On 09/08/2012 11:41, chris derham wrote: > | I am 95% certain of what is happenning but haven't tested this. It will > help your understanding if you read the EL specification, specifically the > section in the Javadoc regarding how implementations are discovered. > | Tomcat implements these rules. It is possible for an EL implementation to > be loaded from an application. Everything will be fine EL-wise for the > application that shipped with the EL implementation but things will > | fall apart for any other application that tries to use EL since the > implementation classes will not be visible. > > Mark, > > I don't understand one key component of your response. I don't pretend to > understand the internals of tomcat, but I thought that I had a good grasp > on how things were supposed to work. Please forgive me if I ask something > stupid. > > You seem to say that the EL implementation will be loaded from the app2, as > it has an implementation present in its lib. You then say that this won't > be visible to app 1, presumably because this class is loaded into the > app2's class loader. That bit matches up with my understanding on app class > loaders in web containers - each app has its own, and they are separate > preventing leakage from one app to another. The bit I don't really > understand is when app 1 tries to run some EL code, surely the same > mechanism should kick in and load an EL implementation.
I strongly suspect - but haven't run a test to confirm - that the implementation obtained from app2 is cached by the EL discovery mechanism so that app1 is given the same instance. It is pretty much a given that app1 will then break. The only question is how long it will take before app1 hits a point where it realises it is working with an instance of a class it can't load. I'd need to do some testing to confirm exactly what is going on but the reported behaviour is - so far - consistent with what I think is happening. HTH, Mark --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
