Cannot import cv2 in Python tests despite following knowledge base

[jwodder]

Conda-forge documentation

• I could not solve my problem using the conda-forge documentation.

Installed packages

N/A

Environment info

Conda forge feedstock CI

Issue

(I'm filing this issue here largely because I don't know of a better place to file it; feel free to let me know of one.)

I'm a maintainer of https://github.com/conda-forge/dandi-feedstock, which uses opencv in some of its tests, and the recipe/ folder contains a yum_requirements.txt file for the opencv deps as described at https://conda-forge.org/docs/maintainer/knowledge_base/#libgl. However, ever since the feedstock's v0.61.0 PR on March 19 (the previous update PR, which was successful, being on February 29), the opencv-using tests have been failing with "ImportError: libEGL.so.1: cannot open shared object file: No such file or directory".

I asked about this on Gitter on March 21, but the only advice I got was to add mesa-libEGL to yum_requirements.txt; at the time, I failed to rerender after applying this change, and so I didn't get anywhere. Re-attemping it now, when I add mesa-libEGL and rerender, this just results in "libEGL.so.1" in the error message being replaced with "libOpenGL.so.0". I tried blindly adding mesa-libOpenGL to yum_requirements.txt and rerendering, but that didn't help, and now I don't know what to do.

[traversaro]

See conda-forge/opencv-feedstock#401 for a related issue.

[jwodder]

Based on that link, I was able to get the tests passing by adding libopencv *=headless* # [linux] to test.requires.

Are there any plans to add any of this information to the Knowledge Base?

[hmaarrfk]

It generally takes a little time for this information to "trickle up" into the knowledge base.

I've been trying to monitor if people have been getting issues or not, and addressing things as they come up.

qt6 is still "new" and many projects still aren't compatible, so I don't want to suggest too broadly (yet).

But it my now be the time to add this documentation.

PRs welcome if you want to summarizing your learnings in the documentation!