Hi,
I saw this old bug report yesterday whilst looking for the convention
for bug subjects used in dpkg.
* Changwoo Ryu [2004-07-04 06:57 +0900]:
> In my ko_KR.UTF-8 locale and with the new Korean translation added
> (#254590),
> dpkg-query --list prints each fields in wrong columns (see attached
> dpkg-list-before.png file).
>
>
> This is because dpkg assumes that a string's length (strlen) is the
> width it occupies
> in text terminals, which is not true in UTF-8 locales. (A Korean hangul
> character
> is 3 bytes long but width is 2.) I fixed this problem by adjusting the
> string lengths
> in *printf() formats according to the their actual widths in text
> terminal (see the
> attached patch).
>
> With attached patch applied, it displays each fiels with Hangul
> characters correctly.
> (attached dpkg-list-after.png file)
Just for the record, I wrote a similar function as used in the patch to
align the columns for (non-dpkg) --help output. I wonder if the rest of
the patch can be simplified. Maybe I'll have a look at this during the
next release cycle.
The code of my function is:
static size_t
mbswidth(const char *s)
{
#if ENABLE_NLS
size_t rv, len, wlen;
wchar_t *wstr;
int wcsw;
rv = len = strlen(s);
if ((wstr = (wchar_t *)malloc((len + 1) * sizeof(wchar_t))) != NULL
&& (wlen = mbstowcs(wstr, s, len + 1)) != (size_t)-1
&& (wcsw = wcswidth(wstr, wlen)) >= 0)
rv = (size_t)wcsw;
free(wstr);
return rv;
#else
return strlen(s);
#endif
}
Unlike the function in the patch with the same name, mine returns
strlen(s) as fallback on errors (for example ENOMEM). Both are not
thread-save nor fast when used in a loop. Mine assumes that strlen(s)
>= mbstowcs(NULL,s,0). Both require POSIX 2001 + XSI extensions.
Instead of ENABLE_NLS, dpkg would of course use HAVE_MBSTOWCS or
similar.
Regards
Carsten
--
To UNSUBSCRIBE, email to debian-dpkg-bugs-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
