<wdanilo>
Hi guys! I've got a question. When writing nix package expression I see that in the "header" we define all dependenceis (libs) that are needed for the app to be run. However we do not define the verisons of these libs. On the other hand I know nix is always reproducible, so where we provide information that this app works with this and this version of dep lib ?
<emily>
you pin the version of nixpkgs you use, say
<wdanilo>
emily: oh, so nix somehow discovers the versions I use and it hardcodes them down somewhere?
<wdanilo>
emily: but does it mean that when I install a package foo-5.0 it could have other version of deps than if you install it ?
<samueldr>
is what is highlighted yellow what you called "the header"?
<wdanilo>
samueldr: yeah! I know these are the function arguments in the nix syntax
wdanilo has quit [Quit: Page closed]
<samueldr>
good, that's simply it, they are filled-in by callPackage, (e.g. in top-level/all-packages.nix) using the version currently in the nixpkgs version being used
<samueldr>
so the versions are not pinned across revisions of nixpkgs
<samueldr>
but inside one nixpkgs revision, they are coherent together^[1]
wdanilo has joined #nixos
<sphalerite>
samueldr++
<{^_^}>
samueldr's karma got increased to 23
<wdanilo>
samueldr: I'm sorry. My irc client crashed. Im back
<sphalerite>
although I think wdanilo missed a couple of those messages lol
<samueldr>
oh, I have joins/parts disabled due to the traffic
<sphalerite>
wdanilo: we have logs, see the topic :)
<samueldr>
[1]: the callPackage call can allow overriding specific attributes for others, e.g. qt5 could be specified as qt56 with `callPackage ../... { qt5 = qt56;}`
<samueldr>
(it helped though that you knew that it was a function call, less to go over)
<wdanilo>
samueldr: makes me think that all other packages installed on my system would just share all the same versions of `babl`, `gegl` , `gtk2` etc. Is it the case ?
<samueldr>
kind of, kind of not
<sphalerite>
all other packages built from the same revision of nixpkgs, yes
<samueldr>
^
<samueldr>
sphalerite++
<{^_^}>
sphalerite's karma got increased to 22
<sphalerite>
unless they're explicitly given something else in all-packages.nix
<sphalerite>
(unfortunately github chokes on the size of that file…)
<sphalerite>
or overridden manually from outside nixpkgs
<wdanilo>
sphalerite: ok! And who controls the "revision of nixpkgs"? Is it some kind of nix committee or something? Or users are choosing it ?
<samueldr>
you!
<samueldr>
(and the developers, by introducing new revisions)
<sphalerite>
revision ~= git commit
<sphalerite>
well no actually it *is* the git commit
<samueldr>
you can control it a couple of ways, they all end up being somehow used as NIX_PATH
<wdanilo>
samueldr: ok, so my current undrstanding is that this "revision" is some kind of global config for common packages versions shared across other packages. If so, does it mean that some of the packages cannot override only small part of these versions ?
<gchristensen>
it doesn't have to be global
<sphalerite>
ah crap is that the time? I really need to get my sleep rhythm sorted…
<wdanilo>
gchristensen: ok, all clear. So when I install gimp it depends on the version of gegl specified in this version of nixpkgs. The version of nixpkgs is one, for all the packages installed in my system right
<wdanilo>
? :)
<gchristensen>
no, you can have many versions of nixpkgs at once
<gchristensen>
a typical user: yes, has one, but there is no effort required to have multiple. just clone a different version of nixpkgs and install something.
<wdanilo>
gchristensen: ok! here was my misunderstanding. Than to have a completely clear picture of everything the last question! If I install gimp, how do I tell nix what version of nixpkgs to use and how can I be sure that other users would use the same version when I give them gimp pkg config ?
<gchristensen>
sorry, I have to go -- hopefully someone else can help -- otherwise I'll be back in ~10hrs :)
<samueldr>
wdanilo: usually it's your nix channel
<samueldr>
nixpkgs* channel
georges-duperon has quit [Ping timeout: 252 seconds]
<samueldr>
while it's not limited to nixpkgs, let's simplify by assuming it does for now
<wdanilo>
gchristensen: sure, thank you so much and good night!
<gchristensen>
good night!
<{^_^}>
Night!
<gchristensen>
thanks for taking it on, samueldr
<samueldr>
well, thanks for jumping in :)
<samueldr>
wdanilo: do you have nix or nixos installed? a hands-on example can sometimes help
<wdanilo>
you are all incredibly nice and helpful, thank you guys <3
<samueldr>
can't catch flies with vinegar... though not sure what we'd do with flies :/
<wdanilo>
samueldr: no, not yet.
<samueldr>
ah okay, it'll make more sense with a hands-on approach in the future
<wdanilo>
samueldr: I can install nix fast, give me 3 minutes and it will be here
kreisys has quit [Ping timeout: 244 seconds]
<samueldr>
but when you install nix or nixpkgs, you will have a channel configured, which is "like a pointer to a revision to nixpkgs"
<samueldr>
it can't readily be moved around freely, but it can be synchronized (via --update) to the current up-to-date state
<wdanilo>
samueldr: ok, I've got nix installed!
<samueldr>
when you rebuild a nixos system, it will use the (root's) current revision of the nixos channel as the revision
<wdanilo>
samueldr: (Im on a gentoo linux)
<samueldr>
so everything built *for nixos* using the declarative configuration will be coherent and cohesive
<samueldr>
you can, and this will apply to you outside nixos, also install stuff through non-declarative means, `nix-env -i`
<samueldr>
this will too, use a channel, this would usually be through something like nix-env -iA nixpkgs.hello
<samueldr>
where `nixpkgs` is the channel name
<samueldr>
when installing, it will install using the current revision where your channel points to
<wdanilo>
ok, so I can configure many "channels" and install packages using one of the configured channels. Where a channel is a set of coherent packages, sometihng like Haskell's stack LTS (of course you might not hear about it, but in case you did it seems related to me)
<samueldr>
if you update your nixpkgs channel, through nix-channel --update, it wouldn't touch the currently installed software, it'd need to be updated using nix-env -u
<samueldr>
you could, though I'm not knowledgeable in haskell things, sorry :)
<wdanilo>
sure thing!
<samueldr>
e.g. I am always building my systems against the current stable branch's channels, but my user on my system has the unstable branch channel to install random misc. things
<wdanilo>
So, using other words, a channel is a set of coherent package versions. So I can set up a channel and install everything knowing that the veversions "speak to each other". I can have different channels and install different packages in different channels. The purpose of such channel is to create a common gorund for different packages
<samueldr>
so my base system is probably not going to fail with random changes and version upgrades, but if I need something fresh, I can cheat a bit installing it non-declaratively
<wdanilo>
Do I get it right?
<samueldr>
channels are *one* way to track nixpkgs, but yes, your phrasing seems right
<wdanilo>
samueldr: awesome! Thank you so much for this clarification. It makes **much sense** now
<samueldr>
now, when using nix to build things outside of nixpkgs, you can do many other things; e.g. you can have a shell.nix or default.nix for a project which *itself* manages where the packages come from by pinning a nixpkgs revision
<wdanilo>
samueldr: Ok, might I have yet another question to you? Because I'm asking about nix for a very specific reason. We're making an open source project and we are thinking about using nix for managing packages for our users and Im looking for the best possible solution here
<samueldr>
do ask :)
<wdanilo>
samueldr: Thank you! <3
<wdanilo>
samueldr: So, we are doing Luna (http://luna-lang.org) - an Open Source, visual platform allowing you to process data and have an interactive visual repsponse on each step of computations
<wdanilo>
samueldr: Luna is an declarative, purely functional programming language, however for our end users (data scientists) it just look like an easy to use tool with nice graphical interface, interactive widgets etc
<wdanilo>
samueldr: We allow people to install luna libraries (new functionalities). So you can open Luna and tell - please install me new nodes I will be using in my grah and these nodes will speak with tensorflow or any otheer library
<wdanilo>
such Luna library sometimes requires shared C libraries available in your system. But we do not want our users to struggle with it and install it in their systems, we woudl liek to provide a full y automatically managable environemtn
<samueldr>
(still listening)
<wdanilo>
so the idea is that we can bundle nix with Luna and if our user installs a library and this library requires some system shared libraries, we jsut ask nix to install them for us
sir_guy_carleton has quit [Quit: WeeChat 2.0]
<wdanilo>
This way we could provide an awesome environment where all system libraries would be automatically managed, will not conflict with system wide librares, would be sandboxed and would work on any linux box
<samueldr>
sandboxed only at the filesystem level (something to keep in mind)
<emily>
(modulo some complications with OpenGL)
<wdanilo>
samueldr: nix seems like the perfect candidate for it. Please correct me if im wrong here. If im not wrong, we have one major issue iwth it right now. Namely, we were unable to run nix without root privileges and we've tried literally everything so far
<wdanilo>
samueldr: yeah, im looking for exactly this level of sandboxing. So if you are a Luna library author and you configure that it requires this and this shared library then it should work on your mahcine only if you provide the right lib configuratino (which nix provides)
<samueldr>
(don't quote me, the general response is true, but finer details may be inaccurate) to use nix and nixpkgs *with* the binary caches, it's generally assumed that the user will be able to at least make an initial setup with root privileges
<samueldr>
I think proot at one point worked, if it doesn't anymore, but at the cost of speed
<samueldr>
nix itself won't really care otherwise, but the store isn't much portable, so it kinda needs to exist at /nix/ or otherwise you need to rebuild everything to point to the right store paths
<wdanilo>
samueldr: we cannot require our users to have access to root. I've seen that we can ask nix to create the lcoal nix store. We were even able to download nix binary, use patchelf and other magic and make it moveable between linuxes, however when creating the local store
<wdanilo>
nix still triesto access the /nix/... libs
<wdanilo>
samueldr: the section "nix 2.0's native method"
<samueldr>
hmm, that's a hard question to answer fully, the best I can do is say "I don't know" :)
<wdanilo>
samueldr: I was not wexpecting full answer tbh. First I wanted to better understand how nix manages the packages (btw I'm slwoly moving form gentoo to nixos, so I hope I'll controbute back sooner than later!)
