False positive warning "Potential resource leak" inside "try with resources"

[iloveeclipse]

Given this example and enabled org.eclipse.jdt.core.compiler.problem.potentiallyUnclosedCloseable=warning option in .settings/org.eclipse.jdt.core.prefs

import java.io.FileWriter;
public class DemoNonCloseableWarning {
    public static void main(String[] args) throws Exception {
        try (FileWriter writer = new FileWriter("/dev/null")) {
            writer.append("\n");
        }
    }
}

we report false positive warning Potential resource leak: '<unassigned Closeable value>' may not be closed.

Of course this is option is not enabled by default, and of course it is potential, but here we have undoubtedly closed writer, so not idea where the "potential" warning is coming from.

@stephan-herrmann : FYI.

[iloveeclipse]

Fun fact: it worked at least in 3.8.2 Eclipse on Java 1.7 :-)

[stephan-herrmann]

This looks spooky on first glance:

----------
1. WARNING in /tmp/DemoNonCloseableWarning.java (at line 5)
        writer.append("\n");
        ^^^^^^^^^^^^^^^^^^^
Potential resource leak: '<unassigned Closeable value>' may not be closed
----------
1 problem (1 warning)

Not allocating the writer but calling append() triggers the warning. But wait: append returns a Writer (which could in theory be a fresh new resource).

We do have provision for this situation, see org.eclipse.jdt.internal.compiler.lookup.TypeConstants.FLUENT_RESOURCE_CLASSES. Something confuses the detection of this "fluent" method.

[iloveeclipse]

But wait: append returns a Writer

OK, I see it now where AI hallucinates :-). Compiler assumes the return value has to be closed...

[stephan-herrmann]

OK, I see it now where AI hallucinates :-). Compiler assumes the return value has to be closed...

Don't worry: with AI in the picture any causalities between input program and warnings would be much more obscure.. ;-P