With this fix, gist (https://gist.github.com/malinoff/148507bf51ff8b9a0a947c178aa0ac3a) output is

$ python example.py
False
False
False
False
True

Errors do not seem to be related to my change.

Thanks @malinoff and @RonnyPfannschmidt, I will take a more detailed look at the PR and the CI errors but probably will only have time for it later in the week though.

I pushed some commits to the PR:

• The errors were related to the new asyncio import, which in turn imports logging which breaks an acceptance test for len(list) should be summarized in most cases #8. I had to resort to copy the relevant 2-lines function to circumvent that. It is simple enough so IMHO it shouldn't be a problem.
• Updated the CHANGELOG;
• Moved the "compat" tests to a single module and used testdir on the tests with newer Python syntax. While I usually prefer unit-tests instead of integration tests when possible, I think it is acceptable in this case to avoid importing asyncio into pytest's process and the flake8 errors.

Thanks again to @malinoff for pursuing this. 👍

Coverage remained the same at 92.832% when pulling caee5ce on malinoff:fix-2129 into da40bcf on pytest-dev:master.

Hmmm now py27-trial is also breaking. Perhaps it is related to #1989?

@nicoddemus could you please merge this?

@malinoff I would like to take a look at the trial failures first. While they seem unrelated, I'm not sure.

Are you using pytest in development mode?

The trial failures are certainly unrelated since they are happening on master as well and this PR has been open long enough, so I'm merging this.

Thanks @malinoff for the patience! 🙇

Yay! Thanks @nicoddemus