<samueldr>
hmm, maybe there were improvements with nix 2.0 that made this more possible (checking into what you linked)
hitchhikingcoder has joined #nixos
<wdanilo>
the hard question I've asked is the "root cause"" of the research so I wanted to show the whole picture hoping that maybe someone would point me somewhere :)
<wdanilo>
samueldr: anyway, using the link I've provided I was able to create local store in my home! But only having previously installed nix globally :(
<wdanilo>
samueldr: That's a good idea! If no one would know the answer here I'll create the topic there as well! :)
<samueldr>
wdanilo: yeah, it could probably work
<wdanilo>
samueldr: the problem is that in order to create local strone you need nix and the command creating the local stroe tries to access '/nix/store/mxg4bbblxfns96yrz0nalxyiyjl7gj98-nix-2.1.2'
<samueldr>
if I were looking into your issue, wdanilo, I probably would first check into building nix outside of nix, so not installing using the curl|bash way, and not a nix installed through nix
<samueldr>
I would bet that by building somewhat statically, or "more classically", it would help with most issus
<samueldr>
issues*
<wdanilo>
samueldr: we've been able to download just a binary from nix servers, patch it using patchelf and just run it. However it seems that the "/nix" path is hardcoded somewhere in the code of the command making local store
<wdanilo>
ok, I think I'll ask about it and write down everythign we've found out on the discourse then
<samueldr>
hmm, it's possible that it still needs "a store", and the default store is a compile time option...
<samueldr>
I don't know which mecanism is used for `nix run`, my bet without looking would be a user namespace
<wdanilo>
samueldr: yep! I was thinking exactly this. And yep, user namespaces are needed for local store
<emily>
wdanilo: I assume the problem here is just that your Nix binary expects to be installed into /nix/store
<emily>
wdanilo: it might be possible to produce a more portable binary for Nix itself and then use --store
<emily>
(but then you could also just compile your own Nix and set the default store path too?)
<wdanilo>
emily: that would solve all the problems I've got in my life!
lassulus has quit [Ping timeout: 245 seconds]
lassulus_ is now known as lassulus
<wdanilo>
emily: regarding setting my own default store path - Wwe want it to be in user's home directory, so we do not know upfront the absolute path
<samueldr>
I would bet keeping /nix as a store path, but using another path using --store has the advantage of using the binary cache (haven't verified)
<samueldr>
while setting another store path would mean compiling the whole world
<emily>
well but the fundamental problem is that a non-root user is never going to be able to mkdir /nix.
<samueldr>
though, I wonder if using --store and a portable-enough nix, it probably could work
<emily>
either you can chroot (possibly emulated with proot) or you're stuck recompiling the whole world, I think
<wdanilo>
samueldr: are you sure that using local store needs to compiel the whoel world? Basically I was just ble to create local store and a lot of packages were jsut downlaoded tere
<samueldr>
(assuming user namespace thing is good enough)
<emily>
wdanilo: right. so you just need a nix that looks for its libraries in, say, ../lib
<emily>
which ELF supports, I believe
<emily>
so you should set about trying to compile your own portable Nix.
<samueldr>
wdanilo: use nix run --store ~/tmp/foo and then list the contents of /nix/store
Supersonic has quit [Disconnected by services]
Supersonic112 has joined #nixos
<emily>
this is just what I'd do, and I'm no expert; YMMV and posting on the discourse with full details is a good idea
Supersonic112 is now known as Supersonic
<wdanilo>
samueldr: sure thing, brb
<samueldr>
it's mounted to /nix using what I bet (still haven't verified) is a user namespace
jperras has quit [Ping timeout: 252 seconds]
<emily>
samueldr: oh huh
<emily>
fancy
<samueldr>
"downloads (or if not substitutes are available, builds) the GNU Hello package into ~/my-nix/nix/store, then runs hello in a mount namespace where ~/my-nix/nix/store is mounted onto /nix/store.
<emily>
most distro kernels don't have user namespaces enabled though
<wdanilo>
what is then the point of having local namespaces ?
<samueldr>
it's kind of halfway between a chroot
<samueldr>
the process and all its children will have the path of --store mounted as /nix
<samueldr>
(which in theory should work even if /nix doesn't exist, if the feature is enabled for the end-user)
<wdanilo>
samueldr: oh, ok, but then it should work without having real `/nix` in the system right ?
jit10 has joined #nixos
<wdanilo>
samueldr: exactly
<samueldr>
I would *bet* that centOS and such distros have it disabled
<jit10>
hello does u2f work under google chrome for anyone in nixos?
<samueldr>
I wouldn't know the real marketshare of users that can use user namespaces
<samueldr>
jit10: not a user personally, but I think it should, and there are users; but it may need an option
<wdanilo>
samueldr: Hmm, you know - it is ok for us to tell that if you use such kernel (and as we've checked fedora, ubuntu have it enabled) then it works out of the box, if not you have to manage deps manually
<samueldr>
wdanilo: if you can ask the end-user to have user namespaces enabled, then it might be fine :)
<wdanilo>
samueldr: however if it works in the most popular distros, its ok for us
<samueldr>
wdanilo: there's the last question of whether nix can be portable or not
<jit10>
samueldr: that option is not really not relevant to u2f, it just adds bunch of udev rules that i have done in my configuration already
<samueldr>
(whether nix can work without the /nix store)
<emily>
wdanilo: user namespaces are not enabled anything close to universally across distro kernels I don't think
<samueldr>
jit10: then I don't know much more, but glad you had the udev rules already :)
<wdanilo>
samueldr: yep exactly. This is the problem we are strugglign with. How to make nix portable and hwo to make local storage without nix trying to access /nix
hitchhikingcoder has quit [Remote host closed the connection]
Mic92 has quit [Quit: WeeChat 2.2]
Supersonic has quit [Ping timeout: 260 seconds]
Supersonic has joined #nixos
<wdanilo>
samueldr, emily: anyway, guys, thank you so much for your time and help. I trully highly appreciate it. Thank you <3
oldandwise has joined #nixos
oldandwise has quit [Client Quit]
init_6 has joined #nixos
<Ralith>
jit10: having the correct udev rules istablished is extremely relevant to making u2f work
<Ralith>
established*
<jit10>
Ralith: it works in firefox
<Ralith>
great!
<jit10>
i needed chrome because i couldn't register key in firefox for google account
<vodurden>
Howdy Nixos channel. I've got a tricky problem that I'm not sure how to solve. I'm trying to package a proprietary binary into nixpkgs, I've managed to do the usual patchelf dance to get it linking and finding the shared libraries but the binaries also search for each other using a hardcoded path. Essentially all the binaries assume that they are installed under /opt/mudfish/4.4.6/ and attempt to launch eachother using that hardcoded
<vodurden>
path. Is there any way I can intercept this and make it point to the nix store instead?
<vodurden>
Currently I'm using symlinks which works but it's an ugly hack since if you install this package you'd have to manually set up the symlinks
<vodurden>
The other tricky bit is that the application attempts to write some state to /opt/mudfish/4.4.6/var, so ideally I'd like to redirect that as well to a location outside of the nix store
jit10 has quit [Ping timeout: 256 seconds]
<wdanilo>
I've got a little bit related question to the one vodurden just asked. Basically we all know that if we just change the nix store place without using kernel userspace support or chrooting, all the nix libraries would need to be recompiled. My question is why. Why we would not only need to use patchelf on them? Are many of the libraries using some deeply hardcoded paths or what is the reason ?
mizu_no_oto has quit [Quit: Computer has gone to sleep.]
DigitalKiwi has joined #nixos
Mic92 has quit [Quit: WeeChat 2.2]
Mic92 has joined #nixos
mizu_no_oto has joined #nixos
Mic92 has quit [Client Quit]
Mic92 has joined #nixos
<elvishjerricco>
wdanilo: In order to patchelf the outputs, you'd have to change the build script to do so. Changing the build script means the derivation is changed, meaning it won't be on cache.nixos.org. You'd have to introduce a phase outside the build process, in Nix itself, that did this work. And even then, the contents of the paths would be changed, meaning that any cryptographic signatures used to trust a path downloaded from
<elvishjerricco>
the internet would be useless.
silver_ has quit [Read error: Connection reset by peer]
abueide has quit [Ping timeout: 244 seconds]
<wdanilo>
elvishjerricco: yep, I was asking theoretically about an alernative (or enhanced) nix design. So sure, nix could have just a "pass" that does it. However I dont understand the second part of your sentence - because such patching with patchelf could be made after all signatures all checked, right ?
<elvishjerricco>
wdanilo: Signatures don't only need to be checked when you do the download. You need to be able to verify the current contents of your store at any time
<Ralith>
vodurden: some sort of user namespacing may be in order, as used by FHS envs
<Ralith>
that can let you fake paths with impunity
abueide has joined #nixos
<wdanilo>
elvishjerricco: hmm, ok, right. Still I feel this particular problem could be overcomed. I might be wrong though. Is this the only problem we are facing here or there are some libraries that even with using such special pass with patchelf would simply not work ?
<vodurden>
Ralith: Yeah that sounds like a good path to follow. Is there anywhere I can read up on how the FHS envs work or should I dive into the nixpkgs code?
