From: Holger Schemel <info@artsoft.org>
Date: Sun, 19 Feb 2023 00:30:03 +0000 (+0100)
Subject: fixed bug that could cause an endless loop (freeze) in the MM engine
X-Git-Tag: 4.3.5.1^0
X-Git-Url: https://git.artsoft.org/rocksndiamonds.git/?a=commitdiff_plain;h=443ff08ee0c81730c4ff198b16bafc4199696724;hp=487e1b8ecea9121a3408e5d4f968945e071aa4aa;p=rocksndiamonds.git

fixed bug that could cause an endless loop (freeze) in the MM engine

This is a workaround for a regression bug of commit f8aec447 (that
fixed overloading the laser due to duplicated tiles in damage list).

The problem that could cause an endless loop already existed before
the above commit, but without that commit, scanning the laser was
stopped by the damage list completely filling up, stopping the loop.
The above commit prevents adding the same tile again and again to the
damage list, therefore not stopping the endless loop anymore.

This commit should prevent the endless loop from happening at all.

The problem (and how to reproduce it) was described by Eizzoux in the
following post in the R'n'D web forum:
https://www.artsoft.org/forum/viewtopic.php?p=18214
---

diff --git a/src/game_mm/mm_game.c b/src/game_mm/mm_game.c
index 9a7ad5e6..2805fb77 100644
--- a/src/game_mm/mm_game.c
+++ b/src/game_mm/mm_game.c
@@ -1496,6 +1496,8 @@ boolean HitElement(int element, int hit_mask)
   // this is more precise: check if laser would go through the center
   if ((ELX * TILEX + 14 - LX) * YS != (ELY * TILEY + 14 - LY) * XS)
   {
+    int skip_count = 0;
+
     // prevent cutting through laser emitter with laser beam
     if (IS_LASER(element))
       return TRUE;
@@ -1505,15 +1507,18 @@ boolean HitElement(int element, int hit_mask)
     {
       LX += XS;
       LY += YS;
+
+      skip_count++;
     }
     while (ELX == LX/TILEX && ELY == LY/TILEY && LX > 0 && LY > 0);
 
-    if (LX/TILEX > ELX || LY/TILEY > ELY)
+    if ((LX/TILEX > ELX || LY/TILEY > ELY) && skip_count > 1)
     {
       /* skipping scan positions to the right and down skips one scan
 	 position too much, because this is only the top left scan position
 	 of totally four scan positions (plus one to the right, one to the
 	 bottom and one to the bottom right) */
+      /* ... but only roll back scan position if more than one step done */
 
       LX -= XS;
       LY -= YS;