From: matz@... (Yukihiro Matsumoto)
Date: 2001-09-08T22:46:50+09:00
Subject: [ruby-talk:20998] Re: Bug in latest snapshot, regarding Float#to_s

Hi,

In message "[ruby-talk:20994] Bug in latest snapshot, regarding Float#to_s"
    on 01/09/08, "Aristarkh A Zagorodnikov" <xm@w3d.ru> writes:

|	Recently I installed the latest snapshot (ruby 1.7.1 (2001-09-03)
|[i686-linux]), and after some time of work found that it has a bug. After 2
|hours of tracing the bug (all started from random Apache hangs), I traced it
|down to Float#to_s and discovered that "ruby -e 'p 0.0'" puts computer in
|and endless loop. The following patch to numeric.c (by no means said to be
|100% correct or extensively tested) fixed the situation for me.

Here's my patch:

diff -u -1 -r1.30 numeric.c
--- numeric.c	2001/09/03 05:37:42	1.30
+++ numeric.c	2001/09/08 13:46:23
@@ -218,3 +218,3 @@
     double value = RFLOAT(flt)->value;
-    double d1, d2;
+    double avalue, d1, d2;
 
@@ -225,4 +225,8 @@
     
-    if (value < 1.0e-3) {
-	d1 = value;
+    avalue = fabs(value);
+    if (avalue == 0.0) {
+	fmt = "%.1f";
+    }
+    else if (avalue < 1.0e-3) {
+	d1 = avalue;
 	while (d1 < 1.0) d1 *= 10.0;
@@ -231,4 +235,4 @@
     }    
-    else if (value >= 1.0e10) {
-	d1 = value;
+    else if (avalue >= 1.0e10) {
+	d1 = avalue;
 	while (d1 > 10.0) d1 /= 10.0;