<elvishjerricco>
wdanilo: Well ELF dependencies aren't the only thing you'd have to correct, and I'd feel FAR less confident attempting to automatically patch all-the-things. Chroot is just way more realistically, and perfectly effective
<Ralith>
vodurden: I don't know details, unfortunately; someone else here likely does, but the code is always the ultimate authority
<vodurden>
Ralith: No worries, thanks for the pointer regardless :)
<Ralith>
hopefully it leads somewhere mapped ^^
jperras has quit [Quit: WeeChat 2.2]
mizu_no_oto has quit [Quit: Computer has gone to sleep.]
<wdanilo>
elvishjerricco: the problem with chroot is that it requires root privileges. You can use buggy proot instead with decreased performance instead or require users to have kernels with userspace support. Every such option has some downsides so im just looking for other theoretical possibilities
<elvishjerricco>
wdanilo: Nix doesn't actually use chroot, as I understand it. It uses libseccomp to basically intercept syscalls.
<Ralith>
elvishjerricco: that doesn't sound right
<elvishjerricco>
Then I can be of no further help :P
<elvishjerricco>
I know Nix depends on libseccomp
<elvishjerricco>
and I know `nix run --store` works without root
<elvishjerricco>
and without multi-user nix
<elvishjerricco>
`nix run --store /some/absolute/path`, that is
<clever>
mount namespaces and usernamespacing
<Ralith>
nix loves it some fancy modern user namespacing support
<elvishjerricco>
ah
<elvishjerricco>
what's libseccomp for then?
<elvishjerricco>
sandboxing?
<wdanilo>
elvishjerricco: nix run --store uses unshare and kernel userspace support to mimic the behavior of chroot if im correct. it does not have performance overheads like interpeting the syscalls
<clever>
wdanilo: using mount namespaces without root, requires you to also use a user namespace
<clever>
which means the uid 0 inside the namespace, isnt the real root
<clever>
so setuid binaries are not actually getting root
<clever>
so even if you fool a setuid binary with a fake /etc/shadow, it gains you nothing
<wdanilo>
clever: Yep! thanks for the clarification!
<wdanilo>
Btw, clever, I remember some time ago you've answered a question of mine and you always know all the answers <3 I'd love to ask you if you have any receipt for trully installing and using nix without root privileges? I was able to download nix binary, patchelfd it and run it, however when trying to run `nix run --store` nix still tries to access `/nix`. It seems like something hardcoded in the sources :(
<clever>
i checked `mount` while inside the above `nix run` sandbox, and it only mounted /nix/store
<clever>
so /nix itself, along with profiles and the db are missing
<clever>
you need to use nix-user-chroot and/or nix-bundle
<clever>
this was an experiment i was doing, to enter a full nixos container (i think it even had systemd)
<clever>
and i was avoiding use of systemd-nspawn, so i could enter it without any special deps
mayhewluke has quit [Ping timeout: 240 seconds]
<clever>
nix-user-chroot from above would probably have helped a lot in those experiments
<elvishjerricco>
clever: Whoa
<wdanilo>
clever: do you know if it would work on macos
<wdanilo>
?
<elvishjerricco>
wdanilo: There's also issues with `nix copy` if you use different a different for the store. You won't be able to copy paths anywhere.
<elvishjerricco>
except other places using the same root path
<clever>
wdanilo: the stuff above is all using linux specific syscalls
<wdanilo>
clever: I was afraid of that
<wdanilo>
clever: is there **any** solution that would work on macos too ?
mayhewluke has joined #nixos
<elvishjerricco>
wdanilo: Make everything a wrapper to a Linux virtual machine? :P
<clever>
wdanilo: install docker, and then use a docker image with nix?
<wdanilo>
elvishjerricco, clever : actually I need a full speed there, so no virtualization is allowed :(
<elvishjerricco>
Docker being the somewhat serious version of my joke :P
<clever>
wdanilo: dual-boot to nixos
init_6 has quit [Ping timeout: 252 seconds]
<wdanilo>
clever: I'm trying to bundle nix with ourp roject. I cnnot tell our macos users to dual boot to nixos :P
<clever>
research what namespacing options darwin has, now the nix sandbox works, and write a variant of nix-user-chroot for darwin
<elvishjerricco>
wdanilo: On macos, doing the single user install requires root?
<wdanilo>
elvishjerricco: yes, as it needs to create /nix
<clever>
elvishjerricco: you need ownership of /nix/ to install nix without root
<clever>
wdanilo: you can manually create and chown /nix, and then the install script wont try to sudo
<elvishjerricco>
I didn't know macOS doesn't let users make things in /nix
<elvishjerricco>
in /, that is
<clever>
its the normal root owns / thing
<jasongrossman>
elvishjerricco: MacOS has a concept of "administrative" users, who I believe can modify things in /.
<elvishjerricco>
Ah
<jasongrossman>
elvishjerricco: I don't know for sure, but I don't think new users default to being Administrative.
<elvishjerricco>
wdanilo: Don't most package managers on darwin require root anyway? Is it too hard a sell that users should provide root access just once simply to create a directory?
hakujin has quit [Ping timeout: 272 seconds]
hakujin has joined #nixos
jtojnar has quit [Ping timeout: 244 seconds]
<jasongrossman>
elvishjerricco: Yes!
<jasongrossman>
elvishjerricco: I mean no! I mean most package managers on MacOS do indeed require root.
<jasongrossman>
MacOS programs are meant to go into /Applications, which is only writable by Administrative users.
<shreyansh_k>
Hi, are there any security implications of sharing configuration.nix file with hashedPassword for users? I mean, is it possible to decrypt passwords is people know hashedPassword field for users?
<emily>
depends how good the hash used is and how many rounds are used and how good the password is
orivej has quit [Ping timeout: 245 seconds]
<oldandwise>
chrony or systemd-timesyncd?
Fare has joined #nixos
<shreyansh_k>
emily: password is strong enough, hash generated with "mkpasswd -m sha-512".
<shreyansh_k>
emily: Is it safe?
<emily>
shreyansh_k: maybe pass --rounds too, I think the default is pretty low.
<emily>
at that point it's up to you how comfortable you feel with it.
<emily>
I think the risk is pretty low (as long as you're right that it's strong enough).
shreyansh_k has left #nixos ["Leaving"]
shreyansh_k has joined #nixos
Zer000 has joined #nixos
<shreyansh_k>
emily: Oh! cool. I'm sharing the config with friends so I'll be making the config publicly available in a github repo.
hakujin has quit [Ping timeout: 252 seconds]
<pie_>
i keep forgetting how to deal with this: i put "with import <nixpkgs> {}; python36.withPackages (ps: with ps; [ pyqt5 pyqtgraph ])" in my shell.nix but running it with nix shell does not yield a python with pyqtgraph in its import path??
<emily>
shreyansh_k: you can also do hashedPassword = readFile ./password-hash, I think?
<pie_>
and add the password file to your gitignore or something
Fare has quit [Ping timeout: 252 seconds]
<shreyansh_k>
I see. I'll do that. Thank you. :)
<pie_>
and make sure your password file has proper permissions :P
abueide has quit [Ping timeout: 260 seconds]
<shreyansh_k>
pie_: Yes
<shreyansh_k>
Also, for the currently visible hashed password, I've come to know its 5000 rounds.
<shreyansh_k>
So, does that mean I don't have to panic about what is already shared?
<pie_>
well it's a combination thing as emily said, if your password is 5 alphabetic characters it will be fast to crack even with a strong KDF (key derivation function)
<pie_>
ok so it looks like the issue is that the nix-shell is using the same python as the one in my env....how do i get the nix-shell python to take priority?
<pie_>
i mean i know that the end result needs to be that it comes first in $PATH but...
<shreyansh_k>
pie_: It's strong enough. That is one thing I know.
<pie_>
i guess its fine then? i'd still do the readfile thing though call me paranoid
<pie_>
or you mean you already shared something?
<pie_>
idk
<pie_>
you could change it to be safe
jedahan has joined #nixos
<pie_>
ok so it looks like nix-shell 's paths take precedence...which makes sense...so im lost, i guess im just not getting a new python??
<pie_>
but i did what the manual said..
<shreyansh_k>
pie_: I'll change it. Thanks.
Fare has joined #nixos
<pie_>
(i mean, its probably fine *shrug*)
hamishmack has joined #nixos
Fare has quit [Ping timeout: 252 seconds]
jasongrossman has quit [Ping timeout: 260 seconds]
<adisbladis>
pie_: You want something like https://ptpb.pw/T0km/nix in your shell.nix/default.nix
<pie_>
adisbladis, yeah sorry it wasnt clear, I know how to do that I just totally forgot that I need to
jedahan has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<pie_>
this while nix-build/nix-shell/nix-env with -p or from a file or not -p or heck knows distinction constantly messes me up
<pie_>
can we not make everything expect the same kind of input?
jasongrossman has joined #nixos
<clever>
pie_: nix-shell -p '(import ./shell.nix {})' would work with your previously wrong file
oldandwise has quit [Quit: leaving]
<pie_>
clever, yeah thats equivalent to what i had working with -p earlier
wdanilo has quit [Ping timeout: 256 seconds]
lopsided98 has quit [Quit: Disconnected]
lopsided98 has joined #nixos
init_6 has joined #nixos
dozn has quit [Remote host closed the connection]
iyzsong has quit [Ping timeout: 260 seconds]
<{^_^}>
[nixpkgs] @andrew-d opened pull request #47160 → libsndfile: Add patch for CVE-2018-13139 → https://git.io/fAdZS
<vodurden>
So, it turns out with enough patience it is possible to redirect hardcoded paths a binary is trying to use. I've got a working example using unshare, mount and a bunch of hacks
<vodurden>
Thanks again Ralith for the pointer in the right direction :)
<vodurden>
Aha! Where were you five hours ago :p That's much nicer
<vodurden>
I'll give that a go right now. Thanks!
