Patch for setup.py egg_info issue

[matteius]

It was discovered that requirementslib uses the sys.executable python rather than the pipenv environment python. Which as @oz123 reminded me should be the same, but due to some weird code paths, it wasn't always the same in my testing when the resolver got invoked as subprocess.

The issue

Fixes #5753

The fix

I don't love the use of the environment variable, but I figured out this was the original intention of how the pipenv <--> requirements lib interface and at some point it got hard-coded to sys.executable in the resolver code which was getting invoked with a different python version at points.

This fixes the problem by ensuring that the correct python of the pipenv environment is being used.

The checklist

• Associated issue
• A news fragment in the news/ directory to describe this fix with the extension .bugfix.rst, .feature.rst, .behavior.rst, .doc.rst. .vendor.rst. or .trivial.rst (this will appear in the release changelog). Use semantic line breaks and name the file after the issue number or the PR #.

[cclauss]

Suggested change

	stdout=sp.PIPE,
	stderr=sp.PIPE,
	capture_output=True

---

% MY_CODE="import subprocess as sp ; sp.run([python, "setup.py"] + args, cwd=target_cwd, stdout=sp.PIPE, stderr=sp.PIPE)"

% echo $MY_CODE | ruff --select=UP022 --fix -

import subprocess as sp ; sp.run([python, setup.py] + args, cwd=target_cwd, capture_output=True)

% ruff rule UP022 # https://beta.ruff.rs/docs/rules/#pyupgrade-up

replace-stdout-stderr (UP022)

Derived from the pyupgrade linter.

Autofix is always available.

What it does

Checks for uses of subprocess.run that send stdout and stderr to a
pipe.

Why is this bad?

As of Python 3.7, subprocess.run has a capture_output keyword argument
that can be set to True to capture stdout and stderr outputs. This is
equivalent to setting stdout and stderr to subprocess.PIPE, but is
more explicit and readable.

Example

import subprocess

subprocess.run(["foo"], stdout=subprocess.PIPE, stderr=subprocess.PIPE)

Use instead:

import subprocess

subprocess.run(["foo"], capture_output=True)

References

• Python 3.7 release notes
• Python documentation: subprocess.run

[cclauss]

Was the removal of cwd=target_cwd intentional?

[matteius]

@cclauss intentional, as it seemed unnecessary with the cd(target_cwd) context manager -- should already be in that directory.

[matteius]

Well its finally having a good run -- some just turned green.

[cclauss]

Thanks massively. This fixed the problem for me.