Re: Unicode changes in CVS. Please test.

[Pawel Kot]

On Fri, 27 Sep 2002, Markus Plail wrote:

> Please test the new unicode code in CVS. I succesfully wrote chinese
> characters to my phonebook (which it obviously couldn't display) and
> read them back. Hu Gang's test also worked fine.
> So please test SMS sending and all other stuff you can think of.

More thoughts on the fix. For now just on decoding (encoding seems also a
bit wrong).
unsigned int char_decode_unicode(...)
{
        int i, length = 0, pos = 0;

        for (i = 0; i < len / 2; i++) {
                length = wctomb(dest, (src[i * 2] << 8) | src[(i * 2) + 1]);
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
The problem here is that we assume that the wide character is 2 bytes
wide. I'm afraid it is more complicated. We should do something like:
        d = dest;
        for (i = 0; i < len; i++) {
                wchar = 0;
                old = i;
                for (j = 0, j < MB_CUR_MAX && length == -1 && i < len; j++) {
                        wchar = (wchar << 8) | src[i++];
                        length = wctomb(dest, wchar);
                }
                if (length == -1) {
                        i = old;
                        *dest = '?';
                        length = 1;
                }
                dest += length;
        }
        return dest - d;

Description:
we may have different width of the characters. If wctobm() fails, we
should try to add one more character. If it fails with the maximal length,
we rewind to its length 1 and insert '?'.
Again, this is completly untested, do think about this and then test it
:-)

pkot
-- 
mailto:address@hidden :: mailto:address@hidden
http://kt.linuxnews.pl/ :: Kernel Traffic po polsku