The search for pyvenv.cfg doesn't match PEP 405

[mattheww]

BPO	33302
Nosy	@mattheww

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2018-04-17.21:48:27.965>
labels = ['interpreter-core', 'type-bug', '3.7']
title = "The search for pyvenv.cfg doesn't match PEP 405"
updated_at = <Date 2018-04-17.21:48:27.965>
user = 'https://github.com/mattheww'

bugs.python.org fields:

activity = <Date 2018-04-17.21:48:27.965>
actor = 'mattheww'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Interpreter Core']
creation = <Date 2018-04-17.21:48:27.965>
creator = 'mattheww'
dependencies = []
files = []
hgrepos = []
issue_num = 33302
keywords = []
message_count = 1.0
messages = ['315424']
nosy_count = 1.0
nosy_names = ['mattheww']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue33302'
versions = ['Python 3.5', 'Python 3.6', 'Python 3.7']

[mattheww]

PEP-405 says that the pyvenv.cfg file is found as follows:

« a pyvenv.cfg file is found either adjacent to the Python executable or one directory above it (if the executable is a symlink, it is not dereferenced), »

But in cpython if the executable is a symlink, it _is_ dereferenced before searching for pyvenv.cfg .

I've checked this behaviour with Python 3.5, 3.6, and today's 3.7 tip.

Looking in 3.7's getpath.c, calculate_path_impl() calls, in order:

   calculate_program_full_path()
   calculate_argv0_path()
   calculate_read_pyenv()

It looks like the symlink resolution happens near the end of calculate_argv0_path().

I think this means that the 'home=' line in pyvenv.cfg files generated by the 'venv' module is irrelevant (and seems likely to confuse anyone who might try to construct a Python environment by imitating a built venv).

[mattheww]

I think this was fixed by #29041 (for Python 3.11) and Python now behaves as PEP 405 said it should.