Re: tlug: XEmacs & XIM, redux

["Stephen J. Turnbull" <turnbull@example.com>]

--------------------------------------------------------
tlug note from "Stephen J. Turnbull" <turnbull@example.com>
--------------------------------------------------------
Executive summary:

(1) Using XIM with XEmacs requires libc locale support.  Cheat with the
    liblocale package for Netscape (URLs below).
(2) You need to set some environment variables for XEmacs and invoke
    kinput2 with switches (see below).
(3) XEmacs seems to handle locale OK, and tries to open the IM
    (kinput2), but fails.
(4) This looks like a program bug, but which one is unclear.  (See
    below for details.)

I'd like to hear from somebody who's got Wnn installed if Xwnmo works
with this configuration.

Details:

>>>>> "Steve" == Steve Dunham <dunham@example.com> writes:

    Steve> Craig Oda <craig@example.com> writes:

    >> Thanks for the tip on --with-xim.  Where did you find this?
    >> Was there something like --with-kinput2 or something?  I
    >> wouldn't have known about --with-mule if Steve Dunham hadn't
    >> told me about it.

    Steve>   configure --help

That's the "right" answer, and it's reliable with GNU software.  (It's
a standard of sorts for GNU software.)  My answer was `fgrep -i xim
configure'.  For binaries, `strings binary | fgrep xim' might work if
`--help' doesn't.

However, the INSTALL document should explain this, and point to Info
docs on how to use the new feature.  I suspect this feature should
still be considered alpha since it is not visibly documented.  :-(

    Steve> I'm currently compiling 20.2 (which is released) with XIM
    Steve> and wnn enabled.  I'll let you know how it turns out.

Are there Info docs for Mule yet, or just the stuff in ./mule-doc?  I
don't have the release version I suspect, must be a prerelease version
(it doesn't have a 'beta' on it, though) but it's substantially bigger
than b6 or whatever the last one was, but no Mule docs.  (The stuff in
./mule-doc pointedly ignores XIM.)

    Steve> It should compile ok. AFAIK, --with-wnn just links in
    Steve> libwnn and defines some new lisp functions.  --with-xim
    Steve> should be completely independent.

I'm currently getting `unknown signal' faults "--with-mule --with-xim"
but I think this is a `simple' bug.  Before the fault I get a
complaint about unknown locale, using C, then an immediate fault in
one of the X text routines.  I need to tell XEmacs where to find the
locale database.  (But XEmacs shouldn't fault on this error, it should 
barf an error message and exit gracefully.)  (* This is wrong; in fact 
you need to supply a correct locale routine.  See below. *)

    Steve> I got kinput2 compiled (although I'm using it with wnn, and
    Steve> it isn't very comfortable in conjunction with wnn - I much
    Steve> prefer egg-wnn.el) and tried it out with kterm.  It works
    Steve> on Linux (I'm having problems on my Ultrasparc at work -

Are you using "kinput2 +kinput -xim"?  If not, kinput2 defaults to
kinput2 protocol, and you don't have any problems.  Using kinput2
compiled with "-O2 -mi486 -DDEBUG" and Canna as "kinput2 +kinput +ximp
- xim -canna -trace" then opening the input method in a kterm results
in a trace of XIM protocol errors from kinput2 and the kterm freezing;
destroying the kterm window segfaults the kinput2 process.  As far as
I can tell, kterm does not have command-line options to select which
IM to use although there are probably resources.  The normal procedure
is to set "LANG=ja_JP" and "XMODIFIERS='@example.com=kinput2'" in the environment
before running kterm.

BUT this procedure is inappropriate for Mule which is explicitly
multilingual; you ought to be able to switch among native cWnn
Chinese, XIM Canna Japanese, Quail Thai, x-compose.el French, and
direct ASCII input on the fly using Mule menus or keyboard shortcuts.

This doesn't seem to be documented anywhere.  It shouldn't be too hard 
to implement by using Lisp facilities to manipulate the environment
and I hope to open and close input methods and input contexts
(although current GNU code does this when a frame is created; I
haven't found the corresponding parts of XEmacs code but I suspect
it's the same from the code I have found).

    Steve> kterm segfaults in OpenIM()).

Verrrry interesting.  I think kterm is probably buggy.  It doesn't
work with Xwnmo, either.  (The Xwnmo code is pretty ugly though.)

I still don't have working XIM.  Here's what I've got to this point.

(1) XEmacs depends on libc locale support.  I don't know if this has
    been added to recent libc releases; my current is 5.3.12.  If your 
    libc doesn't have locale (XEmacs complains about missing locale,
    sets it to C/POSIX, and immediately dumps core), you can "cheat"
    with liblocale.so, available from the Japanese localization of
    Netscape kit (see

    http://www.twics.com/~craig/linux-nihongo/web/index.html or
    http://www.bpel.tutics.tut.ac.jp/~take/Netscape/Linux.html).

