<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JTkPB
<iwq>
software that I'm trying to package is running a bash script after `make install`, but it seems to be failing at /usr/bin/env with `bad interpreter: No such file or directory`, why is that? is /usr/bin/env not available inside the build environment?
kirc has joined #nixos
kirc has quit [Client Quit]
<hodapp>
gah, running into the dreaded issue in which JACK "works" with PulseAudio but there is no sound from apps using it
<infinisil>
iwq: Indeed it isn't
<infinisil>
Run `patchShebangs` on the script before you run it
<samueldr>
zecnate: if it runs during the build, native[to the builder]BuildInputs; otherwise buildInputs
<zecnate>
Also, I don't understand how "bare" derivation attributes map to the named dependency attributes, like `depsHostHost` or `nativeBuildInputs`. If my derivation says `perl = nixpkgs.perl;` as an attribute, what kind of dependency does that become?
<samueldr>
that's what you need to get unless you deal into building a compiler
<zecnate>
samueldr: Thanks. Maybe that section needs to say that as a simplistic heuristic right at the top.
<samueldr>
probably :)
<samueldr>
I think my interpretation might be _too_ simplified though
<zecnate>
In my case: my runtime is a python file (not a python "package" or "module", and it has no dependencies). So what kind of dependency is the python interpreter?
<samueldr>
but for the majority of what you're probably going to end up writing, it should hold
<zecnate>
I would guess it's: propagated (since every downstream needs the interpreter), and it has to run on the target host…
<samueldr>
python is a buildInput since it's used on the host that executes code from the derivation
<samueldr>
that executes the output of the derivation**
<zecnate>
The derivation of the python script is unaware/insensitive to build/target hosts.
<zecnate>
Doesn't it need to be a propagated build input?
<samueldr>
propagated makes it so that it's in the build inputs of whatever includes that derivation as an input
patagonicus1381 has joined #nixos
<samueldr>
so if you put python into propagatedBuildInputs of derivation A; then make a derivation B using builtInputs = [ A ]; python will be put in there like you wrote buildInputs = [ A python ];
<samueldr>
(not really, but simplified)
<zecnate>
And that's exactly what I need right? B needs that python at runtime.
<samueldr>
but python, being an input of A (not propagated) should be enough to make A know about python for runtime uses*
<samueldr>
* (assuming the build properly handles it)
<samueldr>
uh, not sure, I don't do python stuff, and the specific builders for the different ways python can be packaged probably would handle that for you really
<zecnate>
Is this because after building A, nix scans hashbangs and sees a reference to the python interpreter derivation output, so automagically includes it?
<samueldr>
that's one of the way it can be done, yes
<zecnate>
I can't seem to use `filterSource` to exclude `.git`. Currently I get this error:
<zecnate>
The path name '.git' is invalid: it is illegal to start the name with a period. Path names are alphanumeric and can include the symbols +-._?= and must not begin with a period. Note: If '.git' is a source file and you cannot rename it on disk, builtins.path { name = ... } can be used to give it an alternative name.
<zecnate>
But I want to _exclude_ that name, not include it. :-/
<clever>
zecnate: what expr did you use?
<zecnate>
It's ... way over complicated. :-)
<clever>
zecnate: and did you try lib.cleanSource ?
<clever>
zecnate: lines 7-12 will create nix paths, which will then try to copy the entire .git into /nix/store
<clever>
zecnate: and then your comparing the path to the original, and the path to the copy, which will never match
<zecnate>
Thanks!
Chiliparrot has quit [Ping timeout: 244 seconds]
sangoma has quit [Ping timeout: 240 seconds]
<clever>
zecnate: throwing a `toString` on line 7 might fix it, but there should be cleaner ways, check the src of cleanSource
<zecnate>
So, unlike `lib.cleanSource` I want to filter out explicit paths, so not any file/directory anywhere named "nix" but literally the single nix directory in the repo root.
