gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
<ldlework> no help in #pulseaudio
<ldlework> big surprise
<energizer> i wish you could cash in the help that you give in one channel in exchange for getting help in another channel
<ldlework> seriousl
<ldlework> y
<ldlework> why isn't audio solved in 2020
<bbigras> pipewire is coming
<bbigras> did you try it? we have it. The devs were quick to fix that one bug I had once.
<colemickens> I think I have it enabled but not replacing pulseaudio yet
<ldlework> I don't know anything about audio. I don't /want to know anything about audio/
* colemickens shrugs
<ldlework> I want to flip a switch in my nix config and have it work :/
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-chat
<ldlework> How there is not a standard sink/source graph-editor-wire-up thing that all linux audio uses is beyond me.
<ldlework> Right click, this is my default sink.
<ldlework> Right click, this is my default source.
<ldlework> This timeline sucks.
<lukegb> Pulseaudio 14 at least fixes my major annoyance with default sinks
<lukegb> Which is, well, changing the default didn't change where things were going
<ldlework> PA isn't the answer since it doesn't have a "wire things up" metaphor.
<lukegb> I'm sure this will break someone's workflow buuuut it makes my life easier
<lukegb> ldlework: oh sure
<lukegb> Clearly everything should be one big Gstreamer pipeline
<energizer> JACK is the wire-things-up tool atm
<ldlework> Yes, but it isn't standard and the integration in Nix is broke af.
<lukegb> That way it can hang and segfault at random inopportune moments
<ldlework> The need to integrate with PA is what breaks Jack.
<energizer> i dont understand how the jack nixos module was supposed to work so i wrote my own that's probably terrible in a million ways...but it did work
<ldlework> I assume my problem is something like, some buffer size, or flush frequency, in the pulseaudio sink is just totally inappropriate.
<ldlework> So while it technically works, it's just produces garbage audio.
<ldlework> But that's literally just a shot in the dark.
<ldlework> We need to start a fund...
<ldlework> energizer: did you implement PA integration?
<energizer> ldlework: i dont think i did
<ldlework> how do you use spotify and so many other things?
<energizer> it was for a specific audio situation, not my daily driver
<energizer> i think simpson has a jack config, not sure if it's public
<ldlework> The NixOS config sets up a cloop and ploop alsa loop back devices
<ldlework> here's simple_jack_client piping tones into ploop
<ldlework> and cloop piped into my speakers
<ldlework> no audio
<ldlework> here's simple_jack_client pipng tones into the pulseaudio sink
<ldlework> and the source piping into my computer's speakers
<ldlework> hit Discord's monitor, no audio
<ldlework> like what the actual hell
<ldlework> How about we take Jack support out if it doesn't work?
<ldlework> Or rather, the PA/Alsa support
<ldlework> if I pipe simple_jack_client directly into the speakers, i hear the tones
<ldlework> If it actually does work, then it isn't properly documented as while I'm not the smartest, I have been touching computers for over 20 years
<ldlework> Audacity doesn't even get any signal. Not even the BOOOP BOOOP BOOOP from the volume going up and down even though it's a constant tone.
rajivr has joined #nixos-chat
<bbigras> "PipeWire is a project that aims to greatly improve handling of audio and video under Linux.
<bbigras> It provides a low-latency, graph based processing engine on top of audio and video devices that can be used to support the use cases currently handled by both pulseaudio and JACK. "
<bbigras> sounds like the holy grail of audio
<bbigras> no pun intended
<bbigras> pipewire has a pulseaudio compatibility thing but they changed it. I think we need to wait for our module to support it.
<lovesegfault> Pipewire is also, AIUI, promising to fix the screencasting mess that currently exists
<lovesegfault> although that's such a hefty promise that I don't know what will be of it
<ashkitten> bbigras: pulseaudio compatibility works great since #105362
<{^_^}> https://github.com/NixOS/nixpkgs/pull/105362 (by gebner, 4 days ago, merged): pipewire: 0.3.16 -> 0.3.17
<ashkitten> it now provides a pipewire-pulse systemd service
<bbigras> oh yeah pipewire helps with screencasting on wayland. No idea how an audio thing does that.
<ashkitten> and i've been using it as my daily driver for a while
<ashkitten> it's not just an audio thing
<ashkitten> it handles video too
<bbigras> ashkitten: "Upstream has deprecated the libpulseaudio.so shim and provides a systemd service to load pulse-bridge now". are you using the bridge yet? I'm not sure if we are still waiting for the module to support it.
<ashkitten> it can also transport dmabuf fds for copyless screencasting
<ashkitten> bbigras: we are as of the pr i linked
<ashkitten> it's probably not in nixos-unstable quite yet
<ashkitten> since it was merged 9 hours ago
<bbigras> ashkitten: oh sorry I thought it was just a version bump. oh sweet. I'm going to use it as my daily driver asap too. thanks
<energizer> is anybody using only pipewire?
<ashkitten> me
<energizer> already
<energizer> not jack and not pulseaudio?
<bbigras> does program work without the pulseaudio bridge? do you need alsa or jack?
<ashkitten> yes
<energizer> ldlework: ^
<ldlework> it's not just one program, i use a lot of programs
<ashkitten> bbigras: the pipewire module has also, jack, and pulse options that can be separately enabled
<ldlework> ashkitten: this is a thing in nixos already?
<ashkitten> bbigras: also note that the alsa support has a support32Bit option
<ashkitten> ldlework: it's been merged into master but 0.3.17 is not in nixos-unstable yet
<bbigras> ashkitten: yeah I saw the alsa and jack support but I mean, if I enable pipewire and disable pulseaudio and jack support, will sound work in every program?
<bbigras> I don't mind enabling them I'm just curious.
<ashkitten> if you disable pulseaudio and jack, then programs that use those apis will not work
<bbigras> that's what I thought. and I expect everything to work if I enable alsa and pulseaudio support.
<ashkitten> yeah
<bbigras> sweet
<lovesegfault> ashkitten: is there a pavucontrol for pipewire
<ashkitten> if you do enable jack support you will be able to use programs like qjackctl to connect nodes
<ashkitten> lovesegfault: it's called pavucontrol
<ashkitten> :)
<lovesegfault> O.o
<ashkitten> there are no gui programs i'm aware of that currently support pipewire itself natively
<ashkitten> but alsa, pulse, and jack control programs should all work
<lovesegfault> pretty cool
<energizer> ashkitten: can you do bluetooth voice calls, like windows/macos/ios/android can?
<ashkitten> here's a screenshot of qjackctl showing my microphone and firefox both plugged into discord's voice input
<lovesegfault> ashkitten: pretty darn dope
<ashkitten> energizer: supposedly pipewire's plan is to have significantly better bluetooth support than pulseaudio, but i don't know anything about that since i don't use bluetooth
<ashkitten> lovesegfault: in the screenshot both firefox and discord think they're talking to pulseaudio, and they're being displayed in a jack patchbay
<ashkitten> for the record, i've not gotten pipewire screen sharing working yet (but it's not that important to me)
<lovesegfault> That's what I'm most excited about
<ashkitten> it depends on xdg-desktop-portal being configured correctly plus some other stuff
<lovesegfault> I never got xdg-desktop-portal working well in sway
<lovesegfault> All I want is zoom screensharing to work
<ashkitten> it should work fairly flawlessly if you use gnome-shell, i hear
<lovesegfault> It works okay under gnome, yeah
<lovesegfault> but for some reason gnome is _really_ slow for me
<lovesegfault> like, moving the mouse is slow
<ashkitten> gnome's xdg desktop portal is way more complete than the other implementations
<ashkitten> xdg-desktop-portal-wlr might get output selection soon, if progress is actually made on that pr
<ashkitten> but there's no support for capturing individual windows in wlroots yet
<ashkitten> (there is in gnome)
<ashkitten> xdg-desktop-portal-wlr also doesn't support dmabuf yet, so capture might be somewhat laggy as it has to copy buffers instead of passing around gpu memory handles
<ashkitten> no idea if gnome supports that
<pie_> ashkitten: so is this jack but better and not limited to audio?
<ashkitten> sorta
<energizer> pie_: pipewire is intended as a replacement for jack+pulseaudio
<pie_> uhuh
<pie_> th
<pie_> x
andi- has quit [Ping timeout: 272 seconds]
<colemickens> twilio locked me out of my account again and told me they'd have to escalate to engineering to unlock <email that is not my email> account. holy wtf.
<abathur> lolwut
<colemickens> if I wasn't already questioning some security competency based on the initial lockout....
