[ft-devel] Re: [ft] Bytecode hinter producing bad results, seems to be a regression

[Werner LEMBERG]

[About B/W hinted rendering of the font `Helvetica Neue LT Com 65
 Medium']

> I was testing this on 96dpi, with smoothing turned off entirely (if
> enabled, there's the option between 'standard' and 'cleartype').
> I've attached the image for the font you supplied.

With the help of Greg Hitchcock from Microsoft I was able to fix some
`bugs' in the FreeType raster and bytecode engine.  Regarding the
rasterizer, the MS people have implemented some special handling of
border cases which aren't documented (drop-out pixels outside of the
bbox are moved into the bbox; the bbox for B/W rasterizing is computed
using ordinary rounding, not flooring and ceiling), and within the
bytecode engine, there is an error in the specification of the MIRP
instruction.

There are still some differences I have yet to analyze, but in most
cases the output is identical now.

However, if you ignore the `gasp' table and instead rasterize this
font at the particular ppem values with the grayscale renderer (while
hinting is active), the results are still extremely bad.  This can't
be fixed.


    Werner