On Sat, Apr 05, 2025 at 10:42:11AM +0000, Lucas Gabriel Vuotto wrote:
> On Sat, Apr 05, 2025 at 11:33:31AM +0200, Walter Alejandro Iglesias wrote:
> > On Sat, Apr 05, 2025 at 08:57:51AM +0000, Lucas Gabriel Vuotto wrote:
> > > On Sat, Apr 05, 2025 at 09:36:07AM +0200, Walter Alejandro Iglesias wrote:
> > > > Hi again Lucas,
> > > >
> > > > This time I paid a little more attention :-). Maybe I'm missing
> > > > something, but it seems to me that, in your patch, the skip_utf8_cont
> > > > variable is unnecessary.
> > > >
> > > > Anyway, at first I'd also tried doing something similar to what you
> > > > suggested, I still think it doesn't require so much fuss. Let's see if
> > > > I don't make any stupid mistakes with this new version of mine:
> > >
> > > anton pointed the same out in a private email and also requested
> > > regress modifications. The current patch is
> > >
> >
> > I don't want to waste your time giving me C lessons, but if it's not too
> > long to explain, I would appreciate it if Anton, you or anyone else
> > could tell me what the problem (regression?) could be with just removing
> > the conditional, as I did in my last example:
>
> domove is an internal function to vi.c, called exclusively by vi_cmd.
> It either calls it with the last param (sub) set to either 0 or 1. I
> believe one way you can run into the latter is by doing `ye`, for
> example, which is valid command (copy from cursor to end of word) that
> isn't a move command straight off the bat. I believe that, without the
> conditional, the `e` in `ye` will move the cursor, thing that shouldn't
> happen.
>
That doesn't happen. One thing I observed that happen after removing
the conditional is that when you paste the yanked text, the last byte of
the last UTF-8 character is eaten. The following solves that:
Index: vi.c
===================================================================
RCS file: /cvs/src/bin/ksh/vi.c,v
diff -u -p -r1.60 vi.c
--- vi.c 12 Mar 2021 02:10:25 -0000 1.60
+++ vi.c 5 Apr 2025 11:19:57 -0000
@@ -1195,8 +1195,7 @@ domove(int argcnt, const char *cmd, int
return -1;
ncursor = (*cmd == 'e' ? endword : Endword)(argcnt);
if (!sub)
- while (isu8cont((unsigned char)es->cbuf[--ncursor]))
- continue;
+ isu8cont((unsigned char)es->cbuf[--ncursor]);
break;
case 'f':
I discovered another issue (in the unpatched ksh). Given for example:
$ word €€€ word €€€
If you yank the first €€€ (with `ye`), then put your cursor in the last
character (the last € of the last word) and paste the text (hitting `p`),
the string gets corrupted.
Well, it seems that ksh vi mode UTF-8 support needs a lot of work. :-)
--
Walter
