Comment 3 for bug 420738

I fixed the one that Michael refers to (bug #396838). It is unclear how to reproduce this one. As Michael says "its different in the details".

In the bug I fixed, it was an exception being raised as part of cleanup() [which like all destructor type code, really shouldn't be raising... :)]. So I trapped stuff during cleanup() and make sure to preserve the data structure.

The traceback here doesn't include any information about how to reproduce, so I don't really know what needs to be done. I'm guessing there is, indeed, a bug in the code wrt invariants being preserved in the face of exceptions.

However, it looks like we'd have to find it by inspection, which can certainly be tricky.