<gchristensen>
unfortunately you can't interact with pre-push hooks
<fpletz>
that would drive me insane
<fpletz>
oh
<gchristensen>
I never push more than one at a time, otherwise it is a mistake
<ndowens08>
fpletz: reason I pushed that latest update is b/c someone commented that the hash was wrong and I noticed "chec" in that line some how, guess i was trying to type here :)
<fpletz>
gchristensen: thanks! I'm thinking about using pinentry or something similar
<gchristensen>
oh please send back your patches when you figure that out :)
markus1209 has joined #nixos
markus1219 has joined #nixos
kampfschlaefer has quit [(Ping timeout: 268 seconds)]
<ndowens08>
fpletz: I dont try to push bad things :p
<gchristensen>
same :)
<ndowens08>
well i cant really push but open PR atleast
viric_ has joined #nixos
filterfish has quit [(Remote host closed the connection)]
filterfish has joined #nixos
markus1189 has quit [(Ping timeout: 240 seconds)]
markus1199 has quit [(Ping timeout: 258 seconds)]
viric has quit [(Ping timeout: 240 seconds)]
<fpletz>
gchristensen: ehrm, there is no git push --no-check, right? %)
<gchristensen>
d'oh
<gchristensen>
no-verifdy
<gchristensen>
--no-verify
<fpletz>
oh
<fpletz>
then that would actually fit my use case
<gchristensen>
great :)
<fpletz>
as it's very easily overrideable
<fpletz>
thanks again :)
<gchristensen>
you're welcome
<gchristensen>
I made this after there were reports of mistaken branch & tag pushes
<fpletz>
I also like those other checks
viric_ has quit [(Ping timeout: 256 seconds)]
viric has joined #nixos
<gchristensen>
I have some other handy tools, like `try-pr` which applies a PR and then guesses the package to try and build based on the commit message.
digitus has quit [(Quit: digitus)]
<ndowens08>
fpletz: can you look at https://github.com/NixOS/nixpkgs/pull/23407/files dont have to review or push, Mic said I need to rebase it and put platforms.unix back in, in which i do see, but my problem is it seems I deleted my copy of the branch so I can't update the PR that I can figure out
ericsagnes has joined #nixos
filterfish has quit [(Remote host closed the connection)]
filterfish has joined #nixos
viric has quit [(Ping timeout: 260 seconds)]
Kendos-Kenlen has quit [(Quit: Konversation terminated!)]
griff_ has joined #nixos
<ndowens08>
But yea I cant redo it that I know of, and it shows platforms.unix in meta so idk
eacameron has quit [(Remote host closed the connection)]
<gchristensen>
something about nixops + nix + remote builders is doing very stupid things, in that it seems to only be building 1 or 2 packages at a time, where it could be building many
eacameron has joined #nixos
ilja_kuklic has quit [(Ping timeout: 260 seconds)]
stepcut has joined #nixos
eacameron has quit [(Ping timeout: 268 seconds)]
<spacekitteh>
bleh i can't figure out how to compile a custom version of git ;_;
<spacekitteh>
nix-shell seems to be ignoring everything i say and just uses the default git
griff_ has quit [(Quit: griff_)]
Rotaerk has joined #nixos
sellout- has quit [(Quit: Leaving.)]
mudri has quit [(Quit: WeeChat 1.7)]
k0001 has quit [(Ping timeout: 256 seconds)]
<clever>
spacekitteh: what have you tried?
<clever>
spacekitteh: and how do you want to use that custom git?
<spacekitteh>
clever: i'm working on a new feature for it
<clever>
can you gist the command you used and the nix expressions its referencing?
<spacekitteh>
i've tried like 100 variations
<spacekitteh>
i'll show you my current one i guess
<spacekitteh>
i've triiiiiiieeeeeed: nix-shell, nix-shell -A myGit, nix-shell -p myGit, and various other combinations i can't remember
<clever>
spacekitteh: "nix-shell" and "nix-shell -A myGit" will give you an env for building that git, so you have to run configurePhase and buildPhase
<clever>
spacekitteh: and -p wont work at all
eacameron has joined #nixos
<spacekitteh>
configurePhase and buildPhase are not found and not env vars
<clever>
can you pastebin the output of running nix-shell and the prompt that came up?
filterfish has quit [(Remote host closed the connection)]
<clever>
nix-shell -A myGit
<clever>
for this command
filterfish has joined #nixos
<clever>
spacekitteh: do you want to manualy build that new git, or let nix build it, and provide a git command in $PATH?
steshaw has quit [(Quit: Connection closed for inactivity)]
<sphalerite>
Profpatsch: use lib.copyPathToStore
<dfordivam>
Hello, I need some help ... I am trying to use haskell 'stack' tool to build a package. and while the compilation is running I hit an issue due to insufficient space in /run
derjohn_mob has quit [(Ping timeout: 246 seconds)]
<dfordivam>
Currently this is the size of /run "tmpfs 1.5G 4.0M 1.5G 1% /run"
thc202 has joined #nixos
<dfordivam>
And I think this is dependent on the size of ram the system has... is it possible to increase this?
eacameron has joined #nixos
paperd has quit [(Ping timeout: 260 seconds)]
MercurialAlchemi has joined #nixos
<dfordivam>
This is the error " /run/user/1000/ghc11637_0/ghc_47.s: hPutBuf: resource exhausted (No space left on device)"
<NixOS_GitHub>
[nixpkgs] peti deleted haskell/justStaticExecutables at f67a89e: https://git.io/vyumD
<Mateon1>
That's odd. Does anybody know, why I could be getting "bash: ./filename: No such file or directory" when running a file, even though it exists?
<viric>
When perror() shows "Interrupted system call"... that's EINTR or ERESTART?
<Mateon1>
The executable bits are set, file ./filename gives ELF 64-bit LSB executable x86-64
<viric>
man errno says "Interrupted function call" for the former and "Interrupted system call should be restarted" the latter
<Mateon1>
`ldd ./filename` gives No such file or directory error
<viric>
Mateon1: The ELF interpreter is not found
<viric>
Mateon1: readelf -a file | grep ld-linux
<Mateon1>
viric: Ah, how can I fix that?
<viric>
Mateon1: that's set at link time
<viric>
but niksnut wrote a utility to change that on an ELF: patchelf
<Mateon1>
viric: Yep, it needs ld-linux
<viric>
Mateon1: OR you can directly call your interpreter, with the file as parameter
<viric>
my-ld-linux.so file
<Mateon1>
Hm, what's the interpreter location in nixos?
<viric>
there is not one
<viric>
every binary has a reference to its interpreter
<viric>
apply the readelf -a to readelf, for example. :)
<viric>
you will see its.
<Mateon1>
Hm, so it's in the nix store location
<NixOS_GitHub>
[nixpkgs] peterhoeg pushed 1 new commit to master: https://git.io/vyuYi
<NixOS_GitHub>
nixpkgs/master 6f68eb7 Peter Hoeg: msmtp: use netcat-gnu to unbreak darwin
<sphalerite>
I want a nixos sticker for my laptop, is there an "official" source for merchandise?
ericsagnes has quit [(Read error: Connection reset by peer)]
<flyx>
hi folks, I get this when trying to update my nixos: „The option `services.xserver.displayManager.kdm' defined in `/etc/nixos/configuration.nix' does not exist.“
<flyx>
has this moved somewhere? it is still listed on the website
<sphalerite>
flyx: which version are you on?
