<MichaelRaskin>
In principle, a CoW file system without extreme space pressure could write new content at a leisurely rate, receive a confirmation from the drive that the content is safely written (or does non-lying hardware still exist?), then in the next TXG make new content the actual content of the file.
<MichaelRaskin>
So it is not an unimaginable behaviour, especially if documentation promises it.
parsley936 has quit [Remote host closed the connection]
<elvishjerricco>
Well I can't reproduce it now... Guess I was imagining things
<elvishjerricco>
However
<elvishjerricco>
It does seem that it'll return the write call having only written some of the data. That write call is atomic, but the user has to continue writing more
<MichaelRaskin>
Race conditions on power loss are indeed hard to reproduce reliably
<elvishjerricco>
Because write is allowed to write only some and return an integer indicating how much was actually written
Jackneill has quit [Ping timeout: 264 seconds]
<elvishjerricco>
So maybe I misinterpreted that with a worse test previously
<elvishjerricco>
I thought I'd read that the zil was updated record by record somewhere... Oh well
<elvishjerricco>
Gotta go read about the zil again now :P
<gchristensen>
elvishjerricco: incoming code! :)
<cole-h>
Man, it's so fun to be here. I learn so much stuff just by existing around all you smart peeps. Thank you all! clever++ gchristensen++ elvishjerricco++ MichaelRaskin++
<{^_^}>
MichaelRaskin's karma got increased to 38.99999999999999
<{^_^}>
gchristensen's karma got increased to 305
<{^_^}>
elvishjerricco's karma got increased to 15
<{^_^}>
clever's karma got increased to 445
<andi->
whats happening o.O(
<andi->
the matrix is broken?
<cole-h>
lol
<cole-h>
andi-: infinisil added some spice to the karma messages a while ago, if that's what you're talking about
<andi->
oh, that might be it
<gchristensen>
so zfs has atomic writes. will write(lots-of-data) write fewer bytes than I asked it to, returning a count of written bytes fewer than I sent? rsp: This is a good question. It depends on the recordsize and how much you are writing. https://github.com/openzfs/zfs/blob/master/module/os/linux/zfs/zfs_vnops.c#L721
<cole-h>
gchristensen: Do you have a link to that sweet ZFS article series you sent a while ago? It's bookmarked in my Firefox profile, but I haven't copied that over because tomorrow is The Day™.
<cole-h>
I'm in the mood to read
<elvishjerricco>
gchristensen: Yea I just came to the same conclusion. Discovered there's a whole other transaction concept called itx (I assume "intent transaction" or something)
Jackneill has joined #nixos-chat
<elvishjerricco>
And the ZIL is a linked list of these things. Kinda weird how this linked list gets updated; there seems to be a block that gets updated in place to point to the next itx or something like that.
<elvishjerricco>
I thought the linked list worked differently than that
<cole-h>
Thanks, sorry to make you dig it up again :P
<gchristensen>
if I tye "Zfs adm" it is the first thing mybrowser history suggests
<ashkitten>
> You will see further Cosmo perks being "locked" soon after this quality check. In fact, we hope that all Cosmo perks will be "locked" this week, meaning that all Cosmo perks in this campaign have been produced.
<{^_^}>
error: syntax error, unexpected ',', expecting ')', at (string):312:88
<ashkitten>
augh finally
<cole-h>
What's this?
<gchristensen>
the communicator?
<ashkitten>
yeah
<gchristensen>
nice
<ashkitten>
i'm so tired of waiting for it
<ashkitten>
this still doesn't mean it's even shipped yet though, once it's locked
<ashkitten>
it means the unit has been quality checked and is ready to ship
<gchristensen>
just that they won't take any more money?
<gchristensen>
ah
<gchristensen>
okay, step one for writing the talk for Thursday: install the program to write the talk
<pie_>
so it begins
<ashkitten>
i'm assuming there will be shipping delays as the units from the last batch haven't reached their warehouse in HK yet
<ashkitten>
and they ship from the warehouse
<MichaelRaskin>
ashkitten: don't say finally until there is a tracking number
<ashkitten>
plus i'm sure there will be delays out of HK
<ashkitten>
yeah
<cole-h>
gchristensen: lol :D
<MichaelRaskin>
(also waiting)
<ashkitten>
it's been what, 6 months?
<ashkitten>
more than that
<ashkitten>
oh, you ordered one too?
<MichaelRaskin>
Well, a bit above 6 months since they have definitely surely started the first production run
<ashkitten>
some people got theirs in november i think
<cole-h>
gchristensen: btw offer still stands for more proofing if you feel like it'd help
<MichaelRaskin>
Apparently that requires a perfect combo
<MichaelRaskin>
I only ordered in November
<ashkitten>
i ordered probably a couple months before that
<MichaelRaskin>
And I ordered with Cyrillic keyboard
<ashkitten>
dvorak for myself
<gchristensen>
okay wth I spent all this time not using spotifyd
<MichaelRaskin>
I do believe that _eventually_ I will get Cosmo. And to get something pocketable with Cyrillic hardware keyboard and an option to boot Linux, I am not sure there is anything but Planet…
<ashkitten>
isn't there a pinephone keyboard?
<adisbladis>
MichaelRaskin: I had NixOS booting on the Gemini
<MichaelRaskin>
ashkitten: I think «in progress»
<ashkitten>
heh
<ashkitten>
iirc the pinephone and cosmo keyboards are based on the same original design
<MichaelRaskin>
adisbladis: I am not even sure if I want to have NixOS and not Nix-on-Debian
<adisbladis>
MichaelRaskin: The debian setup had pretty horrible performance (read almost unusable)
<MichaelRaskin>
How did they manage, I wonder
<adisbladis>
At least on the Gemini
<MichaelRaskin>
Kali too?
<adisbladis>
I think it had to do with X
<adisbladis>
I had pretty decent performance on Sailfish, which is the only wayland compositor I've managed to get running yet
<ashkitten>
hopefully i can use my cosmo to help push mobile nixos development
<ashkitten>
the dual boot will be good for that
<MichaelRaskin>
multiboot limited by where they ran out of hardware key combinations
<ashkitten>
sorry?
<MichaelRaskin>
I think at some point they planned to support quad-boot and maybe one slot must be recovery
<ashkitten>
hmm
<adisbladis>
ashkitten: I guess the Cosmo is the same as the Gemini in that you need hybris for the gpu drivers?
<ashkitten>
probably
<ashkitten>
mobile-nixos is planned to use hybris for most devices
<MichaelRaskin>
Yeah, they would probably write about it otherwise
<adisbladis>
ashkitten: Yeah, but there is still a ways to go on that front
<MichaelRaskin>
I am surprised about their latest update: it is very out of character for them to write about a problem before it stops having an effect
<adisbladis>
But sure, you're gonna need hybris on most phone hardware regardless
<adisbladis>
Maybe not for the gpu but for other stuff
<ashkitten>
did they ever actually solve the issue with the backdoored ota updater?
<adisbladis>
I don't know
<ashkitten>
i know they wrote that the company removed their malicious code from it, but they're still using that company afaik
<ashkitten>
ugh i'm never ordering shit online from ikea again
<ashkitten>
it took them 6 weeks to tell me the chair had shipped and now on the day it was supposed to arrive they haven't even given the thing to fedex yet
<ashkitten>
is it time yet to call them and find out what the fuck is happening
<MichaelRaskin>
… and _try_ to find out …
<ashkitten>
i know it's all gone to shit because of covid but this is apparently just normal for ikea
<cole-h>
gchristensen: Do you know if skillsmatter will send out an email sometime before the event, so I don't forget? Or do I have to add it to my calendar? :P
<gchristensen>
not a clue!
<cole-h>
:D
<gchristensen>
this is their first one doing this
<cole-h>
Then I'll compromise and add a reminder ;)
<cole-h>
Oh yeah other question: Zoom? Jitsi? YouTube? Something else?
<gchristensen>
a great question!
<cole-h>
loool
<elvishjerricco>
Is there any serious reason to leave atime on?
<gchristensen>
I think systemd-tmpfiles uses it, but I turn it off
<elvishjerricco>
gchristensen: What would systemd-tmpfiles use it for?
<gchristensen>
not deletings thinsg recently accessed
<pie_>
time Linux had been turned into a real OS, and over the next few years it became a "success story" for open source. But that misunderstands the reality of the hundreds of people we had assigned to make it work and to turn Red Hat into a real company. After that, it could be argued that we stayed involved in open source in order to keep it from succeeding too much as part of quietly negotiated side agreements."
