Open sgoings opened 11 years ago
One of the things we get with the current tests, that I don't want to lose, is the ability to debug low-level jit issues using the tests. If we have them running through a gigantic test harness, more likely than not the test harness itself will fail long before the actual test will. This is instrumental in modifying the VM itself.
Sure. This issue is my first way of taking a stab at improving Avian's structure, testing, and build system...
I think there's a need to test Avian at different levels, and we'll definitely need to use the appropriate tools for each component, but that's why I entered this issue - because I personally want to see improvement in this area. To your point, JUnit won't necessarily be the solution for all of Avian, but it should be a tool that we use for a significant number of our classpath tests.
Just to put in my three cents: there is no improvement I'd expect from switching to JUnit. I still will develop and run the tests in Eclipse, or on the command-line, and would not benefit from using another test infrastructure than the RuntimeException based one we have right now.
I'd like to see Gradle take control over the pure Java build portions of this project's build infrastructure in order to reap signifcant benefits in code quality and structure flexibility.
Specifically: