Community
Participate
Working Groups
import java.util.Collections; import java.util.List; import org.eclipse.jdt.annotation.NonNull; public class C { void loop() { List<String> list = Collections.emptyList(); for(@NonNull String s: list) { expectNonNull(s); } } void expectNonNull(@NonNull String s) {} } gives Potential type mismatch: required '@NonNull String' but nullness of the provided value is unknown on expectNonNull(s);
It would appear the right behavior is for this warning to not show up. But there should be a warning in the preceding line since there is no syntax to annotate a collection as containing only @NonNull elements and so as the cursor advances there is no non null guarantee that can be enforced by the null annotation processor.
The other construct that allows a local to be declared works alright: //-------------- import org.eclipse.jdt.annotation.NonNull; public class X implements AutoCloseable { @NonNull X[] x = null; public static void main(String[] args) { try (@NonNull X x = boo()) { goo(x); } } static X boo() { return null; } static void goo(@NonNull X x) { } @Override public void close() { } }
(In reply to comment #1) > It would appear the right behavior is for this warning to not show > up. But there should be a warning in the preceding line since there > is no syntax to annotate a collection as containing only @NonNull > elements and so as the cursor advances there is no non null guarantee > that can be enforced by the null annotation processor. I agree. Should be easy to fix for M6
Fixed for 3.8 M6 via commit d137239e9d64b43b4573cdfef965a2c47040a54e Normally, the annotation on the local variable declaration is analyzed as part of the initialization/assignment. The bug happened because in this particular case we have no AST representing an assignment to the local variable. As a side effect of fixing the bug, we now see another warning against "for(@NonNull String s: list)", where list is a 'List<String>': Potential type mismatch: required '@NonNull String' but nullness of the provided value is unknown. Until we have JSR 308 we cannot say 'List<@NonNull String>' so the elements have the legacy type String, which is not safely assignable to '@NonNull String'. Just mentioning so nobody is surprised about the new outcome.
Verified for 3.8 M6 using Build id: I20120306-0800
