Unable to abort things like consult-line etc. with C-g consistently: have to use C-]

[ashton314]

I'm running Emacs 31.0.50, and I have the most up-to-date versions of Vertico and Consult installed. This is something I noticed recently. I haven't yet been able to get down to a MRE yet, but I wanted to get some early feedback and/or help because this driving me nuts and I'm having a hard time debugging it myself.

When I do consult-line and then try to abort (with or without typing something) with C-g, sometimes it takes me multiple times to actually quit. When it doesn't work I see [Quit] appear for a second or two after the prompt, and I hear the bell sound. I can keep typing in the minibuffer after this happens and vertico/consult remain responsive. Eventually one of the C-g's will get through and I'll actually exit.

Whenever I hit C-] (abort-recursive-edit) then I exit the interface immediately.

What could be causing this? How would I go about debugging such a thing? I'll be trying to bisect my config down to a MRE eventually, but this seems to not happen immediately after startup and only manifests after a few minutes into my editing session, so it's going to be a massive pain to reliably trace down the bug.

[ashton314]

I did a little bisecting and it seems to be something with typst-ts-mode. I'm still having some difficulty reproducing it exactly. The problem does seem to go away if I switch from typst-ts-mode to text-mode and then back to typst-ts-mode again. I will experiment to see if this happens in any other treesit-based modes.

[minad]

I don't really know, and I've never bothered to debug such problems, since it is wild west. C-g is a bit of the weak spots of Elisp imo. It can happen at arbitrary places, potentially leading to inconsistent state (asynchronous exceptions), but it can also be caught anywhere, leading to blocking.