Does the finally block execute if the thread running the function is interrupted?

Question

If I have a function with a try/finally section, and the thread running it is interrupted while in the try block, will the finally block execute before the interruption actu

Answer 1

Many of Oracle's Java tutorials are helpful (I have answers referencing the guarded blocks page and the SAX introduction), but they are not necessarily authoritative, and some of them have mistakes or are incomplete. The quote referenced in the question conflates interruption with the JVM exiting, which is confusing.

First, thread interruption in Java has nothing to do with OS-level interrupts. Sharing a name creates opportunities for confusion but there is no connection.

Next, JVM exit obviously kills the thread without an opportunity to do any cleanup. If the process dies before the thread has gotten as far as the finally block, too bad. But there's no comparison to interruption. Nothing about interruption prevents finally blocks from completing.

A design principle of interruption is that acting on the interruption requires the cooperation of the thread being interrupted. The thread interrupted responds at its discretion, the interruption doesn't compel the thread to do anything. All calling Thread#interrupt() does is set a flag on the thread. Blocking methods like wait or sleep check the flag to see if they should wake up early. (InterruptedException is a checked exception so you can tell who throws it when, and your Runnable can plan for it.) Also any code can use Thread#isInterrupted() to check whether its thread has had the flag set.

When Thread#sleep() recognizes the interrupted flag is set, it clears the flag before throwing InterruptedException. When your thread catches an InterruptedException it's good manners to restore the flag using Thread.currentThread().interrupt(), just in case there is any other code running in that thread that needs to know about the interruption. This comes into play when you have more complex situations with nested synchronizers where, for instance, some deeply nested component could get its sleep interrupted, letting it stay cleared could prevent higher layers from knowing about the interruption. In a simple toy example like the ones in other answers here, it doesn't matter if the flag is restored or not, nothing checks it again and the thread terminates.