From: Mattias Engdegård <mattiase@acm.org>
Date: Thu, 4 Jul 2024 15:48:16 +0000 (+0200)
Subject: ; * src/fns.c (fixnum_float_cmp): Explain argument constraint.
X-Git-Url: http://git.eshelyaron.com/gitweb/?a=commitdiff_plain;h=5906fc6dc9835240e48591e23e3763d8fd98cb5b;p=emacs.git

; * src/fns.c (fixnum_float_cmp): Explain argument constraint.

Suggested by Pip Cet.

(cherry picked from commit fa6f088a483f1f2e19863800c6215a8136288b8f)
---

diff --git a/src/fns.c b/src/fns.c
index 07afcada62f..22947b2649f 100644
--- a/src/fns.c
+++ b/src/fns.c
@@ -3012,16 +3012,18 @@ bool_vector_cmp (Lisp_Object a, Lisp_Object b)
   return (d & aw) ? 1 : -1;
 }
 
-/* Return -1 if a<b, 1 if a>b, 0 if a=b or if b is NaN.  */
+/* Return -1 if a<b, 1 if a>b, 0 if a=b or if b is NaN (a must be a fixnum).  */
 static inline int
 fixnum_float_cmp (EMACS_INT a, double b)
 {
   double fa = (double)a;
   if (fa == b)
     {
-      /* This doesn't mean that a=b because the conversion may have
-	 rounded, but b must be an integer that fits in an EMACS_INT
-	 and we can compare in the integer domain instead.  */
+      /* This doesn't mean that a=b because the conversion may have rounded.
+	 However, b must be an integer that fits in an EMACS_INT,
+	 because |b| ≤ 2|a| and EMACS_INT has at least one bit more than
+	 needed to represent any fixnum.
+	 Thus we can compare in the integer domain instead.  */
       EMACS_INT ib = b;		/* lossless conversion */
       return a < ib ? -1 : a > ib;
     }