gchristensen changed the topic of #nixos-chat to: NixOS but much less topical || https://logs.nix.samueldr.com/nixos-chat
cosimone has quit [Ping timeout: 268 seconds]
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-chat
supersandro2000 has quit [Read error: Connection reset by peer]
supersandro2000 has joined #nixos-chat
supersandro2000 has quit [Client Quit]
supersandro2000 has joined #nixos-chat
spudly has joined #nixos-chat
spudly has quit [Changing host]
<colemickens> Emantor: I'd prefer a matrix room but I'd certainly join. I don't know why I made a room for my tiny Nix project that no one knows about but not nixpkgs-wayland.
<colemickens> actually, I had made a nixpkgs-wayland matrix room, was playing with sending CI status there, but I need a good matrix cli app for that first.
<colemickens> anyway, we can bridge an irc room
cole-h has joined #nixos-chat
<cole-h> First attempt is a failure, thanks to some flakes weird stuff. Apparently `--store` doesn't want to work :(
<colemickens> isn't it just a `nixos-install --system` ?
<cole-h> Is it?
<colemickens> ya
<cole-h> I used `--flake .#name`
<colemickens> mount to /mnt, build the toplevel with `nix build`, run nixos-install --system /path/to/the/built/toplevel
<cole-h> which was throwing errors about the storepath
<colemickens> with nixos-install? sounds like a bug or something mayhaps?
<colemickens> flakes doesn't really come into the install process though
<cole-h> ".... not in store" and stuff
<cole-h> probably a nixUnstable bug somewhere
<colemickens> there's been some lately. there was a week I was having to bounce to nix master to get something to evaluate
<cole-h> Welp, guess time to try again. Thanks colemickens. I'll report back soon (hopefully)
cole-h has quit [Quit: Goodbye]
<colemickens> good luck
tilpner_ has joined #nixos-chat
tilpner has quit [Ping timeout: 240 seconds]
tilpner_ is now known as tilpner
supersandro2000 has quit [Quit: The Lounge - https://thelounge.chat]
supersandro2000 has joined #nixos-chat
LnL has quit [Quit: exit 1]
cole-h has joined #nixos-chat
<cole-h> colemickens++
<{^_^}> colemickens's karma got increased to 52
<cole-h> `--system` worked wonders, thanks.
<cole-h> Took a little longer than expected cuz I needed to make a "bootstrap" profile for the first time cuz my flash drive was too small (ran out of space) lol...
ece has joined #nixos-chat
ece has quit [Read error: Connection reset by peer]
ece4 has joined #nixos-chat
cole-h_ has joined #nixos-chat
cole-h has quit [Ping timeout: 240 seconds]
cole-h has joined #nixos-chat
cole-h_ has quit [Ping timeout: 260 seconds]
<supersandro2000> the unstable word belongs into pname or version?
<cole-h> IMO, version. This has been a subject of debate, though.
supersandro2000 has quit [Quit: The Lounge - https://thelounge.chat]
<cole-h> ,tell supersandro2000 But apparently it should be in the pname -- https://github.com/NixOS/nixpkgs/issues/68518
<{^_^}> cole-h: I'll pass that on to supersandro2000
supersandro2000 has joined #nixos-chat
<samueldr> wrong channel, but it's defined here https://nixos.org/manual/nixpkgs/stable/#sec-package-naming
<samueldr> >> If a package is not a release but a commit from a repository, then the version part of the name must be the date of that (fetched) commit. The date must be in "YYYY-MM-DD" format. Also append "unstable" to the name - e.g., "pkgname-unstable-2014-09-23".
