Re: [GNU ELPA] New package: caffeinate-mode

Stéphane Marks <[email protected]> writes:

> On Mon, Jul 20, 2026 at 1:56 PM Philip Kaludercic <[email protected]> wrote:
>
>  Lucas Christian <[email protected]> writes:
>
>  > Philip Kaludercic <[email protected]> writes:
>  >
>  > The main challenge here is that I as a user have to remember what
>  > setting the mode was last in. E.g. if I last used caffeinate with
>  > display sleep blocking, but now I just want system sleep blocking, I
>  > would need to run M-x caffeinate-mode followed by M-x
>  > caffeinate-toggle-inhibit-display-sleep. Whereas with the separate minor
>  > modes, I just call the one I want. Perhaps that's not that big of a deal
>  > though--if someone really feels strongly they can add a few trivial
>  > functions to their init to turn on caffeinate to the preference value
>  > that they want.
>
>  I am not familiar with the system-sleep API, do you think it is feasible
>  to have the option change the behavior of an existing process?
>
> FWIU, caffeinate-mode is not process specific, it's a user-level all-or-nothing election.  To switch from its global display
> allowed to sleep to display not allowed to sleep it has to cancel its sleep token and request a new one with the display option
> specified.  I believe it does that now.  The caffeinate-mode UI for selecting between display sleep or not is the open
> question, I think, not the mechanism of action.

This is correct--to switch between blocking only system sleep, or
blocking system + display sleep, you have to drop the system-sleep
assertion and acquire a new one.

> For integrations like gnus, it relies only on being notified that the
> system has gone to sleep or awakened and does not itself dictate any
> policy.

True, it's worth calling out that there are two functions of the
system-sleep package:

1. The package provides notifications when the system is about to go to
   sleep, and when the system has re-awakened from sleep.

2. The package provides a "power assertion" API, that allows packages to
   request that the system does not go to sleep.

caffeinate-mode is only focused on the second function.

For what it's worth, on my branch of gnus, I integrated both. I use 1 to
disable the gnus demon from attempting to check for new news when the
system is asleep or right after waking (when network access is usually
still re-establishing, so TCP connections almost always stall). I also
use 2 to keep the system from going to sleep mid-operation (fetching new
news or downloading articles offline with the gnus-agent).

-- 
Lucas Christian
[email protected]