<flyx>
unstable channel
<sphalerite>
Ah right, well the options on the website don't reflect the unstable channel
<flyx>
okay, could have guessed that. how do I find out where it moved?
derjohn_mob has quit [(Ping timeout: 240 seconds)]
johann__ has joined #nixos
markus1189 has joined #nixos
spinus has joined #nixos
markus1199 has joined #nixos
<NixOS_GitHub>
[nixpkgs] peti pushed 1 new commit to master: https://git.io/vyusc
<NixOS_GitHub>
nixpkgs/master 060f7cb Peter Simons: haskell-generic-builder: remove unused nodePackages argument
teknico has joined #nixos
chakerbenhamed has joined #nixos
jophish_ has joined #nixos
ericsagnes has joined #nixos
ThatDocsLady has joined #nixos
takle has joined #nixos
takle has quit [(Ping timeout: 260 seconds)]
yenzenz has joined #nixos
ilja_kuklic has joined #nixos
takle has joined #nixos
MoreTea has joined #nixos
thetet has joined #nixos
goibhniu has joined #nixos
takle has quit [(Ping timeout: 260 seconds)]
civodul has joined #nixos
tippenein has quit [(Ping timeout: 245 seconds)]
johann__ has quit [(Quit: Leaving.)]
johann__ has joined #nixos
tippenein has joined #nixos
Guest93826 has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
maninalift has joined #nixos
filterfish has quit [(Ping timeout: 258 seconds)]
eacameron has joined #nixos
seppellll has joined #nixos
RchrdB has joined #nixos
eacameron has quit [(Ping timeout: 260 seconds)]
freusque has joined #nixos
<sphalerite>
what does the "cannot refer to other paths" error thrown by builtins.toPath actually mean? I don't really get it, and reading the source hasn't helped. I might dig into it deeper, but maybe someone already knows the answer...
martinb1 has quit [(Ping timeout: 240 seconds)]
Wizek_ has quit [(Ping timeout: 246 seconds)]
takle has joined #nixos
takle_ has joined #nixos
takle has quit [(Remote host closed the connection)]
rmeis_ has joined #nixos
seppellll has quit [(Ping timeout: 256 seconds)]
rmeis has quit [(Ping timeout: 256 seconds)]
filterfish has joined #nixos
tomsmeets has quit [(Read error: No route to host)]
ericsagnes has quit [(Ping timeout: 240 seconds)]
seppellll has joined #nixos
rmeis_ has quit [(Ping timeout: 260 seconds)]
<NixOS_GitHub>
[nixpkgs] peterhoeg pushed 1 new commit to master: https://git.io/vyuCr
<NixOS_GitHub>
nixpkgs/master 2c99575 Peter Hoeg: neovim-qt: 0.2.4 -> 0.2.5
rmeis has joined #nixos
zraexy has quit [(Ping timeout: 260 seconds)]
Wizek_ has joined #nixos
rmeis has quit [(Ping timeout: 240 seconds)]
filterfish_ has joined #nixos
roconnor has quit [(Quit: Konversation terminated!)]
<goibhniu>
sorry ekleog, the wiki is locked down at the moment
<goibhniu>
(there has been some discussion about it recently)
<ekleog>
well, some people do have rights to edit it, would it be just to remove pages successfully moved to the manual -- then, it's not a huge issue I guess, given how old this page is
<MoreTea>
I was wondering, do we have a way of creating a "view" of the store that only contains the exact dependencies required by a derivation?
<MoreTea>
I'm thinking about this from the perspective running nix containers in kvm VM's via k8s.
<Dezgeg>
if you enable sandboxed builds that happens with bind mounts
c0ff33 has joined #nixos
<goibhniu>
ekleog: if it's something you could add to the manual, that would be even better
<c0ff33>
Hi guys
<MoreTea>
@Dezgeg, yep, that's why I asked if there is a way to have this view outside of the nix-build daemon
<ekleog>
goibhniu: it's not, I have no knowledge about cross-compilers on nixos (have been trying to get an arm gcc running on x64 since yesterday without any success...)
<c0ff33>
Trying to update my nixos, but due to this AWS issue `nixos-rebuild switch --upgrade` dies on checking cache server. Can I use some other means to bump version of my system/packages?
<goibhniu>
ah, fair enough (I'd also like to try that myself soon)
<ekleog>
well, then you now know where the first gmane link pointed to before gmane closing ;) (and it seems useless from a "how do I do that" pov)
<goibhniu>
c0ff33: does `nix-channel --update` work?
<arianvp2_>
So Nixos selling point you can run multiple versions of same software in parallel right?
<arianvp2_>
but how do I specify a version in my nixos config?
<arianvp2_>
I can only say " pkgs.vim" not "pkgs.vim.6.0"
ThatDocsLady_mtg is now known as ThatDocsLady
johann__ has joined #nixos
<ben>
The selling point is more like "whatever you somehow install, it won't interfere with anything else, b/c it's all isolated", but it doesn't automatically come with recipes for installing every version of everything ever
<ben>
I think you'd look at the definition of the vim package and make a copy that overrides the version and the source and the hashes and then see if that installs
<arianvp2_>
I see
<arianvp2_>
and how does it work with for example pkgs.haskellPackages ?
<ben>
I'm not really familiar with their magic :)
<arianvp2_>
what packages are included? the latest ones that build together? Is it like stackage or something? but except Hydra is the CI?
<ben>
....I think they have a tool to generate package specs from cabal files or cabal names and versions or something...
<RchrdB>
Sometimes nixpkgs has multiple very similar derivations for different versions of the same package
<bennofs>
arianvp2_: haskellPackages is based on stackage LTS
<bennofs>
arianvp2_: and for packages not in stackage, it uses the latest that is available on Hackage I believe (there may be some manual overrides for some packages, not sure, but all core packages are taken from Stackage LTS)
<RchrdB>
like nixpkgs.llvm_36, nixpkgs.llvm_38 and so on
<RchrdB>
other times the benefit of stuff going into unique directories shows up as the fact that you can have packages present from two different versions of nixpkgs itself at the same time.
<RchrdB>
e.g. you can run stable NixOS, clone the nixpkgs repo and work on it locally and have packages from nixpkgs HEAD alongside your stable distro.
<arianvp2_>
bennofs: are you sure? I cannot really find evidence anyware that they follow the Stackage LTS
mudri has joined #nixos
<ben>
arianvp2_: the commit log has a bunch of messages like "LTS Haskell $version", I assume that's them making nixpkgs catch up
<arianvp2_>
so does it only follow the latest snapshot. or is it possible to use the older snapshots as well?
dfordiva_ has joined #nixos
derjohn_mob has quit [(Ping timeout: 240 seconds)]
deepfire has joined #nixos
<bennofs>
arianvp2_: it only follows the latest stackage LTS snapshot. you can check out older versions of nixpkgs though or generate your own configuration-hackage2nix.yaml and then run hackage2nix to make a package set
<gchristensen>
the reason we use(d) the other list, though, is the collected and collated issues
nwspk has joined #nixos
contrapumpkin has quit [(Quit: Page closed)]
k0001_ has joined #nixos
dfordiva_ has quit [(Ping timeout: 260 seconds)]
k0001 has quit [(Ping timeout: 246 seconds)]
jsgrant has quit [(Quit: Peace Peeps. o/ If you need me asap, message me at msg[at]jsgrant.io & I'll try to get back to you within 24 hours.)]
johann__ has quit [(Quit: Leaving.)]
<NixOS_GitHub>
[nixpkgs] fooblahblah opened pull request #23636: Add curl dependency to Slack for calls (master...slack-libcurl) https://git.io/vyuy5
<NixOS_GitHub>
[nixpkgs] domenkozar closed pull request #23636: Add curl dependency to Slack for calls (master...slack-libcurl) https://git.io/vyuy5
ekroon_ has quit [(Quit: Connection closed for inactivity)]
<MoreTea>
domenkozar, zimbatm: Robert started working a tool called "np" (nix project). The idea is that it will wrap around all kinds of tools (like bundix, node2nix, stack2nix etc) to provide a uniform experience. See https://github.com/roberth/np
<musicmatze>
does someone know bash well? Because I have a very weird behaviour with a vim alias
<musicmatze>
I have `alias vim="vim -g --servername VIM --remote-tab-silent"` in my bashrc, but when I now do `vim foo.txt -g`, it opens "foo.txt" and "-g" as files... which is _really_ weird IMO.
<qknight>
avn
<musicmatze>
When I do `\vim -g -g`, it works as expected.
<musicmatze>
(not nixos specific I guess, but I ask here anyways :-/)
<avn>
qknight: ?
<musicmatze>
avn see above
<musicmatze>
my question on bash aliases
<musicmatze>
I guess... Don't know whether qknight pinged you because of my question ... :-/
<bennofs>
musicmatze: try vim -g foo.txt -g
<avn>
musicmatze: no, I think he pinged me by other reason ;)
<musicmatze>
bennofs: opens "-g" "foo.txt" and "-g"
<bennofs>
musicmatze: oops I meant \vim -g foo.txt -gf
<bennofs>
s/f//
rmeis has joined #nixos
<musicmatze>
bennofs: opens "foo.txt"
<musicmatze>
not more
<bennofs>
musicmatze: this is a vim weirdness. try \vim -g --servername VIM --remote-tab-silent -g foo.txt
sibi has quit [(Quit: Connection closed for inactivity)]
<manveru>
man, i really wish there was a node2nix version that didn't take 3 minutes :P
<qknight>
3 minutes for what?
