qyliss changed the topic of #spectrum to: A compartmentalized operating system | https://spectrum-os.org/ | Logs: https://logs.spectrum-os.org/spectrum/
<qyliss> Sommelier supports the pointer_constraints_unstable_v1 extension now
<IdleBot_5e50c57d> qyliss: I guess chromerelease.top can be had cheaper
<IdleBot_5e50c57d> Argh, rust-9p needs unstable compiler (or RUSTC_BOOTSTRAP=1) and uses deprecated features at the same time
<IdleBot_5e50c57d> Seems to build, will see
pie_ has quit [Ping timeout: 268 seconds]
<IdleBot_5e50c57d> OK, so with similar external isolation and similar options Qemu boots the NixOS test VM fine and CrosVM causes the weird stuff; maybe it is indeed about the finer details of 9P (and then overlayfs comes and it is known to be pretty fragile…)
<IdleBot_5e50c57d> BTW, what is the rationale for mktuntap instead of tunctl + exec 3< ?
<IdleBot_5e50c57d> Looking-at-Sway-wait-what-of-the-day: I cannot choose _which_ WAYLAND_DISPLAY Sway will service?
<IdleBot_5e50c57d> (also, default split orientation cannot be changed in runtime?)
<IdleBot_5e50c57d> Hm, what is the longest possible kernel command line? Reading documentation does not fill me with faith in feasibility of network-based root without initramfs, and initramfs+9p sounds better than ext2+9p or ext4+9p
pie_ has joined #spectrum
pie_ has quit [Ping timeout: 258 seconds]
pie_ has joined #spectrum
<IdleBot_5e50c57d> Ah, tunctl lacks IFF_VNET_HDR. Annoying. Hm, there is also ip tuntap
<qyliss> The advantage to mktuntap, I think, is that the taps it creates aren't persistent
<qyliss> They get cleaned up as soon as the fd is closed
<qyliss> You can't do that if you have one command to open it and then another to put it on an fd
tilpner has quit [Quit: tilpner]
<IdleBot_5e50c57d> Indeed. I guess ip tuntap is good enough as long as I tun the VM in a socat-proxied netns anyway.
pie_ has quit [Ping timeout: 268 seconds]
pie_ has joined #spectrum
samueldr has quit [*.net *.split]
samueldr has joined #spectrum
<IdleBot_5e50c57d> Kernel does not share my enthusiasm about putting the entire init script into kernel command line
<qyliss> lol
<IdleBot_5e50c57d> I actually got pretty far with that strategy, but there is a limit
<IdleBot_5e50c57d> BTW, rust-9p seems to work fine on host; now to setup the networking without giving up on my containment-of-VM ideals
<IdleBot_5e50c57d> squashfs-tools-ng only convert from tar, but not cpio, what a pity
<IdleBot_5e50c57d> OK, no problem. Key part of the boot process: source /dev/vdb