ChanServ changed the topic of #asahi-alt to: Asahi Linux: porting Linux to Apple Silicon macs | User-contributed/unofficial distribution ports | Logs: https://alx.sh/l/asahi-alt
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
KxCORP5894000 has quit [Quit: Bye!]
KxCORP5894000 has joined #asahi-alt
Dan___ has joined #asahi-alt
ellyq has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
ellyq has quit [Ping timeout: 480 seconds]
Dan___ has quit [Ping timeout: 480 seconds]
zerdox has quit [Ping timeout: 480 seconds]
tobhe_ has joined #asahi-alt
tobhe has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Halian has joined #asahi-alt
chadmed__ has quit [Remote host closed the connection]
Dan___ has quit [Ping timeout: 480 seconds]
chadmed__ has joined #asahi-alt
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
zerdox has joined #asahi-alt
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
aead_ has joined #asahi-alt
aead has quit [Ping timeout: 480 seconds]
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
chadmed__ has quit [Quit: Page closed]
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
hightower2 has quit [Remote host closed the connection]
hightower2 has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
JayBeeFOSS has quit [Ping timeout: 480 seconds]
JayBeeFOSS has joined #asahi-alt
pjakobsson_ has quit []
Dan___ has joined #asahi-alt
Dan___ has quit [Ping timeout: 480 seconds]
Dan___ has joined #asahi-alt
chadmed_ has quit [Read error: Connection reset by peer]
Dan___ has quit [Ping timeout: 480 seconds]
zerdox_ has joined #asahi-alt
zerdox has quit [Ping timeout: 480 seconds]
zerdox_ is now known as zerdox
qyliss has quit [Quit: bye]
qyliss has joined #asahi-alt
Dan___ has joined #asahi-alt
zerdox_ has joined #asahi-alt
zerdox has quit [Ping timeout: 480 seconds]
zerdox_ is now known as zerdox
<leio>
chaos_princess: x86_64-multilib-toolchain build fails for me with https://dpaste.com/F7EYRRCSP (config.log https://bpa.st/UE6DS ), as somehow strip-unsupported-flags isn't getting rid of -march=native
<chaos_princess>
leio: fix or delete dnssec on dpaste.com
<leio>
that's just reporting gcc can't compile in configure.log, the interesting stuff is in bpa.st
<chaos_princess>
yea, yea, i already disabled validation, but still
<leio>
it's my default go-to when doing manually because bpa.st doesn't like longer stuff, albeit that wasn't long
<chadmed>
i think we should probably just hard code some bumass basic flags for the glibs c
<chadmed>
0x0.st
<leio>
yeah, right now the only way to build this is to do a package.env that forces -O2 -pipe or similar
<leio>
which is user intervention
<leio>
if I set some x86 -march there, it'll fail in binutils build (and probably gcc stage1 and more), as that's done with native compiler
<chadmed>
we can pick up whichever freaky flags folks want for gcc and its binutils friends but the glibsc should just be -O2 -pipe -fomit-frame-pointer etc etc
<leio>
-fomit-frame-pointer is the default on x86, unfortunately; I'm having to override it everywhere :(
<leio>
(and default on arm64)
<chadmed>
99% of users dont care about it, dont need it, and we want the fex stuff to just be as fast as possible while staying correct (hence no O3)
<chadmed>
buuuut we also have to be very careful with what optimisations we let gcc do, especially in the fex rootfs
<chadmed>
eg we should prevent it from emitting AVX and x87 because these are both dog slow
<chaos_princess>
interesting that it worked here, since i have -march=armv8.4 in cflags
<leio>
I set -march=native -mcpu=native in my gcc flags override, which I need for this due to the glibc build
<leio>
built without any -march/-mcpu/-mtune via the env override, with https://bpa.st/WB3IO
<chadmed>
im buying one of those massive movie theatre popcorn makers for when we release all of this to the general public
<chaos_princess>
same-ish
<leio>
that fails to apply to the overlay version
<chaos_princess>
idk, delete resolvconf from /usr/bin or something
<leio>
I'd rather we improved the out of the box experience :)
<chaos_princess>
well, i am working on it :P
* leio
backports
<chadmed>
i am genuinely 100% of the opinion that the only way this isnt going to suck arse is for linux to learn about folios and mixed-pagesize userspace
zerdox has quit [Remote host closed the connection]
<chadmed>
FEX itself is incredible and Just Works(tm), its all the faffing about with krun and passt and sommelier and...
<leio>
meh, I don't see any net.rs changes in those lines there
n3ph has joined #asahi-alt
<leio>
oh, just confused, was another file
<chaos_princess>
you need to add -e resolvconf=does-not-exist to dhcpcd arguments in crates/krun-guest/src/net.rs
<chaos_princess>
if you keep --nodev, you do not need the udev changes
zerdox has joined #asahi-alt
<leio>
overlay snapshot misses the "Add support for input device passthrough" commit, if that'll matter
<leio>
which is mostly why it doesn't apply
<chaos_princess>
that is correct, it is not packaged yet
<chaos_princess>
since that needs one more package
<leio>
grabbed it for user patches too to just apply the PR as-is, but I gather I might have a broken krun then with the missing package now
<chaos_princess>
nah, it is fine, it will just skip doing that bit if hidpipe is not installed
<leio>
still fails to get rootfs list
<leio>
at the beginning it now says "Using default interface naming scheme 'v253'.", which it didn't before these patches
<chaos_princess>
cat /etc/resolv.conf inside the vm?
