<clever>
pikajude: --option can be used to override any field in nix.conf
<clever>
DariusTh`: domenkozar[m]
<clever>
fresheyeball: is it being ran on a git repo?
<clever>
fresheyeball: oops, fetchGit
<clever>
fresheyeball: builtins.fetchgit ./. will get you that
<clever>
fresheyeball: did you get new .drv files, and are you using cleanSource?
<clever>
DariusTh`: that sounds like the nix expr is just broken
<clever>
DariusTh`: you have to restart nix-daemon if you change nix.conf
<clever>
DariusTh`: is cachix configured in /etc/nix/nix.conf ?
<clever>
DariusTh`: ls -lh ~/.config/nix/nix.conf
<clever>
2019-08-23 01:13:55< ryantm> "We do not recommend using ~/.config/nix/nix.conf unless you are a power user. It's simpler to have only one global config file."
<clever>
DariusTh`: in your home dir
<clever>
DariusTh`: check your local machine to see if the file i just pasted exists
<clever>
DariusTh`: does that file exist?
<clever>
2019-08-23 01:13:55< ryantm> "We do not recommend using ~/.config/nix/nix.conf unless you are a power user. It's simpler to have only one global config file."
<clever>
fresheyeball: cleanSource will already ignore .git by default
<clever>
fresheyeball: its in the nixpkgs manual i believe
<clever>
cleanSource can clean most of those up
<clever>
fresheyeball: also, try using `diff -r /nix/store/ia83szf79darxc2m1275f0vx1h6ykix2-kassir /nix/store/9srk6j3zx2qhbaasy0v9dnv7pgh0yf35-kassir`
<clever>
fresheyeball: your ./. doesnt match perfectly between the 2 machines, so its considered a different version
<clever>
fresheyeball: you need to use lib.cleanSource when doing ./.
<clever>
and matrix is back to matrix is once more!
<clever>
fresheyeball: seperately, do `nix-instantiate something.nix -A something` on your project, on both machines, then do `nix-copy-closure --from 192.168.2.10 /nix/store/something.drv` to get both versions of the drv on the same machine, then use nix-diff to compare the 2
<clever>
fresheyeball: first, do `nix-store --query --binding out /nix/store/foo.drv` to get the $out, so they are the same type
<clever>
fresheyeball: if you have a /nix/store/foo and a /nix/store/foo.drv, dont bother trying to compare them
<clever>
you want the output path inside the drv
<clever>
the hashes will never match
<clever>
dont bother comparing drv and non-drv files
<clever>
fresheyeball: use nix-copy-closure to get the .drv from both of them onto the same machine, then run nix-diff on both drv files
<clever>
fresheyeball: is nix-build creating the exact same $out ?
<clever>
fresheyeball: check `nix-store -r -vvv` again on a path it should pull
<clever>
fresheyeball: that means it worked, try with nix again
<clever>
fresheyeball: make a second key for the cache usage
<clever>
fresheyeball: root must have an ssh key in its ~/.ssh, that lacks a passphrase
<clever>
fresheyeball: check its timestamps
<clever>
2019-08-23 01:13:55< ryantm> "We do not recommend using ~/.config/nix/nix.conf unless you are a power user. It's simpler to have only one global config file."
<clever>
fresheyeball: so you lied :P
<clever>
fresheyeball: or just: ls -lh ~/.config/nix/nix.conf
<clever>
fresheyeball: `find ~ -name nix.conf`
<clever>
fresheyeball: do you have a nix.conf in $HOME?
<clever>
fresheyeball: do they agree?
<clever>
fresheyeball: compare /etc/nix/nix.conf to `nix show-config | grep substituters`
<clever>
fresheyeball: or you have a 2nd nix.conf in $HOME, or you used mkForce
<clever>
fresheyeball: try adding a -vvv, and use a pastebin for the full log
<clever>
and we got matrixed!
<clever>
and on the other one?
<clever>
fresheyeball: what happens if you run `nix-store -r /nix/store/891q5c3xk5ipdpx89zvwah381vnq2m9f-kassir` ?
<clever>
fresheyeball: find a path on one machine, and not another, preferably also not in the upstream cache (like `ls -l /run/current-system`) and then run `nix-store -r /nix/store/thatpath` on the other machine
<clever>
fresheyeball: what was the error message?
