<simpson>
No worries. I was quite ready to learn that there is a "quemu" package which is totally unrelated and does something like, I don't know, queue management for some database.
<gchristensen>
mupdf queue
<simpson>
¡Que Música Universal! Spanish-language music player.
fenedor has quit [Read error: Connection reset by peer]
<Church_>
simpson: Heh that wouldn't surprise me
erictape1 has quit [Ping timeout: 250 seconds]
shibboleth has joined #nixos
m0rphism has quit [Quit: WeeChat 2.2]
endformationage has quit [Ping timeout: 250 seconds]
<Guanin>
What can I do to speed up kernel compilation time? I've set nix.maxJobs to the number of threads. Are there other options that might speed it up?
m0rphism has joined #nixos
<gchristensen>
maxJobs doesn't impact build time
<gchristensen>
maxJobs is the simultaneous independent nix builds
<rycee>
jomik: "Anyone have an idea why "we" build a package here and then source it from init.el, instead of just setting text to be initFile (skipping the package)" Because it will byte-compile and optimize the init file and only the small shim will be loaded as a text file. Saves a bit in the start up time.
<rycee>
jomik: I like to keep my Emacs starting in well below 1 second, ideally below 500ms and compiling the init file help quite a bit.
<rycee>
jomik: The emacs-init module might make it into Home Manager at some point but it needs some polishing first. I put it up in that repo for the brave and/or foolish :-)
jasongrossman has joined #nixos
fusion809 has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @schmittlauch opened pull request #53256 → [WIP] ktouch: init at 18.12.0 → https://git.io/fhmGM
endformationage has quit [Ping timeout: 246 seconds]
shibboleth has quit [Remote host closed the connection]
<rycee>
jomik: Oh, and the `programs.emacs.finalPackage` option should be accessible in your configuration (but you need to pull in the config argument attribute)
eadwu has joined #nixos
<rycee>
I.e., refer to the option value as `config.programs.emacs.finalPackage`.
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fhmZ8
<Church_>
samueldr: Hey mind taking a look at a basic deriv for me? Failing when trying to find pcap.h even though I include libpcap. http://git.chatsubo.cafe/noah/Ubridge
<siraben>
But looks like the makeicecat script in the build phase uses "wget" to fetch another tar, why is it failing?
<siraben>
wget: unable to resolve host address 'ftp.mozilla.org'
<siraben>
even when manually wgeting it works fine
<samueldr>
Church_: maybe if you share something to look at
<Church_>
Fair enough, one sec.
<siraben>
samueldr: does wget not work within a build script?
<Church_>
samueldr: The source with the default.nix and builder.sh is above
<samueldr>
siraben: yes and no, it depends whether it's a fixed-output derivation or not; and more succintly, wget always works: network access is restricted
<Church_>
Thinking I might just have to fix the Makefile so it finds libpcap's path automatically
<Church_>
Think that might be it.
<jackdk>
Church_: it appears that libpcap provides a pcap-config program, so you need to patch calls to `pcap-config --cflags` and `pcap-config --libs` into the makefile
<siraben>
samueldr: not sure why it would be failing download inside a build script
<Church_>
jackdk: Well that's awful
<samueldr>
Church_: using a custom `builder` generally isn't the way to go; you lose a bunch of the magic sauce making building in nix painless, you would need to somehow re-do what's lost
<Church_>
Ah. I got that from following nix pills.
<Church_>
Troublesome.
<samueldr>
Church_: ah, nix pills are really about the basics of nix itself; I have to be frank, still haven't done 'em, but they're there so you better understand the underpinnings
<Church_>
Gotcha.
<samueldr>
siraben: it's an uncommon concept with nix; builds are sandboxed, no network access, but that would be inconvenient to fetch source files. So what's going on is that you can give a derivation network access, only and if only you can figure out a way to give it a fixed output; that is, an output which is bit-for-bit reproducible
<samueldr>
siraben: so `src = ...` is generally using a fixed output derivation of something that download a known source file
<siraben>
Right
<siraben>
Looks like Icecat has a source tarball too, I'll try that
tertl3 has joined #nixos
<samueldr>
siraben: when your main derviation is built, that source derivation is already built and complete, so you're given an environment with that source file extracted, as a convenience
<tertl3>
should I install nixos on a pixelbook?
<Church_>
You can't.
<Church_>
Well
<Church_>
Eh
<samueldr>
Church_: removing `builder` (and `args`) your build almost succeeds
<Church_>
I take that back
<samueldr>
tertl3: you probably can, you probably should, but I'm jealous of you now :)
<tertl3>
why cant you
<Church_>
tertl3: You can do it inside a lxc container inside a sandboxed vm
<Church_>
But otherwise linux hasn
<Church_>
But otherwise linux hasn't been ported to the chipset fully yet
<siraben>
I'm getting that error in the ./configure step
<{^_^}>
[nixpkgs] @marsam opened pull request #53266 → qrencode: fix darwin build → https://git.io/fhm8G
<samueldr>
,locate bin whic
<samueldr>
,locate bin which
<{^_^}>
Found in packages: which, busybox, debianutils
<{^_^}>
Found in packages: which, emboss, busybox, easyrsa2, debianutils
<samueldr>
busybox is probably not the one you want, which is the one you want
<siraben>
Ok, adding to nativeBuildInputs
<siraben>
I'm still not clear on the difference between nativeBuildInputs and buildInputs
<samueldr>
the wrong explanation, but works for 80% of the case is: is it a thing that runs during build? native, is a thing that's linked, like a library, or used at runtime by the software? buildInputs
<samueldr>
native is native to what's doing the build :/
<siraben>
I see.