<samueldr> now, does it _belong_ there? dunno, but that's what the docs says :)
<cole-h> ^
ece has joined #nixos-chat
ece4 has quit [Read error: Connection reset by peer]
ece has quit [Read error: Connection reset by peer]
ece has joined #nixos-chat
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos-chat
<ashkitten> ugh i don't wanna report a wine bug
<ashkitten> bugzilla sux
waleee-cl has quit [Quit: Connection closed for inactivity]
ece has quit [Read error: Connection reset by peer]
ece has joined #nixos-chat
rajivr has joined #nixos-chat
<supersandro2000> Why did I not earlier use nix-shell to debug build failures easily with a changing environment but kept build progress
<{^_^}> supersandro2000: 1 hour, 18 minutes ago <cole-h> But apparently it should be in the pname -- https://github.com/NixOS/nixpkgs/issues/68518
<aterius> supersandro2000: I wish it was a bit easier to know which phases to run though
endformationage has quit [Quit: WeeChat 2.9]
<supersandro2000> aterius: buildPhase for cmake
<supersandro2000> could have saved me already hours today
<supersandro2000> Now I only need to configure it to run some shell commands in the opened nix-shell and then keep it interactive..
<aterius> supersandro2000: I've kinda memorized the respective order of the phases and now just try them out sequentially until something fails :P
<supersandro2000> Can I pipe commands to a shell with <<?
<samueldr> supersandro2000: I don't know if you skipped my message, but the docs state where it should be https://nixos.org/manual/nixpkgs/stable/#sec-package-naming
<samueldr> what you linked to is the opinion of a dev
<supersandro2000> yeah I did skip it..
<supersandro2000> so I still don't know the correct answer
<supersandro2000> thanks anyway
<cole-h> The correct answer is "whatever is in the manual", no?
<supersandro2000> the manual does not state if unstable belongs to pname or version
<cole-h> "If a package is not a release but a commit from a repository, then the version part of the name must be the date of that (fetched) commit. The date must be in "YYYY-MM-DD" format. Also append "unstable" to the name - e.g., "pkgname-unstable-2014-09-23". "
<cole-h> "Also append "unstable" to the name - e.g., "pkgname-unstable-2014-09-23""
<cole-h> (From the link samueldr posted)
<supersandro2000> but does this mean we do name = "pkgname-unstable" and version = "2014-09-23" or name = "pkgname and version = "unstable-2014-09-23"?
<supersandro2000> s/name/pname/g
<supersandro2000> pname = "pkgname-unstable" and version = "2014-09-23" or pname = "pkgname" and version = "unstable-2014-09-23"?
<cole-h> I read that section to mean that unstable should be part of the pname. ("append "unstable" to the name", where "the name" is the package name -> pname)
<supersandro2000> okay
<samueldr> 12.8 GiB [##########] /.git
<samueldr> gigabytes of history
<samueldr> (multiple remotes for kernel history from misc. OEMs and projects)
<lovesegfault> Just took some apple pie out of the oven
<lovesegfault> it smells so damn good
<samueldr> now 15.5 GiB with motorola's fork added :|
cole-h has quit [Quit: Goodbye]
<abathur> oh hmm
<abathur> git in your prompt? :)
<samueldr> nope
<samueldr> ~ $
<samueldr> that's my prompt in my home dir
<samueldr> and it's set to show up to three path components
<samueldr> with the return code of different from 0 (in red)
<samueldr> ~/.../linux/arch 1 $
<samueldr> and, as I'm not an animal, with a leading \n
<lovesegfault> Guess I'm an animal :P
<aterius> I just find it easier having a uniform start position for the line along the horizontal axis
<samueldr> it was obviously phrased that way because many find it unconventional
<abathur> samueldr: ah, was just curious how slow your prompt would be if so; we've discussed git prompt speed in nixpkgs over the past few days
<samueldr> yeah, I assumed :)
<abathur> how long does git status take on it?
<samueldr> → user: 0.39, system: 0.19, elapsed: 0:00.50
<samueldr> I don't have a "clean" git checkout of the kernel to compare with
<abathur> huh
<samueldr> I mean, without a bajillion zillion forks
<abathur> do you have a nixpkgs checkout?