<clever>
heh, your the one that started the previous convo on namespaces
<clever>
didnt notice what the original need for namespacing was
<vodurden>
No worries, at least I got to learn a bunch about how namespacing works
<clever>
also fun, /proc/<pid>/root is a magic symlink, that points to the root of that processes mount namespace
<vodurden>
Unfortunately the thing I'm trying to package still isn't working. It runs but it breaks the network setup such that the internet doesn't work. But I think that'll be tomorrows problem
<clever>
even if that dir is outside your own mount namespace
<clever>
it can also let you peek at private /proc mounts inside a namespace
<clever>
via /proc/<pid>/proc/net/foo
<vodurden>
Hah, I guess that's part of why you wouldn't want to use just namespaces for running untrusted code
endformationage has quit [Quit: WeeChat 1.9.1]
<vodurden>
Hmm, I don't seem to be able to get the redirect to work correctly with libredirect. It still fails to find the process
<alex``>
What is the difference between `nix build` and `nix-build`?
jD91mZM2 has joined #nixos
<alex``>
and what is the `nix build` equivalent to `nix-build --expr {expr}`
hakujin has joined #nixos
<cocreature>
alex``: iirc it’s "nix build (expr)"
<alex``>
same for `nix edit` ?
vandenoever has quit [Ping timeout: 244 seconds]
hakujin has quit [Ping timeout: 260 seconds]
<alex``>
with `nix build` I get the error
<alex``>
> warning: Nix search path entry '/home/alex/.nix-defexpr/channels' does not exist, ignoring
<{^_^}>
error: syntax error, unexpected $undefined, expecting ')', at (string):195:32
<alex``>
and `nix-build` no warning
init_6__ has joined #nixos
init_6__ has quit [Client Quit]
init_6 has quit [Ping timeout: 260 seconds]
hotfuzz_ has joined #nixos
hotfuzz has quit [Ping timeout: 272 seconds]
logzet has joined #nixos
NightTrain has quit [Ping timeout: 245 seconds]
<{^_^}>
[nixpkgs] @etu opened pull request #47162 → php: Refactor so we can upgrade PHP per platform → https://git.io/fAdcY
<alex``>
What is the sudo configuration for `sudo --login`?
<alex``>
sudoers
alex`` has quit [Quit: WeeChat 2.0]
Izorkin has joined #nixos
polman has quit [Ping timeout: 264 seconds]
ckauhaus is now known as ckauhaus[afk]
Ariakenom has joined #nixos
haitlah has quit [Remote host closed the connection]
ixxie has joined #nixos
abueide has joined #nixos
Zer000 has quit [Ping timeout: 246 seconds]
thc202 has joined #nixos
georges-duperon has joined #nixos
orivej has quit [Ping timeout: 245 seconds]
georges-duperon has quit [Ping timeout: 252 seconds]
ixxie has quit [Ping timeout: 252 seconds]
[Leary] has quit [Quit: [Leary]]
echel0n_ has joined #nixos
iyzsong has joined #nixos
orivej has joined #nixos
<shreyansh_k>
tab completions for nix commands (nix-env, nixos-* ... etc) don't work in fish or is it just me? google doesn't seems to return anything.
<sphalerite>
shreyansh_k: looks to me like nobody's written any
<Myrl-saki>
Welps, there goes my idea of installing fish. :P
Lears has joined #nixos
arjen-jonathan has joined #nixos
<arjen-jonathan>
Hi guys; I'm trying to work with a python virtualenv under a nix shell to get the right version of python. But I'm having trouble with mysqlclient.
<arjen-jonathan>
Which cannot find libmysqlclient.so.18
<arjen-jonathan>
I've added mysql.connector-c to the buildInputs of the shell.
<arjen-jonathan>
help?
<srhb>
arjen-jonathan: If you can share something minimally reproducible, more people may be inclined to help.
<{^_^}>
[nixpkgs] @vincentbernat opened pull request #47163 → znc: require network-online.target to start → https://git.io/fAdWx
pointfourone has joined #nixos
<pointfourone>
So, the first thing that the Nix manual says that it support multiple versions of a package. Say that I'm using a package 'A-1.0' and 'A-1.1' is released. I want to keep both the versions of that package but they depend on different versions of 'B'. How would that work out? Unless the name of the binaries of both A and B aren't different, how will these packages coexist?
<pointfourone>
I mean they can coexist due to the naming scheme of Nix but how will I choose a version to use? Would I need to refer to the absolute path everytime?
<symphorien>
you can only have one on your PATH. but for example with nix-shell it is easy to tell "drop me in a shell which has all the tools for this specific project on PATH", so you can have one PATH for each project and their specific dependencies. A bit like python virtualenv
iyzsong has quit [Read error: Connection reset by peer]
abueide has quit [Ping timeout: 245 seconds]
jtojnar has joined #nixos
<pointfourone>
symphorien: So, is it possible to switch to the different binary without invoking nix-shell?
iyzsong has joined #nixos
<symphorien>
can you give a concrete case ?
<symphorien>
you can invoke them by absolute path or by symlink
mayhewluke has quit [Ping timeout: 272 seconds]
<symphorien>
you may create a package which only contains a symlink called foo-other-version pointing to the other version
<symphorien>
or if they have different names, you can have both
<symphorien>
just like python2 and python3
<symphorien>
but python can only point to one version
adam-6520493 has quit [Ping timeout: 250 seconds]
<pointfourone>
symphorien: Ok, so I can create a package which just servers as a symlink to a different version of 'A' and call that instead. Got it. Thanks
mayhewluke has joined #nixos
pointfourone has left #nixos [#nixos]
logzet has quit [Remote host closed the connection]
<adisbladis>
kandinski: Installing python libs globally like that pollutes the environment
<adisbladis>
I think that's why you get some libs loaded from python2 and some from python3
hakujin has joined #nixos
<kandinski>
can you explain not so much what you're doing as what I was doing wrong and how your code fixes it?
<adisbladis>
kandinski: With they way I sent you you'll get a wrapped python interpreter that has the desired libs in it's closure
<adisbladis>
So each python interpreter has it's lib isolated
Lears has joined #nixos
<kandinski>
adisbladis: hmm, I'll copy it now, and try to understand it later
<kandinski>
what does the "ps: with ps;" mean?
hakujin has quit [Ping timeout: 272 seconds]
<adisbladis>
kandinski: python.withPackages takes a function which in turn takes an argument, `ps` is just a name I use as convention. `(ps: with ps; [])` is a function body. The `with` keyword scopes an attribute set in nix, so instead of having to write `(ps: [ ps.flake8 ps.virtualenv ])` you can reference them by their name in the ps attribute set.
<kandinski>
adisbladis: thanks, btw. I forgot to say that.
<adisbladis>
rawtaz: Merged :) I'll backport to 18.09.
<adisbladis>
pointfourone: Cache maybe?
<srhb>
pointfourone: (oh and the cached results only update if you use -u, but it will still just show what's in your version of nixpkgs, not what's in master, unless that's what you're actually using, and it probably is not)
<pointfourone>
srhb: I don't have nix-info. I'm using nix in a Ubuntu VM right now.
<symphorien>
pointfourone: then `nix-shell -p nix-info --run nix-info`
<{^_^}>
[nixpkgs] @peti pushed 2 commits to haskell-updates: https://git.io/fAdB3
slyfox_ has joined #nixos
<pointfourone>
A bunch of packages got downloaded but I still don't have nix-info.
<rawtaz>
adisbladis: nice, thank you. dont you want it to be tested on 18.09? i only tested it on master. although there's very limited testing, just making sure the vmware service starts, that copy/paste works, and that manually mounting shared folders work
<symphorien>
did you run nix-info from the shell nix-shell got you ?
palo has joined #nixos
<symphorien>
or from outside ?
<symphorien>
hum scratch that
<srhb>
pointfourone: That command should just spit out the output of nix-info
slyfox has quit [Ping timeout: 260 seconds]
<srhb>
Not put it on your PATH.
<{^_^}>
[nixpkgs] @peti pushed 2 commits to haskell-updates: https://git.io/fAdB4
<pointfourone>
srhb: I just wanted to know why 'nix search' doesn't show the name of the available package and it's version number like every other package manager does. How should I know which version of weechat is present in the repositories right now? The nix-env command is too slow and doesn't seem to work all the time.
<adisbladis>
rawtaz: If you feel like doing additional testing on top of 18.09 that would be great. Typically we (nixpkgs commiters) do some testing for backports but of course it's always better if actual users of the packages do as much testing as possible.
<srhb>
pointfourone: I am trying to explain that to you by showing you what's going on :)
<srhb>
pointfourone: The first thing I was going to show you was to find the commit of nixpkgs you are actually using
<srhb>
Let's say you're using the nixpkgs-unstable channel for instance.
<adisbladis>
rawtaz: Since the branch-off point for 18.09 was not that long ago master and release have not diverged too much yet
<srhb>
The next time you nix search (with -u) it would update the search results to have the packages from the now-updated nixpkgs-unstable
<srhb>
pointfourone: That doesn't change the output of nix search though, as you want
<pointfourone>
srhb: That's what I said before :)
<pointfourone>
I still don't get the version number of a package in the repo.
<srhb>
pointfourone: Yup. Unfortunately there's not a good switch for that. --json will help you
<srhb>
pointfourone: It is a bit of an ommission imo. I don't think there's an issue for it though, maybe you can make one?