<samueldr>
(the nixpkgs manual has the full explanation)
<esmerelda>
neat! I now have pywikibot building! couldn't get the tests to work but that seems like a crapshoot anyway (are theyeven bundled in PyPI?)
<esmerelda>
I should probably submit a PR to merge this into nixpkgs... need to read up on the submission guidelines
ddellaco_ has joined #nixos
griff_ has joined #nixos
* esmerelda
feels accomplished
<{^_^}>
[nixpkgs] @jethrokuan closed pull request #53271 → Move aliases in all-packages to aliases.nix → https://git.io/fhm4Y
ddellacosta has quit [Ping timeout: 244 seconds]
sanscoeur has quit [Remote host closed the connection]
iqubic has quit [Read error: Connection reset by peer]
iqubic has joined #nixos
FRidh has joined #nixos
<siraben>
How do I fast forward my fork of the nixpkgs?
<siraben>
I created a pull request that was "merged" but according to GitHub it was closed, and my branch currently diverges from the main branch https://github.com/NixOS/nixpkgs/pull/51735
<ezrakilty>
Hello; hoping for some help with the install process; I was directed here by the installer itself.
<ezrakilty>
I ran the single-user installer fine and was able to play around with nix-env and friends
sanscoeur has quit [Ping timeout: 250 seconds]
<ezrakilty>
Then I ran the —daemon version of the installer and (after arduously cleaning out remnants of the old install) it told me I could open a new terminal and run nix commands
<ezrakilty>
but even in a new window, my shell doesn’t have a $PATH pointing at nix-env, nor do I have a ~/.nix-profile
<wedens>
etu: big thumbs up for attempt to make nix-mode indentation useful :)
boothead has quit [Ping timeout: 256 seconds]
dermetfan has joined #nixos
<ezrakilty>
To the problem I described earlier: I found the solution was to run `source /nix/var/nix/profiles/default/etc/profile.d/nix.sh`, then `/nix/var/nix/profiles/default/bin/nix-env -i nix`.
<jmarin217>
Does anybody know how to set up custom display managers? I'm trying to use ly.
<wedens>
I can only recommend looking at how existing DMs are implemented
<jmarin217>
Yeah I've been looking at how slim does it and I tried using displayManager.job.execCmd but that just conflicts with lightdm even though I haven't enabled it
jomik has joined #nixos
<wedens>
are you sure it's not enabled? try to explicitly disable it
<wedens>
I don't see how lightdm can define displayManager.job.execCmd without being enabled
<jmarin217>
I displayManager.lightdm.enable = false;?
<wedens>
`services.xserver.displayManager.lightdm.enable = false;` to be more precise :)
<jmarin217>
Yeah I've got that inside of services.xserver = {};
<wedens>
how do you know it conflicts with lightdm? does it try to launch lightdm?
<jmarin217>
Ah! Thanks for your help but I've just found lightdm-mini-greeter. I'm gonna end up using that as it looks like it's easier to set up.
<jmarin217>
Thanks again for your help :)
orivej has quit [Ping timeout: 246 seconds]
<mtesseract>
Hi. I am new to Nix. Currently trying to become familiar with writing simple expressions for nix-shell. I have encountered a problem which I am not sure how to debug, though. If somebody could provide me with some pointers into the right direction, this would be great. :) So, here is a small shell.nix: https://pastebin.com/uXUSc8TM What I am trying to do is the following: I would like to work on a Haskell package (fractals.nix). For
<mtesseract>
nix-shell, I understand that shell.nix should evaluate to the fractalsPkg.env attribute and this works -- the Haskell dependencies of the fractals package are correctly installed. Now, I do not only want the fractalsPkg.env environment available in the nix-shell, but also some other programs, e.g. 'cowsay'. So, I need to combine fractalsPkg.env with cowsay in some way. I thought since the fundamental block here is a derivation, that I
<mtesseract>
should write a new derivation with mkDerivation that has appropriate buildInputs specified. Listing cowsay in buildInputs works as expected, but including fractalsPkg.env there prevents nix-shell from starting. It complains "bash: /nix/store/<some-hash>-ghc-8.4.4-with-packages: Is a directory".
griff_ has joined #nixos
periklis has joined #nixos
juhe has joined #nixos
<wedens>
I think the problem is that fractalsPkg.env is not a derivation
<m15k>
I guess nixos-containers is based on systemd-nspawn?!
<wedens>
yep
<m15k>
Is there any details why a root container access has access to host system? Is that caused by systemd or has something todo with the way nixos-container is organized?
<mtesseract>
wedens: well, (import <nixpkgs> { }).haskellPackages.callPackage ./fractals.nix { } does yield a derivation, according to nix repl.
<ezrakilty>
I have another newbie question: Is it expected that two different users can use the same profile (e.g. default)?
<ezrakilty>
I am using the daemon and each of two users can install packages just fine with their own profiles, but when I switch to the “default” one, they get “permission denied” errors on the /nix/var/nix/profiles/default.lock file
<Myrl-saki>
I honestly do not understand Nix's "stack" trace.
<Myrl-saki>
wedens: I think it's something you added to home manager htough.
<Myrl-saki>
Maybe this.
