From: Matthew Fernandez <matthew.fernandez@gmail.com>
Date: Sun, 28 Aug 2022 20:11:26 +0000 (-0700)
Subject: sfdpgen check_real_array_size: abbreviate no-op 'MAX' call
X-Git-Tag: 6.0.1~14^2~5
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=1c5e605e590409a77a835db1d57bd3b669442ebe;p=graphviz

sfdpgen check_real_array_size: abbreviate no-op 'MAX' call

1b847b5abf50d2ab0701523b90a20306d0ee5528 seems to have incorrectly assumed
multiplication has a higher precedence than casts. In reality, it is the
opposite, meaning the first parameter to this `MAX` call was always 0. So the
entire expression would evaluate to 10. Empirically this seems to have been fine
since this code has been in use for over a decade with no specific problems
blamed on this area. So lets just abbreviate it into what it evaluates to.

Gitlab: #2269
---

diff --git a/lib/sfdpgen/spring_electrical.c b/lib/sfdpgen/spring_electrical.c
index e316c46d6..e7e87cd20 100644
--- a/lib/sfdpgen/spring_electrical.c
+++ b/lib/sfdpgen/spring_electrical.c
@@ -308,7 +308,7 @@ static double update_step(int adaptive_cooling, double step, double Fnorm, doubl
 
 static void check_real_array_size(double **a, int len, int *lenmax){
   if (len >= *lenmax){
-    *lenmax = len + MAX((int) 0.2*len, 10);
+    *lenmax = len + 10;
     *a = REALLOC(*a, sizeof(double)*(*lenmax));
   }