justanotheruser has quit [Ping timeout: 260 seconds]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
emmanuel_erc has quit [Read error: Connection reset by peer]
emmanuel_erc has joined #nixos
vikanezrimaya has joined #nixos
justanotheruser has joined #nixos
sjlnk has quit [Remote host closed the connection]
sjlnk has joined #nixos
lsix has quit [Quit: WeeChat 3.0.1]
mcornick has joined #nixos
<matthewcroughan_>
viric: I actually succeeded. I just added -D BUILD_FLAT_MESH:BOOL=ON to the cmake args and got it working just fine.
<matthewcroughan_>
I'm wondering whether this should be on by default, because my friend who's using the appimage seems to expect it
<matthewcroughan_>
they actually release it with that in their appimage
<matthewcroughan_>
supersandro2000: you are correct, I read the commit and forgot it would not be squashed
<matthewcroughan_>
all I did was `git log --oneline | grep cryptpad` and expected to see **one commit** that added both the service *and* the top level line. This sent me on a diversion.
<matthewcroughan_>
because PRs are not native to git, I keep forgetting this because I"m a noob.
<cwprobablydead[m>
Where can I find the current system configuration.nix?
<cwprobablydead[m>
I can't find any .nix files in `/nix/var/nix/profiles/system`
<vancoding[m]>
In /etc/nixos/
<cwprobablydead[m>
That isn't updated when for example switching profiles
<cwprobablydead[m>
or deploying via nixops
emmanuel_erc has quit [Read error: Connection reset by peer]
<cwprobablydead[m>
So when the system boots to old configs it doesn't save a copy of hte nix expression that built that config?
<vancoding[m]>
Most people manage their configuration with git for that reason
<cwprobablydead[m>
* So when the system boots to old configs it doesn't save a copy of hte nix expression that built said config?
vikanezrimaya has quit [Quit: Connection closed]
vikanezrimaya has joined #nixos
<cwprobablydead[m>
I know how/when/why to use git. I'm just surprised that in a distro designed to build literally everything from nix expressions that the base nix expression isn't saved somewhere
<cwprobablydead[m>
* I know how/when/why to use git. I'm just surprised that in a distro designed to build literally everything from nix expressions that the base nix expression from which everything else is built isn't saved somewhere
emmanuel_erc has quit [Read error: Connection reset by peer]
emmanuel_erc has joined #nixos
<cwprobablydead[m>
<vancoding[m] "Ah, you mean when you have syste"> I guess thats what this is for
<cole-h>
It only symlinks the value of the `NIXOS_CONFIG` env var (or `<nixos-config>` from NIX_PATH)
<cole-h>
If you have `imports`, it won't copy those
mcint has quit [Ping timeout: 240 seconds]
<matthewcroughan_>
cole-h: I need to fix up a package to link its name to lowercase. What phase should I use?
<cwprobablydead[m>
so there's no copies of the original expression saved?
<cole-h>
The copy of the config is where you're deploying from.
<matthewcroughan_>
I've tried installPhase and fixupPhase, but they both incur a complete recompilation of the package, the package in question is FreeCAD.
<matthewcroughan_>
freecad is installed to /usr/bin/FreeCAD, I want to fix that like so: fixupPhase = '' ln -s $out/bin/FreeCAD $out/bin/freecad '';
<matthewcroughan_>
30 minutes of recompilation later..
<djwf>
Greetings: I am using nix 2.3.10 on Devuan (debian derivative) and am trying to install profanity from the nixpkgs-unstable channel, but am getting an error that I don't understand how to troubleshoot: "installing 'profanity-0.10.0'
emmanuel_erc has quit [Read error: Connection reset by peer]
emmanuel_erc has joined #nixos
leex has joined #nixos
<leex>
On my fresh nixos install pulseaudio does not start on boot anymore, I checked /etc/pulse/client.conf and it contains autospawn=no which I think is the reason that it doesn't automatically start anymore, what's the easiest way to get it to start on boot?
gustavderdrache has joined #nixos
gustavderdrache has quit [Client Quit]
crenfrow has quit [Ping timeout: 256 seconds]
emmanuel_erc has quit [Read error: Connection reset by peer]
<slack1256>
I thought that was normal, but my pc is from 2010 :-/
<{^_^}>
#6784 (by lethalman, 6 years ago, closed): Problem attempting to mkMultiInstance
emmanuel_erc has joined #nixos
<matthewcroughan_>
>Closing the issue after almost 2 years of no discussion. FWIW, I've implemented duplicate services by using nixos containers. This has worked fairly nicely for me.
