ExceptionMapper not thread safe

[DanyalBurke]

ExceptionMapper is accessed from multiple request threads but the underlying Map is a HashMap; could be refactored to use a ConcurrentHashMap or methods could be made synchronized

[jakaarl]

I can't claim to be very knowledgeable about the inner workings of Spark, but from a quick read of the sources, I think this is not necessarily a big problem. The exception mappings are done in SparkInstance.exception(), which is synchronized (and AFAIK, multiple instances are not intended to be used). MatcherFilter accesses the mappings without synchronization, but in a read-only fashion. As long as exceptions are mapped before SparkInstance.init()is called, we're good - of course, nothing prevents users from adding routes or filters before exception mappings, which could cause a race condition. I guess the solution would be to either synchronize access (or use a concurrent map) or require that exceptions are not mapped after initialization.