<abathur> hehe
<colemickens> I just spammed back "DO NOT GRANT ME ACCESS TO THAT" as fast as I saw it.
<abathur> if it was a real human, I guess they may have still had something from the previous ticket on their clipboard
<ashkitten> they're social engineering themselves!
<abathur> turing-complete support
<V> colemickens: my goodness
<samueldr> whew
andi- has joined #nixos-chat
<siraben> wowza
supersandro2000 has quit [Read error: Connection reset by peer]
supersandro2000 has joined #nixos-chat
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-chat
<bbigras> energizer: you can track https://github.com/NixOS/nixpkgs/issues/102547 . There was also a comment about Bluetooth 4 days ago
<{^_^}> #102547 (by jansol, 4 weeks ago, open): Pipewire status tracking issue
<energizer> bbigras: not sure that supports voice
<samueldr> are compositor developers allergic to screenshots?
<bbigras> energizer: oh yeah no idea. I'll try to remember to try it when nixos-unstable moves forward
supersandro2000 has quit [Quit: The Lounge - https://thelounge.chat]
<bbigras> oh it did. 4 hours ago.
supersandro2000 has joined #nixos-chat
<samueldr> hmm... running Xephyr in wayland is causing troublesome because some apps "escape"
<samueldr> how do wayland apps know of wayland?
<samueldr> env | grep -i way # gives nothing
<samueldr> $XDG_RUNTIME_DIR/wayland- apparently
<samueldr> :/
<ashkitten> samueldr: WAYLAND_DISPLAY is set on sway
<samueldr> I'm not even running the apps from the same process tree and it escapes into wayland-land
<samueldr> so it's definitely not environmental
<ashkitten> interesting
<samueldr> indeed, and not good for me
<samueldr> not sure it's actually using those sockets, and what is the method it decides to use them or not
<samueldr> because "assuming compositors are using the wayland- prefix"
<samueldr> and how does it cope with running two concurrent compositors anyway?
endformationage has quit [Quit: WeeChat 2.9]
<samueldr> yep, fresh ssh session, mpv [...] and it just launches in wayland
<abathur> meh, I have things to do and there's a new CS:GO operation
<samueldr> indeed it goes throuhg the socket, connect(7, {sa_family=AF_UNIX, sun_path="/run/user/1000/wayland-0"}, 27) = 0
<samueldr> don't tell me every toolkit will implement weird varied methods to detect wayland automagically?
* samueldr clones mpv
<samueldr> export WAYLAND_DISPLAY=foobar # and it seems to block mpv
<samueldr> interesting though, as it would allow singling out mpv
<samueldr> by making it live in a wrapper that can undo a neuter
<samueldr> it's documented
<samueldr> >> The protocol is sent over a UNIX domain stream socket, where the endpoint usually is named wayland-0 (although it can be changed via WAYLAND_DISPLAY in the environment).
<samueldr> but seemingly shouldn't act that way
<abathur> the way of the shell
<abathur> find an arbitrary artifact
<abathur> exalt it
<abathur> assume it is eternal
<abathur> codify it
<samueldr> I really don't like how wayland already has legacy :/
<samueldr> especially given how idealistic the design is
<samueldr> while it almost works, having wayland magically be escapable into really makes this a non-starter
<samueldr> cage that starts Xephyr, that is
<samueldr> redshift (relatively understandbly) can't work as-is, so I'd have to figure out if cage does that
<samueldr> but apps being wayland-promiscuous isn't great
<samueldr> even hiding the socket using the environment var i3lock somehow sees it's running under wayland (??)
<samueldr> (yes, I'm using it wrong)
<samueldr> (or am I?)
<ashkitten> why are you using i3lock in wayland
<samueldr> because I want to have my cake and eat it too
<ashkitten> what does that mean
<samueldr> I was typing more :)
<samueldr> I don't actually want wayland (yet)
<samueldr> I don't have the time to get into a new DE
<samueldr> but on a particular computer it could help with perfs for a *specific* use
<samueldr> mpv videos
<ashkitten> so just use swaylock?
<samueldr> does swaylock work in cage?
<ashkitten> what is cage
<samueldr> I don't want to use sway
<samueldr> cage is a compositor
<samueldr> it runs a single app
<samueldr> well
<samueldr> more subtle than that
<samueldr> single maximised window at a time
<samueldr> with that I can run X11 through Xephyr, most things working
<samueldr> things broken being those that are understandable like manipulating the server Gamma (redshift)
<samueldr> but i3lock somehow going out of its way to detect there's a wayland somewhere it... not good
<samueldr> so I was wrong, I had the env variable for neutering up
<samueldr> it has a naïve check
<samueldr> now, how can I even know which wayland protocols a compositor implements, if it's not documented clearly
<samueldr> ashkitten: yeah, I just correct myself
<ashkitten> use wayland-info
<samueldr> thanks
<samueldr> ,locate bin wayland-info
<{^_^}> Found in packages: wayland-utils
<samueldr> I guess my cache is really stale
<ashkitten> it used to be called weston-info in the weston package
<samueldr> btw, using xephyr in cage is a suggestion I've seen repeated often as a solution to the (hostile) "deprecation" of the xorg server
<samueldr> and I can see it's really not
<ashkitten> i don't know why you want to do that, though
<ashkitten> doesn't xwayland work fine for you?
<samueldr> it's not usable to run a WM
<ashkitten> oh
<ashkitten> i mean
<samueldr> circling back to my wanting to continue using my setup
<samueldr> since it seems everyone wants to make X11 a persona non-grata
<samueldr> cage implements 0/4 protcols
<samueldr> (needed for swaylock)
<ashkitten> doubtful
<samueldr> or wayland-info is not the tool to get that info :)
<ashkitten> i mean, you could always just try it
<samueldr> really, I'm only a couple hours of wayland experience in
<samueldr> I'm sure if my workflow existed already in wayland things would feel better
<ashkitten> wayland-info | grep zxdg_output_manager_v1
<ashkitten> samueldr: what are you trying to do, anyway?
<samueldr> use a computer
<ashkitten> i'm trying to be helpful
<samueldr> I know
<samueldr> there's no exact end-goal other than being able to use that specific computer
<samueldr> where mpv through X11 is terrible
<samueldr> (and other X11 use)
<ashkitten> why would mpv through x11 through wayland be any better?
<samueldr> at least I can confirm what was said online, it *does* show that the intel X11 driver has terrible issues that aren't seen in wayland with DRM
<samueldr> I want to allow only select applications (here mpv) to break-out
<samueldr> so in this instance it would be mpv on wayland
<ashkitten> "break-out"
<samueldr> in this particular situation, that's it
<ashkitten> do you know that's a thing that's possible?
<samueldr> cage -> xwayland -> xephyr -> awesome
<samueldr> yes
<samueldr> when I start WAYLAND_DISPLAY=:0 mpv [...] it works, obviously mpv is not managed by awesome
<samueldr> it's really close to being a workable workaround
<ashkitten> isn't way-cooler a clone of awesome for wayland
<samueldr> isn't it abandoned?
<ashkitten> idk
<samueldr> can't quickly find the announcement, but it was, their repo is currently archived
<ashkitten> oh
<samueldr> and it's not only the WM that's part of my workflow
<samueldr> hm
<samueldr> that's not it
<samueldr> that is it
<samueldr> so yeah, when I'll want to use wayland, I'll have to move to a completely new workflow, WM, and tooling around
<samueldr> so it won't be a single evening project for me
<samueldr> I somehow think that last result was inferred because way-cooler is in Rust
<ashkitten> anyway, idk just try swaylock and see if it works
<ashkitten> otherwise i guess i3lock will probably lock your nested x11 as well as any x11 screen locker can...
<samueldr> `yep
<samueldr> i3lock indeed just does that
<samueldr> swaylock: Aborted (core dumped)
<samueldr> oddly appropriate
<ashkitten> fwiw i agree that x11 being pretty much maintenance-mode is a shame for certain devices
<ashkitten> wayland is definitely getting there but the ecosystem is not quite complete yet
<samueldr> yeah :/
<ashkitten> i have to switch to x11 for vr games still
<samueldr> I was hopeful it was an actually usable solution for transplanting an X11 workflow into a non-xorg
<samueldr> because if it did work, I'd have another device that might have benefited from it
<ashkitten> xephyr in cage?
<samueldr> yeah
<ashkitten> why isn't it working for you?