<matthewcroughan_>
Have you got any of your examples with nixos-containers? :D
<Guest76>
Hey guys, can anyone help please? I am trying to install vscodium + extensions with home manager, but getting "got value instead of list", I guess home manaager expects list, copied this from the wiki don't know how to translate it to home manager https://0bin.net/paste/+zOfS-4E#CQaHfW57ElnS9JMu4VgbiK16K2GCvL9vUomsWyDVamn
<thoughtpolice>
q3k[m]: hi, sorry, no i haven't tried packaging bdw yet. honestly a lot of the tcl stuff wrt bluespec is a bit of a mess still. happy to give ideas though
domogled has quit [Remote host closed the connection]
<thoughtpolice>
i think one issue i ran into is that the tcl stuff in nix isn't really integrated fully, and incrtcl is basically the only library that wants infrastructure like a setup hook to populate search paths, etc. there's not much documentation on this stuff and it's unfortunately essential to having everything working properly, and i think multiple distros vary a bit in how they lay things out of course.
marusich has joined #nixos
monotux has quit [Remote host closed the connection]
oxalica has quit [Remote host closed the connection]
oxalica has joined #nixos
devhell has joined #nixos
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage2nix: update list of broken builds to fix evaluation errors on Hydra »: https://git.io/JmuD1
ahmedelgabri has joined #nixos
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/Jmuyi
<Guest93377>
I'm new to nix and trying to build a python env. I've got one package that isn't in Nixpkgs -- https://pypi.org/project/wandb/#description. I'm defining it using buildPythonPackage and losing my mind. If I choose format = 'wheel', I get no module found for pip. If I choose format = 'setuptools' I get no module found for setuptools.
<{^_^}>
[nixpkgs] @peti pushed 3 commits to haskell-updates: https://git.io/Jmzkh
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JmzIc
zupo has joined #nixos
<aforemny>
Guest93377: I am not very familiar with python, but removing the python27 package from propagatedBuildInputs resolves the error of setup tools not being found (possibly you also do not need setuptools in propagatedBuildInputs). I would try packaging subprocess32 for python3
<aforemny>
Guest93377: Or is subprocess32 a built-in module with python3? Could it be then that wandb requires to be packaged with python39, since it complains about the version bonud not being satisfied with python38. You might also try patching wandb to relax the version bound on subprocess32
<matthewcroughan_>
does anyone know how I can configure vim without home-manager?
<vikanezrimaya>
matthewcroughan_: see https://nixos.wiki/wiki/Vim, section "Custom setup without using Home Manager"
<matthewcroughan_>
yeah I'm seeing that now actually
<matthewcroughan_>
so messing with `vim_configurable.customize` is the way?
<vikanezrimaya>
pretty sure home-manager does the same thing
<patagonicus>
vikanezrimaya: home-manager writes a config file into the XDG config directory
<patagonicus>
matthewcroughan_: If you're ok with one config for all users of the system, you can throw a configuration file into etc via environment.etc."vim/vimrc".text = '' your vim settings '';
<matthewcroughan_>
vikanezrimaya: the syntax in that thing looks so busted
<patagonicus>
Or …vimrc".source = ./vimrc; if you rather get it from a file next to your configuration.
<matthewcroughan_>
vim_configurable.customize {
<matthewcroughan_>
instead of vim_configurable.customize = {
<vikanezrimaya>
It's because it's a function that produces a vim package
<matthewcroughan_>
I can't quite get it to work :V
<matthewcroughan_>
that did it.. I've used nix outside of nixos context, it's weird how it discriminates between nixos. vs nixpkgs., despite only having nixos in nix-channel --list
<matthewcroughan_>
actually, that doesn't even work outside of nixos
<matthewcroughan_>
wait.. `nix edit nixpkgs.vim_configurable` just takes me to `pkgs/applications/editors/vim/common.nix`
<matthewcroughan_>
wtf
<matthewcroughan_>
So. If 'vimrcFile' at /nix/var/nix/profiles/per-user/root/channels/nixos/pkgs/misc/vim-plugins/vim-utils.nix:186:15 called with unexpected argument 'customRc' DOESN'T WORK ANYMORE, what do I do to figure out what **is** the expected argument?
<vikanezrimaya>
the `nixpkgs.config` option family is relatively undocumented compared to NixOS modules
<matthewcroughan_>
Doesn't seem to have it
<matthewcroughan_>
Okay, so what's happening right now? It's compiling vim? Why?
<vikanezrimaya>
because that rudimentary support was built right into the package, and it'll recompile vim to get rid of it
<matthewcroughan_>
but it never worked with it set to true
<matthewcroughan_>
vim syntax highlighting didn't work with ftNix = true
<matthewcroughan_>
I have to set it to false in order to allow nix-vim to be loaded at all
<vikanezrimaya>
I definitely remember them conflicting
mkaito has quit [Quit: WeeChat 3.1]
<matthewcroughan_>
Ok, so when I look at nix files, that weird bash-like highlighting is actually ftNix = true doing its evil work?
orivej has joined #nixos
<matthewcroughan_>
it's not just misinterpreting the nix file as a bash script?
<vikanezrimaya>
it might be syntax=conf actually
<matthewcroughan_>
so I'm having to recompile Vim just to get rid of the bad syntax highlighting that Nix has compiled into it?
vikanezrimaya has quit [Quit: vikanezrimaya]
mkaito has joined #nixos
<matthewcroughan_>
Who made the decision to munt the Vim that ships with Nix by default?
<matthewcroughan_>
like, why ruin it like that?
<gchristensen>
"munt"? O.o
<matthewcroughan_>
This is surely a bad decision? To make it so that only a bad syntax highlighter can be used to view nix files, without recompiling Vim?
<matthewcroughan_>
gchristensen: aussie slang :D
<matthewcroughan_>
Oh, god. That's racist.
