[ft] Freetype 2.4.4 -> freetype 2.4.12: FT_Get_Advance changed behaviour?

[Gregor Mückl]

Hi!

I've upgraded a project of mine from using freetype 2.4.4 to freetype 
2.4.12 and now I experience totally broken text layouting (the layouting 
code is my own). Horizontal spacings between characters are now either 
far too large or far too small, depending on the font. In one case the 
glyph distances are at least double what they should be. None of the 
tested fonts give anything close to believable reasults. Linking against 
the old version of freetype without any further code changes fixes this 
issue.

I seem to get vastly different values for the horizontal advance. If I 
uncommend the if block around FT_HAS_KERNING in the code below, the 
result does not change noticably. This is the code I use to compute the 
horizontal layout of a string of glyphs:

typedef struct {
        unsigned int glyph;
        float posX;
        float posY;
        int lineNumber;
        int font;
        int size;
} TL_GlyphPosition;

void _TL_ActivateFontSize(int font, int size)
{
        if(TL_Globals.fonts[font].currentSize==size) {
                return;
        }

        FT_Set_Pixel_Sizes(TL_Globals.fonts[font].face,0,size);
        TL_Globals.fonts[font].currentSize=size;

        int height=TL_Globals.fonts[font].face->size->metrics.height;
        
if(height<TL_Globals.fonts[font].face->size->metrics.ascender+TL_Globals.fonts[font].face->size->metrics.descender)
 {
	 
height=TL_Globals.fonts[font].face->size->metrics.ascender+TL_Globals.fonts[font].face->size->metrics.descender;
        }
        TL_Globals.fonts[font].currentLineHeight=height;
}

int _TL_CalculateRunWidth(TL_GlyphPosition *glyphs, int length, bool 
firstInLine)
{
        int width=0;
        FT_Vector kerning;
        FT_Fixed advance;
        int i;

        for(i=0;i<length;i++) {
                glyphs[i].posX=(float)width;
                glyphs[i].posY=0;

                // get character width of current character
                _TL_ActivateFontSize(glyphs[i].font,glyphs[i].size);
	 
FT_Get_Advance(TL_Globals.fonts[glyphs[i].font].face,glyphs[i].glyph,FT_LOAD_DEFAULT,&advance);
	 
advance=(FT_Fixed)(64*advance/(float)TL_Globals.fonts[glyphs[i].font].face->size->metrics.x_scale);
                width+=advance;

                // check whether this is the first character in the string/line
                if((i==0 && !firstInLine) || i!=0) {
			// respect proper kerning (if possible, i.e. both glyphs are from the 
same font and size)
			if(FT_HAS_KERNING(TL_Globals.fonts[glyphs[i].font].face) && 
glyphs[i-1].font==glyphs[i].font && glyphs[i-1].size==glyphs[i].size) {
                                // Note: glyphs[i-1] may point beyond start of 
passed glyphs array,
				// but it's only a section of a larger string, so this unchecked i-1 
is fine
			 
FT_Get_Kerning(TL_Globals.fonts[glyphs[i].font].face,glyphs[i-1].glyph,glyphs[i].glyph,FT_KERNING_UNFITTED,&kerning);
                                width+=kerning.x;
                        }
                }

        }

        return width;
}

Am I doing something obviously wrong in this snippet? Was there an 
interface change in freetype that I should be aware of?

Regards,
Gregor