<Myrl-saki>
while evaluating 'mkCmd' at /nix/store/wzggdqn7njsrxwpgg8yf76czav0bq4dh-source/modules/home-environment.nix:316:17, called from undefined position:
<wedens>
I'll try commenting my home-manager line by line :/
<wedens>
wtf. it happens even with a whole hm config being commented
ferriswheel has quit [Quit: Leaving]
<Myrl-saki>
Maybe something happened with home-manager itself?
orivej has quit [Ping timeout: 268 seconds]
<Twey>
Is there a trivial builder in nixpkgs that copies a path into the store under a given path? E.G. I have a directory configs/ and I want to create a derivation whose $out just contains share/configs/
<etu>
siraben: But it doesn't pass contributing guidelines, couldn't you use the template for PR's and check boxes there? Also, commit messages doesn't follow the guidelines stated in there.
<siraben>
etu: ok will fix
<siraben>
Should I reopen the request?
<{^_^}>
[nixpkgs] @das-g opened pull request #53309 → davmail: fix homepage domain name → https://git.io/fhmQ3
<wedens>
what ends up being cached in official binary cache?
griff_ has joined #nixos
hyper_ch2 has quit [Quit: Page closed]
Boomerang has joined #nixos
<gchristensen>
everything built on hydra.nixos.org
<Boomerang>
In nixpkgs, for the Haskell packages that come from Hackage, how is the version specified? I see hackage-packages is auto-generated, how should I go about updating a version number?
<gchristensen>
hydra restricts access to the internet to increase reproducibility
<samrose>
gchristensen: it's weird as the the same hydra server will run the same line on a different project/jobset the only major difference being the first runs locally, the 2nd runs on a remote build machine
<samrose>
(locally to the hydra machine that is )
freeman42x[nix] has quit [Ping timeout: 272 seconds]
<samrose>
searching seems to reveal a pattern of people trying to pin nixpkgs triggering this issue of "access to URI is forbidden in restricted mode"
<schopp0r>
hi. I have built a package from a repo, using nix-build -A ...; it somehow placed this package not only into the result/-folder, but also in /nix/store. How do I remove it from /nix/store again? (I want to build it again…)
<samrose>
gchristensen: I figured it out: I needed a build input in hydra to checkout the https://github.com/NixOS/nixpkgs-channels.git repo. I will have to review the hydra source code, but this was the difference
<clever>
schopp0r: a: nix-build always puts the results in /nix/store/ and result is just a symlink to that
<schopp0r>
clever: how can I remove it?
<clever>
schopp0r: b: nix will detect if any inputs have changed, and rebuild it when they have changed, why do you want to force a rebuild?
<schopp0r>
clever: I want to see the build log. Because IMO the build should fail because the checksum should be false.
<clever>
schopp0r: nix-store -l ./result
sanscoeur has joined #nixos
<exarkun1>
ooh nice
<schopp0r>
clever: the thing is, already the fetchFromGitHub-step should fail, because I increased the version, and I didn't change the checksum
<ottidmes>
schopp0r: and if you really want to remove it, there is: nix-store --delete $(readlink -f ./result)
<clever>
schopp0r: if the sha256 is the same, nix assumes the output is identical, and wont even do the fetch
<clever>
schopp0r: what you want to do, is zero out a few digits of the hash, to force a re-fetch
cyounkins has quit [Remote host closed the connection]
sanscoeur has quit [Ping timeout: 250 seconds]
cyounkins has joined #nixos
rauno has quit [Ping timeout: 240 seconds]
c00w has joined #nixos
mtesseract has joined #nixos
freeman42x[nix] has joined #nixos
mtesseract has quit [Client Quit]
cyounkins has quit [Ping timeout: 246 seconds]
<samrose>
anyone ever run across error: "command `git rev-parse master' failed with exit status 32768 at /nix/store/<hash>-hydra-2017-11-21/libexec/hydra/lib/Hydra/Helper/Nix.pm line 428" ?
<mog>
gchristensen, what is the limiting factor for nixos-channel rebuilds right now? like if we wanted to have nixos-18.09 channel to always be concurrent with git what would that take?
<simpson>
Maybe that doesn't affect oldness, but I thought that we only bumped those -unstable heads when the tests pass.
<mog>
simpson, you are right at least currently there have been 6 build attempts that have failed
<samrose>
I wonder if my problems are that 1. I am running nixos 18.03 on build machien 2. I am runnign an older version of hydra "2017-11-21 (using nix-2.0.4)"
<mog>
but i wasnt sure if that was real failure or related to build resources as a build for unstable did succeed in that time
agander_ has joined #nixos
<gchristensen>
sometimes that is hardware being overloaded, other times it is real stuff -- more hardware and a smarter scheduling algorithm would help there.
<samrose>
simpson: what would source of data be for how old a channel is at any point in time?
<gchristensen>
running the full nixos test suite on every PR would be one way to do it, but man oh man
<simpson>
samrose: Scraping howoldis or similar. gchristensen has a bit of a Prometheus setup, and might already be collecting some similar metrics.
<mog>
gchristensen, is that something that is easily parrelizable to infinity, or is there a real time cost to doing that?
<gchristensen>
mog: which part?
<mog>
running the full test suite on a tag
<simpson>
gchristensen: Can't we cache test results? Or is the problem that the tests will *never* cache?
cyounkins has quit [Remote host closed the connection]
<gchristensen>
it could be parallelized to infinity, but hydra couldn't handle that load as-is
<gchristensen>
simpson: we do, but the cache is rarely hit :)
agander has quit [Ping timeout: 272 seconds]
<samrose>
I wonder if there is an existing scheduling software that could trigger builds via hydra api based on criteria?
<simpson>
gchristensen: Yeah, that. 'Cause we have a top-level symlink farm for the NixOS configuration which is gonna generally be different every time. Makes sense, sort of.