<samueldr> main reason is that wayland apps can break-out
<samueldr> without my intervention
<ashkitten> they shouldn't be able to if they can't see the wayland socket
<samueldr> because of that legacy thing with WAYLAND_DISPLAY
<samueldr> if it's unset they assume the socket is usable if it exists
<ashkitten> also
<samueldr> maybe I could move the socket by patching cage
<ashkitten> the legacy wayland-0 thing is why most compositors name their default socket wayland-1
<samueldr> :|
<ashkitten> so just change the name to wayland-1
<samueldr> "just" change
<ashkitten> in your compositor
<samueldr> yeah, "just"
<samueldr> it doesn't look like it's an option with cage
<ashkitten> it's gonna be a text string somewhere in the source, i'm sure you can patch it
<samueldr> probably in the lib it builts onto
<samueldr> because it's not part of cage itself
<samueldr> wlroots
<samueldr> might even be wayland itself (??)
<samueldr> wl_display_add_socket_auto
<samueldr> yep
<samueldr> I'm really not fond of that muck about the wayland display socket magic
<ashkitten> ah
<ashkitten> dunno
<ashkitten> what if you set WAYLAND_DISPLAY to a wrong value
<samueldr> depends on the app :)
<samueldr> variable results
<samueldr> but the few wayland apps I looked at will think wayland is unusable
<samueldr> but i3lock here assumes you're running on wayland
<ashkitten> i see
<samueldr> it doesn't sit right :/
<ashkitten> well, this is a pretty weird use case :/
<samueldr> yep, and now I know it doesn't work :)
<samueldr> so this pushes back wayland to whenever I have the time to upend my usage workflow
<samueldr> hmm... could I start cage+mpv from full blown X11?
<samueldr> I think it wants a VT though
<samueldr> (I'm not sitting with the device, so there's a delay until I try things, since I need to get up)
<ashkitten> what does mpv do wrong under x11?
<samueldr> it's more about what xorg does wrong I think
<samueldr> the intel driver doesn't deal well enough with rotation
<samueldr> and maybe isn't as fast with accelerated decoding on cherry trail with xorg
<samueldr> ugh, disregard the past few hours of my life
<samueldr> looks like the options I ended up using under wayland works fine under X11 too
<samueldr> so it looks like my issue all along was the lack of hardware decoding to be turned on
<samueldr> (I have read more about mpv since I started testing in wayland)
<ashkitten> heh
<samueldr> and hwdec has a weird (imo) default where it does none at all
<samueldr> you're supposed to prefer using CTRL+H to toggle it on
<samueldr> and not configure it to hwdec=auto (or auto-safe)
<ashkitten> i was going to suggest patching i3lock to ignore WAYLAND_DISPLAY and set it to an incorrect value
<samueldr> so, mpv on cherry trail, use hwdec=auto-safe and it'll be smooth
<samueldr> I was going to be a bit more round-about and put a dummy file at the usual socket location so cage would start at -1 lol
<samueldr> hopefully nothing fallbacks on wayland-1
<samueldr> but yeah, not needed anymore
<samueldr> (note that I also had to deal with X11 needing correct configuration to not be terrible too)
<lovesegfault> ,locate iw
<{^_^}> Found in packages: iw
<lovesegfault> derp
waleee-cl has quit [Quit: Connection closed for inactivity]
kalbasit has quit [Ping timeout: 256 seconds]
neeasade has quit []
cole-h has quit [Ping timeout: 240 seconds]
<lovesegfault> I need suggestions
<lovesegfault> some time over the past couple months SSH has become unusably slow for any data transfer
<lovesegfault> rsync nets like 600kbps from two local machines
<lovesegfault> No clue what changed
<energizer> if you rsync from A to an external server, does it happen? what about from B?
<energizer> what about to A and to B?
<energizer> can you reproduce it with iperf?
<lovesegfault> I tried, but not really
<lovesegfault> I also have some weirdness where my ssh just hard freezes
<energizer> i've seen that before
<energizer> can you pin it down to one computer?
<energizer> or does it happen on both
<lovesegfault> My laptop seems to be the common factor here
<energizer> can you consistently reproduce it?
<lovesegfault> I can consistently reproduce the following:
<lovesegfault> 1. My laptop, hereby machine A, can ssh into a local server, hereby machine B and it works fine
<lovesegfault> 2. A is connected to the router via 5GHz Wifi. B is connected to the router via 2.5gbps lan. They can both see each other consistently
<lovesegfault> 3. An rsync from A to B using tailscale (i.e. going over the internet) saturates my internet speeds at around 60MBps
<lovesegfault> 4. An rsync from A to B using local IP ( is <1MBps
<lovesegfault> 5. nix-copy-closure from A to C, a machine in a data center, consistently breaks the mux'd ssh connection
<energizer> if you use a small --block-size value does it still get stuck?
<lovesegfault> on what?
<energizer> rsync
<lovesegfault> let me try
<lovesegfault> rsycing this 6.2G test.random
<lovesegfault> also: define small
<lovesegfault> 1M?
<energizer> try 1K
<energizer> 1024
<lovesegfault> 165,249,024 2% 1.19MB/s 1:27:45
<lovesegfault> dips down to 200k at times
<energizer> seems fine
<lovesegfault> now let me show you over tailscale
<lovesegfault> 336,199,680 5% 22.86MB/s 0:04:26
<energizer> also fine
<lovesegfault> This is it going over the internet and with wireguard encryption
<lovesegfault> it's faster than the LAN!
<lovesegfault> how is that possible
<samueldr> it's just that good
<lovesegfault> I can't understand what's going on
<energizer> what happens if you ssh laptop and from within the ssh session, cat /dev/urandom
<energizer> does it hang?
<lovesegfault> you mean from laptop to machine/
<lovesegfault> i.e ssh from machine A to machine B
<lovesegfault> then cat urandom
<energizer> no
<energizer> well try both i guess
<ashkitten> lovesegfault: what happens when you traceroute to the machine that's being slow?
<lovesegfault> ashkitten: trying
<lovesegfault> All I get are * * *'s
<ashkitten> it's a routing issue, then
<energizer> maybe, maybe not
<energizer> lots of machines don't respond to icmp
<ashkitten> then traceroute to it over the other connection
<lovesegfault> both the local IP and the tailscale IP give me * * * all the time
<ashkitten> ah
<lovesegfault> I own the machine, I can make it respond to icmp if that helps
<ashkitten> that could help
<lovesegfault> oh, if I use tracerout -I it works
<ashkitten> hm
<lovesegfault> idea: going to enable wifi on that other machine, see if it's as slow as the ethernet
<energizer> ssh server cat /dev/urandom | pv > /dev/null
<lovesegfault> 55MiB/s
<lovesegfault> which is about saturating the 5GHz wifi, I think
<lovesegfault> way way way more than the 1MiB/s I get on rsync
<energizer> rsync has to do more work
<energizer> checksums and stuff
<lovesegfault> energizer: scp only does 1MB/s as well
<ashkitten> hmm
<ashkitten> lovesegfault: are you getting packet loss?
<ashkitten> maybe try ping flood to check for packet loss
<lovesegfault> ashkitten: how can I check?
<ashkitten> ping -f (as root)
<ashkitten> or you could try iperf3
<ashkitten> but it's very strange that cat over ssh is working fine...
<ashkitten> (since that's essentially exactly what scp does)
<lovesegfault> scp over wifi is faster than over ethernet, lel
<lovesegfault> trying the ping flood
<lovesegfault> do I just ctrl-c the ping after a bit?
<lovesegfault> it's just displaying a lil dot
<ashkitten> yes
<lovesegfault> 36588 packets transmitted, 36588 received, 0% packet loss, time 38204ms
<lovesegfault> rtt min/avg/max/mdev = 0.519/1.034/40.277/1.575 ms, pipe 4, ipg/ewma 1.044/0.781 ms
<lovesegfault> lol
<ashkitten> so it's clearly not the physical link
* lovesegfault nods
<ashkitten> wtf lol
<lovesegfault> Yeah, it's really weird
<ashkitten> it's only scp and rsync that are doing this?
<ashkitten> is there anything in logs?
<lovesegfault> That I've noticed
<lovesegfault> which logs? sshd?
<ashkitten> idk lol like just check all of journalctl
<lovesegfault> checking host machine
<lovesegfault> nothing interesting
<lovesegfault> the server has nothing interesting either
<eyJhb> lovesegfault: could you try iperf3 ?
<lovesegfault> idea: going to comment out chunks of my ssh config
<lovesegfault> sure thing
<lovesegfault> let me enable it on the server
<eyJhb> Just do a nix-shell -p iperf3 and then I think iperf3 -s, no need to enable the module ;D
<lovesegfault> Oh, I had it enabled already
<lovesegfault> trying
<lovesegfault> done
<samueldr> so, laptop[wifi] <-> nas[wired] is slow, right?
