By deleting items moving forward, you're cutting the branch off that you're standing on. :-) The upper bounds of the loop is only evaluated once, before the loop begins, and if you delete items there are now fewer in the list than there were when the bound was calculated.
- Loop limit is evaluated (for example,
List.Count - 1 = 5). Valid indexes into it are [0..4]
- The loop starts, and you retrieve List[0] and delete it. List Count = 4,
bounds is still 5
- The index is incremented, you retrieve and delete List[1]. List Count = 3, bounds is still 5
- The index is incremented, you retrieve and delete List[2]. List Count = 2, bounds is still 5.
- The index is incremented, you retrieve List[3] - Oops! There are only 2 items in the list, now at indexes [0..1] - List index out of bounds(3).
By iterating backwards, even though the bounds is still only calculated at the beginning, you're removing the items from the end and decrementing the count at the same time.
- Bounds is 5, and you retrieve List[4] and delete it. Count is now 4, bounds is still 5
- Index is decremented, and you retrieve List[3] and delete it. Count is now 3, bounds is still 5
- Index is decremented, and you retrieve List[2] and delete it. Count is now 2, bounds is still 5.
- Index is decremented, and you retrieve and delete List[1]. Count is now 1, bounds is still 5.
- Index is decremented, and you retrieve and delete List[0]. List is now empty, but we've reached the terminating condition of the loop (
downto 0) and the loop exits safely.