fix(klc): Correct VK in the KLC

[Geobert]

I’ve refactored the key_codes.yaml as VK are not bound to SC.

There’s also a change to make the double 1dk works again, but it breaks test_intl_deadkeys but I think the test needs to be updated.

test_intl_keymap is also broken, but this time, I’m unsure of the correct path to fix it.
I produce:

56	OEM_5	0	005c	007c	-1	-1	// \ |

when the test is expecting:

56	OEM_102	0	005c	007c	-1	-1	// \\ |

Which means my asumption that a char and a VK have a link is also wrong :-/

Need your insight on this :)

EDIT: I think the intl.toml is wrong. We have \ | two times. From what I’ve seen on pictures, they often change one of them to something else (<> or # or "`")

[fabi1cazenave]

EDIT: I think the intl.toml is wrong. We have \ | two times. From what I’ve seen on pictures, they often change one of them to something else (<> or # or "`")

I agree we could totally change this intl.toml file, but the \| is very common on that physical key. It’s even called NUBS for “non-US BackSlash” in many APIs and apps (e.g. kanata).

[Geobert]

on that physical key.

Which one? The one left of [Z] or the one on the right of the homerow?

[fabi1cazenave]

The ISO key has many names — IntlBackslash, Non-US Backslash, LSGT, 102d… but it’s always the same, i.e. the one between Shift and [Z].

[fabi1cazenave]

This is a valuable move forward and it does solve the issue we’re facing with Ergo‑L at the moment, but I think we should aim for a more “single-responsibility” approach while we’re at it. Namely :

• the klc data in key_codes.yaml should be just a dictionary associating XKB identifiers to scan codes (numbers in hex format would be nice here), like for the other dicts in that data file ;
• the VK thing should probably be implemented as a dedicated KLC function in layout.py, to make it easier to read and debug.

[Geobert]

• the VK thing should probably be implemented as a dedicated KLC function in layout.py, to make it easier to read and debug.

Did you meant "dedicated KLC function in template.py" (not layout.py)

[fabi1cazenave]

I totally meant that, sorry for the confusion.

[fabi1cazenave]

Looks good ! I’ve left a couple comments but nothing worth blocking the PR.