<matthewcroughan_>
Search "munted"
codygman` has joined #nixos
<matthewcroughan_>
"munted" has a very different meaning LOL
<gchristensen>
my brief search for the word says that is an ethnic slur, yeah, whew. would you like to give your question another try?
<matthewcroughan_>
I guess I can't use it like that. Alas ;D
avaq_ has joined #nixos
<tpw_rules>
yeah i have heard Australians use it like that. but probably not a good word for a global audience
<matthewcroughan_>
Is it or is it not the case that without vim.ftNix = false, I cannot use any other Nix syntax highlighter?
<leonardp>
matthewcroughan_: you are very welcome to open an issue or a pull request to change the package
<matthewcroughan_>
Sure, but I'm not actually sure I'm correct, so I need to verify this is the case first.
justanotheruser has quit [Ping timeout: 244 seconds]
<matthewcroughan_>
I must set vim.ftNix = false; else I cannot use better syntax highlighters for the .nix format, is this correct? Does anyone know if that statement is true?
codygman__ has quit [Ping timeout: 260 seconds]
<matthewcroughan_>
bah, it still doesn't seem to work :D
<matthewcroughan_>
has anyone got a screenshot of how nix highlighting is supposed to look
<supersandro2000>
but what I wanted to tell you is that I don't do any special and something is wrong with the nix part
<{^_^}>
[nixpkgs] @flokli opened pull request #116743 → nixos/bird*: enable config files outside the store, propagate reload errors to systemd → https://github.com/NixOS/nixpkgs/pull/116743
<matthewcroughan_>
I just don't get why what I've done isn't the same as what they have done
<supersandro2000>
no 😂
<matthewcroughan_>
the dpaste I shared, it's as far as I can tell doing what they're doing, but I'm ending up with a very different result, no highlighting at all.
<matthewcroughan_>
and worse yet, the default highlighting is still inside of vim despite me setting `nixpkgs.config.vim.ftNix = false;`
ahmedelgabri has joined #nixos
<matthewcroughan_>
and it bloody recompiled it, so what's going on with nixpkgs.config.vim.ftNix = false;?
<matthewcroughan_>
Who is the madman that decided to screw with Vim in this way anyway? ;D
<matthewcroughan_>
Theo nly context I see people setting this in is ~/.nixpkgs/config.nix
<matthewcroughan_>
supersandro2000: Okay, so there's a symlink called `vim-with-plugins` that seems to work
<matthewcroughan_>
what's that about?
<matthewcroughan_>
/run/current-system/sw/bin/vim vs /run/current-system/sw/bin/vim-with-plugins
<yaymukund>
is there a built-in way to get from "foo" to "/nix/store/abcd123-foo-0.4.5/"?
<matthewcroughan_>
Oh.. That's what this does. I had no idea.
<matthewcroughan_>
# `name` specifies the name of the executable and package
<matthewcroughan_>
There's no way to just override the global one then?
stree has quit [Ping timeout: 276 seconds]
<matthewcroughan_>
omg, just setting it to `vim` :D
<matthewcroughan_>
Okay, so this works great, the placeholders are just poorly explained, don't have chevrons around them like the rest of the docs and I'm basically an idiot.
<JaakkoLuttinen[m>
Can I somehow force `nix-build` to build the given package, not get it from the cache. (All the dependencies still should be fetched from the cache though.)
<Synthetica>
nix-build --check?
<JaakkoLuttinen[m>
Hmm.. Interesting, man page doesn't mention such flag
Shiranai has joined #nixos
<Synthetica>
... yet it exists
<Synthetica>
Hmm, doc bug?
<JaakkoLuttinen[m>
But yes, seems to work, thanks!
<sterni>
JaakkoLuttinen[m: note that will not substitute anything so you'll rebuild stdenv if you don't have it in /nix/store
Qwerky has joined #nixos
<sterni>
--check only works if you have already downloaded / built the package before because it is intended to rebuild the package to ensure it is deterministic
<nschoe>
Hey everyone, I'm cross-compiling some Haskell software for the RPi, and I'm doing it with an override: `pkgs.pkgsCross.aarch64-multiplatform.haskell.packages.ghc865.override`. But when I happen to garbage collect, it takes hours to recompile everything. Maybe ghc865 is too old and not in cache anymore. Can I see a list of the ghc that are in binary cache (for arch64)?
<fuzzypixelz>
https://github.com/divnix/devos refuses to build because "no wheel user nor the root user has set a password ..." Usually I don't get these warnings from my regular nix flake config and I just set all my passwords right after the build. Anyone familiar with devos? I even tried adding a `root` file with a password under secrets/ and imported it into user/root/default.nix but ... alas! It can't be that simple can it?
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<fuzzypixelz>
the exact errror is: "- Neither the root account nor any wheel user has a password or SSH authorized key.
<fuzzypixelz>
You must set one to prevent being locked out of your system."
<Guest40>
set one?
Dr8128 has quit [Ping timeout: 265 seconds]
lor|d has quit [Ping timeout: 260 seconds]
devhell has quit [Quit: leaving]
lor|d has joined #nixos
waleee-cl has joined #nixos
<fuzzypixelz>
Guest40: I don't know what that means exactly! even adding a blatant line such as users.users.root.password = "my-very-strong-password"; didn't work
Qwerky has joined #nixos
<Guest40>
fuzzypixelz is this after installation?