<adisbladis>
rawtaz: I did the normal "smell test" stuff like testing if things build and if some user facing applications can be started. If you want to do some more testing I can wait. Just ping me on github/irc when you are done.
<srhb>
pointfourone: there are also other ways to get the "version" once you have the attribute. `nix eval nixpkgs.weechat.name` for instance
<pointfourone>
srhb: Sorry, but the list of open issues in the issue tracker are already more than I'd like to see. I've tried to use Nix more times than I'd like but I've always been turned off by insane defaults and a brain dead UI.
<srhb>
pointfourone: No need to apologize to me. It was only a suggestion on what you could do to ultimately help improve that UI that is frustrating you.
<pointfourone>
srhb: That still doesn't give me the version of the package. It just shows weechat-bin-env.
<srhb>
pointfourone: What doesn't? nix eval nixpkgs.weechat.name ?
<pointfourone>
Yeah.
<{^_^}>
[nixpkgs] @peti pushed 2 commits to haskell-updates: https://git.io/fAdBy
<srhb>
What was the output of nix-info? I'm still not sure which channel you're on.
<adisbladis>
You can specify `--json` to nix search for richer output
<adisbladis>
That includes version number among other things
<pointfourone>
srhb: I'm on nixpkgs-unstable because I'm not using NixOS right now.
<srhb>
pointfourone: Oh, yes, I see.
<srhb>
The weechat derivation apparently changed recently.
<pointfourone>
adisbladis: :D That still doesn't show the version of weechat. It shows the JSON output with a date instead of a version number.
<srhb>
Yeah, the change to master has made it harder to find the version of that package directly unfortunately.
<srhb>
At this point the easiest way to go is probably nix edit nixpkgs.weechat
<srhb>
There is a version attribute there.
simukis has joined #nixos
<adisbladis>
pointfourone: My `nix search` spits out 2.1 for weechat and some dates for related packages. Dates are usually used when a package does not have a stable release and is a git checkout.
<srhb>
adisbladis: Unfortunately, weechat is now a buildEnv
<srhb>
Hence the trouble getting at the version.
<srhb>
It could be enriched with it though, and probably should be
<adisbladis>
srhb: Oh right. I was checking out the release channel temporarily :P
<srhb>
Wow, edit doesn't even find it on master with this change. grr
<{^_^}>
[nixpkgs] @adisbladis opened pull request #47164 → weechat: Add version and meta to `buildEnv` package → https://git.io/fAdBj
<srhb>
adisbladis++
<{^_^}>
adisbladis's karma got increased to 5
<adisbladis>
Yaay, internet points <3
<srhb>
Great, the meta change fixes edit too.
<adisbladis>
srhb: That uses derivation.meta.position
<srhb>
Yeah :)
hexagoxel has quit [Quit: ZNC 1.6.5+deb1 - http://znc.in]
brandon_ has quit [Remote host closed the connection]
<adisbladis>
Probably a whole lot more of these around :/
hexagoxel has joined #nixos
<sphalerite>
ah jeez
<sphalerite>
it feels like I've broken so much stuff with the weechat plugin wrapper xP
<pointfourone>
Umm, what is difference between firefox, firefox-wrapper, and firefoxWrapper? Which should I install to get firefox?
<symphorien>
they are probably legacy aliases of one another
<symphorien>
firefox is fine
<{^_^}>
[nixpkgs] @srhb merged pull request #47164 → weechat: Add version and meta to `buildEnv` package → https://git.io/fAdBj
<rawtaz>
adisbladis: nah, it's fine, i have other things to focus on in that regard, i dont think it wont work, so go ahead when you can backport it :)
<srhb>
notlar: Check how?
<srhb>
notlar: You can open the nix file for most derivations using eg. `nix edit nixpkgs.hello` if that's your question
<sphalerite>
notlar: nix edit nixpkgs.foo
<rawtaz>
adisbladis: thing is, it only works to some extent, so even if we were to test it now it wouldnt be complete. better to do serious testing when we figure out the rest of the problems with open-vm-tools
<zduch4c>
how can I declaratively make Emacs be compiled with xwidgets support in configuration.nix, while also using emacsWithPackages?
<rnhmjoj[m]>
mkaito: i did but it isn't much of an improvement over opengl in my case (RX480 and mesa). i just installed the precompiled binaries with winetricks: nothing special to do on nixos.
b1000101 has quit [Ping timeout: 256 seconds]
<mkaito>
rnhmjoj[m]: oh I didn't know winetricks could do that. I was even trying to make a nixos package for dxvk lol
<zduch4c>
lmao
<zduch4c>
am I the only one seeing the spam?
pie_ has quit [Remote host closed the connection]
zduch4c has quit [Remote host closed the connection]
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
<sphalerite>
etu: zduch4c left
<etu>
:(
<etu>
I'm a bit slow cause I'm on the phone
<rnhmjoj[m]>
mkaito: yeah but i'd rather have a nice nix package that builds a wine prefix with everything i need. i always end up reinstalling the same stuff over and over
<mkaito>
rnhmjoj[m]: oh my winetricks is old, that's why I didn't see it. It was just added in July, my winetricks is from June.
<{^_^}>
[nixpkgs] @peterhoeg opened pull request #47166 → crystal: 0.26.1 and introduce 0.25.1 → https://git.io/fAdEl
<rnhmjoj[m]>
mkaito: you can download the latest release from github: it comes with a winetricks "verb" to install it
<mkaito>
rnhmjoj[m]: I saw those, but I jumped to "nothing ever works on nixos" conclusion, tried to build and package it for nixpkgs, failed, gave up :P
<sphalerite>
mkaito: wouldn't just editing the existing winetricks expression be good enough?
<mkaito>
sphalerite: you misunderstand, I tried to package dxvk from source and failed because I honestly don't really C++ and it wasn't finding some dependencies.
<mkaito>
but yes, now that the latest winetricks includes it, that's pointless.
<srhb>
The unit size is debatable. Interdependent separate commits are usually more of a pain to keep track of through the branches than a single non-dependant unit.
mayhewluke has quit [Ping timeout: 246 seconds]
<mkaito>
true. also, any given standard is better than everyone doing what they personally prefer. the above are my preferences for most things.
<adisbladis>
I guess it's one of those things where there has never been a need for a strict policy
<{^_^}>
[nixpkgs] @joachifm pushed 7 commits to release-18.09: https://git.io/fAdus
<mkaito>
as someone who's had to bisect nixpkgs a few times... I wish there was :P
alex`` has joined #nixos
<{^_^}>
[nixpkgs] @peti pushed 2 commits to haskell-updates: https://git.io/fAduR
<Myrl-saki>
mkaito: Is that from Eve no Jikan??
<mkaito>
yep
<Myrl-saki>
Nice. :D
<Myrl-saki>
Aaaa, where's the Eve no Jikan that I downloaded.
alex`` has quit [Client Quit]
<Myrl-saki>
mkaito: How long does it take for you to build Wine?
<mkaito>
3 minutes or so
<Myrl-saki>
Wat
<mkaito>
ryzen ftw
<Myrl-saki>
What are you building it on?
<Myrl-saki>
Oh, no wonder.
<mkaito>
I had an i5 but work had me debugging haskell builds on nix, which resulted in a lot of 7-8h rebuild cycles, so I got an upgrade that cut it down to 3-4h cycles.
<mkaito>
it's also useful for gaming tho :P
<Myrl-saki>
mkaito: Yeah, I'm actually thinking about getting the R7 equivalent TR.
<Myrl-saki>
OTOH, I feel like once I do, I won't be upgrading for a few years, so I might as well just get the R7.
<Myrl-saki>
Not sure
<mkaito>
I purposefully didn't get a TR because the R7 actually outperforms it quite solidly on anything that uses 6 cores or fewer
<clever>
mkaito: somebody else made a commit adding python3
<mkaito>
> 22 days ago
<{^_^}>
attempt to call something which is not a function but an integer, at (string):195:1
<mkaito>
oh ok
<Myrl-saki>
lol
<mkaito>
lol
<clever>
i guess i dont update often enough :P
<Myrl-saki>
clever: I've had a 3-month old 18.03
<Myrl-saki>
clever: I think there was no package at all that didn't change.
<Myrl-saki>
mkaito: How do you build with tmpfs?
<Myrl-saki>
Aside from a dedicated NixOS build machine.
<mkaito>
I think nixos builds in /tmp, so tmpOnTmpfs option probably.
<Myrl-saki>
Actually, I think a dedicated Nix machine would be better for building than a dedicated NixOS building machine since you don't have to early mount /nix/.
<Myrl-saki>
Oh, true.
<mkaito>
on my old i5, tmpfs cut the wine build time from 50m to 15m
<{^_^}>
[nixpkgs] @joachifm pushed commit from @luke-clifton to master « dvtm: refactor (fix on darwin) (#47103) »: https://git.io/fAdzr
<mkaito>
I've tested #46738 locally. builds and runs correctly. someone make a call on whether or not the commit should be split, then it can be merged if no other concerns crop up.
<i-am-the-slime>
Ah, so I guess I can just try to gradually go backwards and hope that it builds eventually? Is a revision a git commit hash?
<infinisil>
yea
<i-am-the-slime>
can I also do something like HEAD~2?
<i-am-the-slime>
Ah no I see it's a tar.gz
dmc has quit [Quit: WeeChat 2.2]
<rawtaz>
wazzaaaahp?!
dmc has joined #nixos
<i-am-the-slime>
infinisil: It's already going for a while, I'm hoping for the best!
