autowrite when receiving SIGHUP, SIGQUIT about neovim

[dundargoc]

How often does neovim crash that this becomes relevant lol?

Also I don't think it's as obvious as you say that autowrite "should" save on abort. In all cases now autowrite is triggered by a deliberate action by the user, while this will write on any random state it happens to be on. Do the recovery functionality/swap mechanism not essentially fulfill this need anyway?

[justinmk]

preserve_exit is called on deadly signal:

neovim/src/nvim/os/signal.c

Lines 180 to 181 in a2d510e

	// Preserve files and exit.
	preserve_exit(IObuff);

IIUC, that only writes swap files. For SIGHUP/SIGQUIT (but not others like SIGTERM) maybe it makes sense to also write the buffer contents.

[robin-carlier]

@dundargoc

How often does neovim crash that this becomes relevant lol?

It’s not relevant for crashes, but the use case that makes me want such a functionality is the following: I wanted a "quick and dirty" neovim scratchpad windows in my desktop environment, and I’m used to quit windows through my DE shorcuts, which calls a SIGHUP.

I agree that there are multiple workarounds for this particular usage but as @justinmk says for signals like SIGHUP or SIGQUIT (but not SIGTERM, I agree), I thought it would make more sense that neovim goes through its usual exit sequence as a :q would.

[justinmk]

'autowrite' already (should) works like this for :suspend (SIGTSTP). And 'autowriteall' is intended to extend 'autowrite' behavior for "quit" (:help autowriteall mentions "closing the Vim window").

Proposal

• Respect 'autowriteall' for SIGHUP/SIGQUIT.
• Are VimLeave/VimLeavePre triggered on SIGHUP/SIGQUIT? Or should they be? Or should the Signal autocmd allow hooking into SIGHUP/SIGQUIT?

[matveyt]

Are VimLeave/VimLeavePre triggered on SIGHUP/SIGQUIT? Or should they be?

I believe, they should if v:dying == 1 (i.e. only one deadly signal caught), as documented at :h v:dying.