<Guest40>
I think you supposed to set password after installation with passwd command
<Guest40>
or it asks for root password I think
<Guest40>
during installation
<Guest40>
if you can do "sudo su"
<Guest40>
the do "passwd:
<Guest40>
"passwd"
Chiliparrot has joined #nixos
<Guest40>
type password for root account
<fuzzypixelz>
Guest40: I actually already have a working nixos system with my own flake config. My flake doesn't complain and still builds just fine. It's just devos that's throwing that error
<Guest40>
oh in that case I don't know, Im nixos noob
<Guest40>
let me ask you my problem, do you use firefox on nixos?
<simpson>
Guest40: Go to `about:support` and double-check that your GPU's actually being used. A lack of WebGL shouldn't matter, as long as your compositing is accelerated.
<fuzzypixelz>
but I don't experience random freezing either
<redmp>
what's the correct way to instruct nixops to use a custom nixpkgs? the only thing I've found so far says to set NIX_PATH in shell.nix, but that doesn't seem to work when nixpkgs has overlays applied
<bb_>
also I dont see any PKG_CONFIG_PATH variable
Qwerky has joined #nixos
<cole-h>
Note how the direnv export happened *after* you ran a command
<cole-h>
If you try again it should work.
<redmp>
when i do `import <nixpkgs> {}`.path or `import <nixpkgs> { overlays = [ /* some nontrivial overlay */ ]; }`.path i get hte same path.. how can i obtain a path for nixpkgs+overlays?
<bb_>
hmm
respawn_ has quit [Quit: Leaving]
ahmedelgabri has quit [Ping timeout: 244 seconds]
Qwerky has quit [Remote host closed the connection]
<sterni>
redmp: can't get a path for something that isn't a path but a value
<sterni>
redmp: what do you need the path for?
<bb_>
cole-h, thanks for the heads up. Tried it again with nix-shell .project_env/shell.nix
<bb_>
when you have pkg-config in your nix-shell, do you have a PKG_CONFIG_PATH variable?
Dr8128 has joined #nixos
<sterni>
redmp: you can set it in the configuration of whatever you are deploying
`slikts has quit [Quit: No Ping reply in 180 seconds.]
<sterni>
redmp: and should influence the package in the deployment then
`slikts has joined #nixos
`slikts has quit [Changing host]
`slikts has joined #nixos
<redmp>
ah, i see.. in that case each machine configuration would import the flake and set its own nixpkgs.overlays option? are there any efficiency concerns of having those separate imports?
supercoven_ has joined #nixos
mbrgm has joined #nixos
zakame has quit [Ping timeout: 264 seconds]
AlpineLlama has joined #nixos
stree has quit [Ping timeout: 244 seconds]
<bb_>
I also remove the direnv and execute shell.nix directly
<sterni>
redmp: well it will have to be evaluated either way and flakes promise evaluation caching at least, so I'd guess there wouldn't really be a difference
<redmp>
sterni: the nixpkgs-overlays approach didn't work.. it copied my overlays.nix to the store, but since that file contained import ../. to access my flake, it resolved to /nix/default.nix lol..
extrowerk has left #nixos [#nixos]
<redmp>
i'll try the nixpkgs.overlays approach next
<redmp>
it produced a deploy plan at least, showing that it's going to download/build all the stuff from the pinned nixpkgs w/o building the dependencies of the flake
<redmp>
and it didn't give an attr error for the package added by the flake's overlay
bb_ has quit [Quit: Leaving]
bahamas has joined #nixos
missionformilk has quit [Ping timeout: 240 seconds]
ahmedelgabri has joined #nixos
<ploupiboulba[m]>
with awscli2, `$ aws ` prints `cannot import name 'get_event_loop' from 'prompt_toolkit.eventloop' (/nix/store/6cd1kpn48f19677jzj8jh12xiksyi9vz-python3.7-prompt_toolkit-3.0.4/lib/python3.7/site-packages/prompt_toolkit/eventloop/__init__.py)` . Kinda weird since the wrapped executable references prompt_toolkit 2.10 (and the package derivation adds a python overlay to specify propmpt_toolkit 2.10 :s
worldeva has joined #nixos
missionformilk has joined #nixos
everythingTaken has joined #nixos
seku has joined #nixos
Ox4A6F has quit [Quit: authenticating]
Ox4A6F has joined #nixos
hyiltiz has joined #nixos
hyiltiz has quit [Changing host]
hyiltiz has joined #nixos
grimvar has joined #nixos
`slikts has quit [Quit: No Ping reply in 180 seconds.]
<Synthetica>
Gimp fails to start with "GEGL operation missing! GIMP requires the GEGL operation "gegl:introspect". This operation cannot be found. Check your GEGL install and ensure it has been compiled with any dependencies required for GIMP.", any ideas?
