getting a class source regresses in Python 3.9

[t-vi]

BPO	42666
Nosy	@t-vi

^{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 2020-12-17.09:06:46.854>
labels = ['type-bug', 'library', '3.9', '3.10']
title = 'getting a class source regresses in Python 3.9'
updated_at = <Date 2020-12-17.09:11:37.904>
user = 'https://github.com/t-vi'

bugs.python.org fields:

activity = <Date 2020-12-17.09:11:37.904>
actor = 't-vi'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2020-12-17.09:06:46.854>
creator = 't-vi'
dependencies = []
files = []
hgrepos = []
issue_num = 42666
keywords = []
message_count = 2.0
messages = ['383223', '383224']
nosy_count = 1.0
nosy_names = ['t-vi']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue42666'
versions = ['Python 3.9', 'Python 3.10']

[t-vi]

getting a class source regresses in Python 3.9 onwards.
The following worked in Python 3.8, now it doesn't anymore for 3.9.1 and 3.10.0a2:

(save as foo.py)

import inspect

class Foo:
    def spam(self):
        global Bar
        class Bar:
            pass
        print(inspect.getsource(Bar))

Foo().spam()

It seems that getsource is very brittle for classes.

[t-vi]

I might add that this is a case I hit in the PyTorch test suite.

I tried to fix this a long time ago in a pricipled way in
https://bugs.python.org/issue33826

and Guido explained that fixing it was too heavy handed to be worthwile in

https://mail.python.org/archives/list/python-dev@python.org/thread/AC5B5VORS6FF6KJ3EG5DIV3ZLAQB6DIY/#UL376F3Y6FKJUZ2HZDA3ERFUDDOX67X4