tox.ini missing from the distribution package

[sergiopasra]

Hi, I have noticed that tox.ini is not present in the last two releases available in PyPI. Its probably due to this commit 291a502

Best regards

[neutrinoceros]

Yes, that was intentional, albeit perhaps misguided. I didn't expect it would be depended on. What do you need it for ?

[sergiopasra]

I'm the packager of astropy in Fedora. We run the tests when building the package. I bet that all other linux distributions do the same.

[neutrinoceros]

I understand that, though tox isn't really needed for that. Is it deeply ingrained in your workflows ? What's your general approach to testing packages that don't use tox at all ?

[pllim]

@olebole how do you test over at Debian?

I guess if Fedora really need that tox.ini, it does not hurt to include.

[neutrinoceros]

I guess if Fedora really need that tox.ini, it does not hurt to include.

It doesn't, indeed, though I'm trying to understand if we really need to.
I would also note that this specific tox.ini is fairly simple, and could be transparently shipped again if we migrated its content to pyproject.toml

[olebole]

@olebole how do you test over at Debian?

We do this very pragmatically, and in this case without the tox.ini: With an updated astropy-iers-data package, we run the whole test suite of astropy, being its reverse dependency. So, specifically for this package we don't need it.

If there are local tests provided with the package, we would run them, but only in the default configuration, so no multiple Python or astropy versions - they would be tested only with backports.

[pllim]

I opened #70 just so I don't forget but if we agree to do something else, feel free to close it without merge. Thanks, all, and we apologize for any inconvenience caused, @sergiopasra .

[neutrinoceros]

If there are local tests provided with the package, we would run them

FYI this is #21 (WIP)

[sergiopasra]

I understand that, though tox isn't really needed for that. Is it deeply ingrained in your workflows ? What's your general approach to testing packages that don't use tox at all ?

We try to follow whatever practice upstream has, so I was just running the defaults tests with (basically) tox r -e test
Now that I look at the tests in detail, they are just astropy's tests, so I think I can stop using them, at least until #63 is implemented

[pllim]

Should I close this issue and also close #70 without merge then?

[neutrinoceros]

friendly reminder @sergiopasra

[sergiopasra]

Should I close this issue and also close #70 without merge then?

Yes, please, go ahead