<elvishjerricco>
I love `nix run -f "<nixpkgs/nixos>" vm --arg configuration '...' -c run-nixos-vm`
<elvishjerricco>
Such a convenient way to test out tiny system level things
<gchristensen>
`nixos-rebuild build-vm` too
<elvishjerricco>
Basically the same thing, but that forces you to put a symlink down and run a separate command to start it :P
<gchristensen>
:D
<clever>
gchristensen: also, nixos-rebuild is only present on a nixos machine
<clever>
gchristensen: what if you have nix on a debian box?
<gchristensen>
on a what box?
<gchristensen>
;)
<clever>
:D
endformationage has quit [Quit: WeeChat 2.5]
jasongrossman has quit [Quit: ERC (IRC client for Emacs 26.2)]
<elvishjerricco>
Spent part of the day hacking on my custom backup system, and I've got it creating snapshots and sending them to a backup pool. All I need now is a program to prune the expired snapshots, and my thing will be usable with systemd timers for a continuous backup system. Sweet.
<gchristensen>
nice
<gchristensen>
nice!
<gchristensen>
sign me up!
<elvishjerricco>
:) One thing I need to make this practical is a daemon that reports systemd unit failures over Signal or email or something.
<etu>
Hmm, one would think that that exists already...
<elvishjerricco>
I really wish Signal didn't require a phone number to create an account. I don't have another number to use for setting up an automated messenger.
<elvishjerricco>
etu: If it does, I'd love to know. I'd particularly appreciate a solution with end-to-end encryption of the messages.
<gchristensen>
prometheus and alertmanager can do it
Drakonis has quit [Quit: WeeChat 2.4]
jasongrossman has joined #nixos-chat
<ashkitten>
i still need to figure out how to get zpool status into grafana
<ashkitten>
and an alert on drive failure
<ashkitten>
which i think i can do with a systemd timer/service and prometheus and annotations and some other shenanigans
<ashkitten>
but honestly that's sooooo much work to setup
<elvishjerricco>
It'd be so easy if Signal either supported getting notifications from yourself or setting up an account without a phone number. You'd just need to ask systemd for the failed units and pipe that into `signal-cli`.
<elvishjerricco>
It's really annoying that sending a message to yourself doesn't raise a notification on the other devices.
<eyJhb>
elvishjerricco: doesn't `note to self` do that?
<elvishjerricco>
eyJhb: As I understand it, if you use a device to send a Signal to a device on the same account, that message will be added to the "note to self" log, but no notifications will be issued to any device.
<eyJhb>
Somewhat hackish, but get a dump phone number, register a Signal account using that, and use it for notifications
<eyJhb>
I have around 300 Danish Numbers.. :p
<elvishjerricco>
Heh, yea I've considered using a Google hangouts or whatever phone number, since the phone number is literally only used for confirming a verification code at the very beginning. But that just doesn't really scale, and also seems incredibly stupid.
<eyJhb>
Well.. You could do Signal as a service, but it looses its privacy if you setup a mitm service to redirect messages
<eyJhb>
:p
<eyJhb>
but for a personal project, I would do that. I currently just use pushbullet instead
<eyJhb>
As the only notifications I send it `X container is unhealthy`
Jackneill has joined #nixos-chat
<jasongrossman>
eyJhb: Wow. Here (in Australia) we can only have phone numbers we're paying rental on.
<eyJhb>
jasongrossman: I just have a bunch of prepaid sim cards with between 0 and 10 DKK on.. You can basically order them for free
<eyJhb>
There have been times where my mailbox has been completely full with just sim cards :p
<etu>
eyJhb: I guess your trip north didn't happen this week ;)
<jasongrossman>
eyJhb: Great!
<jasongrossman>
eyJhb: Do you recommend any particular company (for SIMs)?
<eyJhb>
etu: noo, I got a doctors appointment next week.. :( So wasn't possible..
<etu>
eyJhb: 16-18th of august I'll be in Malmö:)
<eyJhb>
jasongrossman: basically anyone who will send it for free :p I use Lebara and Lyca. Lyca don't care at all about anything. Lebara will send you angry letters after a coulple of weeks of ordering each day :p
<eyJhb>
etu: I would never be that late, I have a meeting in Billund on the 15th, and Cpae Town the 23th, and then stuff inbetween :/ (And uni the 1st of september)
<jasongrossman>
eyJhb: "angry letters" made me laugh.
