I have been thinking about this recently but have found no good solution so I thought I might post it here to see what other people think:
Consider a cylinder made of iron that is magnetised so that the two poles are on the two opposite circular ends. Now suppose we bend this magnet. Bending the magnet does not change where the poles are ( alternate: or consider a bent version of the cylinder magnetised with the two poles on opposite circular ends ).
Suppose we bend this magnet into a torus, so that the north/south poles are brought into direct contact with each other (as they are opposite they should attract so repulsion is not a problem here) to form a torus. (alternate: or consider a series of progressively more bent cylinders whose limit is the torus).
What I can't seem to figure out is where would the North/South poles of this magnet be or even whether the torus would have a magnetic field!
Any help is much appreciated.