Handle alternate names for UTF-8 encoding

[ob-stripe]

The issue

Currently, pipenv does this:

1. In __init__.py, sys.stdout and sys.stderr are replaced by io.TextIOWrappers with encoding set to 'utf8' (no dash):

   pipenv/pipenv/__init__.py

   Lines 31 to 36 in d9ebad3

   	stdout_wrapper = io.TextIOWrapper(sys.stdout.buffer, encoding='utf8')
   	atexit.register(stdout_wrapper.close)
   	stderr_wrapper = io.TextIOWrapper(sys.stderr.buffer, encoding='utf8')
   	atexit.register(stderr_wrapper.close)
   	sys.stdout = stdout_wrapper
   	sys.stderr = stderr_wrapper

2. In _compat.py, the encoding is forced to UTF-8 by checking if sys.stdout/err.encoding.lower() is equal to 'utf-8' (with a dash):

   pipenv/pipenv/_compat.py

   Line 331 in d9ebad3

   if stdout_encoding.lower() != "utf-8" or stderr_encoding.lower() != "utf-8":

Fixes #3313.

The fix

• Canonicalizes the encoding name using codecs.lookup(). This will always return 'utf-8' for UTF-8 encodings, no matter what alias is used ('utf8', 'UTF-8', 'u8', etc.)
• Catch ImportError when importing ctypes, so that this code path doesn't crash with PyPy (which doesn't support ctypes). @frostming mentioned in Incompatibility with pypy3 #3313 that pipenv doesn't officially support PyPy so I can remove this if you'd prefer, but it seems like a net improvement to me.

The checklist

• Associated issue: Incompatibility with pypy3 #3313
• A news fragment in the news/ directory: 3313.bugfix.rst

[techalchemy]

Thanks! if you run into anything else definitely let us know, I may experiment with adding PyPy to CI if we can

[ob-stripe]

Cool, thanks!

[ob-stripe]

@techalchemy Sorry to bother you, but do you have any idea when the next pipenv version will be released? The issue I fixed in this PR has been affecting our CI pipeline for a few months now.

[techalchemy]

Apologies, I have been forced to split focus on a number of projects. Stewardship here seems to have been excellent so I’m just catching up and getting my own changes together so hopefully soon!