FWIW, here is a script to adapt xmodmap to the keyboard:
Note: the sed(1) command could be simplified with range but since I may
use it (the script) on systems whose sed(1) (Plan9) does not implement
ranges, I have written it in the simplest way.
-----8<-----
#!/bin/sh
# C) 2020 Thierry Laronde <tlaro...@polynum.com>
# Public Domain.
# No warranty! Use at your own risk!
: ${TMPDIR:=/tmp}
cat <<EOT >&2
When the window will appear, click first in the xterm
and then in the window (the square in the blank window).
You can then start to press Meta keys: "Windows",
Alt, AltGr, arrows, home, end, page up, page down...
so that we can generate the xmodmap rules.
Press any key to start. Kill the xev window when done.
EOT
read ans
# Get keycode keysym name from xev.
#
xev -event keyboard 2>&1\
| sed -n -e 's/^Outer window is \(0x[^,]*\).*$/\1/p' \
-e 's/^.*keycode \([0-9][0-9]*\) (keysym
0x\([a-f0-9][a-f0-9]*\), \([^)][^)]*\)).*$/\1 \2 \3/p'\
| sort | uniq\
| while read keycode keysym name; do
test "$keysym" || { wid=$keycode; continue; } # unused
echo "keycode $keycode = $name"
case $name in
Shift_[LR]) echo "add shift = $name";;
Control_[LR]) echo "add control = $name";;
Alt_L|Meta_L) echo "add mod1 = $name";;
Super_[LR]|Hyper_L) echo "add mod4 = $name";;
ISO_Level3_Shift|Mode_switch) echo "add mod5 = $name";;
esac\
>$TMPDIR/xmodmap
done
xmodmap $TMPDIR/xmodmap
----->8-----
On Fri, Nov 06, 2020 at 07:11:03PM +0100, tlaronde wrote:
> Hello,
>
> I have stumbled (it took me hours to find the solution) upon this
> problem with the keyboard keycodes and the xmodmap mapping.
>
> Context (yes! it has something to do with NetBSD; please bear with me):
> Fabrice Bellard has developed a VM in JavaScript and has put a
> demonstration with various systems amongst which AlpineLinux.
>
> A teacher, in the USA, has asked me about the possibility to run kerTeX
> (my TeX small distribution) in this restrained environment, in order to
> be able to teach without requiring any installation from the students: a
> browser is sufficient.
>
> So, as far as kerTeX is concerned, no problem: neither for compilation
> nor for size. And it is made so that even in this constrained
> environment (it runs approximately 200x slower for compilation for
> example...), it manages to work quite fast.
>
> So the NetBSD part: since my main development plateform is NetBSD, I
> tested under NetBSD (8.0), amd64, with native Xorg, almost bare (I use
> twm) and firefox 71.0.
>
> When under jslinux, the Alpine X11 version, I switched the keyboard
> for a french azerty mapping, I was then unable to access the
> characters that are obtained after composing with AltGr. Without pipe
> or backslash, one enters a rough time...
>
> So for US qwerty users, OK, but for others, the experience was not
> great...
>
> I skip the tale of the Net search; I asked to the alpine developers
> and even to Fabrice Bellard.
>
> And I finally found:
>
> 1) For the NetBSD puzzling part:
>
> If I run xmodmap(X1) with the -pm option, I see that AltGr (called
> ISO_Level3_Shift) is associated with the keycode 0x7c.
>
> But if I run xev(X1) to get the keycode of my AltGr, it is 0x71! And
> nonetheless it works ?!!!
>
> Does anyone have any clue about why it simply works?
>
> 2) When running jslinux hence NetBSD->Firefox->JavaScript
> engine->jslinux->Alpine->X11, if I run xmodmap in jslinux, it maps
> ISO_Level3_Shift to 0x5c! So "my" AltGr is not recognized as a
> ISO_Level3_Shift, hence not associated to Mod5, hence no action at all.
>
> If I add the rules (in jslinux):
>
> keycode 0x71 = ISO_Level3_Shift
> add mod5 = ISO_Level3_Shift
>
> then everything works.
>
> So the question: I thought that with USB Hid there was some
> standardization. Why do different systems use different keycodes? And
> does NetBSD (and others) customize the mapping in order to be compatible
> with wscons or whatever?
>
> I have found the solution (well at least for me; if there are 256
> possible variations for every OS...) but I'd like to understand the
> logic.
>
> [I'm not saying that Linux is "right" and the others wrong, or
> vice-versa, I'm just puzzled that the same Xorg code does things
> differently under different OSes and that there is no normalization.]
>
> Thanks for any clue!
> --
> Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
> http://www.kergis.com/
> http://kertex.kergis.com/
> http://www.sbfa.fr/
> Key fingerprint = 0FF7 E906 FBAF FE95 FD89 250D 52B1 AE95 6006 F40C
--
Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
http://www.kergis.com/
http://kertex.kergis.com/
http://www.sbfa.fr/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89 250D 52B1 AE95 6006 F40C