<eyJhb>
jasongrossman: yeah... They got quite annoyed. I gave a invalid email and phone number each time, so they could only contact me pr. my postal address :p
<eyJhb>
They basically had to cancel my order manually each morning
<jasongrossman>
eyJhb: Ha!
<sphalerite>
hmm, is it a bad idea to use the same LUKS header for multiple LUKS devices?
drakonis1 has quit [Read error: Connection reset by peer]
<eyJhb>
jasongrossman: I basically made a deal with them, they send me 50 more, and I would stop doing it :p
<eyJhb>
srhb: I should somewhat say, I enjoy having her alive :p - Btw. you had nothing to do with frontend, did you?
<srhb>
eyJhb: If you mean design, no, nothing.
<srhb>
eyJhb: But tell me anyway, I feed on peoples' criticism ;-)
<jasongrossman>
eyJhb: Thank you!
<eyJhb>
srhb: I can't say much :| I just remembered my instructor saying this weekend, that he did not like the design of DMI, which is why he used yr (combined with like 5 other weather apps). - So just curious if I remembered correct :p
<eyJhb>
Backend 4 life <3
<srhb>
eyJhb: A lot of people prefer yr. There's also an app that aggregates data from dmi and yr. That should definitely be preferable, because the data that yr uses for dk weather is strictly less precise than ours. :)
<jasongrossman>
eyJhb: Awesome.
<srhb>
But people tend to care less about accuracy of weather forecasting and more about layout
<srhb>
Which is sort of hilarious.
<eyJhb>
srhb: I tend to hear the people think yr is more accurate than DMI, which seems quite weird that it would... - Also, we like shiny pretty stuff! That is just the way of life
<srhb>
eyJhb: That's factually wrong. :-P
<eyJhb>
jasongrossman: haha, thanks ;) Been way too long since my last post... Really need to put some time into it
<srhb>
But yeah, I hear a lot of people say the same thing. I think people subconsciously bias the quality of the data upwards because they prefer the design.
<eyJhb>
Isn't yr's design rather.. Basic??
<jasongrossman>
eyJhb: Talking of posts, I love your Twitter feed.
<eyJhb>
Hahha
<srhb>
eyJhb: Yes. As it should be imo :P
<eyJhb>
There is like one callout for a stupid company, right jasongrossman ?
<eyJhb>
Think you can still exploit it
<eyJhb>
Sounds like srhb is not a fan of DMIs new design! :p
<srhb>
eyJhb: In fairness I was not a fan of the old one either :P
<eyJhb>
srhb: what are you a fan of? :p
<jasongrossman>
eyJhb: Yes.
<jasongrossman>
I like that yr.no has weather forecasts for my small town in Australia.
<srhb>
eyJhb: Hmm, the norwegian met service's public APIs? :-P
<srhb>
eyJhb: api.met.no for reference.
<srhb>
I guess I'm also a fan of the fact that we've been mandated to "free" our data.
<eyJhb>
jasongrossman: I should properly talk with my boss at AAU what I am allowed to talk about publicly. If she says okay for everything, I got sooo many stupid and annoying things from the wold of universities...
<eyJhb>
Ah. When I read "met" I thought of "meet" => "dating". And was quite ensure wth :p
<eyJhb>
Do you have a deadline for that srhb ?
<srhb>
Oops, sorry, yeah, it's short for meteorological which is a horribly convoluted word...
<srhb>
eyJhb: There are some phases, I don't recall exactly. The first phase is this autumn, I think for instrumental measurements.
<eyJhb>
So.. That is quite soon?
<srhb>
(That is, make it possible to query each instrument in denmark for, say, measurements over the last 24h)
<srhb>
Yes :)
<eyJhb>
each instrument? So all your measuring equipment ?
<srhb>
As far as I know, yes.
<srhb>
I mean, you won't actually talk to the instrument of course.