<clever>
you want the other option
<clever>
fresheyeball: trustedBinaryCaches does almost nothing, it jsut allows untrusted users to opt-in
<clever>
fresheyeball: and on the other machine, what is in trusted-public-keys on nix.conf?
<clever>
fresheyeball: on a path that is failing to copy
<clever>
fresheyeball: yeah
<clever>
fresheyeball: run this, on a path that is failing to copy, what signatures are on it?
<clever>
but you can just ignore config.nix entirely
<clever>
zajko: you would need to edit it manually, theres not really an automated way to do it yet
<clever>
$ nix-instantiate ~/.nixpkgs/config.nix --eval -A allowUnfree
<clever>
true
<clever>
jonreeve[m]: i can then just programs.vim.fat = false; to trim the fat
<clever>
jonreeve[m]: this is the config for my vim, and it also defines a programs.vim.fat option, that controls if my vim config should include some heavy deps
<clever>
secret-key-files goes into nix.conf, via nix.extraConfig
<clever>
give each machine its own keypair, and trust the public of the other
<clever>
and look at the generate command in `man nix-store`
<clever>
secret-key-files = /etc/nix/signing.sec
<clever>
ah right
<clever>
then just add the ssh://user@host to the binary cache list
<clever>
fresheyeball: nix.sshServe.enable is one option
<clever>
fresheyeball: register each as a binary cache for the other
<clever>
nested virtualization is bad
<clever>
and my main windows system, is in virtualbox
<clever>
another problem, with anything hyper-v based, is that hyper-v needs vm extensions in the cpu
<clever>
and last i looked, you only had 1 rootfs dir, so you cant really play around much
<clever>
then use an atomic rename in explorer, to swap the rootfs dirs
<clever>
so, you must install ubuntu, then use the linux tar to unpack the rootfs within WSL
<clever>
something about how you need a special ntfs attr on every "linux" file, and windows explorer (along with nearly every fs tool out there) cant make them, and destroy them when copying
<clever>
i did briefly look into how the old WSL worked, and how to get other OS's into it, and it was pretty ugly
<clever>
since proper vm's existed by that point
<clever>
nobody bothered to port it to 64bit windows
<clever>
and it only worked on 32bit windows kernels
<clever>
but, it never got multi-core support, so linux could only use a single core of the machine
<clever>
ldlework: so in the end, windows thinks colinux is a network driver, that hogs a lot of cpu time, and needs 3gig of ram
<clever>
when linux is done, and wants to "idle" the cpu, it will restore the state, and just do a normal return, back into the windows kernel
<clever>
ldlework: and because the driver is in kernel space, it will just take the entire CPU hostage, and start running linux!
<clever>
ldlework: and windows things colinux is a network driver, so when packets arrive for linux, it runs a function in the "driver" to handle the packet
<clever>
ldlework: when linux needs to schedule a timer, it will use the windows kernel timer subsystem
<clever>
ldlework: colinux, is basically a fork of linux, that will impersonate a windows driver, without needing any hypervisor (it was made in the days before hardware vm extensions)
<clever>
ldlework: that reminds me of colinux, have you heard of that before?
<clever>
a mount namespace just changes the pointer for what / is
<clever>
so you can only see a subset of them
<clever>
for example, on linux, a process namespace, just changes the table for the process list
<clever>
containers are more about swapping out the pointer to the list of something
<clever>
but they may have had just as many global vars laying around, that forced them to make just as ugly of an implementation
<clever>
basically think of it like running a 2nd Xorg on :1 and :2
<clever>
i can see it being semi-related to containers
<clever>
ldlework: everything gui related, was a global kernel variable, so when you context switch between different processes, it has to swap out every single global var, to deal with multiple desktops (like RDP on another user)
<clever>
ldlework: ive heard horrors about how windows handles multiple desktops
<clever>
growpotkin: id enable and run memtest first
<clever>
spease: docker technically cant run on osx or windows, behind the scenes, its just launching a linux VM, and running docker inside that
<clever>
WhittlesJr: yeah, you can always change the nixpkgs after it installs
<clever>
WhittlesJr: thats strange, definitely sounds like a bug
<clever>
WhittlesJr: for a different path, something not libunistring
<clever>
WhittlesJr: what is said path?
<clever>
WhittlesJr: what is the output for one it cant build?
