<IdleBot_5e50c57d>
Hm, vhost-user might become a good solution. Qemu works on the implementation; it actually makes a lot of sense for Firecracker, too — hopefully CrosVM also gets something
MichaelRaskin has joined #spectrum
tilpner has quit [Remote host closed the connection]
tilpner has joined #spectrum
<MichaelRaskin>
Oh fun, apparently, unlike Xorg, wayland composers are actually annoying to get running without systemd
<MichaelRaskin>
OK, weston can be started but unclear how to stop (well, pkill works)
<MichaelRaskin>
Hm, at the second attempt it reacts to normal ctrl-alt- stuff
MichaelRaskin has quit [Quit: MichaelRaskin]
<qyliss>
As long as you have elogind you should be fine
<qyliss>
I think
<multi>
which bits of (e)logind does wayland require?
<samueldr>
a non-systemd distro that has wayland stuff working is postmarketOS, if it helps
* multi
nods
<multi>
okay
<multi>
i'm totally illiterate when it comes to wayland; i'm a stubborn debian user running i3wm on xorg :)
<qyliss>
multi: I believe it's "seats"
<qyliss>
You can also run it as root, just like X
<qyliss>
But that's a bad idea
<multi>
ah, like how some distributions have a setuid X wrapper to permit the server to open the framebuffer?
<qyliss>
Yeah exactly
<qyliss>
(e)logind handles that with seats
<qyliss>
I don't really know the details
<multi>
hmm
<multi>
i do wonder if it's possible to shim around that
<multi>
any links to some code which i can confuse myself with?
<qyliss>
not really
<qyliss>
I do know that sway will detect if it's running as root and decline to run
<multi>
i messed around with sway a while back, and it certainly seems opinionated on things like that
<multi>
(inb4 running it with elevated capabilities but not full root...)
<Shell>
There is support in wlroots for direct access to the devices without going through logind. I don’t know whether that’s exposed in sway.
<multi>
interesting
pie__ has joined #spectrum
<qyliss>
why would you not just run elogind
<multi>
irrational aversion to extra moving parts
pie_ has quit [Ping timeout: 276 seconds]
<qyliss>
Wouldn't your custom shim that does who knows what be a moving part?
<multi>
see the "irrational" in my above statement. i don't have any particularly well-founded argument against just running elogind.
<Shell>
multi: if you build sway+wlroots without systemd/elogind support, you can setuid sway and it'll work. it'll drop permissions shortly after startup.
<multi>
neat
<Shell>
this is how sway works on FreeBSD.
<multi>
makes sense
<qyliss>
woooo, got the chroot working
<Shell>
\o/
<qyliss>
having to use the prebuilt one though -- couldn't get it to build from source
<IdleBot_5e50c57d>
Well, elogind will do some things wrong exactly in the same way as logind…
<IdleBot_5e50c57d>
I already have permissions-to-vts code in my system, and I want to spawn wayland sessions on the fly in sequentially allocated vts (like I now sometimes do with Xorg), etc.
<IdleBot_5e50c57d>
BTW, elogind means D-Bus, and non-contained D-Bus is also bad
<qyliss>
that is true
<qyliss>
D-Bus is sadly almost impossible to avoid afaik
<IdleBot_5e50c57d>
Well, with Weston it is not necessary
<IdleBot_5e50c57d>
I am OK with contained D-Bus — isolation avoids all those «crazy interaction» issues
<Shell>
d-bus winds up being used for all sorts of stuff. for example, it's used for desktop notifications iirc.
<qyliss>
And portals, right?
<Shell>
yup!
<IdleBot_5e50c57d>
Portals?
<Shell>
with portals there's a dbus server per app, and a shim which proxies calls from that to the desktop dbus.
* multi
blinks
<multi>
what's the proxy for?
<Shell>
so that the app doesn't have access to the desktop dbus server.
<multi>
fair enough
<Shell>
IdleBot_5e50c57d: portals are a system being developed by the desktop linux folks for sandboxed apps to be able to request additional access ("open file", "take screenshot", etc) from the user.
<Shell>
it's being pushed by the flatpak people, supported by the GNOME/KDE lot, and it's the only standard way to do things like remote control and screencasting with Wayland.