<pie_>
"Yes. That is exactly true. When Oracle started working on Linux it was only useful for print servers - and I created a white paper for Larry analyzing its shortcomings (lots of functionality missing from the kernel.) So we, along with IBM, Sun, HP, and others, committed tens of millions in resources to making Linux an actual competitor to Windows. And we kept improving it until Bill made enough concessions, and then we stopped. But by that
<pie_>
though " You've posted so many of these stories over the years that we had to ask you to stop unless you could somehow substantiate them:" "That request is still in effect. I don't have any reason to doubt you, but the stories are so grandiose and inflammatory (and similar) that we need to apply at least a little burden of proof."
<gchristensen>
lol
<gchristensen>
you can't really keep secrets that well
<drakonis1>
lol that post
<drakonis1>
goddamn
<drakonis1>
what a hot take
<drakonis1>
did oracle ever reasonably pump money into linux?
<gchristensen>
they had oracle linux
<drakonis1>
they still do
<drakonis1>
it looks like they have been historically freeriding on red hat's distro
<drakonis1>
the first release was based on rhel
<pie_>
im just scratching my head about if its true, how this guy doesnt get sued into oblivion
<pie_>
on the other hand
<pie_>
wouldnt want to streisand effect yourself
<cole-h>
Probably too busy suing other parties
<pie_>
heh
<drakonis1>
it sounds so incredibly ludicrous
<drakonis1>
and sun didnt invest money on linux
<drakonis1>
the funniest thing is bringing up sun
<drakonis1>
because they absolutely reviled linux and then got destroyed by it
<drakonis1>
red hat built itself with the open source model
<drakonis1>
from the very beginning
<drakonis1>
so this is pretty outlandish
waleee-cl has quit [Quit: Connection closed for inactivity]
<drakonis1>
and then lol
<drakonis1>
concessions lol
<drakonis1>
c'mon wtf is this
<pie_>
i vaguely remember something about sun and linux
* pie_
scratches had
<pie_>
head
<drakonis1>
sun's history is of being voraciously destroyed by linux
<drakonis1>
and a variety of poor decisions that sunk it
<pie_>
its a very conspiracy flavor post
<pie_>
peeking at the wiki page, huh. A 2006 report prepared for the EU by UNU-MERIT stated that Sun was the largest corporate contributor to open source movements in the world.[78] According to this report, Sun's open source contributions exceed the combined total of the next five largest commercial contributors.
<drakonis1>
sun never contributed to linux though
<drakonis1>
this looks like garden variety oracle guy in a bubble
<pie_>
"Following several years of difficult competition and loss of server market share to competitors' Linux-based systems, Sun began to include Linux as part of its strategy in 2002. Sun supported both Red Hat Enterprise Linux and SUSE Linux Enterprise Server on its x64 systems; companies such as Canonical Ltd., Wind River Systems and MontaVista also supported their versions of Linux on Sun's SPARC-based systems. "
<pie_>
"support" is a very vague term though
<drakonis1>
basically
<drakonis1>
they're only doing that to delay sinking
<pie_>
sounds like it
<drakonis1>
the sparc port was done by someone else
<drakonis1>
it led to people dumping sunos on sparc
<drakonis1>
this looks like performative shitposting
<gchristensen>
definitely
<pie_>
lol
<pie_>
ive go one too
<pie_>
"Knowledge can be gain through proven datapoints and unproven datapoints. Insisting on proof reflects a particular type of ignorance -- the ignorance of someone who hasn't deeply reflected on the history of the advancement of knowledge."
<cole-h>
On this piece of spinning rust, it takes me over 10 seconds just to cd into nixpkgs lol
<bqv>
Imported my emacs config into nix, finally
<bqv>
Its actually a smarter move than I realised, cause now my config is tested before running
<bqv>
And on a compartmentalized level too
<bqv>
Hey I can actually tick something off my todo list
<energizer>
bqv: what's it look like?
cole-h has quit [Quit: Goodbye]
<ashkitten>
okay what the heck is up with this
<ashkitten>
i finally got around to wanting to fix pulseaudio's annoying thing, flat-volumes
<ashkitten>
it's apparently set to no by default on nixos which is the weird thing, but it doesn't actually seem to work?
<ashkitten>
i don't see a difference in the behavior when i set it to yes vs no
<ashkitten>
maybe this isn't what i actually wanted, i'm not sure
<ashkitten>
it seemed like what i wanted?
<manveru>
what does it do?
<manveru>
or... what is it supposed to do? :)
<ashkitten>
just tested with gf's alpine machine, it seems broken on nixos
<ashkitten>
flat-volumes=no makes application volume relative to the "master" volume of the device
<ashkitten>
flat-volumes=yes makes it so the device volume is set to the highest volume of any stream
<ashkitten>
so application volumes are absolute under flat volumes
<ashkitten>
and relative without
<ashkitten>
(the default on nixos is flat-volumes=no)
<ashkitten>
it appears that it's simply broken on nixos currently though
<ashkitten>
is anyone using 20.03 and can test if it's broken there?
<ashkitten>
just increase the volume of an application and see if it affects the device volume
<ashkitten>
worldofpeace: you're the one that set flat-volumes=no as default originally so i guess i'll ping you?
<srk>
works for me with pulseaudio 13.0-rebootstrapped (20.09 tho)
<ashkitten>
i have that version of pulse
<ashkitten>
i'm tracking unstable
<ashkitten>
so wtf, it's just ignoring daemon.conf?
<srk>
strace or pulseaudio -vvvvvvvvvvv ?
<ar>
if you start pulseaudio with this many v's, does it start the vvvvvv game? ;)
<srk>
haha, it could since it's opensource now :D
<ashkitten>
oml
<ashkitten>
it was using ~/.pulse
<ashkitten>
you're kidding
<ashkitten>
welp
<ashkitten>
false alarm, i guess
<ashkitten>
i'm going to sleep so hard
<ashkitten>
(i only figured this out with `strace pulseaudio`)
<srk>
impurities :D
<ashkitten>
$HOME is an impurity and must be removed!
* ashkitten
is only sort of joking
<adisbladis>
I also found out pulseaudio prefers ~/ and if the file exists it completely disregards the global configuration the hard way :/
<adisbladis>
Such UX, much discoverable
<lovesegfault>
infinisil: you around?
<lovesegfault>
I'm fiddling with nixus and getting some ❓
<ashkitten>
adisbladis: the problem is ~/.config/pulse existed too, and i wasn't aware of the existence of ~/.pulse
<ashkitten>
so i didn't even consider it might exist until i saw it in strace
<srk>
.config/pulse looks like a legit state storage :D .. I: [pulseaudio] main.c: Using state directory /home/srk/.config/pulse
<ashkitten>
yes, it does doesn't it
<srk>
so if you have daemon.conf there it'll pick it up?
<ashkitten>
and yet it was loading config from ~/.pulse
<ashkitten>
apparently ~/.pulse overrides ~/.config/pulse, or somethinc
<srk>
right, maybe .config/pulse should be in /run/user?
<ashkitten>
or maybe it doesn't load config from ~/.config/pulse at all
<ashkitten>
the manpages only reference ~/.pulse for daemon.conf
<srk>
"PulseAudio will first look for configuration files in the home directory ~/.config/pulse, and if they are not found, the system-wide configuration from /etc/pulse will be applied."
<srk>
hmm :D
<ashkitten>
check `man pulse-daemon.conf`
<ashkitten>
nope, never mind
<srk>
I see, looks like ~/.pulse is a thing of the past
<ashkitten>
so it just uses the now completely undocumented directory first, overriding the new, documented directories
<adisbladis>
Aint software great <3
<ashkitten>
i hate software
<ashkitten>
there's no reason to have that load order anyway
<ashkitten>
i understand having ~/.pulse for legacy reasons but it should be loaded after the modern documented directories
<srk>
or just bail out with an error
<ashkitten>
yeah, or that
<ashkitten>
or literally anything besides its current behavior
<ashkitten>
okay, remember what i said about sleep?
<srk>
try looking for repair guides, I found a few for mine, looks quite educational :D
<eyJhb>
The trouble is srk, the monitor will NEVER look the same again :p
<srk>
yeah, freesync ips is not bad
<srk>
this
<eyJhb>
But no DP
<srk>
or the menu will drive you nuts :D
<eyJhb>
Whichi somewhat annoys me, if I am going to have it another 10 years
danielrf[m] has joined #nixos-chat
puzzlewolf has joined #nixos-chat
<eyJhb>
But tbh... 230 EUR is not much
<sphalerite>
elvishjerricco: no, nothing of the sort
<sphalerite>
elvishjerricco: unmounting certainly, which I guess counts as deleting when you're on a tmpfs, but that's not the case here
<sphalerite>
colemickens: yeah, but this is more compact and requires less building and allows faster iteration (since you don't need to build a fresh tarball each time)
KeiraT has quit [Remote host closed the connection]
<sphalerite>
elvishjerricco: well, unless you load this from itself :D
<sphalerite>
elvishjerricco: in any case, I used it for… things… at some point. Can't remember exactly, it may have been repartitioning.
