Add ability to access env vars that are changed in test-related subprocesses

[duckinator]

What's the problem this feature will solve?

While I was working on #6391 to add a pip cache command, I ran into an issue where environment variables are changed in subprocesses started when a test uses script.pip(...) and similar, but I have no simple way to get this information. See #6391 (comment).

Here's @chrahunt's explanation from #6391 (comment):

The basic issue appears to be that changes to APPDATA and LOCALAPPDATA in the environment of the current process do not impact the returned value of ctypes.windll.shell32.SHGetFolderPathW. Only for sub-processes.

One option could be to take the current value of os.environ['LOCALAPPDATA'] and append \pip\Cache to it, duplicating the logic from appdirs.user_cache_dir.

Another option could be to get the cache dir from a sub-process itself, like script.run('python', '-c', 'from pip._internal.locations import USER_CACHE_DIR; print(USER_CACHE_DIR). This seems like more of a hack, but does avoid the duplication.

In either case, it may be good to make it a property of PipTestEnvironment so it's accessible from the required tests or any others that might need the cache directory.

Describe the solution you'd like

TBD — I'm not sure what approach would be best, and was hoping to get some input on that.

Additional context

PR where this problem came to my attention: #6391

Code where the environment variables are changed: https://github.com/pypa/pip/blob/master/tests/conftest.py#L98-L166

File where PipTestEnvironment, a potential location for tracking this information, is defined: https://github.com/pypa/pip/blob/master/tests/lib/__init__.py

---

cc @chrahunt

[duckinator]

One approach would be to have PipTestEnvironment have a function which returns a dict mapping environment variable names to the new values, for anything that'd be overridden. Then, the current code in tests/conftest.py could iterate over the returned dict and apply the actual environment changes there.

I'm not sure if this is a good approach, however.

[pradyunsg]

Whee! Thanks for filing a new issue for dedicated discussion! :)

get the cache dir from a sub-process itself, like script.run('python', '-c', 'from pip._internal.locations import USER_CACHE_DIR; print(USER_CACHE_DIR). This seems like more of a hack, but does avoid the duplication.

Let's just do this, I don't mind the slightly-fragile nature of this and we can revisit this later if needed.

[duckinator]

@pradyunsg would it make sense to create a helper function that runs that, or do you think that should be delayed until it's needed at least twice?

[pradyunsg]

Let's delay that -- I'm fairly averse to over-engineering things so I don't think there's much point in doing this eagerly.

[duckinator]

@pradyunsg sounds good to me. In that case, should I go ahead and close this issue?

[pradyunsg]

I did that. ;)