<clever>
lejonet: you gave it a storepath, and it printed nothing?
<lejonet>
clever: ah, I need to give it a store path, makes sense xD
Myrl-saki has quit [Ping timeout: 260 seconds]
<sphalerite_>
Well! I've got a full-featured graphical browser running on nixos on my chromebook! … except it's kinda cheating since it's in a debian chroot
<clever>
it will list every root that depends on the given path
<clever>
sphalerite_: lol
<Mic92>
LnL: does those error messages also mean, that nixpkgs's header lack behind os x releases? https://logs.nix.ci/?key=nixos/nixpkgs.35425&attempt_id=3152f62e-57af-40b5-9485-8b5ea9392993
<lejonet>
clever: interesting, the 14 generations before the last one used the same dhcpcd, but something changed with the latest one (yet dhcpcd is the same version)
NixOS_GitHub has joined #nixos
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nixpkgs/staging ae9d4fa Tuomas Tynkkynen: Merge remote-tracking branch 'upstream/master' into staging
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to staging: https://git.io/vAVZo
kelleyNif has joined #nixos
Lisanna has quit [Remote host closed the connection]
Georgyo has quit [Quit: Connection closed for inactivity]
<clever>
if you run ls -lh on that path, how big is the file?
<dimensionless>
i can't really paste the output
<clever>
tab-completion will help a lot with that step
<dimensionless>
28k
<clever>
try running `less` on it
<clever>
what do the contents look like?
<dimensionless>
oh i see
<dimensionless>
its an html file
<clever>
404?
<dimensionless>
makes sense
<dimensionless>
not quite
<dimensionless>
im at uni so its the stupid login page
<clever>
login page?
<dimensionless>
yea
<clever>
thats what my first guess was, lol
<dimensionless>
hmm not sure how to login without installing any more packages
<clever>
but i assumed you had working internet, since you where on irc
<dimensionless>
i guess i can just use tethering for the moment
<clever>
is lynx or links installed?, are you on the graphical iso?
<dimensionless>
i am on the graphical iso, i guess i can just start kde then
<clever>
i think it includes a gui browser
erasmas has quit [Quit: leaving]
<clever>
systemctl start display-manager
<dimensionless>
oh wait im not on the graphical iso anymore bc its installed
<clever>
ah
xvetk has quit [Quit: Page closed]
<clever>
you may need to tether then, to install something that can render the anoying login page
<clever>
the last uni i visited had proper wpa auth, where every person had a unique name&pw, that went into the wpa config
<clever>
no more captive portals
<dimensionless>
they are so annoying
<dimensionless>
so anyways how do i remove the file cache
<clever>
nix refuses to mark it as valid when the hash is wrong
<clever>
so it will just re-download it every time
<clever>
until the hash is right
<dimensionless>
hmm
<clever>
just re-run whatever command you wanted to run
<dimensionless>
oh oops had ethernet plugged in thats why
<dimensionless>
sweet as
<dimensionless>
its workin now
<lejonet>
clever: all wise encyclopedia, how do I find the .drv file for a corresponding nix-store path? It would seem like nix-diff wants to look at the .drv file
<taohansen>
dimensionless: maybe curtail your stream-of-consciousness to just content so people on the channel can more easily help you? :-)
<clever>
lejonet: nix-store --query --deriver
<lejonet>
clever: ah it was deriver, I tried driver... :P
<dimensionless>
yea i get that a lot taohansen sorry
Myrl-saki has joined #nixos
<LnL>
Mic92: possibly, but it doesn't look like a header or linker error. we've had other problems with x509 in go before
<lejonet>
clever: ... "unknown deriver", must be because this system is handled by nixops or?
<clever>
lejonet: its possible that you just GC'd the .drv file that originally made it
<clever>
lejonet: only way to get it back, is to nix-instantiate the old version from nixpkgs
<lejonet>
clever: I haven't GC'd anything knowingly atleast
<clever>
lejonet: luckily, nix also keeps all copies of nixpkgs, in /nix/var/nix/profiles/per-user/root/channels*
<clever>
lejonet: so you could try to nix-instantiate -A dhcpcd, against the nixos in each generation
<clever>
then see if you get the right drv by chance
<lejonet>
clever: well, I am running stuff from a git checkout, so I have a git-revision string ^^
<clever>
lejonet: ah, that also helps
BlessJah has joined #nixos
<lejonet>
probably easier to try and replicate stuff with that :P
<clever>
lejonet: and reminds me, the nixos name tells you exactly which rev you want
<lejonet>
clever: mhm, and its conveniently stored in /etc/os-release of each generation :D
<clever>
that as well
pjan has quit [Quit: さようなら]
<lejonet>
awh crap, the working generation is from a branch that doesn't exist anymore :S
<clever>
did you GC your git repo?
<lejonet>
I cleaned up a bit in it, for doing the PR
k0001 has quit [Ping timeout: 252 seconds]
ckauhaus has joined #nixos
<lejonet>
nixops doesn't store the results somewhere, perchance? :P
<lejonet>
or does it remove that after copying the closure?
<clever>
by default, it doesnt root anything
<clever>
so normal GC can eat it
<clever>
but there are generations on the remote machine
<clever>
that root each version that has been deployed
<Unode>
hi all, with nix what would be the correct way to write a "if -> else if". I find "if then else" in the docs, but nothing explicit about chaining two ifs.
<clever>
nixops also has its own option to keep rollbacks at the deployment server
<lejonet>
yeah, all the generations are still on the remote machine, I haven't GC'd anything there
<clever>
Unode: if condition then a else (if condition2 then b else c)
<lejonet>
Unode: iirc you can nest the if then else just like ternaries
<Unode>
clever, lejonet thanks I'll give that a go.
ckauhaus has quit [Ping timeout: 240 seconds]
<lejonet>
clever: interesting, the non-functioning dhcpcd is a bit smaller than the functioning one
ottidmes has joined #nixos
chocopuff298 is now known as chocopuff
<cheshircat>
hey, does anyone know how to force a derivation to stay in the store through nix-collect-garbage -d operations?
<cheshircat>
I want the install file for Mathematica to stay in the store
digitus has quit [Quit: digitus]
<simpson>
_habnabit: You gotta pin it somehow.
<dimensionless>
so ugh another dumb question, im installing some dotfiles and i have a file called packages.nix, where do i need to put it so nixos-rebuild switch finds it
<simpson>
Oh derp, mistab.
<simpson>
cheshircat: You've gotta pin your file somehow. I'm guessing that you want to avoid GC'ing the installer because it's big?
<dimensionless>
oh im dum
<simpson>
dimensionless: You're not dumb, just new and impatient. Take it slow and you'll do fine.
<dimensionless>
so i just move everything into /etc/nixos right?
<simpson>
That's the easy way, yes.
<taohansen>
dimensionless: are you trying to manage your user environment with Nix?
<dimensionless>
im just stealing my friends dotfiles tbh
azertyieio has joined #nixos
<cheshircat>
simpson: yes
<lejonet>
Well this is annoying, I can run that dhcpcd perfectly fine on the source system (i.e. the nixops machine), maybe I should just go the static IP route with those nodes... atleast for now lol
<simpson>
cheshircat: There's no builtin functionality for this other than making a root (a symlink into the store from nix-build or similar) and keeping it in your home somewhere.
ericsagnes has joined #nixos
ckauhaus has joined #nixos
Yatekii has left #nixos [#nixos]
acarrico has quit [Ping timeout: 256 seconds]
ryanartecona has joined #nixos
ckauhaus has quit [Ping timeout: 240 seconds]
Wharncliffe has joined #nixos
NixOS_GitHub has joined #nixos
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nixpkgs/master cb6172c Tuomas Tynkkynen: fio: 3.4 -> 3.5
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to master: https://git.io/vAVW6
<lejonet>
oh well, here goes nothing, made dhcpcd persistent, so lets hope nothing breaks horribly this time lol
<skrzyp>
I'm also digging into that "home manager" thing, but I can't find anyone with a properly maintained, complete configuration as a reference
<skrzyp>
even the original author does not have any examples on his github
<Mic92>
instead follow documentation on home-manager's README.md on how to update home-manager
<skrzyp>
but I'm kinda worried what would happen when the home-manager won't be merged into base NixOS distribution and then they implement some similar features conflicting with HM
<Mic92>
skrzyp: it won't conflict
<Mic92>
you would have the choice
<aanderse_>
if i need kernel 4.15+ how do i specify that? i'm looking at boot.kernelPackages and pkgs.linuxPackages_4_15* but i'm not sure which one i need
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master b54e5b9 xeji: kubernetes: patch broken go version check...
<taohansen>
I did it guys! I wrote a derivation to`fetchgitPrivate` my org's repo (itself a spec for a NixOS AMI on AWS), a systemd service imports this derivation and runs `nixops deploy` on the repo to send it over to the AWS machine.
<atmegaboy>
Hello, my fellow Nixians. I am attempting to use avr-gcc in a nix-shell, and it appears that some (but not all) of the AVR-related header files are missing. Nix: `nix-shell -p avrgcc avrlibc avrdude simavr stdenv` (I'm really reaching now, pretty sure avrgcc is all I need to compile :p) Compiling with `avr-gcc -g -Os -mmcu=atmega328p -c demo.c` gives me a "fatal error: inttypes.h: No such file or directory".
<atmegaboy>
To get avg-gcc to show me where it is searching, I try: `fatal error: inttypes.h: No such file or directory`, which lists a bunch of into that includes header file paths. They mostly look like good Nix paths, but then there is this: `ignoring nonexistent directory "/nix/store/3y2fazm5i9560d3f974a63wdisgazb5w-avr-gcc-5.4.0/lib/gcc/avr/5.4.0/../../../../avr/sys-include"` and similar for "avr/include". Indeed, these files are truly mis
<taohansen>
One issue remains: every time devs update the code, the sha256 and the rev will change. Is there any way to automate the update of their variables in my fetchgitPrivate derivation?
<taohansen>
Or otherwise ignore them and just fetch the latest.
mbrgm has quit [Ping timeout: 240 seconds]
<atmegaboy>
I'm wondering, perhaps I have accidentally installed libraries as user programs? The nix-shell -A --attr flag confuses me a little, I see it used to fetch libs, but that does not jive with my understanding of "attributes".
ryanartecona has quit [Quit: ryanartecona]
mbrgm has joined #nixos
<clever>
taohansen: nix expects you to give a rev, not a branch, and that the versions of things dont change very often
<yorick>
can I turn builtins.fetchGit into a derivation?
thc202 has quit [Ping timeout: 240 seconds]
<yorick>
(I want to be able to get it with nix build, for unification purposes with my non-fetchgit derivations)
ckauhaus has joined #nixos
horsehorse has joined #nixos
<yorick>
I can do runCommand "terrible copier" "cp -R ${fetchgit...} $out"
ckauhaus has quit [Ping timeout: 240 seconds]
Supersonic has quit [Disconnected by services]
d4g_ has joined #nixos
horsehorse has quit [Ping timeout: 260 seconds]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] boronine opened pull request #35438: perlPackages.TextBibTex: fix dylib path for Darwin (master...bibtex-darwin-fix) https://git.io/vAVBq
NixOS_GitHub has left #nixos [#nixos]
Supersonic112 has joined #nixos
Supersonic112 is now known as Supersonic
atmegaboy has quit [Quit: Page closed]
<clever>
yorick: you might even be able to just symlink $out to the path
<clever>
yorick: inside runCommand
<yorick>
hmm, that's slightly less terrible, thanks
<yorick>
seems to be working, too
d4g has quit [Ping timeout: 240 seconds]
<jtojnar>
Is there some documentation for “installables” in Nix 2.0?
<jtojnar>
especially, if I run `nix build nixpkgs.bash` what is nixpkgs?
<clever>
jtojnar: i'm guessing its the channel name from ~/.nix-defexpr/, same as nix-env
<crucialrhyme>
what's the best place to start debugging a nixos system that won't boot (fortunately, i can go back to the old configuration, thanks to an excellent and robust design)
<crucialrhyme>
this is on unstable after doing nixos-rebuild switch --upgrade
johnsonav has joined #nixos
crucialrhyme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<clever>
step 1, wait for somebody to notice your question
<clever>
step 2, dont disconnect from irc
mk-fg has joined #nixos
mk-fg has quit [Changing host]
mk-fg has joined #nixos
hiratara_ has joined #nixos
hiratara has quit [Ping timeout: 240 seconds]
hiratara_ is now known as hiratara
smichel17 has quit [Quit: smichel17]
garbas has quit [Quit: WeeChat 1.9.1]
taktoa has joined #nixos
orivej has quit [Ping timeout: 265 seconds]
day|flip has joined #nixos
tnks has joined #nixos
<tnks>
Is Nix 2.0 on nixpkgs-unstable yet? I just did a channel upgrade, and all I see is nix-1.11.16.
<clever>
tnks: what about nixpkgs.nixUnstable?
<tnks>
clever: oh, I see, I have to opt in explicitly.
<tnks>
clever: and once I opt in, I'm guessing there's no turning back?
<clever>
tnks: i believe 1.11.16 is compatible, so you can switch back
<tnks>
what's the plan to obsolete nixpkgs.nix?
<clever>
modify nixpkgs.nix to be 2.0
<tnks>
With all the news of a "release" I'd assumed they'd throw the switch in nixpkgs.
lord| has quit [Ping timeout: 240 seconds]
<tnks>
yeah, I can do that, just wondering if there's more testing before that's done upstream.
<clever>
same
<day|flip>
i have this nix-2.0pre5968_a6c0b773 when i did nix-env -qa | grep nix-
<clever>
tnks: i already use nixUnstable on my laptop, since before it was "released"
ckauhaus has joined #nixos
<tnks>
day|flip: I see both nix-1.11.16 and nix-2.0pre5968_a6c0b773 (which makes sense to me... they are on different attributes).
lord| has joined #nixos
<tnks>
clever: okay, then "release" kind of sounds like just a press release, which is odd.
ryanartecona has quit [Quit: ryanartecona]
<tnks>
clever: how long have you been using it? I'll trust it's stable enough, but I feel I should ask about at least one's person's experience.
<tnks>
We're using Nix for a developer environment at work, and I'm wondering whether to switch the automation scripts to use Nix 2.0 at work.
<tnks>
Otherwise, I would worry far less about a personal computer.
<clever>
tnks: only been using it for about a week now, but havent noticed anything that major
taktoa has quit [Remote host closed the connection]
ckauhaus has quit [Ping timeout: 264 seconds]
<tnks>
I guess I have to play around with it locally anyway.
<day|flip>
is there a better option to search for local install package other then this: nix-env -qa | grep packagename
<tnks>
day|flip: coincidentally, that's one thing that the new "nix" command from Nix 2.0 is supposed to offer.
<simpson>
tnks: Since this is such a featureful release, it could take a long time for enough folks to use the new features to get a good impression of stability by anecdotes alone. You might have to set up an unstable pipeline with 2.0 and watch it for a few weeks.
<day|flip>
ahh thank tnks
<tnks>
"nix search" kind of thing, like you would expect with other package managers... and cached results too.
