Re: [ft] rendering of TrueType hints

[Patrick Thomas]

Hi Werner,

I've checked it with ftview and it appears to render the same as in FontForge. I used ftview -r 96 -m / -l n 27 Untitled.ttf, with n = 0,1,2,3,4,5,6. I toggled anti-aliasing off, forced auto-hinting off and glyph hinting on.

Is there a way to check what FreeType generates for the resulting gridfitted outline before the scan conversion is done?

I've attached a minimal version of the font with the glyph in question (forward slash).

Thank you,

Patrick

On Mon, Apr 21, 2014 at 11:21 PM, Werner LEMBERG <address@hidden> wrote:

> I'm using the following instructions in Xgridfit to turn off dropout
> control:
>
> <pre-program>
>     <set-dropout-control flags="0" threshold="0"/>
>     <set-dropout-type value="2"/>
> </pre-program>
>
> My understanding is that a pixel should then be rendered if and only
> if its center falls within or on the resulting gridfitted outline?

Yes, provided Xgridfit doesn't do other nasty things with the
bytecode.  You are checking the resulting instructions with a tool
like ttx, I guess...

> In the attached screenshot from FontForge the coordinates of the
> final gridfitted points are (2,5), (0,7), (10,17) and (12,15) at 27
> point, 96 dpi.  I would then expect that the pixels with centers
> (2.5, 5.5) and (10.5, 16.5), which fall on the resulting outline,
> would be rendered, but they are not.  The same relatively positioned
> pixels are also not rendered at other point sizes.

Similar to Xgridfit, I have no idea how FontForge exactly does its
grid rendering :-)

Please test with a FreeType demo program like ftview and check the
result with a magnifier.  Then you can be certain how FreeType renders
your font using its various rendering modes.


    Werner

Untitled.ttf
Description: application/font-ttf