H. G. Muller wrote on Tue, Nov 22, 2016 11:04 PM UTC:
OK, I see. This is not an engine bug, but a GUI bug. That is, WinBoard was never really changed to handle this complicated form of e.p. capture. It has always performed e.p.captures by the heuristic that a Pawn moving to an empty square not in the same file would as a side effect remove the piece just behind its destination square. Later I coded exceptions for this in variants xiangqi and berolina,and limited this behavior to cases where either legality testing was on, or the Pawn moved to the e.p. file. But in the case of a diagonally capturing Pawn e.p. capturing a diagonally double-pushed Pawn, this heuristic deletes the the wrong victim (e4 instead of f5). The engine knows nothing about that.
I any case WinBoard should not assume anything on pieces with redefined moves, even when legality testing is on. It should only assume e.p. capture when the move redefinition explicitly specifies the piece can e.p. capture, and the given move matches the defined move that does this.
Using the Lance symbol for Pawns would avoid application of the heuristic, but then true e.p. captures would never get their victim removed. The whole issue of e.p. capture is problematic, because the standard move notation always assumed that that the capture is an implied side effect of the move. But the black Pawn in the position you show is Elvish, this is no longer true. There is no general heuristic that would predict whether de6 in this position should remove f4 or not. It depends on the game rules, and it could even be that these allow both, so that the move can no longer be specified by just the start and destination squares.
As to the piece images used in the web page: I have only a limited set of images available, namely the XBoard piece glyphs. The crossed swords are WinBoard's standard representation of the Archbishop. I have only two bishop-like glyphs, one for the orthodox Bishop, the other now used for Hunter. (Well, that is not entirely true. I also have a Bishop without the inscribed cross. But it differs very little from the normal Bishop.Would it be better to use that for the Hunter, and the one now used for Hunter for the Centaur?) The point is that the piece that you call Centaur would be called Elephant in almost any variant it occurs. And it is a quite popular fairy piece. So Chess-variant players associate the Elephant very strongly with exactly the move you use for Hunter. So using a Bishop-like symbol for this piece would be rather confusing.
OK, I see. This is not an engine bug, but a GUI bug. That is, WinBoard was never really changed to handle this complicated form of e.p. capture. It has always performed e.p.captures by the heuristic that a Pawn moving to an empty square not in the same file would as a side effect remove the piece just behind its destination square. Later I coded exceptions for this in variants xiangqi and berolina,and limited this behavior to cases where either legality testing was on, or the Pawn moved to the e.p. file. But in the case of a diagonally capturing Pawn e.p. capturing a diagonally double-pushed Pawn, this heuristic deletes the the wrong victim (e4 instead of f5). The engine knows nothing about that.
I any case WinBoard should not assume anything on pieces with redefined moves, even when legality testing is on. It should only assume e.p. capture when the move redefinition explicitly specifies the piece can e.p. capture, and the given move matches the defined move that does this.
Using the Lance symbol for Pawns would avoid application of the heuristic, but then true e.p. captures would never get their victim removed. The whole issue of e.p. capture is problematic, because the standard move notation always assumed that that the capture is an implied side effect of the move. But the black Pawn in the position you show is Elvish, this is no longer true. There is no general heuristic that would predict whether de6 in this position should remove f4 or not. It depends on the game rules, and it could even be that these allow both, so that the move can no longer be specified by just the start and destination squares.
As to the piece images used in the web page: I have only a limited set of images available, namely the XBoard piece glyphs. The crossed swords are WinBoard's standard representation of the Archbishop. I have only two bishop-like glyphs, one for the orthodox Bishop, the other now used for Hunter. (Well, that is not entirely true. I also have a Bishop without the inscribed cross. But it differs very little from the normal Bishop.Would it be better to use that for the Hunter, and the one now used for Hunter for the Centaur?) The point is that the piece that you call Centaur would be called Elephant in almost any variant it occurs. And it is a quite popular fairy piece. So Chess-variant players associate the Elephant very strongly with exactly the move you use for Hunter. So using a Bishop-like symbol for this piece would be rather confusing.