Hello, My keyboard's dead keys work properly in GTK applications, but not in Qt ones (including the LXQt desktop): those keys have absolutely no effect there. What should I do?

@"bsduck"#p9555 How is the X-server keyboard layout being assigned? Although, I don't use LXQt on a daily basis, I do test the desktop after every update and before merging it. Usually, I also ask one more person to test and review the DE before importing the package and, we've never experienced any issues with the keyboard before.

In my .xinitrc I have `setxkbmap ch` The problem is the same with both "fr" and "de" variants. Using "sundeadkeys", "legacy" or "mac" variants makes no difference either. Using "nodeadkeys" makes the dead keys work as regular keys (diacritics do get printed as standalone characters, while nothing happens with the other variants), but of course not as dead keys.

In my case, Qt applications doesn't recognize accents as in the word "Vídeos". I have a folder with that name and for Qt applications as smplayer and vlc, that folder is invisible. I had not realize the keyboard issue, but the same occurs to me. I am unable to use any dead key with Qt.

> @"jmmarin"#p9559 I have a folder with that name and for Qt applications as smplayer and vlc, that folder is invisible. That's already one step further into weirdness... and I don't have that problem. I created, from xterm (where dead keys do work too, so not only with GTK), a folder named "Vídeos", and it shows up perfectly fine, with correctly displayed "í", in PCManFM-Qt as well as VLC.

> @"bsduck"#p9555 My keyboard's dead keys work properly in GTK applications, but not in Qt ones (including the LXQt desktop): those keys have absolutely no effect there. > > What should I do? Have you checked your desktop's keyboard settings? Desktops can ignore or override X server settings in these cases. In XFCE, there is Settings > Keyboard > ... where I can either use the system defaults or change stuff.

@"rvp"#p9561 You made me notice something I didn't pay attention to until now: `lxqt-config-input` is missing! I don't have any tool available to set keyboard settings inside LXQt. It definitely should be there, and it is in the same lxqt-config version 0.17.1 on FreeBSD. May this be related to the following patch? http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/x11/lxqt-config/patches/patch-lxqt-config-input_keyboardlayoutconfig.h?rev=1.1&content-type=text/x-cvsweb-markup

> @"bsduck"#p9562 lxqt-config-input is missing! ... It definitely should be there, If you look into the package Makefile you will find the following section, ``` .include "../../mk/bsd.prefs.mk" .if ${OPSYS} != "Linux" CMAKE_ARGS+= -DWITH_INPUT=OFF .endif ``` This is because, if you read `lxqt-config/lxqt-config-input/CMakeLists.txt`, you will find, `pkg_check_modules(XORG_LIBINPUT REQUIRED xorg-libinput)`. We don't have `libinput`. But, now that I'm writing this, I've realized my mistake. This has been there all the time and has slipped through several reviews and user reports, until now. I should have used `CMAKE_ARGS+= -DWITH_TOUCHPAD=OFF` instead of `CMAKE_ARGS+= -DWITH_INPUT=OFF`. This should be enough to remove the dependency on `libinput` without switching off the whole input module. I'm in the process of updating LXQt to the newly released 1.0.0 and should be able to finish it today. I'll try to build and test with the correct switch off. If you're running current, it should be there by the end of the week or, by early next week. If you're on the Q-release of pkgsrc it will take about 3 months to get it. I could, if you want, send you instructions on _how to build it_ locally before it is actually merged. Let me know if you want to test it and sorry for the inconvenience.

Thank your for looking into this! I think you're right, because if I'm not mistaken OpenBSD doesn't have libinput either, but lxqt-config-input is available in their LXQt build. > @"pin"#p9565 the newly released 1.0.0 Hey, I didn't notice that yet! 1.0 probably means LXQt feels ready to conquer the world 😎 > @"pin"#p9565 I'm in the process of updating LXQt It's great to see LXQt getting attention. In the meanwhile, OpenBSD is stuck at version 0.15 :) > @"pin"#p9565 I could, if you want, send you instructions on how to build it locally before it is actually merged. Yes please, I'll happily test that.

> @"bsduck"#p9566 Yes please, I'll happily test that. I have three sub-packages left to go, should be finished soon. Send me an e-mail and I'll let you know where to fetch the packages and how to build them when I'm ready.

> @"bsduck"#p9562 You made me notice something I didn't pay attention to until now: `lxqt-config-input` is missing! I'm not so certain that this is needed. LXQT uses whatever [keyboard config](https://man.netbsd.org/NetBSD-9.2-STABLE/amd64/xkeyboard-config.7). you setup in X: I added `setxkbmap -layout 'us(altgr-intl)'` to my `~/.xinitrc` just before running `exec startlxqt` and was able to get accented chars. with R-Alt. My other custom keybindings also come througt unscathed. What does `setxkbmap -query` show?