<samueldr> and was it laptop[wifi] <-> nas[wifi] that was faster?
<samueldr> can you wire-up your laptop?
<lovesegfault> samueldr: laptop[wifi] <-> router <-> nas[2.5gbps eth] is slow
<samueldr> can you test another device on the same wi-fi network to the nas?
<lovesegfault> laptop[wifi] <-> router <-> tailscale <-> router <-> nas[2.5gps eth] is fast
<samueldr> yeah, I skipped over tailscale
<lovesegfault> i.e. going over the internet is faster than doing a simple lan transfer
<samueldr> I wonder if somehow your network situation would make connections between the wireless and wired stuff... break
<ashkitten> lovesegfault: i have an idea. load the connection by doing `ssh server cat /dev/urandom > /dev/null`, and then ping the server to check latency
<lovesegfault> ashkitten: one second, taking iperf over tailscale
<lovesegfault> LOL
<lovesegfault> iperf over tailscale fully saturating my link
<lovesegfault> or, close to
<ashkitten> lol
<lovesegfault> trying your idea
<eyJhb> Your setup is weird lovesegfault !
<lovesegfault> alright, 56MiB/s on the cat
<lovesegfault> now pinging
<lovesegfault> ashkitten: do you want flood ping or normal?
<ashkitten> eh, what the heck. try both
<ashkitten> we're scraping the bottom of the barrel now :p
<lovesegfault> :D
<lovesegfault> transfer did not dip below 56MiB/s
<lovesegfault> eyJhb: it's bizarre!
<ashkitten> ugh
<ashkitten> uhhh
<lovesegfault> and it wasn't always like this, something happened
<lovesegfault> it's a mystery
<lovesegfault> this really befuddles me
<lovesegfault> Oh, I have an idea
<lovesegfault> I wonder what happens if I try to zfs send a phat dataset
* lovesegfault tries
<ashkitten> it's clearly a latency issue
<lovesegfault> Why do you say that?
<ashkitten> rsync and scp are latency sensitive
<lovesegfault> Ah
<ashkitten> but cat over ssh isn't
<lovesegfault> do you want me to try doing an rsync and then checking ping?
<ashkitten> hmmm
<ashkitten> actually
<lovesegfault> transfer is going at ~300kB/s
<lovesegfault> but the ping is still showing good latency
<ashkitten> i'm looking back at that iperf3 test
<ashkitten> do you see the number of retries?
<ashkitten> that's ridiculous
* lovesegfault goes look
<lovesegfault> Oh, yeah
<lovesegfault> it retries _WAY_ more on the local transfer
<lovesegfault> 9613 vs 189
<ashkitten> so there's some packet loss i guess
<lovesegfault> I wonder if something in here is biting me
<ashkitten> have you documented why you set each of these options?
<lovesegfault> ashkitten: but how is it possible that A<->R<->B has more loss than A<->R<->......<->R<->B
<lovesegfault> ashkitten: nope, it's like an old family heirloom by now
* lovesegfault removes it from config
<lovesegfault> deploying to the server and then to my local system, and then rebooting
<lovesegfault> brb
<lovesegfault> back
<lovesegfault> I see about 9MB/s on rsync now!
<lovesegfault> still slow as hell, but it improved!
<lovesegfault> let's see iperf
<lovesegfault> ashkitten: [ 5] 0.00-30.00 sec 275 MBytes 76.8 Mbits/sec 11170 sender
<lovesegfault> still loads of retries
<ashkitten> much better throughput though
* lovesegfault runs through tailscale
<lovesegfault> [ 5] 0.00-30.00 sec 1.68 GBytes 482 Mbits/sec 1273 sender
* lovesegfault scratches head more
abathur has quit [Ping timeout: 256 seconds]
<lovesegfault> ashkitten: could I use wireshark to figure out what's happening?
<ashkitten> i don't think so
<ashkitten> have you tried uh
<ashkitten> a different router :/
<lovesegfault> That's the saddest part
<lovesegfault> it's a really nice router
<lovesegfault> effing netgear, I'll never buy this trash again
* lovesegfault sighs
<ashkitten> :<
<ashkitten> there might be some settings in your router
<eyJhb> There's always #networking, they will netork the shit out of it
<ashkitten> but idk
<eyJhb> QOS
<lovesegfault> eyJhb: hum?
<lovesegfault> QOS?
<eyJhb> `Quality of Service`, some routers have this
<eyJhb> And can make weird things happen at times
<ashkitten> it's basically dynamic packet prioritization
<lovesegfault> Should be off
<lovesegfault> I'll try #networking, let's see
<lovesegfault> I appreciate everyone trying to help :D
<lovesegfault> it's been annoying me for weeks now
<eyJhb> I have lived with a broken gigabit switch for 2-3 years... The port I used was only 100 Mbit/s. I just thought it was normal, and that the local server was slow. I saw the light afterwards!
<eyJhb> I feel your pain lovesegfault
<eyJhb> I will be stalking you in #networking as well ;)
<ashkitten> me too... i have an interest now :p
<lovesegfault> :D
<lovesegfault> just asked the !
<lovesegfault> *Q
<eyJhb> lovesegfault: Does tailscale have some cache?
<eyJhb> Someplace?
<lovesegfault> not AFAIK
<eyJhb> :(
<energizer> i dont think you've shown that it's tailscale related
<lovesegfault> regardless, if I just use normal internet it works the same
<eyJhb> If you don't get any answers, then I guess they are like us "shit fucked yo'" :D
<lovesegfault> :D
<lovesegfault> I asked the SREs at work who told me to turn it off and on again
<energizer> lol not a good strategy
<eyJhb> Site Reliability Engineer?
* lovesegfault nods
<eyJhb> Do you maybe run a lot of Windows at work?
<eyJhb> :p
<lovesegfault> No windows at all
<eyJhb> Poor Linux servers :(
<patagonicus> Oh. Was wondering why the wayland/ozone stuff didn't work for me in Chrome, even though it should be included by default. But Nixpkgs is disabling it by default, at least for Chromium. Let's see how long it takes to build on my laptop.
<lovesegfault> patagonicus: it will take a long time :P
<lovesegfault> and you better have a lot of ram
<patagonicus> lovesegfault: Pff, I'm used to building whole NixOS on armv7 machines, can't be that much worse.
<lovesegfault> :D
<patagonicus> 100GB of disk space? Nice. I wonder if that also applies to the NixOS build, because then I'll have to resize my /
<eyJhb> How much ram you got patagonicus ? :p
<patagonicus> 16GB + 10GB swap currently. The Chromium docs say that it should work with 16.
<eyJhb> _should_ :D
<eyJhb> I could not build FF on my computer with 16 GB of ram. which was weird
<patagonicus> Ah, no, 8GB is minimum.
<eyJhb> Still mad about ITS downsizing my instance, so now we can only run 4 tests, instead of 32...
abathur has joined #nixos-chat
<eyJhb> Tailscale seems nice, I just dislike .. having a account etc.
<patagonicus> Hmm, no, reading more in the nixpkgs, while useOzone defaults to false in common.nix, for chromium it seems to be changed to true in default.nix. Let me check google-chrome, since I didn't use chromium.
<patagonicus> That has a wayland dependency that's not optional, as far as I can tell. So maybe it doesn't work for some other reason.
FRidh has joined #nixos-chat
<patagonicus> Oh, huh, it just seems to work in Chromium, but not in google-chrome. Weird.
<patagonicus> …but youtube playback doesn't work. :(
abathur has quit [Ping timeout: 256 seconds]
<eyJhb> IS it just me, or wasn't .xpi files suppose to work for Chrome and Firefox?
abathur has joined #nixos-chat
<ldlework> heh the little wrapper idea worked like a charm
FRidh has quit [Remote host closed the connection]
<jtojnar> ldlework: #pulseaudio devs are mostly on European time
<lunc> Definition of debugging - Being the detective in a crime movie where you are also the murderer
AMG has joined #nixos-chat
FRidh has joined #nixos-chat
thibm has joined #nixos-chat
<eyJhb> lovesegfault: Not really helpful in there I guess :(
<thibm> hey eyJhb, how the oneplus are doing? Sorry, I was afk last time
<eyJhb> They are all fine, the seller said that he thought that maybe I could get at least two working, so my hopes weren't that high. But I just needed the cover anyways.. BUT!
