gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
<gchristensen> anyone want to bang their head against the wall of my AWS IAM permission problem? :P
<andi-> not my kind of kink, sorry
<gchristensen> lol
<gchristensen> me either :(
<andi-> I would buy a package of tiredness.
<gchristensen> hm?
<andi-> It is 2am and I am very awake :D
<gchristensen> oh
<gchristensen> then this is a perfect fit, it'll make you feel tired immediately
<andi-> Have you tried enabling anonymous write access in your policy? That should fix it, no?
<gchristensen> good idea
<gchristensen> and anonymous reads at the same time :)
<adisbladis> Good for privacy.. Or something
<adisbladis> Stay anonymous
<gchristensen> especially when its a nixops state file
<gchristensen> with all the goodies that means
<samueldr> :| that strace file is now 24GiB and I don't know if that eval is about to be done
<andi-> hu, what are you trying?
<samueldr> I guess the two previous failures were quite lucky to resolve in under an hour
<{^_^}> #83647 (by FRidh, 17 hours ago, open): Evaluation error is blocking jobsets
<{^_^}> hydra#728 (by samueldr, 1 day ago, open): hydra-eval-jobs masking real errors
<samueldr> trying to get the current eval issue that's being munged by hydra
<andi-> nice
<samueldr> at least I'm not actively doing something, it's passive
<cole-h> infinisil: Is there a place where all the ,commands of your nixbot are documented?
<infinisil> cole-h:
<cole-h> Dank
<cole-h> e
<gchristensen> Erik Satie? https://www.twitch.tv/grahamchristensen
<andi-> <3
<gchristensen> andi-: my dog is terrified of the rain, and satie calms her down lol
<andi-> Interesting. I never had a dog. No idea if that is common
<gchristensen> both our dogs did
<andi-> seeing that stream reminds me of trying to by a webcam... Looks as bad as last week
<cole-h> gchristensen: I'm lucky I checked into #-chat. Almost missed the piano again
<gchristensen> :)
<cole-h> Funnily enough, this music makes me think of rain
<samueldr> hmm, it's snowing quite a ton here
<gchristensen> :/
<colemickens> app idea - block all calls on Android, unless I've pushed a button that says "Allow dad to call in next 30 minutes". Force the calls to be pre-negotiated.
<cole-h> gchristensen: Thanks for the music :)
<gchristensen> :)
buckley3107 has joined #nixos-chat
buckley310 has quit [Ping timeout: 256 seconds]
lovesegfault has joined #nixos-chat
<gchristensen> !!! it worked !!!
<samueldr> you seem surprised
<gchristensen> it took me a day and a half to realize I should be setting the variable AWS_SESSION_TOKEN, so yeah I'm a bit surprised I figured it out
<samueldr> :/ I think I started that nix gc at ~20:26
<samueldr> I'm also a bit confused, it may have been something else, but I thought I collected everything yesterday, but somehow my root partition, which mostly only holds /nix/store got filled with about 800GiB of data using the hydra evaluator... is it doing more than instantiating the .drvs ?
<samueldr> (a bit too off-topic for -chat)
<samueldr> 110022 store paths deleted, 727497.28 MiB freed
<gchristensen> wow
<jackdk> is there a high score table for nix gc?
<samueldr> [12:41:23] <samueldr> 224278 store paths deleted, 796274.36 MiB freed
<samueldr> ok, so yeah, I did GC earlier
* samueldr is confused
<samueldr> where did that all come from when today I did nothing involving nix
<samueldr> (except evaluating using the hydra evaluator)
<samueldr> gchristensen: maybe relevant to your desire to run that evaluator on ofborg
<samueldr> aw, I had paused (^Z) the eval that wasn't finished but it ended up collecting the nixpkgs store path it was evaluating
<gchristensen> wow
* samueldr abandons
<samueldr> using strace to suss out the eval failure won't happen
<gchristensen> ouch :(
<samueldr> (btw that strace file was at 24GiB!)
<gchristensen> gross!
<samueldr> grosse indeed, but files are gendered masculine in french so it would be gros fichier :)
<drakonis> is this a store size measurement contest?
