Coverage increased (+0.05%) to 92.517% when pulling dff7b20 on blueyed:fix-docs into 4705fd2 on pytest-dev:master.

For some reason pip install -e . does not install atomicwrites for me, while pip install . does.

There is no mention of "atomicwrites" with pip install -e ..

Seems to be related to my checkout. It works when using a fresh clone.
Very odd, any ideas?

But apart from that I think this PR might still make sense, at least partly (without the usedevelop removal).

It also works for me here. Perhaps you need to force tox to recreate the environment using tox -r? I've noticed it sometimes doesn't notice that the dependencies changed between master and features checkouts for example. This happened quite a few times with me, using -r always worked then.

I could fix pip install -e for my checkout by moving away
./pytest.egg-info (from 2015-11!).

requires.txt in there contained:

py>=1.5.0
six>=1.10.0
setuptools
attrs>=17.4.0
more-itertools>=4.0.0
pluggy<0.7,>=0.5

[:python_version < "3.0"]
funcsigs

[:sys_platform == "win32"]
colorama

Apparently this gets used instead of install_requires then?!

That I wouldn't know. 😁

Yes, looks like something strange/unexpected with setuptools/pip.

Using pip install -e . now creates ./src/pytest.egg-info instead/only, which existed before already.

Maybe there is something that could be specified in setup.py to make it look in src only for the egg-info?
It looks like this is an issue that arised from moving the code to src
/cc @RonnyPfannschmidt

nope, this is and will stay a mess due to setuptools legacy ^^

Amended.

Huh? Why is squash-and-merge in the Github UI not enabled?

@blueyed all history editing should be explicit and consensual - squash merge makes it really easy to do it wrong

Well, I disagree - especially with regard to the Github UI.
It is not history editing (of the target branch) at all, but rather that you do not have the merge commit. It is really good for small commits like this, but I won't argue about it.