Re: [Devel] xadvance descrepancy between bitmpats & glyphs?

[David Turner]

Hello Ken,

Ken Hancock wrote:
> I'm adding code to cache transformed bitmaps, among them italics and
> stretched.
>
> In my iFTFaceRequester I call FT_Set_Transform() to set a global transform
> for any Face_ID which should be transformed.

This is sub-optimal, since you're going to create several copies of the 
same FT_Face, except they'll have distinct transform values.

You could also map each FT_FaceID to a single FT_Face with no transform,
then write custom image and bitmap caches that would call 
FT_Set_Transform both before and after FT_Load_Glyph to set/unset the 
transformation matrix.

You certainly know that the cache isn't completely stable, but you can
have a look at src/cache/ftcbasic.c for an inspiration on how to write
these custom caches.

> I then have two routines which use:
> FTC_ImageCache_Lookup() to get the glyph->advance.x for line-layout and
> FTC_SBitCache_Lookup() to get the sbit->xadvance for blitting

> However, it looks like the advances are computed inconsistently:
>
> In FT_Load_Glyph (irrelevant lines omitted):
>
>     slot->advance.x = slot->metrics.horiAdvance;
>
>     if ( ( load_flags & FT_LOAD_IGNORE_TRANSFORM ) == 0 )
>     {
>         /* transform advance */
>         FT_Vector_Transform( &slot->advance, &internal->transform_matrix );
>     }  
>
> In ftc_snode_load (irrelevant lines omitted):
>
>       xadvance = ( slot->metrics.horiAdvance + 32 ) >> 6;
>
>
> slot->metrics.horiAdvance is UNTRANSFORMED
>
> Shouldn't ftc_snode_load use:
>
>           xadvance = (slot->advance.x + 32 ) >> 6;
>           yadvance = (slot->advance.y + 32 ) >> 6;

That's right, I'll commit the fix to the CVS shortly.

Thanks for everything,

- David Turner
- The FreeType Project  (www.freetype.org)


> Thanks in advance!
>
> Ken
>
> _______________________________________________
> Devel mailing list
> address@hidden
> http://www.freetype.org/mailman/listinfo/devel