Remove None values before invoking setup command

[chrisjsewell]

We think/hope this fixes #4032.
I'm not sure if there is a good unit test to confirm this? (just looking at tests/cmdline/commands/test_setup.py)

[giovannipizzi]

Just tested with Mohammad and it does not seem to be enough.
Maybe you can start adding a test that reproduces the problem, if possible?

[chrisjsewell]

Just tested with Mohammad and it does not seem to be enough.

damn

Maybe you can start adding a test that reproduces the problem, if possible?

well yes the million dollar question: it is a mac only bug and not so easy to reproduce. But I will have a try

[flavianojs]

It did work for me. I just had to make sure to relaunch the environment.

[giovannipizzi]

Just checked - the problem is that indeed the host is set to None, but this makes then SQLA crash. Is it set to None on purpose? Why it's not set to localhost as it should be, I think?

[giovannipizzi]

The problem is here I believe:

aiida-core/aiida/backends/utils.py

Line 28 in b9d4bbe

engine_url = 'postgresql://{user}:{password}@{hostname}{separator}{port}/{name}'.format(

That of course will set the string to ...@None...
I'm not even sure if it's possible to specify the engine string if you don't have a hostname.

So again my question: is it not set to localhost for a specific reason?
Pinging @sphuber and @ltalirz as well

(And who sets it to None rather than localhost? pgsu?

[chrisjsewell]

well this was my initial suggestion that seemed to work for me: #4032 (comment)

[chrisjsewell]

but as mentioned below that by Leo, it may not be a global fix

[chrisjsewell]

I think this can be closed in favour of #4255, @ltalirz and @giovannipizzi?

[ltalirz]

I think so, yes.
The conversion to "None" seems to have happened in the sql connect string, not inside click, see also #4032 (comment)