<gchristensen>
there is, and an experiment was done around this a year a couple years ago
<gchristensen>
it was too much for the hydra system
<simpson>
gchristensen: Interesting pattern: I wanted to build a directory of Monte modules in a way that incrementally cached the builds. This meant *undoing* a bunch of Nix best practices around `filterSource` and processing an entire directory at a time.
<simpson>
Similarly, we could do way better at caching if only we had some top-level *unbundling* of symlinks.
<c00w>
simpson - would content based storage also help with this? (since the symlinks would be more consistent between runs?)
<c00w>
s/storage/hashing
<simpson>
c00w: Normally I'd say no, but in this case, maybe? I think that there might not be any way around the fact that a systemd-oriented distro *has* to do a lot of centralization of configuration, and at the same time Nix won't let us forget our path dependencies.
cyounkins has quit [Ping timeout: 268 seconds]
FRidh has joined #nixos
<exarkun1>
schopp0r: you might want to get nix-prefetch-github
<c00w>
I thought that content based hashing would cause the paths to match?
<wedens>
is setting NIX_PATH enough or do I need both NIX_PATH and `nix.nixPath` to eliminate the need for channels?
<simpson>
Hm. I'm not sure about eliminating channels, but setting NIX_PATH should be enough for doing local work.
<simpson>
Note that, if you need stuff from nixos, you'll have to make sure that that is in your NIX_PATH too.
<obadz>
in git add -p / git checkout -p, do you guys have the ability to hit backspace and not get a ^H instead of a backspace?
<obadz>
isn't it nicer to rollback the nix files in this case?
<obadz>
pls ignreo that 2nd line ^^^
erictapen has joined #nixos
<wedens>
I've seen a lot of discussions about pining everything in nixos, and I'm a bit surprised there is still no simple solution
<obadz>
wedens: setting NIX_PATH is fine, no?
phry has joined #nixos
<wedens>
obadz: I'm trying to figure this out. What should be in NIX_PATH? nixpkgs, nixos-config,nixpkgs-overlays? Where do I set NIX_PATH so that it's being picked up by everything?
Denommus has joined #nixos
<obadz>
I might not be understanding the question right but you can set it as an env variable in whatever shell you're running
<simpson>
wedens: I use $(export NIX_PATH=nixpkgs=/home/simpson/nixpkgs:nixos=/home/simpson/nixpkgs/nixos) but I don't have any overlays or other fanciness.
phry has quit [Client Quit]
<wedens>
I guess I'll also install the configuration using `NIX_PATH=... nixos-install`?
<wedens>
also, can I somehow provide nixpkgs via fetchTarball so that I don't have to clone nixpkgs for each system
kingemp has quit [Remote host closed the connection]
kingemp has joined #nixos
<simpson>
I'm sure you've heard the spiel about nixops. (Solution to, and cause of, all my problems~) There probably is a simple way to get NIX_PATH right.
kreisys has joined #nixos
<simpson>
https://github.com/NixOS/nixpkgs/issues/27994 suggests that maybe setting `nixos` doesn't matter as long as you get `nixpkgs` right and also `nixos-config`, which should point to /etc/nixos/configuration.nix, whereever it lives.
<{^_^}>
#27994 (by puffnfresh, 1 year ago, closed): Document pinning nixpkgs in manual
amfl has quit [Ping timeout: 246 seconds]
<immae>
Hey there, can we do better than ${"''"} when we want to include a literal '' in a multiline string in a nix expression? (also, is there an existing way to use a local file as a template with variables instead of including it directly in the nix expression?)
<symphorien>
,escape'' immae
<{^_^}>
immae: '' two single quotes: ''' bash curly bois: ''${} newline: ''\n tab: ''\t any character x: ''\x ''
<wedens>
simpson: yeah, I've seen this issue. There are some bits and pieces of information in dieer
<wedens>
*different issues
<wedens>
but there is no complete picture
<immae>
Thanks symphorien
<immae>
(symphorien: I looked for it in the documentation in vain, is it documented somewhere?)
amfl has joined #nixos
<simpson>
wedens: I agree that there *should* be a complete picture. It's annoying that nixos-rebuild secretly takes some environment variables transitively without documentation.
<wedens>
simpson: I haven't used nixops yet. Can I just set some attribute to desired nixpkgs revision and press the red button? :)
<simpson>
wedens: With both nixops and disnix, it's a bit more complicated than that. You'd end up *locally* creating the environment you want, including the nixpkgs revision, and building the closure locally too, and then the closure would be instantiated on each target machine without even running nixpkgs.
<simpson>
wedens: You *do*, at least with nixops (I haven't seriously used disnix) end up with basically a configuration.nix that you edit once and deploy to many machines.
<wedens>
simpson: but then again, you still have to set up pinning locally, right? :)
<simpson>
wedens: Yes, with the export I had pasted above. I send stuff upstream to nixpkgs, so it's convenient for me to use my local git checkout... I suppose that if you don't have one of those, then yes, the problem persists to a degree.
<wedens>
so that you don't have just the same environment on all machines, but the same and _specific_ environment
shazow has left #nixos [#nixos]
erasmas has joined #nixos
jabranham has joined #nixos
endformationage has joined #nixos
<samrose>
advice sought for hydra build issue: oraclejdk unfree license refusing to evaluate: is this trying to evaluate on the build machine, or on the hydra server? I think I can resolve with an overlay, but I wasn't sure where this error originates in the hydra build process
joncfoo has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @hansjoergschurr opened pull request #53324 → trx: init at 2018-01-23 → https://git.io/fhYtB
<samrose>
I guess looking at the system logs it appears to be a hydra server evaluation error
lokado has joined #nixos
<lokado>
Hi everyone, do you know how to start a seperate X session on the externel monitor on nixos?
