On 12.02.2012 22:59, Bruno Haible wrote: > Hi Vladimir, > > Thank you for the proposed patch. > >> As already reported several years ago > I cannot find it in my archives. Maybe that discussion already contained > some useful thoughts or arguments? Can you please point me to it? http://lists.gnu.org/archive/html/bug-tar/2009-09/msg00008.html >> argp counts bytes even when >> actually what matters is the display length. This patch improves the >> situation by counting only leading and standalone UTF-8 bytes. It >> doesn't handle the double-width characters like Chinese sinograms > A program that needs to consider display length - for example for > line wrapping - should > 1) work with any locale encoding. Don't assume that the locale encoding > is UTF-8. > 2) work with Chinese ideographs correctly, like it should also work > with Russian (single-width) letters. Here you go. Tested on Cyrillic. Haven't tested with Chinese. > > The easiest way to satisfy these two requirements is to base the code on > either > * the function mbswidth (gnulib module mbswidth) and possibly also mbiter > or mbuiter, or > * the gnulib module unilbrk/ulc-width-linebreaks, it contains a complete > line-breaking algorithm. > > Can you rewrite your patch to this effect? > > Also, such tricky issues should be checked in the test suite. Can you > please also provide a test program, some input data, and the expected > output for this data? We can then turn it into a gnulib test. I can strip down one of our programs to keep just --help when time permits. > Bruno > >
-- Regards Vladimir 'φ-coder/phcoder' Serbinenko
=== modified file 'grub-core/gnulib/argp-fmtstream.c'
--- grub-core/gnulib/argp-fmtstream.c 2010-04-02 22:45:01 +0000
+++ grub-core/gnulib/argp-fmtstream.c 2012-02-12 22:37:14 +0000
@@ -29,6 +29,7 @@
#include <errno.h>
#include <stdarg.h>
#include <ctype.h>
+#include <wchar.h>
#include "argp-fmtstream.h"
#include "argp-namefrob.h"
@@ -116,6 +117,54 @@
#endif
#endif
+
+size_t
+__argp_get_display_len (char *beg, char *end)
+{
+ char *ptr;
+ size_t r = 0;
+ mbstate_t ps;
+
+ memset (&ps, 0, sizeof (ps));
+
+ for (ptr = beg; ptr < end; )
+ {
+ wchar_t wc;
+ size_t s;
+
+ s = mbrtowc (&wc, ptr, end - ptr, &ps);
+ if (s == (size_t) -1)
+ break;
+ r += wcwidth (wc);
+ ptr += s;
+ }
+ return r;
+}
+
+static inline char *
+add_length (char *ptr, char *end, size_t l)
+{
+ mbstate_t ps;
+
+ memset (&ps, 0, sizeof (ps));
+
+ while (ptr < end)
+ {
+ wchar_t wc;
+ size_t s, k;
+
+ s = mbrtowc (&wc, ptr, end - ptr, &ps);
+ if (s == (size_t) -1)
+ break;
+ k = wcwidth (wc);
+ if (k >= l)
+ break;
+ l -= k;
+ ptr += s;
+ }
+ return ptr;
+}
+
/* Process FS's buffer so that line wrapping is done from POINT_OFFS to the
end of its buffer. This code is mostly from glibc stdio/linewrap.c. */
void
@@ -168,14 +217,15 @@
if (!nl)
{
+ size_t display_len = __argp_get_display_len (buf, fs->p);
/* The buffer ends in a partial line. */
- if (fs->point_col + len < fs->rmargin)
+ if (fs->point_col + display_len < fs->rmargin)
{
/* The remaining buffer text is a partial line and fits
within the maximum line width. Advance point for the
characters to be written and stop scanning. */
- fs->point_col += len;
+ fs->point_col += display_len;
break;
}
else
@@ -183,14 +233,18 @@
the end of the buffer. */
nl = fs->p;
}
- else if (fs->point_col + (nl - buf) < (ssize_t) fs->rmargin)
- {
- /* The buffer contains a full line that fits within the maximum
- line width. Reset point and scan the next line. */
- fs->point_col = 0;
- buf = nl + 1;
- continue;
- }
+ else
+ {
+ size_t display_len = __argp_get_display_len (buf, nl);
+ if (display_len < (ssize_t) fs->rmargin)
+ {
+ /* The buffer contains a full line that fits within the maximum
+ line width. Reset point and scan the next line. */
+ fs->point_col = 0;
+ buf = nl + 1;
+ continue;
+ }
+ }
/* This line is too long. */
r = fs->rmargin - 1;
@@ -226,7 +280,7 @@
char *p, *nextline;
int i;
- p = buf + (r + 1 - fs->point_col);
+ p = add_length (buf, fs->p, (r + 1 - fs->point_col));
while (p >= buf && !isblank ((unsigned char) *p))
--p;
nextline = p + 1; /* This will begin the next line. */
@@ -244,7 +298,7 @@
{
/* A single word that is greater than the maximum line width.
Oh well. Put it on an overlong line by itself. */
- p = buf + (r + 1 - fs->point_col);
+ p = add_length (buf, fs->p, (r + 1 - fs->point_col));
/* Find the end of the long word. */
if (p < nl)
do
@@ -278,7 +332,7 @@
&& fs->p > nextline)
{
/* The margin needs more blanks than we removed. */
- if (fs->end - fs->p > fs->wmargin + 1)
+ if (__argp_get_display_len (fs->p, fs->end) > fs->wmargin + 1)
/* Make some space for them. */
{
size_t mv = fs->p - nextline;
=== modified file 'grub-core/gnulib/argp-fmtstream.h'
--- grub-core/gnulib/argp-fmtstream.h 2010-04-02 22:45:01 +0000
+++ grub-core/gnulib/argp-fmtstream.h 2012-02-12 22:18:14 +0000
@@ -335,6 +335,9 @@
return __fs->point_col >= 0 ? __fs->point_col : 0;
}
+size_t
+__argp_get_display_len (char *beg, char *end);
+
#if !_LIBC
#undef __argp_fmtstream_putc
#undef __argp_fmtstream_puts
=== modified file 'grub-core/gnulib/argp-help.c'
--- grub-core/gnulib/argp-help.c 2010-04-02 22:45:01 +0000
+++ grub-core/gnulib/argp-help.c 2012-02-12 22:19:05 +0000
@@ -1448,7 +1448,7 @@
/* Manually do line wrapping so that it (probably) won't get wrapped at
any embedded spaces. */
- space (stream, 1 + nl - cp);
+ space (stream, 1 + __argp_get_display_len (cp, nl));
__argp_fmtstream_write (stream, cp, nl - cp);
}
signature.asc
Description: OpenPGP digital signature
