Nicholas Tollervey has been getting Pynsist to use the new pythonXY._pth file to assemble sys.path. However, I'm not clear on how this behaves with .pth files. The docs on this (https://docs.python.org/3/using/windows.html#finding-modules ) appear to contradict themselves:

> site is not imported unless one line in the file specifies import site
> ...
> Note that .pth files (without leading underscore) will be processed normally by the site module.

I can see two possibilities:

1. .pth files are processed normally if and only if the ._pth file specifies 'import site'. If it doesn't, they are ignored.
2. If a .pth file is encountered, site is imported to process it, but with sys.flags.nosite set, so that site doesn't do the normal stuff on import.

Some packages unfortunately seem to depend on .pth files getting processed, so if 1. is the case, Pynsist needs to ensure site is loaded.

Thanks :-)

1 is correct. I don't see the contradiction - maybe you could purpose a rewording? (Perhaps it currently relies on the knowledge that the site module finds and processes .pth files? -S will also suppress it.)

I'd reword the second sentence like this:

> Note that .pth files (without leading underscore) will be processed only if this file specifies ``import site``.

The current wording sounds (to me) like a guarantee that .pth files will still be handled, despite the other things that this mechanism switches off.

New changeset 15ea3a6a320623328530e5d8da5b83c75766180b by Steve Dower in branch 'master':
bpo-32699: Improves doc for .pth files in presense of a ._pth file (#5399)
https://github.com/python/cpython/commit/15ea3a6a320623328530e5d8da5b83c75766180b

Hopefully that's enough of a clarification.

Thanks Steve, that looks good.

resolution: fixed

assignee: docs@python
stage: patch review -> resolved
versions: + Python 3.7, Python 3.8
nosy: + docs@python
messages: + msg311043
components: + Documentation
type: enhancement