<eyJhb>
Awwwwwwwwwwwwwww
<eyJhb>
I hoped for some hacking fun
<eyJhb>
But that would also kill your instruments. Hoping it is talking to your own cache of it :p
<eyJhb>
I am quite sure I will join you soon srhb with the `take me out back and shoot me` if this PR needs more changes
<srhb>
eyJhb: Such is the life of the OSS contributor ;-)
<srhb>
At least you get to feel good about it and have an automatic CV ;-)
<eyJhb>
True, but normally it isn't that annoying at all, but this is 9 packages/changes. And I haven't found a good way to make a single change for that commit. But that might just be my git foo which need to be trained :p
<eyJhb>
Actually the worst part is, I am 50/50 Gitlab/Github now. So my automatic CV looked much better when it was only Github :p
<srhb>
eyJhb: Yes, probably, what are you trying to do exactly?
<srhb>
Normally I rebase interactively against my base branch and choose whichever commit I want to alter, then recommit it, it's a very pleasant process :)
<eyJhb>
Nothing currently srhb , only if worldofpeace has anything. But if I e.g. wanted to edit a commit in-between them all
<srhb>
Then you pick "edit" on the commit in the interactive rebase against your base branch
<eyJhb>
Yeah, that seems like the way to go! But I did not know that.. And not quite sure I do now. I guess it is just `git rebase --interactive 'hash^'`, change it to edit, make changes, `rebase --continue` after comitting them
<srhb>
Or you can add a new commit with the change, rebase interactively, move the new commit to just after the old commit and choose "Fixup" to squash it into the existing one
<srhb>
I do the latter a lot
<srhb>
It's a very nice process :)
<srhb>
eyJhb: But yes, basically git rebase -i base-branch
<srhb>
git has instructions for all the steps from there :)
<srhb>
Definitely a must-know for making nice, comprehensible git histories for review purposes :)
<eyJhb>
:( Don't you leave me now then, wop just made a bunch of suggestions
<srhb>
But they are a good person, it'll be all the better for the attention ;-)
<eyJhb>
9 commits on the PR, 9 commits on the PR, you take one down, patch it around 8 more commits on the PR. <- My bad spin of 99 bottles on the wall
<srhb>
lol
<eyJhb>
Yeah.. But daaaamn, I hate this PR :p
<eyJhb>
Don't I get something for that srhb ? I think it was a worthy try ! :p
<srhb>
eyJhb: Definitely. Two points for Hufflepuff.
<eyJhb>
^ quite sure the sorting hat is broken, I am at least a Ravenclaw! :(
<srhb>
:D
<eyJhb>
srhb: can one actually check the ++ score, without .. Like... Going
<eyJhb>
instead :p
<srhb>
I think only if it's exposed via the nix plugin, which I'm not sure it is..
<eyJhb>
Still quite curious if it is possible to do RCE on {^_^}
<srhb>
RCE?
<eyJhb>
Remote Code Execution :p
<srhb>
What a weird generic name.
<eyJhb>
Have never heard it in any other context (as far as I remember), but yeah...
<eyJhb>
But I am not a good enough Nix programmer to do anything with it :p
<aria>
yeah i've only ever heard RCE mean remote code exec
<jasongrossman>
eyJhb: Of course there's a workaround. You can ++ yourself (which will show your score and reduce it) and then I can ++ you.
<eyJhb>
jasongrossman: it's okay, I know I am under 0 anyways :D
<jasongrossman>
eyJhb++
<{^_^}>
eyJhb's karma got increased to -2
<jasongrossman>
Ha. So you are.
<eyJhb>
Yeeaaah... Giving myself karma does not work well :p Thanks ;)
<jackdk>
lol eyJhb++
<{^_^}>
eyJhb's karma got increased to -1
<etu>
:D
<etu>
eyJhb++ I leave you there at least :)
<{^_^}>
eyJhb's karma got increased to 0
arahael1 is now known as Arahael
<eyJhb>
Yay, thanks :D
<eyJhb>
Feels good to be back to where I was, even though it soudns ironic :p
<sphalerite>
Any isync (mbsync) users here who know how to delete a mail folder? I've tried just deleting the maildir locally then syncing, but it seems to create it again after finding it on the server, rather than deleting it from the server.
