Runtime refactor #30
Runtime refactor #30
Conversation
- Removed vtable from implementation (though kept definition there for documentation) - Created debug_act and debug_sv structs to stash debug informations. I will probably also want to put trace callbacks in there, but need to figure out how best to do that. - Still need to add better interface to ssm-runtime; check what interface Hans has exposed. - Removed the offset and svt_ptr stuff.
Still need to add target for building and testing queue.
|
I have a couple of failing tests, but these fail on master as well, so I don't think these should prevent this PR from being merged. I've documented the MultOverflow cases in #32, but I just have no idea what FiveForker is, or why it fails, but it's a massive test case with lots of forks. Both the interpreter and the executable reach a |
|
The fiveforker one is super interesting. I'll try to manually massage quickcheck to shrink it a bit, and see if I can figure it out. |
|
Hm QuickCheck won't shrink it more. One positive is that whenever we figure out what the issue is, we'll probably manually come up with a smaller test case that can guide another shrinking strategy that replicates it. |
|
Thanks for the approval @Rewbert ! I'll merge it into master after pushing a couple of more minor fixes. |
I dropped in my refactor of the runtime system from upstream. In addition to what should be a functionally equivalent refactor, it makes the following changes:
struct svandstruct actrather than inlining fields, to avoid some type casting