tuxable-ltd / meta-scipy

OpenEmbedded layer for scipy
MIT License
5 stars 10 forks source link

_ufuncs.cpython-310-aarch64-linux-gnu.so: undefined symbol: log_ndtr #12

Open yrik opened 1 year ago

yrik commented 1 year ago

I've built an image with meta-scipy and meta-rockchip together and somehow getting this error:

[ahrs_debug-7]     from scipy.spatial.transform import Rotation as R
[ahrs_debug-7]   File "/usr/lib/python3.10/site-packages/scipy/spatial/__init__.py", line 108, in <module>
[ahrs_debug-7]     from ._geometric_slerp import geometric_slerp
[ahrs_debug-7]   File "/usr/lib/python3.10/site-packages/scipy/spatial/_geometric_slerp.py", line 9, in <module>
[ahrs_debug-7]     from scipy.spatial.distance import euclidean
[ahrs_debug-7]   File "/usr/lib/python3.10/site-packages/scipy/spatial/distance.py", line 123, in <module>
[ahrs_debug-7]     from ..special import rel_entr
[ahrs_debug-7]   File "/usr/lib/python3.10/site-packages/scipy/special/__init__.py", line 649, in <module>
[ahrs_debug-7]     from . import _ufuncs
[ahrs_debug-7] ImportError: /usr/lib/python3.10/site-packages/scipy/special/_ufuncs.cpython-310-aarch64-linux-gnu.so: undefined symbol: log_ndtr
yrik commented 1 year ago
root@rockchip-rk3588-rock-5b:~# python -m pip install scipy --force-reinstall
Collecting scipy
  Downloading scipy-1.11.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (32.8 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 32.8/32.8 MB 4.9 MB/s eta 0:00:00
Collecting numpy<1.28.0,>=1.21.6
  Downloading numpy-1.25.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl (14.2 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 14.2/14.2 MB 6.5 MB/s eta 0:00:00
Installing collected packages: numpy, scipy
  Attempting uninstall: numpy
    Found existing installation: numpy 1.22.3
    Uninstalling numpy-1.22.3:
      Successfully uninstalled numpy-1.22.3
  Attempting uninstall: scipy
    Found existing installation: scipy 1.8.1
    Uninstalling scipy-1.8.1:
      Successfully uninstalled scipy-1.8.1
Successfully installed numpy-1.25.2 scipy-1.11.2

works fine after such and upgrade, but original yocto build seems to be broken in this combination of layers..