cjpbirkbeck has joined #nixos-chat
dtz has quit [Write error: Connection reset by peer]
thefloweringash has quit [Ping timeout: 250 seconds]
ma27[m] has quit [Ping timeout: 265 seconds]
Ralith has quit [Ping timeout: 265 seconds]
joepie91[m] has quit [Ping timeout: 252 seconds]
domenkozar[m] has quit [Ping timeout: 264 seconds]
colemickens has quit [Ping timeout: 264 seconds]
sphalerit has quit [Ping timeout: 276 seconds]
jasongrossman has quit [Remote host closed the connection]
jasongrossman has joined #nixos-chat
colemickens has joined #nixos-chat
__Sander__ has joined #nixos-chat
dtz has joined #nixos-chat
Ralith has joined #nixos-chat
thefloweringash has joined #nixos-chat
sphalerit has joined #nixos-chat
colemickens has quit [*.net *.split]
colemickens has joined #nixos-chat
endformationage has joined #nixos-chat
<eyJhb>
srhb: I now agree with you, this heat.. Was just at a meeting at AAU. Now it is hot.. and I have "tordenfluer" everywhere..
<manveru>
sphalerite: lemme know if you find out :)
<manveru>
man, why are there almost no nice free japanese fonts for the web...
<eyJhb>
Wooooooooooooooooooooooooooooooow my neighboor is a dick
<eyJhb>
There is a license for fonts?
<samueldr>
more than one?
<samueldr>
SIL too IIRC
<eyJhb>
More than one?
<manveru>
a ton of fonts have pretty restrictive licenses, esp japanese ones since they take so much work to make
<samueldr>
ah, OFL is The SIL Open Font License, so I don't _know_ of more than one
<manveru>
atm i'm moving a site off of weebly to hakyll, and the client wanted to use "Hiragino Kaku Gothic", which you can use for free via typesquare (only one weight), but the result is just so annoying to use due to the FOUC
<samueldr>
(flash of unstyled content for non-web-people)
<manveru>
they use some js compiler to combine two blobs of WOFF into a single font
<sphalerite>
TIL: zfs send -R does not include permissions. Blargh.
* colemickens
loves seeing nixos users in #sway :)
<sphalerite>
oh yeah, maybe I should try sway one day
<sphalerite>
s/maybe//
__Sander__ has quit [Quit: Konversation terminated!]
kraem has quit [Ping timeout: 248 seconds]
kraem has joined #nixos-chat
<infinisil>
Did youtube just remove the option to share a video at a specific time? Because I can't do that anymore
<joepie91>
infinisil: works for me? share -> checkbox at the bottom of the dialog
<sphalerite>
infinisil: same as joepie91
<infinisil>
Huh, it's working again now
<infinisil>
That box was gone for a day, I had to manually enter the ?t=
<joepie91>
got caught on the wrong end of an A/B test?
<infinisil>
I did temporarily enable and disable static.google.com and google.com on another site inbetween I think
<infinisil>
When the checkbox was gone, I saw it flash up really quickly too I think
<infinisil>
Oh well
<infinisil>
javascript shenanigans
<samueldr>
in the right click menu on the video there is also "copy video URL at current time", fwiw
<sphalerite>
I recently got a shock from a power cable plugged into a desktop PSU which had been unplugged for at least 2 minutes. Anyone know why that might happen?
webster23 has joined #nixos-chat
<samueldr>
ghosts
<samueldr>
capacitors not being "bled" by a resistor in the AC-DC thingamabobs
Jackneill has quit [Remote host closed the connection]
<joepie91>
sphalerite: that shouldn't really happen...
<infinisil>
samueldr: Oh neat (the right click thing)
<infinisil>
That will be much faster than the share button
ixxie has joined #nixos-chat
<sphalerite>
samueldr: joepie91: well what I took from it (measured with a voltmeter afterwards) is that I should never use the power switch on the PSU
<joepie91>
sphalerite: how so?
cjpbirkbeck has quit [Quit: Quitting now.]
<eyJhb>
"ghosts" - samueldr 2019
<eyJhb>
sphalerite: was it the actual end of the cable that you normally plug into a wall that shocked you, while still being in the PSU?
<eyJhb>
What was taht website called with IRC jokes/chat logs?
<eyJhb>
bash.org, got it
<colemickens>
I wonder how much faster chromium would build if the interpreter replacement were done with ripgrep and didn't print to console... 🤔
<colemickens>
gosh darny google just sucks these days. Search for any repo not on GitHub and it just links to a random damn commit in the repo. Gitlab, sr.ht (if it shows up at all), etc.
<joepie91>
I find it fascinating how in the past 6 months or so, I've suddenly started seeing everybody everywhere complain about Google
Jackneill has quit [Remote host closed the connection]
<andi->
colemickens: I have been thinking about replacing bash with dash in the stdenv to speedup builds. Supposedly it is a few times faster.