KeiraT has joined #nixos-chat
__monty__ has joined #nixos-chat
KeiraT has quit [Remote host closed the connection]
<sphalerite>
elvishjerricco: oh dear. Why though >_<
<elvishjerricco>
sphalerite: Because the initrd root tmpfs can't be unmounted
<elvishjerricco>
the kernel literally just doesn't allow it
<sphalerite>
oh really?
<sphalerite>
okay…
<elvishjerricco>
So you have to delete it all to save memory
<sphalerite>
that seems weird but I guess there's a reason
<MichaelRaskin>
initrd could be unmounted, initramfs no
<elvishjerricco>
Yea I'm not sure what the technical reason is
<MichaelRaskin>
You also cannot pivot_root from it
<MichaelRaskin>
So there is a special dance to mount new root over / and chroot to /
<elvishjerricco>
systemctl switch-root does at least *try* to use pivot_root, but yea you can't do that from initramfs
<MichaelRaskin>
(it has to be chroot to notice new /, and it has to be at / otherwise user namespaces won't work inside)
<elvishjerricco>
sphalerite: Oh so I guess since it tries pivot_root, your thing wouldn't delete everything :P
<sphalerite>
eyJhb: I bought a Dell P2720DC recently, it feels very futuristic what with being able to charge a laptop, connect USB devices and another display all via one cable (once I have a new laptop Soon™)
<elvishjerricco>
MichaelRaskin: I've been wondering... I thought there were cheats to break out of chroots. Why can't you do that to get back into initramfs's root tmpfs?
<elvishjerricco>
since chroot is how switch_root works
<MichaelRaskin>
Well, technically as you mount over /, there is no higher directory to move to
<eyJhb>
sphalerite: the linking of displays requires the computer supports it right?
<eyJhb>
But 27" :(
<eyJhb>
Looking at 24", as 27" gets too big for me
<elvishjerricco>
MichaelRaskin: Oh I see. That's super weird
monsieurp has quit [Remote host closed the connection]
<elvishjerricco>
sphalerite: I used to think of that as the dream but then I got a powerful desktop and my desire to use my laptop at a desk went to zero :P
<sphalerite>
eyJhb: only in the sense that it has an appropriate displayport version and a GPU with a sufficient framebuffer size
<sphalerite>
eyJhb: certainly no displaylink bullshit
<sphalerite>
eyJhb: and there is an equivalent 24" monitor, P2421DC
<sphalerite>
elvishjerricco: well, my work computer is a laptop and I'm currently working from home a lot for some reason… :)
<sphalerite>
elvishjerricco: re going back into initramfs: systemd has support for that, it can move the initramfs to /run/initramfs instead of deleting it. That's used for clean shutdown with rootfs on network or something, can't remember the exact scenario it's used for
<elvishjerricco>
sphalerite: Hm I'll have to figure out how that works.
<sphalerite>
elvishjerricco: oh lol I didn't realise that
<sphalerite>
elvishjerricco: I thought it uses the same one used for boot
<elvishjerricco>
Seems not
<eyJhb>
The difference is 40 EUR :p But I could do so yes
<elvishjerricco>
Which could actually be a problem. If you update, and e.g. your ZFS user space tools are updated, but your kernel module needs a reboot because you have / on ZFS, then the user space tools in that initramfs may not be compatible with the kernel they're trying to export the pool from
<sphalerite>
I'll probably get a P2720D if I start feeling a desperate need for a second external monitor
<sphalerite>
elvishjerricco: yeah, that's what I thought
<elvishjerricco>
Well that was an interesting little sidetrack :P
<eyJhb>
Well yeah, I can save 70 EUR total if I get two of those without USB-c
<eyJhb>
Ohh the deciscions
<eyJhb>
Lets check the bank sphalerite :p
<sphalerite>
eyJhb: if you'll ever be using them with a laptop, I'd say get one USB-C, but if they're only for a desktop without is cheaper therefore better
<eyJhb>
I have a laptop, and only use them with my laptop. But it is my good old x230 :p
<sphalerite>
and you won't be getting a USB-C laptop anytime soon? Then yeah also no-C
<eyJhb>
I am unsure if the resolution might be annoying to look at, with everything getting smaller. They will be combined with one 1920x1080 monitor
<eyJhb>
I am not sure, but these monitors should last me for another 10 years I hope
<eyJhb>
So in that timeframe, maybe :p
<sphalerite>
eyJhb: hm, does the x230 have DP 1.2?
<sphalerite>
and I'm guessing you have a dock? Or do you connect one monitor via HDMI and the other via DP?
<eyJhb>
Using everything from the dock, and unsure, let me see
<sphalerite>
ah right, if you have a dock there's also less reason to get a dock monitor :D
<eyJhb>
I have 3-4 docks, I think I have all the different ones they have made for the 30 series :p
<MichaelRaskin>
It would be more interesting if they could stack
<eyJhb>
The.. Docks?
<MichaelRaskin>
Of course!
<MichaelRaskin>
Then you could have even more USB ports!
<eyJhb>
sphalerite: Displayport 1.1 support :( No daisy chain
<eyJhb>
All the USB ports!
<sphalerite>
eyJhb: but if you have a dock that doesn't matter — you can use both monitors with the dock, then when you get a USB-C laptop switch to daisy-chaining
<{^_^}>
SEK = x : toString x + " SEK (weakest of the Norse Borthers) = " + toString (0.68*x) + " DKK"
<eyJhb>
> NOK = x : toString x + " NOK (second weakest of the Norse Borthers) = " + toString (0.69*x) + " DKK"
<{^_^}>
NOK defined
<eyJhb>
> SEK = x : toString x + " SEK (weakest of the Norse Borthers) = " + toString (0.71*x) + " DKK"
<{^_^}>
SEK defined
<eyJhb>
Perfect, updated to reflect current value :p
<etu>
:D
<adisbladis>
DKK = throw "rødgrød med fløde"
<eyJhb>
NO NO :O
<eyJhb>
DANISH = throw "potato"
<eyJhb>
More like it
<adisbladis>
Anyone can pronounce potato
<eyJhb>
But not many can speak it!
<adisbladis>
DKK = throw "Kamelåså!"
<adisbladis>
eyJhb: There we go
<joepie91>
eyJhb: Borthers?
<joepie91>
:P
<eyJhb>
Noooo
drakonis_ has joined #nixos-chat
<eyJhb>
Damn it joepie91 , do not spellcheck me
<adisbladis>
eyJhb: Danish people can neither speak nor spell ;)
<eyJhb>
> SEK = x : toString x + " SEK (weakest of the Norse Brothers) = " + toString (0.71*x) + " DKK"
<{^_^}>
SEK defined
<eyJhb>
> NOK = x : toString x + " NOK (second weakest of the Norse Brothers) = " + toString (0.69*x) + " DKK"
<{^_^}>
NOK defined
<joepie91>
lol
<eyJhb>
I think I need srhb to give me some backup at some point. She is a much better dane than me..
<adisbladis>
I really like the scandinavian sibling rivalry we got going here <3
drakonis has quit [Ping timeout: 246 seconds]
drakonis has joined #nixos-chat
<eyJhb>
That is just the life of being from Scandinavia :D
drakonis_ has quit [Ping timeout: 260 seconds]
<adisbladis>
There are not that many norwegians in here though :/
<adisbladis>
Despite there being an active NixOS community there!
<eyJhb>
True, missing some norwegians
<eyJhb>
Reminds me, censor/co-examinator for my project exam lives in Stockholm
drakonis has quit [Ping timeout: 272 seconds]
<eyJhb>
He has taken the same education as me at Aalborg University, and then moved to Sweden afterwards. Kinda fun to think of.
drakonis_ has joined #nixos-chat
<eyJhb>
The less fun part is them considering making the exam physical.
<srhb>
eyJhb: Wait what am I backing up :P
<srhb>
adisbladis: Spoke to one Norwegian at NixCon last year, but haven't seem him around despite some encouragement. It is a bit of a shame. :)
drakonis has joined #nixos-chat
drakonis1 has joined #nixos-chat
drakonis has quit [Read error: Connection reset by peer]
drakonis has joined #nixos-chat
<eyJhb>
srhb: you just need to show that Danish people knows how to spell :p Because my track record is not good to say the least
<srhb>
Ah!