<eyJhb> The first three worked fine, the last one had a display that did not show anything. But I reseated the cables, and then it worked. I just need to change the battery in it now, then they all work :D
<lunc> i have a story
<eyJhb> But it is amazing having 1. A complete antenna 2. A power button that works. Mine broke after 6+ years, like a week ago
<lunc> there was a guy contributing to the webkit open source project from the UK
<lunc> we decided to hire him and bring him to the US to work on webkit full time
<lunc> he was just doing it in his own time for fun, mainly working on the JS engine
<lunc> not long after, google announced chrome, and they asked us if we'd like to replace the JS engine in webkit with a new engine that they had developed
<thibm> eyJhb: wow you get lucky, cool
<lunc> theirs was a bytecode interpreter that was much faster, maybe 30-40% faster, than the interpreter we had in webkit, which was just an abstract syntax tree interpreter
<lunc> but it would've required significant changes to webkit, because the JS engine has a strong relationship with the html/css engine through the DOM
<lunc> so our new engineer, who'd only worked for us for a week or two at this point, went home over the weekend, and he wrote a new JS bytecode interpreter for webkit
<lunc> over the course of 3 days
<lunc> and it beat google's new interpreter by maybe 10% our the js benchmarks people were using at the time
<lunc> on*
<lunc> and didn't require significant changes to the memory model for the DOM
<eyJhb> thibm: yeah! So I am not sure what I should do with the rest. I'm considering selling 1-2 to even out the cost. But the real issue as I discussed with samueldr yesterday is, that in the future we will mainly use arm64, which OPO is not...
<lunc> so we switched to it instead, and that started the JS wars of the late 2000s / early 2010s
<ashkitten> my ps5 controller is having lag issues over bluetooth :(
<lunc> chrome added a bytecode interpreter, then we did, the mozilla did. then chrome added a jit, then we did, then mozilla did, etc etc
<ashkitten> maybe i need a better bluetooth adapter
<thibm> eyJhb: yeah I was frustrated that it's not 64 bits too
<eyJhb> I feel like joepie91 could wage in on this lunc :p
<eyJhb> Where the hell do you work now lunc ?
<lunc> i currently dont work heh
<lunc> oh some of my sentence got deleted. i'm half asleep. "competition made all the difference. if they just used google's V8 then JS wouldn't be nearly as fast as it is today"
<eyJhb> thibm: Yeah, it really is. I hoped to rock this phone till it really dies, or becomes really obsolete.. Annoying that it is just 32bit/64bit, as it has everything you need. Like, NFC, BT, 5GHz WiFi, etc... I really do not need much else, except 5G mobile networking
<eyJhb> lunc: retired? :p (I am just spitballing here)
<adisbladis> eyJhb: My gen 1 google pixel is starting to show it's age... I really don't want to get a new phone..
<eyJhb> adisbladis: How is it showing?
<adisbladis> eyJhb: It's often dying with 50% battery left
<lunc> hehe kinda
<adisbladis> And I hate this phone, so I don't want to invest money into changing the battery either
<eyJhb> adisbladis: new battery ?
<eyJhb> Ohh..
<eyJhb> lunc: lottery? :p
<eyJhb> adisbladis: If you hate it, then a new phone might be in place. OnePlus Nord N10? :D
<lovesegfault> eyJhb: Yeah :(
<lovesegfault> I'm going to ask flokli later
<adisbladis> eyJhb: I'm gonna hate that too :P
<lovesegfault> he might know
<flokli> all this asking flokli
<lovesegfault> :D
<lovesegfault> You have a minute?
<flokli> I can't recommend new phones, if that's the question
<lovesegfault> it's much worse than that
<flokli> uh oh.
<adisbladis> flokli: No one can
<flokli> yes :-)
<adisbladis> Because they all suck
<flokli> I can't carry around my workstation to pick up a call unfortunately.
<lovesegfault> flokli: Two machines, A and B. A is a laptop with wifi, B is a NAS with 1.5Gbps eth. I have A[wifi] <-> router <-> B[2.5G] and it's _SLOW_
<lovesegfault> if I use a VPN and jump through a host outside my LAN it's fast (as fast as my internet)
<lovesegfault> so, doing A <-> R <-> Internet <-> R <-> B is faster than the local only thing
<lovesegfault> running iperf3 shows loads of retransmissions, which we suspect might be the issue
<flokli> wait. it's A <---wifi----> R <----2.5Gethernet----> B
<lovesegfault> but we don't know what's causing it
<lovesegfault> Yeah
AMG has left #nixos-chat [#nixos-chat]
<flokli> wait, is it that new aarch64 nas?
<lovesegfault> Nope
<lovesegfault> All amd64
<flokli> hmmh
<lovesegfault> iperf3 from internet host
<lovesegfault> that internet host is the same as the local host, just going through tailscale
<lovesegfault> in other words: local machines are slow talking to each other, but if I use a VPN it's fast again
<lovesegfault> eyJhb and ashkitten have been scratching their heads too :P
<flokli> what are your mtus on these links set to?
<lovesegfault> 1500
<lovesegfault> well, the tailscale0 iface has a 1280 mtu
<ashkitten> lovesegfault: you *did* reboot the router, right?
<lovesegfault> ashkitten: oh yeah, like 5 times
<flokli> lovesegfault: is there some firewalling on the router?
<lovesegfault> Not AFAIK, it's a normal home router though
* lovesegfault goes check
<lovesegfault> I have NAT Filtering set to "secured"
<lovesegfault> that's it
* flokli tilts head
<lovesegfault> under security that's all I have :P
<flokli> hmmh.
<flokli> try with another router you can better control ;-)
<lovesegfault> :P
<flokli> or take a nap and start fresh.
<lovesegfault> I might try setting the router to AP mode
<lovesegfault> then using a switch I have to do the actual switching
<lovesegfault> and my ISP's thing as the gateway/DHCP server
<thibm> We don't have a compose operator in Nix, right? I miss it all the time :(
<lovesegfault> has anyone played with building a Nix C package with LTO?
<lovesegfault> I thought I just needed NIX_CFLAGS_COMPILE and NIX_CFLAGS_LINK to contain `-flto=12`
<lovesegfault> oh, I guess I need gold
<patagonicus> thibm: compose operator?
<thibm> function composition operator
<patagonicus> lovesegfault: "Buy your Nix Gold subscription now! No more ads while evaluating derivations and unlimited cache downloads!"
<patagonicus> Ah, function composition
<thibm> yup, c = f: g: x: f (g x)
<eyJhb> lovesegfault: you should just have a nix router
<eyJhb> :D
<thibm> aoc day 4 was super boring
<etu> thibm: agreed
<etu> thibm: Day 3 was much more fun
<thibm> it was more fun, I wouldn't say "much". I don't find them really challenging for now
<etu> thibm: Still doing it in nix?
<thibm> yes
<thibm> It was definitively the worse tool for the job today
<thibm> as it was 99% parsing
<joepie91> patagonicus: cursed
<joepie91> very, very cursed
<patagonicus> That was my intention. :D
<supersandro2000> and stupid tensorflow again almost crashed my machine
<supersandro2000> I want a blacklist for nix which packages it can build
<supersandro2000> so that it never ever tries to build pytorch and tensorflow on my machine
<supersandro2000> they either crash or are clogged up for 12h+
<thibm> supersandro2000: overriding their meta in an overlay?
<thibm> or just overriding to {} or something that does not evaluate in an overlay
<adisbladis> thibm: I guess you want it evaluated, just not built locally
<adisbladis> The closest thing I can think of to achieve this is requiredSystemFeatures
<supersandro2000> thibm: to easy...
<supersandro2000> I can just do an overlay which adds big-build to them
<supersandro2000> or something which I don't have
<supersandro2000> thibm++
<{^_^}> thibm's karma got increased to 9
<thibm> adisbladis: hum if you want to evaluate them but prevent building them it's a bit more trickier indeed
<adisbladis> supersandro2000: The annoying thing is that it will change the hash
<adisbladis> As requiredSystemFeatures is not a meta thing, but a thing that actually ends up in the drv
<adisbladis> So overriding it will result in always getting cache misses
<adisbladis> And I believe big-parallel is default on all nix systems?
<thibm> You could add `meta.timeout = 1`
<thibm> It will be evaluated but will not "really" be built
<thibm> (but the dependencies will be)
<thibm> Well, this meta attribute seems to be completly ignored
<adisbladis> Yeah of course it's not working :P
<thibm> ?
<adisbladis> Once a drv is passed to the builder anything not explicitly in the drv file is ignored
<thibm> sure
<thibm> why is the timeout supposed to work then?