<zecnate>
I basically want to look at the "relative path" from `rawSrcDir` and compare against that.
<zecnate>
Looking.
<clever>
zecnate: this one is using a strict whitelist instead of blacklist, but it should still serve as a good example
aasgweb has joined #nixos
<zecnate>
Yeah, thanks. The `sansPrefix` looks like exactly what I specified.
<zecnate>
+1 to strict whitelists. ;-) I'm not there yet, though.
sangoma has joined #nixos
<clever>
i tend to write my whitelist by following the errors
<clever>
start with nothing, what fails? no Makefile/configure, add them, repeat
hmpffff has joined #nixos
sangoma has quit [Read error: Connection reset by peer]
sangoma has joined #nixos
Chiliparrot has joined #nixos
ky0ko has joined #nixos
jani1234 has joined #nixos
sangoma has quit [Read error: Connection reset by peer]
jani1234 has quit [Ping timeout: 265 seconds]
sangoma has joined #nixos
cYmen has joined #nixos
<zecnate>
Good idea. I'll start with that approach.
<zecnate>
Can I explicitly name the result of `filterSource`? I have multiple `filterSource` expressions selecting different files from the same repo root. The default name is based on the repo root.
tabor has quit [Remote host closed the connection]
<alexarice[m]1>
Anyone have an idea about how I can take all changes but one from an overlay?
<alexarice[m]1>
I want to get same packages from wayland overlay but want to test using a stable sway
Chiliparrot has joined #nixos
lunik1 has joined #nixos
<NobbZ[m]>
Depending on how exactly the attributes in that overlay work, you might be able to override and pass stable away in to them
<alexarice[m]1>
At the moment I'm trying to make another overlay that saves `sway` to `sway-stable` and the put another overlay afterwards which replaces the modified `sway` by the `sway-stable` I saved before
<lightandlight>
I want to know the proportion of the (yet unreleased) nixos-20.09 branch on nixpkgs has been copied to cache.nixos.org. Is there a fast way to do that?
<clever>
lightandlight: everything hydra builds is pushed to cache.nixos.org once the build finishes
<clever>
so you can just pull up the hydra job for that branch, and look at how many jobs are pending vs complete
maxdevjs has quit [Ping timeout: 272 seconds]
cantaloupe has quit [Ping timeout: 240 seconds]
<{^_^}>
[nix] @regnat opened pull request #4139 → Always keep the CA derivations around → https://git.io/JTInw
<{^_^}>
[nixpkgs] @timokau pushed commit from @samuela to master « spotify-unwrapped: 1.1.26.501.gbe11e53b-15 -> 1.1.42.622.gbd112320-37 (#99924) »: https://git.io/JTIlH
taylan has joined #nixos
sangoma has joined #nixos
supercoven has joined #nixos
sascha__ has joined #nixos
<taylan>
hi all. at my workplace we're looking forward to setting up a kubernetes cluster. I love the principle behind Nix, but don't know how it fits into the picture. (only ever used Guix as my regular PC OS.) is there an advantage to using NixOS instead of e.g. Debian as the OS below kubernetes? or can Nix serve as a (better) alternative to
<taylan>
kubernetes?..
<dminuoso>
taylan: they are completely different mindsets.
<dminuoso>
And address different problems
jakobrs has joined #nixos
<dminuoso>
Kubernetes, at its core, is about automated deployment and automated scaling.
<dminuoso>
Most shops that run k8s dont even use k8s because they need it, but because they think its fancy.