<srhb>
I think I've got that down, though the (totally real and not just some excuse) recent laptop problem I have with keypresses being eaten somehow doesn't show it well. :P
drakonis_ has quit [Ping timeout: 260 seconds]
<eyJhb>
Whe needs those keypresses as well, it is just in-voluntary compression :D
drakonis1 has quit [Ping timeout: 272 seconds]
hax404 has quit [Ping timeout: 260 seconds]
hax404 has joined #nixos-chat
hax404 has quit [Ping timeout: 272 seconds]
hax404 has joined #nixos-chat
<MichaelRaskin>
Sooo, a second time when I do not receive my Cosmo before I need to start coming to work again. (First time was that theoretical possibility that they would ship before Christmas)
<ar>
MichaelRaskin: i got mine before christmass, and i'm looking to get a different phone, so that i can hack this one
<MichaelRaskin>
Well, I ordered pretty late and with Cyrillic keyboard
<ar>
keycaps are easily replacable
<MichaelRaskin>
Yeah, maybe should have found out whether Gemini keycaps are easy to replace and bought a gemini
<ar>
also, at least the polish keyboard layout has characters pointlessly swapped with regards to the english one
<ar>
for example, ~ and ` are, respectively, fn-1 and fn-2 on english layout, but fn-2 and fn-1 on the polish one
<MichaelRaskin>
Cyrillic has this bonus of _definitely_ needing another layer, so I expect the Latin part to be US English layout
<ar>
also, there's a bunch of keyboard shortcuts that aren't present anywhere in the manual or - apparently - anywhere in text form
<ar>
and the only source for them is a youtube video, without timestamps, posted by planetcom
<MichaelRaskin>
Wait, I think _many_ people have received theirs by Christmas, aren't these shortcuts transribed somewhere on forums?
<ar>
also, astro seems kindof interesting, but i'm hestitant towards throwing more money at them, since they provide almost no software updates
<MichaelRaskin>
Astro looks like something even harder to get right mechanically than fxtec pro 1
<adisbladis>
ar: Still those pesky Mediatek chipsets :/
<ar>
yeah
<emily>
"It is now possible[1] to perform chosen-prefix attacks against the SHA-1 algorithm for less than USD$50K. For this reason, we will be disabling the "ssh-rsa" public key signature algorithm by default in a near-future release." https://www.openssh.com/txt/release-8.3
<bqv>
oh wow
<emily>
it's not clear to me reading this whether ssh-keygen -t rsa gets you ssh-rsa or ssh-rsa-sha2-(256|512) with a "recent" openssh
<emily>
if the former, :popcorn:
<bqv>
oh boy
<emily>
maybe I should buy yubico stock for the incoming replacement purchases
<emily>
I guess they're not even public
<MichaelRaskin>
ssh-rsa
<emily>
:')
<__monty__>
Tried connecting with the flag disabling ssh-rsa and it connected fine. Does that mean my keys already use rsa-sha2-256?
<emily>
pass a -v or two to find out
<emily>
that's just for host key algorithms though
<emily>
most distros generate host keys with multiple algorithms by default, including nixos
<emily>
it's the client keys you need to worry about
<emily>
if you .ssh/authorized_keys doesn't have any non-ssh-rsa lines you have issues
<emily>
(on remote hosts you connect to)
<__monty__>
Oh ok, then I'm completely fine. No need for anyone to test my keys >.> <.< >.<
cole-h has joined #nixos-chat
neeasade has quit [Remote host closed the connection]
<cransom>
i wonder if amazon's codecommit still requires rsa keys
waleee-cl has joined #nixos-chat
drakonis has quit [Read error: Connection reset by peer]
drakonis has joined #nixos-chat
<cole-h>
gchristensen: btw skillsmatter did indeed send me an email reminder :^)
<gchristensen>
nice
<gchristensen>
just got off the test call
<cole-h>
:D
<cole-h>
Looking forward to it. T-24h10m!
drakonis1 has joined #nixos-chat
drakonis_ has joined #nixos-chat
drakonis has quit [Ping timeout: 272 seconds]
<cole-h>
And to answer my question from yesterday: it appears it'll be on Zoom
<gchristensen>
oh, yeah, it is on zoom
<gchristensen>
I learned that today on the test call
KeiraT has quit [Remote host closed the connection]
<gchristensen>
any spotifyd users?
KeiraT has joined #nixos-chat
<bqv>
it's in my config but i bet you know more about it than me
<bqv>
i just added it as a bookmark
<gchristensen>
ah
<gchristensen>
I'm wondering if there is an information leakage
<gchristensen>
ie: it stores files like this: ~/.cache/spotifyd/files/a8/baa9820053699a9c6c696c1b13c9683a1ebbbb and I'm wondering if you could use this data to determine what people listen to
<gchristensen>
(that said, ~/.config/spotifyd/spotifyd.conf has your username and either the pasword or a way to get the password -- so if you can enumerate those files, you could just look at the password)
<adisbladis>
gchristensen: spotifyd can shell out to another command to read the password?
<gchristensen>
yeah
<gchristensen>
password_cmd = pass spotify.com | head -n1
<adisbladis>
Also you don't strictly need to log in with spotify iirc
<gchristensen>
oh?
<__monty__>
I wouldn't expect spotify to consider song history all that private?
<adisbladis>
gchristensen: I seem to recall it showing up on the local network
<adisbladis>
So you could just tell your phone to play on your desktop for example
<gchristensen>
cool
<evanjs>
when you get karma for spamming random crap you're working on :D
<evanjs>
looks like my AD(H)D/Autism/Aspergers is paying off lol
<gchristensen>
__monty__: I don't like talking about the music I like. nobody has any business to know just how much I listen to taylor swift
<bqv>
lmao
<LnL>
gchristensen: will that be recorded?
<gchristensen>
LnL: I think it will be!
<LnL>
great :)
<__monty__>
I'm not saying it shouldn't be private. I'm only saying I'd be surprised if spotify put effort into making it private.
<gchristensen>
yea
<evanjs>
sorta the opposite for me. I look for chances to share how "odd" my music tastes are. Though I guess they're not too out there lol
<evanjs>
mostly just HARDCORE TANO*C lol
<evanjs>
(JCore)
<gchristensen>
(to be clear, I think taylor swift's music is fun and good)
<bqv>
having a well defined taste in music is beyond my ability
<cole-h>
gchristensen: Good news: while Zoom didn't work on my non-NixOS system, it works just fine on NixOS :D I won't have to boot into Windows to attend
<Valodim>
I had to join a zoom call recently, but didn't have it installed. and I noticed that while doing "nix run nixpkgs.zoom-us" and watching it start to build stuff, I was nearly 0 concerned that it would work out in the couple of minutes I had before the call :)
<MichaelRaskin>
cole-h: wait, how NixOS Chromium is different from Chromium on any other Linux?
<cole-h>
MichaelRaskin: I meant the `zoom-us` package
<MichaelRaskin>
Ah
<cole-h>
It would segfault for me
<cole-h>
Probably due to the impurity of opengl
<bqv>
adisbladis: running emacsclient -t inside vterm seems to crash the vterm for me :D can you reproduce?
<adisbladis>
bqv: Lol
<adisbladis>
Wth are you doing :D
<bqv>
i was just curious what it would look like, honestly :p
<adisbladis>
This reminds me of pairing with a colleague the other day
<adisbladis>
Using EXWM, with a terminal inside running a terminal emacs, inside that is a vterm running vim
<bqv>
wow
<bqv>
that's impressively insane
<adisbladis>
Oh right, over tmate
<adisbladis>
So tmux was in there too
<adisbladis>
All the good stuff
<cole-h>
Did you have a screen session running in the background for some long-running process? ;)
<sphalerite>
no, inside vim
<bqv>
sins have been committed
<eyJhb>
Sooo, now I have the screens, time to set them up
<adisbladis>
evanjs: What do you say about clown core?
<evanjs>
adisbladis: *googles* lol what even
<sphalerite>
eyJhb: what, already??
<sphalerite>
that was fast
<adisbladis>
I can't even find my favourite clown core song online :/
<sphalerite>
gchristensen: buy CDs with cash, destroy them after ripping, block reading the directory for everything other than the player itself
<sphalerite>
gchristensen: actually, destroying them would be environmentally irresponsible. Mail them anonymously to randomly chosen charity shops, and make sure not to leave any fingerprints
<gchristensen>
hehe
<gchristensen>
I'm just safely storing my backups with this charity shop
<sphalerite>
no, once you've got the music off the CD it's finished serving its purpose for you
<sphalerite>
you back it up like you'd back up anything else
<sphalerite>
on magnetic tape in an arctic cave or something
<adisbladis>
I don't normally run a web server there
<cole-h>
Darn, I was planning on scraping it every day :( (/s)
<cole-h>
OK, I think it's time to install to the SSD. I'm tired of cding into nixpkgs and having to wait for more than 10 seconds before I can do stuff (because I have git status in my prompt)...
<emily>
ultrabook thermal performance sure is a lot worse in the summer
neeasade has joined #nixos-chat
<__monty__>
Even my desktop has a hard time with the temps.
<ashkitten>
hmm async reprojection in vr is like, the equivalent of async scrolling in 2d applications right?
