[Python-Dev] Re: Pre-existing bug of the millennium (so far)

[Tim Peters]

[Anthony Baxter]
> It's not clear to me as to how severe a problem this is -

It's a critical problem -- that's why I set the priority to 9.  It
affects all versions of Python since weakrefs were introduced.  If you
don't use weakrefs, or use any libraries that use weakrefs, it won't
bite you.  If you do, symptoms range from nothing, through "impossible
AttributeError"s at seemingly random times, to segfaults.

> given that beta2 is scheduled for next Wednesday, should I be delaying it?

I don't think so.  On my home box, the Python test suite contains new
tests for all the scenarios in the bug report, and I have a Python
that passes the -uall test suite in debug and release builds.  So it's
looking good.  I'm too punchy to check it in now, but if more tests
pass I'll upload a patch to SF before I go to sleep, and assign it to
Neil for review.

The interactions between cyclic gc and weakrefs are arguably delicate <heh>.