<tnks>
day|flip: yeah, check out the release notes for Nix 2.0.
schoppenhauer has quit [Ping timeout: 260 seconds]
<clever>
day|flip: nox, nix-repl and tab completion
<tnks>
simpson: I'm happy to test it out... even at work. What do you mean by setting up an unstable pipeline with 2.0?
<tnks>
one gets "n = import <nixpkgs> {}" lodged into muscle memory.
schoppenhauer has joined #nixos
<clever>
tnks: nix-repl '<nixpkgs>'
<tnks>
clever: the new "nix" command should completely replace "nox", right?
<clever>
i type that almost daily
<tnks>
Nox seems really simplistic.
<clever>
tnks: yeah, i think nix search replaces nox
<tnks>
clever: that's nicer that the binding I do.
<clever>
tnks: you can also :l <nixpkgs> after launching the repl
<clever>
nixer101: it only takes a minute or 2 to register on github, and i dont have push access
Mateon3 has joined #nixos
Arcaelyx_ has joined #nixos
ckauhaus has quit [Ping timeout: 252 seconds]
Mateon1 has quit [Ping timeout: 264 seconds]
Mateon3 is now known as Mateon1
Arcaelyx has quit [Ping timeout: 276 seconds]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] svsdep opened pull request #35447: keepass: bring back and update plugins load patches (master...keepass-plugins-patch-back) https://git.io/vAVws
NixOS_GitHub has left #nixos [#nixos]
<nixer101>
Can someone with git push access improve the description of "boot.loader.grub.device" option? Here are details: https://pastebin.com/tSdMnszK Thanks!
<NixOS_GitHub>
[nixpkgs] rycee pushed 1 new commit to master: https://git.io/vAVKI
<NixOS_GitHub>
nixpkgs/master cf10ed5 Robert Helgesson: perl-Encode: 2.96 -> 2.97
NixOS_GitHub has left #nixos [#nixos]
Sonarpulse has quit [Ping timeout: 252 seconds]
<hyper_ch>
hmmm, what good is /var/tmp actually for?
<nixer101>
hyper_ch: its stored in RAM, for any semaphores or anything that is small and must be changed fast (if you are on SSD)
<nixer101>
Can someone with git push access improve the description of "boot.loader.grub.device" option? Here are details: https://pastebin.com/tSdMnszK Thanks!
<etu>
I'm packaging a software that needs libgda as well as libgda-postgres and libgda-mysql. I have found libgda, it's gnome3.libgda. But that one doesn't seem to provide -postgres or -mysql... Any ideas?
<hyper_ch>
nixer101: I see... well, I'm currently pondering to store stuff like /tmp in a zfs dataset thta I don't backup.... so I'm lookin ghwat I want/need to backup and what is "transient"
asuryawanshi has joined #nixos
<etu>
I tried to add "--with-postgres=yes" and "--with-mysql=yes" to the libgda expression, didn't help.
<nixer101>
hyper_ch: maybe its not stored in RAM and its /tmp/ that is. I dont remember exactly, check cat /etc/mtab|grep tmp
<nixer101>
hyper_ch: eitherway, there are two global tmps. One slow ondisk, other fast in RAM.
<hyper_ch>
nixer101: well, as said, I'm pondering to move /tmp to a seperate dataset that I don't backups.... possibly also /var/log and pondering what other stuff I could add there
dan_b has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/vAVK9
<NixOS_GitHub>
nixpkgs/master aafd5a6 Jörg Thalheim: buildGoPackage: disable go 1.10 cache...
<sphalerite>
hyper_ch: dropping /tmp from bakcups would make a lot of sense, plenty of churn there and not usually anything you'd want to keep
<sphalerite>
Not so sure about /var/log, it's often useful data, and it tends to be appended to primarily (so COW works well) and often highly compressible too
simukis has joined #nixos
pxc has joined #nixos
<sphalerite>
If not, export your commit using git and submit it via the mailing list, or get someone else to submit it for you
<sphalerite_>
is there a way to trace all the instantiations that happen on a nix command to their sources, or otherwise to trace a .drv file to its source? I'm getting a build failure on expand-response-params and can't find the source expression for it.
<sphalerite_>
oh boy the matrix bridge is slow again
pxc has quit [Ping timeout: 264 seconds]
robstr has quit [Remote host closed the connection]
<hyper_ch>
sphalerite_: what other folders are there that could be dropped from BUs?
vidbina has quit [Ping timeout: 260 seconds]
Itkovian has joined #nixos
robstr has quit [Ping timeout: 240 seconds]
asuryawanshi has quit [Remote host closed the connection]
<sphalerite_>
hyper_ch: /nix/store :D
<hyper_ch>
isn't that like super important?
<sphalerite_>
nope. You can get everything in there from your configuration + nixpkgs
<sphalerite_>
And it tends to get quite large as well so there's lots of space savings to be made by dropping it
<hyper_ch>
sphalerite_: hmmm... hmmm.....
<sphalerite_>
it does make sense to keep it on a mirrored dataset and stuff of course since it's also quite convenient not to have it corrupted. But it's not valuable data that's hard to get back if it does get damaged
<sphalerite_>
nixpkgs and the configuration are your source code, teh stuff in /nix/store is your binaries. You need to back up your source code. Your binaries, not so much
<hyper_ch>
but easier to restore
<hyper_ch>
nah, I do look nix store as kind of important
<hyper_ch>
it's not that big
f0i has joined #nixos
<sphalerite_>
actually you shouldn't back /nix/store up separately from the rest of /nix though
dan_b has quit [Ping timeout: 248 seconds]
<sphalerite_>
since the database is critically related to the contents of /nix/store
ckauhaus has joined #nixos
f0i has quit [Ping timeout: 252 seconds]
ckauhaus has quit [Ping timeout: 256 seconds]
mrkgnao has quit [Ping timeout: 264 seconds]
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
hask_bee_3 has joined #nixos
<hyper_ch>
still continue to backup /nix and /nix/store
<hask_bee_3>
The Nix docs say: "Once a package is built, runtime dependencies are found by scanning binaries for the hash parts of Nix store paths (such as r8vvq9kq…). This sounds risky, but it works extremely well."
<hask_bee_3>
Can you elaborate on this? I'm not sure what they mean here. Who scans the binaries, which binaries are scanned, and why would we need to scan the binaries? I thought Nix was based on just looking into symlinks to those obscure paths.
<sphalerite_>
hask_bee_3: nix walks through all the files in the path (not just binaries, that's wrong) looking for store paths
<sphalerite_>
to establish what the runtime dependencies of a path are
akr has joined #nixos
<hyper_ch>
sphalerite_: btw, rss2email works really well :)
<sphalerite_>
good to know!
slyfox_ is now known as slyfox
JosW has joined #nixos
<akr>
Hello, I'm trying to install Nix separately for my user only without root privileges, following https://nixos.wiki/wiki/Nix_Installation_Guide Now unfortunatelly I have to use the last, PRoot option. Problem is, the command line to run proot seems to be broken in the tutorial.
<akr>
what I got to is `./proot_5.1.1_x86_64_rc2 -R ~/.nix -b /bin` but even then I get an error: "proot error: execve("/bin/sh"): No such file or directory"
<sphalerite_>
since /bin/sh is a dynamic executable and needs to be loaded by ld-linux which lives in /lib
<akr>
yeah, it exists
<clever>
nixos symlinks /bin/sh to ${bash}/bin/sh, and then ensures the closure of bash is in scope
<sphalerite_>
clever: but we're talking about RHEL here
<clever>
that might be simpler then getting the redhat /bin/sh to work
<clever>
if your already using proot to fake the root
<sphalerite_>
clever: but we don't have nix running yet
<clever>
ah
<clever>
ive been messing with nix-bundle lately, and its namespacing stuff is pretty handy
<sphalerite_>
I'm guessing namespacing isn't an option in this case since akr isn't using nix-user-chroot
<clever>
maybe tomorrow i could try throwing something together where it will persist the store long-term, to fix that "i dont have root" issue forever, lol
<clever>
you still need a tar of the nix store to start using nix-user-chroot, i think
<hask_bee_3>
sphalerite_ why does it need to hardcode those store paths into binaries (and other files)? I thought it was all based on symlinks.
<clever>
hask_bee_3: its a mix of symlinks and hard-coding paths into binaries
<akr>
sphalerite_: hmm, that didn't help. It seems that `/bin/sh` is a link to `/bin/bash` which is an actual executable
<sphalerite_>
akr: and with -b /bin -b /lib it doesn't work?
<akr>
sphalerite_: yeah I tried that first, didn't work
<sphalerite_>
huh.
<clever>
akr: what if you run "ldd /bin/bash", what does it report?
<akr>
nope
<sphalerite_>
what if you add /bin/bash at the end of the proot command?
<akr>
oh
<akr>
okay, it's supposed to be /lib64
<sphalerite_>
ooooh >.>
<akr>
yeah now it seems to work
mrkgnao has joined #nixos
<akr>
although for some reason I don't have any commands available in the opened shell: sh, ls etc.
<etu>
Bah, I manage to build libgda with postgres but not with mysql. It says that the headers are missing :/
Itkovian has joined #nixos
<sphalerite_>
akr: export PATH=/bin in it
<akr>
oh yeah, thanks
<akr>
looks like I also need to make a bunch of other stuff available, like SSL certificates for curl
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
ckauhaus has joined #nixos
<akr>
ok, looks like it managed to install Nix. Thank you for the help :)
ckauhaus has quit [Ping timeout: 240 seconds]
jensens has joined #nixos
Wharncliffe has quit [Quit: Lost terminal]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dywedir opened pull request #35455: exiftool: 10.60 -> 10.80 (master...exiftool) https://git.io/vAVPD
<hask_bee_3>
sphalerite_ well, when you build, you're using a certain nixpkgs revision. so your dependencies are already known from the packages collection. that's why i thought just the dependency name only would be enough.
<hask_bee_3>
hmmm
<sphalerite_>
hask_bee_3: I guess the essence of the answer is — if dependencies can be swapped out just by changing a symlink, you have a lot less reproducibility, which is one of the properties nix aims to provide. Additionally, you don't usually have a single place where your stuff goes
<sphalerite_>
on non-nixos, you have user profiles which aren't at some fixed path that you can just point programs to — they're in $HOME/.nix-profile where $HOME may vary so you'd have to patch every single program you want to use with it to look there
<sphalerite_>
on nixos you have the system profile in addition to that, which is active in up to two versions (/run/current-system and /run/booted-system) at any given time
<sphalerite_>
both of the nixos things you don't have on non-nixos, so you'd have to take that into account…
<sphalerite_>
since there are multiple profiles, which one do you look at first?
vidbina has joined #nixos
<sphalerite_>
It gives you all these variables that make it harder to reproduce a configuration and any bugs that might occur in it
<sphalerite_>
That's why it makes sense to bake the exact version of a dependency in: it gives you a lot less variability and makes stuff easier to reproduce
<hask_bee_3>
this makes a lot of sense. it sounds like we're decoupling the final program from the nixpkgs mechanism. decoupling is good.
<sphalerite_>
we actually have a mechanism in nixpkgs to replace a dependency in an already-built derivation: pkgs/build-support/replace-dependency.nix
akr is now known as akr`
<hask_bee_3>
When I run nix-shell, I often see Nix downloading all the dependencies first. So these dependencies were actually found by searching for the Nix store paths in the binaries?
<sphalerite_>
So you can apply "hot patches" and still have reproducibility. But it's easier and less painful to manage to just rebuild stuff
<sphalerite_>
yep
jooni has joined #nixos
<hask_bee_3>
nice
<sphalerite_>
well, they were declared at build time, but only a subset of the build-time dependencies will typically be necessary at runtime
<hask_bee_3>
isn't it just the dynamic libraries? dylibs. any other possibilities?
<sphalerite_>
we could of course pull in all build-time dependencies, but then you'd need gcc to get coreutils which wouldn't be fun
<hask_bee_3>
(or .so)
<sphalerite_>
dynamic libraries are the most prominent case, but there's also stuff like runtime data.
<sphalerite_>
Runtime data is more often actually used indirectly. Fonts for instance
<hask_bee_3>
ah ok
ckauhaus has joined #nixos
<sphalerite_>
in those cases it's often a matter of "this application depends on this specific data" (e.g. game data), in which case it would be baked in, vs "this data is a user preference" (like fonts) in which case it would be searched for in profiles.
<sphalerite_>
In the latter case, this is ideally done through upstream support for the XDG basedir spec, and the data dirs are just discovered through the $XDG_DATA_DIRS environment variable.
<genesis>
svn: E175002: Unexpected HTTP status 429 'Too Many Requests' on '/p/sdcc/code/!svn/rvr/9559/trunk/sdcc/device/non-free/include/pic16/pic18lf23k22.h' :')
<hask_bee_3>
sphalerite_ things are starting to make sense. although it seems like it's super important to make sure no global libraries or binaries (/lib, /usr/lib) etc. are available to the compiler at build time.
asuryawanshi has quit [Ping timeout: 248 seconds]
<hask_bee_3>
(i understand this is harder to pull off on macOS)
<hask_bee_3>
(since it had worse --pure support)
goibhniu has joined #nixos
<sphalerite_>
hask_bee_3: that's what the build sandbox is for. I think it actually works quite well on macos now, not sure though
<srhb>
Okay, so, the channel you want is from the channels remote, right?
<srhb>
nixos-unstable-small ?
<hyper_ch>
I first have to setup the vm for testing
<srhb>
... ok :P
JosW has quit [Quit: Konversation terminated!]
anderslu1dstedt has quit [Quit: leaving]
simukis has quit [Quit: simukis]
<hyper_ch>
how to match a channel root with a commit?
<srhb>
anyway, git remote add channels git@github.com:NixOS/nixpkgs-channels.git && git fetch channels && git checkout nixos-unstable-small then do whatever you like pointing nix at that checkout with -Inixpkgs=...
goibhniu1 has joined #nixos
<sphalerite_>
hyper_ch: the path typically includes the git hash. nix-instantiate --find-file nixpkgs should give you the path
<NixOS_GitHub>
[nixpkgs] aristidb pushed 1 new commit to master: https://git.io/vAVM4
anderslundstedt has joined #nixos
<sphalerite_>
not really related to anything, but I really like `units`
kelleyNif has joined #nixos
<srhb>
sphalerite_: It's great :)
<ottidmes>
Would https://nixos.org/nixos/options.html#nix.distributedbuilds allow my laptop or server to build things on my more capable desktop? Would it always build via my desktop the packages it would normally try to download from a cache if there was a cache miss?
<sphalerite_>
ottidmes: yes and no respectively
<sphalerite_>
err
<nixer101>
Can someone with git push access improve the description of "boot.loader.grub.device" option? Here are details: https://pastebin.com/tSdMnszK Thanks!
<sphalerite_>
at least if I interpreted your second question correctly
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dywedir opened pull request #35459: sxiv: 1.3.2 -> 24 (master...sxiv) https://git.io/vAVMa
NixOS_GitHub has left #nixos [#nixos]
<sphalerite_>
nixer101: did you see my reply earlier?