<samueldr> it wasn't supposed to be
<drakonis> there was someone with a 150GiB store in the telegram group
<samueldr> but I apparently accrued 727497.28 MiB in less than 12 hours
<drakonis> someone else with somehow 250GiB
<drakonis> i'm not sure how they managed these feats
<samueldr> mine gets filled quickly with disk image builds, sometimes going a couple GiB at a time
<drakonis> now, that's productive, those other folks aren't devs as far as i'm concerned
<drakonis> that was a joke mind you
waleee-cl has quit [Quit: Connection closed for inactivity]
<samueldr> hmmm, could it be that all the source files were being downloaded, that the hydra evaluation filled >700GiB of nix store?
slack1256 has quit [Read error: Connection reset by peer]
KeiraT has quit [Write error: Connection reset by peer]
endformationage has quit [Quit: WeeChat 2.6]
KeiraT has joined #nixos-chat
drakonis has quit [Quit: WeeChat 2.7.1]
lovesegfault has quit [Ping timeout: 252 seconds]
cole-h has quit [Quit: Goodbye]
MichaelRaskin has quit [Ping timeout: 264 seconds]
__monty__ has joined #nixos-chat
<tilpner> Oooh, time namespaces in 5.6 :o
<tilpner> Not a proper solution, but Nix sandbox support for could help workaround all those "embed date in $out" instances
<__monty__> It allows spoofing the time date(1) reports?
<tilpner> It's not limited to date(1), but that was probably just an example
<adisbladis> tilpner: I wonder if it would actually help though..
<adisbladis> I think not
MichaelRaskin has joined #nixos-chat
<tilpner> Why not?
<tilpner> It's not a proper solution, in the sense that it will do nothing for Darwin users and users of Linux machines with kernel versions <5.6
<tilpner> So it can't be relied on, and may mask unreproducability
<adisbladis> tilpner: Because it's not a fixed clock
<adisbladis> You're still gonna have varying build times
<adisbladis> Even if you set the start time of every build to the same thing they are not gonna reliably take the same amount of time
<adisbladis> I don't think it's useful at all for our purposes.
<tilpner> Oh no, you're right. I hadn't looked at them yet, just assumed it was the same API I had seen before
<tilpner> (But that might be memory corruption, I can't find them now)
<MichaelRaskin> Darwin already has completely different snadboxing, so why not improve things for NixOS
<MichaelRaskin> But yes, 5.6+ will take some time to become majority of systems
<adisbladis> I'm not saying we shouldn't use time namespaces :) I'm just not sure they bring any value.
<MichaelRaskin> But maybe fixing the start _date_ of the built is not making anything worse even if it rarely helps
<adisbladis> ++
<adisbladis> It probably doesn't hurt
<MichaelRaskin> Actually, having an ability to build in different virtual centuries might help with reproducibility testing…
<adisbladis> Pretty cool idea!
<tilpner> I wonder how many builds would still work if time was frozen at timestamp 1
<tilpner> Some build processes surely rely on a duration >0 somewhere
<MichaelRaskin> Well, they might still converge but very inefficiently…
<__monty__> /topic Brainstorm: Let's figure out weird obscure ways to break builds that are passing without problems rn
<tilpner> What more could you want, than deterministically broken builds?
<adisbladis> __monty__: :D
<adisbladis> Haha
<adisbladis> Broken 100% of the time, totally reproducible
<adisbladis> Reproducability nirvana
<MichaelRaskin> Like true nirvana, looks like death from the other side
<joepie91> wait, so what you're saying is that I can make the r13y percentage for nixpkgs shoot up by prepending an `exit 1` in the default builder?