<bqv>
i've compromised and just removed all but the most simple of git info from my prompt
<bqv>
partially not by choice, tbh
<bqv>
but it's a lot faster and at least i'll get better cli habits
* emily
finds the git info in fish's default prompt both fast and sufficient
<ashkitten>
is cole-h on some sort of ridiculously slow uncached hard drive?
<bqv>
a stone tablet, perhaps
<emily>
nixpkgs git status is pretty slow even on my PM981
<ashkitten>
my zpool is on spinning rust but zfs is very good at aggressive caching i guess
<eyJhb>
adisbladis: yes :p
<emily>
ok, I guess it's only 1s measuring just now, but I swear I've seen it take several seconds before
<emily>
it has a warning it prints out if it's being too slow and I've seen it
<eyJhb>
But one of them has a big scratch....
<eyJhb>
And I cannot run them in 2k, or else I have to sacrifice a monitor
<cole-h>
ashkitten: Maybe. Once I get in there, it's really fast, but first entry after a long suspend or period doing something else has me wait a looooong time
<cole-h>
LnL: Just to clarify on your ZFS automount setup: if I have a /home ZFS partition, should I legacy mount that too, or will ZFS/systemd/friends handle that for me if I have / and /var legacy mounted?
<lovesegfault>
But I find that something isn't quite right as if I `nix build -f. foucault` and `nix build -f deployment.nix` a very different set of things gets built
<lovesegfault>
cole-h: morph was getting on my nerves and NixOps isn't ready so I want to try nixus :P
<bqv>
what's up with morph?
<infinisil>
Hm I see
<infinisil>
I'll check if maybe it passes the wrong pkgs
<lovesegfault>
<3 infinisssssssssssssssssssssssil
<lovesegfault>
oops
<bqv>
lol?
<lovesegfault>
<3 infinisil
<{^_^}>
infinisil's karma got increased to 302
<bqv>
guess someone's in slytherin
<cole-h>
🐍
<lovesegfault>
My keyb got stuck :P
<cole-h>
emilazy: Do you know if it's possible to have the default fish prompt show if there are any dirty files and/or if the current branch is ahead or behind? That's really all I use the pure prompt for
<cole-h>
emily: ^ (sorry)
<emily>
"any dirty files" is the expensive operation
<emily>
since it needs to scan the whole tree
<cole-h>
Ah
* lovesegfault
nods
<emily>
you could write an inotify thing to approximate it, maybe
<cole-h>
True. Then I'd settle for ahead/behind indicators x)
<lovesegfault>
you can use an async prompt with zsh
<cole-h>
Yeah, that's the only thing I miss from zsh
<cole-h>
powerlevel10k is such a nice prompt because of how fast its git status is
<lovesegfault>
I use starship and it's meh
<cole-h>
But I muuuuchhh prefer fish's autocomplete and autosuggestions
<lovesegfault>
I like how it looks
<lovesegfault>
but it's slow because they are OCD about supporting different shells
<lovesegfault>
which IMHO is a huge waste of time
<lovesegfault>
if they were just a zsh prompt they could benefit from async stuff
<infinisil>
lovesegfault: Try adding an overlay like `self: super: builtins.trace super.path {}`, what path does it output?
<infinisil>
Or does it not output anything? (-> the overlay doesn't even get applied)
<lovesegfault>
infinisil: Where should I add it? to that `thinkpad-p1.nix`?
<infinisil>
Yeah probably
<lovesegfault>
Added, building
<lovesegfault>
I get this over and over : trace: /nix/store/dxflrip8139q3x4mlf5nm5bjfmk06a9m-source
<cole-h>
And if you `ls` that, is it your nixpkgs?
<lovesegfault>
and if I use my default.nix it works as well
<lovesegfault>
wat
<infinisil>
It should be yeah, it's not a nixpkgs version that nixus internally uses
<infinisil>
Also from my own testing, overlays with nixpkgs.overlays do indeed get applied to the pkgs in that scope
<lovesegfault>
I'm doing nix-collect-garbage to test
<lovesegfault>
going to `nix build -f.` then `nix build -f deployment.nix`
<infinisil>
lovesegfault: You should be able to just do `nix-store -qR $(nix-instantiate)` both ways, then compare the result with e.g. diff, to see which paths differ
<infinisil>
Then no need for building or gcing
* lovesegfault
tries
<infinisil>
lovesegfault: You could disable the overlay you suspect doesn't get applied for nixus, then check if that makes both ways result in the same (that would confirm it)
<lovesegfault>
Alright this is even weirder
<lovesegfault>
I did the nix-store thing you suggested infinisil and dumped to a file. I did it on my laptop and on my build server and now I am diffing them
<lovesegfault>
there are a bunch of differences
<lovesegfault>
both checkouts are on the same commit
<lovesegfault>
Let's see, building on the remote and then nix copying to my box
<ashkitten>
hmmmm run2batch1 cosmos being shipped out to backers, run2batch2 quality checked this week
<ashkitten>
i think i'm probably in batch 2
<ashkitten>
i mean, since my contribution hasn't been locked and they said batch 2 is everyone who's still left
<bqv>
ldlework: about to launch (i think)
<ldlework>
bqv: just hatch close
<bqv>
oh ok
<__monty__>
bqv: Nah, T-minus 2h.
<bqv>
guess i got 2 hours of sweet nerdy background noise
<ldlework>
lol
<MichaelRaskin>
ashkitten: did they say batch 2 or run 2, though?
<ashkitten>
i thought it was this next batch that would fulfill everyone's perks?
<MichaelRaskin>
I think they promised that the run 2 will complete their promises, and the batches seem to be defined ad-hoc
<ldlework>
dying on a spaceship would be so much cooler than dying of asphyxia from a cop's knee on your neck
<cole-h>
s/.*/Dying in space would be so much cooler than literally anything else/
<ldlework>
cole-h: I'm just amazed at humanity's range :P
<sphalerite>
cole-h: dying literally any other way. Being in space and not dying is still cooler than dying in space.
<ldlework>
sphalerite: more pleasurable. but cooler?
<ldlework>
:P
<__monty__>
cole-h: Are you sure? Neil Armstrong is quite a bit more well known than "the astronauts who died."
<MichaelRaskin>
Yes, cooler. Some people manage to _complete work_ in space.
<ldlework>
that is pretty cool i guess
<__monty__>
Well, when work is dropping a hammer and a feather : )
<MichaelRaskin>
__monty__: are there any astronauts who have died outside the atmosphere?
<sphalerite>
and he's still alive to enjoy being cool!
<cole-h>
__monty__: Not being well-known doesn't mean it's not cool!
<__monty__>
MichaelRaskin: Didn't apollo 13 not return?
<sphalerite>
err nvm
<sphalerite>
but he lived a long time to enjoy being cool*
<MichaelRaskin>
__monty__: sorry?
<__monty__>
Oh, they returned.
<sphalerite>
it was buzz aldrin who's still alive
<ashkitten>
MichaelRaskin: ah, well i hope mine is in this batch though
<MichaelRaskin>
I have given up on understanding this batch structure, but I do hope whatever batch it is they do something with this entire run in the next couple of weeks…
<__monty__>
MichaelRaskin: Crew of Soyuz 11 were killed after undocking from Salyut 1. As of 2020 the only human fatalities above 100 km.
<ashkitten>
MichaelRaskin: they said today the first batch reached their warehouse and are being shipped out
<MichaelRaskin>
ashkitten: yes, I have read this email
<ashkitten>
sorry what are we talking about?
<MichaelRaskin>
ashkitten: what you are saying was in their update that they have also sent via email
<ashkitten>
oh
<ashkitten>
sorry, my short term memory issues are being exacerbated by a lack of restful sleep
<ashkitten>
i lost the train of thought corresponding to this conversation
<MichaelRaskin>
__monty__: Hmmm, actually it is not clear how long did they survive.
<__monty__>
MichaelRaskin: What do you mean? They survived depressurization in space?
<MichaelRaskin>
It was a leak, not an instant-to-zero
<MichaelRaskin>
They did survive long enough to reach the valves, but unfortunately the air was leaking too far for them to re-check and force-close both valves (and apparently they first checked the one that stayed closed)
<MichaelRaskin>
leaking too fast
<__monty__>
I assume that was still in space though?
<MichaelRaskin>
Needs some calculation
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos-chat
<MichaelRaskin>
Hmm, yes, probably still in space
bqv has quit [Quit: WeeChat 2.8]
bqv has joined #nixos-chat
<gchristensen>
btw ~1h before spacex launches people
<jtojnar>
did github change icon set recently?
<adisbladis>
jtojnar: I just noticed they changed the bell icon
<adisbladis>
Like 30 seconds ago
<adisbladis>
Must have been within the last hour
<adisbladis>
Or even less
<FireFly>
I get the same ol' bell? I think
<jtojnar>
yup, it was filled and now it is outline
<FireFly>
I get a filled bell still, even if I hard reload.. unsure if something's cached somewhere, or random a/b test
<lovesegfault>
infinisil: is there a way to only build one host in my deployment file if I have many?
