> The problem is that makes the stack consumption of END_FINALLY variable.
> How about always consuming 6 items, even when most of them are unused
> padding?

Right. Though the value of the stack increment makes sense only for the case 
1, because in other cases the next executed instruction is not the one 
following END_FINALLY, but we need to reserve the stack for the case 3, and 
counting additional "virtual" items  in PUSH_NO_EXCEPT and END_FINALLY will 
help to calculate the maximal stack consumption, even when in runtime they 
push/pop only one item. This is yet one argument for special purposed opcode 
PUSH_NO_EXCEPT (maybe I'll rename it to START_FINALLY or BEGIN_FINALLY). Other 
reason -- this will help to determine bounds of finally blocks in the 
frame.f_lineno setter.