Re: OpenType Patches

[Eric Mader <mader jtcsv com>]

At 09:07 AM 9/6/2002, Owen Taylor wrote:
> Since taking all the intermediate glyphs into account shouldn't be
> hard, I'd rather we did that then count on all the intermediate
> widths being zero.

Done.

> > > > +           glyphs->glyphs[i].geometry.x_offset +=
> > > PANGO_UNITS_26_6(x_pos) - x_adj;
> > > > +           glyphs->glyphs[i].geometry.y_offset += 
> PANGO_UNITS_26_6(y_pos);
> > > >
> > > > -           for (j = i - outgpos[i].back; j < i; j++)
> > > > -             glyphs->glyphs[i].geometry.x_offset -=
> > > glyphs->glyphs[j].geometry.width;
> > > > -
> > > >             if (outgpos[i].new_advance)
> > > >               /* Can't set new x offset for marks, so just make
> > > sure not to increase it.
> > > >                  Can do better than this by playing with ->x_offset. */
> > > > -             glyphs->glyphs[i].geometry.width = 0;
> > > > +             glyphs->glyphs[i].geometry.width  =
> > > PANGO_UNITS_26_6(outgpos[i].x_advance);
> > >
> > >Do you understand the FIXME here? I don't, really. The comment applies
> > >to the '= 0', so it certainly isn't literally relevant with your changes.
> > >
> > > >             else
> > > >               glyphs->glyphs[i].geometry.width +=
> > > PANGO_UNITS_26_6(outgpos[i].x_advance);
> > > >           }
> >
> > No, I don't understand the comment here. I was tempted to remove it,
> > but figured I should leave it in case someone else knows what it
> > means. My understanding of the new_advance flag is that it means that
> > x_advance is a new advance width rather than an adjustment to the
> > existing advance width. (This is done because it makes some of the
> > GPOS calculations a bit simpler)
>
> Lets just kill the comment then; an inaccurate comment is worse than
> no comment.

Done. Patch is enclosed.

> Regards,
>                                         Owen

Regards,
Eric

Index: pango-ot-ruleset.c
===================================================================
RCS file: /cvs/gnome/pango/pango/opentype/pango-ot-ruleset.c,v
retrieving revision 1.4
diff -u -p -r1.4 pango-ot-ruleset.c
--- pango-ot-ruleset.c	7 May 2002 20:38:50 -0000	1.4
+++ pango-ot-ruleset.c	6 Sep 2002 22:56:13 -0000
@@ -247,18 +247,26 @@ pango_ot_ruleset_shape (PangoOTRuleset  
 	{
 	  for (i = 0; i < result_string->length; i++)
 	    {
+	      FT_Pos x_pos = outgpos[i].x_pos;
+	      FT_Pos y_pos = outgpos[i].y_pos;
+	      int back = i;
 	      int j;
 
-	      glyphs->glyphs[i].geometry.x_offset += PANGO_UNITS_26_6 (outgpos[i].x_pos);
-	      glyphs->glyphs[i].geometry.y_offset += PANGO_UNITS_26_6 (outgpos[i].y_pos);
+	      while (outgpos[back].back != 0)
+		{
+		  back  -= outgpos[back].back;
+		  x_pos += outgpos[back].x_pos;
+		  y_pos += outgpos[back].y_pos;
+		}
+
+	      for (j = back; j < i; j += 1)
+	        glyphs->glyphs[i].geometry.x_offset -= glyphs->glyphs[j].geometry.width;
+
+	      glyphs->glyphs[i].geometry.x_offset += PANGO_UNITS_26_6(x_pos);
+	      glyphs->glyphs[i].geometry.y_offset += PANGO_UNITS_26_6(y_pos);
 
-	      for (j = i - outgpos[i].back; j < i; j++)
-		glyphs->glyphs[i].geometry.x_offset -= glyphs->glyphs[j].geometry.width;
-	    
 	      if (outgpos[i].new_advance)
-		/* Can't set new x offset for marks, so just make sure not to increase it.
-		   Can do better than this by playing with ->x_offset. */
-		glyphs->glyphs[i].geometry.width = 0;
+		glyphs->glyphs[i].geometry.width  = PANGO_UNITS_26_6(outgpos[i].x_advance);
 	      else
 		glyphs->glyphs[i].geometry.width += PANGO_UNITS_26_6(outgpos[i].x_advance);
 	    }
@@ -272,7 +280,6 @@ pango_ot_ruleset_shape (PangoOTRuleset  
   last_cluster = -1;
   for (i = 0; i < result_string->length; i++)
     {
-      glyphs->glyphs[i].glyph = result_string->string[i];
       glyphs->glyphs[i].glyph = result_string->string[i];
 
       glyphs->log_clusters[i] = result_string->logClusters[i];