<clever>
or look at the .drv's that have .lock files, but run the query on the version without .lock
<clever>
ctrl+c it right after it starts to build, and scroll up some
<clever>
try again with a .drv its actually building
<clever>
its already built
<clever>
then thats not something it needs to build
<clever>
WhittlesJr: what happens if you run `nix-store -r /nix/store/q45f8b6wp4bynxbrlw2rjqjq9g5k47yb-libunistring-0.9.10` ?
<clever>
WhittlesJr: i just need the output for a single drv, ignore and lock and tar files
<clever>
WhittlesJr: add --store local?root=/mnt/
<clever>
WhittlesJr: what path does it output?
<clever>
WhittlesJr: grab one of the .drv files it wants to build, and run `nix-store --query --binding out /nix/store/foo.drv`
<clever>
WhittlesJr: that definitely sounds like the binary cache didnt work
<clever>
WhittlesJr: can you give an example of a storepath its trying to build?
<clever>
WhittlesJr: did you mess with the binary cache config any? is the date set right?
<clever>
WhittlesJr: does the machine have network access?
<clever>
danbst: which also lets you fake the player name
<clever>
danbst: multimc has an option to play offline
<clever>
danbst: multimc handles forge, mods, and auth for you
<clever>
wedens: yeah, it wont help with boot times, but it will help with normal running
<clever>
wedens: the main downside, is that the l2arc cant persist between reboots
<clever>
wedens: the l2arc is just an extension of the l1arc(the one in ram), which will use a combination of recently and frequently used things, and adjust the ratio to get a high hit rate
<clever>
wedens: zfs will basically use the ssd as a read-cache to speed up reads from the hdd
<clever>
wedens: zfs, with the ssd as a L2arc may be a good option
<clever>
philipp[m]: pkgs.makeDesktopItem
<clever>
craige: sounds good
<clever>
craige: read the nixos module to see what file the strings winds up in, then check the postgres docs to see what that file does
<clever>
craige: i think your best bet is the official postgresql docs
<clever>
selfsymmetric-mu: if you run $$, then it will look for the pid of bash in PATH, and then say command-not-found
<clever>
selfsymmetric-mu: if you eval it, yes
<clever>
Fare: yeah, the expr to build the app itself can be shared, youll just need some differences to make an image around it
<clever>
Fare: docker doesnt use image files much, and systemd doesnt really work in docker
2019-09-04
<clever>
fresheyeball: any time you do src = ./.; it will copy . to the store
<clever>
c: nix can cross-compile to windows!
<clever>
b: death to windows :P
<clever>
a: windows users will need to continue to build things the old non-nix way
<clever>
somebody in here yesterday had something in ~/.ghc that was causing every single package from ghcWithPackages to be hidden
<clever>
and baring weird stuff in your ~/.ghc/, it entirely gets rid of the "but it works for me" problem
<clever>
there is also yarn2nix for nodejs
<clever>
Trubydoor: you would need to do vulkan-headers = master.vulkan-headers; in an overlay
<clever>
Trubydoor: you would make an overlay to change vulkan and vulkan-headers
<clever>
Trubydoor: have you tried the nixos-unstable-small channel?
<clever>
Trubydoor: remove the thing that depends on vulkan from your systemPackages
<clever>
boxdot: and filter it to greater then 256mb
<clever>
sqlite> select path, narsize/1024/1024 as narmb, datetime(registrationTime,'unixepoch') from ValidPaths where narsize > 256 * 1024 * 1024 order by registrationTime;
<clever>
boxdot: that will list everything in /nix/store (run it after the warning is done) and sort it by when it was added, and filter oh wait, wrong sql
<clever>
sqlite> select path, narsize/1024/1024 as narmb, datetime(registrationTime,'unixepoch') from ValidPaths where path like "%BindingAttributes%" order by registrationTime;
<clever>
equivrel: i had a PR on nixpkgs to fix that
<clever>
beyuu: run pavucontrol, is the VU meter moving when you play things, is it going to the right device?
<clever>
beyuu: run `alsamixer -c 0` and check if things are muted
<clever>
:D
<clever>
ar: 37
<clever>
__monty__: can you make a gist showing the output of `list` and `status` from each machine, and also the ip's of everything, and what directions work where?
<clever>
__monty__: if there are nat troubles, toxvpn will use a tcp relay as a fallback