[Bug-gnubg] Bug in importing .mat files

[Holger]

Since no one replied to my first post, I suppose it was because of the
length of the message due to the (text) attachments. So here it is again
without attachments. If someone needs them, they are in the post of
Saturday 5th Oct.

-- 

Hi GNUbg team,

First of all I'd like to congratulate and thank you for making a great
program and offering it for free. But like every non-trivial program it
might contain bugs. And (I think) I found one.

I'm writing a match converter (only for personal use so far) that directly
converts JavaFIBS' internal format into SGF FF4. As JavaFIBS comes with mat
export, that I believe is correct, it's possible to compare the results
after importing in GNUbg.

The problem occurs with resignation. GNUbg doesn't give the correct result
for the game if a player resigns when it's not his turn. At FIBS this is
possible.

I have attached an example played by me on FIBS. It's a 3 pt match, where
in the last game my opponent resigned :-) when I had rolled already. GNUbg
shows an incorrect win of 1 pt for white. (I suppose because white made the
last move.) It should be 4 pts for black or in sgf-notation: RE[B+4R].

About the last 2 lines of my output (BGconvert-correct.sgf):
;DI[22]PL[B]
;W[resign])
The first line is probably a matter of taste (or the capabilities of the
reading program; GNUbg does understand it) not to write ";B[22]". As I read
the docs of SGF "B[]" would be a move, that in this case couldn't be made.
But this would mean predicting an information and might influence the
decision whether to double or not to.

The last line is not covered by the SGF standard. However, imo it's
important for evaluation, especially if one adds information about whether
a gammon or a backgammon is offered. Besides, then the other player might
reject the offer to resign, for which there is neither a property in the
SGF standard.

If you could comment on these thoughts and how to solve it, please?

TIA and regards,

        Holger