The HTML spec describes two ways to encode an attribute
value that contains a URI with an ampersand.

http://www.w3.org/TR/REC-html40/appendix/notes.html#h-B.2.2


>>> from HTMLParser import *
>>> class P(HTMLParser):
...   def handle_starttag(self, tag, attrs):
...     print attrs
...
>>> P().feed("<tag attr=\"&\">")
[('attr', '&')]
>>> P().feed("<tag attr=\"&\">")
[('attr', '&')]

It seems that each string should produce the same
parsed value.  I would hazard a guess that the easiest
way to make this happen is to extend the current
unescape() to unescape character references.  Is there
any reason not to do that?  I'll provide a fix if that
sounds like a reasonable answer.

Maybe the charrefs were lost in the SF -> Roundup transition?

unescape() already converts named, decimal and hexadecimal entities, so this can be closed.

New changeset 3c3009f63700 by Ezio Melotti in branch '2.7':
#1745761, #755670, #13357, #12629, #1200313: improve attribute handling in HTMLParser.
http://hg.python.org/cpython/rev/3c3009f63700

New changeset 16ed15ff0d7c by Ezio Melotti in branch '3.2':
#1745761, #755670, #13357, #12629, #1200313: improve attribute handling in HTMLParser.
http://hg.python.org/cpython/rev/16ed15ff0d7c

New changeset 426f7a2b1826 by Ezio Melotti in branch 'default':
#1745761, #755670, #13357, #12629, #1200313: merge with 3.2.
http://hg.python.org/cpython/rev/426f7a2b1826

There was actually a bug with entities in unquoted attribute values. I fixed it and added tests for all the cases (quoted and unquoted).