Re: Composing Hebrew diacriticals

[Eli Zaretskii]

> From: Kenichi Handa <address@hidden>
> Cc: address@hidden, address@hidden
> Date: Wed, 12 May 2010 17:04:09 +0900
> 
> I found that cursor positioning sometimes failed even for a text
> that doesn't involve composition.  For instance, when I have this
> text in a buffer or bidi-display-reordering t,
> ----------
> A א B.
> בג
> ----------
> and point is at (i.e. before) "A", typing C-e moves point to
> "ג".
> 
> This happens with the vanilla trunk code.

Yes, cursor motion is a constant source of trouble with bidi
reordering.  Too many places assume strictly unidirectional display
and rely on the fact that the first (leftmost) character of line N
follows the last character of line N-1 in buffer's logical order.  It
takes time to learn all of the hidden assumptions and fix them.  C-a
and C-e are even trickier than C-f and C-b, because they move by
buffer positions assuming that screen positions correspond to those
1:1.

I already refactored the related functions twice, and recently came to
a conclusion that I need to refactor them yet again, sigh...

Thanks for the test case, I will add it to my other tests.