<samueldr> $ git remote | wc -l
<samueldr> 36
<samueldr> much fewer than I thought, but then again I forgot I actually restarted from fresh on that computer
<samueldr> on nixpkgs: → user: 0.04, system: 0.16, elapsed: 0:00.15
<samueldr> that machine has 64GiB of RAM, things never get evicted from cache, and SSDs (not specially fast, just not bad)
<abathur> guessing that's mostly the ram then
<samueldr> probably helps a ton
<samueldr> I know that after a reboot I can feel it much more sluggish, which is not that sluggish, but that in-memory cache does wonders
<abathur> my nixpkgs on my desktop with an intel optane 900p is stil .1s with a warm cache
<abathur> ah, doh, I wasn't reading that very carefully, our times are actually fairly close now that I'm not assuming they're in the same order and paying attention :P
<abathur> ~ real 0.1s, user 0.04s, sys 0.14
<samueldr> yeah, my time is customized
<samueldr> TIME=\n → exit status: %x\n → user: %U, system: %S, elapsed: %E\n → cpu: %P, max mem: %M, avg mem: %t, unshared: %D\n → I/O: <I:%I O:%O>, sock: <r:%r s:%s>
<lovesegfault> here it is for me
<lovesegfault> using hyperfine to run it a bunch of times
<abathur> on my desktop hyperfine says ~99.5±1.8
<abathur> on my air it's more like 388.8 ms ± 10.4 ms :)
<lovesegfault> Time (mean ± σ): 191.0 ms ± 14.0 ms [User: 71.6 ms, System: 228.6 ms]
<lovesegfault> over 1000 runs
<lovesegfault> trying my home server
<lovesegfault> (this is nixpkgs)
<samueldr> Time (mean ± σ): 118.4 ms ± 1.7 ms [User: 46.8 ms, System: 124.2 ms]
<lovesegfault> ❯ nix run nixpkgs#hyperfine -- -w 100 -m 1000 "git status"
<lovesegfault> Time (mean ± σ): 59.5 ms ± 0.6 ms [User: 27.0 ms, System: 67.7 ms]
<lovesegfault> home server
<lovesegfault> This is with XFS and a PCIe 4.0 NVME SSD
<lovesegfault> my laptop times were encrypted ZFS on a PCIe 3.0 NVME SSD
<samueldr> ext4 fs on LUKS, probably not even the better settings for speed, I just used whatever it automatically chose
<samueldr> and SATA SSDs
<samueldr> because NVME wasn't really a thing when that old machine was made
Jackneill has quit [Ping timeout: 256 seconds]
Jackneill has joined #nixos-chat
raboof has quit [Ping timeout: 246 seconds]
steveeJ has quit [Ping timeout: 246 seconds]
raboof has joined #nixos-chat
steveeJ has joined #nixos-chat
LnL has joined #nixos-chat
<viric> MichaelRaskin: this is the plumbing thing I needed for the video https://github.com/aler9/rtsp-simple-server
<viric> I'm surprised 'tinc' recommends to have VPN interfaces and hardware interfaces have te same IP (but different mask) http://tinc-vpn.org/documentation/Example-configuration.html#Example-configuration
<sphalerite> viric: makes sense to me I think?
<viric> I think so
<viric> but I wouldn't expected it to be recommended
<lukegb> I mean, for the specific use (branch offices with a single flat global 10.0.0.0/8 space) it makes total sense
cosimone has joined #nixos-chat
__monty__ has joined #nixos-chat
<viric> I see I see
<srhb> Has anyone had problems with kernel > 5.4 and amdgpu / missing resolutions? I think I'm in for a bit of bisect hell...
<eyJhb> At least you can turn off your radiators if you have to do some bisecting srhb
<srhb> Don't think I've had them on at any point this year, not like it's winter yet :P
<eyJhb> srhb: Is there no wind where you live ?! :D It starting to get very could around here, so we have to turn up the radiators..
