UK Extended keyboard problems

awilkins commented 3 years ago

Expected behavior

MiniBLE types circumflex (U+005E) correctly in UK Extended layout, as Shift + 6

Actual behavior

MiniBLE types circumflex modifier AltGr + 6

UK Standard result is ¾
UK Extended result is that the circumflex modifier is set
- This usually results in the next keystroke being "eaten" or modified

Should produce the desired result (a ^) regardless of whether you have UK Standard or UK Extended selected as input method on target machine.

Step by step guide to reproduce the problem

Create a credential with a ^ in it
Set MiniBLE keyboard layout to UK Extended
Type credential

Moolticute Version - If Involved

( You could call this a UX bug as well - all the UK layouts are bunched together except the standard one. )

UK Extended
UK Extended (MacOS)
<some more layouts, you have to scroll the radio to expose the last entry which is ...>
United Kingdom

This makes it likely that people will pick "UK Extended" when they don't need it.

Firmware Version

AUX MCU version: 0.62
Main MCU version: 0.73
Bundle : 2

Operating System

Ubuntu 20.04

Correct use example of this key combo

Ŵ should translate to AltGr + 6 Shift + W

Created a test cred
Set password to ^
Typed the password into a xev window with MiniBLE HID
- This was set to UK Extended
- I have reconfigured to UK Standard which fixes my problem but not this one

Log below.

What is required for this keymap is keycode 50, or 60 (Shift_L or Shift_R), not keycode 92 (ISO_Level3_Shift or AltGr).

MappingNotify event, serial 37, synthetic NO, window 0x0,
    request MappingKeyboard, first_keycode 8, count 248

KeyPress event, serial 37, synthetic NO, window 0xac00001,
    root 0x1e9, subw 0x0, time 25683209, (112,-11), root:(2032,45),
    state 0x10, keycode 108 (keysym 0xfe03, ISO_Level3_Shift), same_screen YES,
    XKeysymToKeycode returns keycode: 92
    XLookupString gives 0 bytes: 
    XmbLookupString gives 0 bytes: 
    XFilterEvent returns: False

KeyPress event, serial 38, synthetic NO, window 0xac00001,
    root 0x1e9, subw 0x0, time 25683224, (112,-11), root:(2032,45),
    state 0x90, keycode 15 (keysym 0xbe, threequarters), same_screen YES,
    XLookupString gives 2 bytes: (c2 be) "¾"
    XmbLookupString gives 2 bytes: (c2 be) "¾"
    XFilterEvent returns: False

KeyRelease event, serial 38, synthetic NO, window 0xac00001,
    root 0x1e9, subw 0x0, time 25683240, (112,-11), root:(2032,45),
    state 0x90, keycode 108 (keysym 0xfe03, ISO_Level3_Shift), same_screen YES,
    XKeysymToKeycode returns keycode: 92
    XLookupString gives 0 bytes: 
    XFilterEvent returns: False

KeyRelease event, serial 38, synthetic NO, window 0xac00001,
    root 0x1e9, subw 0x0, time 25683240, (112,-11), root:(2032,45),
    state 0x10, keycode 15 (keysym 0x36, 6), same_screen YES,
    XLookupString gives 1 bytes: (36) "6"
    XFilterEvent returns: False

KeyPress event, serial 38, synthetic NO, window 0xac00001,
    root 0x1e9, subw 0x0, time 25683256, (112,-11), root:(2032,45),
    state 0x10, keycode 36 (keysym 0xff0d, Return), same_screen YES,
"   XLookupString gives 1 bytes: (0d) "
"   XmbLookupString gives 1 bytes: (0d) "
    XFilterEvent returns: False

KeyRelease event, serial 38, synthetic NO, window 0xac00001,
    root 0x1e9, subw 0x0, time 25683272, (112,-11), root:(2032,45),
    state 0x10, keycode 36 (keysym 0xff0d, Return), same_screen YES,
"   XLookupString gives 1 bytes: (0d) "
    XFilterEvent returns: False

limpkin commented 3 years ago

Wow, this is one of the best bug reports I have ever seen. So in theory the mini BLE should be aware of dead keys. In the case of the ^, it will type a space to get rid of the modifier (which leads to some issues on Linux systems were the compose key isn't set, see #263 where you already contributed). I'm guessing in your case it isn't enabled? Looking at the extended layout in https://en.wikipedia.org/wiki/File:KB_United_Kingdom_Ext.png, it seems to me that both altgr and shift combinations are OK (not sure though, I'm typing on a different layout right now).

