Simon Conseil

@KathleenLabrie - Yes please open a new issue, with the example to reproduce :+1:

I don't think this has anything to do with tox. `holoviews` is trying to get its own version number from git if you use a git checkout (many packages do...

`param`'s version detection is just broken : https://github.com/holoviz/param/issues/468

1. You mean the `.data` attribute? It will still be a plain Numpy array, so indeed the problem is maybe not that worse. All the places where we use `ext.data`...

Tests are passing after setting the unit to ADU in `_append_nddata` and with https://github.com/GeminiDRSoftware/AstroFaker/pull/9. For `.append` I'm not sure what's the best way: we could add a `unit` argument that...

Ok, it works and I replaced various places where BUNIT was parsed. I'm not completely happy with the changes in `fluxCalibrate` but it's a bit tricky because `.uncertainty` has its...

Yes, that's the part I'm not happy with, but the problem with the previous version is that as it does the computation without units it doesn't update `.uncertainty.unit`. So I...

In the previous version `ext.multiply` was called with a Numpy array, without units, so NDData was not updating the unit. `ext.unit` was updated manually, but `ext.uncertainty.unit` was not so it...

So the other option I was thinking is https://github.com/GeminiDRSoftware/DRAGONS/pull/140/commits/ae59e963acd6396b0a8c487cc8839f035f5ef921 which I just pushed. I'm happy to revert to the original version if you prefer it, and setting manually `ext.uncertainty.unit`. Not...

That's also an option, but then something must be done for the setter method as well.