IJ09575: INCORRECT RESULT WHEN USING JAVA.LANG.MATH.MIN OR MAX ON 31-BIT JVM

APAR status

• Closed as program error.

Error description

• Error Message: The java.lang.Math.min or java.lang.Math.max may
  give an incorrect result on 31-bit JVM on z/OS or Linux on Z
  with some very special values.
  .
  Stack Trace: Java callstack:
      at java/math/BigDecimal.powerOfTenBI(BigDecimal.java:7902)
      at java/math/BigDecimal.setScale(BigDecimal.java:5156)
      at java/math/BigDecimal.setScale(BigDecimal.java:5019)
      at
  java/math/BigDecimal.divideToIntegralValue(BigDecimal.java:2680(
  Compiled Code))
      at
  java/math/BigDecimal.divideAndRemainder(BigDecimal.java:2612)
  .
  

Local fix

• This issue can be worked around using the
  -Xjit:disableMaxMinOptimization option.
  

Problem summary

• The JIT compiler was in some very special cases not handingling
  values in register pairs correctly on 31-bit JVM on z/OS or
  Linux on Z when intrinsically transforming
  java.lang.Math.min/max calls into hardware intrinsic
  instructions.
  

Problem conclusion

• The JIT compiler has been updated to fix the problem.
  .
  This APAR will be fixed in the following Java Releases:
     8    SR5 FP21  (8.0.5.21)
  .
  Contact your IBM Product's Service Team for these Service
  Refreshes and Fix Packs.
  For those running stand-alone, information about the available
  Service Refreshes and Fix Packs can be found at:
             https://www.ibm.com/developerworks/java/jdk/
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  JIT

• Fixed component ID

  620700124

Applicable component levels