<sterni>
everythingTaken: no idea, use firefox myself on wayland and never had this issue
<{^_^}>
nixops#452 (by expipiplus1, 4 years ago, open): Documentation is not clear on how to fix nixpkgs version
bahamas has quit [Quit: leaving]
ris has joined #nixos
Athas has joined #nixos
<Athas>
I am disappointed to discover that /bin/sh is bash on NixOS. I found a way to change this to be a plain POSIX shell instead, but also advice against doing so, as apparently much of NixOS assumes /bin/sh is bash. Is this true, or has anyone here successfully used e.g. dash as /bin/sh?
<ajs124>
Athas: which kind of problems are you seeing? I seem to remember somebody using something else there.
<sterni>
everythingTaken: sway yep
<sterni>
redmp: well flakes are experimental after all and nixops doesn't even have a stable release that supports them
<Athas>
ajs124: I haven't tried it yet. I would prefer /bin/sh to be dash, but it's not so important to me that I'm willing to risk random breakage in core OS systems.
<everythingTaken>
thanks
<Athas>
So basically I'm asking if anyone already took a trip into the minefield.
emilsp has joined #nixos
everythingTaken has quit [Quit: everythingTaken]
<redmp>
sterni: not flakes, i meant just pinning nixpkgs in nixops deployment
<redmp>
sterni: it several sources recommend setting NIX_PATH in the shell environment where you do nixops commands
<redmp>
sterni: a few other sources suggest using -I to specify the nixpkgs path
<redmp>
sterni: both options seem somewhat substandard.. the first works, but it's a bit of action at a distance.. what if something overwrites my shellenv before i call nixops?
<redmp>
sterni: not expecting you to have a solution; just explaining what i've understood in case i'm wrong :)
<ajs124>
Athas: Ah, ok. I assumed you did, since you said "apparently much of NixOS assumes /bin/sh is bash"
<simpson>
Athas: The build shell is always going to be bash, but I don't know what specifically breaks if /bin/sh links to dash instead. Feel free to try it and see, and report bugs.
Guest40 has joined #nixos
hiro98 has quit [Remote host closed the connection]
<{^_^}>
#74627 (by 06kellyjac, 1 year ago, open): `/bin/sh` & `/run/current-system/sw/bin/sh` is symlinked to a shell with bash-isms
<ajs124>
The comment in the option (environment.binsh) that warns about using it is also from 2013 and since the option isn't visible, so it's probably not used much.
<Guest40>
andi- andi[m] which one is firefox maintainer?
<ajs124>
neither
<andi->
neither
<sterni>
Athas: I think there is hope since it is not always bash, but sometimes busybox sh in nixpkgs bootstrapping
<sterni>
Athas: and most nixpkgs shellscript should have patched shebangs
<Athas>
simpson: I will try! I don't mind the derivation build shell being bash, but I just got bitten by accidentally using bashisms in a random shell script, and I would rather blame the operating system than myself, as is tradition.
<simpson>
Athas: I know that feeling. I've become more reliant on bash the past few years, but not for any essential reasons. It just kind of piles up.
<fuzzypixelz[m]>
hello again. I'm trying to use https://github.com/colemickens/nixpkgs-wayland in my nixos flake, it looks like this: https://github.com/divnix/devos. If you bear with me (: ... I added nixpkgs-wayland to the overlays declared in extern/default.nix. Problem: this didn't overwrite the input of nixpkgs-wayland and tried to build it against my nixpkgs, which is release-20.09 ...
redmp has quit [Ping timeout: 265 seconds]
<fuzzypixelz[m]>
this is basically me being unfamiliar with how devos works ... yay
<cole-h>
Guest40: Nixpkgs changes rapidly, so some info may be out of date; the 2 current maintainers of pkgs.firefox are Eelco and lovesegfault
jess has quit [Quit: update innit]
szicari has joined #nixos
ml| has joined #nixos
j has joined #nixos
<fuzzypixelz[m]>
colemickens: maybe you're here ? :P
<simpson>
Guest40: Did we establish whether you're using Xorg or Wayland? Does anything else (e.g. PDF viewer, terminal emulator) have this behavior as well?
<Guest40>
cole-h oh, thanks for the info
neiluj has joined #nixos
neiluj has joined #nixos
neiluj has quit [Changing host]
domogled has quit [Quit: domogled]
lsix has joined #nixos
<Guest40>
simpson I am on Xorg, everything else works fine. If you read github issue one guy saying it happend on Plasma moving to Gnome fixed it, below it second guy saying issue was on Gnome moving to plasma fixed it. I am on Xorg + Xmonad, tried Xorg + dwm too
<fuzzypixelz[m]>
cole-h: you're actually here haha I wasn't paying attention to the chat. Any ideas about why the overlay packages build against release-20.09?
<cole-h>
fuzzypixelz[m]: Wrong cole ;) colemickens and I are different people (that share the same awesome first name)!
<Guest40>
I am dedicated to fixing this issue as I like NixOS very much so far :)
* lovesegfault
reads the issue
<fuzzypixelz[m]>
cole-h: TIL
<lovesegfault>
Guest40: I cannot reproduce on Wayland with AMD gpu
<Guest40>
lovesegfault I installed it yesterday, NixOS 20.9 stable + firefox 86 (Xorg) + tried firefox-bin 86 too
<fuzzypixelz[m]>
cole-h: still if you're even marginaly familiar with how devos Extern works please give it a look
<lovesegfault>
Guest40: Out of curiosity, have you tried `firefox-bin`?
