https://bz.apache.org/bugzilla/show_bug.cgi?id=56530
--- Comment #20 from Christopher Schultz <ch...@christopherschultz.net> --- (In reply to Konstantin Kolinko from comment #19) > (In reply to Christopher Schultz from comment #18) > > > > 2. If the branch is not tolerable at runtime, perhaps the branch can be > > executed a single time to install a delegate adapter, something like this: > > > > This cannot be solved with an adapter that is installed atop the current > WebappClassLoader class. > > See [1]. The class itself and all of its parent classes must be registered > as parallel-capable. Aah, yes; I had forgotten about that. Well, that makes things ugly. > This needs > 1. Refactoring, involving extraction of common parent class. > > 2. Implementation of alternative class loader class that is parallel-capable. > > The default implementation (WebappClassLoader) must stay with being non > parallel capable. The common parent class must be registered as > parallel-capable. > > If ClassLoader.registerAsParallelCapable() is called via reflection, I > wonder whether the registerAsParallelCapable() method can correctly > determine the calling class. We are not calling the method directly, but via > reflection APIs. It needs a proof of concept. If registerAsParallelCapable can't be called via reflection, then it means that both Java 6 and Java 7 are required to build Tomcat 7. We already require Java 7 for building the (optional) WebSocket components. Could this simply be added to that list? > 3. Testing, testing, testing... Since the "regular" WebappClassLoader is still available (and the default CL), I think it's okay to mark it as an experimental feature and allow users to try it out. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
