gc assertion failure

[Michael Hudson]

Todd Miller <jmiller at stsci.edu> writes:

> Michael Hudson wrote:
> > Todd Miller <jmiller at stsci.edu> writes:
> > 
> >>I recently discovered an assertion failure in the Python garbage
> >>collection system when scripts using our C extension (numarray)
> >>exit. The assertion is activated for Pythons configured using
> >>--with-pydebug. I have a feeling I may be doing something wrong with
> >>garbage collection support for some of our c types,  but I'm not sure
> >>exactly what.
> >>
> >>Here is the assertion output:
> >>
> >>python: Modules/gcmodule.c:231: visit_decref: Assertion
> >>`gc->gc.gc_refs != 0' failed.
> >>Abort (core dumped)
> >>
> >>
> >>Here's the traceback from gdb:
> > [snip]
> >>Can anyone give me any insight into what needs fixing?
> > Oh good grief, it could be anything (I think).
> 
> Sorry...

It seems I was a bit extreme, reading followups.  But I think that
assertion failure has been a symptom of some scary bugs in Python
itself.

> I have two theories myself:
> 
> 1) I need to implement explicit support for  GC for some of the
>    numarray c types.  I was hoping to avoid this by virtue of the
>    unlikeliness of reference cycles in the types in question.
> 
> 2) There are other reference counting errors in numarray which are
>    being exposed during gc.  In particular,  it appears the tuple type
>    is being traversed when the assertion fails.

I know you found your error, but yet another tool for finding things
like this is the TrackRefs class from Zope's test runner.  I found a
bunch of refcounting bugs in Python 2.3(.0) with it a couple months
back.

Cheers,
mwh

-- 
  About the use of language: it is impossible to sharpen a
  pencil with a blunt axe.  It is equally vain to try to do
  it with ten blunt axes instead.
      -- E.W.Dijkstra, 18th June 1975. Perl did not exist at the time.