<simpson>
Guest40: In `about:preferences`, look for smooth scrolling. (Search "scroll".) Try toggling it. Things *will* behave differently, but possibly more comfortably.
<Guest40>
lovesegfault yes, currently using it
Qwerky has joined #nixos
<lovesegfault>
Interesting, what about the non-bin pkg, same issue?
<Guest40>
simpson it's enabled by default
meh` has joined #nixos
<Guest40>
lovesegfault yep, exactly as other people reported in github issue
<jeaye>
How can I specify the run-time dependencies of a derivation?
<jeaye>
Not the build-time dependencies.
<cole-h>
fuzzypixelz[m]: Not at all, unfortunately :D devos, while cool, is extremely over-engineered for my needs :P Maybe it would help if you post what you've tried somewhere, to see if you've only made a simple mistake?
<Guest40>
lovesegfault I checked irc logs, I think some guy reported this starting from 70 something version of firefox
<Guest40>
72 I think
<Guest40>
what might be done differently, is it the wrapper? or how xorg server + drivers setup in nixos? I really want to solve this issue as I like nixos very much so far. Let me know which logs you want to look at etc
<Guest40>
lovesegfault
<lovesegfault>
Guest40: I don't know, I'm only the FF maintainer so someone reviews the submitted patches, as Eelco never does. I don't really have bandwidth to chase after this issue :/
<lovesegfault>
My first step to debugging it would be to strace it
<lovesegfault>
and see if when it freezes strace shows anything useful
<Guest40>
lovesegfault understandable, how to 'strace' it
<lovesegfault>
If that doesn't yield any useful info, then you might want to build it with debug symbols and reproduce the issue under `rr`
<lovesegfault>
and chase after it in a debugger
<lovesegfault>
to be clear, it's a lot of work
<lovesegfault>
You should also check bugzilla and see if the issue is reported there
<Guest40>
lovesegfault thanks, I will get back to you, one more thing trying to create new profile also segfaults for me, you seem to love segfaults so might be your are of interest :D
<lovesegfault>
and if it isn't, report it
<Guest40>
thanks cole-h
<lovesegfault>
`coredumpctl debug` and submit the stacktrace in a new issue
<Guest40>
lovesegfault I did strace -yy firefox and it launched firefox then strace command finished executing
<lovesegfault>
You might need to make strace follow child processes
<lovesegfault>
I don't recall the cli flag for that, it's in the manpage
<simpson>
...Hm, I hope that they tried *toggling* smooth scrolling; the typical pattern is that it's enabled by default, it lags the GPU every few moments with too many draw commands, and disabling it makes for an overall improved experience.
<fuzzypixelz[m]>
the relevant issue has to do with the `nixpkgs-wayland` overlay under extern/default.nix
szicari has joined #nixos
<fuzzypixelz[m]>
the overlay is building against `release-20.09` for some reason
<cole-h>
and what are you expecting / desiring to happen?
dev_mohe has quit [Quit: dev_mohe]
rj_ has quit [Remote host closed the connection]
<simpson>
Guest40: That's really curious and irritating. I wonder what's causing this, then. We've gone through all the usual suspects.
monotux has joined #nixos
<fuzzypixelz[m]>
cole-h: I'm expecting to build against `master`... The thing is, the build fails for `release-20.09` mason throws an error because it's supposed to be `=> 0.56` which the unstable version, 20.09 is on `0.55.1`.
<fuzzypixelz[m]>
Even more, "Automated, pre-built packages for Wayland (sway/wlroots) tools for NixOS (nixos-unstable channel)." is the first sentence on https://github.com/colemickens/nixpkgs-wayland
<Guest40>
simpson yes, I am on this issue since yesterday lol
<Guest40>
simpson do you have some experience with strace gdb debugging?
<cole-h>
fuzzypixelz[m]: The nixpkgs-wayland overlay uses `prev.callPackage`, so that's probably why it's not working as you'd expect.
<simpson>
Guest40: Sure, although that's not how I'd continue to investigate this. I'd start doing a differential analysis instead: Does this stuttering happen with other GL apps? With other XRender apps?
<Guest40>
simpson i have vscodium does that count? the things is it doesn't freeze right away, only when opening new tab and focusing on address bar and preferences scroll, segfaults when trying to create new profile
<Guest40>
so I am not sure which part of vscodium I could check unless poking here and there
<simpson>
Guest40: Huh. I wonder if this is a problem in nouveau? You should know that reporting bugs to GPU drivers usually doesn't end well; it's very hard to reproduce what exactly is wrong.
<simpson>
Still, if you have crash reports, you can send them along to whoever says that they'll read them.
<Guest40>
simpson this is when trying to create new profile (crashes)
<{^_^}>
[nixpkgs] @MatthewCroughan opened pull request #116783 → freecad: symlink uppercase executable names to lowercase in postFixup → https://github.com/NixOS/nixpkgs/pull/116783
stree has joined #nixos
<Guest40>
simpson what is libpci, is it something in nouveau missing/
<simpson>
Guest40: libpci's part of how to talk to GPUs. You could check demsg for GPU resets or crashes or oops.