<{^_^}>
[nixpkgs] @bennofs opened pull request #53325 → ###### Motivation for this change → https://git.io/fhYt7
tilpner has joined #nixos
sir_guy_carleton has quit [Ping timeout: 246 seconds]
<tilpner>
All of this nonsense started after my last upgrade of nixos-unstable. Now the server dies every few hours. Has anyone else had problems (stability-wise) with recent unstable?
Mateon1 has quit [Ping timeout: 246 seconds]
<tilpner>
(Yes, yes, my fault for running unstable on a server, and for breaking my own rollback. I needed the per-interface firewall fixes)
<{^_^}>
[nixpkgs] @domenkozar opened pull request #53326 → Elm: automate packaging with elm2nix → https://git.io/fhYtj
<immae>
tilpner: just to check that it’s indeed nixos related, how does it "dies" ? Does it simply reboot? I’ve had problems lately related to that issue https://www.mail-archive.com/search?l=debian-bugs-dist%40lists.debian.org&q=subject:%22Bug%23904822\%3A+linux\-image\-4.17.0\-1\-amd64\%3A+Freeze\%2Ftiming\%2Finterrupt+problems+since+4.17%22&o=newest&f=1 which made my system hang for some minutes or
<immae>
indefinitely every now and then. Nixos had nothing to do with it, only changes in the kernel
<tilpner>
immae: I don't really now. I have prometheus and atop recording, but they (and journald) stop recording whenever this happens
<immae>
ok
agander_ has quit [Ping timeout: 240 seconds]
<tilpner>
I suspect it's dead, because it wasn't making progress on raid1 resync-ing during the time of the incident
<tilpner>
Sometimes it comes back on its own, sometimes only after a remote reboot
<immae>
Ah you have a RAID
<immae>
So you’re probably in the same situatoin yes
<immae>
Can you try adding dm_mod.use_blk_mq=0 scsi_mod.use_blk_mq=0 to your kernel command line?
<tilpner>
I had support replace one of the disks yesterday. After it started freezing, I ran self-tests and one of the disks had read errors
<tilpner>
I will definitely try that after these memory tests finish
<immae>
Sorry I forgot one: ahci.mobile_lpm_policy=0 dm_mod.use_blk_mq=0 scsi_mod.use_blk_mq=0
<immae>
If I understand correctly, it’s related to power management and it makes problems with RAID arrays
<immae>
You have the same symptoms as me, so hopefully this will fix that
<immae>
(to be honest, maybe only one of the three may be necessary, but I was so desperate to have such problems that I didn’t even try to find which one was necessary, feel free to make tests and report on your own)
<{^_^}>
[nixpkgs] @FRidh opened pull request #53327 → stdenv: assert whether `name` or (`pname` and `version`) are given → https://git.io/fhYqr
<tilpner>
immae: I'm very glad you were reading along, I'll let you know if it helps :)
<schopp0r>
hm. I get "store path mismatch in file downloaded" … what does this mean?
<schopp0r>
when trying to build.
jonaswouters has quit [Quit: Connection closed for inactivity]
<tilpner>
Where the second buildFHSUserEnv refers to the real thing
<obadz>
so p1.override { p2 = … } overrides package p1's dependency p2 with a modified version of p2. How can I express p1.overrideDeep { p2 = … } where p2 is overriden in the transitive closure of p1?
<wedens>
obadz: perhaps it can be overwritten for nixpkgs that'll be passed to the derivation? or something like that. I'm not sure
pietranera has quit [Quit: Leaving.]
<ottidmes>
I am getting an error while building a nodejs package via node2nix: /nix/store/3zal9zfqb41fsbmwl5ba7sxh3h2ngvxi-git-2.18.1/bin/git ls-remote -h -t https://github.com/duosecurity/duo_web_sdk.git (Could not resolve host: github.com), any idea how to work around this?
<obadz>
wedens: I don't wish for it to be overriden for all of pkgs, just for the dependencies of 1 package.
<wedens>
obadz: yes. that's why I said "nixpkgs that'll be passed to the derivation" ;)
<obadz>
nixpkgs isn't really passed to the derivation, there's an odd fixed point relationship between nixpkgs and the derivation (the derivation is part of nixpkgs)
<ottidmes>
obadz: I agree with wedens, that seems the most obvious way to do so, even something like this might work: (pkgs // { p2 = ...; }).callPackage ... (where the callPackage arguments will have to be looked up for p1 in most likely all-packages.nix)
<infandum>
tilpner: I get the error of value is a function while a list is expected
<exarkun1>
obadz: Make a new nixpkgs with the override you want, use callPackage from that nixpkgs to create p1
<nbp>
obadz: nowhere
<exarkun1>
I guess ottidmes beat me to it
<obadz>
ottidmes / exarkun1: hmmm agree that using callPackage on an overriden version if nixpkgs would work but that feels wrong for some reason. I'll have a think.
<wedens>
obadz: if the derivation is in nixpkgs, it'll be `xyz = (import <nixpkgs> { overlays = [ p1_overlay ]; }).xyz;` where `xyz` is your derivation
<wedens>
if I'm not missing something
<nbp>
obadz: The problem is that it implies that we would have to clean up a few things in Nixpkgs to be useful :/
<Church_>
Or would I want to break out a custom build script then?
