Fix bug: DOUBLE_TO_FIXNUM_P was using limits that assumed the

conversion used rounding; in fact, it uses truncation.

diff --git a/v7/src/microcode/object.h b/v7/src/microcode/object.h
index 2d077446ccddc756eddb2dc6e7856a3ff3427510..0e37499c941390a632587585c9e108320a9cf6c6 100644 (file)
--- a/v7/src/microcode/object.h
+++ b/v7/src/microcode/object.h
@@ -1,6 +1,6 @@
 /* -*-C-*-
 
-$Id: object.h,v 9.54 2003/02/28 04:34:38 cph Exp $
+$Id: object.h,v 9.55 2003/11/26 05:01:25 cph Exp $
 
 Copyright 1986,1987,1988,1989,1990,1992 Massachusetts Institute of Technology
 Copyright 1993,1995,1997,1998,2000,2001 Massachusetts Institute of Technology
@@ -399,8 +399,8 @@ extern SCHEME_OBJECT * memory_base;
 #define FIXNUM_TO_DOUBLE(fixnum) ((double) (FIXNUM_TO_LONG (fixnum)))
 
 #define DOUBLE_TO_FIXNUM_P(number)                                     \
-  (((number) > (((double) SMALLEST_FIXNUM) - 0.5)) &&                  \
-   ((number) < (((double) BIGGEST_FIXNUM) + 0.5)))
+  (((number) > ((double) (SMALLEST_FIXNUM - 1)))                       \
+   && ((number) < ((double) (BIGGEST_FIXNUM + 1))))
 
 #ifdef HAVE_DOUBLE_TO_LONG_BUG
 #define DOUBLE_TO_FIXNUM double_to_fixnum