<matthewcroughan_>
supersandro2000: hopefully I'm not annoying you with my PRs, I'm getting better I hope :D
<matthewcroughan_>
clever: a friend and I want to make a function that pulls keys from Github, it will be something like fetchSshPubKeyFromGithub
<matthewcroughan_>
when he runs nix-build against it, the function has no networking
<matthewcroughan_>
how do the fetchers get around this problem? We just want to run curl inside a function.
<clever>
matthewcroughan_: you either need to use builtins.fetchurl (impure) or make it a fixed-output drv
<clever>
matthewcroughan_: if you make it a fixed output drv (declare the hash of $out), then nix will unlock networking, and fail the build if you dont match that hash
andoriyu has quit [Ping timeout: 265 seconds]
<matthewcroughan_>
clever: is there a small example using curl for this? O.o
<numkem>
I'm currently getting some std::bad_alloc errors in my hydra-queue-runners logs when it tries to send a build job to a remote node... anyone else having this? Using latest unstable nixFlakes
<sparogy>
matthewcroughan_: you would use nix for this presumably
<clever>
matthewcroughan_: or the user changed his keys
<matthewcroughan_>
yup, in which case you'd want it to let you know
<matthewcroughan_>
omg maybe this should even be part of the flake
<clever>
matthewcroughan_: if a derivation has the above 3 attrs, it is fixed-output, and it will get network access
<matthewcroughan_>
then you could update your inputs, screw fixed output derivations
<clever>
matthewcroughan_: outputHashMode=flat means that $out must be a single file, outputHashMode=recursive means $out can be anything (file, directory, symlink)
<clever>
when in flat mode, it is just the raw hash of the file, thats how things like yarn2nix translate a yarn.lock into a drv
<matthewcroughan_>
the flake.lock would contain all of the user's hashed keys
<clever>
when in recursive mode, its the hash of `nix-store --dump $out`, which serializes it as a NAR
<redmp_>
i'm trying to import the overlays only once in a large nixops deployment ... currently they're imported in each machine configuration, but it's very slow to import many times
bebarker has joined #nixos
<matthewcroughan_>
clever: " nativeBuildInputs: A list of dependencies whose host platform is the new derivation's build platform, and target platform is the new derivation's host platform. This means a -1 host offset and 0 target offset from the new derivation's platforms."
z has joined #nixos
<matthewcroughan_>
This doesn't make English sense to me. Can you translate?
<sterni>
redmp_: it doesn't have a path in that case
gustavderdrache has quit [Quit: Leaving.]
<matthewcroughan_>
In fact, can anybody explain what that means to me
alexherbo2 has quit [Remote host closed the connection]
<clever>
matthewcroughan_: anything your running at build time, which must be for the host cpu
<clever>
matthewcroughan_: cmake, pkgconfig, and so on
z is now known as Guest95310
<sterni>
redmp_: a path in that case needs to be import-able
<matthewcroughan_>
Correct, but what I quoted has nothing to do with what you said.
<matthewcroughan_>
I understand that it means "build time dependencies"
<matthewcroughan_>
But what I quoted. What does that mean?
<clever>
"host platform is the new derivation's build platform"
<sterni>
(with some extra quirks I think it can download stuff and unback tarballs)
<matthewcroughan_>
"This means a -1 host and -1 target offset "
gianarb- has quit [Ping timeout: 246 seconds]
<matthewcroughan_>
What does this mean?
attila_lendvai_ has joined #nixos
zal has quit [Remote host closed the connection]
<clever>
host platform (what the binary was made for) is the derivations build platform (where the drv will be built)
attila_lendvai has quit [Read error: Connection reset by peer]
gianarb has joined #nixos
<clever>
matthewcroughan_: those offsets are something internal to setup.sh in the stdenv, ive not figured them out either
<matthewcroughan_>
We are clearly of a lesser intelligence. Nixinator says "bless you" :D
<redmp_>
sterni: is there some way i could make an importable path for a flake's overlay?
<redmp_>
sterni: i'm using flake-compat and so default.nix is a stub that gives me access to most things
<cole-h>
fuzzypixelz[m]: prev.callPackage uses the callPackage from *your* Nixpkgs, so it won't hit the cache (since your callPackage is 20.09, and nixpkgs-wayland uses nixos-unstable)
alexherbo2 has joined #nixos
realrokka has joined #nixos
alexherbo2 has quit [Client Quit]
zupo has joined #nixos
codygman` has quit [Ping timeout: 256 seconds]
<sterni>
redmp_: you can probably just put (import ./default.nix {}).flake.attr in a file
slack1256 has joined #nixos
acarrico has joined #nixos
<viric>
hm does anybody know what are the challenges to have a program linked with two versions of the same lib on purpose?
<viric>
let's say I want to write a program that compares two libpng versions.
<Miyu-saki>
I think NixOS's container module assumes that overrides don't exist.
<redmp_>
sterni: yeah, that doesn't work i think.. i just get a file in the nix store that says "import ./default.nix"
<redmp_>
sterni: i'm using it like this `NIX_PATH = "nixpkgs=${pkgs.path}:nixpkgs-overlays=${../overlays.nix}";`
andymandias_ has quit [Ping timeout: 246 seconds]
<viric>
clever: at least I hope dlopen will work. Or not?
