Closed qbx2 closed 5 years ago
I actually see this manifest as a downstream dependency not working as intended when using the pypi version. Going to deploy and see if it resolved that issue for me also.
I have merged some changes to get the build automation back up an running. If you rebase the branch that would be helpful as there is no way through the github ui to retrigger the build. Going to rebase this branch locally and build a release.
Did that work?
The force push triggered the build, but looks like you haven't rebased the branch so it built using the broken config fixed by #55. I ended up building it locally to test and uploaded it to Pypi, as wanted to get rid of the universal wheel that was already there as we don't support py2 anyways.
I did some additional local testing with Blender upstream using the new version as a dependency and it failed miserably. This is not entirely unexpected as Blender is coupled to an older version anyways. Two things of note
Anyways this PR is certainly not the cause of the above issues so going to merge it in.
This is comparison between what blender is using and this version - https://github.com/niftools/pyffi/compare/niftools:887fe13610e8cdec96d1f67993acae1159e2e440...niftools:develop#diff-0e48de255873b7fbfcd0c291d00203e7R87
I think I have found the enum issue. Will require futher investigation to see if there is an underlying nif.xml issue. https://github.com/niftools/nifxml/compare/959cb9c3dd59a319e60e819fc8a1402f821f3684...f265c56482c728c6877e45d5b5993d3bff83670a I suspect the version that pyffi is currently using is just broken. It should really be updated to latest version but updates to meta-data parsing is required.
@niftools/pyffi-reviewer
Overview
setup.py is not including the package data
Fixes Known Issues
n\a
Documentation
[Overview of updates to documentation]
Testing
[Overview of testing required to ensure functionality is correctly implemented]
Manual
[Order steps to manually verify updates are working correctly]
Automated
[List of tests run, updated or added to avoid future regressions]
Additional Information
Caused by #53 update which didn't reconnect the package_data attribute