<leio>
search lan
<leio>
nameserver 192.168.1.1
<leio>
what does "inside the vm" mean btw, I'm just prefixing with krun - does it have a default VM or smth?
<chaos_princess>
check if internet is working at all there, curl google.com or something
<chaos_princess>
krun starts a vm for your commands, yes
<leio>
could not resolve host
<chaos_princess>
ifconfig?
<leio>
seems to match my host one, except no ipv6
<chaos_princess>
and also curl 1.1.1.1
<leio>
that gets a 301 back with text fine
<chaos_princess>
so it is dns
<chaos_princess>
just to check, 192.168.1.1 is a valid nameserver on your network, yes?
<leio>
krun curl -4 google.com doesn't work either, fwiw
zerdox has quit [Ping timeout: 480 seconds]
<leio>
that's what I use on host, yes
<chaos_princess>
yea, its dns
<chaos_princess>
just to check, what version of passt do you have
<leio>
2024.06.07
<chaos_princess>
what is gentoo package version
<leio>
the same
<chaos_princess>
update it
<leio>
if it helps, please minversion it after appropriately :)
<chaos_princess>
sure
<chaos_princess>
i am still looking at what went wrong with x86-multilib-toolchain, but i'll get to it
<leio>
will take time to clean up my ::gentoo local tree to pull updates in
<leio>
and running out of todays poking time for now soon
<leio>
chadmed: I guess if it's some mesa from COPR, I won't have vulkan in there
<chadmed>
nope
<chadmed>
doesnt really matter anyway like i said in the dms sommelier is so crap that its effectively useless
<chadmed>
steam looks awful, games run in a 1280x720 window with no decorations and smudged by crap scaling
<chaos_princess>
chadmed: how brave are you? :P
<chadmed>
i played i won the loudness war on all of my macs and compiled the kernel for 12 hours on a loop with the fans forced to off
<chadmed>
i can open glxgears but it doesnt render, its just the first frame
<chadmed>
the window has decorations and behaves like a normal window in kwin though!
<chaos_princess>
then dax is not working
<chaos_princess>
give me your mount output inside the vm
<chadmed>
let me try something
<chadmed>
bash: /usr/bin/sudo: Permission denied
<chadmed>
your libkrun branch, your krun branch
<chaos_princess>
right, sudo does not work in a vm
<chaos_princess>
period
<chadmed>
i did chmod u+s x112virtgpu too
<chaos_princess>
no, i know the problem, it is due to cross-vm mmap not working, you need to check if vm mounted the fs properly
<chaos_princess>
make sure the root mount has dax=always in options
<chadmed>
in the vm?
<chaos_princess>
yes
<chadmed>
how do i check that without root access lmfao
<chaos_princess>
cat /proc/mounts
<chadmed>
*facepalm*
<chadmed>
yeah it has dax=always
<chaos_princess>
can you give me all the commands you are using to launch all that
<chadmed>
krun bash
<chadmed>
unset WAYLAND_DISPLAY
<chadmed>
export DISPLAY=:0
<chadmed>
x112virtgpu &
<chadmed>
glxgears
<chadmed>
dmesg has a dax related splat in it
n3ph has quit [Ping timeout: 480 seconds]
<chaos_princess>
did glxgears detach from the terminal?
<chadmed>
yeah
iyes has quit [Quit: Konversation terminated!]
<chaos_princess>
no output from x112virtgpu?
<chadmed>
nup
<chaos_princess>
can you also do a ls -la /tmp/.X11-unix/ inside the vm
<chaos_princess>
and outside too
<chaos_princess>
actually, nvm, shouldnt be the issue
<chadmed>
the vm has X0 and X1, host has X0 only
<chadmed>
perms etc all look good
<chaos_princess>
then do ls /dev/shm, and also try resizing the window of glxgears, just in case
<chadmed>
resizing the window just makes it black
<chadmed>
nothing in /dev/shm
<chaos_princess>
can you try with LD_PRELOAD then?
<chadmed>
do i preload that into glxgears or into x112virtgpu
<chaos_princess>
glxgears
<chadmed>
bc if the former, already tried and didnt help
<chadmed>
oop now it works
<chadmed>
wtf
<chadmed>
nice!
<chaos_princess>
wtf
<chaos_princess>
i am now more confused than ever
<chadmed>
oh now it exploded when i resized
<chaos_princess>
with our without preload?
<chaos_princess>
and how exactly did it explode
<chadmed>
with preload, and the window just went black and did not recover
<chadmed>
glxinfo also takes out x112virtgpu
<chadmed>
this is on the m2 air btw, ill try with the studio tomorrow
<chaos_princess>
yes, resize is part where most bugs are expected, since this results in me having to do some super nasty things with cros-vm mmaps and all of this is just horrible
<chaos_princess>
i was testing it on m1 pro, maybe it could be an issue?
<chaos_princess>
idk
Dementor9 has quit [Ping timeout: 480 seconds]
<chadmed>
i can test with that too tomorrow
<chadmed>
cbf going back downstairs now though
<chadmed>
killing a window also doesnt signal its process to die but thats chill who cares this is already 9999999x better than sommelier
<chaos_princess>
No, that one is specifically glxgears
<chaos_princess>
same under sommelier
<chadmed>
everything just looks so much better
<leio>
curl still does not resolve after passt upgrade
<chaos_princess>
krun bash, kill dhcpcd, run it with -d, see how it fails