[PATCH v2 7/7] gdi32: Improve the determine code whether the chracter is the fullwidth.

[Akihiro Sagawa]

On Wed, 30 Jan 2019 02:12:41 +0900, Byeongsik Jeon wrote:
> Signed-off-by: Byeongsik Jeon <bsjeon at hanmail.net>
> ---
> At the MS Korean dual-width fonts embedded bitmap,
> full-width character's advance width is smaller than expected in wine.
> We can solve this problem by slightly changing the logic.
> 
> Ex.
> gulim.ttc:GulimChe EM=1024 AvgWidth=512
> ppem=19px half-width character advance = 10px
> ppem=19px full-width character advance = 18px
> But, Wine expects 19px.
> 
>  dlls/gdi32/freetype.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/dlls/gdi32/freetype.c b/dlls/gdi32/freetype.c
> index bddcd56837..cb804f8e5c 100644
> --- a/dlls/gdi32/freetype.c
> +++ b/dlls/gdi32/freetype.c
> @@ -6839,9 +6839,7 @@ static FT_Vector get_advance_metric(GdiFont *incoming_font, GdiFont *font,
>          em_scale = MulDiv(incoming_font->ppem, 1 << 16,
>                            incoming_font->ft_face->units_per_EM);
>          avg_advance = pFT_MulFix(incoming_font->ntmAvgWidth, em_scale);
> -        fixed_pitch_full = (avg_advance > 0 &&
> -                            (base_advance + 63) >> 6 ==
> -                            pFT_MulFix(incoming_font->ntmAvgWidth*2, em_scale));
> +        fixed_pitch_full = avg_advance > 0 && (base_advance + 63) >> 6 != avg_advance;
>          if (fixed_pitch_full && !transMat)
>              adv.x = (avg_advance * 2) << 6;


In my opinion, why don't we test the following condition for fixed_pitch_full?
fixed_pitch_full = avg_advance > 0
    && (base_advance + 63) >> 6
        >= pFT_MulFix(MulDiv(incoming_font->ntmAvgWidth, x, y), em_scale);
The x and y are heuristic ratio. I think they will be 1.5~1.8.

Because base_advance can be based on linked font, i.e. another font
replaced by get_glyph_index_linked(), the glyph might be a proportional
typeface. In that case, "(base_advance + 63) >> 6 != avg_advance" is
very loose restriction. A narrow character may be treated as a
Full-width character in that case.

Regards,
Akihiro Sagawa