<gchristensen>
whoa!
<andi->
I tried grapsing how to properly inject it into stdenv ~ZuriHac but haven't investigated further.
<andi->
so many little things to work on but no time /o\
<infinisil>
joepie91: You sure it wasn't just me every time?
<infinisil>
Oh, didn't see colemickens there heh
<joepie91>
infinisil: yes :)
<sphalerit>
andi-: isn't dash a pure Unix shell though?
<sphalerit>
That won't play well with all the bash features we're using
<samueldr>
I think dash tries to taste like bash
<sphalerit>
eyJhb: yes that one
<samueldr>
though I'm unsure now that I check
<andi->
Yeah, Debian made the switch many moons ago. Dash has like 1/4th of the size of code (and complexity?) and tries to be bash compatible
<joepie91>
infinisil: seen it in multiple channels, too
<andi->
I just wanted to give it a try and see if it makes a difference for us
<sphalerit>
joepie91: as soon as I switched it on, the voltage at the terminals disappeared
<andi->
We could use it as standard shell and still provide `bash` as real bash for scripts that want it
<sphalerit>
joepie91: that is, the lingering voltage disappeared when I set the switch to the "on" state
<joepie91>
sphalerit: wut.
<joepie91>
that sounds bad :P
<sphalerit>
joepie91: I'm guessing some capacitors still get mains voltage when the switch is off, then when it gets unplugged they discharge into the mains pins, unless the power supply is switched on in which case the motherboard and other things attached to the PSU drain them quickly
magnetophon has joined #nixos-chat
<andi->
argh, dash didn't support `source`... That is why I stopped trying. Not sure what else will fail. Probably still worth trying to figure out if there is a large enough speed difference to justify changing "everything"
<samueldr>
andi-: source is a bashism
<samueldr>
:)
<andi->
yep
<samueldr>
I'm sure there's going to be a bunch others
<samueldr>
like arrays
<samueldr>
arrays in bash are a big bashism
<gchristensen>
oh and we use those a lot
<andi->
:'(
<samueldr>
doesn't sound bad, sphalerite, it's weird though that the PSU would have a switch with between the caps and the circuit without a bleeding resistor
<samueldr>
I still want to make one, though in-kernel, which is basically "tmpfs, but only for directories"
<samueldr>
so you couldn't have stateful _files_ in / unless you mount something for statefulness
<samueldr>
you couldn't accidentally write files there
<gchristensen>
oh dang
<andi->
symlinks?
<gchristensen>
can / be a fuse fs?...
<samueldr>
hm, that would be allowable
<samueldr>
I think so, but fuse has costs
<samueldr>
not sure if it'd matter for mounted locations
<andi->
I could think of /.foo -> /nix/store/…; /etc -> /.foo/etc
<samueldr>
like, if I mount something to /nix
<samueldr>
does accessing /nix/store go through the fs in / ?
<samueldr>
andi-: not sure I follow
<andi->
probably not very relevant if you still want to mutate some data. I just thought you could do atomic swaps of everything kinda like /run/current-system just on the root level
<samueldr>
ah!
<samueldr>
I see it now
<gchristensen>
:o
<gchristensen>
I feel myself slipping closer and closer to michaelraskin every day.
<samueldr>
is it wrong? :)
<gchristensen>
not at all
<samueldr>
I wonder (1) if the kernel maintainers would be interested and (2) how easy a first draft of such an fs would be
<samueldr>
wondering if it should be tmpfs -o nofiles
<andi->
yeah a tmpfs option would be cool.
<andi->
If the vfs interface is anything like the interface for a virtual file it shouldn't be too hard
webster23 has quit [Ping timeout: 264 seconds]
<samueldr>
btw, what _is_ hugefs? I looked at it quickly the other day but couldn't figure it in the short time I looked
<gchristensen>
it is a content addressed filesystem which transparently stores things locally and/or remotely on s3
<samueldr>
right, then it was what it looked like
<gchristensen>
(after the file is closed it becomes immutable)
<Shados>
gchristensen: / can be a fuse fs, but does mean you need to pivot from some other / originally. Although that's not a huge deal given we already do use an initramfs.
