Are you sure that this PR is related to https://bugs.python.org/issue35537?

@vstinner drat, copied and pasted the wrong issue number. It's actually https://bugs.python.org/issue35321. I updated the commit message, GitHub issue title & message.

Thank you @nnja for this PR!

I added the backport tags for 3.7 and 3.6, but I'd like to get @ned-deily 's take on that.

@warsaw: Please replace # with GH- in the commit message next time. Thanks!

Thanks @nnja for the PR, and @warsaw for merging it 🌮🎉.. I'm working now to backport this PR to: 3.6, 3.7.
🐍🍒⛏🤖

Sorry, @nnja and @warsaw, I could not cleanly backport this to 3.7 due to a conflict.
Please backport using cherry_picker on command line.
cherry_picker 69091cb497b2f0fe7e2789b30b43cf78caf9de9b 3.7

Sorry, @nnja and @warsaw, I could not cleanly backport this to 3.6 due to a conflict.
Please backport using cherry_picker on command line.
cherry_picker 69091cb497b2f0fe7e2789b30b43cf78caf9de9b 3.6

AFAICT this is not a 3.6 regression: the behavior has been around for a long time. Even though it is minor and the risk of breakage is small, the benefit seems to be also small. My take is that this doesn't meet the criteria for a security-fix-only branch so it should not go into 3.6 or earlier. Sorry!

Thank you @ned-deily - we'll just back port this to 3.7.

I have a question about how to backport this into 3.7. (maybe @warsaw or @ned-deily can help?)

I used the cherry_picker tool to try to apply this commit to 3.7. When I looked at the merge conflict
I realized it's in Python/importlib.h which is the generated header file for this frozen module.

As expected, this file has irresolvable conflicts.

What's the process for this type of backport?

I'm not sure this is the right way to do it (does the devguide provide any help?). What I generally do is to use the tool to do the merge in the older branch, then revert Python/importlib.h, then make regen-importlib to get a proper importlib.h. Then that's the branch I propose for merging.

I wonder if @Mariatta has any thoughts on adding this to @miss-islington by default?

you can resolve the conflict manually (regenerate the files) and the do cherry_picker --continue.

Currently miss-islington can't do this automatically. There is open ticket here: python/miss-islington#41.

I'll try this approach and follow up if I run into any issues. Thank you @Mariatta and @warsaw!