<joepie91> cc gchristensen, we've found a solution :P
<MichaelRaskin> Well, you might have some problems with percentage of reproducible packages as a part of the total
<MichaelRaskin> NaN can be nasty to handle correctly
<MichaelRaskin> * as a part ot the total succesful builds
<gchristensen> yay
MichaelRaskin has quit [Read error: No route to host]
nix-build has joined #nixos-chat
{^_^} has quit [Read error: Connection reset by peer]
bridge[evilred] has quit [Ping timeout: 240 seconds]
yorick has quit [Ping timeout: 240 seconds]
thefloweringash has quit [Ping timeout: 240 seconds]
cocreature has quit [Ping timeout: 240 seconds]
talyz has quit [Ping timeout: 240 seconds]
waleee-cl has joined #nixos-chat
talyz has joined #nixos-chat
cocreature has joined #nixos-chat
bridge[evilred] has joined #nixos-chat
yorick has joined #nixos-chat
tokudan has quit [Ping timeout: 240 seconds]
thefloweringash has joined #nixos-chat
atopuzov[m] has left #nixos-chat ["Kicked by @appservice-irc:matrix.org : Idle for 30+ days"]
ldlework has left #nixos-chat ["WeeChat 2.5"]
julm has joined #nixos-chat
avn has quit [Ping timeout: 260 seconds]
avn has joined #nixos-chat
edef has quit [Ping timeout: 252 seconds]
Irenes[m] has quit [Ping timeout: 240 seconds]
worldofpeace has quit [Ping timeout: 240 seconds]
yegortimoshenko has quit [Ping timeout: 240 seconds]
colemickens has quit [Ping timeout: 240 seconds]
hl has joined #nixos-chat
Irenes[m] has joined #nixos-chat
colemickens has joined #nixos-chat
worldofpeace has joined #nixos-chat
MichaelRaskin has joined #nixos-chat
yegortimoshenko has joined #nixos-chat
<aanderse> gchristensen: do i recall correctly when you once talked about using your computer with dictation software instead of a mouse/keyboard for some activities?
ar has quit [Ping timeout: 260 seconds]
<gchristensen> I tried it a while, but it was not good :?
cole-h has joined #nixos-chat
edef has joined #nixos-chat
ar has joined #nixos-chat
savanni has quit [*.net *.split]
tokudan[m] has quit [*.net *.split]
sphalerite has quit [*.net *.split]
savanni has joined #nixos-chat
tokudan[m] has joined #nixos-chat
sphalerite has joined #nixos-chat
tokudan[m] has quit [Ping timeout: 240 seconds]
yegortimoshenko has quit [Ping timeout: 252 seconds]
Irenes[m] has quit [Ping timeout: 252 seconds]
aanderse has quit [Ping timeout: 246 seconds]
jtojnar has quit [Ping timeout: 246 seconds]
emily has quit [Ping timeout: 246 seconds]
colemickens has quit [Ping timeout: 240 seconds]
worldofpeace has quit [Ping timeout: 252 seconds]
thefloweringash has quit [Ping timeout: 256 seconds]
dtz has quit [Ping timeout: 260 seconds]
arcnmx has quit [Ping timeout: 260 seconds]
vaibhavsagar has quit [Ping timeout: 260 seconds]
leons has quit [Ping timeout: 260 seconds]
rycee has quit [Ping timeout: 260 seconds]
<pie_[bnc]> aszlig: now that you brought it to my attention i will be putting more notes in commit messages
<pie_[bnc]> or at least they should refer to a PR #
<pie_[bnc]> but i havent actually done much PRwise so maybe thats why i didnt think about it :p
<aszlig> pie_[bnc]: i'd actually use the full URL to the PR and only if it's not _that_ relevant
<aszlig> eg, i've seen lots of commits with messages like "See #12345"
<nix-build> https://github.com/NixOS/nixpkgs/issues/12345 (by davidak, 4 years ago, closed): HiDPI support
<aszlig> like literally just the ID without any explanation whatsoever
<pie_[bnc]> i probably would have done that
<aszlig> this has a few issues: first of all, it introduces an additional indirection where people need to resort to a web page, second it's even worse if you're offline or with bad connectivity and lastly: if we one day move away from github we suddenly lose a lot of context for the commit messages in question
<aszlig> so IMHO i think the PR/issue ID/URL is okay as an _optional_ part as long as there is enough context in the commit message
drakonis has joined #nixos-chat
<qyliss> Quote liberally from web pages
<qyliss> It should be possible to get context from just a commit message
<qyliss> PR/issue numbers are a nice bonus but shouldn't be required for understanding
<joepie91> +1
<__monty__> In the hypothetical scenario where issues are migrated I hope issue numbers will be conserved. Not doing so would be horrifying.
<pie_[bnc]> heh
<aanderse> Ok thanks
aanderse has joined #nixos-chat
<cole-h> ,ping
<nix-build> pong
<cole-h> Bot got renamed 🤔
<drakonis> ahhh bannerlord is out
<MichaelRaskin> __monty__: I believe nobody cares as long as there is an available and clonable archive of all the text of discussions by-old-number
<cole-h> Just think about all the commits that directly reference by #, not url
<cole-h> "Is that a GitHub number? Or a GitLab number? Or a BitBucket number?"
