OpenJ9-17 Incorrectly calculates the hashcode value for a String about openj9

[pshipton]

jdk8, jdk11 and jdk21 give the expected result.

@tajila fyi

[tajila]

@theresa-m Please take a look at this.

[theresa-m]

In OpenJ9 Java 11/17/21 all use different variations of the java.lang.String library.

• 11 uses the custom String library in OpenJ9 https://github.com/eclipse-openj9/openj9/blob/master/jcl/src/java.base/share/classes/java/lang/String.java
• 17 uses the String library from OpenJDK with modifications https://github.com/ibmruntimes/openj9-openjdk-jdk17/blob/openj9/src/java.base/share/classes/java/lang/String.java
• 21 also uses the modified OpenJDK String library but does not experience this problem because StringBuilder.toString goes through a different String constructor

The error in 17 is caused due to a missing call to initCompressionFlag(); when the string is created through StringBuilder. When String.hashCode is called the uninitialized compression flag causes hashCodeImplCompressed to be called instead of hashCodeImplDecompressed.

initCompressionFlag(); should be called in this String constructor https://github.com/ibmruntimes/openj9-openjdk-jdk17/blob/openj9/src/java.base/share/classes/java/lang/String.java#L5309-L5312

I will also apply the change to 21 and 22 once 17 is approved because even though its not seeing this failure it is possible that it could occur from other calls to this constructor.

[theresa-m]

Java 17 ibmruntimes/openj9-openjdk-jdk17#335
Java 21 ibmruntimes/openj9-openjdk-jdk21#138
Java 22 ibmruntimes/openj9-openjdk-jdk22#37
Java next ibmruntimes/openj9-openjdk-jdk#766