Open Halbaroth opened 1 month ago
I don't think this is related to Dolmen -- this file is also proved by the legacy frontend. This seems to be due to the Backward
instantiation pass which seems to always compute triggers (ignoring matching) and instantiate all lemmas that can prove things from (P e1 .. en)
whenever (P e1 .. en)
is in the context.
While removing the legacy frontend, I ran our non regression test suite with the SAT solver
Tableaux
. I caught a bug. This test failed:We got
unsat
even if we should not because the trigger prevents us from instantiating the axiom. It seems that the fix in #1051 have not been tested withTableaux
.Notice that the bug affects
Tableaux-CDCL
too.