bug#13084: boyer_moore crashes with certain characters in the case table

[Kenichi Handa <handa@gnu.org>]

In article <831uewa9cq.fsf@gnu.org>, Eli Zaretskii <eliz@gnu.org> writes:

> In addition, I'd suggest that Handa-san (or someone else) takes a good
> look at the code that sets up the simple_translate table in
> boyer_moore, because the constants there, like 0200 and 0x3F, and all
> the talk about characters that belong "to the same charset and row"
> smell of pre-Unicode (a.k.a. "MULE") representation of characters.
> For now, I disabled boyer_moore for unibyte characters beyond 160,
> because my reading of the code is that simple_translate and the
> supporting code cannot handle that.  Maybe I'm wrong.

I have not yet checked the code, but what I remember is that
search_buffer checks the search string and decides which to
use; boyer_moore or simple_search.  If all equivalent
characters of all non-ASCII characters in the search string
are in the same character group, we can use boyer_moore.
Here, A and B belongs to the same character group iff A and
B has the same multibyte sequence except for the last byte.
In this condition, we should be able to use the table
simple_translate.

---
Kenichi Handa
handa@gnu.org