Issues with IntelligentMode, Dummies and condition promonly

[thomas-maeder]

Reported by Andrew Buchanan over in dturevski/olive-gui#29

(1) First: h#6 intelligent problems P1359493 and P1375440 run fine, with promonly Q R S B DU in the second case. Only odd thing is that the solution in the second case begins 1.c2-c1=TI. Don't know what a "TI" is - not in the list of fairy pieces.

(2) Now I try to run 5P2/6B1/8/8/3p4/8/2p2P2/k5K1 with promonly Q R S B P. It basically never comes back.

(3) If instead I try to replace wPf8 with wDU, it finds the solution (beginning 1.c2-c1) and does nothing for a while. When I select the window, an error pop-ups where you can't copy the text comes up, saying olive-1.3 Failed to run popeye executable (pywin64.exe). I am running 64 bit laptop. Hope this is enough for you to diagnose.

[JoshuaGreen]

@thomas-maeder, I don't see the input(s) leading to these issues on either ticket. Could you please attach them?

[dturevski]

@JoshuaGreen
This is the one with "TI":

BeginProblem
Stipulation h#6
Option Variation NoBoard Intelligent
Condition PromOnly Q R S B DU
Pieces
  white  Bf6 duf2 Kg1
  black  Pe7c2d4 Ka1
EndProblem

Popeye Windows-64Bit v4.85 (4000 MB)

  1.c2-c1=TI Bf6*e7   2.d4-d3 Kg1-f1   3.d3-d2 Kf1-e2   4.d2-d1=B + Ke2-d2   5.Bd1-b3 Kd2*c1   6.Bb3-a2 Be7-f6 #

solution finished. Time = 4.677 s

[dturevski]

I would suspect that it has something to do with parsing "Condition PromOnly Q R S B DU"
Allowing reductions in a format that is expected to be extensible...

[JoshuaGreen]

@dturevski, thanks for that example. I'm starting with the "mostly orthodox" one

BeginProblem
    forsyth 5P2/6B1/8/8/3p4/8/2p2P2/k5K1
    condition promonly Q R S B P
    stipulation h#6
    option intelligent
EndProblem

which also triggers bad behavior (e.g., an assertion violation in debug mode).

[JoshuaGreen]

@dturevski (and @thomas-maeder), it seems that much of the intelligent infrastructure assumes that we only have pawns, Knights, Bishops, Rooks, Queens, and Kings, and that promotion to pawn is illegal; I'm getting lots of assertion violations for each of these examples. We have to decide how many of these assumptions we're willing to give up.

[dturevski]

@JoshuaGreen
It would seem that "TI" issue is not really related to the intelligent mode and can be triggered simply by "PromOnly DU", as if there's some problem with printing moves that include "DU":

BeginProblem
Stipulation ~1
Condition PromOnly DU
Pieces
  black  Kh6
  white  Kh8 Pf7
EndProblem

1.f7-f8=TI ! (expecting DU)
1.Kh8-g8 !

[JoshuaGreen]

@dturevski, I'll try to pin that one down as it may very well be a separate issue. Thanks for distilling it.

[JoshuaGreen]

@dturevski, the printing of Dummy pieces as "TI" seems to have been an intentional choice that was made at some point during Popeye's development. I don't know enough about these pieces or concepts to know if this decision was reasonable. Nonetheless, I offer feature/outputDummyAsDU as a change to that behavior, and I invite @thomas-maeder to merge it in if he agrees that it's preferable.

[thomas-maeder]

the printing of Dummy pieces as "TI" seems to have been an intentional choice that was made at some point during Popeye's development.

Not quite.

It is an artifact of the fact that when a problem with Total Invisible pieces is solved, dummies are used as placeholders if the position of a Total Invisible piece is known or assumed.

[JoshuaGreen]

Thanks for that input; it was possibilities like that that prevented me from merging my "fix" directly into the codebase. I guess the right fix for this -- if it's really a problem -- might require some deeper analysis to determine the precise nature of the relevant piece.

[thomas-maeder]

Fixed for 4.87. Popeye will:

• display Dummies as DU in the solution (it hadn't occurred to me that Dummies might be printed in the solution)
• reject promotion to Pawn in Intelligent Mode
• accept Dummies (including promotion into Dummy) in Intelligent Mode
• not crash in Intelligent Mode when there are Dummies (not at this one spot anyway :-) )