Don't crash if an item has no _fixtureinfo attribute

[obestwalter]

fixes #2788

Simplest possible way of making that error go away. I have obviously no idea what I am doing though.

[nicoddemus]

Thanks @obestwalter, that's definitely a good start!

We should add a regression test for it though. I think you can add a new test in testing/python/show_fixtures_per_test.py that contains a docstring with a doctest and a text file, I suggest:

def test_doctest_items(testdir):
    testdir.makepyfile('''
        def foo():
            """
            >>> 1 + 1
            2
            """
    ''')
    testdir.maketxtfile('''
        >>> 1 + 1
        2
    ''')
    result = testdir.runpytest("--fixtures-per-test", "--doctest-modules", "--doctest-glob=*.txt", "-v")
    assert result.ret == 0

    result.stdout.fnmatch_lines([
        '*collected 2 items*',
    ])

This fails for me with the same error you got originally.

Also, please add a new towncrier entry, I suggest 2788.bugfix:

Fix crash when using ``--fixtures-per-test`` in directories with doc test items (docstrings or text files).

[nicoddemus]

This is just me and a question of preference, but I try to narrow down my try/except blocks as much as possible, so I would write instead:

try:
    name2fixturedefs = item._fixtureinfo.name2fixturedefs
except AttributeError:
    return

if not name2fixturedefs:
    return

But really it is up to you. 👍

[obestwalter]

good point. will do that.

[coveralls]

Coverage decreased (-0.02%) to 91.808% when pulling ed2a7cf on Avira:master into bf77daa on pytest-dev:master.

[obestwalter]

Thanks, will fix it up properly tomorrow :)

[nicoddemus]

Awesome, thanks! I like the small refactorings you did there. 👍

[obestwalter]

Thanks.

Bit of a mess up now though: I want to learn to have a cleaner commit history so I squashed the changes, but this has closed this PR. Need to learn how to do this properly ...

[obestwalter]

ah just reopening contains the single commit then ... phew

[coveralls]

Coverage increased (+0.002%) to 91.828% when pulling bf77daa on Avira:master into bf77daa on pytest-dev:master.

[nicoddemus]

I want to learn to have a cleaner commit history so I squashed the changes,

Did you do it locally? I do it frequently and it doesn't need to close the PR. Here's how I usually do it:

git rebase -i ... # squash squash squash
git push --force

[obestwalter]

Yes, locally: I did a soft reset to the last commit before my changes, stashed the changes, then did a hard reset to that commit, then a force push and then unstashed, committed and pushed the changes again ... that was the first thing I could think of how to do this ... Until now I have only really worked on projects where nobody cares much about git history cleanliness, so I never bothered and stacked up commit on commit - working on tox now though I start to care so I have to up my game there :)

Will do some reading about how to do this without upsetting the github side of things.

[obestwalter]

o.k. the way you did it looks much easier anyway - so:

git rebase -i origin/master~<wherever I want to go back to> master
git push origin +master

[nicoddemus]

...reset to that commit, then a force push...

OK, this is what closed the PR then: this ended up pushing a branch with no changes compared to master. A little weird that this closes the existing PR, but at least it is consistent given that you can't create a PR without any changes compared to the target branch.

[obestwalter]

Makes sense.

[coveralls]

Coverage increased (+0.002%) to 91.828% when pulling 2802135 on Avira:master into bf77daa on pytest-dev:master.

[coveralls]

Coverage increased (+0.002%) to 91.828% when pulling 2802135 on Avira:master into bf77daa on pytest-dev:master.

[nicoddemus]

Unless somebody wants more time to review this I will merge this later today! 👍

[nicoddemus]

Thanks @obestwalter!