<gchristensen>
cool, I feel fine doing that
<ashkitten>
still bugs me that we call it an initrd ;-;
<ashkitten>
eventually i'll make a pr to fix that and deprecate the old initrd options
<elvishjerricco>
Ah you just don't want to differentiate between a first boot and a reboot. Makes sense
<gchristensen>
right
<gchristensen>
I should clarify that in the readme
<elvishjerricco>
I wonder what would happen if you tried to `zpool create -f` on some zvols as well as the disks those zvols' pool uses.
<gchristensen>
interesting question
<elvishjerricco>
I'm guessing it would forcibly kill the pool using the disks before trying to do anything, and then would complain that the zvols aren't found.
<gchristensen>
I should actually make it lsblk -d -e 1,7,11,230 -o NAME -n
<ashkitten>
hmm, is it possible to cleanly "shutdown" a running multi-user system and reduce it to a single-user squashfs in memory without rebooting?
<gchristensen>
maybe with systemd kexec
<ashkitten>
i know clever had a kexec thing but it didn't unmount things cleanly
<gchristensen>
`systemctl kexec`
<ashkitten>
yeah, maybe if i let systemd do the kexec for me instead of manually
<ashkitten>
then it'd be like shutting down
<samueldr>
sounds like fun, what kind of use case?
<ashkitten>
clever's thing was more for installing nixos on servers where there's no nixos installation template
<ashkitten>
or yk, an iso boot option
<ashkitten>
where unmounting doesn't matter
<samueldr>
yeah, I know about the general use case of clever's
<ashkitten>
my use case would be for stuff like offline disk repartitioning, etc
<samueldr>
wondering the use case of reducing a running system to another
<samueldr>
hmm
<aleph->
Sup folks
<samueldr>
yeah, sounds like fun :)
<clever>
ashkitten: that sounds like a use for my rescue boot
<clever>
ashkitten: this puts the same kernel/initrd from kexec, into /boot, and gives it its own grub option
<samueldr>
(it works well)
* aleph-
goes to read this beautiful thing
<clever>
then you can do a clean shutdown, and boot that any time you want
<ashkitten>
i just realized, i run nixos. i could have a boot option that has everything i need without having to predownload all the store paths (which would likely be bigger than my esp)
<samueldr>
I think the kexec to squashfs at shutdown sounds neat though
<clever>
samueldr: i did that before when i ran gentoo from a usb stick, lol
<clever>
samueldr: something similar
<ashkitten>
it's possible to have it download store paths at boot time, right?
<aleph->
Yes?
<clever>
ashkitten: you would need nix installed in the initrd, and enable initrd network
<clever>
ashkitten: but that could be enough to nix-build configuration.nix
<clever>
might want to use a custom initrd at that point, *looks*
<clever>
ashkitten: this is an entire "linux distro" in a single ~2mb initrd file
<clever>
(its just a single haskell binary, that runs as pid 1)
<ashkitten>
ooh
<clever>
if you want network and nix, then you would have to add a dhcp client, ip2utils, and nix
<clever>
it wont remain 2mb, but will still be pretty slim
<clever>
ashkitten: if your up for a challenge, you could also re-implement dhcp and `ip` in haskell using ffi!
<clever>
then you dont have to bring in those pesky c binaries!
<ashkitten>
i don't know haskell tho :<
<aleph->
That sounds like a great way to hurt myself
<aleph->
Simply dreadful
<ashkitten>
but either way this might be a cool idea
<ashkitten>
or i could just make my boot partition... bigger
<clever>
*looks*
<clever>
if you set `boot.loader.efi.efiSysMountPoint = "/boot/EFI";` then you can put the ESP at /boot/EFI, and then /boot can be anything else (ext4, a dir on /)
<clever>
then the size of /boot isnt related to the size of the ESP
<ashkitten>
the only two things on that 240gb ssd are my boot partition and 32gb swap
<ashkitten>
i could just like
<ashkitten>
make my boot partition 100 gigs
<ashkitten>
if i wanted
<clever>
lol
<ashkitten>
i could make it 200 gigs
<clever>
i would give it more swap :P
<clever>
i have 64gig swap, and still run out occasionaly
<clever>
2g is enough for /boot/
<ashkitten>
i have never needed that much swap and if i did my computer would be brought to its knees already
<samueldr>
close some tabs
<aleph->
clever: Howwww?
