@corona10 self.memo of PyScannerObject should be traversed too, right?

@pablogsal: Oh, that's an intersting bug. I consider that it should be backported to 3.7 and 3.8.

@pablogsal: Do you think that a NEWS entry is required for such fix?

Yes, as this may fix much more things that just a memory leak. If (unlikely) the memo gets into a ref cycle, some arbitrary set of objects may not be identified as a cycle and the leak will be considerable. This is a user-visible change: a potential leak that they may be experimenting should be no more, being the leak potentially unbounded.

And yes, it should be backported.

Thanks @shihai1991 for the PR, and @vstinner for merging it 🌮🎉.. I'm working now to backport this PR to: 3.7, 3.8.
🐍🍒⛏🤖

Sorry, @shihai1991 and @vstinner, I could not cleanly backport this to 3.8 due to a conflict.
Please backport using cherry_picker on command line.
cherry_picker b709302f3125622986bd458dfb2954fda5e8366d 3.8

Sorry @shihai1991 and @vstinner, I had trouble checking out the 3.7 backport branch.
Please backport using cherry_picker on command line.
cherry_picker b709302f3125622986bd458dfb2954fda5e8366d 3.7

Thanks @shihai1991 for the PR, and @vstinner for merging it 🌮🎉.. I'm working now to backport this PR to: 3.8.
🐍🍒⛏🤖

Sorry, @shihai1991 and @vstinner, I could not cleanly backport this to 3.8 due to a conflict.
Please backport using cherry_picker on command line.
cherry_picker b709302f3125622986bd458dfb2954fda5e8366d 3.8

@shihai1991: @miss-islington failed to backport automatically the change, there is merge conflict. You can please backport the fix 3.7 and 3.8? You can use cherry_picker or git cherry-pick -x b709302f3125622986bd458dfb2954fda5e8366d. Tell me if you need help for the backport.