices,nRows,nCols};
                                       ^~~~~
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:4549:40: note: insert an explicit cast to silence this issue
        const mwSize dims[]={nMatrices,nRows,nCols};
                                       ^~~~~
                                       static_cast<mwSize>( )
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:4549:46: error: non-constant-expression cannot be narrowed from type 'int' to 'mwSize' (aka 'unsigned long') in initializer list [-Wc++11-narrowing]
        const mwSize dims[]={nMatrices,nRows,nCols};
                                             ^~~~~
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:4549:46: note: insert an explicit cast to silence this issue
        const mwSize dims[]={nMatrices,nRows,nCols};
                                             ^~~~~
                                             static_cast<mwSize>( )
/Users/lorenzorapetti/Software/robotology-superbuild/robotology/iDynTree/bindings/matlab/autogenerated/iDynTreeMATLAB_wrap.cxx:88975:5: warning: delete called on 'iDynTree::DiscreteExtendedKalmanFilterHelper' that is abstract but has non-virtual destructor [-Wdelete-non-virtual-dtor]
    delete arg1;
    ^
1 warning and 3 errors generated.
make[2]: *** [bindings/matlab/CMakeFiles/iDynTreeMEX.dir/autogenerated/iDynTreeMATLAB_wrap.cxx.o] Error 1
make[1]: *** [bindings/matlab/CMakeFiles/iDynTreeMEX.dir/all] Error 2
make: *** [all] Error 2

[lrapetti]

CC @claudia-lat

[claudia-lat]

I just tried to build it and it failed in the very same way.

[lrapetti]

The problem was introduced in https://github.com/robotology/idyntree/pull/668

[traversaro]

So, there are two main problems here:

• The actual bug in https://github.com/robotology/idyntree/blob/devel/bindings/matlab/matlab_mat4x4vec.i#L10 , where the arguments nMatrices, nRows and nCols should be declared as mwSize, and the bindings regenerated. @fjandrad as I guess you already have the bindings machinery ready could you provide a PR for that? Otherwise if someone else wants to tackle it, you need to follow the instructions in https://github.com/robotology/idyntree#matlaboctave-bindings-modifications to get the proper SWIG for matlab generation.
• The fact that this was not catched by the CI. Probably to fix this we just need to add brew install octave and pass -DIDYNTREE_USES_OCTAVE:BOOL=ON to the macOS build.

[traversaro]

Let me know if no one is able to handle the fix discussed in the previous comment, then I can handle it. @fjandrad @lrapetti @claudia-lat

[lrapetti]

Let me know if no one is able to handle the fix discussed in the previous comment, then I can handle it. @fjandrad @lrapetti @claudia-lat

I can take care of regenerating the bindings

[traversaro]

Great, thanks! Remember to separate the commit that modifies the .i file and the one that add the generated bindings (as documented in https://github.com/robotology/idyntree/blob/master/doc/dev/faqs.md#how-to-add-wrap-a-new-class-or-function-with-swig). Instead you don't need to update the Changelog, as this fix is fixing a bug that was never included in a released version of iDynTree.

[lrapetti]

PR opened at https://github.com/robotology/idyntree/pull/676

[claudia-lat]

The PR has been merged. Now it works! Thanks @lrapetti @traversaro

[lrapetti]

• The fact that this was not catched by the CI. Probably to fix this we just need to add brew install octave and pass -DIDYNTREE_USES_OCTAVE:BOOL=ON to the macOS build.

I have opened a PR also for adding the test in https://github.com/robotology/idyntree/pull/677

[traversaro]

• The fact that this was not catched by the CI. Probably to fix this we just need to add brew install octave and pass -DIDYNTREE_USES_OCTAVE:BOOL=ON to the macOS build.

I have opened a PR also for adding the test in #677

Thanks!

[lrapetti]

Both #676 and #677 have been merged. We can close this issue.

[traversaro]

Thanks @lrapetti !

[fjandrad]

Sorry for the late response I was without a computer for two days. I'm glad is fixed. Why does it only affect MACOS?

[traversaro]

I guess it is something related to the default settings of Clang (used in macOS) that are different from the one in GCC (used in Linux).