<i-am-the-slime>
ah no, error, let's try an older one
<i-am-the-slime>
Sometimes I wonder what's the problem with just statically compiling things.
iyzsong has quit [Ping timeout: 245 seconds]
<infinisil>
i-am-the-slime: Closure size. The fact that pretty much everything is dynamically linked saves a lot of size when you have lots of packages sharing libraries. And with Nix it's trivial to properly fetch library dependencies for everything to work smoothly
<i-am-the-slime>
Okay so pointing to an old version 17.09, I could install ghc-mod. That one was actually smooth. It was compiled by the wrong ghc version, but who cares.
dmc has quit [Quit: WeeChat 2.2]
sir_guy_carleton has quit [Quit: WeeChat 2.0]
Acou_Bass has joined #nixos
dmc has joined #nixos
acarrico has quit [Ping timeout: 252 seconds]
<Orbstheorem>
Hello, I'm trying to setup wireshark for unpriviliged users on a nixos system. The way it's usually done is by giving capabilities to dumpcap, setting 770 permissions, and moving it to the authorized group. In nixos, given all paths are in the store and they're owned by root, I cant implement this authorization model. What's the correct way to authorize a linux group to realize network captures on nixos?
acarrico has joined #nixos
<teto>
I just want to build packages from another (nixos) VM instead of mine. Can I do this without hydra ?
<Orbstheorem>
teto, yes, there's an option on (`nix build`?) to have remote builders
<zduch4c>
in here I see `assert withXwidgets -> withGTK3 && webkitgtk != null;` so it must be possible, the question is how do I make use of this in configuration.nix
<clever>
zduch4c: you need to install (emacs.override { withXwidgets = true; })
<sphalerite>
zduch4c: you also need to stay in the channel longer if you want your questions answered!
zduch4c has left #nixos ["Killed buffer"]
<clever>
like that? lol
zduch4c has joined #nixos
<clever>
though i have a feeling he is using the emacs irc client, and has to restart to update that
<Orbstheorem>
Where can I find documentation on the order of evaluation of nixpkgs? (e.g. stdenv, then top-level, then modules, then configuration.nix, then overlays, then...)
<zduch4c>
i had to restart because emacs bugged out
<{^_^}>
[nixpkgs] @xeji pushed commit from @rbasso to master « exercism: 3.0.6 -> 3.0.9 (#47170) »: https://git.io/fAdwR
<siers>
How can I get nix version from in nixlang?
<siers>
s/from //
<sphalerite>
Dezgeg: zduch4c: they knew exactly what they were doing xD
<srhb>
siers: builtins.nixVersion
<clever>
> builtins.langVersion
<{^_^}>
5
caglan has joined #nixos
<caglan>
Hello all! I'm a newbie NixOS user. I'm having trouble compiling C++ programs with shared libraries(SDL2). I installed the package but gcc cannot find -lSDL2 in my system. What am I doing wrong?
<{^_^}>
caglan: Don't install libraries through nix-env or systemPackages, use nix-shell instead. See https://nixos.wiki/wiki/FAQ/Libraries for details.
<zduch4c>
caglan: nix-shell -p SDL2
<zduch4c>
then either make or gcc
<caglan>
Wow! It finally works! Thanks a lot everyone.
<{^_^}>
[nixpkgs] @vbgl pushed commit from @Zimmi48 to master « dune: 1.1.1 -> 1.2.1 »: https://git.io/fAdwh
<{^_^}>
"Mount filesystems on demand. Unmount them automatically.\nYou may also be interested in afuse.\n"
<sphalerite>
zduch4c: I'm not sure about automatic mounting without a desktop environment, but I believe udisks is what handles the mount requests from GNOME, and you can also invoke it using udisksctl
<Dezgeg>
I think it's rather like udisks yes
alex`` has joined #nixos
<{^_^}>
[nixpkgs] @pschuprikov opened pull request #47171 → Vim clang fix → https://git.io/fAdr8
arjen-jonathan has joined #nixos
<{^_^}>
[nixpkgs] @bobvanderlinden opened pull request #47172 → [WIP] spotifyd: init at 0.2.2 → https://git.io/fAdrX
caglan has left #nixos ["ERC (IRC client for Emacs 25.3.1)"]
<{^_^}>
[nixpkgs] @xeji pushed commit from @bachp to master « gitlab-runner: 11.2.0 -> 11.3.0 (#47167) »: https://git.io/fAdrA
orivej has joined #nixos
<zduch4c>
so how much does webkit take to compile
<zduch4c>
this is already 2 hours
<adisbladis>
zduch4c: In my experience a webkit compile means it's bed time
<zduch4c>
damn
periklis has joined #nixos
<zduch4c>
they say an i7 compiles it in one and a half hours
<zduch4c>
I have a Centrino Duo…
<emily>
my condolences
<{^_^}>
[nixpkgs] @michaelpj opened pull request #47174 → sddm: link whole sddm directory, not just themes → https://git.io/fAdoO
Fare has joined #nixos
<{^_^}>
[nixpkgs] @orivej-nixos pushed commit from @orivej to master « yed: 3.18.1 -> 3.18.1.1 »: https://git.io/fAdoE
mizu_no_oto has joined #nixos
<avn>
zduch4c: webkit or chromium? Webkitgtk builds ~2 hours, chromium take ~9 on 4 cores of amd and spinning disk.
<avn>
(and it matter, because chromium depends from almost all library on system)
b1000101 has quit [Ping timeout: 256 seconds]
<greymalkin>
So I have a problem -- trying to upgrade a long-standing machine that's running 16.03 (which was unstable at the time) -- now it's telling me that it doesn't know how to upgrade nix.
<jtojnar>
etu: I pushed some fixes, will continue later
<sphalerite>
greymalkin: there is an option that avoids going version by version — boot into a recent installer, mount everything, and rerun nixos-install
mizu_no_oto has joined #nixos
ajs124 has quit [Client Quit]
<clever>
greymalkin: why did the directions i gave not work?
<steveeJ>
seems like qemu (3?) ignores da drive's index field. I'm once again trying to make buildVmWithBootloader work here and this is my most recent conclusion.
<clever>
oh, that explains why i never saw it in the docs, lol
<clever>
steveeJ: and you must have tried -boot order= already?
<steveeJ>
when I flip the two mkDisk... lines the drives are in the correct order here *for boot*, but the rest of the qemu-vm.nix is borked cause it expects it the other way
<steveeJ>
clever: I have. you can only choose in groups of floppy, disk, cdrom where it will select the *first* of each
<steveeJ>
it seems like a qemu bug to me but I'm a bit lazy to dig into the qemu code
<Dezgeg>
maybe setting bootindex=0 to the relevant virtio disk helps
<clever>
steveeJ: i also see a bootindex param in the docs
<steveeJ>
AFAIU that's for controllers
<steveeJ>
but these disks are on the same controller
arjen-jonathan has quit [Ping timeout: 252 seconds]
<steveeJ>
I can try of course.. brb
<Dezgeg>
no, virtio has 1 disk per "controller"
<clever>
steveeJ: your screenshot shows 2 pci devices
patrl has quit [Ping timeout: 252 seconds]
<Dezgeg>
I guess with virtio-scsi you could have multiple LUNs on one virtio-scsi device, but virtio-blk shouldn't work like that
<etu>
jtojnar: Ok, I put a comment in there about some python3 thingy needed
<clever>
etu: steam or osomething else?
<steveeJ>
Dezgeg: buildVmWithBootloader is not using SCSI
<Dezgeg>
I know
<steveeJ>
cool
<etu>
jtojnar: A gnome shell plugin named gsconnect
<etu>
clever: ^
<etu>
clever: We tried to package it up some months ago and gave up, it was not possible at the time. But there's been a major rewrite to make it more friendly for packaging.
<etu>
clever: Partly because ubuntu wanted to ship it in 18.10 but won't because it was impossible to package :D
<steveeJ>
and Block format 'qcow2' does not support the option 'bootindex' is probably not a surprise to y'all either
<clever>
ah
<clever>
steveeJ: one min
<Dezgeg>
maybe you need to split the -drive if=virtio to a separate -device virtio-blk-pci one
<clever>
steveeJ: almost got an example, these logs are a bit slow to search
<jmsb>
hey guys, i am trying to connect to a raspberry pi via a usb-ttl cable. i can vouch for the cable as it works on os x. i see a /dev/ttyUSB0 when i connect the cable, but when i call the screen command im not seeing anything. anyone have any insight as to how i can probe whats going on?
<clever>
jmsb: is the baud rate correct? is the remote end running a getty on the serial port?
<{^_^}>
[nixpkgs] @WilliButz opened pull request #47181 → nixos/release.nix: run test for pgjwt → https://git.io/fAdPO
<jmsb>
baud rate is the same as when i did some trials on osx
<clever>
jmsb: what about with minicom, as root?
<steveeJ>
I'm wondering why we're only installing the i368 qemu manual
<clever>
might even explain why libvirtd went the more complex way
<jmsb>
@clever: also nothing. command im running is `minicom -b 115200 -o -D /dev/ttyUSB0`
<clever>
jmsb: do you have a scope?
<jmsb>
nope :)
<clever>
jmsb: try the menu in minicom to play with some of its options, no hardware flow control
justan0theruser has joined #nixos
<jmsb>
clever: thank you for the guidance!
justan0theruser has quit [Ping timeout: 246 seconds]
<{^_^}>
[nixpkgs] @xeji merged pull request #47181 → nixos/release.nix: run test for pgjwt → https://git.io/fAdPO
<{^_^}>
[nixpkgs] @xeji pushed commit from @WilliButz to master « nixos/release.nix: run test for pgjwt (#47181) »: https://git.io/fAdPQ
<steveeJ>
clever, Dezgeg: yay, that worked! now I'm trying to fix the SCSI code too but it fails in the VM that writes to xchg/in/vm/exit with no error. code is 1
<clever>
steveeJ: thats the 9plan fs sharing stuff
jmsb has quit [Ping timeout: 245 seconds]
<steveeJ>
clever: the sharing works, but the VM exits with 1
<clever>
ah
<steveeJ>
but I can't get it to print the output
mizu_no_oto has quit [Quit: Computer has gone to sleep.]
