[17][Pattern Switch] NPE not thrown for a null with default, (non-total) type pattern case labels

[Brian Goetz]

Agree, this should NPE.

On 8/20/2021 7:37 AM, Manoj Palat wrote:
> Hi,
>  Consider the following case:
> class X {
> public void foo(Object o) {
>          switch (o) {
>            default: break;
>            case String s :System.out.println(s); break;
>          }
> }
> public static void main(String[] args) {
>          (new X()).foo(null); // NPE expected, but prints null
> }
> }
> javac (17+ea35)  prints a null rather than an NPE. I believe it should 
> throw NullPointerException (rather  javac should generate code for 
> Null check to raise NPE at run-time).
> Reasoning behind this is [based on 
> https://cr.openjdk.java.net/~gbierman/jep406/jep406-20210608/specs/patterns-switch-jls.html#jls-14.11.3 
> <https://cr.openjdk.java.net/~gbierman/jep406/jep406-20210608/specs/patterns-switch-jls.html#jls-14.11.3>]:
> section 14.11.3, see the second rule:
> ...
> Otherwise, if the result of evaluating the selector expression is null, and no switch label in the resolved switch block applies then a NullPointerException is thrown...
> I don't think either of the switch labels apply here
> [based on sec 14.11.1
> If the value is the null reference, then we determine the first (if any) switch label in the switch block that applies to the value as follows:
>
> A switch label...]
> Hence I believe an NPE should be thrown for null.
> [A longer version of reasoning at: 
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=575051#c4 
> <https://bugs.eclipse.org/bugs/show_bug.cgi?id=575051#c4>]
> Please let me know your take on this.
> Regards,
> Manoj
> [Note: If the case labels are swapped, ie if its in the order shown 
> below, NPE is thrown at run-time
> case String s :System.out.println(s); break;
> default: break; ]
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.java.net/pipermail/compiler-dev/attachments/20210820/b0e999bf/attachment.htm>