Arch Linux

luna-klatzer

Member

Registered: 2022-11-08

Posts: 3

[SOLVED] JP IME Mozc using ibus only translates one char at a time

Hello everyone,

I am currently trying to set up Mozc as my Japanese IME for my KDE Plasma desktop, which should as far as I know behave similarly to the Google Japanese IME on Windows, which should automatically detect and convert Romanji -> Hiragana/Katakana as well as translate Hiragana -> Kanji using automatic recommendations.

The issue though is that with the current setup I have that Mozc only converts one character at a time and doesn't remember the previous input, so only romanji characters that are one character long, such as あ、お、い、う、え are recognised/converted. To that, the Kimpanel just appears for a few milliseconds, before quickly disappearing again, which leads to me think that the IME has some bug or config issue that makes it not save my input or abort the input process.

I am not sure though what might be causing the issue, but here are the steps I did for installing Mozc and IBus:

1. I installed ibus, mozc (AUR) and ibus-mozc (AUR), as explained in the guide for Mozc.

2. Afterwards, I added the following config to the ./.bashrc file and  ./.xprofile file (It didn't exist before, so I created it):

# Export ibus settings
export GTK_IM_MODULE=ibus
export QT_IM_MODULE=ibus
export XMODIFIERS=@im=ibus

# Run ibus
ibus-daemon --xim --daemonize --desktop=kde --replace

  As well as the following config to the /etc/environment file:

export GTK_IM_MODULE=ibus
export QT_IM_MODULE=ibus
export XMODIFIERS=@im=ibus

3. Then I edited ~/.config/mozc/ibus_config.textproto and set active_on_launch to True. This is the result file:

# `ibus write-cache; ibus restart` might be necessary to apply changes.
engines {
  name : "mozc-jp"
  longname : "Mozc"
  layout : "default"
  layout_variant : ""
  layout_option : ""
  rank : 80
}
active_on_launch: True

4. Afterwards I simply added Mozc as an input method to IBus using the user interface that appears when running:

ibus-setup

  (I instead preferred to use the IBus Preferences entry that appears when searching for ibus in the KDE Plasma search bar, but I don't think that should change the behaviour)

Regarding the config for Mozc, I barely changed any value and only set the input mode to Romanji. I have attempted a few different ways to start or run ibus, but no attempts have helped me fix this issue.

Since I don't know what I can do to fix this issue, I hope someone here can help me. Hope I was able to document my steps well!

PS: I have found this post, which explains another issue for Mozc and IBus, which seems to have been fixed by switching to fcitx, but for now I would like to stick to Ibus if possible, since other than this issue it has worked well for me.

Last edited by luna-klatzer (2022-11-10 10:00:59)

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,676

Re: [SOLVED] JP IME Mozc using ibus only translates one char at a time

Wayland or Xorg session? If Wayland try a xorg session, the input handling in Wayland is significantly different that I'd expect to be some quirks there that you can likely not reproduce on a Xorg session.

luna-klatzer

Member

Registered: 2022-11-08

Posts: 3

Re: [SOLVED] JP IME Mozc using ibus only translates one char at a time

I tried it out now and apparently that was already the fix for the issue since now it works perfectly as expected and the input is handled as it should.

I didn't think that the session could be a problem, but thanks a lot for the help! Marking the forum post as resolved now :)

Last edited by luna-klatzer (2022-11-10 09:55:19)

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 21,676

Re: [SOLVED] JP IME Mozc using ibus only translates one char at a time

I don't consider this "the fix for the issue" it would simply explain the symptoms and in that case we'd need to find out why KWin doesn't interface this properly or whether there are known issues here.

luna-klatzer

Member

Registered: 2022-11-08

Posts: 3

Re: [SOLVED] JP IME Mozc using ibus only translates one char at a time

That's true, but at the moment that seems to be an easy workaround until the issue with Wayland is resolved and I sadly have no clue how the issues with Wayland could be debugged/diagnosed properly.