<adisbladis> Idk if this ever worked or how it was supposed to work
<adisbladis> Maybe in some distant past meta was a part of the derivation?
<thibm> (how*)
<adisbladis> thibm: I went spelunking and found out it's something only supported by Hydra
<adisbladis> Hydra is using it's own evaluator that can take these kind of things into account
<thibm> Yeah that was what I guessed too
<thibm> commit 02e1f00ffdd0 is strange
<thibm> because it's missing its counterpart in hydra I guess
<thibm> so we miss the timeout logic
<adisbladis> Those docs needs rewording
<thibm> totally
supersandro2000 has quit [Quit: The Lounge - https://thelounge.chat]
supersandro2000 has joined #nixos-chat
<supersandro2000> adisbladis: ....
<supersandro2000> come on. Why can't things just work!?
<adisbladis> supersandro2000: We can't have that
<etu> :D
<supersandro2000> Apple: It just works
<eyJhb> supersandro2000: Ohh, you be carefull now
<eyJhb> The company who puts its hot air output against the screen so that the glue comes loose
<eyJhb> :p
<ar> >Explosive hydroforming a steel sphere. https://twitter.com/MachinePix/status/1334559037157994496
<supersandro2000> eyJhb: I was joking
<eyJhb> supersandro2000: Well it is hard to know! A lot of people feel that way :(
<eyJhb> ar: that is fucking cool
<supersandro2000> eyJhb: A lot of people also think that repeating the same steps 5 times in a deterministic system yields any different results
<supersandro2000> or that I can help anyone when they tell that it just does not work
<supersandro2000> or that they can play the system
<eyJhb> supersandro2000: Yeah, but precisely apple is a topic in the kind like emacs vs. vim for most :p
<supersandro2000> emacs does not charge you 25 bucks for the full nix overlay
<supersandro2000> I recently tried to find 2 very simple mac apps which are basically xml editors and they where both paid
<supersandro2000> I could do the same work in Vim and that not slower
<adisbladis> eyJhb: I was forced to use a macbook for work many moons ago
<adisbladis> It had discolouration of the screen near the exhaust :/
<eyJhb> supersandro2000: Well some like being charged for that! https://www.stickycomics.com/wp-content/uploads/update_for_your_computer.jpg
<eyJhb> adisbladis: Damn... Why forced? Did Tweag do this to you?? You can tell if they did
<adisbladis> eyJhb: No, many employers ago :P
<adisbladis> The dev environment was only officially supported on windows or macos
<eyJhb> Well, I heard many and thought, that is at least two
<adisbladis> (it was a modified jetbrains)
<adisbladis> I managed to make a franken-IDE by splicing some components from jetbrains on linux
<adisbladis> So while it was not supported, I could dev on linux
<eyJhb> That must have been nice
<eyJhb> That you at least could. With office365 and shit and stuff, it is impossible...
<adisbladis> It's called that because you turn 365 degrees and walk away
<eyJhb> Like just sharepoint sites might not work, because the policy on the PC is not "correct"
<eyJhb> But... LEGO adisbladis :(
<gchristensen> adisbladis: I'm playing with a Go project and I think I must be missing something, because the author of the library couldn't be this dense. can you help point out the silly thing I'm missing?
<joepie91> eyJhb: that comic, lol
<adisbladis> gchristensen: In ~45 mins OK?
<eyJhb> gchristensen: What what! Go! Can I see?
<gchristensen> for sure (I might be getting a croissant then, but if so, whenever after that would be great)
<eyJhb> joepie91: But it is true
<joepie91> yes :D
<srk> 57
<srk> oop
<supersandro2000> eyJhb: I am being charged for using Linux tools
<supersandro2000> If I find a bug I fix it and upstream it
<supersandro2000> I pay more than enough
waleee-cl has joined #nixos-chat
<adisbladis> gchristensen: I got stuck, but I'm available now
<gchristensen> adisbladis: given this "example" https://github.com/hanwen/go-fuse/blob/master/example/loopback/main.go can I implement my own "loopback" fuse filesystem behavior without duplicating a significant portion of the go-fuse/fs package? I tried copying all the stuff out of fs/fuse.go and then updating references, but ultimately got stuck on loopbackFile being un-exported.
<gchristensen> https://gsc.io/content-addressed/8cdae7882f047dbaa3c538fcbb570a439c0d797bffb2017e926c6b7f82dc7306.go I'm just heading out for that croissant, but I'll be back in a bit to talk about it
rnea has joined #nixos-chat
<adisbladis> gchristensen: I guess you want to cast something to loopbackFile ?
* adisbladis feels like he's missing some context
<bbigras> croissants are great
<joepie91> eyJhb: you work at LEGO, yes? can I file a website bug through you? :P
<joepie91> (from someone I know)
<eyJhb> joepie91: I don't know many UX guys :(
<eyJhb> What kind of bug is it?
<eyJhb> And yeeees, I have worked for them for... a year?
<eyJhb> I have no clue how this is possible, but it is amazing - https://termbin.com/xhmg0
<eyJhb> ''$'\n'*
<eyJhb> as a filename
__monty__ has joined #nixos-chat
<joepie91> eyJhb: it's a UX bug that's probably costing them money, so :P
<joepie91> will PM
<bbigras> lovesegfault: https://github.com/bbigras/nix-config/pull/49 \o/ yay. (it didn't work on the first time since I needed to config git (set name and email)
<{^_^}> bbigras/nix-config#49 (by github-actions[bot], 1 minute ago, open): [automation] update flakes dependencies
<bbigras> I'm going to trigger another build to see if the bot avatar is the same with another email address
<eyJhb> ffs, I checked my email yesterday. I have 33 new ones, and none for me personally
<infinisil> eyJhb: I have a mail sieve that puts all non-personal mails into a separate folder :)
<infinisil> Very useful
<infinisil> Well I use contact@<domain> for personal emails, and <serviceName>@<domain> for services
<infinisil> And the sieve puts those into different folders
<eyJhb> infinisil: I should do that...
<eyJhb> And there is the random phishing internal testing email as usual
<eyJhb> And don't you dare be curious and click on it, they will murder you
<thibm> At my job the phishing was obvious. I opened the link because I was curious and then entered fake credentials. It said I would have been screwed anyway :p
<bbigras> haha
cole-h has joined #nixos-chat
red[evilred] has joined #nixos-chat
<red[evilred]> I've done the phishing testing for an org before
<red[evilred]> got them from > 50% click rate to zero in 6 months
<red[evilred]> Yes - that sounds impossible - but I had "help"
<red[evilred]> because teaching the end user is only a very small part of it, it's the thing you have no control over as new people are hired etc
<red[evilred]> and when people test they focus on that
<red[evilred]> which is bad
* joepie91 is very confused
<red[evilred]> The trick is to accept taht someone will click - and instead train your users to do rapid notification
<red[evilred]> then when you have people doing instant notification on receipt - you can block any IoCs for the badness and scrub all the otehrs from everyone else's mailboxes
<red[evilred]> once you've done that - you can get to the point where you can scrub the phishing from your peeps mailboxes before the first 'inattentive' user has the opportunity to click
<supersandro2000> eyJhb: that is bash for eval to new line
<eyJhb> supersandro2000: it is still evil filenames :D
<JJJollyjim> phishing testing is one of my concerns about getting into infosec professionally tbh
<red[evilred]> Why does it make you concerned?
<JJJollyjim> because it seems inevitable that someone's gonna get fired over clicking something anyone would have clicked on on a bad day
<supersandro2000> joepie91: reminds me of the super conservative anti gay marriage EU Parlament guy which was caught on 25 person gay gang bang party
<joepie91> supersandro2000: well *that* one wasn't a surprise at all
<joepie91> "repressed homosexual" is a meme for a reason :P
<joepie91> there's a long, long list of incidents like this
<red[evilred]> Ah - I refuse to work for a company that will do that\
<red[evilred]> it's a question I ask when I interview for a job
<red[evilred]> I refuse to work in a "blame" environment
<red[evilred]> because if someone clicks - you want them to contact you and tell you everything as quickly as possible
<red[evilred]> if they have fear of being fired - that makes your life in DFIR an order of magnitude harder
<red[evilred]> THAT
<JJJollyjim> oh yeah, but i'm planning to join a consultancy, and idk if they can really stop shitty clients doing that with the results
<red[evilred]> plus - if they have a user that's capable of ransoming all your data with an errant click - I would suggest that they have put zero thought into proper security boundaries.
