HashedWheelTimer does not always respect maxPendingTimeouts

Expected behavior

HashedWheelTimer should respect the maxPendingTimeouts parameter.

Actual behavior

When the value is unlimited (e.g. <= 0), users can still hit an java.lang.IllegalStateException: Queue full from the underlying Queue which comes from PlatformDependent.newMpscQueue() in the constructor.

Steps to reproduce

Create a HashedWheelTimer with unlimited maxPendingTimeouts then add a task that never completes (e.g. while (true) println("hi");). Then keep adding tasks to the queue until you get the queue full exception.

Here is an example (partial) stacktrace:

Caused by [alertable exception] java.lang.IllegalStateException: Queue full
     at java.util.AbstractQueue.add(AbstractQueue.java:98)
     at io.netty.util.HashedWheelTimer.newTimeout(HashedWheelTimer.java:423)
     ...

Minimal yet complete reproducer code (or URL to code)

Netty version

4.1.12

JVM version (e.g. java -version)

OS version (e.g. uname -a)

[kevinoliver]

@Scottmitch nothing else needed from my end. thank you!

[Scottmitch]

@kevinoliver - Now that https://github.com/netty/netty/pull/6295 has been pulled in … and HashedWheelTimer already uses PlatformDependent.newMpscQueue() (unbounded queue), and we enforce the limit via pendingTimeouts … is there any more action required to close this issue?