From: PatR <rankin@nethack.org>
Date: Thu, 27 Oct 2022 22:55:57 +0000 (-0700)
Subject: more knockback induced dismount
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=9682f84aedaeec5a4f86cd62f7ea7ad414670f31;p=nethack

more knockback induced dismount

When looking at the previous commit, I realized that my comment about
radius 1 was wrong.  The original code prefers dismounting to spots
that are orthogonal to the steed's position over diagonal ones.  It
doesn't say why.
---

diff --git a/src/steed.c b/src/steed.c
index 72114dc18..4c3528029 100644
--- a/src/steed.c
+++ b/src/steed.c
@@ -512,11 +512,11 @@ landing_spot(
                 ++viable;
                 distance = distu(x, y);
                 if (min_distance < 0 /* no viable candidate yet */
-                    /* or better than pending candidate (note: distance
-                       is never less than min_distance because we're
-                       limiting search to radius 1; j==0 won't get here
-                       because 'min_distance < 0' will always pass for it) */
-                    || (distance < min_distance || (best_j != -1 && j == 0))
+                    /* or better than pending candidate (note: orthogonal
+                       spots are distance 1 and diagonal ones distance 2;
+                       treating one as better than the other is arbitary
+                       and not wanted for DISMOUNT_KNOCKED) */
+                    || (best_j == -1) ? (distance < min_distance) : (j < 3)
                     /* or equally good, maybe substitute this one */
                     || (distance == min_distance && !rn2(viable))) {
                     /* traps avoided on pass 0; boulders avoided on 0 and 1 */