<nixer101>
sphalerite_: no, sorry, let me check
<sphalerite_>
wait ugh the matrix bridge dropped it
<ottidmes>
sphalerite_: Like I would not like it to start building samba or something else that is somewhat big, on 1.3 Ghz laptop...
<sphalerite_>
ottidmes: If it has preferLocalBuild set, it will build locally (this is typically for trivial derivations like config files). Otherwise, if it's in a binary cache, it will fetch it from there. Failing that, it will try building it
<nixer101>
sphalerite_: erm, its partially cut. But I dont have any git registration, I am new user doing my first install. And this is pretty important and undocumented option. So I thought...
goibhniu has joined #nixos
<sphalerite_>
ottidmes: it will build stuff first on remote builders, and if there are more things that can be built concurrently it will fall back to building them locally if the remote machine is fully busy
<sphalerite_>
Unless you specify --max-jobs 0, in which case it will build everything remotely
hamishmack has quit [Quit: hamishmack]
<ottidmes>
sphalerite_: That last bit is what I would want on my laptop then, thank you!
thc202 has joined #nixos
<sphalerite_>
nixer101: do you know how to use git?
duncan_ has quit [Quit: WeeChat 2.0.1]
<ottidmes>
sphalerite_: Except, does that break preferLocalBuild?
<sphalerite_>
ottidmes: not sure how --max-jobs 0 will interact with preferLocalBuild
<sphalerite_>
yeah I think so
goibhniu1 has quit [Ping timeout: 240 seconds]
<sphalerite_>
not sure though
hamishmack has joined #nixos
<sphalerite_>
since it's *prefer* and not *require* maybe it will build it remotely as well? Haven't looked at the source though.
duncan^ has joined #nixos
<ottidmes>
sphalerite_: I will just try and see :)
<sphalerite_>
clever: this seems like something you might know
<nixer101>
sphalerite_: No, I also don't have no github registrations. In fact, I have even no access to my note database. I am in the middle of setting up nixos desktop...
robstr has joined #nixos
<nixer101>
sphalerite_: I thought that it would be much better for community, if I would share this functionality, instead of leaving the note just for myself.
ckauhaus has joined #nixos
<nixer101>
sphalerite_: "boot.loader.grub.device = /dev/Xda" is dangerously unpredictable. The wwn methods is reliable.
goibhniu1 has joined #nixos
<sphalerite_>
nixer101: yes of course :) I'll PR it for you. Could you give me a name and email address you'd like the contribution listed under?
jensens has quit [Ping timeout: 256 seconds]
goibhniu has quit [Ping timeout: 240 seconds]
<sphalerite_>
Actually maybe just change the example value instead
<obadz>
ok nevermind, upgrading to nix 2.0 seems much harder than I though
<sphalerite_>
obadz: really? It should just be nix.package = pkgs.nixUnstable; on nixos
<obadz>
sphalerite_: that seems to be a relatively old version?
<sphalerite_>
huh, so it is. I thought it was updated
<sphalerite_>
obadz: oh right. That's not all necessary for nixos to work with nix 2.0 though. The only thing that's broken is nixos-install
ckauhaus has quit [Ping timeout: 248 seconds]
goibhniu1 has quit [Ping timeout: 240 seconds]
<sphalerite_>
I'm surprised that nixUnstable in 17.09 is so old though
<ottidmes>
sphalerite_: Is it better on nixos-unstable? I will then just use my unstable checkout, nix.package = pkgs.unstable.nixUnstable; Or is is already called nix on unstable
<nixer101>
sphalerite_: I really don't need any PR. I am thinking about choosing username later and registered to start contributing to nix community in near future though. I am fully okay with you also taking the "medal". The email would be "inbox-dpbofnq aaattt mail ddoott ru", its an alias for d-evelopment.
mrkgnao has joined #nixos
<hyper_ch>
so, can I somehow list all commits made and when they were made?
<nixer101>
sphalerite_: I really don't need any PR. I am thinking about choosing username later and registerING to start contributing to nix community in near future though. I am fully okay with you also taking the "medal". The email would be "inbox-dpbofnq aaattt mail ddoott ru", its an alias for d-evelopment.
<sphalerite_>
ottidmes: yeah that should work. I don't think the nix name has been bumped yet
<sphalerite_>
hyper_ch: git log
goibhniu has quit [Ping timeout: 252 seconds]
bitnotri has joined #nixos
<sphalerite_>
nixer101: no point obfuscating it in here since it will be publicly accessible, is that alright with you?
<nixer101>
why is there sphalerite_ and sphalerite? -__-
<sphalerite_>
and my laptop
<nixer101>
sphalerite_: no idea, will switch the alias, if I get overrun with spam; but I will likely need to set whitelist on it.
<sphalerite>
My phone via matrix's IRC bridge
Izorkin has joined #nixos
<sphalerite_>
nixer101: fwiw my git@sphalerite.org email address only gets a little bit of spam :)
<nixer101>
sphalerite_: I am really ok if you get the medal for it. Well, I am using alias technique to combat it. Its a little bit, but still annoying.
<NixOS_GitHub>
[nixpkgs] lheckemann opened pull request #35461: grub module: suggest by-id paths for devices (master...grub-by-id) https://git.io/vAVDr
NixOS_GitHub has left #nixos [#nixos]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master 6183b1c Gabriel Ebner: libqalculate: 2.1.0 -> 2.2.1
<NixOS_GitHub>
[nixpkgs] gebner pushed 2 new commits to master: https://git.io/vAVDP
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nixpkgs/master f5b173f Gabriel Ebner: qalculate-gtk: 2.1.0 -> 2.2.1
<obadz>
copumpkin: ping
<nixer101>
sphalerite_: Thank you very much. Although its a bit more radical that I suggested (only reminding it in description), but neverless its correct. The only issue I can think off, is trying to work with a backup without changing this value.
<nixer101>
Can someone recommend me a browser, that is already in channels, that would be chromium equivalent of waterfox/palemoon?.. I am thinking of "brave", but its not present in channels.
<sphalerite_>
nixer101: there probably isn't one. Browsers are complicated beasts and a pain to package
<sphalerite_>
nixer101: what are you actually looking for? Why do you not want to use chromium?
<nixer101>
sphalerite_: mainly tracking. I found out how to get addons without google store registration requirement. Currently chrome/chromium are much faster than firefox and firefox has issues with addons thanks to "quantum" thing, so there is virtually no benefit in using it anymore.
<sphalerite_>
should a refactor even be on 17.09 at all?
<nixer101>
sphalerite_: although I am completely white guy, who does nothing illegal, but don't like being datamined.
<nixer101>
sphalerite_: white in sense clean in front of the law, white as piece of paper.
<sphalerite_>
I use firefox with privacy badger and ublock origin myself, but I don't really have any obscure addon requirements either
<hyper_ch>
sphalerite_: how can I list all commits between two commits? this doesn't seem to work git log a3c96fd4de3 b9806751074 > /tmp/log
<sphalerite_>
hyper_ch: put a .. between them instead of a space
<hyper_ch>
and which should be first? the newer or older one?
<ottidmes>
nixer101: Both chromium and firefox have plenty extensions to help you with this, they are better at different things, so I just use both
<sphalerite_>
hyper_ch: the older one, think of it as from..too
<sphalerite_>
s/o$//
<nixer101>
sphalerite_: well, firefox is 3x to 2x slower compared to chrome/chromium and with quantum thing, many addon devs refused to continue development. There is whole branch of comments about it, it has to do with how fast firefox demanded (practically commanded) them to rewrite everything.
<hyper_ch>
sphalerite_: you have a strong git-fu
<nixer101>
ottidmes: I would be most grateful, if you would offer me a track-free chrome or chromium.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/vAVyI
<MichaelRaskin>
The real problem with the extensions on Quantum is that some old APIs used by popular extensions have no replacements in 57 and even in 58.
<srhb>
lejonet: Would you be open to a minor rewrite of your ceph module before merging? I'm considering refactoring it to a more ${cluster}-aware style.
<srhb>
lejonet: Great work by the way :)
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 closed pull request #35447: keepass: bring back and update plugins load patches (master...keepass-plugins-patch-back) https://git.io/vAVws
NixOS_GitHub has left #nixos [#nixos]
iyzsong has joined #nixos
ckauhaus has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bignaux opened pull request #35463: zsdcc: init at Unstable-20180224 (master...zsdcc) https://git.io/vAVyn
NixOS_GitHub has left #nixos [#nixos]
kelleyNif has quit [Quit: Leaving.]
<ottidmes>
MichaelRaskin: But the alternative, supporting all those legacy APIs indefinitely seems insecure and probably limit what speedups they can manage, so I for one am happy about the choice.
<sphalerite_>
ottidmes: AFAIU it's about functionality that just *can't* be accessed now, not about insecure APIs
<sphalerite_>
as in there's no replacement at all for them
<obadz>
sphalerite_: you're probably right
<obadz>
brb
obadz has quit [Quit: WeeChat 1.9.1]
kelleyNif has joined #nixos
<nixer101>
MichaelRaskin: yes, this is correct. There is an extensive comment over it in discussion. They added new API before it is ripe enough and demanded everyone to get onto it.
<MichaelRaskin>
nixer101: that would be half the problem
<nixer101>
MichaelRaskin: in addition Firefox does collect data, it does ignore user opinions, it is a corporation, and it is slower than Chrome/Chromium. So I am really unsure why I should use it further.
<MichaelRaskin>
ottidmes: if there wqs _any_ replacement, even scehduled to be changed again in 60, there would be much less of a problem
<nixer101>
I was thinking about Brave, but its not in channels. If I find nothing, I think I will stick to palemoon.
<MichaelRaskin>
nixer101: its data collection is easier to enumerate and disable, and also Chrome is becoming a new IE
<MichaelRaskin>
Mozilla is like the only organisation on the web that has ever succeeded to annoy people into actually reading the HTML standard.
<ottidmes>
But for them to innovate, I can fully understand that they have to drop a lot of features. And I think they are doing pretty good, they do not make the mistake to try and do a complete rebuild
<nixer101>
I want to clear something up. My inquery about alternative was not designed to be a flamebait or to derail this channel!.. I am really asking for possible alternative.
<ottidmes>
nixer101: About your request for privacy addons, the common ones, which I use or have used as well are mostly already mentioned, ublock origin, privacy badger, or ghostery
<nixer101>
just that I dont end up not beaten with the stick...
roberth has joined #nixos
<nixer101>
ottidmes: Thanks! I mean a chrome-derivate, that has internal tracking disabled.
<timokau>
Is hydra down right now?
<ottidmes>
nixer101: Never used one, have heard about a few in the pasts, but like already mentioned, they are unlikely to be available on nixpkgs. And I doubt you are able to package on yourself if you are new to Nix.
obadz has joined #nixos
ixxie has joined #nixos
<MichaelRaskin>
I think we do have Vivaldi, I am not sure how much it changes tracking
<nixer101>
ottidmes: well, I dont have any choice but to package it, do I? I am just doing the nix-install and I have small pet project in bash, that collects all *nix-* commands into more unified form for most often used cases.
<MichaelRaskin>
(also, it is unfree)
cinimod has joined #nixos
<nixer101>
MichaelRaskin: on www, there were at least two chromium derivates plus Brave, last time I checked; but I couldnt find them in nixos channel.
pie__ has joined #nixos
<MichaelRaskin>
If they rebuild Blink, they should take an eternity to build, which disincentivizes their packaging…
pkill9 has joined #nixos
<timokau>
nixer101: At the risk to go further off-topic, why do you think a closed source nieche browser will support your needs better than firefox?
<ottidmes>
nixer101: But honestly, I think if you really care about privacy, you should not be looking at any chromium based browser to begin with, at that leaves firefox
<timokau>
Nevermind, I think I got that closed-source bit wrong
<MichaelRaskin>
Well, something weird Gecko-based, like Conkeror.
cinimod` has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/vAVSR
<cinimod`>
I want to have nix build a package with "./configure CFLAGS="-DDEBUG""
<sphalerite_>
obadz: does taking out the nixpkgs. from the nix-shell invocation help?
<obadz>
sphalerite_: no
robstr has quit [Remote host closed the connection]
robstr has joined #nixos
<ottidmes>
obadz: I know of a roundabout way to do so, there are probably better ways to do it, you could probably create a new package that just exports the bin folder of nixUnstable, and the load that with nix-shell
<sphalerite_>
obadz: right… so probably not :/ you could do something like export PATH=$(nix-build --no-out-link '<nixpkgs>' -A nixUnstable):$PATH which will give you something similar
<sphalerite_>
cinimod`: the configure script will probably also use the CFLAGS env var so you just put CFLAGS = "-DDEBUG" in the derivation? That would probably be simplest if it works
nh2 has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] rnhmjoj closed pull request #34570: nixos/emacs: set VISUAL along with EDITOR (master...emacs) https://git.io/vNx21
<sphalerite_>
just stick that wherever you'd usually put plain gsl
robstr has joined #nixos
<obadz>
is there no way to build an expression in nix2? ☹
<cinimod`>
sphalerite_: great but now I look at my shell.nix I see I use gsl because I have hmatrix-gsl as a dependency
<cinimod`>
I guess I can add it explicitly?
<obadz>
nix build 'let pkgs = import <nixpkgs> {}; nixos = import <nixpkgs/nixos> {}; in [ nixos.system pkgs.mainEnv ]' ⇒ tries to access a file named after what's inside the quotes
orivej has quit [Ping timeout: 256 seconds]
ckauhaus has joined #nixos
<sphalerite_>
cinimod`: ah right you'd want to use an overlay for that kind of thing
<sphalerite_>
so the overlay would look like self: super: {gsl = super.gsl.overrideAttrs (o: {CFLAGS = "-DDEBUG";});}
<obadz>
nix edit is mindblowing!
palo has joined #nixos
palo has quit [Changing host]
palo has joined #nixos
<sphalerite_>
and you'd either put it in ~/.config/nixpkgs/overlays, if you don't mind it applying globally, or do import <nixpkgs> { overlays = [ (self…) ]; }
lambdamu has joined #nixos
<ixxie>
in my local fork of nixpkgs, to build my new package, I tried nix build -f /path/to/my/package/default.nix
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] yegortimoshenko opened pull request #35466: Build GTK icon cache for most themes (master...20180224.103545/icon-theme) https://git.io/vAV9Z
<ixxie>
sphalerite_: in the event of a very long package name am I free to invent a shortened name?
<ixxie>
sphalerite_: in this case the name is jupyterhub-ldapauthenticator and I wanna make it jupyterhub-ldapauth
<nixer101>
o/
nixer101 has quit [Quit: Konversation terminated!]
ebzzry_ has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] primeos pushed 1 new commit to master: https://git.io/vAVHt
<NixOS_GitHub>
nixpkgs/master 39ccb6a Michael Weiss: libmd: 0.0.0 -> 1.0.0
NixOS_GitHub has left #nixos [#nixos]
<sphalerite_>
ixxie: personally I wouldn't recommend it, the more names you have for something the more confusing it gets. And we do have tab completion for that sort of thing ;) but I don't make the rules, I don't know if there is a rule that swings either way on this, and stuff
<sphalerite_>
so that's just, like, my opinion, man.
