I'm not sure what the right behaviour here should be. In my opinion, the project should stop honoring the "preferred encoding" and instead expect UTF-8 unless otherwise specified, as that's going to become the default behavior in Python 3.14 for most IO operations. I'm unsure of compatibility implications. It does appear as if this test (test_fallback_no_utf8) would no longer be relevant in that regime, so I'd just delete it. Regarding the non-UTF8 mode, that does sound more complicated, although...
EncodingWarnings in io module
The upstream pipx issue is https://github.com/pypa/pipx/issues/1198.
Yes, using the docutils app does work, so rst.* should also work once converted to console entry points. Thanks! I agree the "macOS workaround" could work around the issue, but it's not a very robust workaround, as it requires repairing every shebang on every installation, and moreover, the bug doesn't exist just in macOS or just with pipx. Here's the same bug manifest in Ubuntu using just pip: docker run -it jaraco/multipy-tox bash root@6b1625414874:/# mkdir /foo\ bar root@6b1625414874:/# py -m...
scripts install with broken shebang
narrow: indicated the default of 'Yes' when confirming auto-remove-includes
library: incorporate demandimport into runpy invocation
library: enable runpy invocation on mercurial package