<srhb> A bit yesterday, but no, not very windy either.
<eyJhb> However, my GF seems to think it gets too hot when she cooks food, and will open every door in the apartment until it is freezing...
<eyJhb> Lucky :(
<srhb> Guess I'll have to try and build an early 5.5 kernel and hope that that doesn't display the problem. If even 5.5.0 has it, I'm not sure what to do. Diffing major branch changes sounds like no fun at all.
<ar> 5.4 doesn't work with my renoird laptop (or at least, 5.4 didn't work ootb, and >=5.9 works)
<ar> renoir*
<srhb> 5.4 works here, anything newer is broken (navi)
<MichaelRaskin> Including latest 5.4 patch?
<MichaelRaskin> (no ide where 5.4.x ends)
tilpner has quit [Remote host closed the connection]
tilpner has joined #nixos-chat
endformationage has joined #nixos-chat
Mic92 has quit [Quit: WeeChat 3.0]
Mic92 has joined #nixos-chat
<jtojnar> ugh, I have an old Minecraft acccount (before Mojang account or MS account was a thing) but the Win client only supports Mojang or MS account login and Migration from Minecraft account is disabled because they will migrate Mojang accounts to MS accounts in 2021 😠
<sphalerite> huh? That… doesn't make sense to me
waleee-cl has joined #nixos-chat
<jtojnar> but the windows MCJE launcher only has fields to log in with microsoft account and mojang account
<supersandro2000> jtojnar: MultiMC?
<supersandro2000> try logging in with mojang account and using your MC name
<jtojnar> supersandro2000: yeah, just figured it out
<jtojnar> that's what I get for trusting the form labels
<supersandro2000> 😂 never trust anything
<joepie91> especially not software
<supersandro2000> also when it complains about your mail being wrong try turning on JS
patagonicus has quit [Ping timeout: 264 seconds]
cole-h has joined #nixos-chat
cole-h has quit [Client Quit]
cole-h has joined #nixos-chat
<cole-h> Man, NixOS is the best.
<cole-h> Everything's reinstalled and basically exactly as I left it on my other disk
<bbigras> 😀
ece has quit [Ping timeout: 260 seconds]
cosimone has quit [Quit: cosimone]
ece has joined #nixos-chat
<drakonis> cole-h: haha, that's the magic
<drakonis> i have briefly considered switching to another distro just to check out how things are on the other side but nahhhh
<drakonis> the tradeoffs are too high
<MichaelRaskin> You could try Guix!
<drakonis> that's the one i want to try
<drakonis> guix is fun and games until you have to route around the policies
<drakonis> opening twitter videos leads to distorion
<drakonis> i wonder if that's dealt with
<drakonis> because the audio always sounds like it is skipping
<MichaelRaskin> I heard that pushing Guix for HPC lead to existence of unfree package repos
<joepie91> drakonis: you having pulseaudio issues?
<drakonis> on guix specifically
<drakonis> nixos is fine
<drakonis> nonguix has existed for some time now
<drakonis> but it didnt really have a lot of popular nonfree packages until this year
<drakonis> plus there's still no kde packages, but i can live without it
<drakonis> rather, there's no service for enabling it, the packages are in the repos
<drakonis> there are some pretty interesting things they've been working on these days
<infinisil> Interesting thought: You could theoretically have a planet that's entirely made of water. Just a blob of water in space, holding together by its own gravity
<infinisil> I'd expect so at least
<bbigras> Waterworld
<MichaelRaskin> I guess the question is, is it liquid and if yes,how the hell did it form
<drakonis> scifi has dealt with it
<drakonis> isnt waterworld basically a world where the icecaps melted?