<qknight>
like execution time?
<contrapumpkin>
redownloading everything
iyzsong has joined #nixos
<contrapumpkin>
how can I stop buildPythonApplication from making a wrapper for my executables?
musicmatze_ has joined #nixos
<manveru>
qknight: just updating deps and building the node_modules
freusque has quit [(Quit: WeeChat 1.7)]
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to master: https://git.io/vyuQD
<NixOS_GitHub>
nixpkgs/master c8402dd Tuomas Tynkkynen: libp11: Fetch src from GitHub
<qknight>
manveru: ok
<manveru>
atm we have nearly 1000 packages in node_modules, so things are slow as fuck
<qknight>
domenkozar: could you help me out. i want to override the src of bepasty-server but found no way (except modifying nixpkgs and not using override)
c0ff33 has joined #nixos
griff_ has joined #nixos
<manveru>
last i heard the biggest issue is that you can't just symlink stuff in node_modules, since node doesn't resolve them... otherwise one derivation per npm package would be possible
<qknight>
manveru: couldn't we patch npm to do so?
johann__ has quit [(Client Quit)]
<manveru>
that'd probably be worth a try
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to release-17.03: https://git.io/vyu7U
<NixOS_GitHub>
nixpkgs/release-17.03 c5a10ac Tuomas Tynkkynen: libp11: Fetch src from GitHub...
c0bw3b has joined #nixos
freusque has joined #nixos
exarkun has left #nixos []
<domenkozar>
qknight: sorry really busy toda
<domenkozar>
y
rmeis_ has joined #nixos
cpennington has joined #nixos
rmeis has quit [(Ping timeout: 246 seconds)]
seppellll has quit [(Ping timeout: 264 seconds)]
cpennington has quit [(Ping timeout: 246 seconds)]
<NixOS_GitHub>
[nix] shlevy opened pull request #1265: Add option to disable import-from-derivation completely, even if the drv is already realized (master...allow-import-from-derivation) https://git.io/vyu5D
Itkovian has joined #nixos
takle has quit [(Remote host closed the connection)]
cpennington has joined #nixos
mudri has quit [(Ping timeout: 264 seconds)]
rmeis_ has quit [(Ping timeout: 246 seconds)]
rmeis has joined #nixos
ThatDocsLady_afk is now known as ThatDocsLady
<NixOS_GitHub>
[hydra] edolstra pushed 1 new commit to master: https://git.io/vyuFJ
echo-area has quit [(Quit: Going to sleep, good night.)]
takle has joined #nixos
ludat has quit [(Ping timeout: 268 seconds)]
<HappyEnte>
Heyho, does anybody else has problems with krita on latest unstable channel? On my main system it runs flawlessly, but on my notebook it fails because it can't find qt platform plugin xcb.
tomsmeets has joined #nixos
mudri has joined #nixos
<NixOS_GitHub>
[nixpkgs] PierreR opened pull request #23640: pythonPackages.pepper: init at 0.5.0 (master...master) https://git.io/vyupB
<Profpatsch>
fpletz: Is it normal that git cherry-pick -x fails first because the commit is empty?
zraexy has joined #nixos
tomsmeets has quit [(Ping timeout: 258 seconds)]
griff_ has joined #nixos
<MoreTea>
Hmm. How must $out/init (in a systemconf derivation) be invoked?
<bennofs>
what's systemconf? :o
<NixOS_GitHub>
[nixpkgs] ixmatus opened pull request #23641: security-wrapper: Don't remove the old paths yet as that can create migration pain (master...parnell/fix-wrapper-migration) https://git.io/vyujT
rmeis has quit [(Ping timeout: 260 seconds)]
erasmas has joined #nixos
pepa has quit [(Read error: Connection reset by peer)]
MercurialAlchemi has quit [(Ping timeout: 264 seconds)]
pepa has joined #nixos
<MoreTea>
bennofs, the derivation that NixOS creates for one generation of your system
<MoreTea>
e.g. $ nix-build -I nixos-config=./config.nix '<nixpkgs/nixos>' --set -A system
arianvp2_ has joined #nixos
<arianvp2_>
anybody familiar with vim_customizable.customize?
<arianvp2_>
the vim_customizable package installs both vim and gvim
<arianvp2_>
however if you use vim_customizable.customize, you cannot access gvim anymore
<c0bw3b>
yes it's a base32 SHA256 hash that has been truncated
<c0bw3b>
cf man nix-hash
Itkovian has quit [(Ping timeout: 240 seconds)]
<c0bw3b>
well nix-hash --help rather
tomsmeets has joined #nixos
reinzelmann has quit [(Quit: Leaving)]
johnsonav has quit [(Ping timeout: 246 seconds)]
ThatDocsLady is now known as ThatDocsLady_mtg
ThatDocsLady_mtg has quit [(Quit: Arma-geddin-outta-here!)]
Itkovian_ has quit [(Ping timeout: 246 seconds)]
setgid has joined #nixos
<setgid>
Hi, I bought a Wacom Intuos Pro (new generation) pen tablet and it is only supported by the 4.11 kernel. Is there an easy way to pull in the 4.11 kernel in NixOS stable?
Nazral has joined #nixos
<Nazral>
Hi
<c0bw3b>
setgid : we are not even on the 4.10. kernel branch yet
<setgid>
I assume I will have to set boot.kernelPackages, but there is no 4.11 kernel in nixpkgs.
spinus has joined #nixos
<c0bw3b>
4.11 is not a stable branch yet
ilyaigpetrov has joined #nixos
Mercuria1Alchemi has joined #nixos
<setgid>
I'm aware and willing to experience some breakage. :) The power of nix is how well you can mix and match versions of different packages.
<nixy>
Is there a way to specify which channel you want to pull pkgs from in a packageOverrides?
<c0bw3b>
setgid : TBH it should be possible but I never tried to use a custom kernel myself
<setgid>
nixy: yes! just add the nixos-unstable channel and then put "let unstable = import <nixos-unstable> {}; in" around your configuration.nix expression
<setgid>
nixy: then you can use unstable packages by prefixing them with unstable.
<FRidh>
copumpkin: maybe we should have a `set NIX_PYTHONPATH`. We could add a sitecustomize.py in the folder of the interpreter that checks for this env var. We could use that one for all the wrapping, and users could then use `PYTHONPATH`.
<NixOS_GitHub>
[nixpkgs] rickynils pushed 1 new commit to release-17.03: https://git.io/vyzsz
<xeviox>
hi, maybe someone can help. I try to use a different version of a package in my configuration.nix. Because of that I added http://pastebin.com/4a1Rhkc5 to my configuration. But it still installs the old version :/.
<clever>
xeviox: you need to override name and src, version has no effect on the build
<shlevy>
niksnut: Is there a better way to detect if I'm compiling against nix 1.12 than "#ifdef ENABLE_S3"? :)
<shlevy>
niksnut: I need to know whether to pass one arg or two to nix::EvalState
<xeviox>
clever: I thought it will do the trick as its defined as url = "mirror://pypi/a/awscli/${name}.tar.gz"; in nixpkgs
<clever>
xeviox: the problem is that it evals the ${version} and ${name} before the overrides apply
<shlevy>
niksnut: Also, would be nice if nix's headers didn't conflict with other autoconf packages...
<xeviox>
ah ok, thanks :D
<clever>
xeviox: so it will use the old value of version
<contrapumpkin>
FRidh: \o/ thanks
<bennofs>
shlevy: can't autoconf tell you the version of a dependency? (sorry not familar with autoconf at all)
<contrapumpkin>
FRidh: oh sorry, I missed your comment above
<shlevy>
bennofs: No idea! :)
johann__ has quit [(Quit: Leaving.)]
<shlevy>
bennofs: I have access to a c string "1.12.whatever" or "1.11.whatever" but that's not exactly easy to parse in CPP :D
<contrapumpkin>
FRidh: what benefit would that provide over what we're doing now with --prefix?
