"Moe Elzubeir" <[EMAIL PROTECTED]>:
ME> Seeing that I know _nothing_ about the internals of XFree86, I would
ME> really appreciate alllllllllll the pointers you can offer. Ranging from
ME> pieces of code that affect this to documents and papers that can help.
1. look into include/X11/Xlib.h, and search for the definition of
XFontStruct. You will see that it contains a field called
``per_char'' which points at a linear vector of XCharStruct. Thus,
the per_char vector uses 12 bytes for every codepoint between
first_char and last_char.
2. look into the protocol definition. You will see that the QueryFont
request elicits a reply that contains a linear vector of XCharStruct.
Thus, the 12 * (last_char - first_char) overhead is not only memory
usage, but also protocol overhead.
Note, however, that (2) and, to a certain extent, (1) are solved by
Bruno's BIGFONT extension. Thus, the above is only partly true.
ME> This 'truncation' of the 10x20 for 'optimization' is seriously
ME> hampering our efforts to bring Arabic support on platforms where
ME> XFree86 runs.
The correct solution (TM) is to use Xft rather than the core protocol
for Arabic. You may want to get in touch with Owen Taylor and the
Pango project.
Another solution (hack) would be to provide an alternate font with the
Arabic glyphs. Call it e.g. 10x20ar.
Juliusz
_______________________________________________
I18n mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/i18n
