[OPENJPA-2243] Deadlock involving FormatPreservingProperties

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.2.0
• Component/s: lib
• Environment: All environments
• Patch Info: Patch Available
• Assignee: Albert Lee
• Reporter: Michael Pradel
• Created: 27/Jul/12 10:02
• Updated: 27/Jul/12 15:25

Description

FormatPreservingProperties extends Properties, which is a thread-safe class, but it does not preserve the thread safety. This leads to surprising deadlocks if you have a reference of type Properties, because it may turn out to be not thread-safe (despite being documented as such).

Here's a simplified example that shows how we hit this bug:

final Properties p1 = new Properties();
final Properties p2 = createPropsWithDefault(p1);
p2.put(p2, p2);

Thread t1 = new Thread(new Runnable() {
public void run()

{ p2.remove(p1); }

});

Thread t2 = new Thread(new Runnable() {
public void run()

{ p2.stringPropertyNames(); }

});
t1.start();
t2.start();
try { t1.join(); t2.join();
} catch (InterruptedException e)

{ e.printStackTrace(); }

Properties createPropsWithDefault(Properties p) {
// return new Properties(p); // OK
return new FormatPreservingProperties(p); // leads to deadlock
}

If createPropsWithDefault() returns Properties, everything is OK (as it should be, because Properties is thread-safe). However, if it returns FormatPreservingProperties, we occasionally get a deadlock.

Are you aware of this inconsistency? It seems that the safest way of extending Properties is to make methods that override synchronized methods synchronized.