In the original code, a for-else is used with wrong logic, so that if
there was a response ('no blank line between quoted message and
response') in the section, the loop would break and not execute the
else[1]. Unfortunately, the code that checks whether to include the
quoted text verbatim is inside the else clause.

Attached patch fixes this and reorganizes the code a bit. Also tracked
at http://psf.upfronthosting.co.za/roundup/meta/issue264

Daniel, I've looked into this and also studied issues  264 and 321 in
your meta tracker.
Are we talking about the same settings in config.ini:
- keep_quoted_text = no
- leave_body_unchanged = no

With these I'm not seeing the same results with the original message in 
http://psf.upfronthosting.co.za/roundup/meta/file197/msg99926-orig.txt
and outcome in http://bugs.python.org/msg99926 -- neither with nor
without your patch.

- Without the patch *all* quoted text is stripped -- and I've always
  thought that this is the intention of "keep_quoted_text=no" (that's
  why I'm usually turning it to "yes" in my trackers)
- with your patch the *last line* of quoted text is kept if an empty
  line follows (if the user directly replies to a portion of quoted
  text without an empty line after the quote, the quote is still
  stripped)
So I'm unsure what the patch should have achieved and can't reproduce
the current behaviour of bugs.python.org. I've not yet closely looked at
your code in the patch, though.

I think what really *would* be nice is a heuristic that strips quotes
only if the user did top- or bottom-quoting for the whole message, it
should try to keep quotes in case the user did some effort in citing
specific portions of the original message.

Ah. You're using "keep_quoted_text = yes" and it doesn't -- sometimes.
I see the purpose of your fix but it seems to break keep_quoted_text = no

fixed in git 709f252