Michael König

Pull request #105 was merged. So you could just update to the latest master and retry the build without exporting any `VIRTUAL_ENV` variable.

@aschmu Can you confirm that this resolves your issue?

Ok, maybe @xhochy could have another look at the arrow script? Building arrow from scratch should not be necessary. I don't do it on my local machine, and continuous builds...

Just that I don't get confused: * @aschmu experienced cmake problems with finding NumPy and Arrow in conda environments. @xhochy supposedly fixed that in pull requests #105 and #106. #105...

I'll try to setup a local Fedora VM in the near future to do some debugging. Thanks for reporting back!

Thanks for the clarification. I suspected I may have become confused ;-)

Thanks for reporting!

This particular driver is starting to annoy me ;-)

Yes, currently it is the same, and this seems to violate the specification you referenced (funny thing that the column size is ignored for most data types). Splitting element size...

That's good to hear. Still I'll keep this issue open and see if one can do some generic database driver capability detection with respect to the fraction second precision.