<ixxie>
is it me or is the new nix build less verbose?
<obadz>
looks like for expressions one simply needs to put parenthesizes around the string!
<obadz>
nice
roberth has quit [Ping timeout: 240 seconds]
<jluttine>
where can i find the nixpkgs tree that my system is currently using? i'd like to see some service definitions there
<sphalerite_>
obadz: good to know! I didn't know that :)
<cinimod`>
But I am struggling to get the syntax correct
<cinimod`>
error at column 48
<cinimod`>
unexpectecd :
<pjan>
tnks: no, not yet afaik
<sphalerite_>
cinimod`: put parentheses around the hwole overlay
<sphalerite_>
it's trying to parse it as distinct elements of the list, rather than one big element
<lejonet>
Mic92: will try that then, got it to work with makeWrapper, but if wrapPythonProgramsIn is a cleaner solution, wrapPythonProgramIn it is :)
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] ixxie opened pull request #35467: pythonPackages.jupyterhub-oauthenticator: init at 1.1 (master...ldapauth) https://git.io/vAVH9
NixOS_GitHub has left #nixos [#nixos]
<Mic92>
lejonet: if you need to inject more environment variable, there is an option for that in wrapPythonProgramIn
ckauhaus has joined #nixos
<lejonet>
Mic92: Well its only the PYTHONPATH I've the need for now, because the other modules, that are actually packaged python libs, I just add to the python-env that it uses in buildInput
<hask_bee_3>
After upgrading to Nix 2, "stack build" on my Haskell project fails:
<hask_bee_3>
dyld: Library not loaded: /nix/store/6a9fr14xc9yr4xwlkyaf4zzqmbh3xngy-gmp-6.1.2/lib/libgmp.10.dylib
<obadz>
haha this is amazing I can nix copy a closure that's based on an expression to another machine!
<Mic92>
lejonet: wrapPythonProgramsIn for example also adds PATH of propagated dependencies if necessary, your makeWrapper will also include buildInputs as they are added to PYTHONPATH as well
<ambro718>
I was wondering, does nixpkgs itself have any such support?
necronian1 has joined #nixos
necronian has joined #nixos
<jluttine>
Mic92: this is the command that raises the error: echo 'select count(*) > 0 from information_schema.tables where table_schema = schema()' | /nix/store/fw1zigv2wiai5yxhzlzsnjra11qfq15z-mariadb-10.2.12/bin/mysql -u tt_rss tt_rss
<jluttine>
not sure why it raises permission denied
<Mic92>
jluttine: is this table readable for all users? This is some management table of mysql, right?
<necronian>
are you sure the tt_rss user exists? when I set up ttrss last week I needed to add a postgresql user first
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] obadz pushed 1 new commit to master: https://git.io/vAV5t
<NixOS_GitHub>
nixpkgs/release-17.09 b1f9a73 Frederik Rietdijk: citrix_receiver: 13.7.0 -> 13.8.0...
NixOS_GitHub has left #nixos [#nixos]
<jluttine>
because the command opens tt_rss database
<jluttine>
as far as i understand
<Mic92>
jluttine: seems to be metadata of the database
<necronian>
it's checking to see if the ttrss schema has already been loaded, if the query doesn't return anything then it is going to call the ttrss schema
<jluttine>
but why does it raise access denied error
<necronian>
jluttine: it did it for me because they database/user wasn't properly created, I just manually created the tt_rss user and database and granted the correct permissions and everything worked fine after that
<jluttine>
also, i can do "sudo -u tt_rss mysql tt_rss -u tt_rss" successfully
<jluttine>
so doesn't that mean the user and the database exist?
cinimod` has joined #nixos
<necronian>
jluttine: I don't know what to tell you, I was using postgresql and I kind of detest the nix expression language. I just know that I had the same exact error and that is how I fixed it
<jluttine>
necronian: ok, thanks
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] shlevy pushed 1 new commit to master: https://git.io/vAV59
<necronian>
Does anyone know how I can install a nonfree unstable package in my configuration.nix while I'm running stable? I can install non-free packages fine because I have allowUnfree = true and I can install unstable packages because I import <nixos-unstable> into a variable
<necronian>
but when I try to install a non free package from unstable it complains that I need to set allowUnfree
xeji has quit [Quit: WeeChat 2.0]
nh2 has joined #nixos
<ottidmes>
necronian: Have a look at pkgs/top-level/impure.nix there you can see how it determines the nixpkgs config
<ottidmes>
I have NIXPKGS_CONFIG set to my global nixpkgs config (for nixos-17.09 and nixos-unstable nixpkgs checkouts + nix binaries)
<necronian>
ottidmes: alright I'll try to see if I can work it out
kelleyNif has joined #nixos
<symphorien>
necronian: instead of import <nixpkgs> {} do import <nixpkgs> { config = ... }
<necronian>
I haven't touched NIXPKGS_CONFIG at all, quite frankly I've been running nixos for maybe a year now but I'm still mystified as to how it all works
<lejonet>
Mic92: Hmm, it seems like wrapPythonProgramsIn atleast does the right wrapping for pythonPath, but putting the third-party libs in propagatedBuildInputs didn't make it take them into the python-env, however, that is a easy problem to solve, because its just to move them to the ceph-python-env variable, then they get added to the python and python-env that is later used to replace the /usr/bin/env python in the
<lejonet>
script
<Mic92>
lejonet: was ceph-python-env using for something else?
<Mic92>
I don't remember why it was there
<Mic92>
*used
<lejonet>
Mic92: its used by cmake to generate docs and as the active python for the python tools that is bundled with ceph, (like the main ceph binary and so)
<rgc>
hi,I have a vga SiS on my old asus laptop. What kind of video drivers should I install?
<necronian>
symphorien: Thanks unstable = import <nixos-unstable> { config = { allowUnfree = true; }; }; seemed to do the trick for me
<Mic92>
lejonet: I assume it was not possible to just but it into buildInputs instead?
<ottidmes>
necronian: What symphorien suggests is more flexible, as you can see in pkgs/top-level/impure.nix, the check for NIXPKGS_CONFIG is just the first fallback it tries if this option was not explicitly set
<Mic92>
lejonet: I mean all libs needed for build
<lejonet>
Mic92: that is where ceph-python-env is put :)
<lejonet>
Mic92: and that is exactly what I was proposing :)
<hyper_ch>
why does isl take like forever to compile?
<necronian>
ottidmes: Yes I see that now, and I see why adding allowUnfree to .config/nixpkgs/config.nix didn't work either
<lejonet>
Mic92: that is how I solved it the first time, the hard part was figuring out how to get the pythonpath for the ceph-supplied python stuff, which you guided me to wrapPythonProgramsIn, which works
<ottidmes>
rgc: I am not sure what kind of GPU that is, but you can always fallback on the "vesa" driver
<rgc>
thank u ottidmes
nh2 has quit [Quit: Leaving.]
<rgc>
I'll try
<necronian>
I kind of get tired of having to read through a lot of source code in a lanuage I don't really understand to get anything done though
<ottidmes>
rgc: Did you try nixos-generate-config, it will also try and guess your hardware config
simukis has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] shlevy pushed 1 new commit to master: https://git.io/vAVdS
<Mic92>
lejonet: what does the wrapper content looks like?
<nkaretnikov>
in the light of the 2.0 release, just want to say that it's awesome that nix exists. thanks to all contributors!
<obadz>
I now get: $ nix-collect-garbage
<obadz>
error: Nix database directory ‘/nix/var/nix/db’ is not writable: Permission denied
<Mic92>
and the nix code
<gchristensen>
nkaretnikov: thank you!
<contrapumpkin>
obadz: 2.0 just has a default to hit the daemon if the db dir isn't writeable
<contrapumpkin>
obadz: I don't think leaving that in would hurt anything, tbc
<lejonet>
Mic92: interesting enough, it seems like it didn't create a wrapper or so O.o the only reason why it must've worked for my user is that I forgot to unset python path from my testing earlier
<obadz>
looks like I'm still using /nix/store/ny62xjh4jysdnw0pfpk52d1crvrr92vd-nix-1.11.16/bin/nix-collect-garbage somehow
<obadz>
oh I see why. naughty…
<Mic92>
lejonet: you were in a nix-shell probably
humanoyd has joined #nixos
<contrapumpkin>
obadz: why was it?
<obadz>
I had nix1 installed in my personal nix-env
<lejonet>
Mic92: I'm thinking of adding this just before postFixup: pythonPath = "${out}/lib/python2.7/site-packages:''$PYTHONPATH", think that will work?
<obadz>
contrapumpkin: thanks :)
<Mic92>
lejonet: pythonPath should be there without this hack. There is something else wrong. I will take a look
<lejonet>
Mic92: My guess is that L143 screws it up tho
<lejonet>
I haven't tried compiling it without L143 to test tho, but that line was added by adevress
coot has joined #nixos
<Mic92>
lejonet: what parts of that env are required for building documentation?
<lejonet>
Mic92: whatcha mean? Setuptools should exist at configure time
<Mic92>
lejonet: oh, sorry this was gibberish. I was with my mind between two pull requests. Did cmake already check for setuptools at configuration time and fails?
<Mic92>
now it seems compiling
<lejonet>
Haven't tried to remove that line myself
<zimbatm>
maybe use `nix run` instead of nix-shell
<robstr>
Hey, I'm first time nix user - I want to create a single file containing development setup instructions (like which gcc, cmake, ....) shouldn't this be possible with a default.nix file and nix-shell ?
<zimbatm>
I think it's wrong of nix-shell to expose all the build phases as exported envs
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] primeos opened pull request #1903: doc: Fix a URL in the release notes for Nix 2.0 (master...release-notes-2.0-url-fix) https://git.io/vAVxL
NixOS_GitHub has left #nixos [#nixos]
<manveru>
well, but that's the purpose of nix-shell though
<zimbatm>
robstr: yes, let me fetch an example for you
<manveru>
to go manually through all the build phases
<NixOS_GitHub>
[nix] grahamc closed pull request #1903: doc: Fix a URL in the release notes for Nix 2.0 (master...release-notes-2.0-url-fix) https://git.io/vAVxL
NixOS_GitHub has left #nixos [#nixos]
<lejonet>
Mic92: I'm going to up the cores on my nixops machine to see if I can get it to compile faster than 30 min :P
pmeunier has joined #nixos
<zimbatm>
manveru: it mixes two purposes and it's not good at either of those
<robstr>
zimbatm: thank you, and how can i enter the env ? i tried the same with stdenv.mkDerivation ...
<lejonet>
Mic92: explicitly setting pythonPath var didn't help either, just got that build finished now
<manveru>
robstr: run `nix-shell`
<zimbatm>
1) provide a development environment. this should only pull the dev dependencies and populate the shell env
<zimbatm>
2) debug an existing build. this would be much better achieved by dropping into a sandbox
<Mic92>
lejonet: I am confident, that it works the way I specified it. It might just be not complete (not all libraries)
<Mic92>
zimbatm: I have build something for the latter one
mrkgnao has quit [Ping timeout: 252 seconds]
<zimbatm>
for (2) I would like a `nix build --interactive` that binds stdin/stdout into the sandbox somehow
<Mic92>
something like a breakpoint for nix builds
<zimbatm>
yeah
<zimbatm>
gimme! :)
<robstr>
manveru: thank you
<lejonet>
Mic92: well, the error I'm getting is that it cannot find the rados module, which is created by the ceph build itself (the one in $out/lib/python2.7/site-packages), all the other libraries I've been able to find by starting the cluster and seeing what they whine at (first prettytable, then webob and pecan, then Mako :P )
<zimbatm>
`nix run` is almost there, it needs a few tweaks
<lejonet>
Mic92: and it seems like the wrapPythonProgram doesn't run, because the resulting ceph exec isn't a wrapper, its the same script as before :/
<zimbatm>
the --command doesn't really work as expected
<Mic92>
zimbatm: you also want the network sandbox
<Mic92>
and mount namespace
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] jfrankenau opened pull request #35473: valentina: init at 0.6.0.0a (master...init-valentina) https://git.io/vAVxE
NixOS_GitHub has left #nixos [#nixos]
<zimbatm>
I haven't tried to use `nix run` in a script shebang yet
<zimbatm>
Mic92: for `nix run` or the breakpoint?
<manveru>
well, you'll still need some equivalent of shell.nix for more complex `nix run`s
<Mic92>
zimbatm: the breakpoint.
ixxie has quit [Ping timeout: 252 seconds]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nix/master d9d8a84 Dmitry Kalinkin: Fix a small typo in the release notes
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
[nix] grahamc pushed 2 new commits to master: https://git.io/vAVxo
<NixOS_GitHub>
nix/master 9432f3f Graham Christensen: Merge pull request #1901 from veprbl/patch-5...
<zimbatm>
`nix run` loads default.nix by default
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] grahamc pushed 2 new commits to 2.0-maintenance: https://git.io/vAVxi
<NixOS_GitHub>
nix/2.0-maintenance bc0899d Michael Weiss: doc: Fix a URL in the release notes for Nix 2.0...
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nix/2.0-maintenance a1cc741 Dmitry Kalinkin: Fix a small typo in the release notes...
gfvgfvgfv has joined #nixos
pie__ has quit [Ping timeout: 260 seconds]
<lejonet>
Mic92: hmm, I found the definition of wrapPythonProgramsIn now, I most likely will want to use wrapPythonPrograms only, going to try that
<Mic92>
lejonet: that would be maybe a bit simpler
<lejonet>
Mic92: because I can't find any .wrapped binaries in the result directory, at all
<Mic92>
lejonet: I love that server. I have finished 40%
<lejonet>
Mic92: :D
<manveru>
i think my biggest gripe with `nix build` is the `--no-link`, which basically gives no output of where your result is...
<lejonet>
Mic92: I upped the cores of my nixops machine from 8 to 14, lets see if I can get it to run faster than ~30min per compilation :P
<manveru>
if often used `$(nix-build --no-out-link -A foobar)/bin/foobar`, but i guess that's replaced now with `nix run`
<robstr>
with fetchFromGitHub and custom build script ? :-/
<manveru>
hmm
<manveru>
emsdk would need a package i think
NixOS_GitHub has joined #nixos
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nixpkgs/master 63fa1fc Shea Levy: kexectools: Only build on kexecable architectures.
<NixOS_GitHub>
[nixpkgs] shlevy pushed 1 new commit to master: https://git.io/vAVpw
<gchristensen>
shlevy: weren't you just saying something about disabling pushing directly to master? :)
<shlevy>
gchristensen: Is ofborg ready for that?
<shlevy>
Let's do it if so :D
<gchristensen>
well it probably is. I wish I had a budget of $100/mo to split the core components out from my personal server, though.
