Numpy - Failed compile on 32bit Buster

General Discussion
1

SBC: Odroid xu4
OS: Armbian 21.08.6 Buster
Python: v3.7.3 (OS and Klipper venv)

numpy failed during the first two compilations and had to install a couple packages:

sudo apt-get install gfortran libopenblas-base libopenblas-dev

On to the 3rd time to compile numpy, I decided not to use the cached version, so here is the full download/compile as is. Looking for some direction on how to get past the compile issue.

Issue 1:

  numpy/core/src/multiarray/_multiarray_tests.c.src: In function ‘array_solve_diophantine’:
  numpy/core/src/multiarray/_multiarray_tests.c.src:1353:29: warning: comparison of integer expressions of different signedness: ‘Py_ssize_t’ {aka ‘int’} and ‘unsigned int’ [-Wsign-compare]
       if (PyTuple_GET_SIZE(U) != nterms) {
                               ^~

Issue 2 (may be because of 1):

  creating build/temp.linux-armv7l-3.7/build/src.linux-armv7l-3.7/numpy/core/src/common
  during RTL pass: expand
  numpy/core/src/umath/loops_trigonometric.dispatch.c.src: In function ‘FLOAT_sin_NEON_VFPV4’:
  numpy/core/src/umath/loops_trigonometric.dispatch.c.src:202:20: internal compiler error: in convert_move, at expr.c:218
   NPY_NO_EXPORT void NPY_CPU_DISPATCH_CURFX(FLOAT_@func@)
                      ^
2

I was able to verify that numpy v1.21.5 was being installed by pip, and that version is supported under python 3.7.

Installing a numpy v.1.21.4 is yielding worse results.

3

After more version attempts, here is what worked:

First, make sure these packages are installed:
sudo apt-get install gfortran libopenblas-base libopenblas-dev

Then install numpy v1.19.5:
~/klippy-env/bin/pip3 install -v --no-cache-dir numpy==1.19.5

Compiling takes awhile, but let it go. There was some odd things on the console, but ultimately, it finished:

Successfully built numpy
Installing collected packages: numpy
  changing mode of /home/pi/klippy-env/bin/f2py to 755
  changing mode of /home/pi/klippy-env/bin/f2py3 to 755
  changing mode of /home/pi/klippy-env/bin/f2py3.7 to 755
Successfully installed numpy-1.19.5