From mboxrd@z Thu Jan  1 00:00:00 1970
From: apbianco@redhat.com
To: gcc-gnats@gcc.gnu.org
Subject: libgcj/4583: problems BigDecimal(double) ctor.
Date: Tue, 16 Oct 2001 13:26:00 -0000
Message-id: <20011016201938.25319.qmail@sourceware.cygnus.com>
X-SW-Source: 2001-10/msg00333.html
List-Id: <gcc-prs.sourceware.org>

>Number:         4583
>Category:       libgcj
>Synopsis:       problems BigDecimal(double) ctor.
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Oct 16 13:26:01 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator:     Alex Petit-Bianco
>Release:        unknown-1.0
>Organization:
>Environment:
linux/x86, current trunk.
>Description:
I'm running into the following situation with BigDecimal. Consider
this code:

  private static final BigDecimal minDouble =
    new java.math.BigDecimal(Double.MIN_VALUE);

It ends up throwing a java.lang.NumberFormatException.

It seems to be that Long.parseLong doesn't handle what it is given as
an argument (`50E-324', and `scale' in BigDecimal would be modified
accordingly) as Long.parseLong won't handle anything but digits, and
the exception is thrown from there:

  private static long parseLong(String str, int index, int len, boolean isNeg,
                                int radix) throws NumberFormatException
  {
    ...
        if ((digval = Character.digit(str.charAt(index), radix)) < 0)
          throw new NumberFormatException();

Long.parseLong() complies with the spec, but BigDecimal uses
Double.toString to convert the input number. If the input can't be
represented in a string without using the scientific notation,
Long.parseLong() won't receive something it can deal with. 4.0e-4
instead of Double.MIN_VALUE will trigger the same error; but 4.0e-3
won't.

The attached test case can be use and could be added to our testsuite.
>How-To-Repeat:
Build BIG.java.
>Fix:
Not yet. One can work around the problem by hand crafting
a string representation of Double.MIN_VALUE or Double.MAX_VALUE that doesn't use a scientific notation.

>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: text/x-java; name="BIG.java"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="BIG.java"

aW1wb3J0IGphdmEubWF0aC5CaWdEZWNpbWFsOwppbXBvcnQgamF2YS5tYXRoLkJpZ0ludGVnZXI7
CgpjbGFzcyBCSUcgewogIHByaXZhdGUgc3RhdGljIGZpbmFsIEJpZ0ludGVnZXIgbWluRG91Ymxl
ID0KICAgIG5ldyBqYXZhLm1hdGguQmlnRGVjaW1hbChEb3VibGUuTUlOX1ZBTFVFKS50b0JpZ0lu
dGVnZXIoKTsKICBwcml2YXRlIHN0YXRpYyBmaW5hbCBCaWdJbnRlZ2VyIG1heERvdWJsZSA9CiAg
ICBuZXcgamF2YS5tYXRoLkJpZ0RlY2ltYWwoRG91YmxlLk1BWF9WQUxVRSkudG9CaWdJbnRlZ2Vy
KCk7CiAgcHVibGljIHN0YXRpYyB2b2lkIG1haW4gKFN0cmluZ1tdIGFyZykKICB7CiAgfQp9Cg==