<dminuoso>
It's made for "How do I deploy 5,000 services across tens of thousands of servers"
<dminuoso>
The kind of problems Google runs into
sgrunert has quit [Ping timeout: 256 seconds]
<dminuoso>
The tech you can sensibly compare here, is nixos vs docker
<dminuoso>
Because they solve similar problems
wazdra has quit [Ping timeout: 245 seconds]
<dminuoso>
docker, at its core, is a stack of tech to be able to talk about how to build up a server to provide some service with essentially a shell script, and obtain artifacts you can push around
<dminuoso>
as first class values
<dminuoso>
(sort of a more efficient version of vagrant)
sangoma has quit [Read error: Connection reset by peer]
<dminuoso>
nixos is about declaratively describing your server in a fully deterministic way
supercoven has quit [Ping timeout: 272 seconds]
<dminuoso>
You can however, build up something similar to k8s with nixos
<dminuoso>
Notably you can use nix + terraform, if you combine that with a sensible cloud provider, you can build something similar but with high deterministic guarantees
<taylan>
the more experienced colleague who was planning this approach said, if I understood correctly, that we would be using docker to deploy the individual "nodes" (or w/e they're called) in the k8s cluster. I wonder if Nix could instead take that task?
<dminuoso>
docker is not about deploying, docker is about building oci images
<dminuoso>
and all the infrastructure to run containers off of that
sigmundv has joined #nixos
<dminuoso>
taylan: and no, nix is not suitable for that
<dminuoso>
taylan: If we compare things side to side, it's roughly:
<dminuoso>
nix <> Dockerfile
<dminuoso>
nixos <> docker-compose
<dminuoso>
nixops <> docker swarm
<dminuoso>
(By that I dont mean they are equivalent, but the left and right side take somwhat comparable roles)
supercoven has joined #nixos
<taylan>
I understand, I think. e.g. while nixos and docker-compose do similar things, nixos does it on a whole-OS basis whereas docker-compose does it on a container basis, is that correct?
<dminuoso>
terraform+nix <> k8s
bahamas has quit [Ping timeout: 256 seconds]
<dminuoso>
taylan: the comparisons are difficult and not so useful I think
<dminuoso>
nix is a pure functional language to describe packages in a deterministic way
<dminuoso>
Dockerfile is just a shell script to describe the filesystem of a server
<dminuoso>
(The reason I dont make a direct comparison between Dockerfile and Nixos, is because it's implicitly understood that a container should always talk about a single service/process only)
<dminuoso>
Nixos otoh extends nix to not just talk about packages, but the entirety of the linux. Filesystem, systemd services, kernel, grub, etc..
jakobrs has quit [Quit: WeeChat 2.9]
<dminuoso>
A more useful analogy to nixos is probably a pure functional and whole meal version of Chef/Puppet
<dminuoso>
(Or just Puppet, I guess)
<taylan>
guix has this "guix container" command to start a container with a certain environment, which seems very analogous to Docker, I think? (never really used Docker.) does Nix not have a "nix container" command of sorts?
<taylan>
(I thought Nix would be the way to go over Guix since it's more mature, but maybe there's a thing or two in Guix that's more suitable for our plan, I dunno)
<dminuoso>
well, nixos can run docker containers if you want that
<dminuoso>
But at that point, Im not sure how much value you're gonna get off nixos
<dminuoso>
Ah I think I misread
<dminuoso>
taylan: You can build docker images using nix
<dminuoso>
That gives you the fully deterministic benefits of nix, while obtaining a deployable/runnable artifact from it
supercoven has quit [Read error: Connection reset by peer]
<dminuoso>
So you describe a docker image with nix, rather than Dockerfile, and nixpkgs.dockerTools will take care building a guaranteed deterministic image off of that
<dminuoso>
So this can be useful for your CI
mekster has left #nixos [#nixos]
<dminuoso>
(And if you deploy/manage containers with k8s, this fits very well in)
<taylan>
that sounds like a good idea
<dminuoso>
another big benefit, is that you get a clean development environment to play with the contents on your dev machine without all the container nonsense
<dminuoso>
and still have the guarantee that the container will be *the* *same*
<dminuoso>
(because since nix is deterministic, you can trust that it generates the same exact environment into the container)
<dminuoso>
Well. Not' 100%, but close to it
supercoven has joined #nixos
zupo_ has joined #nixos
sangoma has joined #nixos
vidbina has quit [Ping timeout: 256 seconds]
supercoven has quit [Max SendQ exceeded]
supercoven has joined #nixos
<{^_^}>
[nixpkgs] @toastal opened pull request #100327 → JuliaMono font → https://git.io/JTI0Z
<miyu-saki>
Yeah, makes sense. I guess I'll just make a direct translation of their service file instead. There's some bugs in their application, but their service files worked because it masked the bugs, that's why I wanted to take from upstream instead.