<redmp_>
sterni: that `../overlays.nix` just copies verbatim to the nix store, and so the imports performed therein don't work .. i don't think i can `import ./overlays.nix` because that will return the list of overlays themselves..
<clever>
viric: dlopen, and using dlsym on that handle, might work
<sterni>
redmp_: if you set it in NIX_PATH or -I you don't have to use a nix-style path can just use the normal absolute one
hyper_ch2 has quit [Remote host closed the connection]
andymandias has joined #nixos
<sterni>
redmp_: in nix do ${./..}/overlays.nix, which will copy the directory as well
hyper_ch2 has joined #nixos
<clever>
redmp_: a secondary problem with those ideas above, is that the NIX_PATH will change on every rebuild, but only take effect after shells are re-launched
<clever>
redmp_: you have to also add them to nixpkgs.overlays for each machine in the nixops network
<viric>
clever: what if I dlopen libA and libB? the two libpng symbols won't intermix?
<clever>
viric: not sure
<viric>
clever: I cannot find a document describing how symbols get resolved
<redmp_>
clever: both places? i thought the point of putting it into NIX_PATH="nixpkgs-overlays=..." was to prevent needing to explicitly apply to each machine in the network
<clever>
redmp_: putting it in NIX_PATH only effects the runtime of the given machine, but not how the machine itself is built
<redmp_>
clever: yeah, that's essentially what i had before.. i'm not sure now what benefit there was to adding to NIX_PATH="nixpkgs-overlays=..." if the overlays have to be separately imported by each machine
<redmp_>
clever: the problem i'm having is that importing the overlays takes a couple of minutes _per machine_
<redmp_>
clever: so i'm trying to only import once, but that seems to be not possible
lsix has joined #nixos
<clever>
redmp_: have you seen the defaults key in nixops?
<redmp_>
clever: yeah, i wasn't using that for this package.. i'll try it and see
<sterni>
redmp_: is the flake really necessary? flake-compat + flake evaluation is pretty costly
<redmp_>
sterni: yeah, the flake fell out of discussions last week here on #nixos ... i have several haskell packages split across three repositories with interdependencies and my previous nix code which did ad-hoc imports of repositories was cumbersome and aparrently pulled multiple versions of nixpkgs into deployed machines' stores ..
<redmp_>
sterni: so i learned about flakes and overlays and now i'm back at trying to deploy :)
<redmp_>
clever: yeah, the run times i'm quoting are for --dry-run, so it's all nix evaluation.. i know which flake it is too, but i thought there would be a way of at least only evaluating it once for a deploy
<clever>
if the flake stuff is working fully, it should cache the eval
<clever>
but nixops may not be capable of that yet
<redmp_>
clever: nixops is importing a flake-compat default.nix in my project repository to access the pinned nixpkgs and overlay
<clever>
redmp_: flake-compat wont have any of the flake eval caching
tarruda3 has joined #nixos
<redmp_>
clever: yeah, thats fine.. i'm used to slow build times for this project .. i'm just not sure i want to pay machines*build-time for every deploy :)
lsix has joined #nixos
<clever>
redmp_: --include can save some of that time
leo60228- has joined #nixos
redmp has joined #nixos
leo60228 has quit [Ping timeout: 265 seconds]
philr_ has joined #nixos
redmp_ has quit [Ping timeout: 256 seconds]
<abathur>
<3 supersandro2000
<{^_^}>
supersandro2000's karma got increased to 39
<supersandro2000>
abathur: I was kinda waiting for you that it does not break resholve
<abathur>
nod, it looked okay
<abathur>
I've been kinda noodling that dependency; in theory resholve could probably be used to package bats, but then...
<supersandro2000>
cyclic, right?
<abathur>
yep :)
slack1256 has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @andir pushed to release-20.09 « firefox: drop maintainership »: https://git.io/JmwBr
alexherbo2 has joined #nixos
<redmp>
clever: i'm deploying an integrated distributed system so --include isn't something i'll ever use
addcn has joined #nixos
alexherbo24 has quit [Ping timeout: 264 seconds]
<redmp>
clever: i'm going to try extracting the package from a single `import flake.inputs.nixpkgs { overlays = [ flake.overlay.${builtins.currentSystem} ]; }.blah.blah` and then threading that argument down to the hosts.. it blows up all the nice abstractions, but if it saves me on 20 minute deploys for <20 hosts then i'm happy
<addcn>
hey everyone -- there's a cool project that was posted on hn today called outrun (https://github.com/Overv/outrun#outrun). Based on its caching strategy, I don't think it's immediately usable on NixOS, but it got me thinking. If we pin both the source and the destination machine to the same package versions, then we essentially can just skip the package step. Does anyone know of any projects that do
<redmp>
sterni: it's a big mess; don't bother with it.. point is that this is an interaction thing, not a "flakes are slow all the time" thing .. i think
worldeva has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @AndersonTorres merged pull request #116783 → freecad: symlink uppercase executable names to lowercase in postFixup → https://github.com/NixOS/nixpkgs/pull/116783
<rmcgibbo[m]>
wow, the cool thing about randomly reviewing backlogged PRs to "leaf" packages in nixpkgs... there's a ton of really neat stuff that i never would have known how to search for.