Fix cwd explosion

[marscher]

This contains only a small fix for the bug mentioned in #1235, so it should go to master.

[nicoddemus]

Could you use testdir instead? This way we can be sure the sub-test is actually failing, while the actual test_cwd_deleted then passes.

[nicoddemus]

Thanks for taking the time to work on this! 😁

[coveralls]

Coverage increased (+0.003%) to 92.207% when pulling c3a9700 on marscher:fix_cwd_explosion into 70fdab4 on pytest-dev:master.

[nicoddemus]

Sorry for not being clearer, I meant something like this:

def test_cwd_deleted(testdir):
    testdir.makepyfile("""
        def test(tmpdir):
            tmpdir.chdir()
            tmpdir.remove()
            assert False         
    """)
    result = testdir.runpytest() 
    result.stdout.fnmatch_lines(['* 1 failed in *'])
    assert 'INTERNALERROR' not in result.stdout.str() + result.stderr.str()

This way we make sure that when the test passes we are observing the correct behavior.

[marscher]

Ah okay. Thank you!

[nicoddemus]

While you are at it, please add a and@marscher_ for the PR. 😁

[coveralls]

Coverage increased (+0.003%) to 92.207% when pulling 7c3bbf8 on marscher:fix_cwd_explosion into 70fdab4 on pytest-dev:master.

[marscher]

Now paths are still relative if cwd exists and absolute if it doesn't. Is this really such a huge game changer to target feature?

[The-Compiler]

I think that'd be fine. If you have a pytest crash, $tool isn't going to be able to parse the output either, so the worst that can happen is that it still can't.

[coveralls]

Coverage increased (+0.007%) to 92.212% when pulling 09d163a on marscher:fix_cwd_explosion into 70fdab4 on pytest-dev:master.

[coveralls]

Coverage increased (+0.003%) to 92.207% when pulling cd30f3d on marscher:fix_cwd_explosion into 70fdab4 on pytest-dev:master.

[nicoddemus]

I agree, it can't possibly break anything, so in this case I think it is OK to go to master. @RonnyPfannschmidt do you agree?

Thanks again @marscher for the work and patience! 👍

[RonnyPfannschmidt]

I propose merging to features and doing 2.10 next weekend

[nicoddemus]

No problem in releasing 2.10 soon (although next weekend I'll be in Freiburg already), but do you have anything against merging this to master? I would rather not ask @marscher to rebase on features unless really required.

[marscher]

Do you think rebasing will cause a lot of conflicts? If you prefer this solution I'm happy with reopen this against the feature branch. Although I would say that a deleted cwd is a very rare case ;o) btw: Viel Spaß in Freiburg!

[RonnyPfannschmidt]

Agreed