<bennofs>
shlevy: you could do the parsing in your configure.ac and use AC_SUBST
<contrapumpkin>
ouch :)
<shlevy>
I'd rather do #IFDEF ENABLE_S3 honestly :P
<shlevy>
But that's probably the Right Way
<bennofs>
shlevy: or typical autoconf style, write a feature test where you try to call it with one or two args and then generate a #define based on that
<clever>
shlevy: i usualy see programs that have #define NIX_MAJOR 1, and #define NIX_MINOR 11 #define NIX_PATCH 7
<clever>
shlevy: and then you can do simple int compares in the macro, to detect whatever you want
<shlevy>
clever: Yeah, I'd like to add that to nix if niksnut wants it
pepa` has quit [(Ping timeout: 240 seconds)]
freusque has quit [(Quit: WeeChat 1.7)]
<xeviox>
I'm this far: http://pastebin.com/qEyq4apH but awscli is complaining about the missing dependency to botocore 1.5.21 which I tried to build with the second override
<FRidh>
contrapumpkin: the problem with prefix is that if you have nested Python applications each needing a different PYTHONPATH they will be concatenated. This is especially problematic when one of the applications requires Python 2 and the other 3.
thetet has quit [(Remote host closed the connection)]
<contrapumpkin>
I see, yeah
pallav has joined #nixos
<LnL>
gchristensen: aarch64-linux builds <3
exarkun has joined #nixos
pallav has quit [(Client Quit)]
rmeis_ has quit [(Ping timeout: 268 seconds)]
<exarkun>
I'm having trouble getting stuff cached between builds. My builds happen in a Docker container and I'd prefer to throw the container away between builds. I can't figure out what I should _keep_ to avoid having to re-download everything for every build. I thought I could `nix-push` my derivation and then `nix-build --option binary-caches ...` but that doesn't appear to help.
<ikwildrpepper>
Dezgeg, gchristensen, niksnut: awesome that the arm stuff is starting to build now :)
<NixOS_GitHub>
[nixpkgs] globin pushed 4 new commits to master: https://git.io/vyzWz
<exarkun>
bendlas: I don't really know how all this stuff works but there's ... "profiles"? that work by having symlinks that go through /nix/var, for example.
<globin>
contrapumpkin: I didn't see a mass rebuild when starting to build firefox..
<contrapumpkin>
globin: perhaps ICU is only in stdenv in darwin?
<globin>
seems to be darwin only
<contrapumpkin>
either way, it's miserable :)
<sphalerite>
Is there a way to stop nix-build from trying to use a binary cache? I have an unreliable internet connection and I don't like how it hangs when I try to build something I know won't be found in the binary cache anyway
<bendlas>
exarkun: have you been talking to bennofs by chance?
<bendlas>
exarkun: OK, then please start from the beginning, I'm failing to recollect a prior conversation, right now ...
<exarkun>
oh blech
<sphalerite>
contrapumpkin: good thing there's a quality zsh completion for nix :D
<exarkun>
Sorry
<contrapumpkin>
sphalerite: it even completes --option values? I wonder how that works
<setgid>
Is there some good writeup that helps nix newbies understand how all these recursive definitions in nixpkgs fit together and how they relate to the configuration.nix?
<sphalerite>
contrapumpkin: black magic. I was surprised too
<exarkun>
bendlas: yes I was talking to bennofs, I misunderstood
<gchristensen>
"Load average: 3030 1228 839"
<sphalerite>
gchristensen: o.o
<contrapumpkin>
sphalerite: I'm assuming someone must've listed them out for the completion, since nix doesn't have a programmatic way to query them
<exarkun>
dtzWill: I want caching for build dependencies too. nix-copy-closure seems like it takes steps to avoid those?
<contrapumpkin>
globin: will you revert or is that a pain?
<gchristensen>
sphalerite: that isn't so bad, 31pts per core
<contrapumpkin>
I can trigger a hydra build I guess and reduce pain
<Dezgeg>
did you remember to grow the root filesystem btw? :P
<sphalerite>
gchristensen: 100 cores? Nice
<gchristensen>
Dezgeg: I did :) conveniently swap being sda2 and old-root being sda3 made that easy
<Dezgeg>
yes, I wouldn't have picked the swap partition as the target otherwise
goibhniu has quit [(Ping timeout: 264 seconds)]
<globin>
contrapumpkin: I'd actually like that in asap, probably need those in 17.03, too, as the firefox ESR support will run out during the support period otherwise..
<contrapumpkin>
alright, the usual "darwin relegated to second-class citizenry" complaints apply though
<contrapumpkin>
it's pretty miserable developing darwin in nix when linux folks clobber us regularly
<contrapumpkin>
I don't know what to do about it, of course
<sphalerite>
Switch to linux :D
<gchristensen>
sphalerite: don't be mean :(
<contrapumpkin>
not helpful
<dtzWill>
exarkun: ah, probably, not sure about those. Maybe --include-outputs similar to nix.conf's keep-outputs? but sounds like not what you want after all :)
pallav has quit [(Remote host closed the connection)]
<sphalerite>
Sorry, I had to :p
<gchristensen>
I don't think think "Don't merge mass-rebuilds" solution is very helpful
<contrapumpkin>
no you didn't
<contrapumpkin>
gchristensen: if we had a PR hydra, then I'd be tempted to lock master
<contrapumpkin>
and have a process manage master
<gchristensen>
LGTM
<contrapumpkin>
not that we have the infrastructure in place to do that realistically today
<contrapumpkin>
but in my ideal world
<gchristensen>
the hardest part for darwin is the limited build resources
<gchristensen>
and the difficulty in getting more
<contrapumpkin>
well, the hardest part is that linux developers can't easily figure out if they're breaking darwin
<contrapumpkin>
(and vice versa)
sphaugh has joined #nixos
<LnL>
^ yeah, mass rebuilds are more annoying for us because of that
<gchristensen>
did it break darwin, or rebuild darwin?
<MoreTea>
contrapumpkin, gchristensen what about creating a RFC for that?
<contrapumpkin>
rebuild
<gchristensen>
right
<contrapumpkin>
but often it also breaks
<contrapumpkin>
same kind of thing really
<gchristensen>
I have found that to be true, too :(
<contrapumpkin>
the issue is that they can't run anything on darwin
endformationage has joined #nixos
Arjun has joined #nixos
<contrapumpkin>
we could be stricter about nox review if it somehow did some darwin smoke tests
<LnL>
yeah adding glibc or systemd dependencies without a conditional happens frequently
<contrapumpkin>
I'd use it more if it told me stuff about other platforms
c0bw3b has left #nixos []
<gchristensen>
travis is 90% useless already
<contrapumpkin>
MoreTea: yeah, need to think about how to actually do it
Arjun is now known as Guest68339
<setgid>
Can I enable '--option build-cores 8' globally?
<sphalerite>
setgid: yes
<LnL>
setgid: yes in nix.conf
<sphalerite>
setgid: on nixos, use the option nix.buildCores
<setgid>
nice, ty :)
<gchristensen>
contrapumpkin: PR build infrastructure is almost here, the limiting factor is having a secure hand-off between github PR being approved & hydra
<gchristensen>
(approved meaning a Member says "hey hydra build this PR at this hash")
<MoreTea>
gchristensen, ok, so we need a github bot
<gchristensen>
yeah and it somewhat exists already
<gchristensen>
but the important thing is the hand-off can pin the sha to build and triggers it in hydra
<LnL>
gchristensen: contrapumpkin: I think what would also help a lot is to have a tool that evaluates everything for the different platforms
<MoreTea>
I started working on extracting a mapping of files -> maintainers based on the pkg.x.meta.maintainers
<contrapumpkin>
LnL: nox could be that I think
<gchristensen>
that is only tangential to the problem, but seems good
<MoreTea>
we probably should try to add the maintainers stuff as well to the github bot.
<MoreTea>
gchristensen, so this GH bot should should create a jobset in hydra for each PR, with an input to pin the commit SHA?
xeviox has quit [(Quit: Leaving)]
<gchristensen>
MoreTea: yeah, when a Trusted User (has commit rights to NixOS/nixpkgs) adds a special comment (like "+build", or "@nixosbuildbot build this thing") this bot should check the current sha of the PR, merge with the target branch, and then create-or-update a corresponding jobset in hydra
<MoreTea>
gchristensen, we could probably do something interesting there with allowing maintainers to OK stuff for their packages directly in V1.1
<gchristensen>
could be
<gchristensen>
I'd be happy with a v0.1 to start with :P
<MoreTea>
did you have a set of .nix expressions to setup a hydra locally with e.g. nixops?
<LnL>
hydra is starting to send emails about the arm builds :)
<globin>
contrapumpkin: after the release I'll continue the work on the hydra pr stuff, sry for creating the pain..
