gh-70331: Protect IDLE's imports from user files in the current directory

[serhiy-storchaka]

A user file such as random.py, threading.py or tkinter.py in the directory from which IDLE was started (on Windows, "Edit with IDLE" starts it in the file's directory; on macOS, IDLE.app starts in ~/Documents) shadowed the stdlib modules imported by IDLE, in the IDLE process (python -m idlelib) or in the user process (started with python -c), and IDLE failed to start.

Now the user process is started with -P, so the current directory is not on sys.path while idlelib.run and its dependencies are imported. sys.path of the user process is set by transfer_path() later anyway, so user code sees the same sys.path as before. python -m idlelib removes the current directory from sys.path before importing idlelib.pyshell, which imports almost all of IDLE.

pyshell.main() still adds the current directory or the directories of the files to be edited to sys.path of the IDLE process, so the few modules imported later (pydoc) can still be shadowed. Changing this would change sys.path seen by user code, so it is left for a separate change.

🤖 Generated with Claude Code

• Issue: Ensure that IDLE's stdlib imports are from the stdlib #70331

[terryjreedy]

I once did some research on on what sys.path looks like after various methods of starting python without IDLE. Probably on some other issue. I should repeat and save somewhere in idlelib.

[bedevere-app]

When you're done making the requested changes, leave the comment: I have made the requested changes; please review again.

[serhiy-storchaka]

Done for idlelib.idle: the same three lines at the top, before it imports os.path, guarded by __spec__ is not None so that running idle.py as a script keeps its own directory in sys.path. The test now shadows os.py too and checks both -m entry points.

idle.pyw and pyshell need nothing. .pyw can only be run as a script, so sys.path[0] is its own directory, like for idle3 and idle.bat. Every entry point fixes sys.path before importing pyshell, and module-level code there would also run in tests, where deleting sys.path[0] is wrong — that is also why I duplicated the lines instead of importing a shared module, which could itself be shadowed.

Measured sys.path[0]: the current directory for python -m idlelib and python -m idlelib.idle, the script's directory for idle.py, idle3, idle.bat and idle.pyw, and '' for the user process. The user process is affected whichever way IDLE is started, only the moment differs: at startup with Shell, at F5 from an editor. NEWS reworded accordingly.

I have made the requested changes; please review again

[bedevere-app]

Thanks for making the requested changes!

@terryjreedy: please review the changes made to this pull request.

[terryjreedy]

I included pyshell with idle(w) because it is also an entry point. I vaguely remember seeing is documented as such (though not sure if 'officially'). The lines if __name__ == "__main__":\n main() ended the initial version. I'd like to deprecate this, but have no idea how many scripts that would break. idle.py originally consisted of

#! /usr/bin/env python

import PyShell
PyShell.main()

Both were part of the initial commit.

Should it not get the same treatment?

[miss-islington-app]

Thanks @serhiy-storchaka for the PR, and @terryjreedy for merging it 🌮🎉.. I'm working now to backport this PR to: 3.13, 3.14, 3.15.
🐍🍒⛏🤖

[bedevere-app]

GH-158034 is a backport of this pull request to the 3.15 branch.

[bedevere-app]

GH-158035 is a backport of this pull request to the 3.14 branch.

[bedevere-app]

GH-158036 is a backport of this pull request to the 3.13 branch.