<steveeJ>
that would be "vm-run-stage2"
notlar has quit [Quit: notlar]
reinzelmann has joined #nixos
<clever>
steveeJ: i think i saw something about the build-in-vm script leaving a shell script behind in $NIX_BUILD_TOP
<clever>
and if you re-run that after using --keep-failed, it will drop you into a shell in the vm
<steveeJ>
I've tried that, gives me all sorts of permission errors and I'm too lazy to copy the image somewhere writable
<clever>
steveeJ: you can use qemu-img to create a writable qcow2, that uses the readonly qcow2 as a backing file
<clever>
then boot with that instead
justan0theruser has joined #nixos
<steveeJ>
clever: do you know how I can access the serial device it logs to?
abueide has quit [Ping timeout: 245 seconds]
<clever>
steveeJ: when -nographic is used, the serial port links to the stdio of qemu
<steveeJ>
ah right, there was '-serial pty' hidden in the QEMUOPTS
justan0theruser has quit [Ping timeout: 244 seconds]
<steveeJ>
the SCSI code path is broken. qemuCommandLinux in pkgs/build-support/vm/default.nix has virtio hardcoded and breaks if grub tries to install to /dev/sda
periklis has joined #nixos
slack1256 has quit [Remote host closed the connection]
justan0theruser has joined #nixos
periklis has quit [Ping timeout: 240 seconds]
zduch4c has joined #nixos
<zduch4c>
AHAHAHAHAHA GUYS
<zduch4c>
clever:
<zduch4c>
MY WACOM TABLET WORKS NOW
<zduch4c>
it was all the fault of the pesky GNOME
<zduch4c>
now that I'm not using GNOME, it all works!
<{^_^}>
[nixpkgs] @bhipple opened pull request #47182 → mkl: init at 2019.0.117 → https://git.io/fAd1i
<pointfourone>
Which branch does nixpkgs-unstable use?
<etu>
In which repo?
<pointfourone>
etu: I'm using Nix in an Ubuntu VM right now. nix-channels --list says that I'm using nixpkgs-unstable. Am I tracking master or some other branch?
justanotheruser has joined #nixos
patrl has joined #nixos
<clever>
patrl: the nixpkgs-unstable channel is the latest version of master to pass a certain set of tests
<EternalZenith>
Do I have to manually override the variables mentioned in the package definition?
<JonReed>
EternalZenith: In configuration.nix put (polyBar.override { i3Support = true; })
<JonReed>
EternalZenith: In packages
<JonReed>
EternalZenith: s/polyBar/polybar
<JonReed>
Same for all other support that you need
<JonReed>
Make sure that you `nix-env -e polybar` if you have installed it using `nix-env`, otherwise you can launch by accident polybar without support.
deltasquared has left #nixos [#nixos]
<EternalZenith>
Does it not automatically detect that you have those things installed?
<JonReed>
EternalZenith: No, it doesn't.
<EternalZenith>
I'd have thought that "assert i3GapsSupport -> ! i3Support && jsoncpp != null && i3-gaps != null;" would enable support automatically
<EternalZenith>
Unless that means something else
<clever>
EternalZenith: that just makes things like jsoncpp being null a fatal error, if i3GapsSupport is true, i believe
<EternalZenith>
Oh, that makes more sense
<ixxie>
anybody got a working config with xmonad + gnome?
<{^_^}>
[nixpkgs] @xeji opened pull request #47185 → nixos/tests/prosody: use sqlite database → https://git.io/fAdDy
<JonReed>
EternalZenith: I think it checks if support is enabled that nixpkgs also needs to contain support for it. Since the expression is called using `callPackage` it would automatically supply that support or it would default to null. If it defaults to null and support is enabled, then assert breaks execution.
<{^_^}>
[nixpkgs] @xeji pushed commit from @Mic92 to master « luaPackages.luadbi: 0.5 -> 0.6 (#47156) »: https://git.io/fAdDF
<{^_^}>
[nixpkgs] @xeji pushed commit from @Mic92 to release-18.09 « luaPackages.luadbi: 0.5 -> 0.6 (#47156) »: https://git.io/fAdyL
<EternalZenith>
Is there anything I should do aside from reading the NixOS/Nix/Nixpkgs manuals that will help make NixOS easier to maintain and configure on a laptop?
<steveeJ>
clever, Dezgeg: any idea why /dev/sda doesn't exist even though "[ 3.456208] scsi 0:0:0:0: Direct-Access QEMU QEMU HARDDISK 2.5+ PQ: 0 ANSI: 5" is in the dmesg?
<clever>
steveeJ: you may need to run something like udev trigger or settle to update /dev/
<EternalZenith>
clever: I don't see anything about my laptop on there
<clever>
EternalZenith: youll probably want to file a pr then, once you find out what special stuff it needs
<{^_^}>
[nixpkgs] @samueldr opened pull request #47187 → U-Boot: Add Orange Pi Zero Plus2 (H5) build → https://git.io/fAdyl
<EternalZenith>
clever: I will once I know if any issues appear
<pointfourone>
So, the nix-env man page says that if I don't mention the -P flag before installing a newer version of a package, it will be removed. However, if I mention the flag, the build may not succeed. So if I want to preserve an older version of a package, what should I do before installing a newer version?
<steveeJ>
clever: the system is using devtmpfs for /dev, that should do the trick
<{^_^}>
[nixpkgs] @tomberek opened pull request #47188 → gnuradio: reverse PYTHONPATH order → https://git.io/fAdy0
<EternalZenith>
clever: I'm using a Precision 3510, and haven't had any weird issues with it before aside from failures to resume from suspend at times
<clever>
pointfourone: all nix-env operations are atomic, if the command fails, it will not make any changes to the profile
<pointfourone>
clever: But what if I want to keep both the versions?
<clever>
pointfourone: why do you want to install both?
<pointfourone>
clever: I think one of the primary highlights of Nix is the ability to install and use multiple versions of packages at the same time. It's one of the main reasons why I've started with Nix.
<clever>
pointfourone: nix run or nix-shell would be better for that i think, since nix-env will have collisions on the binaries
<pointfourone>
clever: But nix-shell is used for testing packages in an isolated build environment right?
<JonReed>
EternalZenith: I would leave comments about parts of the config that is specific to laptop/system. This will make it easier to switch when a year later you'll try to figure out what you need or don't need anymore. You can even put it in a separate configuration as `imports = [ ./hardware-configuration.nix ./configuration-laptop.nix ]`.
<steveeJ>
clever: any idea which udev trigger util I can use in that minimal VM?
<clever>
pointfourone: that is one thing it can be used for
<clever>
steveeJ: cant remember off the top of my head, check the stage-1-init.sh in nixpkgs
i-am-the-slime has quit [Quit: Konversation terminated!]
<pointfourone>
clever: So, nix-env won't even succeed if I try to preserve an older version of package while doing an upgrade?
<clever>
pointfourone: yeah, it wil likely fail because there are 2 binaries with the same name in the profile
<clever>
pointfourone: the conflicting version stuff only really plays nicely when dealing with multiple users, or in the build or shell environments
mayhewluke has quit [Ping timeout: 245 seconds]
Fare has joined #nixos
EternalZenith has quit [Remote host closed the connection]
klntsky has quit [Ping timeout: 256 seconds]
mayhewluke has joined #nixos
<pointfourone>
clever: Let's say I'm using emacs-25.3 and emacs-26.1 comes out. I want to keep emacs-26.1 for a few days but I don't want to remove emacs-25.3. Is this possible with Nix? I mean if name of the binaries contains the version of the package, it can coexist right?
<{^_^}>
[nixpkgs] @LnL7 opened pull request #47189 → ffmpeg: don't use gcc on darwin → https://git.io/fAdyx
<alex``>
How to get the nix packages for a command?
<clever>
pointfourone: rollbacks and generations already do exactly that
<clever>
pointfourone: when you install or remove anything with nix-env, it keeps the old profile, and you can use `nix-env --rollback` to undo the last operation
<alex``>
For example: notify-send, when the command is not available, Nix will tell me the packages providing the command, but only for Bash.
<alex``>
I want it for Elvish
<clever>
alex``: nix-locate is the backend for that, and bash has a command-not-found script somewhere that runs it
klntsky has joined #nixos
<pointfourone>
clever: rollbacks are just for going back to an older version. It won't let me open and use both the versions at the same time right?
<alex``>
And for the PATH?
<alex``>
I need to set it from Elvish
<clever>
pointfourone: if you run `realpath $(which emacs)` you can see the absolute path for emacs
<clever>
pointfourone: and you can use that path to run that version, at any time
<alex``>
or, environment
<alex``>
in configuration.nix
<alex``>
How bash PATH is set?
<alex``>
and why is it bound to Bash instead of environement option from Nix?
haslersn has joined #nixos
<steveeJ>
clever: /nix/store/70k1sw727s6lcwgfaff02axbnsia4rz6-stdenv-linux/setup: line 1261: /proc/sys/kernel/hotplug: No such file or directory
<steveeJ>
this system doesn't have anything to make it work :-D
<clever>
steveeJ: you may need to mount /proc
<alex``>
I also have issue with file which doesn’t exist, but it exists actually
<pointfourone>
clever: that would just give me the absolute path of the currently installed emacs binary, i.e., 25.3. What about 26.1? How do I use that? You're saying that nix-env won't even be able to successfully install emacs-26.1.
