Explain use of global thread-local variable in repr.

[thetorpedodog]

It's not completely obvious to a reader why we need a thread-local
"global" variable to track reference cycles in __repr__ calls,
and the test does not currently contain one. This change adds a comment
explaining the need and also adds a non-attrs value in the reference
cycle of test_infinite_recursion.

---

This was inspired by a bug that was opened where the reporter couldn't see why the global variable was needed in the __repr__ process. Initially I couldn't either, so I figured it was a good idea to add an explanation of why it is there—turns out it is an excellent solution to a real problem. I will gladly accept input about the nature/phrasing/placement of the commentary.

This is essentially a docs-only change (with a minor test change attached) so I don't think the full checklist applies in this case.

[hynek]

Can you please fix the line lengths? See https://github.com/python-attrs/attrs/runs/4056466583?check_suite_focus=true#step:5:43

I wonder if we should maybe add a blurb to the repr argument to explain, that setting it True involves threadlocals?

[thetorpedodog]

Can you please fix the line lengths?

79 characters! My worst enemy!

I wonder if we should maybe add a blurb to the repr argument to explain, that setting it True involves threadlocals?

I actually have an idea that may solve the cloudpickle incompatibility (and doesn’t involve any special handling on either side), which I am going to try once I get into work tomorrow. If that does the trick, I can create a new change with the fix, but if not I’ll add that note to the docs.

[thetorpedodog]

I actually have an idea that may solve the cloudpickle incompatibility (and doesn’t involve any special handling on either side), which I am going to try once I get into work tomorrow. If that does the trick, I can create a new change with the fix, but if not I’ll add that note to the docs.

Update: #857

[hynek]

Closing thanks to #857!