gchristensen changed the topic of #nixos-dev to: NixOS Development (#nixos for questions) | https://hydra.nixos.org/jobset/nixos/trunk-combined https://channels.nix.gsc.io/graph.html | 18.09 release managers: vcunat and samueldr | https://logs.nix.samueldr.com/nixos-dev
drakonis__ has joined #nixos-dev
lassulus_ has joined #nixos-dev
lassulus has quit [Ping timeout: 252 seconds]
lassulus_ is now known as lassulus
drakonis__ has quit [Remote host closed the connection]
fpletz has quit [Ping timeout: 260 seconds]
orivej has joined #nixos-dev
sir_guy_carleton has quit [Quit: WeeChat 2.0]
goibhniu has joined #nixos-dev
orivej has quit [Ping timeout: 246 seconds]
orivej has joined #nixos-dev
<Enzime> did anyone notice/fix this?
rsa has joined #nixos-dev
jtojnar has quit [Remote host closed the connection]
jtojnar has joined #nixos-dev
init_6 has joined #nixos-dev
orivej has quit [Ping timeout: 260 seconds]
orivej has joined #nixos-dev
orivej has quit [Ping timeout: 245 seconds]
init_6 has quit [Ping timeout: 260 seconds]
orivej has joined #nixos-dev
drakonis_ has joined #nixos-dev
nwspk has quit [Quit: Quit: *.banana *.split]
<nbp> I really do not see why we need a `builtins.fromTOML` function as part of Nix, when it can be implemented in Nix. :P
<gchristensen> I really do not see why we need Nix when we can just use assembly
<gchristensen> :)
<nbp> I totally agree!
<nbp> We just need a side-effect-free implementation of Assembly.
<ekleog> easy: include the world in a monad
<nbp> and be doomed to encode your understanding of Monad with non-collapsed wave function.
<ekleog> a non-collapsed wave function is just a burrito in the category of burritos
<nbp> How large is the category of burritos, I bet this is not a finite set.
<nbp> "Buritto" would have been a great particle name for some heavyweight particle, I guess.
nwspk has joined #nixos-dev
<infinisil> Maybe there should be a parser-combinator library in Nix, hmm..
thoughtpolice has joined #nixos-dev
<niksnut> nbp: Nix is not a general-purpose language
<niksnut> you shouldn't write parsers in Nix
<simpson> nbp: If a category still fits in sets, even transfinite ones, we generally consider it to be "small". The first non-small category is then Cat, which cannot possibly store its objects in anything smaller than a proper class.
<simpson> niksnut: Could there be a brighter line on what kinds of things are inappropriate to do in Nix? Stuff like nix-kubernetes is basically using Nix as a nicer-than-jq JSON generator, and I'm not sure whether this is a road I want to go down.
<ekleog> OTOH, sometimes there just is no choice but IFD, and IFD is… not really well-supported, afaiu
<niksnut> simpson: obviously people are free to do whatever they want, but if somebody wants to generate some JSON there are more light-weight ways to do that
<nbp> niksnut: I agree, we should not write parsers, but we should have enough features to show-case the usefulness of having a specific parser.
<thoughtpolice> simpson: I think the JSON-style usage is more like a "happy feature" that just happens to work out well. Like, if you're already using Nix to control "periphery" stuff it's probably good. Occasionally some things want configs written in JSON (don't ask me _why_) for example, so if you're already using Nix to package it/a systemd service, using Nix to generate the options is spiffy.
<thoughtpolice> simpson: Some people are also using Dhall for stuff like this too, if you're more a type-minded person. You can also use that to replace even parts of Nix, although then you have to go down the IFD route/precompile step. I think that's maybe the main disadvantage?
<thoughtpolice> Well, and also, you might not like Dhall. Can't say anything about that, haven't used it.
<simpson> Well, it's more that I've already spent the time learning jq. jq's only weakness here is that templating is kind of a metafeature, in that I'd get to invent my own "$syntax$" for template holes, but otherwise it's much stronger and more direct than Nix.
<thoughtpolice> And actually, JSON is just terrible to hand author for a configuration. If I wrote something and say, it was a service, and it *only* used/supported Nix, and I only committed to something like Nix/NixOS modules for creating those JSON configurations -- it wouldn't be that bad. Because I'd write Nix, not JSON. Sort of like a compile step.
<thoughtpolice> I mean, it's still sort of a weird choice. You also lose the module system enforcing some things. But that's actually not a very unrealistic scenario either. Tons of people deploy private NixOS-only services, of course (where you often don't have to worry about multi-use modules, option merging issues, etc)
<simpson> Sure. It's important to remember that modularity only exists within a given Nix incantation, and that the system being modified, whether Linux+systemd or k8s or etc. usually isn't modular in the same way.
<thoughtpolice> simpson: jq is definitely more robust in a lot of situations, yeah. Nix feels sort of square peg/round hole here but it's not too bad. I'd say the amount of times I use toJSON for example is otherwise pretty limited.
