On Wed, Aug 10, 2022 at 08:53:28AM -0700, Junio C Hamano wrote:
> Torsten Bögershausen <tboegi@xxxxxx> writes:
>
> > git log --stat
> > [snip]
> > Arger.txt | 1 +
> > Ärger.txt | 1 +
> > 2 files changed, 2 insertions(+)
> >
> > From this very first experiment I would suspect that we use
> > strlen() somewhere rather then utf8.c::git_gcwidth()
>
> Yeah, that does sound like the case, and quite honestly, knowing
> that the diffstat code is way older than unicode-width code, which
> was added by you in mid 2014, I am not all that surprised if we used
> to use strlen() throughout and we still do by mistake.
>
> Thanks for a doze of sanity.
Some 2 updates here:
- The strlen() needs a replacement.
It looks as if the following patch helps:
/* somewhere in diff.c */
static size_t screen_utf8_width(const char *start)
{
const char *cp = start;
size_t remain = strlen(start);
size_t width = 0;
while (remain) {
int n = utf8_width(&cp, &remain);
if (n < 0)
return strlen(start); /* not UTF-8 ? Use strlen() */
width += n;
}
return width;
}
@@ -2620,7 +2635,7 @@ static void show_stats(struct diffstat_t *data, struct diff_options *options)
continue;
}
fill_print_name(file);
- len = strlen(file->print_name);
+ len = screen_utf8_width(file->print_name);
if (max_len < len)
max_len = len;
@@ -2743,7 +2758,7 @@ static void show_stats(struct diffstat_t *data, struct diff_options *options)
* "scale" the filename
*/
len = name_width;
- name_len = strlen(name);
+ name_len = screen_utf8_width(name);
if (name_width < name_len) {
=====================================
Let's see if I can make a proper patch out of it.
The second problem, and I hoped it wasn't, seems to be related to what
you had digged out earlier.
>Sounds like a symptom observable when the width computed by
>utf8.c::git_gcwidth(), using the width table imported from
>unicode.org, and the width the terminal thinks each of the displayed
>character has, do not match (e.g. seen when ambiguous characters are
>involved, https://unicode.org/reports/tr11/#Ambiguous).
That needs a second patch, probably after some more digging,
how unicode is rendedered on the different systems
