Gili Tzabari

The only way I found to make this work is using the following `spy.properties` configuration: `modulelist=com.p6spy.engine.logging.P6LogFactory,com.p6spy.engine.outage.P6OutageFactory,\ foobar.LogExceptionsFactory` where `foobar.LogExceptionsFactory` is defined as: ``` public final class LogExceptionsFactory implements P6Factory {...

To clarify, I am asking for an extension point (allow users to pass in a ClassLoader to use) or an outright feature.

Tibor, Most code makes use of static variables somewhere. So long as you run inside the same ClassLoader you run the risk of one test influencing another. I have run...

It would work like this... Given: Thread 1: T1 Thread 2: T2 T1 creates a new ClassLoader C1 T2 creates a new ClassLoader C2 T1 loads class A using C1...

Bytecode intrumentation is evil and results in many annoying problems (especially when it comes to debugging and dealing with PermGen leaks). Using separate Classloaders solves all that.

In an ideal world, yes. Every @Test should run in its own ClassLoader. More realistically, applying concurrency on a per-class level should be a good first step. So all tests...

I understand, which is why I filed this issue in the first place. It would be useful to crunch some numbers to check whether this is even reasonable to implement....

I can't speak to the JUnit API. I've used it for years but I'm sure you know its inner workings more than me. The incentive behind this RFE is to...

Even Guice has a problem with static context. You're simply never going to get them all. I understand reloading classes will slow things down, but it doesn't really matter. AMD...

The interesting bit is if you remove module1, module2 from the testcase (convert it to a single-module project) then Surefire returns the compiler error to the console without crashing. The...