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

Re: Identifying remaining core font users



Le mercredi 11 novembre 2009 à 13:22 +0000, Richard W.M. Jones a écrit :

> I guess it falls to me (with Debian folk) to do this one, since
> upstream are unlikely to care enough to change this old, working code.
> 
> Here is the code at issue:
> 
> http://caml.inria.fr/cgi-bin/viewcvs.cgi/ocaml/trunk/otherlibs/graph/text.c?rev=6171&view=markup
> 
> Some questions since I know very little about this:
> 
> (1) Are there any recipes / guides / tutorials for changing basic X
> core fonts calls into whatever has replaced them?  Note that using a
> library like gtk is not an option.
> 
> (2) Will the replacement work on other Unix platforms (eg Solaris,
> AIX)?

The right person to ask this would be Behdad, as he's the Fedora/Red
Hat/upstream maintainer of most core components of our current text
stack. IIRC his advice last time I asked the question was to avoid
accessing fontconfig directly, but to pass through gtk2/pango, QT, or
pango-cairo in cairo (those are the three main ways to do text nowadays
and modern text requirements are complex enough they're all trying to
converge behind the scenes instead of using different implementations of
text libs. Too difficult to do alone, even for projects their size)

Note that pango-cairo in particular will work on recent versions of
other Unixes, and even on Windows and OSX. In fact OO.o is currently
migrating this way mainly for OSX compat.

If depending on this level of libraries is out of the question for you
I'd advise ripping the text parts from those modules. Text is much more
complex than just drawing a simple form like a triangle, it is getting
more complex every year (as new font format specs and encoding standards
revisions are released, and support is added to more minority languages
with strange requirements). I don't think you'll find maintaining text
support sustainable without depending on common text libs.
QT/GTK/Mozilla/OO.o tried, and concluded convergence was the only path.

But feel free to ask Behdad directly, I'm sure anything he tells you
will prove valuable.

Regards,

-- 
Nicolas Mailhot

Attachment: signature.asc
Description: Ceci est une partie de message numériquement signée


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