From: H.G.Muller <hgm@hgm-xboard.(none)>
Date: Tue, 20 Oct 2015 15:30:34 +0000 (+0200)
Subject: Fix sortng of mate scores
X-Git-Url: http://winboard.nl/cgi-bin?p=xboard.git;a=commitdiff_plain;h=f49e7e17e3b4e66fb70f0a9416ed65b5e4516d7d

Fix sortng of mate scores

With the new convention for mate scores (100000+DTM), lower numbers
represent faster mates and thus better scores, and the sorting of PVs
in the Engine Output window now takes that into account.
---

diff --git a/engineoutput.c b/engineoutput.c
index d6fde3c..0bfd8b5 100644
--- a/engineoutput.c
+++ b/engineoutput.c
@@ -402,6 +402,14 @@ SetEngineColorIcon (int which)
 // [HGM] multivar: sort Thinking Output within one depth on score
 
 static int
+MateFlip (int n)
+{   // map mate-score to monotonous scale, so sorting compares them correctly
+    if(n >=  MATE_SCORE) return 2*MATE_SCORE - n;
+    if(n <= -MATE_SCORE) return -2*MATE_SCORE - n;
+    return n;
+}
+
+static int
 InsertionPoint (int len, EngineOutputData *ed)
 {
 	int i, offs = 0, newScore = ed->score, n = ed->which;
@@ -425,7 +433,7 @@ InsertionPoint (int len, EngineOutputData *ed)
 		keys[i+n+2] = ed->moveKey;
 		fail[i+n+2] = failType;
 		if(ed->moveKey != keys[i+n] && // same move always tops previous one (as a higher score must be a fail low)
-		   newScore < scores[i+n] && fail[i+n] == ' ') break;
+		   MateFlip(newScore) < MateFlip(scores[i+n]) && fail[i+n] == ' ') break;
 		// if it had higher score as previous, move previous in stead
 		scores[i+n+2] = ed->moveKey == keys[i+n] ? newScore : scores[i+n]; // correct scores of fail-low/high searches
 		textEnd[i+n+2] = textEnd[i+n] + len;