as i recall, this and several other peculiar effects
were the consequence of the process
by which the tree was rewritten.
it was along these lines. (i could of course go
off and look at the source but is it worthwhile?)
you might have thought it would leave the types
alone and then when pointer-to-function appeared
as the type of an expression to which () was applied
it would insert a dereference.
but instead it applied the f to &f transformation
so *****f involved not a while loop but repeatedly
introducing &f from the bottom of the tree and having
adjacent *& cancel out (similarly, &*). thus
******f
******&f
*****f
*****&f
****f
****&f
***f
etc (i might have got the wrong number of * at some points)
little fleas indeed!