<JJJollyjim> yep
<red[evilred]> and are vulnerable to insider attacks
<red[evilred]> and yes - that's another reason I don't work for a consultancy
viric has joined #nixos-chat
<viric> Hello nixosers
<bbigras> o/
<viric> Today I dove into the world of CI software...
<viric> I'd say all is an enormous pile of shit
<viric> emphasis on enormous.
<red[evilred]> that's because everyone insists on reinventing make
<red[evilred]> (badly)
<red[evilred]> :-)
<viric> Isn't there anything simple that has a coordinating server, and agents playing the checkout, build, run tests, store results?
* joepie91 looks at challen name
<joepie91> channel*
<joepie91> viric: Hydra? :P
<viric> Hydra is very tuned to nix
<JJJollyjim> yeah i was about to say lol joepie
<bbigras> is jenkins "simple"?
<JJJollyjim> nix is definitely just make
<red[evilred]> "no"
<viric> jenkins sounds terrible
<viric> I'd go with some scripts I think. Even without web interface, if results are just fine in a file tree.
<thibm> viric in 1 year: https://xkcd.com/927/ :)
<bbigras> is drone.io open source?
<viric> :)
<viric> well, I expected someone here will say "yes, you have not yet found this great CI"
<viric> or "bunch of scripts that do CI"
<thibm> nice try :D
kalbasit has joined #nixos-chat
<viric> I tried to look at how some people that do things easy deal with that (sqlite, fossil) and... they seem not to use any CI.
<hexa-> wow, the firefox fckp on master just bit me :(
<cole-h> fckp?
<hexa-> cole-h: all my addons and their data is gone
<hexa-> #105796
<{^_^}> https://github.com/NixOS/nixpkgs/pull/105796 (by Luis-Hebendanz, 1 day ago, open): Fix firefox wrapper
<hexa-> guess I'll restore last nights firefox profile
<hexa-> hah, last backup is 3h old
thibm has quit [Ping timeout: 256 seconds]
thibm has joined #nixos-chat
<gchristensen> adisbladis: I want to take the loopback example and change its behavior
<Taneb> Can anyone recommend a VPS provider that supports NixOS?
<eyJhb> hexa-: Guessing a lot of profiles will die because of that...
<gchristensen> adisbladis: but their "example" isn't just an example I can copy-paste and toy with (which is the traditional thing for examples to be I thought)
<hexa-> eyJhb: this will be a lection in having proper backups
<gchristensen> so I'm left wondering if maybe I just don't understand Go enough to know that I can in fact take the loopback example and mess with it.... or they spoiled the example by Engineering It To Death
<bbigras> Taneb: I'm using digital ocean for a cheap vps. They don't support nixos, but nixos supports DO.
<eyJhb> hexa-: welllll! It could have been avoided :p
<eyJhb> But yeah
<eyJhb> Luckily I don't use FF! Or I do, but I do not have it persistent... So..
lunc has quit [Ping timeout: 260 seconds]
<siraben> colemickens: even with `hardware.cpu.intel.updateMicrocode = true;` will turning off spectre mitigations still take effect?
lunc has joined #nixos-chat
kalbasit_ has joined #nixos-chat
kalbasit has quit [Ping timeout: 256 seconds]
<cole-h> hexa-: Oh yeowch. Guess it's a good thing I haven't bumped unstable recently...
<siraben> good reminder to take regular backups
<hexa-> or snapshots
<cole-h> Well, I do have snashots every 5 minutes so I wouldn't be hurt too badly...
<siraben> ZFS?
<cole-h> Yep
<siraben> Heh nice, so you don't need to use other problems to back up?
<siraben> I'm considering switching to ZFS after finals are over
<cole-h> I don't, no.
<joepie91> "other problems" :)
<bbigras> isn't snapshots not backups?
<siraben> Oops, Freudian slip
<joepie91> hehehe
<siraben> s/problems/programs
<bbigras> I would use snapshots and a backup
<cole-h> bbigras: The way I use them, they're the same.
<siraben> snapshots are filesystem level, right?
<cole-h> I send my snapshots to an external disk every once in a while
<siraben> Incremental backups, encryption, deduplication and all?
<cole-h> and since the snapshots are incremental, I get a full backup from that point in time, basically
<bbigras> I guess some people might not trust them. but I mean with restic I still have to trust the deduplication
<cole-h> No dedupe, not enough RAM for that. Encrypted and incremental, though.
<bbigras> or maybe the saying is that "don't just use snapshot on your computer". while copying them to an external disk is fine
<cole-h> Snapshots won't help if my SSD fails, indeed.
<bbigras> I should set automatic snapshots with btrfs. I tried using snapper once but it didn't clear my old snapshots automatically.
<nicolas[m]> IIRC snapper supports timers
<bbigras> yeah snapper did take the snapshots. and it has a setting to clear old ones. but I think I wasn't saving them to the right place or something.
<nicolas[m]> Is there a user contributed NixOS module that does automatic snapshot when doing a rebuild?
<bbigras> I bumped nixos-unstable. 14.6GB to download. 🤔
<bbigras> no. 4GB to download. 14.6GB might be to install.
<bbigras> or copy
<red[evilred]> Wow - an AWS engineer just tried to call me out on a sales call. Trying to tease me about something.
<red[evilred]> It did not go well for him.
<bbigras> cole-h: do you snapshot your nix store, everything or only home?
<cole-h> I snapshot rpool/user and rpool/system (and children), but not rpool/local (basically only /nix)
<bbigras> thanks
<bbigras> oh yeah in my case I should only snapshot the stuff I persist with "impermanence"
cole-h has quit [Ping timeout: 240 seconds]
endformationage has joined #nixos-chat
rajivr has quit [Quit: Connection closed for inactivity]
thibm has quit [Quit: WeeChat 2.6]
KREYREEN has quit [Remote host closed the connection]
KREYREEN has joined #nixos-chat
tilpner has quit [Remote host closed the connection]
tilpner has joined #nixos-chat
<lovesegfault> bbigras: nice!
<bbigras> yep
Guest67 has joined #nixos-chat
<bbigras> there was this rant about flakes being not ready https://twitter.com/tpflug/status/1334829899996160001
<bbigras> I'm going to ignore it for now
<bbigras> but maybe you would want to know
<lovesegfault> I mean, he's right they aren't ready
<lovesegfault> and the RFC thing was a joke
<lovesegfault> but it's what we have
<siraben> Original size Compressed size Deduplicated size
<siraben> All archives: 3.43 TB 2.97 TB 503.93 GB
<siraben> This archive: 27.12 GB 17.54 GB 3.90 GB
<siraben> according to borgbackup, deduping saves a lot
<__monty__> Flakes are still an experimental feature, of course they're not ready yet.
<__monty__> I don't like the privileged treatment they're getting but that's an entirely different matter altogether.
<bbigras> It seems to work well enough for me. I must say I didn't try many things yet.
Guest67 has quit [Quit: Connection closed]
<lovesegfault> the `nix` tool is what annoys me the most
<lovesegfault> they've been experimental since forever
<__monty__> If it's not going fast enough you can throw time or money at it. Otherwise, it's what we have.
<nicolas[m]> On the topic of new nix commands, how do you get the store path of running `echo "$(nix build 'nixpkgs#something' --no-link)` ?
<__monty__> That sounds like a question for #nixos.
<nicolas[m]> People are often quicker at answering questions here
<__monty__> Doesn't mean you get to abuse the offtopic channel ; p
<bbigras> people using btrfs with compression. https://github.com/NixOS/nix/issues/3550 should be in nixos-unstable now.
<{^_^}> nix#3550 (by vcunat, 31 weeks ago, closed): btrfs compression ignored
<bbigras> well the fix
<nicolas[m]> tpilner: thanks, JQ it is
<tilpner> Also look at the discussion and linked issues. It's intentionally not supported
<tilpner> "That looks good overall, except that it brings back the intentionally removed behavior of printing the built store paths thinking"
KREYREEN has quit [Remote host closed the connection]
KREYREEN has joined #nixos-chat
FRidh has quit [Read error: Connection reset by peer]
lunc has quit [Ping timeout: 265 seconds]
omnd has joined #nixos-chat
dada_da has quit [Ping timeout: 240 seconds]
dada_da has joined #nixos-chat
ajs124 has quit [Quit: killed]
das_j has quit [Quit: killed]
das_j has joined #nixos-chat
ajs124 has joined #nixos-chat
<colemickens> Ugh. I didn't mean to send that?
<colemickens> Not even sure it sent, gosh darn the Matrix clients still need so much polish
<bbigras> Send what?
