Stack traps

A chip has been running flawlessly for hours, then it suddenly stops with StackOverFlow. Nobody changed anything. The bug was in the code from the start, it just needed 512 ticks to show.

StackOverFlow and StackUnderFlow#

The classic case: one pass puts three values onto the stack but only takes two off. Every tick sp grows by one. After a little over 500 ticks, a little over four minutes, the stack is full.

push a
push b
push c
pop x
pop y          # the third pop is missing: sp grows every tick

The rule: in every pass as many pop as push. If you are unsure, put move sp 0 at the start of the main loop. Then every pass begins with an empty pile, even if a pop is missing somewhere.

Overwriting sp by accident#

sp is a perfectly normal register, number 16. Every instruction that writes into it moves the pile: move sp 7, add sp sp 1, an alias on r16. Afterwards pop reads from a wrong place, without an error message. A register plan with alias only for r0 to r15 protects against that.

Memory after a restart#

In the game code, uploading new code and re-inserting the chip only set sp and the program line to 0. Registers and stack contents stay. They are stored in the save as well.

For the stack that means: after a restart sp is 0, but the addresses still hold the old values. A table you fill with poke is back immediately. A counter you keep in an address carries on with the old value. If you want a clean start, clear with clr db before the start label.