<bbigras> yeah probably. there's jet skis and fuel and a tiny bit of dirt
<bbigras> and humans
ece1 has joined #nixos-chat
ece has quit [Ping timeout: 260 seconds]
ece1 is now known as ece
kalbasit has joined #nixos-chat
<drakonis> MichaelRaskin: https://gitlab.com/nonguix/nonguix this exists
<drakonis> there's even a nvidia driver now
<bbigras> "Please do NOT promote or refer to this repository on any official Guix communication channels." this is a bit silly
<viric> syncthing is one of those programs with a terrible UI. Specially when it stops synchronising and asks "Override changes?"
<viric> without any clue of what changes override what others
<joepie91> my biggest complaint about syncthing is that its model of error reporting is basically "Error: It's fucked"
<joepie91> can be a huge pain to figure out why two devices are not syncing
<bbigras> haha
<joepie91> spent half an hour two days ago on exactly this issue, before finding out - from one log line buried in many others - that it has a default "do not sync" threshold of 5% of your free disk space or some shit
<joepie91> it doesn't say "not enough disk space", it just... gets stuck on syncing
<bbigras> uh I thought I saw some errors about not having enough space before
<drakonis> bbigras: silly, yeah.
<joepie91> well I sure didn't! :D
<kraem> anyone running an "erase your darlings" setup and persisting /var/log ? can't seem to symlink the dir to a persisted location with tmpfiles.rules..
<bbigras> kraem: I think so
<drakonis> tghere are some cool as heck stuff that guix has been doing
<bbigras> do you use impermanence?
<bbigras> for some reason I disabled it on my desktop. 🤔
<kraem> i don't. should probably start though since i've been struggling with tmpfiles for this :p
srk has quit [Write error: Broken pipe]
srk has joined #nixos-chat
kalbasit_ has joined #nixos-chat
rajivr has quit [Quit: Connection closed for inactivity]
kalbasit has quit [Ping timeout: 256 seconds]
* colemickens is amazed people trust two way sync tools
<gchristensen> frightful
<cole-h> What would you suggest instead?
<cole-h> I use Syncthing to have access to stuff on my phone as well as my various desktops. If there's a better tool, I'm all ears.
kalbasit_ has quit [Ping timeout: 256 seconds]
kalbasit has joined #nixos-chat
<MichaelRaskin> I basically use rsync for stuff that is large and has a clear master copy location, and DVCS for the text
<cole-h> rsync is hard to do on mobile. Syncthing "just works" (no need to remember local IP, no need to set up forwarding so I can rsync from abroad, etc).
<gchristensen> for that sort of thing I'd just do it
<gchristensen> (syncthing)
* joepie91 also uses syncthing for that purposed
<joepie91> purpose*
<bbigras> syncthing is not bad. but the UX could be better when dealing with errors.
<gchristensen> one place I was at used https://csync.org/ to sync many TB of customer data between two fileservers in active-active mode
<bbigras> I wish syncthing had the option to use an encrypted node like resilio sync has.
<samueldr> been using syncthing, with different targeted synchronizations
<samueldr> e.g. a bunch of markdown files as notes
kalbasit has quit [Remote host closed the connection]
<samueldr> phone photos to a specific computer, where the computer is not actually syncing (not removing when it's removed on the phone)
<samueldr> but for files I work on, repositories and stuff, I don't use a synching tool like that; I use unison which is more of a manual tool
<samueldr> I don't want to deal with conflicts for those bigger projects, and no churn
<joepie91> bbigras: not that huge tbh
kalbasit has joined #nixos-chat
<joepie91> encryption is a complex topic that requires specialized knowledge to get right, and this would be a bunch of work to implement, also considering the UX aspects of it
<joepie91> $5k doesn't get you that far at commercial rates :P
da_dada has joined #nixos-chat
<samueldr> slap a token gnupg call somewhere in the synchronization
<samueldr> get bounty
<f0x> bbigras: could use borgmatic for that specific node
<bbigras> it's better than nothing. which is often the open souce way. but yeah it's a complex feature. maybe syncthing would need to be rewritten from scratch. who knows.