<samueldr>
the only reason to break out a custom builder is if the build is wholly alien to anything remotely sensible on linux; even other build systems than the classic "./configure; make; make install" are handled through the generic build in nixpkgs
sanscoeur has joined #nixos
<samueldr>
e.g. cmake or meson work fine
<samueldr>
(once their hook used)
Xiro` has joined #nixos
<Church_>
Gotcha, so I should probably just run sed on the makefile during the configure phrase then?
<Church_>
Need to edit out some lines.
<samueldr>
yeah, that's what's usually done
<samueldr>
or a patch
Xiro` has quit [Remote host closed the connection]
elgoosy has joined #nixos
elgoosy has quit [Remote host closed the connection]
tzemanovic has quit [Remote host closed the connection]
ubert has quit [Quit: Leaving]
tzemanovic has joined #nixos
sanscoeur has quit [Remote host closed the connection]
sanscoeur has joined #nixos
<Church_>
samueldr: So this is the modifcation I made, added a configPhase. https://paste.rs/QJM The error I get is this: https://paste.rs/ez5 Getting a permission denied error despite not leaving the source directory
<Church_>
Verified that I was in the dir by running ls during the configPhase as well/
<Church_>
Not sure what temp file it's trying to open since I explicitly name Makefile as an argument.
<samueldr>
(you probably want to prefer using `postPatch` to sed things out, because configurePhase when there's a ./configure script is used to run it; though it doesn't affect your current derivation)
<Church_>
Yep
m0rphism has quit [Quit: WeeChat 2.2]
m0rphism has joined #nixos
cyounkins has joined #nixos
<Church_>
samueldr: Okay so here's an odd error, in the configure phase I decided to run "mkdir -p $out/bin" And then append /bin/ to my Makeflags so the binary would be placed in $out/bin/
<samueldr>
Church_: looking at other `makeFlags` in nixpkgs I think I see; looks like everything is escaped. using $(out) instead would work AFAIUI
<samueldr>
that's because instead of being bash that does the expansion, it's make
<samueldr>
(TIL)
<Ericson2314>
Nbp you mean my big list of steps?
<Ericson2314>
I tried to put people's priorities besides mine in that last :)
<nbp>
Ericson2314: yes. I mean to do write Nixpkgs overlays with NixOS modules.
<nbp>
Ericson2314: NixOS module have much-more overhead compared to Nixpkgs.
<Ericson2314>
Sure
<Ericson2314>
What do you think of the PR itself?
cyounkins has quit [Ping timeout: 250 seconds]
<Church_>
samueldr: So tried passing both ($out) and {$out}
<Church_>
The latter worked but still no output in the store.
mtesseract has joined #nixos
<nbp>
Ericson2314: I have no look deep inside the logic, but from my point of view "monoid" can be renamed into "blue-duck-with-dark-feathers" and that would carry the same level of knowledge to me.
dermetfan has joined #nixos
<Ericson2314>
Yeah all the monoid stuff should just be implementation detail
<Ericson2314>
The end user interface is 2017's {{...}}
<Ericson2314>
But with self and super
<nbp>
Ericson2314: I have no look that deep yet
<Ericson2314>
The monoid part is refers to getting the underlying (self: super: ...)
<Ericson2314>
And combining them with composeExtensions
<samueldr>
Church_: and what about `$(out)`?
<samueldr>
because here it looks like it's working
<Ericson2314>
And making a new attrset out of that
<Ericson2314>
Sure NW
mtesseract has quit [Client Quit]
<nbp>
Ericson2314: Find a name which is meaningful in Nixpkgs terms.
<nbp>
Ericson2314: from a highlevel overview this is my conclusion.
<Ericson2314>
Just trying to give you a little flavor now; I hope you can soon because I think we both agree we an incremental approach is needed to fix nixpkgs's organization
<{^_^}>
To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000) then replace it with the correct hash Nix expected.
<disasm>
Is there a way to preserve .git with src in a derivation? Essentially, the derivation in parts uses git to pull the rev to embed it in a document generated by the derivation.
lawlesseel has joined #nixos
<gchristensen>
it is not possible to do that in a stable way, because the .git dir is not reproducible
<gchristensen>
you're using fetchgit?
<FRidh>
disasm: we always patch thhose kind of things out
<disasm>
no, it's not using fetchgit
griff_ has quit [Quit: griff_]
<disasm>
that's what I though FRidh, just wanted to check if there was some work-around
<exarkun1>
(that is, it still fails during install with "/nix/store/n002dadikki28mjprlcsyhvbavxqn1ih-python-2.7.15-armv6l-unknown-linux-gnueabihf/bin/python2.7: cannot execute binary file: Exec format error")
<samueldr>
exarkun1: I believe #53123 is working to fix those issues
<disasm>
latex with a makefile that uses texlive.gitinfo2
<exarkun1>
yesterday I booted a c5d.2xlarge to do my building and made a lot of progress but then all of my packages are on the opposite side of a slow link from me ... :/
<gchristensen>
disasm: not sure whats up there, but Nix definitely doesn't automatically exclude .git
<exarkun1>
samueldr: thanks for the link to that commit. it helped a bit. the build still fails because of a python 2.7 / python 3.6 mixup. but maybe if make the use of buildPackages consistent within the installPhase script...
ris has joined #nixos
phreedom has joined #nixos
<samueldr>
yeah, I think I was lucky in my use case, I was 50% sure the `buildPackages.python` was wrong there
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/fhY4z
<dtz>
looking at the gitinfo bits, is there a particular reason to not use the sample post-commit hook (even if invoking manually or w/e) instead of what's in the gist?
ferriswheel has joined #nixos
<Twey>
Myrl-saki: Ahhhhhh linkFarm
ferriswheel has quit [Client Quit]
ferriswheel has joined #nixos
<Myrl-saki>
Twey: Oh. :P
<Myrl-saki>
Twey: It's not exactly what you really want though, since the files can move, right?
tzemanovic has joined #nixos
ferriswheel has quit [Client Quit]
<Twey>
Myrl-saki: What do you mean?
