Thomas Dickey-san wrote (12/25/08 11:11 PM):
On Thu, 25 Dec 2008, Thomas Dickey wrote:
@@ -952,6 +952,20 @@ int UCTransUniCharStr(char *outbuf,
HTSprintf0(&tocode, "%s//TRANSLIT",
LYCharSet_UC[charset_out].MIMEname);
cd = iconv_open(tocode, "UTF-16BE");
FREE(tocode)
+ if (cd == (iconv_t)-1) {
+ HTSprintf0(&tocode, "%s", LYCharSet_UC[charset_out].MIMEname);
I'm puzzled here - since TRANSLIT is less stringent, I would expect
that the check would be first made on the exact translation rather
than on the transliteration. Any comments?
looking a little further, I see that the issue is that TRANSLIT is
a glibc extension, so it's tried first in case it works. (google didn't
show much comment about whether it's more/less likely to succeed than the
plain POSIX version).
Yes, you're right.
This means Solaris iconv always failed the "$ENCODING//TRANSLIT".
So if iconv_open("$ENCODING, "UTF-16BE") is added without "//TRANSLIT", it
works on none GNU iconv.
> - if ((strcmp(LYCharSet_UC[charset_out].MIMEname, "shift_jis") == 0) ||
> - (strcmp(LYCharSet_UC[charset_out].MIMEname, "euc-jp") == 0)) {
> + if (LYCharSet_UC[charset_out].codepage == 0 &&
> + LYCharSet_UC[charset_out].codepoints == 0) {
> iconv_t cd;
This means I'd like to use iconv for CJK but not Japanese only.
_______________________________________________
Lynx-dev mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/lynx-dev