<lukegb> *slaps side of sync engine* this baby can hold so many crypto operations
<bbigras> hehe
<f0x> syncthing didn't work for me, horrendously slow even over local network, and messing up git repos :/
<f0x> so back to sshfs + manual rsync
<bbigras> f0x: does it support mobile? my usecase for the encrypted node is having my phone and my desktop in sync without the need for both of them to be online at the same time. using another node, that I don't want to trust.
<flokli> I fear I'll just wait for mobile nixos before trying to sync media and state from the phone
<flokli> It's all very sad and manual
<bbigras> for nixos as your phone os or nix on your phone? there's nix-on-droid
<MichaelRaskin> To be fair, Debian chroot with SSH running is good enough to run rsync
<MichaelRaskin> (As well as some Dynamic DNS script on a short command or once in 15 minutes or something)
<flokli> MichaelRaskin: I'm more concerned about "appdata"
<samueldr> flokli: AFAIUI syncthing is the "same" build on android as on other platforms
<samueldr> flokli: you can get into the same web interface if needed
<flokli> like, backing up messenger history etc
<flokli> which isn't really "files" anymore
<samueldr> flokli: but yeah, you can't really sync everything
<MichaelRaskin> Ah OK. Of course, if we are talking Debian chroot, you can just bind mount whatever inside it because root is needed anyway
<flokli> yeah. I mean, I don't even need dynamic DNS, but could just run wireguard
<bbigras> or tailscale
<flokli> well, that's also wireguard, isn't it? ;-)
<MichaelRaskin> Tailscale is a wireguard wrapper anyway, no?
<flokli> -ish
<bbigras> yes
<bbigras> but for lazy people
<bbigras> which I am
<flokli> :-)
<bbigras> well with tailscale you don't need a dynamic dns
<bbigras> that's nice
<flokli> I might just wait till samueldr fixes all the mobile nixos, and then run some android emulator on top, which I just snapshot entirely :-P
<samueldr> flokli: might I suggest YOU contribute TOO
<samueldr> :)
<lukegb> haha
<flokli> But I need my phone, and have no other hardware
<bbigras> yeah I could only install that on my old phone
<flokli> I considered getting a pinephone or so, but the whole "automotive modem draining all the battery" situation scared me
<samueldr> it's easy enough, just yesterday I went to my mail, picked-up a cheap phone I got off ebay, and in a couple of hours had it working¹!
<flokli> samueldr: does it crosscompile?
<samueldr> the Mobile NixOS specific parts, yes
* joepie91 spots a superscript
<samueldr> the NixOS system (stage-2 and up) as well as NixOS upstream
<joepie91> where's the fineprint :P
<flokli> sweet :-)
<samueldr> 1: working as well as other qualcomm devices of the same vintage
<joepie91> aha :)
<bbigras> I wonder if "normal" linux apps on the phone is a good idea? don't android apps have things like wakelock or whatever to use less battery?
<bbigras> well wakelocks might be the opposite
<samueldr> those things are part of the mainline kernel now
<flokli> I spent quite some time fixing various parts in nixpkgs to cross-compile my nixos christmas tree
<samueldr> so apps ecosystem could use those features
<bbigras> do they use it?
<samueldr> I actually don't know :)
<samueldr> spent almost all of my time still only up to stage-1
<bbigras> well it's a good thing if they could. there's hope at least. and they would probably accept PRs
<samueldr> though, mobile-nixos#3
<{^_^}> https://github.com/NixOS/mobile-nixos/issues/3 (by samueldr, 2 years ago, open): Add AArch64 QEMU target
<samueldr> oops
<samueldr> though, mobile-nixos#263 really was made in a few hours, including research
<{^_^}> https://github.com/NixOS/mobile-nixos/pull/263 (by samueldr, 8 hours ago, merged): motorola-surfna: init
<bbigras> Does it work on mako yet?
