Closed ehildenb closed 3 years ago
Since this passes with a version of the backend which is ~1 month old, I'm not that sure anymore that this is related to the predicate simplification loop (this is an older issue). So I think we could take a look at this before we get the predicate simplifier fixed. We should at least identify the version of the backend this starts failing, or check that this isn't actually an issue with the spec/semantics itself.
This issue was introduced by commit 67d4d23698551ba29228864fdff55a15d86f3d55
Hmm, that's weird! Could something have been broken in Bool
unification?
@ttuegel Suggests to run with the equation debug logs and see if it gets stuck while simplifying the side condition.
On K master (* 72d33ae0e0 - Update dependency: haskell-backend/src/main/native/haskell-backend (#2016)), which is immediately after merging this PR: https://github.com/kframework/k/pull/2016
I have a proof that loops. This comes from KEVM, prior to this it takes about 25s for this proof to run, but now it times out with a time limit of 50min.
Here is the minimized example
test.k
:To reproduce the looping behavior, run:
Note the 120s timeout here. If you remove either of the lemmas in
INFINITE-GAS-SPEC
, it fails very quickly (4s or so).On KEVM master, the proof that inspired this minimized example (it's been mangled a bunch to minimize it) passes in about 25s.