RFR for 8043740 (Doubles with large exponents overflow to Infinity incorrectly) (original) (raw)
Sandipan Razzaque me at sandipan.net
Mon Jun 2 14:08:06 UTC 2014
- Previous message: RFR for 8043740 (Doubles with large exponents overflow to Infinity incorrectly)
- Next message: RFR: 5043030 (reflect) unnecessary object creation in reflection
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hi all -
I've made a quick revision to that last patch. Please find inline the latest link + patch.
http://www.sandipan.net/public/webrevs/8043740/webrev.01/
Cheers, SR
--- old/src/share/classes/sun/misc/FloatingDecimal.java 2014-06-02 09:32:20.000000000 -0400 +++ new/src/share/classes/sun/misc/FloatingDecimal.java 2014-06-02 09:32:19.000000000 -0400 @@ -1994,19 +1994,29 @@ } int expLimit = BIG_DECIMAL_EXPONENT+nDigits+nTrailZero; if ( expOverflow || ( expVal > expLimit ) ){
//
// The intent here is to end up with
// infinity or zero, as appropriate.
// The reason for yielding such a small decExponent,
// rather than something intuitive such as
// expSign*Integer.MAX_VALUE, is that this value
// is subject to further manipulation in
// doubleValue() and floatValue(), and I don't want
// it to be able to cause overflow there!
// (The only way we can get into trouble here is for
// really outrageous nDigits+nTrailZero, such as
2 billion. )
//
decExp = expSign*expLimit;
- // There is still a chance that the exponent will be
- // safe to use: if it would eventually decrease
- // due to a negative decExp, and that number is below the limit.
- // We check for that here.
- if ( ( expSign == 1 && decExp < 0 )
- && ( expVal + decExp ) < expLimit ) {
- // Cannot overflow - adding a positive and negative number.
- decExp = expVal + decExp;
- } else {
- //
- // The intent here is to end up with
- // infinity or zero, as appropriate.
- // The reason for yielding such a small decExponent,
- // rather than something intuitive such as
- // expSign*Integer.MAX_VALUE, is that this value
- // is subject to further manipulation in
- // doubleValue() and floatValue(), and I don't want
- // it to be able to cause overflow there!
- // (The only way we can get into trouble here is for
- // really outrageous nDigits+nTrailZero, such as 2 billion. )
- //
- decExp = expSign*expLimit;
- } } else { // this should not overflow, since we tested // for expVal > (MAX+N), where N >= abs(decExp)
--- /dev/null 2014-06-02 09:32:21.000000000 -0400 +++ new/test/java/lang/Double/Bug8043740.java 2014-06-02 09:32:20.000000000 -0400 @@ -0,0 +1,41 @@ +/*
- Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
- DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- This code is free software; you can redistribute it and/or modify it
- under the terms of the GNU General Public License version 2 only, as
- published by the Free Software Foundation.
- This code is distributed in the hope that it will be useful, but WITHOUT
- ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
- version 2 for more details (a copy is included in the LICENSE file that
- accompanied this code).
- You should have received a copy of the GNU General Public License version
- 2 along with this work; if not, write to the Free Software Foundation,
- Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- or visit www.oracle.com if you need additional information or have any
- questions.
- */
- +/*
- @test
- @bug 8043740
- @summary Test for Double.parseDouble incorrect normalization of exponent.
- */ +public class Bug8043740 {
- public static void main(String[] args) {
String text = "00.000000000000000001e326";
- double expected = Double.parseDouble("1e308");
double value = Double.parseDouble(text);
if (Double.compare(expected, value) != 0) {
throw new RuntimeException("String [" + text +
"] was incorrectly parsed to [" + value + "]");
}
- }
- +}
Sandipan Razzaque | www.sandipan.net
On Sun, Jun 1, 2014 at 10:55 PM, Sandipan Razzaque <me at sandipan.net> wrote:
Hi Brian/all,
Fix for the above-mentioned bug is complete. If you (or anyone else as appropriate) could please review the changes and let me know of your comments that would be great! Link for the webrev: http://www.sandipan.net/public/webrevs/8043740/webrev.00/ The patch is inlined below my signature. Cheers, SR
--- old/src/share/classes/sun/misc/FloatingDecimal.java 2014-06-01 22:33:34.947651253 -0400 +++ new/src/share/classes/sun/misc/FloatingDecimal.java 2014-06-01 22:33:34.827651251 -0400 @@ -1992,7 +1992,10 @@ break expLoop; // stop parsing exponent. } } - int expLimit = BIGDECIMALEXPONENT+nDigits+nTrailZero; + + int expLimit = BIGDECIMALEXPONENT + nDigits + nTrailZero + - (nLeadZero - decPt); + if ( expOverflow || ( expVal > expLimit ) ){ // // The intent here is to end up with --- /dev/null 2014-05-28 21:22:08.061671015 -0400 +++ new/test/java/lang/Double/Bug8043740.java 2014-06-01 22:33:35.227651260 -0400 @@ -0,0 +1,40 @@ +/* + * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + * @test + * @bug 8043740 + * @summary Test for Double.parseDouble incorrect normalization of exponent. + */ +public class Bug8043740 { + + public static void main(String[] args) { + String text326 = "00.000000000000000001e326"; // expect 1e308 + double value326 = Double.parseDouble(text326); + if (Double.isInfinite(value326)) { + throw new RuntimeException("String [" + text326 + + "] was incorrectly parsed to Infinity."); + } + } + +}
- Previous message: RFR for 8043740 (Doubles with large exponents overflow to Infinity incorrectly)
- Next message: RFR: 5043030 (reflect) unnecessary object creation in reflection
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]