On Jun 24, 2014, at 2:42 AM, Mike Duigou <mike.dui...@oracle.com> wrote:
> Hello all;
>
> This changeset corrects a reported problem with the lists returned by
> Collections.checkedList(). Since Java 8 the replaceAll() method on checked
> lists has erroneously allowed the operator providing replacements to provide
> illegal replacement values which are then stored, unchecked into the wrapped
> list.
>
> This changeset adds a check on the proposed replacement value and throws a
> ClassCastException if the replacement value is incompatible with the list.
That seems like a reasonable approach and it's in sync with the behaviour of
Map.replaceAll.
> Additionally the javadoc is updated to inform users that a ClassCastException
> may result if the proposed replacement is unacceptable.
>
No users will see the JavaDoc on Collections.CheckedList since it is package
private, plus i think it redundant. Any such associated documentation would
need to be on the public static method, and i am not sure we really need to say
anything more than what is already said:
3388 * Any attempt to insert an element of the wrong type will result in
3389 * an immediate {@link ClassCastException}. Assuming a list contains
> Note that this changeset takes the strategy of failing when the illegal value
> is encountered. Replacements of earlier items in the list are retained.
>
> jbsbug: https://bugs.openjdk.java.net/browse/JDK-8047795
> webrev: http://cr.openjdk.java.net/~mduigou/JDK-8047795/0/webrev/
>
Are there existing tests for the checked Map.replaceAll (for keys and values)?
Paul.
> This change will be backported to Java 8.
>
> Mike