<MichaelRaskin>
I wouldn't say it is ready before there has been a succesful roundtrip of «@borg merge-if-x86_64-ok nmap» or something like that
Myrl-saki has quit [Ping timeout: 256 seconds]
nuncanada has quit [Ping timeout: 252 seconds]
<gchristensen>
hmm it'd only be gating on evaluation checks at the moment, exactly what the pull request status shows
Myrl-saki has joined #nixos
sud0x3 has quit [Quit: WeeChat 2.0.1]
<MichaelRaskin>
gchristensen: I would still have tools for the natural workflow first
<MichaelRaskin>
But OK, «@borg merge» issued before the evaluation completes and executed after (conditional on success) would also convince me.
<manveru>
robstr: seems like a PITA to package that...
ryantm has joined #nixos
simukis has quit [Quit: simukis]
<gchristensen>
that seems like an awfully high bar
Arcaelyx_ has joined #nixos
simukis has joined #nixos
<robstr>
manveru: what is PITA ? :-/ currently trying to use fetchFromGithub inside a stdenv.mkDerivation but it is strange
<MichaelRaskin>
gchristensen: blocking the direct pushing should require a reasonably high bar on tooling, no?
* ryantm
Is there an easy way to do something like "nix-prefetch-url -A blah.src" where it doesn't look for it in the local store? My current strategy is to make the hash wrong before running the command.
<ryantm>
Gah. sorry about the /me. I need an new IRC client.
<genesis>
lejonet : i'm trying to package sdlmame , it takes 4 hours each try ;)
<lejonet>
genesis: yay, because of bad buildsystem or?
<genesis>
no 35K machines supported
<lejonet>
haha :P
Arcaelyx has quit [Ping timeout: 252 seconds]
<genesis>
best emulator ever!
<gchristensen>
MichaelRaskin: oh you're saying a way to automatically merge if evaluation passes, not necessarily blocking on a build too.
<MichaelRaskin>
We-ell, if these pieces of code are reasonably independent you could try to create a store path per a group of hardware
<MichaelRaskin>
gchristensen: as the lowest bar, yes.
<MichaelRaskin>
I would like an option to wait-and-gate on build, but that can wait.
<gchristensen>
and doesn't need to gate? ;)
<MichaelRaskin>
I mean, before considering gating, I want the expected workflow used on simple version bumps.
Itkovian has joined #nixos
<MichaelRaskin>
Well…
iyzsong has quit [Ping timeout: 248 seconds]
<MichaelRaskin>
I want to ask Borg to merge and receive either a merge or a ping when the evaluation finishes.
<MichaelRaskin>
Because that is what I am supposed to use with my own PRs after direct pushes are blocked.
jensens has joined #nixos
<lejonet>
genesis: what machine you build that on? Kindof feels like the build should be highly parallel in nature :P
<genesis>
lejonet : old intel i5 laptop (x201)
<MichaelRaskin>
35K files sounds like more code than Chromium, though.
<lejonet>
genesis: that explains the 4h :P
<lejonet>
The VM I'm using for nixops has 14 cores @ 3.4GHz and 20GB of RAM, lets hope that can reduce the time :P
<manveru>
ryantm: there was supposed to be `nix build --hash`...
<lejonet>
Kindof sucks that computers weigh a bit, I have a unused FM2 based computer that is just collecting dust that probably would compile that faster for you
<genesis>
lejonet it's very rare to have to compile such big big software, but i want it in nixos to provide it to every linux user that would like to use nix
<lejonet>
genesis: ^^ the best reason ever :)
<genesis>
lejonet : and compiling on gentoo for nothing more that 1 people is worthless ;) 15 years i use gentoo, i can't deal with hours of useless compilation.
gfvgfvgfv has quit [Quit: leaving]
<lejonet>
genesis: haha, I know what you mean there, atleast you could've done binpkg and atleast share with your own machines :P That is what I do
<genesis>
that is very limited, i know some gentoo-based distro, russian one used binary port, but hey, we need a real infra and ... nixos has hydra and that works.
<lejonet>
or do like a gentoo user in netherlands, she would compile libreoffice back-to-back on 4 P4s to get heating in the winter, because the place she lived at didn't supply heat, but electricity was "free"
<lejonet>
genesis: you mean sabayon?
<lejonet>
Yeah, the lack of a real tinderbox except diegos really is a problem for gentoo
<lejonet>
I wonder how far Zorry got with his project
<genesis>
lejonet : calculate
<lejonet>
genesis: ah, didn't know that was russian
<genesis>
Easy update via Portage throughout the distribution life cycle."
<avn>
lejonet: btw, do you know -- gentoo accepted libglvnd or not?
FRidh2 has quit [Quit: Konversation terminated!]
<lejonet>
avn: libglvnd?
^il is now known as vil
<avn>
runtime GL driver dispatcher
<lejonet>
avn: I can't find it in the official tree, but there is a layman overlay that has it, so I guess that is a no
<genesis>
lejonet : does it work for u ?
<lejonet>
genesis: libglvnd? Haven't tried it
self has joined #nixos
<genesis>
i've some difficulties with my gl too, spends hours without a proper fix
<genesis>
oki
self is now known as Guest51991
<avn>
debian (and I believe ubuntu as well) adopted it
<lejonet>
I would rather run SCO Linux than have nVidia in my computers, honestly xD
<avn>
and arch adopted as well
acarrico has joined #nixos
<lejonet>
so I just have to deal with AMD drivers and such anyway
<avn>
lejonet: agreed, but I curious, how to run tensorflow and other cuda stuff on my amd card (with open drivers if possible)
<lejonet>
avn: I'm the wrong person to ask that about, I've not used those things myself
<lejonet>
But isn't there a opencl compatible binding for tensorflow?
<avn>
idk actually
<avn>
even not sure, if my amd cards capable to compute
<lejonet>
which ones you got?
<avn>
TURKS and MULLINS
<lejonet>
OpenCL on the free driver is borked, tho I'm waiting for ROCm to be merged in 4.16
<lejonet>
awh, seems like tensorflow is CUDA-only :/
<lejonet>
(for its GPU support that is)
robstr has quit [Remote host closed the connection]
<avn>
I heard about some patch, and about stuff like coriander (which is true hackish ofc)
robstr has joined #nixos
yann-kaelig has quit []
<lejonet>
I hope that more projects go over to OpenCL/vulkan once ROCm hits mainline
<lejonet>
A fully open-source GPGPU stack is possible then, which would be awesome
<avn>
yep
robstr has quit [Ping timeout: 240 seconds]
palo_ is now known as palo
<lejonet>
lol, adding 75% more cores just got the time down to 26min instead, meeeh :P
<simpson>
lejonet: How much speedup did you expect?
Guest18665 has quit [Ping timeout: 260 seconds]
<lejonet>
simpson: with a proper buildsystem (because this is a compilation thot has a lot of parts that in theory can be compiled simultaneously) I was hoping at atleast a 20% speedup
<ij>
How do I trim whitespace?
<simpson>
lejonet: Ah, so assuming parallelism and applying Amdahl's Law?
<lejonet>
Mic92: with wrapPythonPrograms, it properly created the wrapper, so now it seems to work
<lejonet>
simpson: yep
yann-kaelig has joined #nixos
<lejonet>
Sure, a bit naive, but still, a proper tuned buildsystem can achieve a near-linear speedup with more resources
<lejonet>
(to a degree ofc)
coot has joined #nixos
<robstr2>
is there a problem with nix-build and ssl cert ? I receive `<urlopen error [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed `
Guest51991 has quit [Ping timeout: 252 seconds]
kelleyNif has joined #nixos
<gchristensen>
there is a guy who maintains some open source software and wrote their own teeny-tiny tool to setup recurring donations via Stripe. anyone know what that is?
Myrl-sak1 has joined #nixos
robstr has joined #nixos
Myrl-saki has quit [Ping timeout: 248 seconds]
<lejonet>
Mic92: trying now without git to see, if you read install-deps.sh from ceph source tree, it pulls in git, but my guess is that, after rg'ing through the source, that git is only needed if you intend to run the test or use some of the scripts
jensens has quit [Ping timeout: 252 seconds]
<lejonet>
(as in the more maintaineresque scripts, not the scripts that are tools for handling a ceph cluster, like ceph, rbd or ceph-mgr)
<gchristensen>
"Companies not always comfortable donating to campaigns" is a big one there
<timokau>
I read somewhere that selling trivialities for a lot of money can be a way around that, because its easier to buy than to donate
mrkgnao has joined #nixos
<samueldr>
e.g. swag or more like services?
<timokau>
Like a "license to use" even though your project is open source anyways or stuff like that. IANAL of course. I've read about that multiple times in hacker news comments I think
<lejonet>
Yeah, usually the bookkeeping rules around donations is a mess, whereas "buying goods" is usually straightforward
<MichaelRaskin>
I think Vim has a subscription to become «registered Vim user/a company with N registered Vim workplaces». This comes with voting on features (but no specific obligations about this voting having any outcome except published results)
<lejonet>
Mic92: git can be thrown out of the buildInputs, I successfully compiled it without git in buildInputs and atleast the basic cluster functionality of ceph-mgr, ceph-mon, ceph-osd and rbd works without any hitch
<Mic92>
lejonet: me too. I only messed up a loop variable
woffs has joined #nixos
kelleyNif has quit [Quit: Leaving.]
<lejonet>
Mic92: so now that we've got the executables to work as they should on nixos, should I file that PR, after cleaning up the branch? Then I'll look into the test harness to see about making that test for the service module
<lejonet>
just doing wrapPythonPrograms worked beautifully
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] jfrankenau opened pull request #35476: udisks2: add gobjectIntrospection to nativeBuildInputs (master...udisks2-gobjectIntrospection) https://git.io/vAwvz
NixOS_GitHub has left #nixos [#nixos]
<Mic92>
lejonet: what code did you now use for building?
<robstr>
lejonet: thank you, but how can I tell nix to install python with ssl support?
<lejonet>
robstr: no clue, for me it seems to already be done, I can both "import ssl" and "import socket; socket.ssl" when I just plainly use python
<lejonet>
robstr: does that executable maybe have an embedded python or so, that doesn't have SSL?
asivitz_ has joined #nixos
<robstr>
Please take a look at: Changed in version 2.7.9:
<robstr>
This class now performs all the necessary certificate and hostname checks by default. To revert to the previous, unverified, behavior ssl._create_unverified_context() can be passed to the context parameter. You can use:
<robstr>
`
<robstr>
lejonet: I'm really new - I get this error when i run `nix-build emsdk.nix` which is provided in the gist link, maybe this is interesting
<robstr>
whops ....
yann-kaelig has quit []
tg has quit [Quit: Leaving]
dan_b has joined #nixos
<asivitz_>
I just installed nixos, and I'm a bit confused about the channels concept. I switched to nixos-unstable because I wanted newer packages, but now I'm wondering if that was a mistake (should most people stick with stable?). Also, how do I get back to stable? Remove unstable, or add stable, or both?
<lejonet>
robstr: keep in mind that you need to get the /usr/bin/env python rewritten to a /nix/store path, I'm trying to find the builtin you can use for that, its in the nixpkgs manual
<asivitz_>
symphorien: thanks
<robstr>
lejonet: why is this _working_ on my machine, i mean it compiles everythink until ssl cert problem
<symphorien>
!channel
<symphorien>
!channels
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<lejonet>
robstr: you running something other than nixos? If so, /usr/bin/env most likely does exist :P
<robstr>
yes I'm on osx. I thought nix-build /nix-shell are independent from my os
<lejonet>
LnL: thank you, was wracking my brain to remember that name
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] boronine opened pull request #35477: Fix Darwin shebang issue in buildPerlModule [WIP] (master...perl-darwin-fix) https://git.io/vAwJv
NixOS_GitHub has left #nixos [#nixos]
<symphorien>
you can use sudo nix-channel --list to check what you currently have
asdfjkljkl has joined #nixos
asdfjkljkl has quit [Client Quit]
<lejonet>
robstr: I get "urlopen error [Errno -2] Name or service not known" now that I've gotten it to patch the shebang
<MichaelRaskin>
Is it during the build? Because during the build network access is not supposed to work
<lejonet>
MichaelRaskin: which was going to be my next thing to check, then the answer is there :)
vidbina has joined #nixos
ryanartecona has joined #nixos
<robstr>
lejonet: i dont get it ... looks like we receive different output
<robstr>
lejonet: i will start reading nix documentation from the beginning, maybe i do thinks wrong
cmacrae has joined #nixos
<lejonet>
robstr: I'm running em on nixos, I dunno the subtle differences in nixos and nix outside of nixos :)
<robstr>
lejonet: i see, thank you anyway. Would you recomment nixos or is it more like day to day fiddling to get normal stuff working ? :D
<lejonet>
robstr: so far I've only used nixos for a nixops machine and a server, so I honestly wouldn't know :P
<lejonet>
But for those usage I would warmly recommend it :)
<robstr>
yea maybe i will give it a try too, i like the idea but i have the feeling documentation lacks behind, entry is a bit hard
<MichaelRaskin>
Well, if you have a finite amount of day to day stuff, the idea of NixOS that the risk of breaking things by unrelated changes is lower
<symphorien>
robstr: are you trying to reach internet from the build script of a derivation ?
<symphorien>
this is prevented by the sandbox
<cmacrae>
Hi guys - I want to use make-system-tarball.nix to generate a root tarball. Where exactly do I start? Are there any examples of its use?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] jtojnar pushed 1 new commit to master: https://git.io/vAwJS
<NixOS_GitHub>
nixpkgs/master 114a8f1 Miguel Madrid Mencía: gthumb: fixed mime types
NixOS_GitHub has left #nixos [#nixos]
<MichaelRaskin>
You should remember that one reason of a hard entry is the need to learn a new programming language. When you are comfortable with it, though, you get the nice stuff like no path dependency
<lejonet>
MichaelRaskin: that is my thought too, that you need to segment your work into different envs, if you're handling stuff that assumes FHS rootfs, it could potentially just improve workflow imo :P
<nixy>
clever: lol I didn't expect it to be that simple, thanks
<robstr>
symphorien: thank youj i will check soon, just dinner :P
gemishhood has joined #nixos
<gemishhood>
Hey
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] lejonet opened pull request #35479: ceph: Modify ceph package so that it compiles and runs fine on nixos (master...ceph-patch-python-env) https://git.io/vAwTG
<catern>
why did people even try this whole "dynamic loading of dependencies" thing
<clever>
catern: that reminds me of a fun problem i had with flash years ago
<catern>
we know that it's a bad idea and that you should just have a good package manager/build system which lets you select your dependencies at build time
<clever>
i was on gentoo, and had always used wget, so i made a point of purging curl from the system, lol
<catern>
and static linking is also what people did originally
<catern>
but why did people do dynamic linking in the middle?
<clever>
and firefox could be compiled without curl support (it removed error reporting)
<lejonet>
catern: because you can load code on the fly! :P
<catern>
were they just idiots?
