FULL PRODUCT VERSION :
java version "1.8.0"
Java(TM) SE Runtime Environment (build 1.8.0-b132)
Java HotSpot(TM) 64-Bit Server VM (build 25.0-b70, mixed mode)


ADDITIONAL OS VERSION INFORMATION :
RHEL 6.4 (but that's irrelevant)

A DESCRIPTION OF THE PROBLEM :
The code for the method hash() does not match the comment and contains a simple bug.

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Look at the source code for the method hash() in class java.util.IdentityHashMap.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
Should calculate (-127*hashCode)&(length-1)
ACTUAL -
Actually calculates (-254*hashCode)&(length-1)

REPRODUCIBILITY :
This bug can be reproduced always.

---------- BEGIN SOURCE ----------
Here is the source for the method hash():


    private static int hash(Object x, int length) {
        int h = System.identityHashCode(x);
        // Multiply by -127, and left-shift to use least bit as part of hash
        return ((h << 1) - (h << 8)) & (length - 1);
    }


Note that it says "multiply by -127", but the effect of the operation ((h<<1)-(h<<8)) is in fact "multiply by -254".

This is significant because it means that odd-numbered buckets are simply never accessible, so average occupancy in buckets will be nearly 2 by the time of a resize.

Also the implementation critically depends on the assumption that identityHashCode() produces a well-distributed range of values. I don't know that this is really the case.
---------- END SOURCE ----------

CUSTOMER SUBMITTED WORKAROUND :
Try modifying the source to say (-127*h)&(length-1). This is correct, not obfuscated and will probably produce better code anyway as it allows the JIT to take a crack at the code.