<bbigras> nexus 4
<samueldr> that surfna phone doesn't have custom ROMs or even a known working TWRP build
<samueldr> bbigras: armv7l, for now assume it won't be a good experience
<samueldr> (no cached builds)
<flokli> samueldr: hey, my christmas tree is armv6 :-P
<bbigras> oh yeah. thanks
<samueldr> flokli: yeah, but you don't run a GUI app in visual basic to trace it
<samueldr> there's also the issue of the vendor kernel being too old (IIRC), so you'll need a mainline kernel
<samueldr> (for mako)
<flokli> samueldr: it runs docker, does that count? :-P
<samueldr> flokli: no
<samueldr> nothing running docker counts
<samueldr> ;)
* flokli shakes fist
* lukegb runs podman instead
<lukegb> oh, and lxd apparently
<flokli> well, docker was an intermediate step. It's gone now ;-)
<lukegb> heheh
<bbigras> how is podman?
<flokli> turns out, glibc on raspberry pi os is OLD
<flokli> and if you have CGO stuff, you end up being dependent on the glibc of the host system
<flokli> so throw some file system namespace in between and you're good :-)
insep_ has left #nixos-chat ["User left"]
cole-h has quit [Quit: Goodbye]
cole-h has joined #nixos-chat
<bbigras> oh, on the syncthing issue, a recent comment says: "a solution for untrusted/encrypted devices is being tested and improved at the moment. A wider call for testing will likely happen in on of the next releases (and will also be announced in this issue)."
cosimone has joined #nixos-chat
<energizer> bbigras: if you're using syncthing on an untrusted device you must set the ui password, but the nixos module doesn't have a setting for that
<energizer> nixos really should have a setting for that imo
<bbigras> energizer: do you mean with the current syncthing or with the new untrusted devices feature?
<energizer> bbigras: with the current version
<energizer> but the nixos module doesn't have an option for it either way
<bbigras> energizer: is the ui password just to prevent people from editing the config? I'm guessing the data is not encrypted at rest unless you use disk encryption or something
<energizer> currently anybody on your machine can look at your files just by going to localhost:8384
<energizer> because the unix user security model doesn't apply to web interfaces
<bbigras> I didn't know we can look at the files with the web app but I see that anyone could add a sync to a remote host or folder. my usecase is more to use a 3rd party node that doesn't have access to the file's content.
<kraem> anyone know if i can create a blank zfs snapshot after i've populated the dataset with data already in an easy way? (otherwise i guess i'll have to create a new dataset and migrate everything over to that..)
<kraem> bbigras: thanks btw.. took the plunge and went for impermanence :)
<bbigras> kraem: 👍️
<bbigras> you're welcome
<cole-h> kraem: Pretty sure the answer is no.
<cole-h> It's why I always create a blank snapshot right after creating all my datasets lol. In case I ever wish to make the jump to impermanence ;)
<kraem> guessed so.. i chickened out and thought i wouldn't blank /home on boot :p
<eyJhb> kraem: Do it, do it, do it! :D
<energizer> kraem: impermanence is new software. be careful and make sure you have backups before doing anything dangerous with your filesystem.
<eyJhb> ^ but then afterwards
<eyJhb> Do it, do it :D
<eyJhb> It is really fucking nice...
<kraem> energizer: thanks. i got a new desktop which made me start cleaning up my nixconfigs, backup everything and wipe my laptop as well. so i feel safe (at least as safe as one can feel) :)
<kraem> eyJhb: yeah i'm gonna wait until tomorrow with doing it. my experience tells me i'll make a small mistake this late in the evening which means i'll be up all night tearing my hair off :D
<abathur> aw; resholve's first tweet-thread; they grow up so fast https://twitter.com/tgamblin/status/1343630959673950208
<gchristensen> nice
<gchristensen> oh the spackguy
<eyJhb> kraem: yeah, best not end up wiping it all by a mistake
<abathur> gchristensen: yeah, he's tentatively crediting you for mentioning it somewhere on irc
<Taneb> I've just impulse-bought myself a domain and now I'm thinking about the nix + pandoc static site generator I started writing a while back
<__monty__> Styx?
