In our many small issues to refactor the landmark code of Fast Downward, we often find places where derived landmarks (i.e., landmarks containing derived predicates) are treated in a special way. Most often, they are simply excluded which is probably because nobody ever bothered to think through what should be done in that case. One such example is in the dead-end check based on landmarks with empty achievers (see issue1049). The goal of this issue is to come up with a more sophisticated treatment of these variables (or to make sure there is none). This is closely related to the findings documented in issue247. Especially Malte's msg10605 could be a starting point for this issue.
|