ghostty +show-keys

[00-kat]

Related: #1230 is an old PR for this functionality. (cc @rockorager as you were the author of that PR and you may have opinions about this feature request.)

I couldn't count the number of times I have asked people to install kitten and show what it sees for debugging some issue in a Discord help post or discussion; it would be nice if this could be in Ghostty itself, especially because Kovid does not aim to ensure that his kittens work on other terminal emulators (kovidgoyal/kitty#8058 (comment)). Any such CLI would have to have a mode without the Kitty keyboard protocol too, as that would be very helpful with debugging Fixterms-related issues, the sequences sent when using the legacy encoding, or determining the sequences to use in a text: action when you want to remap a keybinding to another keybinding (related: #5891).

#1230 was closed because:

I'm considering abandoning this - the main motivation was so we could direct people to a ghostty command instead of a kitten, but that was before I realized this information was already in the inspector. Seems like we should just point people there for debugging these things.

However, it has become exceedingly common for the actual keypresses a terminal application receives not being the same as what the inspector shows, often due to the application not supporting the Kitty keyboard protocol.

Mitchell also mentioned:

its not that critical since I don't mind using the Kitty kitten either.

However, asking people to install kitten often results in quite a lot of friction, especially since some users don't know which binary to download (amd64 vs arm64, darwin for macOS, etc.) and some don't know the difference between the GPG key and the binary too.

[mitchellh]

I agree, I think the implementation @rockorager originally did was a bit blunt too (recreated a lot of things we do in core, if I recall). I think if we do this we should be more mindful to refactor or reuse core elements, though I'm not sure what those are yet. I like the simplicity of Kitty's tool, but I think if we were to do it I'd prefer a TUI approach with libvaxis so we can change modes, clear the history, etc. from within the CLI.

[00-kat]

I'd prefer a TUI approach

On this note, I want to mention notcurses-input as prior art.

[rockorager]

I think a TUI for this would be pretty easy. I would need to do some plumbing in libvaxis to expose raw sequences from the parser, but it wouldn't be too bad. Interactive to clear, change mode and flags would be really nice to have too.

[rockorager]

Could change the name and expand scope to let it be a query tool, too.