Open heidihoward opened 2 years ago
We will have a closer look. State and action properties are treated slightly differently. There may be a one-off issue somewhere.
Cheers! The issue seems to be that state invariants are checked upto the length (inclusive) whereas the unprimed variables in action invariants are checked up to length (exclusive)
Description
I would expect that if one of two invariants, A and B, are found to be incorrect when separately model checked in Apalache (same length, same spec etc) then
A /\ B
should also be incorrect. However, this is not always the case if one is a state invariant and one is an action invariant.Input specification
Invariants.tla from the provided examples with the following addition:
The command line parameters used to run the tool
Expected behavior
Test 1 and Test 2 to both find a counter example
Log files
And
System information
apalache-mc version
]: 0.24.0Additional context
I'm not sure if this is a bug or by design but if the latter users like me might benefit for some more documentation.