*Unless* you fear that this would destabilize things (which are coming together so nicely), I'd suggest to check this in before beta2 goes out!
Accepted and Assigned to Neil for this reason.

It doesn't *look* like there's anything tricky going on, but I'll leave it to your judgement. If you don't want to check it in after all, please log in to sourceforge and change the status to Postponed.

Is there a reason this patch is just sitting here?  *Looks* like all the concerns were addressed two weeks ago.

- regrtest will print the repr of cyclic garbage found if -l option is specified and -q is not
- put else on newline in accordance with Python style
- constrain debugging options during test_gc (don't enabled DEBUG_LEAK)
- fix test_gc to remove garbage from gc.garbage

Is this patch is okay for 2.0?  If so I can check it in and close the patch.  I was planning to wait for post 2.0.

Checked in.  Guys, _please_ review my diffs.  The original patch did not apply 
cleanly.  Also,  I found some problems with the patch which I tried to fix.

Looks good. I approve the str fix. One nit: "else" clauses usually go
to a new line in the Python source. Make it "else if" on a new line.

Questions:

- DEBUG_LEAK now includes SAVEALL. The regression test suite sets
   DEBUG_LEAK by default. It used to report a bunch of objects during
   test_gc. Now it will collect the detected cycles in the suite. Should
   regrtest.py report them somehow?

- After the regression test, gc.garbage contains an A instance and an
   empty list. I think they're created by test_gc. Can't they be cleared?
   We can assign gc.garbage = [], but this doesn't clear anything.

- test_gc/test_saveall has gc.set_debug(0), followed immediately by
   gc.set_debug(gc.DEBUG_SAVEALL).
   Is this on purpose or is it a typo? BTW, I don't quite understand why
   this function does what it does... Could you elaborate on it?