Hi,
You have interesting points. Since this is continuation of the
discussion on bug report, I will CC to original bug report and
Awashiro-san who is scim-anthy maintainer and who is also concerned
about how all these fit together.
(I have touched-up your statement in quote to be suitable for the public
listing.)
On Sun, May 21, 2006 at 11:12:39AM +0200, Jan Willem Stumpel wrote:
> I think that in a UTF-8 environment in Linux, GTK_IM_MODULE must
> always be xim. If it has this value, all characters can be entered
> in all programs. If it has another value (like uim), some
> characters cannot be entered in Mozilla.
GTK IM Modules interface are still buggier than XIM interface I guess.
Ming's SCIM also recommend to use GTK_IM_MODULE=xim .
> To test this: try to enter ŏ (o with breve) by means of
> Compose-U-o. This works in all programs (xterm, Openoffice, QT
> programs, Mozilla) when GTK_IM_MODULE=xim, but it does not work in
> Mozilla when GTK_IM_MODULE=uim. You could also try, for instance,
> ® ("registered" sign, made by Compose-o-r).
I do not have compose key here en_US. Maybe I better rewad your web
and lean. http://jw-stumpel.nl/stestu That is good place.
> So I have "export GTK_IM_MODULE=xim" in /etc/environment. IMHO
> this should be a Debian default; but I do not know where I should
> talk about this issue, as there seems to be no Debian mailing list
> where UTF-8 and "m17n" issues are being discussed.
Not anything I know, except [EMAIL PROTECTED] is good but unless we
are prepared for long discussion...
> But because of this entry in /etc/environment, all the new
> versions of im-switch (including 1.10) fail, because of the test:
>
> # Check other configuration has not configured IM
> if [ -z "$XIM" -a -z "$XIM_PROGRAM" -a -z "$XIM_ARGS" \
> -a -z #$XMODIFIERS" \
> -a -z "$GTK_IM_MODULE" -a -z "$QT_IM_MODULE" ]; then
>
> This test fails because GTK_IM_MODULE is already set. Now the
> whole 90im-switch script does nothing.
>
> I do not understand why you put this test in.
If someone set any of these values, he knows what he is doing and he
does not want to be interfared by im-switch. There was similar code
before I hacking. It seemed they did it to solve conflict with m17n-env
package behavior. That is solved now since m17n-package does not touch
IM any more.
> If someone (user or
> administrator) has selected an IME using im-switch, then surely he
> wants that IME to be used, not whatever was the default before?
> The older versions (1.5 and older) were better in this respect.
As I reread source of 1.4, if any of IM variable was set before this
package, that value overides im-switch script. Yes it was better.
I will make change.
[...]
> IMHO the safest is to always export GTK_IM_MODULE=xim
> ("hard-wired") in 90im-switch. It never does any harm, and it is
> necessary for good keyboard input in Mozilla.
That has varied opinion. Most of the problem has been solved for SCIM
since the hook scripts are GTK_IM_MODULE=xim for SCIM. I still have
GTK_IM_MODULE=uim in UIM side. (If there is compelling discussion, I
will NMU again. Awashiro-san (scim-anthy maintainer) thinks immodule
should be used instead like GTK_IM_MODULE=scim. I am just following
previous hook script for UIM and package maintainer comment for SCIM.
See discussion on
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=368173
Skip last one in Japanese. I mistakingly wrote it in Japanese.
I say that there is no issue with SCIM while setting GTK_IM_MODULE=xim
for oppenoffice-gnome, firefox, and gnome-terminal.
> I ran im-switch -s uim. This works fine. I start
> uim_gtk_toolbar_systray from the icewm startup script.
I think the idea for im-switch is to help user. In that weay, we should
recommend action whern we run command. That is wishlist bug :-) Once I
fix major issues, I will add it.
> For scim, even the XIM_PROGRAM should be removed, and the scim start
> command (scim -d) run from the icewm startup script (i.e. after the
> window manager has started, not before).
But im-switch can not do this easily since there are far too many window
managers. It is a fundamental problem here. We may think a bit
differently. If we are to ask user to do something it should be simple.
One idea is to run a command which points to set of programs to be run
from these session start up script which is also maintained like these
environment values. To do this cleanly, it will involve moving
directory. A bit messy preinst script is needed.
/etc/X11/xinput.d/environment/*
/etc/X11/xinput.d/session-startup/*
/usr/bin/im-startup managed by update-alternative which user inset into
session manager's start up script.
> This solves the "scim not available on X restart" problem.
Yes, it sounds right for session managers.
Thanks for detailed analysis.
Osamu
PS; Time to upload another im-switch after testing.
