2.0.3...2.2.2: IllegalArgumentException: Not an array type for method like void method(@Nullable Object[])

[vlsi]

Sample:

public interface EqualityComparer<T> {
  boolean equal(T v1, T v2);

  int hashCode(T t);
}

  private static class ArrayEqualityComparer
      implements EqualityComparer<@Nullable Object[]> {
    @Override public boolean equal(@Nullable Object[] v1, @Nullable Object[] v2) {
      return Arrays.deepEquals(v1, v2);
    }

    @Override public int hashCode(@Nullable Object[] t) {
      return Arrays.deepHashCode(t);
    }
  }

Error (in jandex 2.0.3.Final and 2.2.2.Final):

Caused by: java.lang.IllegalArgumentException: Not an array type!
  at org.jboss.jandex.Type.asArrayType(Type.java:254)
  at org.jboss.jandex.Indexer.resolveTypePath(Indexer.java:621)
  at org.jboss.jandex.Indexer.resolveTypeAnnotation(Indexer.java:593)
  at org.jboss.jandex.Indexer.resolveTypeAnnotations(Indexer.java:478)
  at org.jboss.jandex.Indexer.index(Indexer.java:1458)
  at com.github.vlsi.jandex.JandexWork.execute(JandexWork.kt:52)

See https://issues.apache.org/jira/browse/CALCITE-4459

[vlsi]

Smaller case:

public interface EqualityComparer<T> {
  boolean equal(T v1, T v2);

  int hashCode(T t);
}

  private static class ArrayEqualityComparer
      implements EqualityComparer<Object[]> {
    @Override public boolean equal(Object[] v1, Object[] v2) {
      return Arrays.deepEquals(v1, v2);
    }

    @Override public int hashCode(@Nullable Object[] t) {
      return Arrays.deepHashCode(t);
    }
  }

javac 11, javap:

  public int hashCode(java.lang.Object[]);
    descriptor: ([Ljava/lang/Object;)I
    flags: (0x0001) ACC_PUBLIC
    Code:
      stack=1, locals=2, args_size=2
         0: aload_1
         1: invokestatic  #4                  // Method java/util/Arrays.deepHashCode:([Ljava/lang/Object;)I
         4: ireturn
      LineNumberTable:
        line 497: 0
      LocalVariableTable:
        Start  Length  Slot  Name   Signature
            0       5     0  this   Lorg/apache/calcite/linq4j/function/Functions$ArrayEqualityComparer;
            0       5     1     t   [Ljava/lang/Object;
    RuntimeVisibleTypeAnnotations:
      0: #29(): METHOD_FORMAL_PARAMETER, param_index=0, location=[ARRAY]
        org.checkerframework.checker.nullness.qual.Nullable

  public int hashCode(java.lang.Object);
    descriptor: (Ljava/lang/Object;)I
    flags: (0x1041) ACC_PUBLIC, ACC_BRIDGE, ACC_SYNTHETIC
    Code:
      stack=2, locals=2, args_size=2
         0: aload_0
         1: aload_1
         2: checkcast     #5                  // class "[Ljava/lang/Object;"
         5: invokevirtual #6                  // Method hashCode:([Ljava/lang/Object;)I
         8: ireturn
      LineNumberTable:
        line 490: 0
      LocalVariableTable:
        Start  Length  Slot  Name   Signature
            0       9     0  this   Lorg/apache/calcite/linq4j/function/Functions$ArrayEqualityComparer;
    RuntimeVisibleTypeAnnotations:
      0: #29(): METHOD_FORMAL_PARAMETER, param_index=0, location=[ARRAY]
        org.checkerframework.checker.nullness.qual.Nullable

[vlsi]

I guess the issue is caused by ACC_BRIDGE, ACC_SYNTHETIC method.
https://bugs.java.com/bugdatabase/view_bug.do?bug_id=6695379 Copy method annotations and parameter annotations to synthetic bridge methods might be relevant.

[n1hility]

Thanks for the report. I'll need to research various compiler outputs. It looks like the generated bridge is referring to a type path that is not valid with this signature. This will probably need to be gracefully ignored.

[vlsi]

What do you think if a GitHub Action was added that tried to parse Apache Calcite codebase with the current jandex? Calcite has a significant amount of type annotations added in apache/calcite#2268

It won't make debug easier, however, it might help to prevent regressions.

[n1hility]

@vlsi i think thats a great idea. I would be happy to regression test with calcite.