<clever>
samueldr: the last incident was building 2 years worth of data into a single list in haskell :P
<aleph->
Eesh
<samueldr>
I do tabs "bankruptcy" regularly (not often enough)
<clever>
a single process was using ~80gig of ram+swap
<ashkitten>
i always just make swap the size of ram in case i want to hibernate, but i will never use hibernation because it's super insecure
<samueldr>
I kill any window I know I won't work on realistically
<samueldr>
those that are still useful but far get onetab'd
<clever>
samueldr: onetab has repeatedly lost its entire database on me
<samueldr>
aw
<clever>
samueldr: ive switched over to sessionbuddy
<ashkitten>
when are we gonna get encrypted hibernation
<samueldr>
I guess when someone writes the needed bits?
<samueldr>
(sorry)
<clever>
ashkitten: there are 2 ways to resume from hibernation, one is to just set a special kernel param, and the kernel reads the swap back into ram
<samueldr>
though I don't think I've seen people talking about doing it
<clever>
ashkitten: the 2nd is to run a special util on a block device, which might support luks?
<clever>
ashkitten: i think this is the new way, you write the block dev info to a magic file, and it resumes
<ashkitten>
hmm
<ashkitten>
i'll try that out
<aleph->
clever: Always something interesting you pull out of that toolbox
<ashkitten>
got a feeling nvidia card still won't be happy
<aleph->
ashkitten: Heh what's the issue with it?
<ashkitten>
nvidia is super buggy wrt resuming from hibernation
<ashkitten>
doesn't reinit properly
<aleph->
Huh is it? Works fine with me
<aleph->
Or well, I haven't noticed any issues
<ashkitten>
might've been fixed since last i checked
<clever>
ive played with pci passthru and xen before
<clever>
sending my AMD gpu to a windows guest
<aleph->
Ugh never again
<clever>
and if linux even touched the gpu once, it would fail, so i had to blacklist it extremely early in the boot
<clever>
which results it in appearing to hang the instant it starts booting (zero gpu output)
<ajs124>
I've been meaning to ask, has anyone here ever seen "!!! Unknown header type 7f" on a PCIe device before?
<aleph->
Somehow despite provisiong a windows vm with 7 cores, 12GB ram and passed through my nvidia gpu it ran at a crawl
<ashkitten>
i think my motherboard has a really bad iommu
<ashkitten>
i couldn't get pci passthrough to work for my gpu
<ashkitten>
gave initialization errors in the guest even though the driver was blacklisted in the host
<ashkitten>
and i gave it the vfio_pci driver
<aleph->
Huh, could I see your config?
<aleph->
What computer model anyway?
<clever>
and once windows has touched the card once, its "spent" and i have to reboot the host before windows will work again
<clever>
(or just ditch windows)
<clever>
we need declarative windows config :P
<clever>
and after reinstalling with a bigger disk, it never worked once
<clever>
then i realized the windows disk was only 7gig, and had no room for games
<ajs124>
It used to work for me, but I installed an UEFI upgrade, because my MB didn't recognise the new RAM I was installing and now I get that weird error I pasted above :(
<samueldr>
I finally got pci passthrough going on mine, but the card can't reset, so one vm boot per boot
<clever>
usb and audio cards hotplug just fine
<ashkitten>
clever: since valve has been focusing on linux gaming i really haven't felt the need to run windows at all tbh
<aleph->
clever: Heh, I have a wine config that will only work if it's the first program opened at boot by me
<clever>
i could dynamically eject them from linux, and hotplug them into windows, and it just worked
<clever>
but the GPU, no way :P
<samueldr>
(tbf I never ran windows on the gpu, only linux, to test it, and never did anything afterwards)
<aleph->
ashkitten: Incidentally I have played far too much FF14 now. This was a mistake
<ashkitten>
aleph-: but it's so good!
<aleph->
It is!
<ashkitten>
i just finished stormblood yesterday
<ashkitten>
im so ready to get into shadowbringers, stormblood has some pretty janky pacing
<aleph->
And reading this reddit thread I'm now remembering when this startup/consultancy wanted to hire me for an architect position at only 1y of exp
<aleph->
My god the fact I made it several interviews is horrifying in retrospect.
<ashkitten>
shadowbringers was written by the same woman who wrote the azim steppe portion of stormblood which imo is the best part, and it apparently follows up on the heavensward storyline so yeah i'm super excited to dive in
<aleph->
ashkitten: How is the NieR crossover content?
