Re: [ft] Regression in rendering quality with subpixel antialiasing

[Bernd Schubert]

Hello David,

> I must say I've never heard of Christian Marillat, or even know who Steve
> is, but, hey, any way to fix the situation would be appreciated :-)

oh, sorry. Christian Marillat provides inofficial archives for multimedia 
programs not provided officially by Debian. If you are interested see
http://www.debian-multimedia.org/

Steve Langasek is the present Debian maintainer of libfreetype6.

[... many many usefull information ]

> > Hmm, a 'ldd soffice.bin' shows me that it is not linked to libXft. So
> > either
> > my guess to the function call order is wrong or OpenOffice really calls
> > freetype functions without libXft.
> > Is there a typical function call to libfreetype? I would just like to
> > compare
> > what openoffice does differently.
>
> first, don't always trust the output of ldd, because some programs open
> some of the system libraries through explicit calls to the dynamic linker
> (i.e. dlopen) in order to support a variety of versions of a given library
> installed on the system. I know that at least FireFox and Thunderbird do
> that.

Yeah I know, for my (quantum chemical, not not console-based) programs I do 
the same and call the .so files plugins :)
Though, that time I was not thinking off it, soffice.bin is even linked 
against libld.so.

>
> does OpenOffice links to Cairo ?

Yes and it seems it even uses it for font rendering.

>
> Another point to note is that an application is free to change a fontconfig
> pattern freely, for example to enfore its own hinting style that may be
> different from the default in fonts.conf.

Yep, found it. They are presently hardcoding the font settings :(

http://qa.openoffice.org/issues/show_bug.cgi?id=64508

> Hope this helps,

Oh yes, your explanations did help indeed very much. At least I could identify 
the problem and come up with a small patch for OpenOffice patch restoring the 
old font rendering. Though, since its still hard coded, some people again 
won't like it.

http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=385798


Thanks a lot for all your help,
Bernd

-- 
Bernd Schubert
PCI / Theoretische Chemie
Universität Heidelberg
INF 229
69120 Heidelberg