BACKPORT: Introduction of Config.invocation_args

[fermezz]

This change was first introduced in #5564 but, logically, only since the 5.1.x version. I'm back-porting it into the 4.6.x branch since many people still use it in the meantime they migrate to Python 3.

I was also able to test this backport to be working successfully locally.

[fermezz]

I'm not sure if this the thing to do when back-porting things. If it's not, let me know! Just following the steps-to-contribute 🙂

[nicoddemus]

It is definitely the right thing! 👍

[fermezz]

Since there is not an actual 4.6-maintenance branch, I changed it for what I believe is correct. This should probably be in master as well. I can make that happen if I get confirmation that that's ok.

[nicoddemus]

No need to port to master, we have changed that significantly already: https://docs.pytest.org/en/latest/py27-py34-deprecation.html

[asottile]

imo this patch extends beyond "bugfix" and probably shouldn't be backported -- curious to hear others thoughts though

[fermezz]

Yeah, the way I think about it is that this change, although feature, was specifically introduced to fix a bug in the interaction with pytest-xdist. So while technically it is a feature for pytest, it's a bugfix for the pytest-ecosystem.

That said, I'm open to suggestions. I can maybe find a workaround for my specific use case, but I'm probably not the only one still not being able to update to 5.1 (because of Python 2).

What do you think?

[blueyed]

@fermezz
FWIW I think it makes sense, given that it is a bugfix after all.

For the Travis config check https://github.com/blueyed/pytest/blob/my-4.6-maintenance/.travis.yml, where it still works.

[fermezz]

Thanks @blueyed. I'll give it a shot at fixing that up, although I don't see a lot of differences beyond having fewer jobs to build 🤔.

[fermezz]

I have a separate PR to fix CI by migrating it to GitHub actions #6884. I appreciate input there, especially since I'm going to need assistance on modifying the required checks. Also, see #6821

[nicoddemus]

imo this patch extends beyond "bugfix" and probably shouldn't be backported -- curious to hear others thoughts though

While I agree that technically it is a (small) new feature, the main reason this was created was for a bug fix in xdist as @fermezz mentioned. Given that this allows to use a more recent pytest-xdist version which fixed a serious config problem, I think we can open an exception. 👍

[nicoddemus]

Let's add a remark here:

Suggested change

	New ``Config.invocation_args`` attribute containing the unchanged arguments passed to ``pytest.main()``.
	New ``Config.invocation_args`` attribute containing the unchanged arguments passed to ``pytest.main()``.
	Remark: while this is technically a new feature and according to our `policy <https://docs.pytest.org/en/latest/py27-py34-deprecation.html#what-goes-into-4-6-x-releases>`_ it should not have been backported, we have opened an exception in this particular case because it fixes a serious interaction with ``pytest-xdist``, so it can also be considered a bugfix.

@asottile what do you think?

[asottile]

👍 makes it much more clear that this is one-off and not a greenlight to resume features on 4.6

[fermezz]

Remark added, it makes a lot of sense.

[nicoddemus]

Oh just noticed this was not added in the end.

No worries, I will update directly on #7199. 👍

[fermezz]

I committed it locally and never pushed 🤦. Sorry about that

[nicoddemus]

No problem 😁

4.6.10 has been released btw! 👍

[fermezz]

I added a remark to be clear that this is a one time thing about backporting a feature. We should rebase this branch after #6884 is merged so that we can have a successful travis run before merging it.

[fermezz]

Thoughts about this one?

[nicoddemus]

@fermezz sorry about the delay, we were resolving some internal issues.

Thanks a lot for the PR!

[fermezz]

@nicoddemus I heard, sorry you had to figured that out, but I'm glad y'all are back! And thanks for assisting!!