<clever>
catern: then i wanted flash, and the install script was just a dumb curl -> ~/.something/
<clever>
after a lot of googling, i found many confused people, that discovered using the bash script as-is magically fixed the problem
<clever>
i refused to accept that answer, and dug deeper
goibhniu1 has joined #nixos
<clever>
turns out, the bash script wasnt magic, installing curl was the solution :P
<gemishhood>
Hey
<clever>
and they unknownly fixed the proble, while trying to fix that bash script
<clever>
and re-running it did nothing
<lejonet>
catern: glhf doing that when you want plugins, even tho it was meant sarcastically, I think the main reason initially was probably saving on memory footprint because dynamically loading made so that you could share the lib in memory
<lejonet>
gemishhood: o/
<catern>
lejonet: the question is, do you load *arbitrary* code on the fly, or do you load fixed code on the fly? i.e., are you loading plugins or just lazily loading things from disk so you don't have to load your entire program at once
<clever>
lejonet: except, chromium/electron dlopen()'s a crap-ton of things on startup, including udev and opengl
<gemishhood>
lejonet: If you have a moment, I need some quick help.
<catern>
lejonet: sure, Nix has dynamic linking and everything, that's all good and nice
<catern>
but it still hardcodes the location to dynamically link from
<lejonet>
gemishhood: its better to state your question and see if people can help :)
goibhniu has quit [Ping timeout: 260 seconds]
<catern>
so you get the performance and security benefits of dynamic linking, without actually dynamically selecting libraries at runtime
<gemishhood>
lejonet: Sure. I'm trying to install lutris, and made a quick .nix file:
<lejonet>
catern: I dunno, I'm partial to statically linking stuff, even tho I can understand people wanting to use dynamic linking for enabling plugins as a way to extend a program
<lejonet>
catern: but some people just go overboard with it, like chromium for example
<catern>
chromium doesn't actually have plugins though, it doesn't need to use dynamic linking
<lejonet>
Sure, you _CAN_ just make your core code tinneh tinneh and dlopen in everything you need, but the harness you need to properly dlopen stuff and such usually add more complexity than just statically linking together different parts at compile time intsead
<catern>
though i suppose it does have extensions, which is an entirely different question...
<catern>
I guess I think plugins are an anti-pattern?
<catern>
instead of plugins, you should let the user control the top-level of your application, so that the user can customize what different modules are built-in at build/link time
<lejonet>
catern: it highly depends on what you define as plugins tho, plugins in the context I'm using, means something that actually adds new functionality to a program, i.e. the way claws-mail uses plugins
<catern>
lejonet: yeah, that's what I mean by plugins too
<catern>
lejonet: but you could get away without having runtime plugins by instead "building" your own claws-mail that just statically links in your plugins
<MichaelRaskin>
I guess for Chromium it is also about their sandboxing games.
<catern>
(and if you wanted to load the plugins lazily, do the same Nix trick where you dynamically link but hardcode the path)
<lejonet>
catern: yep, and I am in favour of doing that way instead
<clever>
catern: that gives me an idea....
<clever>
catern: patchelf can manipulate the DT_NEEDED field of an elf file...
<lejonet>
xD
<catern>
MichaelRaskin: oh?
<clever>
if i forcibly inject a binary it will later dlopen()
<clever>
i believe dlopen is also smart enough to find an already loaded library and return the existing handle
<lejonet>
gemishhood: cool project
<MichaelRaskin>
Plugins add functionality… I remember using Vacuum IM for some time. As a GUI XMPP client it is actually nice. The funny part is that Vacuum IM binary knows how to do 1 thing — load plugins. I think it loads plugin manager plugin by default…
<catern>
clever: oh! nice trick!
<catern>
clever: we currently just have LD_LIBRARY_PATH for this right?
<clever>
catern: you can also use rpath, but you have to be careful
<catern>
clever: this would allow getting rid of some program wrappers?
<lejonet>
:P
<MichaelRaskin>
catern: Well, if you want different parts of the code to be in different sandboxes and only give them the required libraries to reduce fault propagation
<clever>
catern: the stdenv runs patchelf --shrink-rpath, which basicaly checks ldd, and removes rpath entries you "dont need"
<clever>
catern: but its not aware of dlopen based things, and breaks it
<MichaelRaskin>
catern: I think you might want LD_PRELOAD
<MichaelRaskin>
As usual, there is that pesky setuid topic
<clever>
catern: so you have to use LD_LIBRARY_PATH or careful rpath'ing to ensure its in the search scope
<catern>
clever: ah, so why do we need the DT_NEEDED trick then? can't we just do the careful rpath?
<clever>
catern: this is the script responsible for the shrink
<catern>
clever: so this raises a question: why do any (proprietary crappy) programs in Nix needed to use LD_LIBRARY_PATH at all? can't we just edit rpath everywhere?
<clever>
and since patchelf is in the stdenv, that always gets sourced when things start up in the stdenv
ryantm has quit [Quit: Connection closed for inactivity]
jensens has quit [Ping timeout: 240 seconds]
<clever>
catern: the people who packaged them didnt know about shrink-rpath breaking things
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master b7a6c54 Miguel Madrid Mencía: nautilus: fixed mime types
<NixOS_GitHub>
[nixpkgs] jtojnar pushed 1 new commit to master: https://git.io/vAwk8
NixOS_GitHub has left #nixos [#nixos]
<clever>
catern: and just fell back to LD_LIBRARY_PATH
<catern>
good, so there's no technical reason :) simple bad packaging can be (relatively) easily fixed :)
<clever>
yeah
pie__ is now known as pie_
<catern>
clever: so what exactly would be the benefit of the DT_NEEDED trick? wouldn't it just remove some of the nice laziness and thereby slow down startup?
<clever>
catern: but it might break the program, because the library init functions run much sooner then the code expects
<catern>
also, a slightly unrelated question (also to lejonet) does anyone know a good name for the alternative to plugins where the user controls the top-level of the application, and the user can customize what different modules are built-in at build/link time and how they are used?
<robstr>
symphorien: whooo, looks like this fixes it, atleast this error :D - "Writing .emscripten configuration file to user home directory /homeless-shelter/" it looks like a few steps later, the file doesnt exists under this path ...
<catern>
it's a model that xmonad and some other things as well use, as well as any extension handling stuff in Nix (kind of), but I don't know what to call it
<clever>
robstr: nix builds arent supposed to store any state in $HOME
badi has quit [Ping timeout: 268 seconds]
vidbina has quit [Ping timeout: 240 seconds]
<lejonet>
catern: modules perhaps?
<lejonet>
I dunno of any other, or better, alternative name of such a system
<robstr>
clever: ty, what is the workaround ? I just want this package to be accessible and the installer script is given :-/
badi has joined #nixos
<simpson>
catern: "Compile-time configuration", "components", "compile-time features", as well as plenty of not-so-nice terms. It's a really bizarre antipattern.
<lejonet>
ah yeah, components is probably not a bad name
badi has quit [Read error: Connection reset by peer]
nh2 has joined #nixos
<clever>
robstr: do you care about anything its writing to $HOME?
<catern>
simpson: why do you say it's a bizarre antipattern :)
<catern>
personally I think it's better in a lot of ways
<robstr>
clever: I don't know but it looks like, yes
<simpson>
catern: Certain packages, like st and xmonad, come to mind as really user-hostile with this requirement.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] megheaiulian opened pull request #35481: Upgrade gitkraken to 3.3.4 (master...feature/gitkraken-3.3.4) https://git.io/vAwIk
NixOS_GitHub has left #nixos [#nixos]
<catern>
doesn't seem user-hostile to me, it seems user-empowering
<clever>
robstr: mkdir fake-home ; export HOME=$(realpath fake-home), try adding this somewhere like preConfigure
<robstr>
thank you
<catern>
the user can control whatever goes on, the "application" is actually just a library
<simpson>
catern: Good concept, bad execution, IMO.
<catern>
I think that's the right way to interpret it: you don't build an application, you build a library, and you give that to the user
<catern>
simpson: that's fair enough :) I think it's possible to do a lot better, it's pretty nice in Nix for example
<catern>
lejonet: I mean, I don't usually think of "modules" or "components" as generally referring to this pattern
goibhniu1 has quit [Ping timeout: 240 seconds]
<catern>
simpson: also, it's not compile-time configuration that I really care about - some runtime configuration is fine - it's compile time composition and wiring things up
<MichaelRaskin>
catern: I think st is a case of philosophy chimera. If you go the Unix way of programs, you don't ask people to recompile everything; if you give libraries — you want an environment of long-running processes with dynamic partial code replacement.
goibhniu has joined #nixos
<lejonet>
catern: well, a system is made of components, and a application could be said to be a system :)
badi has joined #nixos
<simpson>
catern: I don't care, but if this isn't Nix-related, we should move it to #proglangdesign or some other argument clinic.
<lejonet>
In a very broad and generic sense atleast
<MichaelRaskin>
#nixos-blah ?
<catern>
it's definitely Nix-related :) it's a question of correct software deployment
<catern>
it's a pattern that occurs repeatedly in Nixpkgs - the declarative specification of plugins for vim, emacs, Firefox, etc
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] fgaz opened pull request #35482: init bulletml and rrootage (master...kenta-cho) https://git.io/vAwI4
NixOS_GitHub has left #nixos [#nixos]
<simpson>
Those plugins are discoverable during run-time, aren't they?
ckauhaus has joined #nixos
<catern>
simpson: sure, they're wrappers to provide an interface like this
<clever>
simpson: but they also have to be patched to find their deps in the store
<clever>
and then it becomes a pain in the ass to manage imperatively installed plugins
<clever>
so just let nix take over :P
<catern>
MichaelRaskin: that's an interesting point, but on the other hand you could call Unix as a whole, the environment which allows dynamic partial code replacement
kelleyNif has quit [Quit: Leaving.]
<simpson>
clever: What I mean is that catern's trying to mix examples like st and xmonad with examples like vim, emacs, or firefox.
<catern>
simpson: no, I'm mixing examples like st and xmonad with examples like Nix's packaging for vim, emacs, or firefox
<lejonet>
dynamic partial user replacement! (yes, I'm not trying to participate in the discussion, it just poped into my head xD )
<catern>
"Nix's packaging for X" and "X" are separate
<simpson>
Okay, then pay to Nix what belongs to Nix. Are you trying to improve some upstream package's packaging, or some part of nixpkgs, and what are you proposing?
Fare has joined #nixos
<gemishhood>
Hey, I've been trying to install Lutris(a game manager) and have had no luck. Relevant Details: https://pastebin.com/rhc7z1Qj
<Fare>
When is the last call to get something in the 18.03 release ?
<catern>
MichaelRaskin: I mean, one question is, in the Unix way of programs, how do programs find the other programs that they need and depend on? in practice they use PATH, but you could also argue that static linking against programs is a very Unixy thing to do.
goibhniu1 has joined #nixos
<Fare>
is 18.03 automatically cut from unstable on March 1st at midnight, or something?
<clever>
Fare: pretty sure its manual
goibhniu has quit [Ping timeout: 248 seconds]
ckauhaus has quit [Ping timeout: 260 seconds]
<gchristensen>
Fare: what do you want in?
<Fare>
gchristensen, my latest updates to Gambit and Gerbil :-)
<lejonet>
gemishhood: is lutris a python application/library?
<gemishhood>
lejonet: Yep
<Fare>
but I suppose what's in from earlier this month is pretty good
<MichaelRaskin>
catern: if the quantum of easy replacement it a program, ship a program. If you ship a library of functions, make replacing functions mid-way easy
<lejonet>
gemishhood: have a look at buildPythonPackage and section 9.11.1.2 in the nixpkgs manual :)
<Fare>
one thing though: the static compilation feature of gambit and gerbil isn't working, because I couldn't find how to get static libraries from some of the dependencies
<gemishhood>
lejonet: Thanks!
<lejonet>
gemishhood: Np :)
<MichaelRaskin>
Fare: which one?
<lejonet>
gemishhood: its most likely a lot easier to use that tooling to build a python package, as its hooked into how python work, pip and all :)
goibhniu has joined #nixos
kelleyNif has joined #nixos
<Fare>
I think openssl, sqlite, libxml2, yaml and lmdb were missing static libraries
<Fare>
I haven't tried postgresql, for which gerbil added support since.
<gemishhood>
lejonet: Yep, thanks again
<clever>
Fare: when adding them, try to add a static output, and put them in $static/lib/
goibhniu1 has quit [Ping timeout: 260 seconds]
<Fare>
clever: same $output that also includes all the rest, or a separate output with just the static library?
<lejonet>
gemishhood: np, good luck!
<clever>
Fare: seperate, so it can just be ignored by users not wanting it
<Fare>
is there a standard for what e.g. makeStaticLibraries libyaml should look like?
<catern>
MichaelRaskin: actually, wait a second, I don't believe that it's good to have dynamic partial code replacement; a piece of code should be a closure over all the other code it depends on; it should all be static. if you want to replace something dynamically, I think you need to also replace everything that depends on it. that maybe means converting data to new formats, which in a powerful language you could do
<Fare>
should it be called ${makeStaticLibraries libyaml} or libyaml-static ?
<clever>
Fare: i dont think makeStaticLibraries does what i said, i think it forces static-only
<clever>
Fare: what you really want, is a derivation that makes static, dynamic, headers, and bin, all at once
<Fare>
well, I'm willing to fix the libraries at some point, but I'd like to have a clear spec for what the convention is.
<clever>
Fare: just set outputs = [ "dev" "lib" "out" "static" ]; and then copy things to $dev, $lib, $out, and $static
<clever>
Fare: then you can refer to each output as an attribute on the resulting object
<Fare>
having a reference implementation would be nice
<clever>
Fare: let me see..
<simpson>
${libyaml.static} in this case, I think?
<MichaelRaskin>
For example, zlibStatic is a separate package
<clever>
MichaelRaskin: thats not proper multi-output
alfie has joined #nixos
<MichaelRaskin>
Ouch. There is zlib.static and zlibStatic. zlibStatic also has static output. zlib, too
<Fare>
also, since nix 2.0 is released... for nix 3.0 could we offer a standard library / builtin set with consistent naming and calling conventions, and deprecate the legacy names?
<simpson>
Fare: Sounds gratuitous. What'd you have in mind?
<Fare>
simpson, I don't have a proper replacement in mind, but the documentation for the primitives available to nix is somewhat ridiculous.
mkg has joined #nixos
<simpson>
Fare: So perhaps those docs need improvements.
<clever>
Fare: basically, any derivation can define an outputs attribute, which is a list of strings, like outputs = [ "dev" "out" ];
<MichaelRaskin>
Or creation, for some builtins…
<Fare>
not just the documentation, but the inconsistency between the builtins, primitives, library functions, etc.
<mkg>
I am looking for help with nixos-container networking
<clever>
Fare: nix will then define $dev and $out to point to paths, and the derivation must create and populate those paths
gemishhood has quit [Quit: leaving]
<simpson>
clever: So, why should zlibStatic still be a thing? Should we refactor to have zlib.static everywhere?
