[java] Unable to parse some Java9+ resource references

[jsotuyod]

Affects PMD Version:
7.0.0

Description:

Since Java 9, PMD has allowed for concise resources (ie: not variables declared within the try-with-resources), but some constructs are currently rejected by the parser even if valid Java code.

Code Sample demonstrating the issue:

class Foo implements AutoCloseable {
      void testThis() throws Exception {
          try (this) { // unsupported
          }
      }

     void testQualifiedThis() throws Exception {
          try (Foo.this) { // unsupported
          }
      }

     void testArrayAccess(AutoCloseable[] c) throws Exception {
          try (c[0]) { // unsupported
          }
      }

      @Override
      public void close() throws Exception {
      }
}

Probably others too…

Running PMD through: Any

Probably, the saner fix would be to simply remove this check altogether, and continue under the premise that the code compiles…

pmd/pmd-java/etc/grammar/Java.jjt

Lines 2868 to 2873 in a28943a

	{
	Node top = jjtree.peekNode();
	if (!(top instanceof ASTVariableAccess || top instanceof ASTFieldAccess))
	throwParseException("Expected a variable access, but was a " + top.getXPathNodeName());
	}
	{}

[oowekyala]

Interesting... We would need to make this method nullable and adapt the callers: https://github.com/pmd/pmd/blob/master/pmd-java/src/main/java/net/sourceforge/pmd/lang/java/ast/ASTResource.java#L52

Especially EmptyControlStatementRule needs new tests to disallow these constructs. Writing

try (this) {}

is weird and should be this.close(). I opened #4930.

[oowekyala]

Also note that c[0] is not a valid resource. Javac errors with the following message:

the try-with-resources resource must either be a variable declaration or an expression denoting a reference to a final or effectively final variable.

The only expressions allowed here are field accesses and variable accesses according to the JLS 14. It seems javac also accepts this, but that's all. So while we should support this corner case, I don't think we should remove the parser check altogether.