<MichaelRaskin> Meh. year-to-system list would take a century to become longer than a page
mrCyborg has joined #nixos-chat
<cole-h> https://youtu.be/Eh-2Lk87Gms Factorio meme
sphalerite has quit [Ping timeout: 240 seconds]
jtojnar has joined #nixos-chat
NinjaTrappeur has quit [Ping timeout: 272 seconds]
NinjaTrappeur has joined #nixos-chat
<infinisil> I can't remember last time I got a non-callcenter call on my phone..
<infinisil> Maybe whitelisting callers could be good
<gchristensen> my strategy is to have a non-local phone number, so if somebody local does call it is obviously a person
<cole-h> That's actually really smart
* cole-h moves to another area code
<infinisil> gchristensen: Not getting that
<infinisil> You mean you can see on the number prefix whether it's local or not?
<gchristensen> phone numbers here start with a local prefix. so say it started with 512-...-.... that would mean it was issued in Austin, TX, but I live in Berkshire County, MA whose phone numbers start with 413-...-....
<ashkitten> augh
<gchristensen> if someone calls with a 413-...-.... number, I can be quite confident they are actually a person I want to talk to
<infinisil> Ah I see
<ashkitten> did you know, restarting services on a local machine doesn't work if you put the commands into an ssh session?
<tilpner> They can, if you ssh back to your local machine!
<ashkitten> point being, i accidentally just rebooted my desktop because i thought systemd was messing up on my laptop
<ashkitten> at least i didn't reboot my server, i guess?
<tilpner> I did that the other way around once, wanted to restart the server and suddenly my laptop shut down
<ashkitten> lol
<tilpner> The solution would be color-coded terminals, but I still haven't gotten around to that
<MichaelRaskin> On my laptop, the easiest way to reboot requires physical presence confirmation
<cransom> i know if someone calls from $localareacode it probably is a person, but that still doesn't mean it someones i really want to talk to.
<MichaelRaskin> UID does not matter, password does not matter, physical access does
<samueldr> how do you assert physical presence?
<samueldr> (I see many ways, but still curious)
mrCyborg has quit [Ping timeout: 240 seconds]
<MichaelRaskin> VT switch
<ashkitten> switch to tty, ctrl+alt+del
dtz has joined #nixos-chat
rycee has joined #nixos-chat
Irenes[m] has joined #nixos-chat
colemickens has joined #nixos-chat
thefloweringash has joined #nixos-chat
worldofpeace has joined #nixos-chat
vaibhavsagar has joined #nixos-chat
emily has joined #nixos-chat
tokudan[m] has joined #nixos-chat
arcnmx has joined #nixos-chat
leons has joined #nixos-chat
<samueldr> ah, I thought you meant the `reboot` command didn't go through until you asserted presence, e.g. pressing the power button once
yegortimoshenko has joined #nixos-chat
<MichaelRaskin> Well, the VT switch for asking a confirmation is automated
<samueldr> ah, so it's that
<MichaelRaskin> reboot command does nothing, but the similar reboot request switches and asks the user whether reboot is intended
<MichaelRaskin> Same for brightness or network reconfiguration
<samueldr> yep, sounds about right
<samueldr> or I should say, desirable
<MichaelRaskin> Somehow it is not how things are normally done for presence-expected systems, but oh well
<MichaelRaskin> I was breaking all conventions anyway by that moment
<samueldr> what is normally done?
<MichaelRaskin> Well, there is sudo, sometimes taking into account which users have logged in via a local session
<MichaelRaskin> Of course, if a user is logged in both via a console and via SSH these checks are not guaranteed to make sense
<MichaelRaskin> Re: colour-coded backgrounds — it is easy to set up, but doesn't work for everyone. I failed to learn to notice the colours
<ashkitten> i just answer phone calls no matter what, idgaf about spammers/scammers
<ashkitten> re: gchristensen's technique
<gchristensen> impressive, how many spam calls do you get?
