freetype-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Devel] inaccurate letter y-position


From: Werner LEMBERG
Subject: Re: [Devel] inaccurate letter y-position
Date: Sun, 25 Jul 2004 00:48:43 +0200 (CEST)

> > > since changing from freetype 2.1.8 to 2.1.9 I discovered an
> > > inaccuracy in the y-positioning of single letters when the
> > > patented or unpatended hinting is activated.  The effect is
> > > obviously recognizable - so I suppose it could be a fault.
> > 
> > I can't see a problem (testing with ArialNb.TTF from 1998 -- I
> > think I don't have Futura Light Condensed).  Please try the
> > current development snapshot and send small images if the problems
> > remain.
>
> The Arial font shows the effect only at the letters "X" and "h"
> (e.g.  sizes 9, 10 and 20 pts).  Other fonts have even more such
> faces.
> 
> To be more exact: This effect occured already in freetype 2.1.8 after 
> 2003-11-27.
> 
> example1.c (angle set to 0 degrees, font "Arial Narrow Bold" 20 pts)
> 
> freetype 2.1.7:
> 
>            **                ***     ***
>            **                +**+   +**+
>            **                 +**+ +**+
>            **                 +**+ +**+
> ***   *** **+****+ ***   ***  +**+**+  ***   ***
> +**+ +**+ ******** +**+ +**+   *****   +**+ +**+
>   +**+**+  ***+++**  +**+**+    +***+    +**+**+
>    *****   **+  +**   *****     +***+     *****
>    +***+   **+  +**   +***+     *****     +***+
>    +***+   **    **   +***+    +**+**+    +***+
>    *****   **    **   *****   +**+ +**+   *****
>   +**+**+  **    **  +**+**+  +**+ +**+  +**+**+
> +**+ +**+ **    ** +**+ +**++**+   +**++**+ +**+
> ***   *** **    ** ***   ******     ******   ***
> 
> freetype 2.1.9:
> 
>            **                ***     ***
>            **                +**+   +**+
>            **                 +**   **+
>            **                 +**+ +**+
> ***   *** **+****+ ***   ***  +**+**+  ***   ***
> +**+ +**+ ******** +**+ +**+   *****   +**+ +**+
>   +**+**+  ***++***  +**+**+    +***+    +**+**+
>    *****   **+  +**   *****     +***+     *****
>    +***+   **+  +**   +***+     +***+     +***+
>    +***+   **    **   +***+    +**+**+    +***+
>    *****   **    **   *****    *** ***    *****
>   +**+**+  **    **  +**+**+  +**+ +**+  +**+**+
> +**+ +**+ **    ** +**+ +**++**+   +**++**+ +**+
> ***   *** **    ** ***   ******     **+***   ***
>            ++    ++          +++     +++

Thanks, no I see the problem (for all others: Try `ftstring -m xhxXx 7
arialnb.ttf').  I think it isn't a bug at all.  Reasons:

  . The hints in the font are not intended for anti-aliased rendering.
    If you switch it off, all sizes render perfectly.  Looking at the
    instruction dump created with TTX, the font calls GETINFO[] only
    to test for rotation (if I interpret the code correctly), but not
    for grayscale rendering.  Have you actually compared the FreeType
    results with Windows output?

  . If you use the autohinter, the results are good also.

  . Rogier van Dalen has fixed some bugs in the TT interpreter which
    makes your special case worse.

My conclusion is that you must not use native TrueType hinting for
anti-aliased rendering.  While George has recently added support for
handling vertical phantom points, we still don't have code for
rendering natively hinted grayscale.  Volunteers welcome.


    Werner



reply via email to

[Prev in Thread] Current Thread [Next in Thread]