bug#74016: 29.4; emacs-pgtk doesnt receive numpad events on wayland(sway)

[joakim@verona.se]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: joakim@verona.se
>> Date: Fri, 25 Oct 2024 21:35:48 +0200
>> 
>> 
>> Emacs-pgtk doesnt seem to receive numpad events, instead number events
>> are received.
>> 
>> There was a thread about this on emacs-help:
>> 
>> https://lists.gnu.org/archive/html/help-gnu-emacs/2024-10/msg00094.html
>> 
>> I have attempted to summarize the thread here, wherein Stefan Monier
>> recomended I report this as a bug, and I also received knowledgeable
>> advice from several others.
>> 
>> Heres the bug:
>> 
>> I used to use a numpad to bind to do things in emacs.
>> 
>> This used to work really well on the i3wm window manager and a gtk emacs
>> build.
>> 
>> Then I bought a nice new framework laptop which has a hidpi screen, so I
>> switched to using the sway wm, and emacs-pgtk.
>> 
>> Now the numpad no longer sends keypad events in a way that emacs
>> recognizes, instead just plain numbers are received.
>> 
>> below is an example of the code I use, but normally I just test with c-h
>> k to see which key emacs receives.
>> 
>> I also tried xev, and there it seems I get kp events, at least under
>> xwayland I guess.
>> 
>> (define-key cider-mode-map (kbd "<kp-enter>") 'sylt-stop)
>> (define-key cider-mode-map (kbd "<kp-separator>") 'sylt-play-metro)
>> 
>> (defmacro sylt-song-event-map-key (key num)
>>   `(define-key cider-mode-map (kbd ,key)
>>                (lambda  () (interactive)   (cider-nrepl-sync-request:eval  
>> ,(concat "(sylt.core/song-event \"" num "\")")))))
>> 
>> 
>> 
>> (sylt-song-event-map-key "<kp-0>" "init")
>> (sylt-song-event-map-key "<kp-1>" "1")
>> 
>> 
>> Instead of xev, I tried wev, and got this, which seems expected:
>> 
>> Using wev, installed from Fedora repos I get, for numpad 4, and normal 4:
>> 
>> [14:     wl_keyboard] key: serial: 60293; time: 61646320; key: 83; state: 0 
>> (released)
>>                       sym: KP_4         (65460), utf8: ''
>> [14:     wl_keyboard] key: serial: 60294; time: 61651642; key: 13; state: 1 
>> (pressed)
>>                       sym: 4            (52), utf8: '4'
>> 
>> That is, the numpad and the number pad sends distinct events that are
>> received by wev.
>> 
>> using (read-key) in emacs-pgtk, both the numpad and the number row are
>> received as the same event:
>> 
>> 52 (#o64, #x34, ?4) [2 times]
>> 
>> I get the same in a emacs gtk session.
>> 
>> Maybe gtk is doing something weird?
>> 
>> so now I tried emacs-29.4-gtk+x11 -Q
>> 
>> and then I get "4 (translated from <kp-4>)"
>> on c-h k, numpad 4, so xwayland is doing something correctly in this
>> case, which pgtk isnt doing in the same way(imho incorrectly but what do
>> I know)
>> 
>> (and as a reminder, I run emacs-pgtk because it looks super nice on
>> wayland, with the scaling hints I use on my framework hidpi screen, but
>> emacs-29.4-gtk+x11 doesnt look good.)
>
> Po Lu, is this another manifestation of problems with keyboard input
> under PGTK?  If so, should we add something to etc/PROBLEMS?

FYI I found a "workaround" to this problem. I now run Sway with scaling of
1 rather than 1.25.

Then I have to scale each application independently, so I set the
corresponding scale factor in sway window decorations, waybar, firefox
etc.

(the problem with this aproach is that a number of x11 apps cant be scaled)

Lastly in emacs I just scale the fonts and run emacs-gtk-x11. Now things
work adequately.

emacs-pgtk has other interesting features, but I guess they will have to
wait a while more.

Thanks everyone!

>
>
>
-- 
Joakim Verona
joakim@verona.se