Dave Hirschfeld

I tested with the frontend implementation in this repo and am seeing the same issue with not being able to access the `repodata.json` file

I just tested with a package I uploaded to a new channel (i.e. not a proxy channel) and got the same error.

My current working assumption is that the issue (https://github.com/mamba-org/quetz/issues/583#issuecomment-1369412074) with the `quetz-client` creating a `null` user might be stuffing up the channel access as well: 

In a new instance with no `null` user: ``` quetz.sqlite> select * from users +------------------------------------+-------------------+-------+ | id | username | role | +------------------------------------+-------------------+-------+ | 0xd4e3705a759741919a6ed680cec74593 | dhirschfeld | owner |...

I guess it might be useful if the error message `[quetz] User not found` had a bit more context to aid debugging. Mentioning *which* user was not found might be...

I guess it's one of these two failing: https://github.com/mamba-org/quetz/blob/6cdf96e69ecae68fd9e7fe9b0efebe5ac5957f06/quetz/dao.py#L194-L198 https://github.com/mamba-org/quetz/blob/6cdf96e69ecae68fd9e7fe9b0efebe5ac5957f06/quetz/dao.py#L200-L204 ...but I'm in there so I'm not sure where it's getting it's `user_id` from: ``` quetz.sqlite> select * from users...

Thanks, I'll definitely give that a go! I've been meaning to pick this back up...

***xref:*** https://github.com/astropy/pytest-doctestplus/issues/14

I'm not fussed *where* it's implemented - I just opened an issue here since pytest already has some doctest functionality and this would be a great enhancement IMHO. I'd be...

`dask` has a function for calculating the *actual* CPU available: https://github.com/dask/dask/blob/1615724e22206112a6864bcd8b0d3f6afa07204e/dask/system.py#L15