[MATH-721] building common-math on Solaris SPARC gives "error: floating point number too small"

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.2
• Fix Version/s: 3.1

Environment

Build environment is:
mvn --version
Apache Maven 3.0.3 (rNON-CANONICAL_2011-10-13_22-23_builder; 2011-10-13 22:23:49+0200)
Maven home: /usr/local/share/maven-3.0.3
Java version: 1.7.0-internal, vendor: Oracle Corporation
Java home: /usr/local/share/jdk1.7.0_1/jre
Default locale: de_DE, platform encoding: UTF-8
OS name: "sunos", version: "5.10", arch: "sparc", family: "unix"

• Assignee: Unassigned
• Reporter: Jörg Prante
• Created: 08/Dec/11 23:51
• Updated: 04/Mar/13 18:53
• Resolved: 21/Apr/12 12:16

Description

It seems the assumption in MathUtils.java

/** Safe minimum, such that 1 / SAFE_MIN does not overflow.

• <p>In IEEE 754 arithmetic, this is also the smallest normalized
• number 2^-1022.</p>
  */
  public static final double SAFE_MIN = 0x1.0p-1022;

does not work on my openjdk 1.7.0_1 on Solaris SPARC, because I get the following build error:

[INFO] Compiling 457 source files to /opt/builder/projects/commons-math-2.2-src/target/classes
[INFO] -------------------------------------------------------------
[ERROR] COMPILATION ERROR :
[INFO] -------------------------------------------------------------
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/main/java/org/apache/commons/math/util/MathUtils.java:[42,42] error: floating point number too small

[INFO] 1error
[INFO] -------------------------------------------------------------
[INFO] ------------------------------------------------------------------------
[INFO] BUILD FAILURE

and also in the tests I encounter the error when parsing floating point constants

[ERROR] Failed to execute goal org.apache.maven.plugins:maven-compiler-plugin:2.1:testCompile (default-testCompile) on project commons-math: Compilation failure: Compilation failure:
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1050,69] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1055,28] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1055,69] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1056,28] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1062,70] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1063,70] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1068,70] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/util/FastMathTest.java:[1069,70] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/optimization/general/LevenbergMarquardtOptimizerTest.java:[503,51] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/optimization/general/LevenbergMarquardtOptimizerTest.java:[503,66] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/estimation/LevenbergMarquardtEstimatorTest.java:[583,52] error: floating point number too small
[ERROR]
[ERROR] /opt/builder/projects/commons-math-2.2-src/src/test/java/org/apache/commons/math/estimation/LevenbergMarquardtEstimatorTest.java:[585,52] error: floating point number too small

I suggest using java.lang.Double.MIN_NORMAL for a platform normalized minimal floating point value.

A quick program for printing Double.MIN_NORMAL gives the following result.

class test {
public static void main(String[] args)

{ System.out.println("min float = " + Double.MIN_NORMAL ); }

}

java test
min float = 2.2250738585072014E-308

Could it be an openjdk glitch? Thank you for any comments and help how this could be fixed.