<miyu-saki>
(I copied their service files, but made it more Nixy, so avoided environment variables and stuff)
<miyu-saki>
(In favor of evaluation-time replacement)
vidbina has joined #nixos
griff__ has joined #nixos
gueorgui_ is now known as gueorgui
griff_ has quit [Ping timeout: 244 seconds]
griff__ is now known as griff_
alp has quit [Remote host closed the connection]
alp has joined #nixos
vidbina has quit [Ping timeout: 256 seconds]
<{^_^}>
[nixpkgs] @ttuegel closed pull request #70691 → Abort Qt builds that do not use a supported deriver → https://git.io/JeWUD
<{^_^}>
[nixos-weekly] @domenkozar pushed 10 commits to production: https://git.io/JTIHw
<Raito_Bezarius>
exarkun: can you tcpdump on your side?
<{^_^}>
[nixos-weekly] @domenkozar pushed to master « Call for Content: 2020/09 »: https://git.io/JTIHr
<{^_^}>
[nixos-weekly] @domenkozar opened pull request #135 → Call for Content: 2020/09 → https://git.io/JTIHo
<Raito_Bezarius>
can you make it work on another OS ?
<exarkun>
What should I dump with it? traffic on the wireguard interface? traffic on the uplink interface?
jmeredith has joined #nixos
<Raito_Bezarius>
the uplink
<exarkun>
beats me, I've never tried setting up wireguard on any other OS
<Raito_Bezarius>
what about trying to use wg-quick rather than manual wg?
flythief__ has joined #nixos
<exarkun>
I'm not sure. I tried declarative wg-quick in configuration.nix and I couldn't get that to work either.
<exarkun>
I didn't try imperative wg-quick.
alp has quit [Ping timeout: 240 seconds]
<Raito_Bezarius>
imperative wg-quick could be used for debug, but I have some large WireGuard setup in NixOS (20.03 and beyond) and they're working quite well
<Raito_Bezarius>
(non trivial ones with routing stuff, etc)
<Raito_Bezarius>
So I think it's either a problem on the endpoint side or on your local side
<Raito_Bezarius>
(but it could be also a mix of bugs, etc.)
<exarkun>
tcpdump says when I run ping a bunch of packets go out from me to the endpoint but nothing ever comes back
<exarkun>
oops not nothing
<exarkun>
just far fewer than go out
<exarkun>
the return traffic appears to be wireguard keepalives
<flythief__>
hi, I just follow nix-pills course (pill 10) and I run installPhase in nix-shell and get error: mkdir cannot create directory /nix/store/hash**-hello. anyone know this? It works when I just run nix-build. thanks firstly :).
<Raito_Bezarius>
exarkun: and on the wireguard layer?
<Raito_Bezarius>
on the wg layer, you should see those ICMP echo request I suppose
griff__ has joined #nixos
<exarkun>
Raito_Bezarius: yes, just those
<Raito_Bezarius>
did this commercial VPN already worked before with WireGuard?
<Raito_Bezarius>
if you try wireguard on mobile or something like this
<Raito_Bezarius>
can you ping an internal known IP rather than Internet?