<lovesegfault>
Like, I don't want to deploy to all my boxes all the time
<infinisil>
Yeah, I have `host = root@<ip>` specified in my config
<infinisil>
Hm though the bash error is concerning
<lovesegfault>
I have root ssh disabled everywhere :(
<infinisil>
Hm ideally it would work with that
<infinisil>
lovesegfault: You mean it should `ssh` as a normal user and then `sudo` on the target host?
<infinisil>
Because you need to be root on the target to switch systems
<lovesegfault>
Yep, that's what I mean
<lovesegfault>
escalate privilege after sshing is always easier as I think "no root ssh" is a pretty common config
<infinisil>
Shouldn't be too hard to implement, I'll take a look at it later
<lovesegfault>
infinisil++
<{^_^}>
infinisil's karma got increased to 303
* sphalerite
has a great trick for finding people's pronouns
<sphalerite>
grep for \*\tname in the weechat logs and look for a /me where the person uses their pronoun :D
<eyJhb>
sphalerite: it works, the hell
<energizer>
that's how i end up calling people "myself"
* joepie91
does runtime tracking for this
<sphalerite>
joepie91: ?
<joepie91>
I remember people's pronouns when they use them :P
<sphalerite>
oh right
monsieurp has quit [Remote host closed the connection]
<sphalerite>
but I don't always actually see the messages where they use them otherwise
<joepie91>
or try to, anyway
<joepie91>
right :P
<joepie91>
logs are definitely a useful fallback
<cole-h>
Alright people, it is time. I am wiping my Arch install and switching it to NixOS.
<joepie91>
if unclear, I just default to they/them
<colemickens>
cole-h: you fixed your client!
<joepie91>
cole-h: huzzah, welcome to the dark side, cookie jar's over there on the shelf
<samueldr>
cole-h: great!
<cole-h>
colemickens: Wait, what was broken?
<sphalerite>
joepie91: yes same, though I try to find one first :D
<cole-h>
joepie91: :D Although, I have been running NixOS since ~last Sunday
<eyJhb>
Damn it, need some quick overlay support... Anyone up for it?
<colemickens>
cole-h: it sent something weird about not being on a nixos system earlier
<eyJhb>
Maybe not
<cole-h>
colemickens: Hahahahaha
<cole-h>
colemickens++ lmao hahaha
<{^_^}>
colemickens's karma got increased to 26
<sphalerite>
cole-h: did you migrate your previous nixos installation using zfs send? :D
<cole-h>
sphalerite: No, I'm starting from a mostly-clean slate :P
<cole-h>
Real quick: can I rename my zpool from `tank` to `panzer` or something? I figure having two pools with the same name is a bad idea?
<cole-h>
(Since I'm not wiping my test disk just yet)
<gchristensen>
you can
<cransom>
witnessing the transition of servers as pets to file systems as pets.
<gchristensen>
though `tank` is a person
<cransom>
iirc you have to export the pool and then reimport with the desired name
<sphalerite>
yep
<sphalerite>
also adjust your mounts :D
<gchristensen>
so maybe renaming tank to switch would be more appropriate
<sphalerite>
I generally name the pools after the hosts. Am I the only one?
<samueldr>
what's the preferred holidy of ZFS users?
<samueldr>
tanksgiving!
<cole-h>
samueldr++
<samueldr>
holiday*
<{^_^}>
samueldr's karma got increased to 234
<sphalerite>
is there any practical reason to name pools rpool or tank?
<gchristensen>
lol
* colemickens
should've used host-based names so the backup pool names could just be the same
<gchristensen>
sphalerite: copy-pasta from the docs
<cole-h>
^
<samueldr>
good ol' cargo cult
<cransom>
tank is just my default for a zpool that's not root. if it's root, it gets root. no need to follow a hostname imo.
<samueldr>
that's how we got the FHS and FHS defenders' league!
<cole-h>
gchristensen: I like how you mixed rpool + tank in your "ZFS Datasets for NixOS" blog post :P
<gchristensen>
:P
<sphalerite>
we have a backup server at work, `rpool` is a single-device SSD pool with the root on it and `tank` is the big one that has all the backups on it
<cole-h>
sphalerite: btw, naming after host wouldn't work, since this disk is from the same host :P
<gchristensen>
my big pool at home is "big"
<sphalerite>
cole-h: well, I have one host named thirtythree, and it has two pools, "thirtythree" and "thirtythree-ssd"
<sphalerite>
This rocket stream is annoying, I can't play it at 2× speed!
<cransom>
but re tank, being a person. i think i mentioned before someone being at a company where their windows domain controller was renamed by their boss one day to 'ne0' from 'neo' without mentioning it to anyone and the few days their machines were busted
<gchristensen>
lol
<energizer>
what's the "cattle not pets" technique for filesystems?
<cole-h>
I don't get the "tank is a person" thing pls explain
<bqv>
aw, no launch, probably :(
<gchristensen>
cole-h: a long time ago, ZFS devs called their pool "tank" after The Matrix character
<cole-h>
Been so long since I watched the Matrix, I didn't even notice lol
<cole-h>
OK, y'all have swayed me. rpool for root it is :P
drakonis has joined #nixos-chat
<cole-h>
What's teh command to basically zero my SSD? Something like blkdiscard?
<cole-h>
s/teh/the/ (darn laptop keyboard)
<bqv>
abort launch!
<bqv>
epic anticlimax
<Taneb>
No space today
<Taneb>
Next window is Saturday I think
<samueldr>
missed the reason
<sphalerite>
cole-h: yes, blkdiscard will make it discard all the blocks on an SSD
<sphalerite>
samueldr: weather
drakonis_ has quit [Read error: Connection reset by peer]
<bqv>
literally ^
<samueldr>
alright
<energizer>
can i reproducibly configure my filesystems?
<gchristensen>
that is just a typical Florida afternoon
drakonis2 has joined #nixos-chat
<hexa->
:)
<cole-h>
sphalerite: Now the question becomes: should I blkdiscard my SSD, or should I just install over whatever's already there?
<gchristensen>
I grew up about 30min from titusville and playalinda beach
drakonis has quit [Ping timeout: 252 seconds]
<samueldr>
(that's where you go to play a linda?)
<gchristensen>
hehe
<sphalerite>
cole-h: doesn't really matter. `zpool trim` will discard unused space later anyway, and since 20.03 it's run automatically periodically by nixos
<cole-h>
OK, then I'll just not do anything unnecessary. Thanks.
<gchristensen>
I mean, you might as well, but also like, don't have to
<eyJhb>
*sigh* back on my addiction srk and sphalerite.... Just nice to know that it works
* cole-h
just `echo 1 > /sys/block/sdX/device/delete` to all unused disks to be safe
<gchristensen>
I never have because I neve remember to. but I always think "ah, next time"
<colemickens>
some tools will insert an extra prompt if they find an old fs signature :/
<colemickens>
was scripting my partitioning last night actually and noticed that might be a reason to wipe a bit harder
<colemickens>
:|
<gchristensen>
oh yeah
<gchristensen>
erasing all the filesystems is surprisingly tricky
<sphalerite>
wipefs -a is good for that
<sphalerite>
just wipefs won't actually do any wiping, funnily enough
<cole-h>
Alright alright alright, now ladies. It is time to press `o` in gdisk.
<gchristensen>
or cover secure erase
<cole-h>
Yeah, my SSD is frozen according to hdparm rn, so I don't think I can secure erase :P
<colemickens>
sphalerite: see, I swore I did that, and still got a warning from fdisk about an ntfs partition after I wiped+recreated and the first partition aftert the ESP was on the same sector.
<cole-h>
The curses interfaces are too much for me
<cole-h>
This is one of the few times I actually prefer the text-driven stuff
<samueldr>
I personally don't see any reasons to continue recommending fdisk/gdisk; if you want to script, it's not the good choice, if you want to interactively use it, it's a bad UX
<sphalerite>
colemickens: ooooh, because wipefs deleted the partition table but not the filesystem signatures within the partitions… you probably need to wipefs the partitions before wipefs'ing the whole device
<samueldr>
though even as a commands-based UX, it's a pretty terrible one :/
<colemickens>
sphalerite: wipefs -r plz
<samueldr>
options have changed many times in the past
<sphalerite>
colemickens: that would be nice.
<colemickens>
samueldr: I started out with sfdisk, but honestly I am happier with "scripting" fdisk instead. it's way less work, it will do the "right thing" even on differently sized disks, I don't have to script around /dev/nvme0n1p2 naming weirdness.