<Twey>
You can give a path to link it under
<Twey>
That's what I was looking for
<ezrakilty>
Hi all. I know about the “single version policy” for the main nix channel. But for demo purposes, is there any package you can think of that does have multiple versions in the channel? I want to show some people how nix can handle multiple versions for different users, but without waiting for the channel to update I don’t know how to do that :-)
<bbarker>
ezrakilty, off the top of my head, ocaml
<ajs124>
ezrakilty, do you mean something like boost155 vs. boost168?
<bbarker>
^ ocaml would be the same as the boost example
<Mic92>
gchristensen: on the 35c3 I got a signature for my gpg key with a guy that tries to establish a strong set of web-of-trust between all maintainer. Our key that is signs the nix release is not in good shape.
<Mic92>
We should cross-sign it some time on the next nixcon.
<Mic92>
sphalerite: will it stop, once I install some yubikey software?
<gchristensen>
Mic92: nice
juhe has quit [Remote host closed the connection]
<Mic92>
I am too scared to set yubi-key up, without making a backup on my harddrive that is at home.
<gchristensen>
like moving your gpg key over to it? I agree
<gchristensen>
I hate GPG.
<Mic92>
I don't like it either
<Ericson2314>
FRidh: so you never need `buildPackages.` in any `*Inputs` or `dep*` list, except for `depsBuildBuild = [ buildPackages.stdenv.cc ];`
<Mic92>
but I don't see an alternative
ezrakilty has quit [Ping timeout: 272 seconds]
ezrakilty_ is now known as ezrakilty
<tilpner>
immae: I verified /proc/cmdline contains ahci.mobile_lpm_policy=0 dm_mod.use_blk_mq=0 scsi_mod.use_blk_mq=0, but it just locked up again. You wouldn't happen to have any debugging tips?
sir_guy_carleton has joined #nixos
arjen-jonathan has joined #nixos
<FRidh>
Ericson2314: ok, good to know. I need to read up on the splicing first before making more changes I think
<{^_^}>
[nixpkgs] @globin pushed commit from @eonpatapon to master « nixos/prometheus-bind-exporter: add module »: https://git.io/fhY0M
<pie__>
sigh, looks like the binary cache is still unreliable
<Ericson2314>
you can add a `strictDeps = true;` if you want that to always be the case
<FRidh>
Ericson2314: I am a bit skeptical now that it was correct, considering I had some build-time dependencies such as `setuptools_scm` as `buildInputs`, not `nativeBuildInputs`. As long as they're Python-only, things will likely be fine, but it's not correct. Also, about the setup hook
<pie__>
have to restart the update process every cople tens of packages
<Ericson2314>
oh yes, that was incorrect
<FRidh>
addEnvHooks "$hostOffset" addPythonPath
<Ericson2314>
but cateogrizing the deps correctly is orthogonal to the splicing stuff i was talking about
sir_guy_carleton has quit [Ping timeout: 258 seconds]
<{^_^}>
esmerelda: There are multiple ways of managing declarative profiles. 1) Attrset, compatible with imperative use of nix-env https://git.io/fAQHW ; 2) buildEnv, providing more control over the paths that are linked into the profile https://git.io/fp0aU ; 3) home-manager, providing nixos-like config for your ~ https://github.com/rycee/home-manager
FRidh has quit [Quit: Konversation terminated!]
<sphalerite>
esmerelda: note however that this won't be able to create dotfiles and stuff for you
<esmerelda>
sphalerite: oh? I have an overlay that configures neovim for me
<Ralith>
I'm not super up on FIDO2 but I assume it's good also
<Ralith>
legacy yubikey functionality is mostly pretty ugly hacks
<{^_^}>
[nixpkgs] @astro opened pull request #53346 → libxslt: add libgcrypt to buildInputs for cryptoSupport → https://git.io/fhYuW
<esmerelda>
buildenv looks perfect for me, thanks sphalerite!
<ottidmes>
If you set say installPhase in a derivation, should you add runHook preInstall/postInstall? I got this as feedback, but I was wondering what was best practice in this regard, checking nixpkgs I do see some packages doing this, but a minority
<sphalerite>
esmerelda: right, but that works by creating a wrapper which calls neovim with the approrpiate config
<sphalerite>
esmerelda: not by creating a ~/.config/neovim/init.vim
sanscoeur has quit [Remote host closed the connection]
<{^_^}>
fusion809: Setting stateVersion to the latest release doesn't update anything and can only break your setup at best. If you want to regardless, Ctrl-F for "stateVersion" in https://nixos.org/nixos/manual/release-notes.html to see things that need to be manually upgraded with the new value
<Baughn>
So... how do I log in to a system that isn't responding to network, where there are no local passwords set?
<Baughn>
Normally I'd use `init=/bin/sh`.
<exarkun1>
I put m4, autoconf, etc in `nativeBuildInputs` and cross-compiled for armv6l and I got armv6l versions of m4, autoconf, etc, in the build environment. what the heck? I thought dealing with that was the whole point of `nativeBuildInputs`?
<dgarzon>
Hi, given I have an attribute set `project { name = “foo”; version = “latest” }`, is there any way I can override the `version` when calling `nix build -f .`? I tried doing `nix build -f . my-package --argstr project.version 0.0.1` with no success. Any ideas?