<clever>
Fare: you can then do derivation.dev and derivation.out to get paths to them
<clever>
Fare: "${foo}" is the first output (dev in this case), "${foo.dev}" and "${foo.out}" refer to a specific output
<clever>
Fare: and nix can garbage collect, or not even download, the outputs you dont need
<clever>
simpson: yeah, id change that to just be an alias for zlib.static
<MichaelRaskin>
simpson: not sure if there are some additional tricks related to stdenv bootstrap…
badi has quit [Read error: Connection reset by peer]
<Fare>
if there is a one well-done library that I can emulate, I can reproduce the pattern for all my dependencies. But it would be nice to have a showcase library that does the right thing.
<Fare>
For instance, zlib
chocopuff298 has joined #nixos
<simpson>
Fare: Why not assume that everybody does it wrong and you'll do it wrong too? It sounds like you're expecting perfection from a Unix ports tree.
<Fare>
not perfection *everywhere*
<Fare>
just perfection *somewhere*. Just like the GNU hello that other people can copy.
<clever>
Fare: hello lacks libraries and headers, so it cant really make use of this feature
<obadz>
so in the new progress bar, what's the blue number? number of derivations that haven't been found on cache so far?
<Fare>
I don't need a fully realized ideal, just a well-enough-defined ideal that I can aspire to and make progress toward.
mkg has quit [Client Quit]
chocopuff has quit [Ping timeout: 240 seconds]
<Fare>
for instance, for lisp libraries (not nix related), I recently published a "best practices" document telling people how to properly use ASDF (the CL build system). Some kind of style guide for Nix recipes would be nice.
<MichaelRaskin>
Well, you also did tell me what not to do when packaging ASDF libraries for Nix…
<clever>
nix-repl> systemd.outputs
<clever>
[ "out" "lib" "man" "dev" ]
<Fare>
one that would cover libraries, applications, services
<clever>
Fare: maybe the systemd package?
<clever>
Fare: oh, another one i wrote, one sec
<clever>
is missing the output i added
<simpson>
Fare: There's the entire nixpkgs manual, and we're here to discuss whatever's not in there.
badi has joined #nixos
<MichaelRaskin>
Actually, openssl is pretty good
<MichaelRaskin>
Just replace removing static libraries with moving them to a different output
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] zimbatm pushed 4 new commits to release-17.09: https://git.io/vAwLQ
<clever>
cmacrae: this generates a tarball with a /kexec_nixos symlink pointing to the storepath of the kexec_script, and also includes that script, and the rest of its closure
<cmacrae>
clever: Thank youuuu! Much appreciated!
<samueldr>
in nixpkgs, nixos/modules/profiles/docker-container.nix and nixos/modules/installer/cd-dvd/system-tarball.nix make use of it
<zimbatm>
obadz: classic. I'm surprised you didn't switch earlier :)
<cmacrae>
clever: This is exactly what I needed :)
<obadz>
zimbatm: I was on master for the longest time but now I like to be on stables & betas.. I guess it's close enough to beta at this point..
<zimbatm>
I always think I will use the stable branch, but revert to master after a few month when things get out of date
<Fare>
simpson, there are many things the manual does not cover. Pointing at particular packages as examples to follow would be great.
<genesis>
it does
<Fare>
(which means, first, making sure that said packages are and do stay in great shape)
raynold has joined #nixos
sigmundv_ has joined #nixos
f0i has quit [Read error: Connection reset by peer]
reinzelmann has quit [Quit: Leaving]
<simpson>
Fare: Yeah, I understand your desire, but I'm wary of ever expecting nixpkgs to have such a thing. In other communities, this has led to all kinds of disasters, from copy-and-paste cargo-culted PHP nightmares to Wikipedia's Featured Articles political arena.
<chisui>
hey, does anyone here have experience in running games through wine od nixos?
<Fare>
simpson, there is STILL copy-paste nightmare without good examples
<simpson>
chisui: Sure.
<Fare>
even WORSE nightmare, because instead of one good example, it's lots of bad examples
<simpson>
Fare: I don't know what to say; that's not my experience. I don't copy and paste Nix code between expressions anymore; I learned enough of the Nix language to be able to write my own phrases.
nuncanada has joined #nixos
<Fare>
Also, nix has abstraction capabilities that PHP does not, so I'm not too worried about copy-and-paste, as long as there is a good discipline of OAOOM
<chisui>
whenever I try to run something wine says that several drivers are missing: r600_dri.so, swrast_dri.so
<avn>
32bit dri not checked on
<simpson>
chisui: 32-bit game? Did you `hardware.opengl.driSupport32Bit = true;`
<Fare>
I'm just saying, if you want more contributors, you have to work to lower the barrier to entry --- something that the PHP community is good at (maybe the only thing)
<Fare>
hopefully, I can turn one of my dependencies into a good showcase.
<simpson>
If. I think we need more maintainers and code-reviewers.
<Fare>
... in my copious free time
gemishhood has joined #nixos
chisui_ has joined #nixos
<chisui_>
simpson: hm, I set the flag, upgraded, rebooten and set WINEARCH=win32. I still get the same errors
b has joined #nixos
<MichaelRaskin>
Fare: judging by the growth of the open-PR count, we need de-bottlenecking the processes more than just attracting new contributors, so I agree with simpson
<gchristensen>
indeed, making our process a bit better would be really good
<clever>
chisui_: did you nixos-rebuild switch?
<simpson>
chisui_: Hm, that's normally what does it. If you wanna share which game you're experimenting with, some folks may have game-specific advice. Also if you need to winetricks, don't forget that.
<chisui_>
Oblivion
chisui has quit [Ping timeout: 260 seconds]
Guest24326 has quit [Ping timeout: 248 seconds]
<MichaelRaskin>
Fare: and of course, trying to agree on what is the proper style leads to getting one more example that you can never ever get anything looking like a consensus in Nix* project.
self has joined #nixos
self is now known as Guest27394
<cmacrae>
clever: Once this tarball is built, where can it be found? :)
<clever>
cmacrae: as always, nix-build creates a result symlink pointing to the result
<cmacrae>
clever: Sorry, I'm a novice,
<cmacrae>
haha
<cmacrae>
Thanks!
<clever>
yep
obadz has quit [Quit: WeeChat 1.9.1]
<cmacrae>
Okay, got it working. Thanks again clever! Really appreciate it
<ottidmes>
Is anyone using the MIT version of Kerboros in Samba? I had to remove "--with-system-mitkrb5" in samba/4.x.nix to get my Active Directory Domain Controller setup working. In turn I had to remove some of the Samba services, because after the change the builtin Heimdal Kerboros is used and those services are loaded by Samba itself by demand
sigmundv_ has quit [Ping timeout: 256 seconds]
justanotheruser has quit [Ping timeout: 265 seconds]
<ottidmes>
I would like to make a pull request for this, because I think this is the better approach (using the builtins), but I do not want to break stuff from others, and am not experienced with Samba (this is my first setup)
<clever>
chisui_: looks like the package is broken
aarvar has quit [Quit: Leaving.]
<samueldr>
oooh, nix 2.0 on 17.09 will be possible :)
Thra11 has joined #nixos
Guest27394 has quit [Ping timeout: 240 seconds]
<chisui_>
Do I need the unfree AMD drivers to be able to run games?
<clever>
chisui_: it should work with the open source amd drivers as well
<ottidmes>
chisui_: What GPU card do you have?
<chisui_>
Radeon HD 6870
<ottidmes>
chisui_: Have you gotten it working already with one of the drivers? ("ati", "ati_unfree", "amdgpu")?
<chisui_>
after disabling the unfree drivers nixos-rebuild switch show the following warning: "cannot copy /nix/store/45x393csygnals1www37m03qxfx3b1n4-initrd/initrd to /boot/kernels/45x393csygnals1www37m03qxfx3b1n4-initrd-initrd.tmp"
<chisui_>
ottidmes: the manual only mentions "ati_unfree"
<clever>
chisui_: what does "df -h /" and "df -i /" report?
<ottidmes>
chisui_: I have only gotten "ati_unfree" to work under kernel 4.4
<chisui_>
Filesystem Size Used Avail Use% Mounted on /dev/disk/by-uuid/c1adc6b1-7a9a-4b9b-8938-eb4b659c8ff2 469G 45G 401G 10% /
Fare has quit [Ping timeout: 260 seconds]
<aristid>
ottidmes: isn't 4.4 a bit old?
Rusty1_ has joined #nixos
<aristid>
i think if i was using an AMD GPU, i would try to always use the latest kernel
<ottidmes>
aristid: Tell me about it! But it is the only one that works for the "ati_unfree" driver, since it is no longer updated apparently
<aristid>
ottidmes: sounds like circular reasoning :D
<ottidmes>
aristid: Wut?
<samueldr>
I have ~4059 packages I want to download to get *some* files from, 1) how can I get the total size it would download? 2) can I download them without the dependencies *via nix*
<chisui_>
is it safe to reboot if initrd can't be copied after a switch?
<aristid>
ottidmes: "i use ati_unfree because it works on kernel 4.4 which i use because ati_unfree only works on 4.4"
yann-kaelig has joined #nixos
<aristid>
ottidmes: at least that is how i understood your statements :)
<clever>
samueldr: just refuses to allow a storepath to exist without its dependencies
pie_ has quit [Read error: Connection reset by peer]
pie__ has joined #nixos
<clever>
samueldr: s/just/nix/
<samueldr>
aw, then it's going to be done outside nix :)
coot has quit [Ping timeout: 252 seconds]
<samueldr>
(I want to see if it's feasible to gather all manpages)
<clever>
samueldr: thats where split outputs come in handy
<samueldr>
yes!
robstr has quit [Remote host closed the connection]
<clever>
samueldr: if the man pages are in the .man output, and it depends on nothing, then you can cheaply download just all the .man's
<samueldr>
though, they aren't right now, right?
<clever>
samueldr: not all packages use split outputs fully
<sphalerite_>
only for some packages
<samueldr>
(nix-locate gets me mostly .out and .bin)
<samueldr>
yeah
robstr has joined #nixos
<samueldr>
I could comb the ~4059 derivations... but it would be quite some work!
<ottidmes>
aristid: I only said I use "ati_unfree" and that I am stuck with kernel 4.4 because of it. I never said I used "ati_unfree" because I am on kernel 4.4
<aristid>
ottidmes: ah, then i misunderstood :)
chisui has joined #nixos
<samueldr>
looks like only 168 are split as of the update to the nix-locate database
ryanartecona has quit [Quit: ryanartecona]
chisui_ has quit [Ping timeout: 260 seconds]
<chisui>
after rebooting rebuild switch still says: cannot copy /nix/store/45x393csygnals1www37m03qxfx3b1n4-initrd/initrd to /boot/kernels/45x393csygnals1www37m03qxfx3b1n4-initrd-initrd.tmp
<chisui>
how do I fix that?
<clever>
chisui: what about df -h /boot/
<chisui>
ahhh, ok, the boot partition if full
<chisui>
I thought 100M was enough
<clever>
i tend to make mine 500mb to 1gig
robstr has quit [Ping timeout: 252 seconds]
<ottidmes>
aristid: Apparently I just had to had the one card that never supported any of the open source drivers, I have tried any combination of any flag or other hints I could find to make it work under the open source drivers. And "ati_unfree" will build fine on my system under 4.9 and 4.14, but will just fail on boot with kernel API bugs
kitemikaze has quit [Ping timeout: 268 seconds]
<ottidmes>
chisui: You have to clean your /boot sometimes, nixos garbage collect wont collect everything on /boot
<samueldr>
and, contiuing my idea, can I *somehow* know the archive (nar?) URL from the nixpkgs tree? nix-repl a .out shows me «derivation /nix/store/...», which I'm wondering if from that I can either derive or get the URL of such file
<clever>
samueldr: you will see a URL field in there
<clever>
samueldr: download that file, uncompress, and pipe it into `nix-store --restore HASH1-foo`, and it will create a HASH1-foo in the current directory
<samueldr>
okay, so not the hash of the derivation, but the hash from *what's inside* the derivation
<clever>
yeah, the hash from its output path
<clever>
which is basically the result of hash("out" + serialized_drv)
<clever>
or "man" + serailize...
<clever>
so each output within a single drv, has a unique hash
<samueldr>
and using nix(-repl), that hash() function would be?
* samueldr
is looking at the nix manual
<clever>
samueldr: "${hello}" gives you the path
i-am-the-slime has joined #nixos
<clever>
no need to manually hash things
<gchristensen>
samueldr: so ... what're you doing? :)
<samueldr>
ah!
<clever>
gchristensen: i think its something along the lines of download all the docs!
<i-am-the-slime>
I would like to change one file in hipchat.
<gchristensen>
neat!
<i-am-the-slime>
I suppose what I should do is copy the existing package definition and make some adjustments to it.
<samueldr>
I though nix-repl would toString
<samueldr>
gchristensen: a thought I had not long ago, there are those man repos for *other distros* on the web
<samueldr>
well, man web interfaces
Thra11 has quit [Quit: IRC for Sailfish 0.9]
<samueldr>
would be neat to have those from nixpkgs
<clever>
samueldr: toString foo and "${foo}" behave very differently
<samueldr>
now that I have the pieces, should be trivial enough to cheat
<samueldr>
ah, then does nix-repl toStrings ?
<clever>
samueldr: if you try to treat certain things as a string (deriations, paths, and special attrsets), it will automatically turn into a string for you
<clever>
samueldr: but if you call toString on it, you can get a totally different string out of it
<i-am-the-slime>
So I need to change one line in one file. How should I do that?
<samueldr>
okay, so nix-repl is *helpful* for some types, if I understand correctly?
<clever>
yeah
<samueldr>
good to know, thank you very much!
asuryawanshi has joined #nixos
<chisui>
ottidmes: I tried setting the boot.loader.grub.configurationLimit=5, ran sudo nix-collect-garbage --delete-older-than 1d, sudo nixos-rebuild boot and sudo nixos-rebuild switch but /boot is still full
<chisui>
Do I have to manually clean /boot?
<clever>
chisui: yeah, you need to manually delete one or 2 initrd's
<samueldr>
infrastructure-wise, I'm assuming it isn't rude to fetch a bunch of narinfo
<clever>
chisui: nixos-rebuild doesnt delete the garbage until it has finished copying in the new files
<ottidmes>
ij: Not sure, you could always use the import from derivation trick if you need some more advanced features, I used it to convert a decimal netmask to a ip netmask
<catern>
I'm writing a C library that also provides a Python module... should they be in the same repository? assuming they should be in the same repository, is it possible to get a single nix-shell session for developing both?
Guest87904 has quit [Ping timeout: 248 seconds]
Myrl-saki has joined #nixos
Itkovian has joined #nixos
<catern>
hmm I guess that's really a cffi question
<clever>
catern: i think it would have to be under pythonPackages, so it can properly deal with the python2 vs python3 stuff
<clever>
and maybe split it into 2, the c library that doesnt care about python
<clever>
and then the python module, that links to the 1st, and the right python version
<avn>
and btw, you can write 2 expressions for single source (repo)
knupfer has joined #nixos
cmacrae has quit [Ping timeout: 252 seconds]
<chisui>
How do I check which optional packages where used building a derivation?
