bpo-43988: Add test.support.check_disallow_instantiation

[erlend-aasland]

https://bugs.python.org/issue43988

[erlend-aasland]

cc. @vstinner

[vstinner]

You should use f"xxx\.xxx" or fr"xxx.xxx". Or maybe use re.escape().

[erlend-aasland]

Since these tests are decorated with cpython_only, it should be ok to match the exact string, as produced by Objects/typeobject.c? Just matching the type module/name could in theory generate wrong result; maybe not in practice though.

[erlend-aasland]

What about setting the match pattern depending on what test.support.check_impl_detail() says?

[vstinner]

Enough nitpicking, you can always enhance the code later ;-) I merged your PR, thanks. I guess that the next step is to modify existing tests to use it.

[erlend-aasland]

Backport to 3.10? I guess not.

[vstinner]

Backport to 3.10? I guess not.

It seems like more changes are coming. Once they all landed into main, maybe write a backport to 3.10 PR including all of them at once (git cherry-pick -x all commits in the right order). Or we can leave 3.10 as it is: it's more up to you.

[erlend-aasland]

Lets leave 3.10 as it is.

[erlend-aasland]

@pablogsal, should we backport this? grep -r disallow_insta Lib/test | wc -l gives me 14 hits on the 3.10 branch, including the recently backported sqlite3 test.

[pablogsal]

@pablogsal, should we backport this? grep -r disallow_insta Lib/test | wc -l gives me 14 hits on the 3.10 branch, including the recently backported sqlite3 test.

I would prefer to backport it, yep

[erlend-aasland]

Once they all landed into main, maybe write a backport to 3.10 PR including all of them at once (git cherry-pick -x all commits in the right order)

@vstinner: I've opened GH-26885 which is a backport of this PR, GH-26394 (docs for this PR), and GH-26392 (apply this PR to Lib/test).