Re: case-table functions clobbering extra slots

[Dave Love]

Richard Stallman <address@hidden> writes:

>     In the case of sigma, it should be that both ? and ? upcase to ?, and
>     ? downcases to ?.  [I hope I got those right, typing unicodes without
>     being able to display them.  Obviously likewise for the iso8859-7
>     charset.]
>
> It should be possible to do that by setting up the upcase table and
> the downcase table explicitly to do those things.  I think the code
> that computes the other two tables can handle this case, but maybe it
> doesn't, so it would be useful to find out.

As far as I could tell, what I did originally worked, as below.  I'm
not sure that's what you're addressing, though.

> Handa, can you help here?
>
>       I set up entries in the case table interactively to get
>     that right and then found they were clobbered by subsequent
>     case-table.el functions being called.
>
> I can see that subsequent calls would clobber this, because
> they would clear out the downcase table.  They do that
> so as to force recomputation.

Yes.  I don't understand the recomputation business, not that I tried
very hard.

> Which case-table functions were called, and when?
> Were they called by part of Emacs setup, or by some specific
> library that you loaded, or what?

As far as I remember, after I tested the setting the case table
interactively with success, I put the same thing in the appropriate
place in characters.el and then wondered why it didn't work in the
dumped Emacs.  It's possible it could go at the end of characters.el,
but I don't know whether that would cause trouble and it wouldn't work
if anything subsequently called set-case- functions.  (In Emacs 22,
for instance, switching in and out of Turkish language mode does that
to treat `i' correctly.  It would do the same in Emacs 21 if
caseification could change the byte length, as required.)