Re: [Bug-gnubg] Amazing GNU bug - Related to August 3rd change

[Michael Petch]

On 17/09/2011 2:28 PM, Michael Petch wrote:
> On 17/09/2011 1:31 PM, Timothy Y. Chow wrote:
>> I then decided to perform a 2-ply evaluation of one of the moves that I 
>> had *not* rolled out.  I scrolled down to find "20/14* 3/1" and clicked on 
>> it, and then I clicked the "2" button to get a 2-ply analysis.  I got the 
>> 2-ply analysis I wanted, but (here's the bug) the dice switched color!  
>> That is, instead of two black dice with 2's on them, there were now two 
>> *white* dice with 2's on them.
>>
>> The most amazing thing, though, was that I then clicked on some other play 
>> and clicked the "2" button to get a 2-ply analysis, and now it's somewhat 
>> hard to describe what happened.  The computer seems to have treated the 
>> play as a play by *White* (after all, the dice are white!) and tried to 
>> evaluate it accordingly.  But 20/18(2) is illegal for White to play here 
>> because Black owns that point.  Undaunted, GNU relabels the moves with 
>> 20/18(2) in them as "20/18*(2)".  Astounding!  What is it doing?
>>
>> I'm curious whether you can reproduce my experience.
>>
>> Tim
> With your file I followed the instructions exactly. Originally I
> couldn't reproduce it but then discovered I was using the 20110209
> build. I then tried with the latest 20110822 version and it was
> reproducible. Although the 20110209 version didn't seem to have the same
> issue I did note that in both cases selecting 20/14* 3/1 (28th in the
> analysis pane) causes the cursor in the game (move) list to shift from
> the right column to the left column (So the cursor is now sitting on the
> blank) . In newest releases the dice change color and the state seems to
> get mixed up as well. It is definitely a bug. I am currently busy on
> something else at the moment but I will put it on the mailing list if
> someone wished to investigate.

Howdy Philippe,

Earlier today I posted a bug reported to me by Timothy Chow regarding
opening of an SGF file (attached to the previous email) and the unusual
state selecting on one of the moves in the analysis pane produced.

I have discovered that this bug seems to be specific to a patch made at
the beginning of August 2011. Specifically this code this causing the
problem:

2011-08-03  Philippe Michel  <address@hidden>
       * gnubg.c, gtkboard.c, play.c, set.c: More fixes to swap players
       feature.

Would you be kind enough to see if you can see why this is occurring? I
haven't looked closely at this code section and figure you probably know
it better than I do.

Thanks,

-- 
Michael Petch 
CApp::Sysware Consulting Ltd. 
OpenPGP FingerPrint=D81C 6A0D 987E 7DA5 3219 6715 466A 2ACE 5CAE 3304