<immae>
tilpner: sorry, if you have it in /proc/cmdline then you did it correctly (at least as I intended), so it means only the symptoms were the same, but the cause is different :/
<{^_^}>
[nixpkgs] @matthewbauer merged pull request #52760 → lib/make-ext4-fs: more efficient store maker → https://git.io/fhUOU
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to master: https://git.io/fhY2H
<immae>
Then I have no clue
<tilpner>
immae: I also checked /sys/module/scsi_mod/parameters/use_blk_mq and the other paths, they contain N or 0, as intended
Acou_Bass has quit [Ping timeout: 246 seconds]
<tilpner>
immae: How did you first find that mailing list thread?
mtesseract has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<dgarzon>
Doing `nix build '(with import ./default.nix {}; project.overrideAttrs (old: { version = "0.0.1"; }))’ my-package` yields `error: attempt to call something which is not a function but a set, at (string):1:7`
<catern>
is there a way to do nix-store --import (or something similar) such that if there's a reference to a path which is not found in the store, Nix attempts to substitute that path, and only fails the import if it fails to substitute?
<catern>
that would allow distributing a small NAR which mostly pulls things from the binary cache
pie__ has joined #nixos
shibboleth has joined #nixos
ddellaco_ has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 merged pull request #53103 → pysnmp: add missing dep on pycryptodomex → https://git.io/fhLdZ
<yl[m]>
I thought `gccStdenv.mkDerivation` is supposed to give a C environment
griff_ has joined #nixos
Thra11 has joined #nixos
<{^_^}>
[nixpkgs] @KAction opened pull request #53355 → ###### Things done → https://git.io/fhYiX
<samueldr>
yl[m]: guessing here, but could it be because bundlerEnv isn't in "gccStdenv"?
<yl[m]>
oh
<yl[m]>
I'll try inlining the bundlerEnv in the buildinputs
rprije has joined #nixos
jackdk has joined #nixos
<yl[m]>
samueldr: it did not work :( (I updated the Gist if you want to take a look at the new shell.nix)
jackdk has quit [Remote host closed the connection]
jackdk has joined #nixos
<samueldr>
I'm not entirely confident in how those work, but I don't think "simply" putting it there changed it, see, I'm not sure, but there is no re-compilation for anything in buildInputs when put in that derivation, right?
sanscoeu_ has quit [Remote host closed the connection]
arjen-jonathan has quit [Ping timeout: 250 seconds]
dgarzon has quit [Quit: dgarzon]
Judson has joined #nixos
<Judson>
Did anything change e.g. with fetchGit recently? I'm getting "fixed-output derivation produced path '/nix/store/0fpa1wn561glgn39lhgly9hi1lb1vid7-puppet-lint-security-plugins-bd308dd' with sha256 hash '0x98xxns62yhf13irpljbjrnbn1pyzj09r9gf4a9n2cv2xa4a2ga' instead of the expected hash '144wykkacmvrn00qqlkb0wflq13hnyhsfvrd9jfxm7lwkqb5rs85'"
<{^_^}>
[nixpkgs] @pukkamustard opened pull request #53357 → ocamlPackages: Use lib.makeScope → https://git.io/fhYXx
<Judson>
Which is fetching from a digest ref, not a branch or tag
<gchristensen>
are you keeping .git?
<Church_>
Anybody who's used r2 before. Have you had problems with it attaching to a debugger?
<Judson>
Am I keeping .git? I haven't changed anything here - it's a shell.nix that includes a gemset that comes from a Gemfile that points to a git repo + ref.
<Judson>
(I know I know - it's not mine)
<gchristensen>
Judson: is keepDotGit (or something like it) passed to fetchGit ?
<Judson>
I don't believe so.
ddellaco_ has quit [Remote host closed the connection]
<Judson>
The gemset doesn't have anything like that, and IIRC the nix builders for gems don't
<Judson>
Is that a new flag?
<gchristensen>
no
cyounkins has joined #nixos
caryoscelus has joined #nixos
sigmundv_ has joined #nixos
Guanin has quit [Ping timeout: 252 seconds]
lord| has joined #nixos
BlessJah has quit [Remote host closed the connection]
<Church_>
samueldr: After I add something to my nixstore how would I make sure it gets picked up in the system path?
<Judson>
I use `bundle lock` first, to go from Gemfile to Gemfile.lock, then bundix for the gemset.
<Church_>
nix store*
<yl[m]>
Judson: thx
<yl[m]>
let me try that
<Judson>
yl[m], I assume that's too spare an answer. Let me know what I've missed and I'll elaborate
<grp>
hi, I'm having troubles with nixos-unstable: nix search -u testdisk is not listing anything. The package is there, and it's not allowUnfree related. I could nix-env -iA testdisk just fine, but I'm still wondering why won't it list with nix search
<yl[m]>
Judson: can you post your shell.nix?
<Judson>
See the gist above
<grp>
must get this working since trying to find stuff will be a pain if it's not listing everything there is
cyounkins has quit [Ping timeout: 244 seconds]
<yl[m]>
Judson: sorry about that, my network dropped and the message arrived too late :)
<Judson>
n/p
jtojnar has joined #nixos
<Church_>
Hmm binary still isn't being picked up. >_>
<grp>
and by the way, in case I should want to downgrade to nixos stable, should I just change the nix channel and rebuild? there's a different store version, unstable is at 19.03. Would that break stuff?
<yl[m]>
`rubyEnv.env` does not work
BlessJah has joined #nixos
<yl[m]>
this shell does provide me with a working ruby env