bpo-45873: Restore Python 3.6 compatibility (GH-29730)

[tiran]

https://bugs.python.org/issue45873

Automerge-Triggered-By: GH:tiran

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @tiran for commit f5818e6 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[gvanrossum]

LGTM, but is it even worth testing for 3.7+? Why not always take the slow path?

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @tiran for commit cb24cf5 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[tiran]

LGTM, but is it even worth testing for 3.7+? Why not always take the slow path?

I left the old implementation in because it's more readable. The fast path is now in a doc string for future reference.

I also added more debug code to configure to figure out what is going on on the Debian buildbot.

[tiran]

Closing and reopening to retrigger Azure.

[gvanrossum]

(Why do we even bother with "python3 python"? For weird platforms that have only one of those but it happens to be 3.6 or better?)

[tiran]

(Why do we even bother with "python3 python"? For weird platforms that have only one of those but it happens to be 3.6 or better?)

I don't know and kept python 3 python just in case.

[tiran]

Broke Azure tests are likely caused by GH-28612GH-29488.

[gvanrossum]

:-)

[gvanrossum]

Broke Azure tests are likely caused by GH-28612.

How? That PR was closed without merging. And if there was a cause outside this PR why would it not have failed before (e.g. when I submitted #29717)?

Are you suggesting there are some cached files from running that PR present?

[miss-islington]

@tiran: Status check is done, and it's a failure ❌ .

[tiran]

I confused GH-28612 with GH-29488. Both are for bpo-39026.

This PR was created after @vstinner merged GH-29488. That's why my PR has failing Azure tests and your PR did not.