I guess I know what's my least favorite update now. Still my favorite tool though.
Edit: Replacing mpaf with mcdauf fixed this.
I completely missed this Comment. Unfortunately the links you give there don't seem to work anymore, so I have no idea what kind of moves you defined for which this could be harmful. Where these cases where the mpaf was supposed to make a move lame versus off-board squares / holes?
As I wrote elsewhere, the alternative would be to only perform the merging of legs on a step with mpo mode. Although the merging also can greatly improve efficiency of the move generator (XBetza like (mpaf)W for a flying Rook would effectively expand to WDHWXDXHXWXX instead of to WmpafWmpafmpafWmpafmpafmpafWmpafmpafmpafmpafWmpafmpafmpafmpafmpafWmpafmpafmpafmpafmpafmpafW, avoiding the need for needless tests that could not possibly fail of dummy intermediate squares), its main purpose was separation of auto-generated mpa legs during expansion of bracket notations like [N-B]. So that these would not be unintendedly subject to hop bans from the capture matrix, which you might want to have to make a move like [pN-B] type specific. If I make these bracket expansions use mpo intermediate dummies, and only merge these, it will likely do less harm to existing Diagrams.
I completely missed this Comment. Unfortunately the links you give there don't seem to work anymore, so I have no idea what kind of moves you defined for which this could be harmful. Where these cases where the mpaf was supposed to make a move lame versus off-board squares / holes?
As I wrote elsewhere, the alternative would be to only perform the merging of legs on a step with mpo mode. Although the merging also can greatly improve efficiency of the move generator (XBetza like (mpaf)W for a flying Rook would effectively expand to WDHWXDXHXWXX instead of to WmpafWmpafmpafWmpafmpafmpafWmpafmpafmpafmpafWmpafmpafmpafmpafmpafWmpafmpafmpafmpafmpafmpafW, avoiding the need for needless tests that could not possibly fail of dummy intermediate squares), its main purpose was separation of auto-generated mpa legs during expansion of bracket notations like [N-B]. So that these would not be unintendedly subject to hop bans from the capture matrix, which you might want to have to make a move like [pN-B] type specific. If I make these bracket expansions use mpo intermediate dummies, and only merge these, it will likely do less harm to existing Diagrams.