Here's a better testcase for zh_CN locale using chinput. :)

-----
chinput (zh_CN)
Steps to Reproduce:
1.toggle on (ctrl-space)
2.enter 'hong' (do not commit)
3.Press up or down arrow key
4.enter 'h'
  
Actual results:
Preedit buffer shows 'hong h'

Expected results:
Preedit buffer shows 'h'

The current event handling also affects the "right arrows" and "left
arrows". When LE is activiated, press right arrow or left arrow once
will be executed twice. (i.e. in oocalc, press right arrow will jump
from A column to C column)

Well, it looks like you contains multiple problem here - for first
testcase (both of zh_CN and zh_TW), it's duplicated bug with Bug#136643.
for second testcase, it's because of iiim immodule doesn't support
gtk_im_context_reset(). neither first testcase nor second testcase is
OOo bugs. and technically both are im-sdk bug.
I'm reassiging this to im-sdk to be tracking for gtk_im_context_reset
stuff.

Marking it as item for RHEL4 U1

Fixed in 12.1.1-6.svn2208 for rawhide and 12.1-12.EL for RHEL4U1.

Tested 12.1-12.EL with openoffice.org-1.1.2-20.6.EL4 and Dan's recent 
libvclplug_gtk645li.so. zh_* and ja are working properly in oowriter. However,
hangulLE seems to be experiencing problem. Downgraded im-sdk to 12.1-10.EL and
hangulLE is working properly again.

Actual Result:
Unable to type korean char continuously. Have to input a single char, type enter
and proceed to the next char.

Expected Result:
Should be able to input korean text continously.

Test string: qkfrkd (ë°ê°)

Well, It's a bad effect after fixing gtk_im_context_reset() stuff and
OOo has that problem IMO. and it should be filed as OOo bug separately.

Reason is that when "commit" is called, iiimf calls "preedit-changed" as well
with an empty string, exactly like XIM does.  This is fine.  But OOo doesn't
treat the committed string and the preedit string differently, so when you do
the "preedit-changed", it erases the previously committed string.

So I fixed the original problem, but there's a new one with Korean:

signalKey: PRESS 114 (Unicode 0x72), time 1820681158.
signalIMPreeditChanged: called with text 'ë°'.
   ^ I just pressed 'r'
signalKey: filtered by IM, returning.
signalKey: RELEASE 114 (Unicode 0x72), time 1820681289.
signalKey: filtered by RELEASE matching, returning.
signalKey: PRESS 107 (Unicode 0x6B), time 1820684218.
signalIMPreeditChanged: called with text 'ê°'.
   ^ I just pressed the second 'k'
signalIMCommit: called with text 'ë°'.
   ^ LE decides to commit 'qkfr'
EndExtTextInput: called.
   ^ EndExtTextInput() calls resetIMContext()
resetIMContext: resetting...
   ^ now gtk_im_context_reset() is called
signalIMCommit: called with text 'ê°'.
   ^ which causes the LE to commit the _next_ text in the preedit buffer
resetIMContext: done
signalKey: filtered by IM, returning.
signalKey: RELEASE 107 (Unicode 0x6B), time 1820684310.
signalKey: filtered by RELEASE matching, returning.
signalKey: PRESS 100 (Unicode 0x64), time 1820709739.
signalIMPreeditChanged: called with text 'ã'.
signalKey: filtered by IM, returning.
signalKey: RELEASE 100 (Unicode 0x64), time 1820709864.
signalKey: filtered by RELEASE matching, returning.
signalKey: PRESS 32 (Unicode 0x20), time 1820711989.
signalIMCommit: called with text 'ã'.
EndExtTextInput: called.
resetIMContext: resetting...

So, it seems that the gtk_im_context_reset() during "signalIMCommit" causes the
Korean LE to commit the text in the preedit buffer that shouldn't really be there.

Ok, seems to be fixed now with changes to OOo to not call gtk_im_context_reset()
during a "commit" signal, but to still EndExtTextInput internally.

Fixed in:

RHEL3: 1.1.2-22.2.2.EL3
RHEL4: 1.1.2-22.6.EL4

Still needs to be forward-ported to 1.1.3 for FC2 & FC3.

eek, back to ASSIGNED since this is for Fedora Core not RHEL.  Also moving back to 
"openoffice.org" component since it's an OOo issue mostly.

So when will you push the updated package then?

FYI, the latest IIIMF which contains the implementation of
gtk_im_context_reset() was pushed for rawhide. what we need to do is fixing OOo
now. reassigning to dcbw.

RHEL4 patch applied to FC3 builds, now doing local build for functional
verification before pushing to beehive

removing self as CC since its my issue...

FC3 testing packages are at:

http://people.redhat.com/dcbw/ooo/

Please verify that these issues have been resolved and that the packages work
correctly with IIIMF that has gtk_im_context_reset()

*** Bug 144557 has been marked as a duplicate of this bug. ***

Warren: can you test and verify the packages in Comment 17?

hmm, my short testing in ja_JP.UTF-8 with im-sdk 12.1-10.FC3 seems to indicate
that resetting is not happening to the preedit string.  More investigation required.

Dan: I will update you with my findings by tomorrow morning.

Tried the OOo package from Comment 17, I have similar situation as well for CJK
locale with oowriter and oocalc. The preedit string actually commits to the
application if I switch off the LE with preedit buffer != NULL.

Seems to work for me when I build the im-sdk in the FC-3/ dir of package CVS. 
Patches for gtk_im_context_reset() are present in the specfile there but not in
current 12.1-10.FC3 packages available for FC3.

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHSA-2005-375.html