To test the runtime and the translator w/o loading modules into kernel
for this, it is desirable that as much of the normal runtime interface as possible still continue to work. of course, kprobes proper can't be used. but otherwise, maps, basic I/O transport calls ... should all work. As for packaging under user/, it would be easier to use if the user level simulations lived within the main directory, and enabled by #ifdef macros (e.g. TEST_MODE set by the translator)
no longer high priority
Closing this. The existing user-level testing for runtime and the translator seem to be adequate. It is not possible to improve much on that without major work and it would never be good enough because we cannot accurately simulate the contention that occurs in the kernel. It is much more important at this point to work on real world tests that actually load modules into running kernels.