Even though pew has virtualenv as a propagatedBuildInputs, virtualenv is not being propagated to pipenv. This leads to: No module named virtualenv when running the following commands on nixos-19.03 (sha256:164g1qg2yp9w5a8klfxpw29xjslbxxz6q4v5d0zsxx6rsf04dzgy):

pipenv --python 3.7.3
pipenv lock
pypenv sync

Furthermore, after testing this against a real project, I found that the PYTHONPATH being set is not correct, as it does not take into consideration the current working directory for a given project. I added a new commit to this PR to reflect the needed changes to make it work. Can we get this merged soon? @zimbatm @flokli . Thanks!

/cc @berdario as the pipenv maintainer and @FRidh as he is maintaining python in general.

@dgarzon the commit should be formatted like: pipenv: <reason>. Other than that it looks ok to me.

I suspect that part of the issue is that pipenv doesn't depend on pew anymore

pypa/pipenv@a5a9583#diff-2eeaed663bd0d25b7e608891384b7298

So, the code that now creates virtualenvs and activates them is different than when it was first packaged (it requires direct access to virtualenv and it otherwise requires other fixes like the PYTHONPATH one)

pipenv usage of pew was also one of the reasons that I cared about it working properly and why I packaged it in the first place. As you can see, I haven't touched it since and in fact I'm not even using it personally anymore.

Sorry for being an absent maintainer.

After unbreaking the tool, we might want to address the above?

Thanks @berdario for the heads-up!

I guess in that case, we should drop pew from propagatedBuildInputs entirely. I grepped through pipenvs source code (master), and it doesn't seem to be used.

Thank you @berdario @zimbatm @flokli I have addressed the commit message structure, and also removed pew from the propagatedBuildInputs 👍