<gchristensen>
I did, MoreTea, but you don't need that for debugging
Yaniel has joined #nixos
<Yaniel>
Is there some requirement for packages to produce a binary with the same name as the package?
<contrapumpkin>
no problem really, I'll just work from HEAD~1
<Yaniel>
I'm trying to package renderdoc and it fails because there is no binary called 'renderdoc'
pallav has joined #nixos
<FRidh>
anyone knows of a program that is wrapped several times? I would like to check what the name is.
<gchristensen>
MoreTea: if you exec `hydra-create-or-update-jobset nixos --triger --pull-request 12345 --description "fiddle the frobs" --sha d4bb1c7d4bb1c7d4bb1c7d4bb1c7` it'll do the right thing
<sphalerite>
Yaniel: no, there isn't
<sphalerite>
Yaniel: What's your expression and the error message?
derjohn_mob has quit [(Ping timeout: 260 seconds)]
<NixOS_GitHub>
[nixpkgs] copumpkin opened pull request #23646: buildbot: fix .withPlugins to propagate inputs (master...buildbot-cleanup-take-2) https://git.io/vyzzc
<contrapumpkin>
FRidh: wrapping multiple times clobbers old wrappers, iirc
<FRidh>
contrapumpkin: Thanks. I know we sometimes have double wrappers that work, I just forgot how exactly their name is. I think it was .foo-wrapped-wrapped with one dot,right.
<contrapumpkin>
globin: note the tons of invalidated builds :(
joneshf-laptop has quit [(Ping timeout: 240 seconds)]
<contrapumpkin>
FRidh: hmm, the python double wrapper seemed to work differently at least. Not sure about others
<gchristensen>
contrapumpkin: that is probably the aarch64 builds?
<contrapumpkin>
FRidh: that is, python has a modified python script wrapper, plus a shell wrapper around ithat
<gchristensen>
sphalerite: no, reactions aren't an appropriate tool here
<contrapumpkin>
gchristensen: green went from 34000+ to 8000
<gchristensen>
ahh right right
chakerbenhamed has quit [(Ping timeout: 240 seconds)]
yenzenz has quit [(Ping timeout: 240 seconds)]
c0bw3b has joined #nixos
<globin>
contrapumpkin: hmm maybe something like systemd then :/ at least not the stdenv, feel free to revert if it causes too much pain, I'm on the tube right now..
<contrapumpkin>
globin: I don't really know or have a linux box to test on hand... if you're okay with it I'll leave it alone but it makes me a bit sad :)
<gchristensen>
contrapumpkin: speaking of linux boxes
<contrapumpkin>
oh shit
<contrapumpkin>
let me fix that now
Wizek has quit [(Ping timeout: 268 seconds)]
<gchristensen>
by the end of the day, anyway :)
sphaugh has quit [(Quit: Page closed)]
uwap has joined #nixos
ertesx has joined #nixos
mudri has quit [(Ping timeout: 260 seconds)]
<exarkun>
I'm having trouble getting stuff cached between builds. My builds happen in a Docker container and I'd prefer to throw the container away between builds. I can't figure out what I should _keep_ to avoid having to re-download everything for every build. I thought I could `nix-push` my derivation and then `nix-build --option binary-caches ...` but that doesn't appear to help.
ertes has quit [(Ping timeout: 240 seconds)]
ertesx is now known as ertes
<contrapumpkin>
FRidh: I really link my simplification to the withPlugins :)
<contrapumpkin>
I hope it's correct
<contrapumpkin>
fridh: I also kind of prefer --set to --prefix because I happened to be running in a nix-shell with a dependency I needed and accidentally omitted it from the expression but nothing broke :(
<contrapumpkin>
(until later when I ran it outside the nix-shell)
ij has joined #nixos
hotfuzz_ has joined #nixos
kampfschlaefer has joined #nixos
hotfuzz has quit [(Ping timeout: 240 seconds)]
<sphalerite>
what does the "cannot refer to other paths" error thrown by builtins.pathExists actually mean? I don't really get it, and reading the source hasn't helped (tbf I've barely read any of nix's source, so that's not really surprising)
fresheyeball has quit [(Ping timeout: 246 seconds)]
<LnL>
gchristensen: my hydra also started to rebuild the world
simendsjo has quit [(Remote host closed the connection)]
<NixOS_GitHub>
[nixpkgs] joachifm pushed 4 new commits to master: https://git.io/vyzVQ
<NixOS_GitHub>
nixpkgs/master adf044e Joachim Fasting: nixos/dnscrypt-proxy: refactoring...
<NixOS_GitHub>
nixpkgs/master e72aaa7 Joachim Fasting: nixos/dnscrypt-proxy: support updating before nss is up...
<NixOS_GitHub>
nixpkgs/master 5f27abe Joachim Fasting: nixos/dnscrypt-proxy: more fs isolation for the updater...
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to master: https://git.io/vyzwm
<NixOS_GitHub>
nixpkgs/master a3a21a7 Tuomas Tynkkynen: delve: Disable on i686
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to release-17.03: https://git.io/vyzwn
<NixOS_GitHub>
nixpkgs/release-17.03 6c0b236 Tuomas Tynkkynen: delve: Disable on i686...
<contrapumpkin>
lol, that packet box now takes up half of the machine status page
<gchristensen>
it is a busy fella :P
<contrapumpkin>
should we add a sponsorship/thanks page to the bottom of the nixos homepage?
<contrapumpkin>
more than promoting companies, I'd see it as a quiet endorsement
<gchristensen>
yeah, we need to add Packet to the site somewhere
<contrapumpkin>
or perhaps on the hydra homepage
<gchristensen>
not sure where exactly
<gchristensen>
ooo! for the first time, the Packet t2a box is running at capacity - it hit 96 concurrent builds.
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to master: https://git.io/vyzrz
<NixOS_GitHub>
nixpkgs/master 8dce538 Tuomas Tynkkynen: gtkglextmm: Mark as broken...
<contrapumpkin>
\o/
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to release-17.03: https://git.io/vyzro
<NixOS_GitHub>
nixpkgs/release-17.03 b89623e Tuomas Tynkkynen: gtkglextmm: Mark as broken...
<clever>
copumpkin: yeah, i'm moving away from it as well, but it would take a while to redo the entire thing
danielrf has quit [(Read error: Connection reset by peer)]
<copumpkin>
I'd make an issue about it and mention it in the ZHF thread
nckx has quit [(Quit: ☭ + 🐧 + GNU Guix)]
bennofs has joined #nixos
mudri has quit [(Ping timeout: 240 seconds)]
Mercuria1Alchemi has quit [(Ping timeout: 246 seconds)]
ertes has quit [(Quit: Bye!)]
nckx has joined #nixos
ertes has joined #nixos
ertes has quit [(Client Quit)]
<contrapumpkin>
ikwildrpepper, niksnut: mac4 seems to be idle despite there being a ton of work for mac hydra
ertes has joined #nixos
<gchristensen>
could it be that they're not yet runnable?
<contrapumpkin>
perhaps, but it seems to be building assorted random stuff at this point, not the conventional bootstrap
<gchristensen>
ah
<contrapumpkin>
maybe still working on an old evaluation though?
<contrapumpkin>
hard to say
<contrapumpkin>
nckx: you're on guix instead of nix now?
<sphalerite>
Has anyone here used guix? How does it compare to nix in terms of usability? Or does anyone know an article comparing them?
<gchristensen>
how do you feel about hard-liner approaches to free software?
<nckx>
contrapumpkin: mainly, yes, although I still have a few NixOS boxes chugging away (GuixSD is still too rough around some edges). And too little free time to contribute much to either, unfortunately.
<contrapumpkin>
ah, that's a pity
* nckx
should take that plug out of his whois, though, that was a half-joke.
<bennofs>
i would like a guix that wasn't so focused on free software
<gchristensen>
bennofs: here we are :P
<contrapumpkin>
I still quite like the lack of power of the nix language :)
<contrapumpkin>
so "a general-purpose language for specifying packages" is negative appeal for me
<contrapumpkin>
but I see why some people would prefer it
<bennofs>
contrapumpkin: well, i would like it if it actually lacked power, but right now it has *just* enough power so that people use it
<contrapumpkin>
I don't really care about power in the theoretical spectrum
<contrapumpkin>
i.e. if it's turing complete or not
<contrapumpkin>
I'm talking about ability to mess with the system evaluating it
<bennofs>
I think an architecture like XML and XSLT would be good, where you have a data-based specification and then a programming language to "transform" the specs into derivs
<contrapumpkin>
probably, as long as you make it look nothing like either of those ;)
<ertes>
my main concern about guile scheme would be the lack of laziness… some abstractions used in nixpkgs rely on that
<bennofs>
:D
* contrapumpkin
's eyeballs bleed at the thought
<gchristensen>
omg bennofs
<nckx>
contrapumpkin: I started getting into Lisp/Scheme by coincidence, and Guix caught my attention at just the right time. So I like it a lot, but I'm hopelessly biased.
