On Thursday 13 October 2005 14:10, Eugeniy Meshcheryakov wrote: > It is somnething wrong with Arabic and Persian there. It looks like > there are isolated variants for some characters that should not be. There may be, but I have no idea as I have no idea how things should look. One possible cause is that AFAICT fontconfig makes a "priority list" with the order in which fonts should be scanned for characters. Currently, with the font FreeSans as default, this list is: $ fc-match --sort "FreeSans" FreeSans.ttf: "FreeSans" "Medium" FreeSansBold.ttf: "FreeSans" "Bold" kochi-gothic.ttf: "Kochi Gothic" "Regular" Gargi_1.7.ttf: "Gargi_1.7" "Medium" gbsn00lp.ttf: "AR PL SungtiL GB" "Regular" bsmi00lp.ttf: "AR PL Mingti2L Big5" "Reguler" FreeSerif.ttf: "FreeSerif" "Medium" FreeMono.ttf: "FreeMono" "Medium" ae_Tholoth.ttf: "ae_Tholoth" "Regular" chandas1-1.ttf: "Chandas" "Regular" sansu___.ttf: "Sans Uralic" "Regular" lohit_pa.ttf: "Lohit Punjabi" "Regular" Saab.ttf: "Saab" "Regular" UnDotum.ttf: "UnDotum" "Regular" FreeMonoBold.ttf: "FreeMono" "Bold" FreeSerifBold.ttf: "FreeSerif" "Bold" FreeSerifItalic.ttf: "FreeSerif" "Italic" As you can see, there is an arabic font in that set (ae_Tholoth), but fairly down the list. As I understand it, if any font higher in the list contains a character for a UTF8-code, that will be taken. The problem may be something else though, maybe shaping functions doing the wrong thing? Something to look into by font experts I guess. > Also it looks like one word is drawn with several fonts. For simplified > Chinese there is only one ideogram in translation, should be two. This is now fixed. See http://people.debian.org/~fjp/d-i/GTK_CJK.png. Cheers, FJP
Attachment:
pgpZGTYF6C5Wl.pgp
Description: PGP signature