Failed to build psbody-mesh

[927087638]

encountered a build failure while compiling the psbody-mesh........

• error message:

    /tmp/pip-build-env-6o0hvk5i/normal/lib/python3.10/site-packages/numpy/_core/include/numpy/ndarraytypes.h:1520:35: note:   initializing argument 1 of ‘npy_intp* PyArray_DIMS(const PyArrayObject*)’
     1520 | PyArray_DIMS(const PyArrayObject *arr)
          |              ~~~~~~~~~~~~~~~~~~~~~^~~
    mesh/src/aabb_normals.cpp:119:37: error: cannot convert ‘PyObject*’ {aka ‘_object*’} to ‘const PyArrayObject*’ {aka ‘const tagPyArrayObject_fields*’}
      119 |     npy_intp* n_dims = PyArray_DIMS(py_n);
          |                                     ^~~~
          |                                     |
          |                                     PyObject* {aka _object*}
          
  ...........
    error: command '/usr/bin/gcc' failed with exit code 1
    error: subprocess-exited-with-error
    
    × Building wheel for psbody-mesh (pyproject.toml) did not run successfully.
    │ exit code: 1
    ╰─> See above for output.
  

  Seems like, the numpy api is not compatible with Mesh……..

• the Cmake cmd

  BOOST_INCLUDE_DIRS=/usr/include/boost make all

• env:

  • Ubuntu 22.04.4 LTS
  • gcc 11.4.0
  • g++ 11.4.0
  • Python 3.10.14
  • numpy 1.23.5

• already have

  • removed ./mesh/requirements.txt
  • apt-get install libboost-dev
  • modified ./mesh/Makefile

---

Emotional damaged, seeking for help........

[YuDeng]

Hi, it seems that the python version is too high for the psbody-mesh library. Can you try to downgrade python to 3.9 or 3.8 and see if it works?

Alternatively, since psbody-mesh is only used for data processing and does not affect inference, you can simply skip this part if the problem still exists.

[Edisoneh]

Cool! When I downgrade python to 3.8, it fixed! Thanks!

Hi, it seems that the python version is too high for the psbody-mesh library. Can you try to downgrade python to 3.9 or 3.8 and see if it works?

Alternatively, since psbody-mesh is only used for data processing and does not affect inference, you can simply skip this part if the problem still exists.