<contrapumpkin>
the way I'd phrase what bennofs is saying
<gchristensen>
bennofs: sounds like the ".fix" files that used to be in Nixpkgs
<contrapumpkin>
is "explicitly model your packages as data"
<contrapumpkin>
which is something we're lacking right now in nixpkgs
<nckx>
contrapumpkin: ...with bash snippets.
<contrapumpkin>
well, I don't want my data model to have bash snippets in it, ideally :)
<bennofs>
nckx: I'd argue that 80% of nixpkgs does not need bash snippets
<contrapumpkin>
bash snippets are in the "massage" layer, which takes the modeled data and transforms it into something practical
<bennofs>
nckx: and the remaining 20% should use python or something more powerful than bash instead, because those are often quite complex bash scripts
<contrapumpkin>
yeah probably
<nckx>
Or Guile. /ducks
<sphalerite>
gchristensen: I feel like they're very important for the community in general, because it's thanks to them that free options exist for many things, but not as easy to use in themselves.
<contrapumpkin>
what's a decent scripting language with the smallest dependency closure
<gchristensen>
sphalerite: sorry?
<dash>
contrapumpkin: openfirmware's forth?
<contrapumpkin>
hah
<exarkun>
bennofs: Hey you dropped off before I could answer before. /nix has too much stuff in it. that said, copying everything out and then back in to it is what I ended up doing. :/
<ertes>
contrapumpkin: there are fairly small schemes
<contrapumpkin>
okay yes let's switch nixpkgs to forth
<sphalerite>
gchristensen: in response to my feelings about hard-liner approaches to free software
<dash>
contrapumpkin: doesnt even need an OS
<contrapumpkin>
excellent :P
<bennofs>
contrapumpkin: nim?
<ertes>
contrapumpkin: i was looking for the same thing a few weeks ago, trying different languages, including many schemes, but ultimately i decided to just use haskell for that as well
<bennofs>
not that I have used it
<contrapumpkin>
perhaps! it's a pity that most of these languages make calling out to external programs rather painful
<ertes>
they just don't try to create the illusion that invoking child processes is a safe, deterministic thing to do =)
<contrapumpkin>
:)
<ertes>
bash tries really hard to create that illusion
<pikajude>
nix is great
<pikajude>
i really appreciate it when i don't have to figure out how the latex packaging system works
<ertes>
(and zsh and …)
<ertes>
in any case, if we replace bash in nixpkgs, please let's not replace it by python ;)
<copumpkin>
agda for all expressions
<copumpkin>
gchristensen: okay I cleared out your box, you can kill it
<copumpkin>
thanks :)
<pikajude>
we should replace bash with haskell ヽ ╭ ° ͜ʖ͡°╮ノ
<sphalerite>
perl everywhere
<pikajude>
a shell scripting dialect of haskell
<pikajude>
called "hash"
<sphalerite>
+1 for relevance, -1 for ambiguity
<nckx>
Haskell closure size ftw.
<gchristensen>
copumpkin: you're welcome :)
<ertes>
"dialect" is almost equivalent to "library" in haskell, and there are actually quite a few "shell scripting" libraries around (none of which i like)
<ertes>
nckx: that can be fixed by multi-output GHC
<gchristensen>
copumpkin: reassigning it to hydra soon enough :P
<ertes>
but it hasn't been done yet
<pikajude>
nah we should just get a quasiquoter that parses bash
<ben>
let's replace it w/ something that looks like bash but where you don't have to be afraid about quoting issues literally every time you use variables
<sphalerite>
wait, ghc's closure includes ghc-doc?
<dash>
ben: oilshell?
<ben>
also maybe something that reads structures values out of env vars, not just strings
<copumpkin>
gchristensen: don't trust me! I added a Ken Thompson-style hack that will pollute all hydra outputs forever, untraceably
<gchristensen>
copumpkin: oh, no way
maninalift has joined #nixos
<ertes>
ben: execline has that property, but it's also a rather awkward "language" for anything more complicated
<ben>
i feel like in the easy case you gotta degrade to basically bash and in the complicated case you want to be haskell and its all about figuring out the syntax for the inbetween parts
<bennofs>
i think you need a language with good meta-programming abilities, that's why I was suggesting nim
<ben>
dash: impressive that you, as a shell yourself, are advertising yet another shell ;)
<dash>
ben: utterly mad project to make a decent language for shell-ish use cases, and to *compile bash scripts to it*
<ertes>
i wouldn't mind scheme as a scripting language, but all of the schemes i've seen lacked *something*, be it a good toolchain, useful libraries, etc.
<dash>
ertes: sanity? good taste?
<ben>
dash: does it look cute tho?
<dash>
ben: maybe use a pink terminal font
<ertes>
dash: hmm?
<ben>
`(), mhm
<ben>
i like $()
<ertes>
i like:
sellout- has quit [(Ping timeout: 240 seconds)]
<RchrdB>
ertes, I think there was one called scheme48 which was trying to be a shell as an explicit project goal or something like that
<ertes>
(in other words: string handling in shell languages is utterly horrible)
<NixOS_GitHub>
[nixpkgs] joachifm pushed 1 new commit to master: https://git.io/vyzDm
<NixOS_GitHub>
nixpkgs/master 74f8e0f Joachim Fasting: torbrowser: 6.5 -> 6.5.1
<ertes>
RchrdB: there was also scsh, but i didn't like the experience
<sphalerite>
I like "${function "string ${function}"}"
<RchrdB>
actually i think I was thinking of scsh, which *implemented atop* scheme48
<bennofs>
I think the places were we currently use bash don't really need super solid type systems or guarrantes, since the code often runs in very constrained environments (always the same inputs) and is tested when the package is first developed
<bennofs>
it's more a documention for the steps you needed to take when packaging the software
<ben>
but I still feel bad every time I interpolate a variable into a string that's actually bash code :(
<sphalerite>
joachifm: for the change to precision strike — I can assume it's GNU sed, right?
<ertes>
bennofs: i use the type system as a development tool more than a correctness check
<sphalerite>
because the -i flag behaves inconsistently across implementations, I remember running into that when I was writing a script that also had to run on OSX...
<clever>
sphalerite: i dont see a pkgs.sed in nixpkgs, but i do see a pkgs.gnused
<sphalerite>
clever: and that's where my sed is from — but I'm not sure if it's okay to make assumptions about it
<sphalerite>
e.g. should it be busybox-proof
sellout- has joined #nixos
katyucha has quit [(Killed (barjavel.freenode.net (Nickname regained by services)))]
katyucha_ is now known as katyucha
katyucha_ has joined #nixos
<NixOS_GitHub>
[nixpkgs] copumpkin pushed 1 new commit to master: https://git.io/vyzyr
<NixOS_GitHub>
nixpkgs/master 4ae5f4d Dan Peebles: pythonPackages.cryptography: fix test on macOS 10.12...
freusque has joined #nixos
tvon has joined #nixos
twopoint718 has joined #nixos
<NixOS_GitHub>
[nixpkgs] shlevy opened pull request #23648: stripDirs: Silence annoying 'File format not recognized' errors (staging...silent-strip) https://git.io/vyzS1
<NixOS_GitHub>
[nixpkgs] teh opened pull request #23649: Pyrax (master...pyrax) https://git.io/vyz9w
mudri has joined #nixos
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 2 new commits to master: https://git.io/vyzHc
<NixOS_GitHub>
nixpkgs/master 5f5b871 Tuomas Tynkkynen: raspberrypifw, linux_rpi: 1.20161020 -> 1.20170303
<NixOS_GitHub>
nixpkgs/master 9188f2e Tuomas Tynkkynen: release.nix: Fix typo
<NixOS_GitHub>
[nixpkgs] dezgeg pushed 1 new commit to release-17.03: https://git.io/vyzHu
<NixOS_GitHub>
nixpkgs/release-17.03 cdd5709 Tuomas Tynkkynen: raspberrypifw, linux_rpi: 1.20161020 -> 1.20170303...
ilja_kuklic has joined #nixos
<NixOS_GitHub>
[nixpkgs] joachifm pushed 1 new commit to master: https://git.io/vyzQO
<NixOS_GitHub>
nixpkgs/master 116953f Joachim Fasting: torbrowser: callPackage can fill in missing params from xorg
griff_ has joined #nixos
<c0bw3b>
Dezgeg: just browsing through your "crash-bandicoot-password-cracking" repo.... ! ^^ crazy stuff to me
<Dezgeg>
hehe
maninalift has quit [(Ping timeout: 258 seconds)]
<Kendos-Kenlen>
What is the initrd used by nixos ?
