Hello Originally function get_word operates with signed char and compares it numerically with SPACE to check, if this character may be part of word. It is true only for 7-bit characters, accented letters (with most significant bit set) will be treated as negative and condition "greater than SPACE" will be always false. In effect, get_word can't recognize word starting with accented letter. You can simply reproduce this bug. If you are non-english speaker and you have console properly working in ISO-1 or ISO-2, simply echo some sentence in your native language with some word inside sentence starting with accented letter. For Polish language in ISO-2 please enter: echo stary pies ?apie lisa If you have keyboard not configured for non-ascii letters, you must have console working in ISO-1 or ISO-2, and then enter something like: echo -e "Linux \xfcber alles" For Cyrillic scripts simply echo some sentence. Then press 'previous line' and try to navigate with previous/next word. You will never hear word starting with accented letter, but previous word instead (for example 'linux' instead of '?ber'). For cyrillic you will always hear first word My patch simply changes signed char to unsigned char. It's not the best solution, as get_word should be simplified (in fact, we do not need to check 'can the character be part of word' for character inside word), but seems to be working perfectly with ISO-2. I also make very short test with ISO-5 (using builtin Milena cyrillic translator instead of russian synthesizer) and also I had no problem with navigation by word. Would you take a look on attached patch? ethanak -- http://milena.polip.com/ - Pa pa, Ivonko! -------------- next part -------------- A non-text attachment was scrubbed... Name: getword.patch Type: text/x-patch Size: 1553 bytes Desc: not available URL: <http://linux-speakup.org/pipermail/speakup/attachments/20120708/d258aefb/attachment.bin>