<haslersn>
Hi :) Does anybody know: How, in my system config, can I install a package from a nixpkgs fork (which is hosted on GitHub)? Can I do it without specifying the commit SHA?
<alex``>
in sddm.extraConfig
<alex``>
I set directory to /home/alex/repositories/github.com/Foo/
<clever>
pointfourone: nix-env can only have one version of emacs installed in the profile at once, but several emacs can live in /nix/store
<alex``>
and theme to Bar
<alex``>
it said /home/alex/repositories/github.com/Foo/Bar/Main.qml doesn’t exist
<alex``>
I don’t understand why
<JonReed>
pointfourone: When you build something, it's stored in `/nix/store`. A binary for that package will be in `/nix/store/<package-hash>/bin`. You can launch it from there manually.
<steveeJ>
clever: proc is mounted. I'm afraid that kernel doesn't have hotplug enabled
<clever>
steveeJ: hh
<pointfourone>
clever: Yes, that's what I want to know. How would that work? The man page says that older versions will be removed unless -P is specified but if I do that, nix-env will fail. Then how can several emacs come in /nix/store?
<clever>
pointfourone: they are only removed from the newly made generation of the profile
<clever>
pointfourone: they remain in the store, and you can use rollback or absolute paths to access them
<pointfourone>
clever: Oh, ok, only the symlink will be removed but the package will be kept?
<infinisil>
(Consider using one of these sites instead)
<infinisil>
That's a weird error
<haslersn>
infinisil: I had a system where your solution worked; at least during nixos-rebuild. Now I run the installer on another laptop and I get that error (during nixos-install)
<clever>
haslersn: that tends to happen if you use builtins.fetchTarball during nixos-install
<clever>
infinisil: only major issue, is that the "rescue" env is updated on every nixos-rebuild
<clever>
so if your main nixos upgrades to a bad zfs that corrupts things, the rescue env does as well
<infinisil>
I guess one could use a pinned nixpkgs to prevent that
<clever>
but if you pinn the pkgs.path on lines 3&5, you could avoid that
<infinisil>
Instead of pkgs.path
<clever>
yep
<infinisil>
Neat
ixxie has joined #nixos
<clever>
the config also has to be altered if /boot is on zfs
<avn>
clever: Just curious, if it can be expanded to provide boot/rescue images by this way.
<ixxie>
anybody go powerline-go running? I can't quite get it enabled on zsh and I am wondering if the official instructions - https://github.com/justjanne/powerline-go#zsh - need to be tailored for NixOS
<haslersn>
clever: Does that also work with disk encryption?
<clever>
haslersn: i originally used rescue_boot.nix on my laptop (with luks encrypted zfs) to move /nix/store into its own zfs dataset
patrl has quit [Quit: WeeChat 2.2]
<clever>
avn: probably
<neonfuz>
I'm getting "glxChooseVisual Failed" when trying to run steam, I use bumblebee for graphics
<clever>
avn: that lets you netboot into the installer
<avn>
Although I like to have my travel notebook have ability make recsue medias, because my main host who pose as builder/netboot server etc sometimes can be down ;)
<clever>
avn: netboot_server in its current configuration, is actually perfect for a laptop
<clever>
avn: it runs a dhcp+ipxe+dns server on the ethernet jack, with NAT into the wifi
<clever>
so you can just jack into any machine and netboot nixos, and leech the laptop wifi
<steveeJ>
why to tests not have to reference "makeTest" by path?
<steveeJ>
s/to t/do t/
arjen-jonathan has quit [Ping timeout: 252 seconds]
ixxie has quit [Quit: Lost terminal]
georges-duperon has quit [Ping timeout: 240 seconds]
<{^_^}>
[nixpkgs] @matthewbauer pushed 2 commits to master: https://git.io/fAdHF
<clever>
steveeJ: i can check it in the morning, heading off to bed now
<steveeJ>
alright. I'm never aware of people's timezone here :D
<clever>
steveeJ: do /ctcp clever time
<pointfourone>
How do I list the packages which are present in my /nix/store but not installed? I can use `$ nix-env -q` to list the explicitly installed packages but what about packages which were installed as dependencies?
<infinisil>
pointfourone: Why would you care about that?
<steveeJ>
clever: tty tomorrow then ;)
<pointfourone>
infinisil: I ... can't think of a reason right now but I can get the list of all the installed packages in Arch.
<infinisil>
pointfourone: These other dependencies aren't installed
<infinisil>
in NixOS
<d1rewolf>
all, I have "(python2.withPackages(ps: with ps; [ dbus ]))" in my configuration.nix, but "import dbus" fails within the python repl. Any ideas what i'm missing?
<steveeJ>
can someone give me a clue how I can reproduce `nixos-rebuild build-vm-with-bootloader` in a nixos test? I want to to use exactly the same command line arguments and image
<infinisil>
steveeJ: Look into the source of nixos-rebuild
<pointfourone>
infinisil: So, say I install libreoffice or firefox and that depends on gcc. So you're saying that gcc isn't installed?
JonReed has joined #nixos
<steveeJ>
infinisil: I know it builds vmWithBootLoader. that emits a script and an image in PWD
<steveeJ>
it's designed to be run by the user manually
<infinisil>
pointfourone: Indeed
<infinisil>
I mean it's there, but not installed, you can't call gcc from the command line
<infinisil>
gcc is then only used by that package as a dependency, and only that, there's no (easy) way to access that packages contents
<d1rewolf>
all, I have "(python2.withPackages(ps: with ps; [ dbus ]))" in my configuration.nix, but "import dbus" fails within the python repl. Any ideas what i'm missing?
<{^_^}>
[nixpkgs] @xeji pushed commit from @dtzWill to master « gnomecast: 1.4.0 -> 1.4.1 (#47191) »: https://git.io/fAdQ2
ixxie has joined #nixos
<ixxie>
in a multiline string, I only need to escape ${ or also other $ signs?
<JonReed>
ixxie: Only ${
<JonReed>
Wouldn't it make more sense to organize fonts under `fonts.`, instead of having them all at root of `nixpkgs`? They would not pollute nixpkgs and it would make it easier to browse them in repl. As more fonts are added it would make it easier to keep sanity.
<ixxie>
thanks JonReed
sigmundv__ has joined #nixos
<{^_^}>
[nixpkgs] @tg-x opened pull request #47197 → ltc-tools: init at 0.6.4 → https://git.io/fAd78
<manveru>
JonReed: that'd be nice, yeah...
<manveru>
at least having a reference to them there, and adding new ones only there :)
sie has quit [Quit: Connection closed for inactivity]
ixxie has quit [Ping timeout: 260 seconds]
jluttine has quit [Ping timeout: 246 seconds]
camsbury has quit [Read error: Connection reset by peer]
hitchhikingcoder has joined #nixos
haslersn has quit [Ping timeout: 256 seconds]
camsbury has joined #nixos
jluttine has joined #nixos
<{^_^}>
[nixpkgs] @andrew-d opened pull request #47198 → xloadimage: init at 4.1 → https://git.io/fAd7A
Boobuigi has joined #nixos
<infinisil>
,tias = Try it and see! (then tell us what you saw)
<{^_^}>
tias defined
<Boobuigi>
Let the record state that Boobuigi first joined #nixos the moment tias was defined.
<{^_^}>
[nixpkgs] @xeji pushed commit from @charles-dyfis-net to master « Introduce desync, an alternate implementation of casync (#47195) »: https://git.io/fAd5O
<jtojnar>
,locate Atspi
<{^_^}>
Found in packages: onboard, at_spi2_core.dev
Anton-Latukha has quit [Quit: Leaving.]
orivej has quit [Ping timeout: 252 seconds]
hitchhikingcoder has quit [Remote host closed the connection]
<jtojnar>
why do python applications propagate meson?
pointfourone has left #nixos [#nixos]
jasongrossman has quit [Ping timeout: 245 seconds]
<infinisil>
jtojnar: Probably because meson needs python at runtime, but there's better ways of doing that than propagatedblabla
<jtojnar>
infinisil: well, meson does not propagate it, buildPythonApplication does
<infinisil>
Oh
<jtojnar>
it makes sense for buildPythonPackage
<jtojnar>
but I do not see the point for apps
<infinisil>
Yeah, seems like something worth fixing
<jtojnar>
maybe there is some reason
<jtojnar>
otherwise, I would expect it to be fixed
thc202 has quit [Ping timeout: 244 seconds]
haslersn has joined #nixos
<infinisil>
jtojnar: Considering that buildPythonApplication is just a small wrapper around buildPythonPackage, this might just be an oversight
camsbury has quit [Read error: Connection reset by peer]
<haslersn>
Hi. I installed NixOS on my new laptop. For doing so, I deleted all partitions and created a fresh EFI partition and an encrypted luks partition which contains two logical volumes (root and swap). After the installation, I tried to boot and the laptop doesn't find any boot device. What could be the reason?
camsbury has joined #nixos
echel0n_ has quit [Quit: WeeChat 2.0]
<haslersn>
I did basically the same installation on another laptop where it worked
<haslersn>
Am I right in that it's not a problem with the disk encryption since if it was, it should still be able to boot into grub?
<infinisil>
haslersn: Did you run nixos-generate-config?
<haslersn>
infinisil: yes and I checked that the uuids in the hardware-configuration.nix are right
Brejic has joined #nixos
<infinisil>
haslersn: Does it work when you select a boot device manually? via F12 or something while booting
<haslersn>
infinisil: No. There is no boot device in that list. If I plug my live usb in, it's the only list entry
<clever>
haslersn: does efibootmgr -v list it? is the partition type code correct?