<ashkitten>
say what?
<aleph->
Yeah doesn't shadowbringers have a NieR raid or something?
<ashkitten>
no idea, i'm still in the beginning of post-stormblood
<aleph->
nod
<clever>
and lately, steam and proton has made gpu passthru a non-issue
<ashkitten>
yeah
<ashkitten>
exactly
<ashkitten>
it's honestly so refreshing to see valve actually committing significant effort to linux gaming
<clever>
ashkitten: i also recently got a steam link box, its basically an embeded device with hdmi, that uses the "steam inhome streaming" service
<Miyu-chan>
I'm quite surprised by Valve's move to promote Linux gaming TBH.
<aleph->
Oh wat, Tetsuyo Nomura made a FF14 raid
<clever>
ashkitten: any usb keyboard/mouse will relay thru it to the host seamlessly
<ashkitten>
clever: yeah, that sounded cool
<samueldr>
clever: that software is available for rpi3
<aleph->
I should grab one
<clever>
ashkitten: but joysticks/gamepads need a $15 app on the remote end, which may not work on nixos
<clever>
samueldr: yep, .deb files exist for it on arm
<samueldr>
they stopped making the steam link
<Miyu-chan>
What's Valve's rationale?
<aleph->
Since I basically just run my gaming/nix laptop closed as a server box behind my old tv
<clever>
samueldr: yeah, i heard it was selling dirt-cheap, getting rid of stock
<samueldr>
ebgames was the sole vendor, and never really had it in stock, except for the first push from valve :/
<aleph->
Speaking of, anyone gotten discord to work on nixos? Can't seem to get v9 to install. >_>
<Miyu-chan>
Is SteamOS even... alive?
<aleph->
Miyu-chan: Sorta, maybe?
<samueldr>
AFAIK yes, though nothing major lately
<clever>
samueldr: one minor issue, is that "focus follows mouse" confuses the in-home streaming
<clever>
samueldr: steam wont forward alt+tab, and wont let the mouse leave the window, but if the remote mouse moves even 1 pixel, it can focus the wrong window, and then steam blindly streams whatever has focus
<Miyu-chan>
clever: I also had problems with different monitors.
<clever>
Miyu-chan: and its a bit anoying to use a keyboard&mouse on a tv, so id rather use a gamepad when possible
<clever>
but it only works in the steam ui, and stops working the instant i launch a game
<Miyu-chan>
Also. Steam actually uses screen capture instead of window capture, so it gets even funnier.
<clever>
Miyu-chan: it also does pulseaudio loopback capture, so youtube and teamspeak leak :P
<ashkitten>
i really wish it used window capture
<ashkitten>
i want to be able to play games while my computer isn't unlocked
<Miyu-chan>
Oh cool. I didn't know that. :D
<Miyu-chan>
OTOH, I guess in a way, that sounds more of a feature.
<ashkitten>
or even better, used a second x server with hardware accel somehow but offscreen rendering
<clever>
Miyu-chan: if you go back to the linux box, you can just alt+tab to chrome, and then the in-home streaming shares youtube! :P
<ashkitten>
that'd be fantastic
<Miyu-chan>
YES
<Miyu-chan>
A second X server would be amazing.
<ashkitten>
are there any x servers with hardware accel and offscreen rendering simultaneously?
<clever>
Miyu-chan: i was thinking, how would i setup a vm like xen, with windows, and hw accel, but no actuall hw output (2nd card maybe?)
<clever>
Miyu-chan: then in-home stream it to linux, so it becomes a proper window
<Miyu-chan>
Hm.
<Miyu-chan>
I'm not sure how this works, but my dGPU is connected to one monitor, and my iGPU is connected to another.
<Miyu-chan>
The iGPU monitor still uses the dGPU for render/compute.
<Miyu-chan>
Just a thought, but maybe a virtual monitor from your hypervisor suffices?
<Miyu-chan>
I really thought that my computer would just crash if an OpenGL application croses monitor boundaries, but that's not what happened. And I don't know enough about graphics architecture to make an educated comment about it.
<clever>
ajs124: ive seen something similar before (years ago), that would basically report the coords of a window inside the guest
<clever>
ajs124: and then the vm software would crop that section out, and render it to its own host window
<clever>
so the guest apps appear to have their own host windows