<colemickens> Element Android is... struggling. that's all.
lukegb has quit [Quit: ~~lukegb out~~]
lukegb has joined #nixos-chat
<bbigras> oh yeah for me too. when I send something, it's sent but the message stays in the text field.
<bbigras> and it's taking a while to refresh
<joepie91> Element Android is buggy as hell
<joepie91> colemickens: can recommend looking into Fluffychat instead
<energizer> is it possible to persist an ssh connection across suspend/resume?
<pie_> vpns might help
<pie_> also setting the server and clients to not time out
<joepie91> energizer: not in a literal sense I don't think, but tmux/screen usage would get you an approximation of that - might be possible to hack something together with mosh also
<pie_> or look at mosh
<srk> or ethernal-terminal
<pie_> whats the use case, sshfs or something?
<joepie91> oh for sshfs you could also just use autossh
<srk> mosh can't scroll :D
<pie_> srk: tmux in mosh?
<srk> but why
<energizer> oh i guess this is an XY problem
<joepie91> heh, yeah
<pie_> "Mosh is a popular alternative to ET. While mosh provides the same core functionality as ET, it does not support native scrolling nor tmux control mode (tmux -CC). " aha. i have no idea what this means but ok
<srk> pie_: I'm xmonad user so I don't do tmux :)
<pie_> srk: hmm
<pie_> i guss that makes more sense
<srk> pretty happy with ET tho, just needs systemd.services.eternal-terminal.serviceConfig.OOMScoreAdjust = "-999";
<srk> :)
<joepie91> I don't understand WINE's versioning scheme
<samueldr> make number bigger when they feel like it
<samueldr> or plain old versioning scheme AFAIUI
<joepie91> semver doesn't make a lot of sense for software that breaks stuff all the time, and sentimental versioning doesn't make sense for this list of changes either
<joepie91> as none of them are particularly notable changes?
<samueldr> >> beginning of the yearly code freeze period
<joepie91> ... that having been said, Foundation support has been solved?
<samueldr> yearly as in duration of a year, or yearly as it happens once every year?
<samueldr> I guess it could be that they switched to timed releases
<__monty__> Yearlong /= yearly in my book.
<samueldr> right
* samueldr still doesn't feel totally awake
<joepie91> hmmm
<joepie91> this does suggest that they just bump the major version every year
<samueldr> there's the explanation I guess
<joepie91> odd versioning scheme
<samueldr> not uncommon
<samueldr> ruby does it
<samueldr> every december 25th they release the next ruby
<joepie91> I don't really see it making sense. it doesn't communicate "we think this is an important set of changes", it also doesn't communicate "breaking changes", and it also doesn't communicate "this is how old your version is" because it doesn't use the year in the release number (like NixOS does)
<samueldr> chrome does it, in a way, with a six week cycle
<samueldr> maybe it's fine if the version number doesn't "make sense" here
<__monty__> It may be as simple as "Users want to see numbers go up."
<samueldr> as long as it's steadily increasing
* joepie91 feels that's just bad communication
<__monty__> I've had people complain about "no new versions" even though they didn't experience any bugs or had feature requests.
<joepie91> sure but that doesn't change by having a different version scheme
<joepie91> if there's no releases, there's no releases
<samueldr> 1.X.0 is not worse communication than X.0 in my personal opinion, unless they have specified they follow a specific versioning scheme
<joepie91> (incidentally I think that that particular issue is illustrative of a really serious cultural problem in software, but.)
<energizer> i dont see a problem with semver even when stuff breaks often
<joepie91> samueldr: it's bad communication because it doesn't communicate anything
<samueldr> look
<joepie91> it's leaving a communication channel unused, making something look meaningful while it isn't
<samueldr> I didn't say "not bad"
<samueldr> it's not worse
<joepie91> I would be making the same complaint if it were 1.X.0
<samueldr> wherever that steadily increasing number is doesn't matter if there is no meaning
<pie_> its extra noise
<pie_> i think is what he is saying
<samueldr> then what should be done?
<samueldr> stick to 0.0.1-pre?
<energizer> joepie91: are you sure it's not semver?
<samueldr> they don't seem to have a development process that allows them to attach any more value to the version number
<joepie91> energizer: sure? no. but afaik it isn't
<samueldr> it's whatever was ready at the moment they decided to release
<joepie91> samueldr: they should use a versioning scheme that communicates *something* relevant :P
<samueldr> my question stands
<samueldr> how?
<samueldr> and they are
<joepie91> even if that means a single-segment version number that only ever increments to signal that there is nothing to distinguish releases
<energizer> it does communicate something relevant, 60 > 59
<pie_> a date does soud more convenirent
<samueldr> it's one version more than the previous
<joepie91> yes, that is not 'relevant'
<samueldr> it depends
<energizer> joepie91: yes it is
<pie_> its not?
<joepie91> with their current development schedule a year number would make sense
<samueldr> is 7.x forked from 6.0 and 6.1+ bug fix releases?
<samueldr> or is 7.x somehow forked from 6.whatever close to next release?
<pie_> so its wine 20 instead of 60 :P next century rthey have field lengh problems :ÍD
<joepie91> honestly this is a discussion that is spiraling into a far more time-and-energy-consuming endeavour than I planned on with my original comment :)
<pie_> yeah this is a classic bikesheddig discussion
<joepie91> ah yes, here's the obvious explanation of their versioning scheme: https://wine.lovetoknow.com/wiki/Wine_Release_Dates
<joepie91> :P
neeasade has joined #nixos-chat
<samueldr> I guess my opinion is it's important only if the project made it important at some point
<samueldr> and only because reality shows us we can't assume the other way around
<eyJhb> ldlework: did you change Github handle?
<ldlework> still dustinlacewell
<eyJhb> I give up, I need a translation table
<eyJhb> Just saw the bump in discord PR as soon as you mentioned it, so I thought it was you
Dotz0cat has quit [Ping timeout: 272 seconds]
<__monty__> The actual problem for today is a simple one-liner.
<__monty__> woops
<gchristensen> anyone interested in a 30 minute project? https://github.com/grahamc/packet-nix-builder/issues/7 a shell.nix, about 100 lines of terraform config
<{^_^}> grahamc/packet-nix-builder#7 (by grahamc, 1 day ago, open): Manage spot market bids with Terraform
<gchristensen> (ltots of which will be copy-paste)
<neeasade> gchristensen: > do you know what the "open source promise" is?
<neeasade> when I read this sentence this starts in my brain:
<gchristensen> oh dear
<eyJhb> Wonder how many got their addons nuked
<samueldr> hm?
tilpner_ has joined #nixos-chat
tilpner_ has quit [Client Quit]
tilpner_ has joined #nixos-chat
<eyJhb> samueldr: Firefox addons
<{^_^}> #105885 (by cmm, 8 hours ago, closed): firefox removed all my add-ons
tilpner has quit [Ping timeout: 260 seconds]
tilpner_ is now known as tilpner
<samueldr> I was lacking some contxt
<samueldr> context*
<eyJhb> Yeah, sorry!
<samueldr> I wondered if firefox had a booboo again
<lovesegfault> supersandro2000: around?
MichaelRaskin has joined #nixos-chat
spudly1 has quit [Ping timeout: 240 seconds]
spudly1 has joined #nixos-chat
Dotz0cat has joined #nixos-chat
<lovesegfault> Hmm, anyone know why gtk3 apps under sway look like windows 95 for me?
<samueldr> because wayland strive for perfection
<lovesegfault> :P
<hexa-> lovesegfault: can't confirm
<lovesegfault> hexa-: try running super-slicer
<lovesegfault> see how it looks
<hexa-> lovesegfault: compare it to dino
* lovesegfault nix-shell -p dino
<lovesegfault> dino looks good
<lovesegfault> lollypop looks good too
Dotz0cat has quit [Max SendQ exceeded]
<lovesegfault> but super-slicer is windows 95
<adisbladis> lovesegfault: Totally not helpful here, but:
<adisbladis> Becaue windows 95 ui was perfection?
<adisbladis> Because*
<hexa-> lovesegfault: maybe superslicers fault?
<lovesegfault> If I run it under gnome-shell it looks like Windows 7 though
<hexa-> hm, ask upstream
kalbasit_ has quit [Ping timeout: 240 seconds]
<samueldr> adisbladis: quite derivative shitposting :)
Dotz0cat has joined #nixos-chat
mtr has quit [Quit: WeeChat 2.9]
<lovesegfault> hexa-: but does it look windows 95-y to you too?
<hexa-> i know what you mean
<hexa-> but does prusa slicer look better?
<lovesegfault> nope