[Uim] Re: GTK+ compose table (bug #377808)

[Jan Willem Stumpel]

Etsushi Kato wrote:

> To use GTK+ immodule installed in /usr/local, please check your
> /etc/gtk-2.0/gtk.immodules file (or equivalent) to have 
> "/usr/local/lib/gtk-2.0/2.*.0/immodules/im-uim.so" (which can
> be generated with gtk-query-immodules-2.0).

This works! Now I get full X11 compose sequences in GTK+ apps with
MODIFIERS=@im=uim and GTK_IM_MODULE=uim (instad of [..]=xim), and
of course now Mozilla can no longer be crashed by "mis-hitting"
keys (the original subject of Debian bug
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=377808).
Congratulations!

OK, back to Compose again (which was the original subject of our
conversation).

The problem of the submitter of bug
https://bugs.freedesktop.org/show_bug.cgi?id=8118 clearly is:

> So GTK+ lacks some of the sequences which X11 has, and has some
> extra sequences like "Compose a a", and even has contradicted
> sequences. 

> [..]

> And I'm now thinking about #3 (though having troubles with some 
> contradicted sequences as you described...) and asking xorg
> about the existence of the standard specification about Compose
> sequence on X desktop at 
> http://lists.freedesktop.org/archives/xorg/2006-September/017874.html,
> but have got no response yet...

In the meantime I exchanged mail with Owen Taylor (author of the
GTK+ compose table). He said that (at that time, of course) there
were inconsistencies inside the X11 Compose file itself. So e.g.
both ª and ā had the same Compose sequence, and GTK+ just had to
make a choice.

But now the X11/Compose file has been cleaned up (I think), and
also it has been extended. A lot of work on the X11/Compose file
was apparently done by Simos Xenitellis (well-known from the
linux-utf-8 list).

[hello Simos! Please see
http://lists.freedesktop.org/archives/uim/2006-September/thread.html
for earlier installments of this discussion]

My impression now is that the GTK+ Compose mechanism, because it
is hard-wired in a C file, is "frozen in time" (i.e. Xfree86 4.0
time). How we (i.e. uim, xorg, GTK+) have to deal with this,
should be discussed. I still think the best way would be to follow
the xorg (X/Compose) road, but still a method must be found to
avoid surprises for GTK+ users (like the original bug submitter).

Regards, Jan