<Taneb> I've not seen that before. It looks... similar to what I want?
<Taneb> Ah, not exactly. I want to lean on Pandoc more and have more "magic"
<Taneb> (some of which I'm worried will require IFD)
<__monty__> Is IFD really so bad?
<gchristensen> "such a nice foot gun"
<samueldr> there's different view points about it, and different reasons to disallow it
<samueldr> one of them is build perfs, as it creates "stopping points" during the build reducing some of the parallelization that can happen
<Taneb> My alternative is to do a lot more lifting in nix itself, which feels like a bad idea
<Taneb> Or make my design slightly less magic
<Taneb> (probably going to end up doing IFD here)
__monty__ has quit [Quit: leaving]
<Gaelan> there's also a big difference from IFD from a download and IFD from the result of a build
<Gaelan> the former can be done up front, and the bulk of your build is "normal"
<gchristensen> one problem I have is I often like to have a fully distinct evaluation phase, copy the drvs somewhere eles, and then build
<Gaelan> what happens if you try to nix-instantiate something involving IFD?
<gchristensen> it tries to build
<gchristensen> if it can't, it fails
<Gaelan> if you just IFD from a file in a repo (say, to get a Cargo.lock), then your instantation phase does some downloads but it otherwise isn't too bad imo
<Gaelan> but yeah, IFD from an actual build would be nasty
<gchristensen> IFD inherently means builds of some sort
<Gaelan> and that still isn't "fully distinct"
<samueldr> if the machine you eval on cannot build it, it won't be able to produce nix from the cargo.lock file
<Gaelan> is fetchTarball architecture dependent? if so, eww
<samueldr> no because it's using FOD
<samueldr> fixed-output derivation
<samueldr> it is guaranteed that the (successful) outcome of fetching is a given hash
<Gaelan> so I guess what I'm trying to say is that IFD from a FOD isn't too bad
<gchristensen> builtins.fetchTarball isn't
<gchristensen> but pkgs.fetchurl is
<samueldr> yeah, forgot to distinguis about the few eval-time fetchers
<samueldr> distinguish*
<MichaelRaskin> We just need to properly separate eval system and build system like build and host and target are already separate
<samueldr> right, if we "just" need to do that, then I guess the PR's in the writing? ;)
<samueldr> (sorry, "just" is just such a loaded word)
<MichaelRaskin> Nah, too many _even clearer_ things got junked, why even bother
<energizer> ,ifd
<{^_^}> import-from-derivation (IFD) is when you evaluate nix from a derivation result, for example `import (pkgs.writeText "n" "1 + 1")` will evaluate to 2. This is sometimes problematic because it requires evaluating some, building some, and then evaluating the build result. It has been described as "such a nice footgun."
<Gaelan> Wouldn't separate eval system mean I couldn't reproduce derivations created on Linux on my mac?
<Gaelan> Currently, as long as you have a remote builder, the system where you eval doesn't matter. That's very nice.
<samueldr> you could, since there would be no way to evaluate from a build output
<samueldr> so things would be built with that in mind
<colemickens> Emantor: #nixpkgs-wayland is bridged to #nixpkgs-way
<colemickens> * Emantor: #nixpkgs-wayland is bridged to #nixpkgs-wayland:matrix.org
<colemickens> whoa kiwi irc looks different than I remember
da_dada has quit [Quit: WeeChat 2.9]
da_dada has joined #nixos-chat
bqv has quit [Quit: WeeChat 3.0]
bqv has joined #nixos-chat
<flokli> samueldr: bbigras:on the "does anyone use the wakelock stuff" situation: https://twitter.com/pid_eins/status/1334253187176468483
<samueldr> neat!
cosimone has quit [Quit: cosimone]