<cole-h>
How is it bad UX, interactively? It serves me just fine. Maybe the discoverability of stuff is rough, but I already know exactly what to do (o, n ..., n ..., n ..., c 3 "label", w).
<samueldr>
I think the only reason I would have a fdisk or gdisk session is on an extremely slow bandwidth link, and in that occasion I'd rather script it with s[fg]disk
<colemickens>
(in retrospect, I wish I would've scripted gdisk, it has a slightly better UI I think, but still)
<cole-h>
And I don't want to script it (yet)
<samueldr>
colemickens: options names are not stable! use sgdisk or sfdisk!
<sphalerite>
cole-h: idk, I think discoverability is pretty good
<sphalerite>
cole-h: since it says "? for help" right in the prompt
<cole-h>
True
<samueldr>
at the very least, you're all already knowledgeable and used to it
<samueldr>
what really grinds my gears
<sphalerite>
what is annoying is that it'll skip things where there's only one choice
<samueldr>
is recommending this to those who are not used
<sphalerite>
fair
<colemickens>
samueldr: too late, I've already written my angry letter for whenever someone finally tries to change it, or even angrier letter if I only notice after it's been changed and shipped :P
<colemickens>
how dare they break my foot pedal
<cole-h>
Well, that was me last week :P The last time I had partitioned was in 2018 when I installed Arch (and I'm pretty sure I used an installer/script to do that for me)
<samueldr>
colemickens: they broke the nixos installation section in the manual in the past
<samueldr>
still a bit too expensive for me to get, but it tempts me even more
<cole-h>
samueldr: Can you make a partition of 16GB with parted?
<samueldr>
colemickens: no, it's forbidden
<samueldr>
you have to use 15.9 or 16.1GiB
<sphalerite>
samueldr: oooooh I already have a remarkable so…
<cole-h>
e.g. you make your boot partition from +1M to 512M, and then a swap partition from that 512M to 16G+512M?
<sphalerite>
samueldr: do I hear mobile-nixos for remarkable?
<samueldr>
sphalerite: someone will need to step up as a maintainer
<samueldr>
sphalerite: so unless I'm getting one, that won't be me :)
<gchristensen>
whoever buys samueldr one, send me one too
* colemickens
imagines being able to code in the park on that screen, omg
<samueldr>
colemickens: pretty sure it's all doable using parted, I don't remember the details, and it's been more than a year since I wrote that section fresh
<energizer>
samueldr: is anbox on the roadmap for nixos-mobile?
<samueldr>
sphalerite: (I wasn't asking for one, but rather nudging you into being the maintainer btw)
<samueldr>
energizer: is anbox on the roadmap for nixos?
<samueldr>
that's not of the domain of mobile nixos
<cole-h>
samueldr: (btw I think you're pinging colemickens when you mean to ping me... :P)
<samueldr>
cole-h: ugh, y'all starting with cole now
<cole-h>
LnL: How should I specify my e.g. /var mountpoint when I zfs mount? If I set it to /mnt/var, won't it mount there when I finish setting up and reboot?
<adisbladis>
But Hybris is just too much of a buzz kill
<LnL>
cole-h: not sure what you mean
<samueldr>
adisbladis: yeah :/ finishing up some yak work to get back into libhybris
<cole-h>
LnL: Well, I figure I need to set my rpool/system/var mountpoint to /mnt/var in order for nixos-generate-config to notice it, right?
<samueldr>
adisbladis: with proper bootloader management on my part it'll be easier to rebuild and reboot
<cole-h>
So I would imagine it would then want to mount at /mnt/var every time, instead of /var
<cole-h>
(This could be me totally misunderstanding zfs mounts and stuff, though)
<samueldr>
adisbladis: my main gripe last time I worked on it was that a dev cycle to test took like 20-30 minutes
<cole-h>
s/could be/probably is/
<samueldr>
adisbladis: so I spent weeks, nay, months, in making it quicker!
<samueldr>
(not literally)
<LnL>
cole-h: I have it just set to / but not sure how the installer, etc. deals with that
<LnL>
cole-h: does the installer even need that?
<cole-h>
Well, I want to separate / from /var
<samueldr>
sphalerite: finally read the instructions... I don't think it should be required to add an sd card slot, I wonder how brick-proof its internal storage is though
<samueldr>
confirmed in the comments, not required to add an SD card slot
<LnL>
cole-h: if the installer touches that might be safer to keep that legacy
<LnL>
but I think all it does is maybe generate some directories
<samueldr>
and looks like it's a u-boot based system, so the real question is what happens if you empty its storage
<samueldr>
(i.mx6)
<LnL>
everything like journald is still running from the installer image itself at that point
<LnL>
so that won't write to /mnt
<adisbladis>
samueldr: Oh, and back when I did that work Plasma on Wayland was broken on NixOS anyway.
<adisbladis>
It may still be
<cole-h>
LnL: Ah, looks like all I had to do was set `mountpoint=/var` and zfs auto-mounted it. Cool.
<cole-h>
We'll see how this goes -- it's not like I have anything important on the disk right now :P
<gchristensen>
don't depend on that
<LnL>
oh, assumed you knew that
<cole-h>
Heh
<gchristensen>
as of now, always use mountpoint=legacy for nixos
<cole-h>
gchristensen: Well, LnL said that legacy isn't always necessary, which is why I thought to try it
<gchristensen>
ewll
<gchristensen>
it isn't necessary if you don't care when it gets mounted :)
* LnL
doesn't use it (except for store etc.)
<gchristensen>
but you don't want, say, journald to open up a big file in /var/log/.... and *then* have /var mount
<gchristensen>
you'll see all sorts of weird behavior
<cole-h>
Hmm... good point...
<cole-h>
I imagine /home will be mounted before I can login at the TTY, so that should probably be safe to not be legacy... right?
<LnL>
yeah I think var might be safer, but I have a bunch datasets in /var/lib for other services which works totally fine
<{^_^}>
#21929 (by ToxicFrog, 3 years ago, closed): `zfs-mount.service` doesn't have the correct dependencies
<LnL>
maybe a year 2.5 years more :p
<ashkitten>
sway has crashed like 7 times today and it's only 14:00
<sphalerite>
ashkitten: that's like, once every 2 hours!
<ashkitten>
assuming i've been awake since 00:00
<ashkitten>
which i haven't
<ashkitten>
i've been awake 4 hours
<ashkitten>
also, it brought down my entire system once today
<sphalerite>
cole-h: no, wantedBy doesn't express ordering at all
<cole-h>
ashkitten: :')
<ashkitten>
i shouldn't've updated it
<colemickens>
I haven't had it take out my system yet but "sway" has been spoken aloud more than usual lately, often with another word or two.
<ashkitten>
it's segfaulted like a billion times
<cole-h>
... I just realized I forgot to move my super-big-VM-image to my external lol
<ashkitten>
i don't like saying this but if it was written in a safe language this literally would not happen
<colemickens>
every time I look at virt-manager it crashes
<cole-h>
Gonna have to restore that :')
<colemickens>
at least obs I can run with xcb and it won't crash, i just can't screencapture with dmabuf :|
<cole-h>
Not looking forward to transferring 400GB from spinning rust :')
<ashkitten>
cole-h: that's only like, an hour of data transfer
<colemickens>
ashkitten: I'm unproud of how often I want to say that in another channels :P
<cole-h>
Depending on how fast the disk is :')
<sphalerite>
ashkitten: wrong! The bugs are mostly in wlroots iirc and sway being written in a safe language wouldn't help with that :P
<cole-h>
Imagine if wlroots was written in a safe language!
<colemickens>
my understanding is that that's hard for reasons
<MichaelRaskin>
Imagine wlroots was written at least in C that is not doing magic tricks with pointers for no reasons except pure chos
<colemickens>
but it does seem like some wlroots issues might've been mitigated earlier on by the guidance on a borrow checker
<colemickens>
^
<ashkitten>
at least with safe languages you can isolate and inspect the unsafe parts
<ashkitten>
if you do have to write unsafe code you know exactly where it is and you can effectively quarantine it
<LnL>
unsafe { main() }
<MichaelRaskin>
I have written interfaces to C code, and I have read battle notes of attempts to wrap wlroots, and well, typical C code does not proudly push that levels of pointers magic to API
<cole-h>
Uh, cool, no space left on device when installing my bootloader?
<ashkitten>
MichaelRaskin: please tell that to drew's face
<ashkitten>
like, legitimately
<ashkitten>
i think he needs more criticism
<cole-h>
Time to remove /sys/firmware/efi/efivars/dump-*
<cole-h>
And hope it doesn't brick my system
<MichaelRaskin>
ashkitten: by now I am «not sure» that will help
<colemickens>
cole-h: I have to delete those anyway (though I'm not sure what makes them) otherwise my XPS13 takes 5 minutes to POST
<ashkitten>
:<
<colemickens>
took me nearly a year to find out why. >_<
<cole-h>
Hey, it worked :)
<cole-h>
If you just hit enter for the root password prompt, does it set an empty password, or disable the password altogether?