<exarkun>
no, this is a new service, I haven't used it before
griff_ has quit [Ping timeout: 260 seconds]
griff__ is now known as griff_
<flythief__>
well, in nix-shell, I shouldn't install it? I find that in pill 11. nix-shell for debuging, nix-build for building it in real environment. :)
ddellacosta has joined #nixos
Rusty1 has joined #nixos
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
vandenoever has quit [Read error: Connection reset by peer]
vandenoever has joined #nixos
<lordcirth>
In nix, how can I remove '.'s from a string?
meh` has joined #nixos
sangoma has quit [Read error: Connection reset by peer]
mthst has quit [Ping timeout: 240 seconds]
<lordcirth>
I tried "${filter (!= '.') version}" but got "syntax error, unexpected NEQ"
<lordcirth>
'${builtins.replaceStrings "." "" version}' gives "value is a string while a list was expected"
mupf has joined #nixos
<lordcirth>
Oh I see
<lordcirth>
Seems a bit overkill of a function
xelxebar has joined #nixos
bahamas has quit [Ping timeout: 272 seconds]
mupf_ has joined #nixos
mupf has quit [Read error: Connection reset by peer]
<flythief__>
hi, I install bsdgames by `nix-env -iA nixpkgs.bsdgames`. but I get "error: attribute 'nixpkgs' in selection path 'nixpkgs.bsdgames' not found". some know this? thanks.
<flythief__>
I find that I can't install all nixpkgs.* directly. but I write it in configure.nix (by `nix-rebuild switch`) will works.
utlaert has joined #nixos
zanc has joined #nixos
<zanc>
how do I enable systemd services but not started on startup, i.e. I would rather start it manually `systemctl start service`?
<utlaert>
Hi. I'm running NixOS on a vmware guest and even though I have virtualisation.vmware.guest.enable
<utlaert>
= true automatic screen scaling doesn't work. Does anyone here have experience with NixOS vmware guests? Thanks in advance.
<utlaert>
zanc, can't you just define the systemd.services set?
<zanc>
it will prevent the service from running on startup, yes but I won't be able to start it manually using `systemctl` because the `.service` file is now symlinked to `/dev/null`
mupf has joined #nixos
<utlaert>
zanc, well if you set the enable flag to true the service will of course start with the system. That's what "enabled" means in the context of systemd.
<utlaert>
Ah I see.
<zanc>
utlaert: wait, I mean `{ enable = false; }`
<utlaert>
rnhmjoj, not sure what you mean. I'm just talking about screen resolution of the vmware guest running nixos. Apparently I have no derivation with 'spice' in its name and there doesn't seem to be anything like that in the closure of the open-vm-tools derivation.
<zanc>
parsnip: you need flakes-enabled Nix to do this `nix show-derivation nixpkgs#emacs` https://nixos.wiki/wiki/Flakes
<rnhmjoj>
utlaert: i don't know how vmware works but there are other programs that use a protocol called SPICE (a kind of remote desktop) to connect to a VM. that supports automatic resolution changes but requires running a daemon on the guest, called vdagent
domogled has joined #nixos
bahamas has joined #nixos
<utlaert>
rnhmjoj, vmware doesn't seem to be using that. When I compare it with my ubuntu machine (where screen resolution change is working), it has vmtoolsd running as vmusr user, which is what all the forum threads say is needed. But that is also the case on my nix machine.
simonpe^^ has joined #nixos
<simonpe^^>
I'm trying to port our build system to WSL2 but when the binaries are downloaded from our cache I'm getting errors about sha256 sum mismatches. Does somebody now why?
<parsnip>
and that clarifies that it's a configure make install sort of thing
<utlaert>
I remember that for whatever reason the package libgtkmm must be installed (that was the root cause when resolution change didn't work on ubuntu). But it is in the closure of the open-vm-tools derivation, so I don't think that's the problem.