<ashkitten> like, i give out my phone number to friends in case they need to reach me, and half the time i don't actually have their phone numbers
<ashkitten> ehh a few a week
<drakonis> lol i get spammers every day
<gchristensen> oh
<drakonis> except during the quarantine because nobody's on the callcenters
<gchristensen> yeah we're at, like, an order of magnitude difference
<ashkitten> if i hear "social security" i hang up
<MichaelRaskin> Yeah, ten a day would be annoying
<drakonis> i get robocalls that shut down in 3 seconds
<MichaelRaskin> ashkitten: don't hang up! just put the phone on the table
<drakonis> imagine getting two dozen calls a day
<ashkitten> protip: the social security administration doesnt use phones, unfortunately
<gchristensen> I very much enjoy messing with fake student loan companies...
<ashkitten> can i speak to the head of the household?
<MichaelRaskin> «we are hiring first-line tech support, how much are you currently making»
<ashkitten> oh that's one i haven't heard before. i only get super generic calls
<MichaelRaskin> No, that's supposed to be _the answer_
<ashkitten> sometimes i get one trying to sell me car insurance
<ashkitten> i don't always tell them i don't have a car
<MichaelRaskin> Because you already have the selection of people doing shit calls anyway!
<ashkitten> the one that annoys me the most is the ones trying to get me to donate to cops
<ashkitten> like, cops can go to hell
<ashkitten> i don't even know if they're scammers or legit
<ashkitten> well, they're scammers either way but
<cransom> at least 4-5 times a week, i have calls that may leave a voicemail in chinese/mandarin? from dhl about an international shipment i should contact them about.
<cransom> *purportedly dhl
<cransom> though my favorite is the vm my spouse has where they threaten 'to report you to the local cops'
<drakonis> lmao
<drakonis> gchristensen: can i PM you real quick?
<gchristensen> sure
<andi-> very excited to see nix dev-shell arrive in Nix (master). nix dev-shell --profile /my/projects/shell-gc-root <3
<nix-build> nix#3460 (by edolstra, 14 minutes ago, open): Backport 'nix dev-shell' from the flakes branch
<cole-h> andi-: Pardon my ignorance, but what is dev-shell's purpose?
<andi-> cole-h: "run a bash shell that provides the build environment of a derivation"
<andi-> basically like nix-shell
<cole-h> OK, I was thinking it was gonna be something revolutionary :P
<andi-> have you ever had a projects shell environment garbage collected while on a very low speed internet connection?
<andi-> (nobody leaves the house these days, but *before* ;))
<cole-h> haha
<cole-h> Before I was stuck at home, I wasn't doing much Nixing
<cole-h> So I can't say that I have ;^)
<cole-h> But I can see how that would be useful
<cole-h> Would that be in the case where you have auto-optimize enabled or something? Or, what would gc a shell environment without an explicit collect-garbage?
<andi-> well it is very common to have daily/regular scheduled GC on all boxes.
<andi-> I run GC every hour on my boxes
<MichaelRaskin> Ouch. I have builds taking multiple hours sometimes!
<joepie91> my GC is every midnight, but it doesn't nuke generations
<andi-> MichaelRaskin: that is what the temproots are for
<cole-h> andi-: Ah, okay, makes sense. Never run into that because I haven't switched to NixOS :P
<MichaelRaskin> andi-: but GC blocks switching to the next derivation
<andi-> I am not sure if there is a racecondition between building and the actual switching. It shouldn't be...
<MichaelRaskin> Not a race condition, but an annoying loss of time due to locking
<andi-> You could disable the GC while you are rebuilding by pausing the timer/cron/…
<__monty__> MichaelRaskin: It seems exceedingly silly to just throw the information in issue numbers away. It's quite easy to retain in many cases. Github -> Gitlab is what I'm most familiar with.
<MichaelRaskin> Throw information _completely_ — sure. But the effort threshold when it is worth maintaining identical IDs is pretty low. Nobody has time to look at the merged PRs one release cycle ago anyway.
<__monty__> I assume activity won't go down so there's thousands of issues and PRs in-flight when you migrate.
<MichaelRaskin> it would be nice if cross-references were converted during migration, but even that might be overrated.
<__monty__> Dumping the IDs to a file is almost the same amount of work as maintaining them in a migration. So I don't get the drama.