<sphalerite>
Is there any sort of order to pkgs/top-level/python-packages.nix, or at least an attempt at one? It seems vaguely alphabetical but I've found at least 4 clusters of Ns
takle has quit [(Remote host closed the connection)]
<bennofs>
sphalerite: there's an attempt at one, per category (search for ##)
<copumpkin>
Kendos-Kenlen: it builds it like any other nix derivation
<copumpkin>
Kendos-Kenlen: you asking where the expression that defines it lives?
<Kendos-Kenlen>
copumpkin: clever I was wondering if it was a classic initrd like Dracut or one specific to NixOS
<copumpkin>
oh, we build everything :)
<bennofs>
sphalerite: it contains lines like ### DEVELOPMENT / LIBRARIES / JAVA
<bennofs>
sphalerite: for me at least
<sphalerite>
bennofs: that sounds like all-packages.nix, not python-packages.nix
<clever>
bennofs: your looking at all-packages.nix, but sphalerite mentioned python-packages.nix
<bennofs>
sphalerite: oh sorry, I can't read
<bennofs>
sphalerite: I don't think python-packages.nix has an ordering, often times, new packages are just added at the end (at least that's what it looked like to me the last time i looked at it)
<sphalerite>
hm, those all seem to be callPackage ones which are presumably more complex. The last one that isn't through callPackage is zipfile36
<viric>
aarch64 in hydra? Great! Dezgeg, is this you?
<sphalerite>
I'll just put nltk somewhere amongst one of the N clusters :D
lucasOfBesaid has quit [(Quit: Page closed)]
<contrapumpkin>
viric: at least partly gchristensen, not sure who else was involved :)
<contrapumpkin>
viric: and not just any aarch64, but 96 cores of it
<contrapumpkin>
:P
<clever>
contrapumpkin: v7 as well?
<viric>
contrapumpkin: wow not bad
<viric>
contrapumpkin: can aarch64 build 32-bit arm? that'd be great
<viric>
(the same kernel, I mean)
<contrapumpkin>
no idea!
<contrapumpkin>
I haven't been involved in ARM in a while
<clever>
i would expect it to be able to do that
<clever>
64bit x86 can do the same
<viric>
that would be of great help
<Dezgeg>
I think it depends on the chip implementation
<NixOS_GitHub>
[nixpkgs] abbradar pushed 1 new commit to master: https://git.io/vyzF4
<NixOS_GitHub>
nixpkgs/master f36b721 Nikolay Amiantov: primus: fix if nvidia is not used...
<contrapumpkin>
last I looked at ARM, they could even have different endiannesses
<contrapumpkin>
admittedly that was a while ago
<Dezgeg>
yes, they deprecated that in aarch64
<contrapumpkin>
yeah, every actual one I saw used LE
danielrf has joined #nixos
<Dezgeg>
I had this plan for setting one of the ARM devices at work in BE mode (to test code on big endian)
<Dezgeg>
but my current plan is buying old powermacs :)
<LnL>
so would nixos now run on a raspberry, or does that still need some work?
<Dezgeg>
pi3 is capable of aarch64
<LnL>
oh, I thought that was armv7
<LnL>
or is that the same? I have no idea :)
<Dezgeg>
it's not the same
<sphalerite>
Dezgeg: How about POWER8/POWER9? :D
<Dezgeg>
raspbian only supports running pi3 in armv7 mode though
<LnL>
cool, I might buy one to play around with then
<Dezgeg>
POWER could work too, but that probably costs more (I offered my friend I trade his G5 for some wine bottles)
<sphalerite>
haha fair enough
derjohn_mob has joined #nixos
JagaJaga has quit [(Ping timeout: 264 seconds)]
AllanEspinosa has joined #nixos
takle has joined #nixos
<clever>
LnL: the pi3 is aarch64, but its also backwards compat to armv6 and armv7
<clever>
the raspbian userland is always compiled in armv6, so it maintains compat with the rpi1
<viric>
but can the aarch64 kernel run armv6 and armv7?
<clever>
but it has 2 kernels, v6 and v7
<LnL>
yeah Dezgeg mentioned that, but apparently we support aarch64 and raspberian doesnt :D
<clever>
LnL: a lot of the gpu accel stuff involves passing userland pointers to the gpu in 32bit fields as tokens
<clever>
LnL: and the gpu then returns those 32bit pointers back in events
<clever>
its not going to be happy when you stick a 64bit userland in the mix
<viric>
oh
<Dezgeg>
I think that only applies to the non-free gpu implementation
<clever>
yeah
<viric>
aarch64 with 32-bit pointers would be fine
<Dezgeg>
because glmark2 runs fine with the mesa vc4 driver in 64-bit mode
<clever>
x86 64bit has a MAP_32BIT flag in mmap
<clever>
which forces the upper 32bits to be 0
<viric>
ok
<clever>
and there was talk of implementing the same flag in the aarch64 kernel
<clever>
then if the non-free gpu userland uses that to allocate memory, it would be safe to truncate the pointers
<LnL>
copumpkin: I tried to reproduce it last week but had no issues after ~20 --check builds
c0bw3b has joined #nixos
nh2 has joined #nixos
setgid has quit [(Quit: leaving)]
<copumpkin>
oh so they only happen on hydra?
<MoreTea>
gchristensen, I hacked a minimal bot to trigger the hydra script
<Ralith>
Yaniel: are you using qt's callPackage?
<LnL>
copumpkin: pretty sure some other people also run into issues when building locally, but I'm not running into any issues
<Yaniel>
qt5.callPackage, yes
<Ralith>
Yaniel: also presumably it's your preFixup that's causing the error
<Ralith>
it sounds like you're just passing it the wrong filename?
<Yaniel>
to what?
<Yaniel>
ohh
<Yaniel>
nvm I'm blind
<copumpkin>
LnL: yeah it built fine for me too
<Yaniel>
thanks ralith
<Ralith>
\o
<Ralith>
thanks for packaging this!
<Ralith>
will definitely be using it
cpennington has quit [(Remote host closed the connection)]
<Yaniel>
I hope someone will gloss it over for any glaring mistakes preferrably before I make a PR
<Yaniel>
btw how soon would I have to get said PR in to have it in the next stable?
<gchristensen>
9 days ago
<gchristensen>
but we may be able to sneak it in, depending
<Ralith>
you can always just copy the package into your config.nix if not
<NixOS_GitHub>
[nixpkgs] rycee pushed 5 new commits to master: https://git.io/vygta
<NixOS_GitHub>
nixpkgs/master 661048c Justin Bedo: platypus: init at 2017-03-07
<NixOS_GitHub>
nixpkgs/master 7f214e6 Justin Bedo: bedtools: init at 2.26.0
<NixOS_GitHub>
nixpkgs/master 4bb50fd Justin Bedo: picard-tools: init at 2.7.1
<Ralith>
having stuff in stable isn't actually that important
<Yaniel>
sure - I mean I wrote it :P
<Yaniel>
well I'd say having it in stable drastically lowers the entry barrier
hiratara has quit [(Ping timeout: 246 seconds)]
<Ralith>
since you asked for review, a few other comments: I think you can set cmakeFlags[Array] in the package directly, rather than going through preConfigure
<Yaniel>
but I suppose I'll leave that to whoever is in charge of those things
<Ralith>
I'm also a bit surprised that there's no dependency on the loader
<Yaniel>
I tried to set cmakeFlags but then it wouldn't accept $out there
<Ralith>
oh, right
<Yaniel>
if there's a nicer way around that I'm all ears
<Ralith>
your current approach makes sense then
<NixOS_GitHub>
[nixpkgs] rycee closed pull request #23585: init a collection of bioinformatics tools (master...bioinformatics) https://git.io/vyRJT
<Ralith>
I mean, there might be some way to define it relative to the prefix or something
<Ralith>
using cmake terms
<Ralith>
but diving into cmake for that sort of thing is a can of worms and there's nothing really deeply bad about what you're doing
<LnL>
copumpkin: it does seem to fail a lot on hydra.org, I wonder why that is
<Yaniel>
yeah cmake has like 5 different prefixes which mean almost the same thing
<Ralith>
I'd be more concerned about the variable expansion rules
<Yaniel>
but use the wrong one and nothing will work
hiratara has joined #nixos
<Yaniel>
I don't need to do anything about the non-qt binary, right?
