[Java] Infer fails to detect a deadlock bug

[RJerrica]

Hi, I found that Infer fails to detect a deadlock in the following code example. It should have reported a deadlock warning at lines 9 and 14, as these two methods acquire locks this and lockB in reverse order, which can lead to a deadlock. This constitutes a false negative in deadlock detection.

Minimized Code Example

import android.support.annotation.UiThread;
import android.support.annotation.WorkerThread;

public class Main {
  private final Object lockB = new Object();

  @UiThread
  public synchronized void annotatedUiThreadBad() {
    lockSync();
  }

  @WorkerThread
  public void annotatedWorkerThreadBad() {
    syncLocks(lockB, this);
  }

  private void lockSync() {
    synchronized (lockB) {}
  }

  private void syncLocks(Object lock1, Object lock2) {
    synchronized (lock1) { synchronized (lock2) {} }
  }
}

Analysis Log

Capturing in javac mode...
Found 1 source file to analyze in /infer-out
0/6 [................................................................................] 0% 126ms
⊢ [ 0.1s][17.8M] idle
⊢ [ 0.1s][17.8M] Main.java: Main.<init>()
3/6 [########################################........................................] 50% 185ms
⊢ [ 0.0s][17.8M] void Main.annotatedUiThreadBad()
⊢ [ 0.1s][17.8M] idle
6/6 [################################################################################] 100% 191ms
⊢ [ 0.0s][17.8M] idle
⊢ [ 0.1s][17.8M] idle
⊢ [ 0.1s][17.8M] idle

  No issues found