The attached file contains the proposal for 

(1) a new circle method (here called circle2)
(2) a new circle function (also called circle2)

Moreover it contains a testscript showing the
respective3 filling behaviour.

Visually they behave differently and it may well
be a point of discussion if this replacement
schould be made.

circle2 has correct filling behaviour and doesn't
show the strange 'not drawing feature' after a
radians() call.

All the other problems with radians()/degrees()
remain. (see #1514685)

Logged In: YES 
user_id=21627

New methods are not acceptable for Python 2.5 at this point.

Can you think of a reason why this should *not* become the
circle method? I understand that this fixes a bug, right?

Logged In: YES 
user_id=1547199

A misinterpretation:
I think it *should* become the circle method. (It were a 
weird idea to introduce a circle2 method anyway). I only
named it circle2 for now and test purposes in order to 
compare the behaviours of the two. If you are content
with the patch, replace circle by circle2 'as circle' - if
not leave circle2 out and let everything as is. Imho the 
the points to think about are:
(i) is the polygon approximation ok? I think, my testscript
gives a positive answer: the polygons are visually nearly
indistiguishable from the canvas circle items. (Moreover I
tested it with radiuses in the range from 1 to 200
successfully) 
(ii)that the new method relies on the forward method and
thus circles are drawn much slower. For me this is quite ok. 

Logged In: YES 
user_id=21627

Thanks for the patch. Committed as r47209.