Myrl-saki has quit [Read error: Connection reset by peer]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vcunat pushed 20 new commits to staging: https://git.io/vAwZH
NixOS_GitHub has left #nixos [#nixos]
<NixOS_GitHub>
nixpkgs/staging 6ad7452 Hamish Mackenzie: fontconfig: Fix for HFS+ 1s date resolution issue...
<NixOS_GitHub>
nixpkgs/staging 965ddac Ambroz Bizjak: cups: Add patch for bug where CUPS fails to save files to disk...
<NixOS_GitHub>
nixpkgs/staging 2c59b40 Jan Tojnar: libxkbcommon: 0.7.2 → 0.8.0...
<samueldr>
ah, I thought you were going to say something along "haskell packages are big"
<samueldr>
thanks, but I (think I) would still need to download the nar.xz
<samueldr>
though, there's probably ways to seek with HTTP?
<clever>
samueldr: its compressed, so seeking isnt useful
<clever>
samueldr: but this would allow you to downloat the nar, uncompress it, stream it thru the parser, and just filter out the paths you dont care about
<clever>
so it never saves the junk to disk
hiratara has joined #nixos
asuryawanshi has joined #nixos
<sudoreboot[m]>
Hmm, curious; I get `mkdir: cannot create directory ‘openvr/headers’: Permission denied` when putting `mkdir -p openvr/headers` in `preConfigure`. I could have sworn it worked just earlier though..
<samueldr>
hmm, I'm dumb, I could run it directly on my "toy" server (which has good throughput) lol
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
asivitz_ has quit [Quit: Connection closed for inactivity]
robstr has quit [Ping timeout: 252 seconds]
chreekat has quit [Ping timeout: 240 seconds]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] jtojnar opened pull request #35492: udisks: point to udisks2 (master...udisks2) https://git.io/vAwch
NixOS_GitHub has left #nixos [#nixos]
gemishhood has joined #nixos
<gemishhood>
Hey
<gemishhood>
I'm trying to install lutris(a game manager) and have had some problem installing it
knupfer has quit [Ping timeout: 252 seconds]
<gemishhood>
It is a python based app, and I get this error
coot has quit [Remote host closed the connection]
<eacameron>
clever or anyone: I have a DigitalOcean ocean droplet that got rebooted due to the spectre stuff (hypervisor update) and the machine no longer could connect to the network. I think it's because I didn't install the DO networking module (I forgot). I'm able to access NixOS drive from a recover OS via a mount. Is it possible to install the DO module from here somehow?
ambro718 has quit [Quit: Konversation terminated!]
<eacameron>
This would be like running the installer via USB and installing a module on the target OS.
<eacameron>
clever: I don't know how to chroot into an install. I'm also worried because this is a nixops machine, but last time I tried to run nixos-rebuild on a nixops deployment it did not go well because the configuration is not in the normal place
mfiano has joined #nixos
<mfiano>
Hello
<clever>
eacameron: ah, nixops complicates it a lot more
<eacameron>
clever: drat
<clever>
eacameron: hmmm, first, lets get the chroot up, mount the rootfs and nix store to /mnt/ and then: cd /mnt ; chroot . bin/bash
<clever>
eacameron: does that give you a shell?
<mfiano>
I'm reading the manual and just starting to make a config. It states: "You must set the option boot.loader.grub.device", but that option isn't even in the default config. Do I write this myself, or has this option been deprecated, etc?
<eacameron>
clever: chroot . bin/sh works
<clever>
eacameron: ok, now read /nix/var/nix/profiles/system/etc/systemd/system/sshd.service to find the args for launching sshd
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] fgaz opened pull request #35494: gnujump: init at 1.0.8 (master...gnujump) https://git.io/vAwCV
NixOS_GitHub has left #nixos [#nixos]
<eacameron>
ahh didn't know chroot just changes the root!
<eacameron>
(heh...hence the name...)
<clever>
eacameron: then carefully kill the main sshd in the recovery (without killing the sshd your using), and start the nixos sshd inside the chroot
<eacameron>
clever: O_o ... ok thinking
<clever>
eacameron: and whatever you do, dont close that ssh, it will be difficult to re-open :P
ajgrf has joined #nixos
<bhipple[m]>
Does anyone know how many GB a full binary cache build of a given nixpkgs sha is on cache.nixos.org?
<eacameron>
clever: I could reboot the droplet and it will come back to recovery OS
<clever>
eacameron: of course, so you can always try again
rgc_ has left #nixos [#nixos]
<clever>
eacameron: once you have an sshd running inside the chroot, fix the deployment file, and `nixops deploy --build-only` to build the fix
<clever>
eacameron: then use nix-copy-closure to push it over manually
<lejonet>
mfiano: You write it yourself, the rebuild will whine at you if its not set to either a device name or efi
<clever>
eacameron: and then run /nix/store/HASH-name/bin/switch-to-configuration boot, inside the chroot, to update the grub config, so the build runs on bootup
<eacameron>
clever: does nixops really not push the actual config at all?
<clever>
eacameron: nope, the config never leaves the central machine
<clever>
eacameron: all it pushes out is finished builds
<eacameron>
clever: Why???
<eacameron>
(But this is for another time...)
<clever>
eacameron: it depends on things the machine wont have, like the IP's of other machines in the clusters
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
nixpkgs/master 3dfe50f Elis Hirwing: sequeler: init at 0.5.3
<NixOS_GitHub>
[nixpkgs] jtojnar pushed 2 new commits to master: https://git.io/vAwCM
<NixOS_GitHub>
nixpkgs/master 55d2daa Elis Hirwing: libgda: Add optional mysql and postgresql support
NixOS_GitHub has left #nixos [#nixos]
<clever>
and other things
<eacameron>
clever: ah
<mfiano>
lejonet: Thank you. Do you happen to have an extensive config made public I can browse to get some ideas while I go?
<eacameron>
clever: How do I kill the ssh of the recover without killing the one I'm on?
<clever>
eacameron: ps -eH x, youll see one sshd is the parent of your shells, that one is important
<clever>
eacameron: netstat -anp | grep sshd
<lejonet>
mfiano: extensive I dunno, but I could pastebin the one I use for my nixops machine, its kindof scaled down tho, because I'm not using X11 on that one
<clever>
eacameron: youll see one is listening on 22, kill that one
hiratara has quit [Remote host closed the connection]
<clever>
mfiano: that info is also in `man configuration.nix`
<eacameron>
clever: Out of curiosity, could nix-copy-closure just point to /mnt/nix instead?
<mfiano>
I saw that. I don't like that it's all clumped together, namespaced or not. I learn better by seeing others' configurations/code.
<eacameron>
And piggy back off the ssh I'm already on?
<clever>
eacameron: only with nix 2.0
<clever>
eacameron: and the flags arent documented as well
<eacameron>
clever: Ah...well nix 2.0 will be great once I'm using it... lol
<clever>
eacameron: you could also run the 2nd sshd on a different port, and convince nix-copy-closure to use that port
<lejonet>
mfiano: true, such a interface is usually best if you're fairly certain of the name of the option, but you're not sure what format of the option and such
<clever>
copying 19 missing paths (11.30 MiB) to ‘root@router’...
<clever>
thats a good sign
day|flip has left #nixos [#nixos]
<MichaelRaskin>
I mostly switch my laptop between non-NixOS Nix-based systems…
<MichaelRaskin>
Hopefully the current iteration can be just evolved from now on… At least I finally have correct /etc/ switching, unlike NixOS
day|flip has joined #nixos
<gchristensen>
MichaelRaskin: maybe one day NixOS can do etc switching properly
<MichaelRaskin>
I am afraid that the most likely case for it to happen is via the fallout scenario…
<bhipple[m]>
Has anyone tried setting up a debian multi-user nix daemon that doesn't run as root?
<ajgrf>
I packaged an executable from source, but it can't find the shared objects it needs at runtime. Do I reach straight for patchelf, or is there something else to look into first?
chisui has quit [Ping timeout: 260 seconds]
<clever>
gchristensen: pulling the switch....
<gchristensen>
* clever has disconnected
<bhipple[m]>
I see some setuid and `assert uid == 0` in the code, but I'm wondering whether it could be run as a `nix` user without too many extra privileges
<clever>
gchristensen: it restarted hydra, dbus, and polkit, the firewall service hasnt even changed, lo
<gchristensen>
nice!
<clever>
gchristensen: basically, i used nix-diff to compare a `deploy --build-only` drv against the old drv, and synced the config up
<MichaelRaskin>
bhipple[m]: I think the global store remounting as R/O and private remounts as R/W would need root
94KAAAH7Q has joined #nixos
<bhipple[m]>
Doesn't it only do the remounting on nixos?
21WAAAXLB has joined #nixos
<94KAAAH7Q>
[nixpkgs] abbradar pushed 1 new commit to master: https://git.io/vAwle
94KAAAH7Q has left #nixos [#nixos]
<94KAAAH7Q>
nixpkgs/master 6850418 Nikolay Amiantov: rpcs3: init at 2018-02-23
<MichaelRaskin>
clever: in the ideal world it would be enough for the new rules and the old rules to intersect enough to keep the connections in the allowed set for the same reason.
<clever>
MichaelRaskin: there is also the interface config, if the wan interface goes down, the iptable rules dont matter
<MichaelRaskin>
timokau: current NixOS has setup-etc.pl which has to walk /etc/ on switch and remove the old symlinks to /etc/static and add the new ones
<clever>
mfiano: a: its nixpkgs.config.allowUnfree, b: its listed under nixpkgs.config
<clever>
mfiano: and the nixpkgs.config is just an opaque attr set, so the documentation doesnt know whats within
<mfiano>
Hmm
<MichaelRaskin>
clever: I think I have in my life re-run DHCP on interfaces without dropping connections
<clever>
MichaelRaskin: as long as it gets the same ip, all is good
<mfiano>
So I have to know nixpkgs internals?
<timokau>
MichaelRaskin: Why isn't /etc/ a symlink?
<clever>
mfiano: linux can handle minutes of outage
<clever>
MichaelRaskin: oops, ^
<MichaelRaskin>
timokau: on my system, it is.
<clever>
MichaelRaskin: but windows, it cant even handle 20 seconds of outage
<clever>
timokau: some files in /etc need to be "private" and not world-readable
<MichaelRaskin>
On a wired connection, it is feasible to reset the link faster…
<clever>
timokau: others have to be writable like /etc/nixos/configuration.nix
<MichaelRaskin>
clever: which is actually not a good enough reason
<lejonet>
mfiano: nah, its more that its an option that is nixpkgs specific instead of being bunched together with the nixos option
<clever>
MichaelRaskin: with my old dsl modem, i had a bug in the pppoe driver, causing only the upload path to 100% die, for upwards of 20 seconds
<MichaelRaskin>
The most annoying thing is that I currently need to chroot passwd
<clever>
MichaelRaskin: windows would mass-disconnect everything, linux didnt even care
<timokau>
clever: So when the private store issues was resolved, /etc/ could be a symlink?
<MichaelRaskin>
I think I have in my life downloaded email over connections with 20s as non-surprising RTT
<MichaelRaskin>
timokau: in reality, it can and should be symlink now
<clever>
timokau: possibly, depends on if you want to make it fully read-only or not
alasi has quit [Quit: Connection closed for inactivity]
<mfiano>
lejonet: Okay, so where is the list of nixpkgs options?
<MichaelRaskin>
You can have writeable files there, they just need to be symlinks
<MichaelRaskin>
The most truly annoying thing is that you need a special wrapper over passwd
<clever>
MichaelRaskin: the oddity in my bug, is that the downstream was 100% working, with no latency, but the upstream had 100% packet loss, not high latency
<timokau>
MichaelRaskin: What about files like /etc/shadow?
<clever>
MichaelRaskin: for 20 seconds at a time
<MichaelRaskin>
Because it wants to use rename("/etc/passwd.new", "/etc/passwd")
<lejonet>
mfiano: some of em are documented in the nixpkg manual, the rest you'll sadly have to look around in the git repo to find :/
<clever>
ahhh, atomic writes
Myrl-sak1 is now known as Myrl-saki
<MichaelRaskin>
timokau: on my system, /etc/shadow -> /var/auth/etc/shadow
<MichaelRaskin>
I tried to use a unionfs for that, but eventually gave up.
<clever>
MichaelRaskin: i also recently found a major problem between nixos and nix-bundle
<MichaelRaskin>
Overlayfs doesn't do the right things, FUSE options can create an annoying situation on suspend…
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] shlevy pushed 1 new commit to master: https://git.io/vAwl6
<clever>
the conditions are a bit vague, but i think the trigger is if the mount namespace doesnt match its parent
<MichaelRaskin>
On the other hand, on _my_ system any user can ask root to setup a fresh UID and a fresh namespace (with mounts subject to a policy, and nosuid on everything)
<MichaelRaskin>
Yes, and I probably will write a separate write-up about the Nix part.
<gchristensen>
thanks
<MichaelRaskin>
Because the paper I am revising to the ELS deadline is more about the Lisp part.
<mfiano>
MichaelRaskin: I thought your name looked familiar. Fellow CL developer here
<MichaelRaskin>
Although, frankly, I am a bit surprised: I write some integration between all the parts, I write a completely natural demo of the integration, I think «neat feature», look for precedents — what do you mean, it is impossible to find any implementation of this natural desire that is also trivial to implement.
<MichaelRaskin>
Yes, my Nix interests and my Lisp intersts came into an actual intersection.
<MichaelRaskin>
If my paper gets accepted to ELS, I should talk to Fare there to add a parallel Gerbil daemon next to my Lisp daemon and a minimal Guile Daemon. I looked whether it is simpler to implement a second daemon in Guile and Gerbil, Guile won for me (sorry Fare, I just happen to already know POSIX and RnRS).
<clever>
gchristensen: and ive borked the router, lol
<gchristensen>
can't be that bad, you're online
<clever>
gchristensen: i removed an "un-needed" nfs mount, which broke an old binary cache http server from nix-push
<gchristensen>
ooops...
<clever>
gchristensen: which then broke the nix on every machine in the house
<gchristensen>
omg
<clever>
gchristensen: and it refused to build the fix
jb55 has quit [Ping timeout: 256 seconds]
<MichaelRaskin>
mfiano: there is package name (part of the path) and attribute name (basically variable name)
<clever>
[root@router:/nix/var/nix/profiles]# ./system-377-link/bin/switch-to-configuration test
<clever>
and now its been rolled back, lol
<MichaelRaskin>
They are similar, but sometimes differ just enough to be annoying
<MichaelRaskin>
I should check if Nix 2.0 supports sane uninstall by attribute name
<mfiano>
MichaelRaskin: Ah. I don't see dejavu_fonts defined in the package's files anywhere.
<mfiano>
How do I determine the font name to list in my config then?
<MichaelRaskin>
Look at nixpkgs/pkgs/top-level/all-packages.nix
<clever>
mfiano: the attribute is pkgs.dejavu_fonts
<MichaelRaskin>
I mean, there are multiple ways, of course. But grep over sources is just so fast and simple when they are already in the cache in the RAM…
<mfiano>
I don't have a filesystem yet
<MichaelRaskin>
You probably still have the source on the install media