Repository navigation
JSpecify: False positive error for nullable source with nullable-param non-null-return method #4086
Description
Expected behavior
When a @Nullable source property is mapped to a @NonNull constructor parameter via a reused mapping method that accepts @Nullable (explicitly handles null) and returns a non-null value, the mapping should compile successfully. The reused method's parameter and return type annotations prove the mapping is safe - null is handled and the result is guaranteed non-null.
Given:
@Mapper
public interface Issue4086Mapper {
@Mapping(target = "payload", source = "value", qualifiedByName = "mapValue")
Target map(Source source);
@Named("mapValue")
default Target.Nested mapValue(@Nullable String value) {
return value == null ? new Target.Nested( "default" ) : new Target.Nested( value );
}
}
public class Source {
private String value;
public @Nullable String getValue() { return value; }
}
public class Target {
private final Nested payload;
public Target(@NonNull Nested payload) { this.payload = payload; }
}Expected: compilation succeeds with:
public Target map(Source source) {
Target.Nested payload = mapValue( source.getValue() );
Target target = new Target( payload );
return target;
}Actual behavior
Compilation fails with:
error: Can't map potentially nullable source property "value" to @NonNull
constructor parameter "payload". Consider adding a defaultValue or defaultExpression.
The current check only inspects the source getter chain nullability, which reports NULLABLE. It does not consider that the assignment was resolved through a reused method whose parameter is @Nullable (no null guard needed before the call) and whose return type is @NonNull (the result is guaranteed non-null).
Steps to reproduce the problem
A @Nullable source property mapped to a @NonNull constructor parameter
via qualifiedByName referencing a method with @Nullable parameter and non-null return. Can be reproduced in this branch: main...RaniAgus:mapstruct:issues/4086
MapStruct Version
1.7.0.Beta2
Activity
tests: reproduce mapstruct#4086
mapstruct#4086: Check nullable status of methode return types
tests: reproduce mapstruct#4086
mapstruct#4086: Check nullable status of methode return types
r0h1tb commentedon Aug 1, 2026
I had a go at this and got as far as pinning the root cause, so posting it in case it's useful to whoever picks it up.
The error is raised in
PropertyMapping.java(PropertyMappingBuilder.build(), ~L286–303):The condition only looks at the source property and the target constructor parameter. It has
assignmentin scope but never consults it — so a value routed through a qualified method whose return type is non-null is treated identically to a value assigned directly. In the reported casemapValue(@Nullable String) -> Target.Nestedmakes the result provably non-null, and the diagnostic still fires.The obstacle to the obvious fix — "skip the error when the assignment's method returns non-null" — is that the return nullability isn't reachable from there.
MethodReferencetakes aMethodin its constructor but only retainsmethod.getReturnType():NullabilityResolver.getNullability(ExecutableElement, ...)already returns the return type nullability for anExecutableElement, so the resolver side needs nothing new — the missing piece is that theExecutableElement(or a precomputed nullability flag) isn't carried onMethodReferenceto the point where the check runs.So the fix seems to need one of:
MethodonMethodReference, orNullabilityat construction and store that flag,then gate the diagnostic on it. (2) is narrower but still touches a core model class on every mapping path, so it's worth a maintainer opinion on the preferred shape before someone writes it.
Worth noting the same gap likely affects #4081 — both are the reused-method return-nullability path not being considered.
I haven't opened a PR; flagging the analysis rather than guessing at the architecture.