Created attachment 220008 [details]
Patch to update math/py-pandas to 1.1.4
The py-pandas port has been held back to 0.2.4 for a long time to maintain Python 2.x compatibility.
Support for Python 2.x ended at the beginning of 2020, so this is no longer necessary, but the maintainer for this port has retired.
Here is a candidate patch to update to 1.1.4, which supports Python 3.9.
Created attachment 220009 [details]
Original version of patch missed new file needed for clang compatibility
Original version of patch missed a new file needed for clang compatibility; they compile with -Werror on clang, but have unreachable code from Cython. They use -Wno-error=unreachable-code on Mac to deal with this, so the patch to setup.py just adds the same flag.
Created attachment 220024 [details]
Additional fix for 3.8
This updated patch incorporates an additional flag needed to keep deprecation warnings in 3.8 from causing the build to fail.
Many thanks for the patch.
There two issues with your patch (test with 'portlint -AC'):
- Don't remove PORTEPOCH, once it exists 
- Your files/patch-setup.py is not generated by 'make makepatch'. For a proper patch format it is recommended to use 'make makepatch' 
 Section 220.127.116.11 from Porters Handbook https://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/book.html
 Section 4.4 from Porters Handbook
The hint in pkg-message about two additional ports is only visible when the port is installed.
Wouldn't it be useful from the user's point of view to include these two ports math/py-statsmodels and devel/py-xarray as RUN_DEPENDS in the Makefile? I would expect a much better user experience.
Created attachment 220077 [details]
Same patch, but restore PORTEPOCH
Here is the patch with PORTEPOCH restored.
I only use poudriere so I don't know how to do "make makepatch" in that context. This patch seems to apply and work, but feel free to modify it as needed.
With respect to math/py-statsmodels and devel/py-xarray, I am a sysadmin not a Python programmer so I know very little about it. But I have encountered that the Python numeric stuff can be incredibly tricky to build and combine, and not everything works in every combination. If that stuff isn't commonly used, blanket adding it to RUN_DEPENDS may have serious unanticipated consequences. It's not something I feel I know enough about to mess with.
Created attachment 221231 [details]
Update to 1.1.5 and reorder Makefile vars
After v1.1.5 is released for some weeks now, I renewd the existing patch.
While here, I also reordered Makefile variables, as suggested by portclippy (ports-mgmt/portfmt) and added USE_GL=glu.
Forgot to mention, that for some days now, also v1.2.0 is released.
I haven't prepared a patch for it because there are some problems with plist that I don't understand at the moment.
^Triage: [tags] in issue Titles are deprecated
Don't know if this is meant with 'needs-qa':
Patch to v1.1.5 is tested on Poudriere (amd64, i386) on HEAD, 12.2, and 11.4.
'portlint -AC' is ok and portclippy has no suggestions for Makefile ;)