<theuni>
I'm currently trying to create disk images using "runInLinuxVM" and I'm trying to create images that include LVM volumes. Obviously the runInLinuxVM environment is very "raw" and so I'm stumbling over the fact that there is no udev/device mapper around. I haven't found any use of runInLinuxVM and am currently trying to research the way that the test runner does it's thing, but I haven't found anything useful yet. Does
<theuni>
anyone know how to simply either use a full config to run a command in runInLinuxVM or maybe knows a light-weight way to get dm/udev started without the rest of the environment?
<zanc>
why does `nix-copy-closure --from user@hostname /nix/store/abcde...` while `nix-env -i --substituters ssh://user@hostname` does not? it gave me an error `cannot connect to '192.168.0.106'`
<flythief__>
hi, I want to use gd lib. But I find that gd in nix-shell only refers gd-bin. how can i get directory of gd-dev which will contain include and lib ?
<clever>
zanc: nix-copy-closure ssh's from the current user, but `--substituters` will try to ssh from root, via the nix-daemon, which lacks your $SSH_AUTH_SOCK
berberman has quit [Max SendQ exceeded]
berberman has joined #nixos
domogled has quit [Ping timeout: 240 seconds]
Fare has quit [Quit: Leaving]
<zanc>
clever: so how do fix it?
berberman has quit [Max SendQ exceeded]
<clever>
zanc: put a private key in /root/.ssh/id_rsa, that lacks a password, and can get into user@hostname
civodul has quit [Quit: ERC (IRC client for Emacs 27.1)]
<rnhmjoj>
zanc: are you on nixos?
<zanc>
rnhmjoj: yes
marcy_ has joined #nixos
philr has quit [Ping timeout: 244 seconds]
berberman has joined #nixos
<zanc>
clever: do I put it `/root/.ssh/id_rsa` in my root or remote?
<clever>
zanc: on the machine where you ran `nix-env -i --substituters ssh://user@hostname`
berberman has joined #nixos
<clever>
zanc: the nix-daemon on that machine, is what did `ssh user@hostname` as root
<rnhmjoj>
zanc: i'm not sure it's the same thing but to let the nix-daemon connect to my remote builder i configured nix.buildMachines.*.sshKey and added an entry in program.ssh.knownHosts
<clever>
rnhmjoj: thats more for building things on a remote machine, rather then just checking if it already has a built copy
<noonien>
clever: is it possible to create a kexec tarball for a running nixos system?
asbachb has quit [Ping timeout: 245 seconds]
berberman has joined #nixos
<clever>
noonien: possibly
<noonien>
i just created the kexec tarball on the target system, this way the /nix/store paths already existed and didn't fail because tar couldn't extract the files
<clever>
noonien: oh, if your on nix, you dont want to create the tarball
<xensky>
not sure if this is a nix problem but it started after i moved over; clicking links in discord doesn't open in my browser, has anyone else experienced this?
<euank>
xensky: to give you some paths to debug with, xdg is generally how applications open those links. `xdg-open "https://example.com"` in a terminal can confirm if it's discord specific, or if your xdg mime setup isn't configured correctly
<xensky>
another counter example is that links work fine in irc (quassel)
<euank>
Ah, yeah, weird. Maybe `XDG_UTILS_DEBUG_LEVEL=2 Discord` from a terminal will print more info on what it's trying, but I'm not sure if it will.
mupf has quit [Client Quit]
<euank>
I assume it's running via a wrapper script, and I'd hope the wrapper isn't clobbering XDG_DATA_DIRS
hnOsmium0001 has joined #nixos
mupf has joined #nixos
justanotheruser has quit [Ping timeout: 265 seconds]
<simpson>
Are you using dockerTools? It is better than `FROM nixos/nix` Dockerfiles at only including what is needed.
<emmanuel_erc>
I am using dockerTools
<simpson>
And the dependencies are correctly ignored in other contexts? I don't see anything in pkgs/build-support/docker which would do this, but I'm not very familiar with this code.