Keyboard dead keys don't work in Qt applications

rvp

pin No the other stuff: adding QTPLUGIN += composeplatforminputcontextplugin to some .pro file.

rvp

bsduck What should I do?

Got to thinking about how QT programs handled, say, Japanese input under KDE. Obviously, they would rely on an IM Engine (which would get installed along with KDE) for this.

Turns out, QT progs. recognize dead-keys just fine in LXQT, too, if you point 'em to one and do some minor config.

Install an IME first (here, UIM):

# pkgin install -y uim

Run IME and point QT to it. In ~/.xinitrc (or elsewhere)

export XMODIFIERS='@im=uim'
# export GTK_IM_MODULE='uim'
export QT_IM_MODULE='uim'
uim-xim &

exec startlxqt

Configure IME for Latin input:

Go to Preferences > Input Method
In Global Settings
Check the Specify default IM box
Set Default input method to Latin characters
Apply; OK

@pin might want to indicate in the LXQT install messages that an IME is required to process dead-keys and intl. char. input.

rvp @pin might want to indicate in the LXQT install messages that ...

Or, just add one to the LXQt meta-package.

Thanks!

bsduck

rvp This works fine, thank you!

nia

pin Maybe it needs to be enabled in the base qt package? But I can't find evidence of other projects needing to explicitly do this so I'm very confused.

nia I've just added inputmethod/uim to the lxqt meta-package for now. I didn't look at it anyfurther but, one thing pops-up. Why is this not the case with Gtk? Has it been built-in?

rvp

pin Why is this not the case with Gtk? Has it been built-in?

This puzzled me too; also the fact that lxqt seems to require an IME for dead-key recognition, but, mentions this nowhere in their official docs. At the time, I rationalized it by saying: OK, they need an IME, but don't supply one by default because (as you would say) choices: there are many IMEs around--UIM, SCIM to name two--and they don't want to override what a user may already be using.

But, this does look like something you should ask upstream. Does lxqt require an IME? Or, does some magic gobbledygook flag have to be set when building lxqt to get dead-keys recognized?

rvp Good suggestion, thanks
https://github.com/lxqt/lxqt/discussions/2130

bsduck

Dear @rvp,

Unfortunately, this doesn't work for Qt6 applications inside the Qt5 LXQt environment. For example, FeatherPad is now provided as a Qt6 build, and it doesn't seem to care about export QT_IM_MODULE='uim'. Is there another line to add for Qt6?

rvp

bsduck I'll take a look, but it won't be soon.

BTW, did you try my patch for your other Featherpad issue?

bsduck Qt6 applications inside the Qt5 LXQt environment

As you might have noticed LXQt has moved to Qt6. Unfortunately, we still don't have a package for kf6, which is required to do the switch.

bsduck

Take your time! As you probably noticed, I'm not in a hurry...

I did test the other issue again and it is fixed now - I'll reply in the corresponding thread.

bsduck

I noticed, indeed, but haven't tested the new LXQt 2.0.0 yet. FreeBSD is still on 1.4.0 too.

bsduck

For information: I don't need export XMODIFIERS='@im=uim' nor uim-xim & in .xinitrc for this to work.

The only line needed for Qt5 applications is export QT_IM_MODULE='uim'.

uim-helper-server is automatically started.

rvp

bsduck So it all works now?

bsduck

Not for Qt6 applications.

rvp

bsduck Can you try after adding this in your ~/.xinitrc or ~/.xsession?

export XLOCALEDIR=/usr/X11R7/lib/X11/locale

You don't need to add any of the other stuff that we've been adding before. Not even QT_IM_MODULE. For dead-key handling, I don't think an IME, even, needs to be installed. You just need to set a keyboard which produces dead-keys codes using either setxkbmap or in the DE own config. In my case `us(intl)' is what produces dead-keys. The German ones should do it by default, I believe.

Briefly, what's happening here, I think, is that libxkbcommon (which both QT5 and QT6 rely on for keyboard management)--as it is currently being built--assumes all the keyboard-related files are in /usr/pkg. If you had modular Xorg, it would be; but for native X it's not correct. A ktrace of featherpad indicated what was happening.

libxkbcommon and whatever it relies on will ultimately have to be adjusted, but, that's for the pkgsrc people to fix...

rvp I can confirm this works. I've added set-env XLOCALEDIR "/usr/X11R7/lib/X11/locale" to ~/.config/elvish/rc.elv and dead-keys are working on FeatherPad.

Thanks!

rvp libxkbcommon and whatever it relies on will ultimately have to be adjusted, but, that's for the pkgsrc people to fix...

Fixed by wiz@, https://mail-index.netbsd.org/pkgsrc-changes/2024/05/20/msg300473.html

rvp

Updated prev. Please try.

« Previous Page Next Page »