Re: [gnugo-devel] cosmic patch

[Stéphane Nicolet]

> I had the impression that the patch cosmic_5_1.2 should have NO
> effect on the regressions if you omit the command line option.
> Is this correct?
>
> However it DOES affect the regressions, for example it causes
> four unexpected fails in owl.tst.
> 132 FAILED: Correct '(1|2) R4', got '1 R1'
> 133 FAILED: Correct '1 M3', got '0'
> 150 FAILED: Correct '0', got '1 G12'
> 151 FAILED: Correct '1 G13', got '1 G12'
> Dan

Incredible as it sounds, this is actually a problem with the
'patch' Unix command, which silently disgard some lines marked
with +  in the patch file cosmic_5_1.2 !

The hunk starting at line 247 and ending at line 320 of the
patch is not applied correctly, for some reason. Specificaly,
the end of the hunk (lines 311-318) is forgotten, which has
the net effect of removing I patterns from the
influence_callback() function...


Lines 834-869 in influence.c should read like that
(maybe you could restore lines 859-868 directly in CVS) :




  /* For I patterns, add a low intensity, both colored, influence
   * source at *.
   */
  if (pattern->class & CLASS_I) {
    int this_color = EMPTY;
    float strength;
    float attenuation;

    if (q->color_to_move == EMPTY || (pattern->class & CLASS_s))
      this_color = BLACK | WHITE;
    else if (q->color_to_move != color)
      this_color = q->color_to_move;

    if (cosmic_gnugo) {
      float t = 0.15 + (1.0 - cosmic_importance);
      t = gg_min(1.0, t);
      t = gg_max(0.0, t);
      strength = t * pattern->value;
      attenuation = 1.6;
    }
    else {
      strength = pattern->value;
      attenuation = 1.5;
    }

    /* Increase strength if we're computing escape influence. */
    if (q == &escape_influence && (pattern->class & CLASS_e))
      add_influence_source(pos, this_color, 20 * strength, attenuation, 
q);
    else
      add_influence_source(pos, this_color, strength, attenuation, q);

    DEBUG(DEBUG_INFLUENCE,
          "  low intensity influence source at %1m, strength %f, color %C\n",
          pos, strength, this_color);
    return;
  }


**********************************************


Stephane, disgusted








... /// ... \\\ ...