Our algorithm at https://github.com/mooltipass/minible/blob/master/scripts/keyboards/cldr_parser.py#L184 is detecting dead keys by detecting that (for example) ^ + = ^ and therefore whenever the mini BLE needs to type a dead key, the latter will add the extra space to type (https://github.com/mooltipass/minible/blob/master/scripts/keyboards/cldr_parser.py#L638). However, in the case of UK extended (I need to look at other layouts, but am fairly sure you don't have this use case for french for example) there is a way of typing a dead key without typing the extra space (the shift combination you mentioned) and it can actually be found in the CLDR: https://github.com/mooltipass/minible/blob/master/scripts/keyboards/cldr-keyboards-37.0/keyboards/windows/en-GB-t-k0-windows-extended.xml#L67 . But that means that depending if you want to type the dead key or the dead key + other vowel (â) then the ^ you're going to type will need to be different (therefore "splitting" the logic at https://github.com/mooltipass/minible/blob/master/scripts/keyboards/cldr_parser.py#L644 ).

I'm wondering if that's quite a tough one to implement without creating all sort of collateral damages on other layouts :/

awilkins commented 3 years ago

On the description above it should work, but what I'm getting is

Input method on computer set to UK Extended
First line MiniBLE UK Standard
Second line MiniBLE UK Extended

test    ^
test

Second line you can watch the MiniBLE type the dead key and then "eat" it with the 0x0d (see xev log above), so the space isn't happening.

Curiously Shift+AltGr Shift+6 Space works for either layout, but this seems to be the compose feature, even though my settings claim I have compose off ; the compose char appears in the editor.

Just to be sure it's not trimming ..

Actual cred ^ prefix, all MiniBLE typing after this line set to UK Extended

test    ¾ prefix  # UK Std. input method
test    ^prefix    # UK Extended input method

Change cred to ^WIBBLE

test    ¾WIBBLE  # Std.
test    ŴIBBLE    # Ext.

Change cred to ŴIBBLE

test    ¾WIBBLE  # Std
test    ŴIBBLE   # Ext

limpkin commented 3 years ago

I think I'm getting close to a conclusion:

we need to handle the transform="no" property of https://github.com/mooltipass/minible/blob/master/scripts/keyboards/cldr-keyboards-37.0/keyboards/windows/en-GB-t-k0-windows-extended.xml#L67 to always use the AltGr + 6 combo to keep our logic the way it is (sending a space after a detected dead key)
for UK standard, ^ is NOT detected as a dead key (see https://github.com/mooltipass/minible/blob/master/scripts/keyboards/cldr-keyboards-37.0/keyboards/windows/en-GB-t-k0-windows.xml): does that mean it is impossible to type ê with a uk standard layout @awilkins ?

However that doesn't explain the second line of your first test.... but anyway we can just add that transform="no" check to our cldr parser and test that way :)

awilkins commented 3 years ago

does that mean it is impossible to type ê with a uk standard layout

Not in an ISO Standard way I think...

On Windows you can do the trick where you hold Alt and type out a number on the numpad (my new keyboard is a TKL, but hey, all the keys on the Mooltipass are pretend ones). e.g. e-circumflex would be Alt+0234

On Linux I think Shift+AltGr e ^ will work, as would Ctr+Shift+u 0 0 e a Enter - as to whether these work reliably as inputs to e.g. passwd or an ecryptfs password prompt, don't know.

limpkin commented 3 years ago

@awilkins did you update to bundle 3 and verified that things are now working? :)

awilkins commented 3 years ago

I've been busy! Happy to try it, but the bundle issuer has given me a 0_to_3 bundle, and I'm already on bundle 2 from the knock detection issue - can't apply it, just getting an error.

limpkin commented 3 years ago

@awilkins did I give you the right upload password? if no, send me a private message :)

limpkin commented 3 years ago

I'm closing this issue as it's converging similar to https://github.com/mooltipass/minible/issues/263 . I therefore suggest we move our discussion over there :)

mooltipass / minible