<MichaelRaskin> wget once, ignore what's lost
<MichaelRaskin> We use git which drops the single most useful piece of information every push, we don't care about history being very usable
<cole-h> Curious what this "single most useful piece of information" is? I'm not a git wizard
lovesegfault has joined #nixos-chat
<MichaelRaskin> Hmm, probably not every push, probably merge+push
<MichaelRaskin> The initial branch the commit was targeted at — this often contains more meaningful information than the commit message…
<cole-h> Got it
<cole-h> So e.g. backports to release-20.03 lose that context
<MichaelRaskin> They actually often do not
<MichaelRaskin> because they are marked as cherry-picks with a commit reference
<MichaelRaskin> But merges to staging and to master are annoying to distinguish in old history
<cole-h> Right, that would have been a better example
ldlework has joined #nixos-chat
<pie_[bnc]> MichaelRaskin: so I looked at my display-manager unit file and its actually pretty bare, basically just calls a script that runs /run/current-system/sw/bin/sddm
<pie_[bnc]> so I guess all the magic is in config files?
<pie_[bnc]> theres some env vars, but i havent looked at sddm.nix yet
<pie_[bnc]> so for X11 I think it was basically just stuff in /run/current-system/share/X11
<pie_[bnc]> and /etc/X11
<MichaelRaskin> No idea where sddm configs go. Never used it
<pie_[bnc]> yeah sure thats fine
<pie_[bnc]> i was just saying
<pie_[bnc]> so where does the whole systemd mess come in? I guess I saw mentions of polkit and logind?
<pie_[bnc]> but also suggestions that once I have a TTY or something stuff should run fine from there
<pie_[bnc]> (modulo X11 configs)
<pie_[bnc]> so I guess another question I really want to ask is what your problem was? I think you said the X11 situation was why you finall decided to switch away
<MichaelRaskin> In an actual system, running Xorg as root is enough for it to grab a fresh console
<MichaelRaskin> With systemd… no
<pie_[bnc]> hm, but i think I was able to run startx "fine" from the tty without root?
<MichaelRaskin> I think it was a later development that it was anywhere near working, and also it takes over the source VT which is also not exactly what I want
<pie_[bnc]> hm
<pie_[bnc]> OK I kind of guessed the former.
<pie_[bnc]> I will continue poking around.
evanjs has quit [Quit: ZNC 1.7.5 - https://znc.in]
evanjs has joined #nixos-chat
<pie_[bnc]> MichaelRaskin: well, this all may have been pointless given that I just realized SDDM allows me to start a second "session"
<pie_[bnc]> but i may as well finish up
<pie_[bnc]> meanwhile, I'm uncertain, do display managers call X11 or does X11 get called and you pass stuff to the display manager? because currently it's looking like the former to me... but I havent gotten far enough along to be sure
<pie_[bnc]> i mean all I see in the display-manager service exec script is a call to sddm, no mentions of xinit or whatever
endformationage has joined #nixos-chat
worldofpeace_ has joined #nixos-chat
__monty__ has quit [Quit: leaving]
worldofpeace_ has quit [Quit: worldofpeace_]
ekleog has quit [Quit: back soon]
ekleog has joined #nixos-chat
<MichaelRaskin> Telegram-in-Russia chronicle: there is a news item in Russia right now with the source being… official Telegram channel of the Moscow coordination center for handling the COVID-2019 situation. No, the corresponding channel entry can _not_ be found on the city government website. The same channel is actually mirrored as a VK.com public group (unlike Telegram channel, this page is visible without logging in), but news s
<joepie91> cut off after "but news s"
<MichaelRaskin> but news sites cite Telegram channel for some reason…
<MichaelRaskin> Hmm, I thought ii does and honest roundtrip
<MichaelRaskin> Silly me
<joepie91> hey, https://howoldis.herokuapp.com/ is broken
<MichaelRaskin> Yes
<MichaelRaskin> status.nixos.org seems to be the thing to use
<joepie91> aha
<joepie91> maybe it should be linked from there :P
<joepie91> oh
<joepie91> that just links back to howoldis lol
<gchristensen> just for the information
<gchristensen> if you'd like to port the information, here is where that page's source is: https://github.com/NixOS/nixos-org-configurations/blob/master/delft/eris/status-page/index.html
<joepie91> I'll have to parse that when I'm not about to sleep :D
<gchristensen> up to you