<emmanuel_erc>
No they aren't
cjpbirkbeck has joined #nixos
<emmanuel_erc>
I've had to manually ignore them elsewhere
<simpson>
Then probably the best course is to figure out why these dependencies keep getting pulled in even though they're supposedly build-time-only.
is_null has joined #nixos
<numkem>
Has anyone played with the latest version of node2nix? I'm using the nodeDependencies that was just introduces that expose what npm install would create. Than I symlink those dependencies into another derivation and than build it from there. The problem is while I no longer need the nodeDependencies, it's still getting pulled when getting copied
<numkem>
over to another machine. Is there way to remove all deps? I've tried setting propagatedBuildInputs = [] but it doesn't change anything
<{^_^}>
[nixpkgs] @AndersonTorres opened pull request #100357 → zegrapher: init at 3.1.1 → https://git.io/JTLzX
<pittma>
I've since installed gpgme, but it seems thunderbird can't find it.
<emmanuel_erc>
simpson: So I realized that there is one particular derivation in question that is common to all my builds. One of the build-time deps that keeps finding it way into the acutal build is the GHC compiler (1.3 GB)
<pittma>
Is there more to do to include libgpgme in LD_PATH?
<simpson>
emmanuel_erc: Fun! At least you've narrowed it down somewhat.
<euank>
I know that's how the nixpkgs go builder keeps the go compiler from being pulled in as a dep of go binaries. Not sure how the haskell builder does it
vidbina has quit [Ping timeout: 260 seconds]
<euank>
or if that problem even is supposed to exist for haskell binaries
<__monty__>
Until recently it was the case for pandoc, for example. So I think it's common or at least semi-common.
<clever>
__monty__: does pandoc still function for you on master? i recently found problems where it blows the haskell heap, no matter how big you make it
<clever>
it started when the default ghc updated a few months back
<emmanuel_erc>
euank: How does this script actually work? I'm confused. Is there supposed to a list it references that keeps away non-runtime deps?
<euank>
there's a description of why/how pandoc uses it emmanuel_erc
knupfer has joined #nixos
knupfer has quit [Client Quit]
<euank>
you give it a bunch of store paths you don't want, and it seds out those store path's hashes for 'eeeeee...' to remove the automatic runtime dep detectoin of them
knupfer has joined #nixos
lsix has quit [Client Quit]
<euank>
And it's also probably a sign that the toolchain in question is doing something kinda weird.
<{^_^}>
[nixos-homepage] @garbas pushed to redesign-governance « switch from flex to grid layout for teams section »: https://git.io/JTLVP
<euank>
In the go case, it's that debug symbols referencing the go stdlib, including filesystem path, are stored in all binaries
knupfer has quit [Client Quit]
knupfer has joined #nixos
<euank>
For haskell, I don't know why a reference to ghc would sneak into the final binary. It may be worth understanding that bit first before blindly replacing it with 'eee...' since there might be a better way to have the build output just not include that entirely
knupfer has quit [Client Quit]
knupfer1 has joined #nixos
knupfer1 has quit [Client Quit]
knupfer has joined #nixos
<emmanuel_erc>
Yeah, that might be hard for me to figure if only because the entire infra for building my package in question is obelisk. Obelisk is a framework for web development in Haskell. So the entire app actually consists of three haskell projects (Haskell-wise)
<emmanuel_erc>
How they string things together is slightly opaque.
<__monty__>
clever: Haven't used it recently.
knupfer has quit [Client Quit]
knupfer has joined #nixos
cr4y1 has quit [Remote host closed the connection]
<emmanuel_erc>
euank: So given what you showed about removing references, that seems to work really well, when the package in question is a dependency. However, what if it is the package/library that you are trying to build that has "Paths_<nameoflibrary>" as specified module in its cabal file?
Mateon2 has joined #nixos
<euank>
emmanuel_erc: I'm well out of my depth in haskell knowledge here. I don't know what 'Paths_' modules are meant to be/do