(2) Start kinput2:

    kinput2 +kinput -xim [-trace | &]

    (use -trace if you've compiled in debugging, otherwise you may as
    well run it in the background).

(3) If you haven't already set the environment, invoke XEmacs as

    LD_PRELOAD=/lib/liblocale.so \
    XMODIFIERS='@example.com=kinput2' \
    LANG=ja_JP \
    /var/tmp/xemacs-20.2/src/xemacs &

and watch XEmacs complain

    Warning: Missing charsets in String to FontSet conversion
    Warning: XCreateIC failed

and watch weird stuff happen in kinput2's trace (stuff I think should
probably not happen is flagged with `->' in the left margin):

    imlib:mainDispatcher(#2)
    checkRequest(): length=28, buflen=40
    ** processing ENCODING-NEGOTIATION request...
    ximEncodingNegotiationProc(#2)
            selected encoding index: 1
    IMSchedule(#2, 1)
    insert into the queue
    imlib:mainDispatcher(#2)
    IMProcessQueue()
    IMFlush(#2)
    imlib:xinput()
    IMDispatch(#2)
    imlib:mainDispatcher(#2)
    checkRequest(): length=8, buflen=20
    ** processing GET-IM-VALUES request...
    ximGetIMValuesProc(#2)
    IMGetIMValues()
    imlib:getQueryInputStyle()
    IMSchedule(#2, 1)
    insert into the queue
    imlib:mainDispatcher(#2)
    IMProcessQueue()
    IMFlush(#2)
    imlib:xinput()
    IMDispatch(#1)
    imlib:mainDispatcher(#1)
    checkRequest(): length=356, buflen=360
    ** processing CREATE-IC request...
    ximCreateICProc(#1)
    IMSetICValues()
    imlib:setICValues()
    imlib:setInputStyle()
            input style: 260
    imlib:setClientWindow()
            client window: 03c00034
    imlib:setFocusWindow()
            focus window: 03c00034
    imlib:setPreeditAttributes()
    imlib:setICValues()
    imlib:setArea()
->          area: 0, 0, 0, 0
    imlib:setSpotLocation()
            spot location: 0, 0
    imlib:setForeground()
            foreground: 1
    imlib:setBackground()
            background: 0
    imlib:setFontSet()
->          fontset: -misc-fixed-bold-r-normal--13-120-75-75-c-70-iso8859-1,-sony-fixed-medium-r-normal--16-120-100-100-c-80-jisx0201.1976-0,
    imlib:setStatusAttributes()
    imlib:setICValues()
    imlib:setArea()
            area: 0, 0, 0, 0
    imlib:setForeground()
            foreground: 1
    imlib:setBackground()
            background: 0
    imlib:setFontSet()
            fontset: -misc-fixed-bold-r-normal--13-120-75-75-c-70-iso8859-1,-sony-fixed-medium-r-normal--16-120-100-100-c-80-jisx0201.1976-0,
    IMValidateICAttributes()
    imlib:validateCommonAttr()
    IMValidateWindow(win: 03c00034)
    imlib:validatePSAttr()
    zero area width/height
    IMSchedule(#1, 1)
    insert into the queue
    imlib:validatePSAttr()
    zero area width/height
    IMSchedule(#1, 1)
->  IMFreeICAttributes()
    imlib:mainDispatcher(#1)
    IMProcessQueue()
    IMFlush(#1)
    imlib:xinput()
    IMDispatch(#1)
    imlib:mainDispatcher(#1)
    checkRequest(): length=12, buflen=20
->  ** processing ERROR request...
->  ximErrorProc(#1)
->  ** XIM_ERROR message received:
->           input-method ID: 1
->          input-context ID: N/A
->                error code: BadProtocol
->             error message: 
    imlib:mainDispatcher(#1)

The setArea calls may be OK since I think mule tries for on-the-spot
interactions.  In that case the preedit and status areas might legally 
be null.  The fontLists are ugly, ending in commas and containing no
kanji.  I think that protocol messages are getting truncated there.
The IMFreeICAttributes means (I think) that it gave up and refused to
create an input context (the approximate equivalent of X refusing to
create a graphics context - you're in deep trouble).  And ERRORs
shouldn't happen (not BadProtocol, anyway...).

I don't know which is wrong, kinput2 or XEmacs.  But this is
definitely not a working system at this point.

-- 
                            Stephen J. Turnbull
Institute of Policy and Planning Sciences                    Yaseppochi-Gumi
University of Tsukuba                      http://turnbull.sk.tsukuba.ac.jp/
Tel: +81 (298) 53-5091;  Fax: 55-3849              turnbull@example.com
-----------------------------------------------------------------
a word from the sponsor will appear below
-----------------------------------------------------------------
The TLUG mailing list is proudly sponsored by TWICS - Japan's First
Public-Access Internet System.  Now offering 20,000 yen/year flat
rate Internet access with no time charges.  Full line of corporate
Internet and intranet products are available.   info@example.com
Tel: 03-3351-5977   Fax: 03-3353-6096