I'm struggling to see any difference in behavior here, so:

• I'm not understanding this, or
• the original code wasn't buggy (I think that is the case), or
• this doesn't fix the bug.

Or is this just to fix the warning and not change the semantics?

The warning causes a flaky test, see https://buildbot.python.org/#/builders/29/builds/7932

I agree the "is" and "is not" would be slightly better then ==. Also, I think it helps to "push down" the code that's inside the outer try/except into a helper method. The triple nested try/except is quite hard to understand.

I agree the "is" and "is not" would be slightly better then ==. Also, I think it helps to "push down" the code that's inside the outer try/except into a helper method. The triple nested try/except is quite hard to understand.

This can be done as a further improvement. In this PR I'd rather make the change more straightforward to see from the diff.

The warning causes a flaky test, see https://buildbot.python.org/#/builders/29/builds/7932

The issue says there is a bug, not just a warning. What's the bug?

The warning causes a flaky test, see https://buildbot.python.org/#/builders/29/builds/7932

The issue says there is a bug, not just a warning. What's the bug?

See discussion on the issue - assumption now is that it's not a bug. I'll fix the title.