<colemickens>
cole-h: whoa, so like, the efi nvram was out of space for the entry for the nixos boot loader? o_0
<cole-h>
colemickens: I have no idea lol :D
<sphalerite>
cole-h: it doesn't set a password
<cole-h>
Ah, it asks me to set it again lol
<cole-h>
--no-root-passwd to the rescue!
<sphalerite>
cole-h: if you do it three times it will give up
<MichaelRaskin>
ashkitten: we don't need fewer people writing weird code, we need fewer people thinking that Wayland is a newer protocol therefore Sway is mor esecure than Xorg
<hyperfekt>
cole-h: although i think upstream might be busted because it's not putting the dmesg into the journal for me, and idk why
<samueldr>
hmm, is the empty bell on github a graphical ui change, or does it mean something different than the previously filled bell?
<ashkitten>
MichaelRaskin: i think wayland has the ability to be more secure than xorg not because it's possible to restrict programs in what they can do, but because it's possible to implement proper user consent flows (which nothing has done)
<MichaelRaskin>
Well, you still need that baseline thing called «not dereferencing broken pointers»
<ashkitten>
yeah
<MichaelRaskin>
At least, not hourly
<jtojnar>
samueldr most icons are now outlines
<samueldr>
yeah, just saw other icons after that
<MichaelRaskin>
Also, with X protocol you could also have proper user consent flows
<Valodim>
what are "proper user consent flows"?
<MichaelRaskin>
Arguably, you would also be forced to implement them _the right way_
<MichaelRaskin>
Well, right now whatever can ask for a global keybinding
<samueldr>
(that's a joke, like the cookie prompts)
<ashkitten>
frankly i just want something that works and doesn't crash
<MichaelRaskin>
Xorg
<ashkitten>
i wanted to use sway because it has some nice features
<ashkitten>
over i3
<ashkitten>
i'm thinking of switching back to i3
<ashkitten>
but that means more work to switch back
<ashkitten>
sigh.. okay
<sphalerite>
I'm using sway because yay rapid adjusting to multiple monitors and multiple scaling factors
<hyperfekt>
i want to use it because i have the inexplicable need to live on the bleeding edge with all my software
<hyperfekt>
wayland is newer ergo desirable
<sphalerite>
what I'm starting to wish I had though is a wayland compositor that I can control with emacs so I can live the exwm dream but also have different scaling factors (disclaimer: I haven't even tried exwm yet)
<hyperfekt>
sphalerite: you mean you wish emacs implemented a wayland compositor
<adisbladis>
sphalerite: If you find this let me know ;)
<MichaelRaskin>
Oooohh
<MichaelRaskin>
Weston now ships libweston
<sphalerite>
oooooooh ×2
<sphalerite>
hyperfekt: emphatically no :D
<sphalerite>
hyperfekt: ideally, the compositor and the emacs would be independent, and the emacs could just restore everything after the compositor crashed.
<cole-h>
Uh, guys, how do I import my rpool from the NixOS installer? I need to change something and `zpool import rpool` says no such pool available...
<sphalerite>
And I could restart the emacs without losing all of the other stuff that's running.
<sphalerite>
cole-h: what does `zpool import` say?
<cole-h>
oh
<cole-h>
lol
<sphalerite>
wrong name?
<adisbladis>
-f ?
<cole-h>
Needed to import as root :D
<sphalerite>
haha
<adisbladis>
^_^
<sphalerite>
that does help
<sphalerite>
adisbladis: what I'd also really like is an emacsclient that implements the gtk stuff, such that the actual emacs doesn't need to care about it…
<sphalerite>
adisbladis: I've lost my emacs servers the past few times sway has crashed
<sphalerite>
which kind of sucks
<adisbladis>
sphalerite: That's not a problem with exwm ;)
<adisbladis>
You just don't use emacsclient
<infinisil>
lovesegfault: I guess you have passwordless sudo?
<MichaelRaskin>
Now we just need to implement a reasonable Xorg-like architecture around libweston, with a WM process and window-decoration processes separate from the process holding the memfd's
<FireFly>
MichaelRaskin: oh that's nice, wlroots could certainly use some healthy competition
<cole-h>
Now, uh, how do I decrypt the pool? x)
<FireFly>
*nod*
<adisbladis>
But yes, it would be nice if you could actually attach fully remote clients
<sphalerite>
cole-h: zfs load-key -a
<cole-h>
<3 sphalerite
<{^_^}>
sphalerite's karma got increased to 0x5c
<sphalerite>
adisbladis: also on my wishlist is multiple-state support for evil-mode, so multiple clients don't have to share a mode in the vim emulation…
<sphalerite>
there are so many nice things that could exist but don't yet!
<sphalerite>
adisbladis: maybe we should reimplement emacs and make everything better.
<sphalerite>
s/and/in rust and/
<sphalerite>
(that's the meme right?)
<lovesegfault>
infinisil: correct
<MichaelRaskin>
sphalerite: Nah, Emacs should be reimplemented in Guile
<adisbladis>
I'm not so sure anything is gonna come of it though
<energizer>
emacs should be reimplemented in C++ and typescript
<bqv>
MichaelRaskin: that's been tried
<bqv>
iirc
<bqv>
guileEmacs
<adisbladis>
I'm pretty excited about native-comp
<MichaelRaskin>
Yes, but I think it is replacing the Elisp engine with Guile, but not all the C code for GUI etc. (which could be replaced with Guile bindings to GTK / native Guile code)
<cole-h>
I sure wish I could `zfs create` multiple filesystems at a time...
<energizer>
make a filesystem creator in nix!
monsieurp has joined #nixos-chat
__monty__ has quit [Quit: leaving]
<sphalerite>
adisbladis: ooh yeah I saw that a while back, I wasn't using emacs as extensively at the time
<sphalerite>
really though? Because then people will start asking questions like "which ones does -o compression=on apply to when I pass multiple args?"
<cole-h>
lovesegfault: So much for your pure system with all that pinning ;)
<lovesegfault>
cole-h: be quiet you! 😭
<infinisil>
You can even check the deploy script yourself, there's not much magic going on. The only thing that could be it is the PATH setting in the beginning
<infinisil>
But that error seems to have come from the target host, so that shouldn't be it
<gchristensen>
I'm *fairly* certain somebody has implemented this before in nixos
<gchristensen>
but they've sort of fallen off the map
<infinisil>
gchristensen: There was a nixcon talk about a company-internal implementation of a simple auto-rollback feature
<infinisil>
Well not about that specifically, but it was mentioned
<gchristensen>
yeah, at lumiguide
<infinisil>
Ah yeah, I guess that's not what you meant?
<gchristensen>
I'm pretty sure there is/was another implementation, though, using motherboard/cpu features and bootloader integration
<infinisil>
Oh even CPU integration :o
<gchristensen>
yeah, for like a "no really, it'll cost $100,000 to fix this if it doesn't come back" deployment
<infinisil>
Arguably having something like that is a bad sign from the start!
<pie_>
one of those issues was me trying to go "guys can someone that knows systemd merge at least a crappy implementation of this thats better than nothing? :P"
<pie_>
back when i was even more of a noob
<infinisil>
:P
<cole-h>
Good news, team: I'm almost done rebuilding the world (my h-m profile has lots of source builds)!
<cole-h>
Soon, I'll be on a shiny, new, and, most importantly, zippy NixOS install :D
<gchristensen>
infinisil: sometimes that is just the cost of doing business :P
<infinisil>
I guess it's a good idea in any case to really make sure you don't lock yourself out, never a good thing to happen
<infinisil>
Though for NixOS this kind of means to reboot every time you rebuild..
<infinisil>
I guess no need for that unless kernel or so change
<infinisil>
(probably)
<lovesegfault>
infinisil: Opening an issue
<cole-h>
I mounted a disk to get something off of it, but I can't remember what it was......
<gchristensen>
"The gamble was that ordinary civilians and rival nation states would lack the sophistication to communicate with Iridium. Fast forward thirty years and we can buy off-the-shelf components, such as the rad1o [13] software-defined radio that the Chaos hackers used to eavesdrop on Iridium’s unencrypted messages. These devices were so small that they were used as badges at the conference."
<infinisil>
Hehe
<cole-h_>
Ahhhhhhhhhhhhhhhhhhhhhhh
<cole-h_>
I'M BACK, BABY
cole-h_ is now known as cole-h
<gchristensen>
woo!
<cole-h>
Man, I just rebuilt my kernel, and then zen has to go release a NEW one... so much kernel building in one day :D
<pie_>
gchristensen: wonder if this means more formal methods jobs :p
<bqv>
adisbladis: wanna see something absolutely terrifying?