<thoughtpolice> There was one use case I had with a NixOS deployment recently, doing a sort of cluster deployment of a database, can't remember what it was...
<aminechikhaoui> niksnut: any chance you can get Nix 2.1 to 18.03 release ?
<aminechikhaoui> or is it supposed to only be available in the next release
orivej has quit [Ping timeout: 240 seconds]
<niksnut> aminechikhaoui: didn't have plans to do that
<aminechikhaoui> niksnut: I could test an 18.03 image with 2.1 to see if it fixes the out of memory issues
<aminechikhaoui> if the test turns out to fix the issue, is it possible to update it in 18.03 ?
{`-`} has joined #nixos-dev
<niksnut> aminechikhaoui: try nixUnstable, I think it's close to 2.1
<niksnut> it should have all the memory fixes
<samueldr> niksnut: what do you think about adding nix_2_1 (or properly named) attribute, and nixUnstable pointing to it on 18.03? this would make nixUnstable point to a stable version at EOL while not regressing to an earlier version
<niksnut> sure
<LnL> niksnut: sorry to ping you, have you had a chance to look at my nix doctor pr? it's still a bit rough but I'd love to hear your feedback
orivej has joined #nixos-dev
phreedom has quit [Remote host closed the connection]
phreedom has joined #nixos-dev
<elvishjerricco> Would there be any downside in having `nixos-rebuild` run `switch-to-configuration` under a `systemd-run -t --wait`?
<elvishjerricco> This way if the terminal closes or something, the activation won't get cut short
<niksnut> elvishjerricco: I think it already ignore SIGHUP
<elvishjerricco> Hm... I seem to remember some people having issues with this... I'll try to find an example
<{^_^}> #39118 (by nagisa, 19 weeks ago, open): `nixos-rebuild switch --upgrade` over ssh may put system into an unusable state
<domenkozar> any news for nixos weekly?
<domenkozar> as usual, it goes out tomorrow: https://github.com/NixOS/nixos-weekly/pull/63
<{^_^}> nixos-weekly#63 (by domenkozar, 1 week ago, open): Call for Content: 2018/08
<elvishjerricco> An additional benefit of doing a `systemd-run` would be that it would prompt users for administrative access if they forget to run with sudo.
<niksnut> wouldn't nixos-rebuild fail before that?
<elvishjerricco> Would it? I only did a brief scan, but I don't think I saw anything that requires root privileges.
<niksnut> installing in the system profile
<elvishjerricco> Hm for some reason I thought that was handled in switch-to-configuration...
<elvishjerricco> Ah, I see it now. Yea that would fail
phreedom_ has joined #nixos-dev
phreedom has quit [Ping timeout: 256 seconds]
phreedom_ has quit [Remote host closed the connection]
phreedom has joined #nixos-dev
<gchristensen> samueldr: https://github.com/NixOS/nixpkgs/pull/45400 this should be addressed (either merged before release, or my original change reverted) before stabilization
<{^_^}> #45400 (by dtzWill, 2 weeks ago, open): NIX_PATH: don't prepend $HOME-based value in session variable, set later
<niksnut> LnL: nix doctor looks good to me
<samueldr> gchristensen: noted
<samueldr> is this what accounts to the weird sudo needs in the past months on unstable?
<gchristensen> not sure?
<samueldr> sudo -i nixos-rebuild switch instead of sudo nixos-rebuild switch (going from memory, didn't look)
<samueldr> I may be conflating two issues, will look after work, nix and web don't match too well :)
<gchristensen> shouldn't make a difference unless they were trying to use their user channel to control the system nixpkgs. which is an interesting question: will this cause unintended side effects of people having confused nixpkgs revisions? maybe we should revert the whole thing.
<samueldr> I don't even know whether what I said is right (right now)
<LnL> niksnut: cool, I'll clean it up a bit then
<LnL> any ideas on the profile liveness check? except for the more naive approach used by upgrade-nix
phreedom_ has joined #nixos-dev
phreedom has quit [Ping timeout: 256 seconds]
<niksnut> LnL: how slow is it?
<LnL> maybe something else is going on
<LnL> on my desktop it's under .5s, but my (slower) laptop needs 3 minutes
<LnL> could linux vs darwin matter?
goibhniu has quit [Ping timeout: 244 seconds]
drakonis_ has quit [Remote host closed the connection]
<domenkozar> huh
<domenkozar> s3 down?
<gchristensen> twitter isn't blowing up so I'd think no?
<domenkozar> seems like some DNS issues
<domenkozar> could be around me :)
<gchristensen> (blowing up -> lots of tweets)
<domenkozar> :D
<srhb> Is there any problem with letting https://github.com/NixOS/nixpkgs/pull/46031 go straight to master first? It shouldn't cause rebuilds, only suddenly succeeding builds, afaik.
<{^_^}> #46031 (by srhb, 13 hours ago, open): go_1_11: Allow only one of GOARM={5,6,7}
<srhb> (And then release-18.09)
<gchristensen> dont see why not
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev