tzemanovic has quit [Remote host closed the connection]
tzemanovic has joined #nixos
lnikkila has quit [Ping timeout: 245 seconds]
Intensity has quit [Ping timeout: 260 seconds]
tzemanovic has quit [Ping timeout: 248 seconds]
tzemanovic has joined #nixos
kreisys has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Leira has joined #nixos
<Leira>
I have nix installed on my machine, I tried to install kakoune, but `nix-env -qa kakoune` cannot find any thing. ~/.nix-defexpr/channels/nixpkgs/pkgs//applications/editors/kakoune shows platform as "platforms = platforms.unix", mac os should be part of it, right?
Intensity has joined #nixos
logzet has quit [Remote host closed the connection]
<rotaerk>
hmm any idea why my master audio in alsamixer is muted and min volume every time I boot?
<Leira>
is there anyway I can verify if my platform is considered as "platforms.unix", like through nix repl?
<gchristensen>
Leira: what about nix-env -iA nixpkgs.kakoune
JasonGrossman has quit [Remote host closed the connection]
JasonGrossman has joined #nixos
Synthetica has quit [Quit: Connection closed for inactivity]
blonkhart has joined #nixos
<Leira>
@gchristensen, it installed~! Maybe it was because of the name I searched, `nix-env -qa 'kakoune.*'` did give me kakoune-unstable-2018-05-21, so the package name is not kakoune, but kakoune-unstable
<gchristensen>
ayee
<gchristensen>
nix-env is frustrating I thihnk
<Leira>
hmmm, I need to keep in mind that it takes an RE to match, not a term to search
andreabedini has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
andreabedini has joined #nixos
Fare has joined #nixos
Supersonic has quit [Disconnected by services]
Supersonic112 has joined #nixos
Supersonic112 is now known as Supersonic
eqyiel has quit [Ping timeout: 248 seconds]
eqyiel has joined #nixos
lassulus_ has joined #nixos
smallville7123 has quit [Quit: rosa]
olto has joined #nixos
oltoAltn has quit [Ping timeout: 276 seconds]
lassulus has quit [Ping timeout: 240 seconds]
lassulus_ is now known as lassulus
<mog>
i am on unstable, and the linux testing kernel was upgraded 10 days ago
<mog>
but i dont see it when i do a nixos-rebuild switch --upgrade
oahong has quit [Ping timeout: 264 seconds]
oahong has joined #nixos
<shlevy>
Anyone here have calibre working? I'm seeing 'This application failed to start because it could not find or load the Qt platform plugin "xcb"' when I try to run it from a nix-shell
<shlevy>
Ah seems I need it in my systemPackages :(
<srhb>
steveeJ: One thing that would be excellent is better and more per-package testing, since that happens continually and is checked by ofborg.
<etu>
But cython on aarch64 won't end up in the binary cache even though I've excluded the failing test #41453 because it seems to take a *very* long time to run their tests on aarch64. Took 14.5 hours on my Raspberry and on Hydra it timed out after 10 hours: https://hydra.nixos.org/build/75636013
<srhb>
steveeJ: A lot of packages do have this, but it would be excellent if we could define something akin to NixOS tests more easily in the actual checkPhase, or post that. Maybe something containerized
<srhb>
etu: It was recently made possible to override the timeout, but that's probably not reasonable in this case.
<srhb>
See 537d14f4e2c8ef908641223bc80fe8e2bca74e90
woodson has joined #nixos
<etu>
srhb: wow, that's handy. But shouldn't really be needed :/
<etu>
Not for this
<srhb>
Agreed. But the builders are overloaded, too, and it's a compounding issue with really slow builds. :(
<srhb>
So most of the time is the tests. I don't know, I'd rule out timeouts in the test at least and then consider bumping meta.timeouts
<etu>
I only know that it takes ~1 hour to build cython on my raspberry and then 14.5 hours to run all the tests on that machine. But that's a raspberry.
<srhb>
It would be nice to try and run it on something faster than an RPi and see how long it takes
<srhb>
Maybe try asking gchristensen or one of the other people who have access to bigger aarch64 machines?
smallville7123 has quit [Remote host closed the connection]
<etu>
Maybe, but I'm off to lunch about now:)
<srhb>
Seeya
atu has joined #nixos
smallville7123 has joined #nixos
smallville7123 has quit [Remote host closed the connection]
blankhart has joined #nixos
asuryawanshi has quit [Remote host closed the connection]
smallville7123 has quit [Remote host closed the connection]
<steveeJ>
manveru: in theory you could register qemu-user for each respective architecture, and then run the foreign binaries directly. in practice this doesn't work with all the things
simukis has joined #nixos
smallville7123 has joined #nixos
ixxie has joined #nixos
smallville7123 has quit [Remote host closed the connection]
NightTrain has joined #nixos
<steveeJ>
srhb: I would not rely on a tool that isn't integrated with nixos which makes nixos-containers the only option I know :D
<fearlessKim[m]>
Myrl-saki: pro tip: nix repl '<nixpkgs>' '<nixpkgs/nixos>' and then you get your config too
<adisbladis[m]>
I meant cross compiling using nixpkgs cross compiling facilities
<Akii>
aha i did that wrong
<Akii>
thanks :)
<steveeJ>
adisbladis[m]: and then what? :D
<Myrl-saki>
Akii: Ping. `nix-build ... --no-out-link 2>&1` shoudl do what you want.
<fearlessKim[m]>
Akii: use the autocompletion, several openssl will popup
<adisbladis[m]>
steveeJ: And then you have whatever arch containers? Isn't that the goal?
<Akii>
I just ran `nix repl`without the <nixpkgs> part
<Myrl-saki>
Yeah, I don't think --arg is working.
<jluttine>
anyone using (spac)emacs? is it possible to have environments defined in shell.nix for projectiles? for instance, i'd like to define my python project environment in shell.nix and be able to run unit tests in emacs
<manveru>
jluttine: well, it's possible, but it's not easy...
<adisbladis[m]>
jluttine: I use emacs-direnv + nix + direnv for that
<jluttine>
:/
<Unode>
hi all, going through available pomodoro-like apps in nixpkgs. Are there any working alternatives to gnome3.pomodoro? I tried haskellPackages.pomodoro but it fails to install/compile.
<manveru>
i tried the direnv route, but for some reason it won't work with a lot of packages
<adisbladis[m]>
Not using projectile specifically but it works for most other modes
<adisbladis[m]>
manveru: Why not?
<manveru>
so i'm using nix-sandbox now and patch whatever package relies on executing stuff directly
<adisbladis[m]>
I do the same thing with some pretty insane projects
<Akii>
so nixos' openssl comes without a /etc/ssl/openssl.cnf ?
Neo---- has joined #nixos
nico202 has quit [Ping timeout: 260 seconds]
<manveru>
Akii: it's in /nix/store/3ag1mbdvkg6jx29dl5wnwypvckkfy3sv-openssl-1.0.2o/etc/ssl/openssl.cnf for me
<manveru>
but not linked into /etc
<Akii>
I have "/nix/store/89p9lf28cyc3dn3p180fv68a9nda97j1-openssl-1.1.0h-bin"
<{^_^}>
[nix] @AmineChikhaoui opened pull request #2215 → use debug instead of printTalkative to avoid spamming hydra queue-runner → https://git.io/vhzpY
smallville7123 has quit [Quit: rosa]
<{^_^}>
[nixpkgs] @woffs opened pull request #41674 → loadwatch: init at 1.1 → https://git.io/vhzpl
genesis has joined #nixos
<__monty__>
I have single user nix running on arch, I have nix-daemon running on mac os. Can I use the mac as a remote build host for the linux box? I don't want to initiate darwin builds on the linux host, I want to offload the building for the linux host to the mac.
selaux has quit [Ping timeout: 276 seconds]
<LnL>
not sure what you're asking, but you can't build linux stuff on darwin
<LoneTech>
a cross compiler should be capable of doing it.. but that doesn't mean nix has the infrastructure to build the same thing using cross compilers as native compilers
<LoneTech>
something like UML or some other virtualization to run a linux layer on the darwin host should work
<__monty__>
The original problem is the linux box runs out of memory trying to build ghc 8.2.2, which I need to rebuild because I need a new dependency.
fendor has joined #nixos
silver has joined #nixos
<LoneTech>
you need a new dependency for ghc? what for?
<__monty__>
Not a ghc dependency. I need a new haskell package as a dependency for my project.
<__monty__>
Nix always rebuilds the entire thing if I update my channel.
NightTrain has joined #nixos
jensens has joined #nixos
ris has joined #nixos
Jetien_ has quit [Ping timeout: 264 seconds]
<ris>
am I going mad or has the phantomjs2 currently in 18.03 not been linked with libQt5Core.so.5?
<LoneTech>
__monty__: https://nixos.wiki/wiki/Distributed_build has notes on remote building, I suspect the easy thing to do is use VirtualBox as there's a starting image of nixos for it
<goibhniu>
Akii: gah, sorry, I got distracted :D
<LoneTech>
ris: is that ldd output? if so, it indicates more that it's missing than that it wasn't linked
<LoneTech>
a nix build should have the full path to it though
<LoneTech>
iirc
iyzsong has joined #nixos
<ris>
right yes but i thought nix used rpath tricks to bind a binary's dependencies together
<ris>
so technically maybe not "linking"
<__monty__>
LoneTech: Thank you.
<ris>
but yknow
<__monty__>
How do I find out what the current default ghc version is? (Considering just using the default since that should have everything cached?)
<goibhniu>
Akii: FWIW, I have the plasma nm-applet working here. I thought you needed to add plasma-nm to your package groups but I don't seem to have that. All I've done is enable plasma and networkmanager.
<gchristensen>
__monty__: for the mac to build Linux stuff you need a Linux VM on the mac
hamishmack has joined #nixos
<goibhniu>
Akii: can you post your configuration.nix to a pastebin please?
<__monty__>
gchristensen: Yeah, looking into the page LoneTech linked. Could nix-darwin run a nixos container?
<gchristensen>
then you'd need to do the work to make that linux system inside mac accessible via your linux system
owickstrom has quit [Ping timeout: 264 seconds]
nh2 has quit [Quit: Leaving.]
<__monty__>
gchristensen: So still in a vm?
<manveru>
well, there's no way to do it without a vm, the question is just which vm you choose :)
<LnL>
yeah, that's what I was saying before
<LnL>
cross compiling is something completely different
Fare has quit [Ping timeout: 240 seconds]
<ryantm>
Do we have anything checking that the src derivations in nixpkgs build later? Do we have automated checks that would catch upstream removing a release after Hydra has built it?
logzet_ has joined #nixos
<gchristensen>
ryantm: I think copumpkin has something about that, but nothing formal
logzet has quit [Ping timeout: 276 seconds]
nh2 has joined #nixos
blankhart has quit [Ping timeout: 256 seconds]
<contrapumpkin>
yeah, I periodically check everything I use at work for validity
<contrapumpkin>
occasionally you'll see a commit or few from me or coworkers fixing up broken source links, but that's obviously not exhaustive
<Unode>
Hi all, trying to create an overlay to override mpv with a newer version and stumbling on newbie mistakes. How do I import fetchFromGitHub in http://dpaste.com/3N3JXZB ?
Neo--- has quit [Remote host closed the connection]
Neo--- has joined #nixos
ris has quit [Quit: Konversation terminated!]
<Unode>
If I add the import to the top of the file I get an infinite recursion error.
<ryantm>
contrapumpkin: Do you have any public code related to this?
<contrapumpkin>
nope, but I could probably clean it up and make it public if you wanted it
<contrapumpkin>
it's basically a fancy --check invocation
<contrapumpkin>
with a workaround for an obscure Nix bug I still haven't figured out
<gchristensen>
I'd love to formalize it in a job runner that runs against all of nixpkgs. I now have five things I want to run like this, this would be a sixth... so it is time to set it uup :P
blankhart has joined #nixos
<contrapumpkin>
I actually do maintain a cache for it ryantm
<contrapumpkin>
so basically only check each FO drv hash (not the output hash, the input hash) once
<LnL>
Unode: import? that looks fine just use super.fetchFromGitHub
<contrapumpkin>
so as long as nobody changes the download URL or the downloader source code, I don't recheck FO derivations
<Unode>
LnL: I still don't quite understand what self/super/<insert_var> represents in different contexts.
<contrapumpkin>
ryantm: you thinking of setting up an automated nixpkgs checker? I'd love to see that
<Unode>
LnL: thanks
<contrapumpkin>
gchristensen: and yes :)
<ryantm>
Yeah, that sounds like what I was thinking of. The part I don't exactly know how to do is how to discover all the FO drvs.
<contrapumpkin>
basically nix-instantiate whatever you use, filter the output for outputHash
<LnL>
Unode: self/super both are the entire nixpkgs set, but super doesn't include your overlays/overrides
<Unode>
I see
<ryantm>
contrapumpkin: Okay maybe that approach could work if I had a copy of the Hydra nix store, I was thinking more like I'd interate over all the packages in all-packages and try to do something.
<gchristensen>
ryantm: sounds awesome :D
woodson has quit [Ping timeout: 255 seconds]
<contrapumpkin>
ryantm: hmm, doesn't seem like you'd really need to do anything special though
alexteves_ has joined #nixos
<contrapumpkin>
`nix-instantiate '<nixpkgs>'
<contrapumpkin>
that'll produce a ton of .drvs which you can now search for outputHash
<contrapumpkin>
I guess the usual hydra concerns of whether packages are marked as broken and such apply
woodson has joined #nixos
<ryantm>
contrapumpkin: You can't instantiate all of nixpkgs that directly.
<gchristensen>
I'm so glad we got to the point where all of nixpkgs can evaluate
<gchristensen>
feels like a real win for tooling
<Unode>
LnL: if you don't mind me bothering, to override 'version' and 'src' should I be using override/overrideAttrs? I tried both but overrideAttrs has no effect (0.27.0 is still built) and override fails with "error: anonymous function at /var/nixpkgs-channels/nixos-18.03/pkgs/applications/video/mpv/default.nix:1:1 called with unexpected argument 'version', at
<clever>
<foo> maps to (__find_file __nixPath "foo") when parsing
<clever>
let and rec are nearby
<Unode>
so rec forces evaluation of recursive references?
freeman42x]NixOS has quit [Remote host closed the connection]
<contrapumpkin>
Unode: rec brings keys of the current map into scope in their values
<clever>
infinisil: look at lines 390 and 392!
<clever>
let creates a recursive set
<contrapumpkin>
a normal attrset: { a = 5; b = a + a; } a isn't in scope in that definition of b
<Unode>
contrapumpkin: and in nixpkgs context this means... variables from the definition?
<Unode>
I mean the derivation
<LnL>
copumpkin: well practical example of where it's not is { stdenv, darwin }: stdenv.mkDerivation { ... meta.platforms = with stdenv.lib.platforms; [ darwin ]; }
<clever>
let a = body; in a
<clever>
> let a = body; in a
<{^_^}>
undefined variable 'body' at (string):43:9
<LnL>
I think you mean this
<LnL>
> let { a = 42; body = a; }
<{^_^}>
42
<infinisil>
Huh
alexteves_ has quit [Remote host closed the connection]
<clever>
there is `let { ... }` and also `let foo = bar; in baz`
* contrapumpkin
runs
<infinisil>
Whoa, are you serious with this body thing..
<infinisil>
I did not know that..
<gchristensen>
let's help people who have questions contrapumpkin :)
<LnL>
yeah there are 2 let syntaxes
<contrapumpkin>
gchristensen: already ran, too late
<clever>
i didnt know that
nschoe has quit [Quit: Program. Terminated.]
<gchristensen>
ij: works for me?
<LnL>
gchristensen: I think it's too late, they will be too confused now ;)
<gchristensen>
ij: what are you man-ing and what are eyou see-ing?
<goibhniu>
Akii: oh, have you got any Qt programs installed in your user or root profiles?
<Unode>
gchristensen: here's one: if I want to override a dependency in http://dpaste.com/2W6DCAA how do I do that?
<LnL>
as in another version?
<gchristensen>
LnL: this is one of the concerns I have with a code evaluation bot, drowning away questions as people get too deep in to esoteric syntax topics
<__monty__>
Question, why does nix have to rebuild ghc every time I add a dependency and update the channel? I assume it's because the derivation's changed because dependencies changed but does it really have to rebuild ghc?
<clever>
__monty__: which ghc is it building?, i would expect it to be on the binary cache
<i-am-the-slime>
...7.3.0/collect2': execv Argument list too long
<manveru>
__monty__: i think that's when you compile with hoogle?
<i-am-the-slime>
So it won't compile.
<clever>
i-am-the-slime: try switching to stack2nix maybe?
<i-am-the-slime>
this is a work project, I don' t know how manageable that will be
<__monty__>
clever, manveru: ghcWithHoogle for ghc-8.2.2
<i-am-the-slime>
but if it allows me to work it would be good
<manveru>
__monty__: if i remember right, if you use ghcWithHoogle it'll recompile everything for some kind of documentation thingy
atu has quit [Ping timeout: 240 seconds]
<clever>
i-am-the-slime: if you run `stack2nix .` it will create a default.nix file that refers to every package defined in your stack.yaml
<clever>
i-am-the-slime: you can then just -A your package name with nix-build and it will build everything under nix
<i-am-the-slime>
-A to which command?
<clever>
i-am-the-slime: nix-build
tzemanovic has quit [Remote host closed the connection]
<__monty__>
manveru: Ah, unfortunate.
<manveru>
i might be wrong though :)
reinzelmann has quit [Quit: Leaving]
<manveru>
it's been ages since i used it
<__monty__>
manveru: Just not using haskell anymore or found a better haskel+nix workflow?
<__monty__>
The local hoogle's pretty useful to me.
<{^_^}>
[nixpkgs] @volth opened pull request #41688 → [staging] perlPackages: convert some modules to aliases → https://git.io/vhgqM
<manveru>
no haskell for me... felt too much like punching my brain through a meatgrinder
<__monty__>
How far'd you take it?
sbdchd has quit [Remote host closed the connection]
<manveru>
i built a parser for some game save files with it
<__monty__>
Haskell's way more than just a brainteaser but to each his own I guess.
<__monty__>
For small projects I hardly notice compile times and it doesn't matter because I use ghci during development. And with ghcid it's basically instant.
<clever>
__monty__: ive found some files that take ghcid 30+ seconds to reload
<infinisil>
clever: Must be a pretty big file then..
<clever>
it must have to do when what the file is doing
<__monty__>
clever: I did say small projects. Did you find out why they took long though? Tons of extensions, TH or something?
<infinisil>
There *are* bigger files that compile faster then, not that bigger files generally compile faster :P
<clever>
__monty__: the cabal file enables a ton of extensions globally, lol
<manveru>
my $DAYJOB atm is 90% with Go, with a little Crystal on the side, and i'm getting nervous if something takes longer than a second to build
<logiq_>
does anybody know if shellcheck is able to detect the shell when my script starts with `#!/nix/store/4sp0sc2wxas7213x59smrwhliqmhc5yk-bash-4.4-p12/bin/bash` ?
<logiq_>
"Tips depend on target shell and yours is unknown. Add a shebang."
<gchristensen>
oh, then no
<infinisil>
> :v weird1
<{^_^}>
weird1 = weird1 is not defined
<logiq_>
gchristensen: what are the alternatives for shebangs?
<__monty__>
manveru: Ghcid checks my projects faster than I can look up at its output.
<gchristensen>
usually I use #!/usr/bin/env bash, shellcheck it, then use patchShebangs to replace #!/usr/bin/env bash with #!/nix/store/4sp0sc..../bin/bash
<__monty__>
clever: Maybe that particular file just used more of them than others? Or it used something iffy like undecidableinstances?
<manveru>
__monty__: my favorite is compiling ghc :)
<__monty__>
manveru: Was just trying to figure out why you didn't like it because many (if not most) people I've spoken to that didn't, didn't based on outdated info about haskell. Haskell isn't perfect by any means, tooling's a huge painpoint still.
<gchristensen>
a major problem is Nixpkgs is one of the biggest projects on github, and so casual alternatives to github aren't necessarily going to work
<manveru>
i use gitea myself, but don't think it'd scale well for our needs...
<samrose>
srk: pagure looks good
<srk>
I've tried pagure, fedora uses it for thousands of repositories. it's backed by git-daemon/gitolite so that's the perfomance "bottleneck" I guess
<srk>
then auth is bit of an issue
<manveru>
what was the system SuSE uses?
<srk>
but again, if you have a way to deliver ssh keys to gitolite all is well
<mpickering>
"cachix use" now just prints some formatted instructions instead of modify nix.conf?
owickstrom has quit [Ping timeout: 260 seconds]
<Mic92>
yes
<Mic92>
on nixos at least
atu has joined #nixos
<mpickering>
What's the reason for this
<maurer>
On NixOS, /etc/nix/nix.conf is autogenerated
sorixelle has quit [Ping timeout: 265 seconds]
<maurer>
I have no idea what cachix is, but if it wants to modify nix.conf, it's going to want you to modify your configuration.nix and rebuild instead
<mpickering>
ok
<LnL>
Mic92: hmm, it was cherry-picked so I'd expect it to be in 2.0.4
<{^_^}>
[nixpkgs] @kirelagin opened pull request #41699 → ihaskell: Do not unset $PATH in the wrapper → https://git.io/vhgnN
<henrycrutcher>
Silly question: Are there any docs on how to set up container networking? I've tried reading the test cases, and I'm just baffled. I just want my container to be able to talk to the internet, and I have no idea how to do that. The host can talk to the internet, and I'm using NixOps
ogotai has quit [Remote host closed the connection]
<__monty__>
How would I go about limiting ghc's memory usage (-M ?) for nix builds of ghcwithhoogle?
Neo--- has quit [Ping timeout: 240 seconds]
matthewbauer has joined #nixos
<henrycrutcher>
It would be if its nat instructiobns worked -- they don't. They also don't talk about bridging at all, which would be another solution. Also, no idea whether the configuration goes in the host or container
<{^_^}>
[nixpkgs] @Mic92 merged pull request #41699 → ihaskell: Do not unset $PATH in the wrapper → https://git.io/vhgnN
<__monty__>
Hmm, ran out of memory just fetching files from cache.nixos.org, what do?
<henrycrutcher>
I tried, but to no avail. The problem isn't that you can't see the container from the outside. It's more that the container cannot see anything except the host
<henrycrutcher>
trying @clever's gist now
<clever>
__monty__: which thing was it downloading when it failed?
<{^_^}>
[nixpkgs] @Mic92 pushed commit from @Sorixelle to release-18.03 « jetbrainsjdk: 152b1136.20 -> 152b1248.6 »: https://git.io/vhglm
<adisbladis[m]>
jbboehr: Ping?
<Jetien>
hi! how to configure the nix-daemon in non-nixos (e.g. ubuntu)? i'd like to add a binary cache
<mpickering>
Is it possible to modify an existing derivation to add an extra output?
<hodapp>
what's the magical way to start building things in my shell.nix, but drop me into the environment at the step where it fails if it does?
<__monty__>
clever: : / Was hoping the latest release had those fixes because the release was after edolstra's comment about landing them in master. But ran into the same issue.
<mpickering>
hodapp: -K and then, source env-vars
<adisbladis[m]>
mpickering: Sure thing! Just do `someDrv.overrideAttrs(oldAttrs: { oldAttrs.outputs ++ [ "man" "info" ]})`
<clever>
__monty__: the nix in nixpkgs needs to update, its a bit old
<__monty__>
clever: It's 2.0.4, that doesn't have those fixes?
<clever>
samrose: each project has its own spec.json for declarative project definitions, which directs hydra to load the nearby default.nix to generate all jobsets
<clever>
so nearly all hydra config is in git and can be tracked and re-configured easily
<adisbladis[m]>
Jetien: Just start nix-daemon as root
<hodapp>
hrmph... can't modify the build directory now though
<adisbladis[m]>
You should only need to adapt the path in ExecStart
<Jetien>
adisbladis[m]: thanks
<mpickering>
What happens if you have "outputs = ["out", "out1", "out1"]"
<clever>
mpickering: something will probably fail
Fare has joined #nixos
<adisbladis[m]>
Just tried what would happen: error: duplicate derivation output 'out1', .....
<hodapp>
uggggh I fucking hate CMake... doing full trace output, and it's looking right at libboost_chrono.so and setting Boost_CHRONO_FOUND, and then ignoring libboost_atomic.so in exactly the same directory and complaining it can't find these (and others)
<{^_^}>
[nixpkgs] @ookhoi opened pull request #41708 → 18.03-pinentry-debian-patch - anonscm.debian.org has migrated to sals… → https://git.io/vhgBj
<hodapp>
hmm... so boost = boost16x, boost16x = boost166, and boost166 = callPackage ../development/libraries/boost/1.66.nix { }, and 1.66.nix has callPackage ./generic.nix blahblahblah, and what I need is to set enableStatic in generic.nix's arguments to true
<hodapp>
how do I go about finding the 'right' override construct for this? this is never clear to me
<clever>
hodapp: it looks like it correctly cascades
<hodapp>
not sure what that means
<clever>
hodapp: an override of boost166 also effects boost
<hodapp>
I don't care if it affects boost; I just want to make a modified version of one of them for which enableStatic is true
<clever>
hodapp: oh, one min
<eeva>
Anyone has experience with closure size reduction with Haskell. justStaticExecutable doesn't seem to cut it : when I `nix-store --query --requisites` is still giving me the whole haaskell universe of deps. My package only requires base and hakyll!
<gchristensen>
infinisil: it is bad enough that I often just switch to text-mode
<infinisil>
gchristensen: Without you pressing <tab>?
<gchristensen>
even without pressing tab it goes to col0
<infinisil>
Huh, <tab> works just like it should for me
schoppenhauer has quit [Ping timeout: 268 seconds]
<infinisil>
Using nix-mode
<LnL>
gchristensen: vim-nix
<Mic92>
emacs has completion, which vim-nix does not have.
<__monty__>
clever: I'm on btrfs so I'd have to mount a swapfile as a loop device. Would a nixos vm as remote build host or a hydra vm even be an easier option?
<infinisil>
LnL: vim-nix does work pretty well, thanks! Much better than nix-mode, so I often end up editing stuff in vim even though I'd prefer to use emacs
<infinisil>
Mic92: Completion for what?
Sonarpulse has joined #nixos
nh2 has quit [Quit: Leaving.]
<clever>
__monty__: nix build slaves would still result in it trying to download ghc locally, then upload it to the slave
<Mic92>
infinisil: I have seen the other day on someones laptop that it can complete nixos options or pkgs.<TAB>
<infinisil>
Mic92: Hmm.. I do have nixos-options and company-nixos-options installed, but have never figured out how it worked
<__monty__>
clever: Hmm, foiled again. Any idea why the download for ghc got bigger? This same machine has successfully installed ghc-8.2.2 before. It has 2G ram + 2G swap.
<{^_^}>
[nixpkgs] @xeji pushed commit from @r-ryantm to master « eid-mw: 4.4.1 -> 4.4.2 (#41571) »: https://git.io/vhg2a
<clever>
__monty__: how much is free in `free -g` right now?
<Myrl-saki>
__monty__: What machine?
<Myrl-saki>
__monty__: Your OS might have a 2GB/2GB split.
<clever>
__monty__: 32bit or 64bit cpu?
<__monty__>
clever: That doesn't seem to work totals are 1 and 1, free is also 1 and 1 for memory and swap.
<__monty__>
64bit arch linux on an old laptop.
<clever>
__monty__: check `top` to see whats using a lot of ram and see if you can free some up
jbboehr has quit [Read error: Connection reset by peer]
<Myrl-saki>
__monty__: Do you have another Nix device?
jbboehr has joined #nixos
<LnL>
gchristensen: Mic92: mostly joking :)
<infinisil>
Oh nice, to get nixos option completion in emacs, you just need company-nixos-options installed and do (add-to-list 'company-backends 'company-nixos-options)
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
<Myrl-saki>
infinisil: :o
<Myrl-saki>
I think I already have that.
<Myrl-saki>
Wait, how would I know
<infinisil>
package-list-packages shows which ones you got
<Myrl-saki>
infinisil: I have some company completion on my configuration.nix, but I think it's just textual completion.
<infinisil>
You might want to make company complete faster
<Myrl-saki>
Oh wait. There's a dedicated package for it. I'ma install it.
<__monty__>
Myrl-saki: I had an arch server I used for remote builds but no access currently. I do have nix on a mac though.
<{^_^}>
[nixpkgs] @Ericson2314 pushed to release-18.03 « haskell infra: Use `self.ghc` rather than closing over initial value, take II »: https://git.io/vhgr1
mekeor has quit [Quit: digital revolution = cryptoanarchism + cybercommunism]
barrucadu has quit [Client Quit]
szicari has joined #nixos
atu has joined #nixos
<mupf>
thblt: I already have a ~/.nixpkgs/config.nix file. Does it automatically install the packages defined in this file after login in with the user?
<mupf>
or how does it work with per-user configuration files?
<phry>
can someone tell me if stdenv.lib.platforms.unix includes darwin?
<mupf>
fyi zsh works perfectly now
nico202 has joined #nixos
<phry>
mupf: it will not automatically install everyhting in ~/.config/nixpkgs/config.nix - but you can define a derivation there that depends on everything you want installed, and then you can just install that "one derivation"
BenceF has quit [Remote host closed the connection]
BenceF has joined #nixos
BenceF has quit [Remote host closed the connection]
netsan has left #nixos ["Leaving"]
nico202 has joined #nixos
knupfer has quit [Ping timeout: 240 seconds]
humanoyd has joined #nixos
sigmundv__ has quit [Remote host closed the connection]
humanoyd has quit [Client Quit]
Sonarpulse has joined #nixos
knupfer has joined #nixos
logzet has joined #nixos
<{^_^}>
[nixpkgs] @Anton-Latukha opened pull request #41722 → [WIP] mbedtls: build with cmake&ninja → https://git.io/vhgXK
<hodapp>
bleh, just ran into what looks like https://github.com/NixOS/nixpkgs/issues/29877 while trying to build Open Shading Language, which I guess assumes it has both GCC and Clang/LLVM
<hodapp>
for whatever reason
johnw has quit [Ping timeout: 245 seconds]
<hodapp>
but I feel like Open Shading Language is a clusterfuck no matter how you build it
<hodapp>
yeah, I just meant pastebin very generically :)
<ldlework>
stepmania
<ldlework>
now that is a name I have not heard in a long time
<hodapp>
I used to play PyDance back when it was PyDDR.
orivej has joined #nixos
<ThatPako>
I usually play osu but that only runs through wine while stepmania is native :^)
<hodapp>
can stepmania read data from DDR?
unacceptable has quit [Ping timeout: 276 seconds]
<hodapp>
so, it's strange that missing files would cause a segmentation fault
<ThatPako>
I'm just as stumped as you are.
<hodapp>
I have to wonder if that's the actual problem, or just *a* problem
<ThatPako>
I assume that it has something to do with the game expecting the Data and Save folder next to the binary
<adisbladis>
jbboehr: Ah.. I had a suspicion this was gonna be the case. It's circular dependencies
<adisbladis>
Something which node package managers happily allow :/
<hodapp>
but it would seem like it can't be *that* fatal, if it proceeds to try to read the next file
<ThatPako>
which is usually the case when just unpacking the game "normally"
<adisbladis>
But in nix we get infinite recursion
Fare has joined #nixos
<ThatPako>
maybe it tries to read, fails, then tries to use the stuff it couldn't read which reads memory that isn't allocated?
<hodapp>
possible I guess
<ThatPako>
Anything I could do to fix that other than having all of stepmania's dependencies in my systemPackages and using it as if I were not using nix?
phdoerfler has joined #nixos
<hodapp>
does said configuration ship with the distribution that's being built, to where it might just be a matter of hacking the derivation a little to copy it alongside the binary?
phdoerfler has quit [Ping timeout: 256 seconds]
civodul has joined #nixos
<ThatPako>
hodapp: it just got weirder. the thing in /bin is just a symlink to the actual binary, which has the Data folder. Inside it is missing some the files though
<ThatPako>
so it _does_ ship with _some_ of the stuff needed, but not everything
<ThatPako>
neither does the file I could download from their site though
matthewbauer has joined #nixos
sigmundv has joined #nixos
<hodapp>
the actual binary has the Data folder? what do you mean?
<ThatPako>
the folder that contains the binary, which the file in /bin symlinks to is next to the Data folder
<ThatPako>
so the folder itself exists, and also contains some stuff
<ThatPako>
but apparently not what the game wants
rprimus has left #nixos [#nixos]
<pie_>
so i started trying to poke at haskell again but whenever i try to import anything it does Failed to load interface for ‘Graphics.UI.GLFW’ when i try to import that package
<pie_>
similarly for other packages
<pie_>
i used nix-shell -p haskellPackages.whatever
unacceptable has joined #nixos
chessai has quit [Remote host closed the connection]
<hodapp>
are you running GHC directly from that? or stack or cabal?
<pacmeister>
no just politics, just want to know nixos' stance
<sphalerite>
ldlework: I'm not sure. I had a brief look when you mentioned it this morning but was moslty just confused by the documentation
<ldlework>
sphalerite: a minimum the docs claim that the displayManager modules use this enumeration method
<ldlework>
but they all have their own enable option
<ldlework>
and no such enumeration support exists on the displayManager module itself
jperras has joined #nixos
szicari has quit [Quit: Leaving.]
aj_ has joined #nixos
<ldlework>
I think the most interesting thing in that conversation is the proposal to breakup nixpkgs, rather than any arguments the proponents of migration make.
<ldlework>
one of the problems is that if you are the maintainer of package "foo" you don't get notifications when there are bugs opened for that package (or any other package) or you get notifications for everything.
<ldlework>
"For example ArchLinux maintains a “core” repository for the base distro and then has individual repos in AUR for all the extra packages. As a package maintainer I am given my own git repo where users can post issues to it.
<ldlework>
Debian has a repository with issue tracker for each package. Canoncial built their own solution as well.
<maurer>
ldlework: We might be able to fix that by adding it to the issue checklist?
<ldlework>
"
<maurer>
ldlework: I will also note that the AUR is the biggest security nightmare known to man
<hodapp>
if one thing comes out of this in general, I hope it's some sort of way to migrate things that sort of have to live alongside a repo - like PRs and issues and their associated discussion - between different providers
<ldlework>
because of this?
<maurer>
Yeah
<ldlework>
I'm skeptical that nixpkgs monorepo is a source of security
<maurer>
basically, AUR is de-facto standard to get a lot of things done in arch
<maurer>
so you end up having to turn it on to install many packages
<maurer>
once AUR is on, you have to trust on an _ongoing_ basis a huge number of people
<ldlework>
same with nixpkgs no?
<maurer>
In the PR monorepo model, you trust random people, but only after verification, and only at discrete points
<ldlework>
not everyone is reviewing each change to each package
<maurer>
Sure, but someone is reviewing each change
<ldlework>
so you still are depending on a huge number of disparate eyes
szicari has joined #nixos
szicari has quit [Client Quit]
<MichaelRaskin>
A hundred of people with some track record
<gchristensen>
and a handful of corporations keeping eyes
<hodapp>
if split up per-package would we be expecting that it would typically be via people getting nix derivations pushed upstream to the project in question?
<MichaelRaskin>
Big Brother watches you? And subscribes to your repositories?
<maurer>
ldlework: Would your notion about package maintainers getting pinged be resolved if we added to the issue checklist "Run nox --get-maintainer packagename, and add @foo to your issue" (after implementing a functionality like that)
__monty__ has quit [Quit: leaving]
<ldlework>
I don't know, I'm an utter neophyte and not a maintainer of nixpkgs.
<infinisil>
mention-bot was a thing..
<maurer>
I don't remember what happened with mention-bot
<ldlework>
I'm still hung up on that bit of the docs that seems wrong regarding enumerations lol :D
<maurer>
In any case, the tl;dr of my position here is:
<infinisil>
ldlework: You can make a PR to fix it :D
<ldlework>
I wanna try out the design pattern but I'm suspicious of it given that the actual displayModules don't utilize the method
<ldlework>
infinisil: yes, maybe, if someone with authority confirms my suspicions
<maurer>
* Each repo you want to add is a trust relationship. If you want to add more than a few to get stuff done, this is going to get vulernable and brittle very quickly
<ldlework>
and if the suspicions are confirmed, then well, we need some other example of the enum pattern being used :D
<ldlework>
because it really seems like displayManager doesn't
<ldlework>
maurer: make it heirarchical
<ldlework>
that's how basically all other domains of human activity handle this problem
<maurer>
ldlework: So now, in order to facilitate a move to a multirepo environment (which none of the maintainers are asking for), you want to add a management layer to a project already short on manpower
<ldlework>
maurer: it is your claim that the management layer is already in place
<maurer>
...
<maurer>
I'm done with this.
<ldlework>
that you have reliable maintainers who are sufficiently capable of reviewing every package
<ldlework>
well, I'm not trying to upset you, sorry.
<ldlework>
just doing my best to reason about the problem.
halfbit has quit [Ping timeout: 256 seconds]
<hodapp>
from my perspective: anything that makes it more difficult for me to grep -R the entire nixpkgs repository to figure out how to do something because it was never really documented is not something I can get behind presently.
<ldlework>
it seems the complaint with separate repos is that you have to trust random package owners and so having a monorepo is better because you trust a known group (to review every single PR). I was simply suggesting that you assign areas of ownership across the package space.
<gchristensen>
hodapp++
<{^_^}>
hodapp's karma got increased to 2
<hodapp>
whoa, how was my karma already 1?
<LnL>
no multiple repositories is harder to maintain
<MichaelRaskin>
Historical observation: NixOS was a separate repository, then we gave up
<LnL>
^^
<LnL>
was just going to give that as an example :)
<hodapp>
still a better love story than Twilight?
<gchristensen>
I think my karma is negative :$
<ldlework>
I'll leave it to the experts. :)
<thekolb>
Did /var/setuid-wrappers move?
<infinisil>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 5
<hodapp>
don't we already have areas of ownership across the package space? we do, at least, have maintainers.
<MichaelRaskin>
Historical observation number two: NixOS in Subversion was only separate-ish, because SVN is really good with the narrow checkout stuff but in a sense everything is simply versioned at once.
<gchristensen>
:o
<LnL>
thekolb: 2 releases ago yes
<infinisil>
gchristensen: You figured out how to get karma xD
<MichaelRaskin>
thekolb: to /run/wrappers/bin I think
<gchristensen>
infinisil: do cool stuff?
<hodapp>
pay people?
<hodapp>
bring cookies?
<MichaelRaskin>
MiсhaelRaskin++
<{^_^}>
MiсhaelRaskin's karma got increased to 4
<infinisil>
gchristensen: Nah, make people wonder what your score is, because the only way to see it is to first increase it
<hodapp>
I do owe the channel several batches of cookies at this point.
<ldlework>
hodapp: then those same maintainers can be responsible for the same separate area-specific repos
<MichaelRaskin>
Cheating is simpler
<thekolb>
MichaelRaskin: Thanks! alien high five!
<MichaelRaskin>
hodapp: giving people cookies over internet is highly regulated nowadays in EU
<infinisil>
,botsnack
<hodapp>
MichaelRaskin: ...I see what you did there
<infinisil>
> botsnack
<{^_^}>
"Thank you very much."
<maurer>
Outside the NixOS sphere, ChromeOS dev started as a bunch of git repos held together by submodule glue
<ldlework>
you hide the complexity of having repos for seperate packages, by having trusted maintainers update pinning merges after reviewing related changes in the specific package repo
<maurer>
they've since moved to a combination of monorepo and the "repo" tool
<gchristensen>
maurer: and have you_SEEN AOSP's `rep....
<samueldr>
:/ I need nix 2.0+ to be deployed automatically, with the daemon, on a non-nixos platform
<hodapp>
ldlework: considering the number of packages and maintainers we're talking, this doesn't seem a tenable transition
<LnL>
thekolb: you might want to reboot if you did an upgrade, otherwise old processes might not work anymore
<gchristensen>
samueldr: systemd?
<samueldr>
is there a reason, gchristensen, the --daemon script doesn't want to run as root
<samueldr>
yes
<aminechikhaoui>
is it rude to ping a pkg maintainers on irc for build issues ? :D
<ldlework>
so each package maintainer gets its own repo, issues, notifications, etc. when they make a release, a maintainer would review the changes, just like they do now apparently, and move some pin
<maurer>
gchristensen: Yeah, AOSP is a mess, but it turns out it's better than full tiny repos :P
<samueldr>
it is automated at provisionning, and sudo/non-root would be highly annoying
<hodapp>
aminechikhaoui: might just ask here before pinging the maintainer
<thekolb>
LnL: no, I am just trying to retrofit a sudo-in-build.nix module from 16.03.
<aminechikhaoui>
yeah I asked :)
robstr has quit [Quit: WeeChat 1.9.1]
<ldlework>
the individual package mainatiner could even be responsible for making the change to the pin on behalf of the reviewer.
<aminechikhaoui>
"folks what's the deal with the mailutils pkg, seems it fails spuriously in one of the tests"
<samueldr>
I previously was using the (unsanctioned) 1.11 .deb that hydra built
<LnL>
thekolb: ah
<aminechikhaoui>
it's probably a known issue but it's causing some random build failures every once in a while
<hodapp>
ldlework: that large a number of repos seems basically unmanageable, combined with the fact that there'd have to be some management of specific revisions of every single repo
<aminechikhaoui>
so if the test isn't reproducible, seems it would be better to disable tests for the pkg
<LnL>
thekolb: was renamed because since 17.09 we also support setcap wrappers
<MichaelRaskin>
If the tests are organised sanely, killing just one test should be a simple operation…
<gchristensen>
samueldr: I don't know :|
<gchristensen>
samueldr: I have terrible hacks we could talk about ....
<samueldr>
it all depends on the height of the pile of hacks :)
<LnL>
I've also done some weird stuff to install nix :)
<gchristensen>
samueldr: very short ...
<samueldr>
the .deb was pretty hands-off, which was fun
<samueldr>
install .deb, create users and groups, sip from cup, all done :)
<thekolb>
so I am one step further, sudo now apparently tries to load libcap-ng.so.0, so I guess I have to try configure it without
* samueldr
might have forgotten the systemd services
<thekolb>
would they call that configure option --without-cap-ng or --without--libcap-ng?
pie_ has quit [Ping timeout: 264 seconds]
<thekolb>
hmm no such option...
pacmeister has quit [Quit: WeeChat 2.1]
<thekolb>
So this nix expression builds a sudo without pam and makes it available during build (along with /etc/passwd and some other files), while relaxing the sandbox. But this injected sudo barfs because it fails to load libcap-ng, do I no try to make that shared library available in the sanbox too? :/
<toonn>
Hmm, trying to install ghcWithHoogle with dhall-text which fails because of the missing dependency dhall >=1.14.0 && <1.15... Would jailbreaking work? (Took hourse to get this far so trial and error doesn't seem like a great idea.)
<ldlework>
Does anyone want to give me an example of adding an activation script that simply copies a directory next to the nix-expression to a directory like /home/ldlework?
<ldlework>
or /etc/somewhere/else
pacmeister has quit [Ping timeout: 256 seconds]
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
goibhniu has quit [Ping timeout: 264 seconds]
toonn has quit [Quit: leaving]
Mr_Keyser_Soze has quit [Ping timeout: 248 seconds]
__monty__ has joined #nixos
sbdchd has joined #nixos
<ldlework>
If you make option.foo.bar = mkEnableOption, can you also option.foo = mkEnableOption?
<ldlework>
I guess I'll just hope the answer is yes for now
sbdchd has quit [Ping timeout: 276 seconds]
<hodapp>
https://github.com/NixOS/nixpkgs/issues/29877 anyone have any creative solutions to work around this? Open Shading Language seems to require clang, and not sure if it also requires gcc but the CMake build tries to use g++
softinio has quit [Quit: Connection closed for inactivity]
nico202 has quit [Ping timeout: 264 seconds]
<hodapp>
the shellHook trick didn't seem to do anything
npmccallum_ has quit [Quit: npmccallum_]
<infinisil>
Hmm, why would one prefer `abort` over `throw`?
<infinisil>
You can't catch aborts, but you can throws
<LnL>
abort should only be used in fatal cases, things like the hydra evaluator will just stop
<hodapp>
so I tried both llvmPackages.clang-unwrapped and llvmPackages.libcxxClang
bennofs has quit [Quit: WeeChat 2.0]
<hodapp>
with the former, clang calls fail because they don't have the required -stdlib=libc++, and with the latter calls to g++ fail because -stdlib=libc++ is passed in
<{^_^}>
[nixpkgs] @xeji merged pull request #41509 → nixos/munge: run munge as user munge instead of root. → https://git.io/vhRnt
<{^_^}>
[nixpkgs] @xeji pushed commit from @markuskowa to master « nixos/munge: run munge as user munge instead of root. (#41509) »: https://git.io/vhgj5
<hodapp>
I guess I could try to build it in Docker or something... I don't know
<infinisil>
My objective was to write a function getSrcs, which you can pass pkgs, and it will return all urls of all derivations that get used as a source in nixpkgs
<infinisil>
This requires error handling of all sorts, because as soon as the first error occurs the evaluation would stop
<infinisil>
But you can *only* catch throws and aborts
<infinisil>
Stuff like `x.y` will throw an uncatchable error when x doesn't have y as an attribute
<infinisil>
> tryEval {}.y
<{^_^}>
undefined variable 'tryEval' at (string):43:1
tusj has quit [Read error: Connection reset by peer]
<infinisil>
> builtins.tryEval {}.y
<{^_^}>
attribute 'y' missing, at (string):43:18
<infinisil>
Can't catch that
lnikkila has quit [Ping timeout: 265 seconds]
<infinisil>
And that's the problem. While you can catch your own uses of attributes by doing `(tryEval x).success && x ? y && (tryEval x.y).success`. It would take a lot of work to do that for all the code in nixpkgs
<infinisil>
I wish that was catchable as well..
<thekolb>
So I am using security.wrappers.*.user = "root"; but it doesn’t change the owner of the resulting binary? :O
<thekolb>
mhm I am doing something wrong
<florianjacob>
grahamc: do I need to be a known / trusted user of grahamcofborg to be allowed to run the tests through the bot? did not find that in the readme.
<rotaerk>
bleh, alsamixer is still muted on boot ... wtf
<samueldr>
thekolb: the binary in /nix/store will not be owned by whoever, the wrapper in /run/wrappers
<thekolb>
does nix.sandboxPaths change the ownership somehow?
<{^_^}>
[nixpkgs] @xeji pushed commit from @andir to release-18.03 « memcached: 1.5.5 -> 1.5.8 (#41727) »: https://git.io/vh2f4
<endformationage>
How does one normally go about testing packages with modules? I wish to use an updated package in my local nixpkgs repo with a service module.
<thekolb>
Hmm, I’m out of idea... anyone got an idea of how to get a working sudo inside nix builds on nixos 18.03?
<clever>
thekolb: sudo should never work inside a nix build
<thekolb>
clever: yeah I kow, we had a hack that worked with 16.03 though... need that back :}
<thekolb>
well first the wrapper now looks for libcap-ng and libcap, and last I end up with sudo-chroot: /nix/store/3xsjm8rfpy0ysfjs1pcybj33fsigszgp-wrapper.c:203: main: Assertion `!(st.st_mode & S_ISUID) || (st.st_uid == geteuid())' failed.