tvon has joined #nixos
<Yaniel>
as in it'll Just Work
<Ralith>
if it's genuinely non-qt, I'd assume you're fine yeah
<Ralith>
note that people do sometimes write command line apps using qt
<Yaniel>
okay, let's see then
<Yaniel>
good point about the loader, it's not strictly speaking necessary but I assume most people who care about renderdoc will want vulkan support anyway
<Ralith>
Yaniel: note that if renderdoc uses dlopen to access the loader you may need to take measures to ensure the rpath entry isn't stripped as unused
jophish_ has quit [(Ping timeout: 260 seconds)]
<Ralith>
in particular, `patchelf --shrink-rpath` will probably erase it; not sure if it's invoked by default or how to disable it or what
<gchristensen>
globin: around?
<Yaniel>
according to baldurk there shouldn't be any problems
<NixOS_GitHub>
[nixpkgs] vbgl pushed 1 new commit to master: https://git.io/vygYK
<NixOS_GitHub>
nixpkgs/master 1fb6fc2 Vincent Laporte: smplayer: 17.2.0 -> 17.3.0
nekroze has joined #nixos
<Ralith>
Yaniel: what does that mean? does it not need access to the loader at runtime?
<Yaniel>
it doesn't dlopen
<Ralith>
was the build system automatically disabling vulkan support wholesale, then?
freusque has quit [(Ping timeout: 256 seconds)]
<nekroze>
Since upgrading to 17.03 my gitlab-runner cannot execute builds. it looks like when it tries to start a build it attempts to switch user via su which it cannot find. I have tried giving the service pkgs.su in the path and even /run/wrappers/bin as an extension to the path variable by overriding the systemd start script with no joy. Anyone know how to resolve this?
<Yaniel>
oh wait, that was librenderdoc.so, not vulkan
shaugh has joined #nixos
mythmon has joined #nixos
fhoffmeyer has joined #nixos
glines has joined #nixos
<LnL>
nekroze: did you do a live upgrade or a reboot?
<nekroze>
LnL: Rebooted
<LnL>
oh wait, pkgs.su is not the same as /run/wrappers/bin/su
<nekroze>
If I ssh to the machine I can see that `which su` is now at /run/wrappers/bin/su
<nekroze>
LnL: I assumed it wasnt but I gave it a go anyways
<nekroze>
My first attempt to fix it was to add export PATH=$PATH:/run/wrappers/bin to the start script of the runner
<LnL>
pkgs.su doesn't have the setuid bit so it won't work, does the gitlab service use the wrapper?
fhoffmeyer has quit [(Ping timeout: 260 seconds)]
<LnL>
nekroze: it would have to be /run/wrappers/bin:$PATH so the correct one gets priority
<nekroze>
LnL: Interesting. I will give that a go. The error says it could not find su in the path though so thats why I assumed appending was ok
<copumpkin>
anyone have any clue why we default to java7 on darwin and java8 on linux?
<copumpkin>
I guess I should check the blame, but running blame on all-packages makes me sad
<shaugh>
aha! thank you. i did manage a work around so it's good this is a known issue
<Yaniel>
ralith: how would a shell.nix look if I want a shell with the env defined by the package?
Sonarpulse has joined #nixos
<Ralith>
Yaniel: what is "the env defined by the package?"
<Ralith>
the original purpose and default behavior of nix-shell is to give you the build environment of a package
tvon has joined #nixos
<Yaniel>
ooh, TIL
<Yaniel>
does it use default.nix or package.nix by default?
civodul has quit [(Remote host closed the connection)]
<Ralith>
I think it searchers first for shell.nix and then default.nix
<Ralith>
a shell.nix that simply invokes callPackage is convenient, iirc
<Yaniel>
well I have a default.nix that invokes callPackage
<Ralith>
that works too
<Yaniel>
guess I could just rename that to shell.nix
<Ralith>
only really matters if you're expecting to call the package from elsewhere, whereas you're presumably planning on committing it to nixpkgs instead
<Yaniel>
is there a way to run all steps up to X?
<Ralith>
just run them by hand, there's not that many
<Ralith>
and you'll presumably want to skip unpackPhase
Shados_ is now known as Shados
thc202 has quit [(Ping timeout: 260 seconds)]
tvon has quit [(Quit: Peace out, y'all!)]
freusque has joined #nixos
griff_ has joined #nixos
ilja_kuklic has quit [(Ping timeout: 246 seconds)]
<Profpatsch>
I’m a bit concerned that the RFC process will create lots of bikeshedding.
<Profpatsch>
And that it’s going to be way too hard to get anything substantial in.
<Profpatsch>
As in “i don’t even want to bother, let’s just deploy on local hydra”-hard.
<gchristensen>
what does "Deploy on local hydra" mean?
athan has quit [(Remote host closed the connection)]
<Profpatsch>
(I’m a fan of loose & anarchistic processes)
<Profpatsch>
Well, just not push upstream.
<Profpatsch>
And use the stuff internally, as it’s not polished enough yet.
<Profpatsch>
Right now I PR stuff rather incrementally.
<Profpatsch>
One library function, another library function.
<Profpatsch>
Once RFCs are in place, I might not feel that it’s enough to create an RFC, get a second person …
<Profpatsch>
And once people think it’s a good idea now I suddenly have pressure to actually implement it.
<bennofs>
Profpatsch: I don't think it can get harder than it is to get big stuff in tbh
<Profpatsch>
If I were to ask for advice about bigger changes, I’d just open an issue and CC a few people right now.
<Profpatsch>
Or ask a few very bright people personally.
<Profpatsch>
Without need of being scrutinized by the whole world.
<Profpatsch>
imho the biggest reason for nixpkgs moving so fast was the very loose process so far.
<bennofs>
Profpatsch: how is nixpkgs moving fast? What are some big changes that got actually merged? The only ones I can remember are overlays and the all-packages restructure
<bennofs>
oh and the security stuff
<stepcut>
Dezgeg, or anyone else, any idea how to work around this OOM error? http://lpaste.net/4043322726880903168 I already have 1.5GB of swap enabled
<gchristensen>
bennofs: not a big change
<Profpatsch>
Also, nix is the perfect platform for decentralized development and spreading of ideas.
<Profpatsch>
It sounds to me as if every library change would have to go through the RFC process.
<dmj`>
anyone know how nixpkgs manages to not get timed out on travis?
<bennofs>
Profpatsch: I don't think that's how it is intended
<dmj`>
is there some kind of under the table dealings
<Profpatsch>
Well, that’s how it could turn out.
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<bennofs>
Profpatsch: at least as far as I understand, it's more intended for stuff like changing nixpkgs architecture (like reworking some framework, changing module system etc)
<Profpatsch>
I’m pretty sure that would need an RFC first.
<Profpatsch>
It just fell out of the practical stuff I tried to do.
<Profpatsch>
If I had to do an RFC, it might have taken thrice the time.
<Profpatsch>
I’m not sure if I would have done it in that case.
fhoffmeyer has joined #nixos
<Profpatsch>
(that’s of course the worst possible turnout, but who knows)
<bennofs>
Profpatsch: how does that change stdenv? :o
<Profpatsch>
Right now it’s: PR, and the people of interest will be mentioned.
<Profpatsch>
The more you change, the more you’ll be asked for advice by the system.
<Profpatsch>
bennofs: Ah, wait, wrong PR.
<Profpatsch>
But it’s a new module in the library.
<Profpatsch>
And adds an interface for generators.
<Profpatsch>
So by the new system I’m pretty sure that would need an RFc.
<bennofs>
Profpatsch: but note that you had to merge the PRs yourself in each time. i myself would have problems with that, I don't feel confident merging such things without an ack from someone
<bennofs>
Profpatsch: it doesn't change anything that already existed, so it cannot break much stuff. Not sure if it needs an RFC in that case
<Profpatsch>
bennofs: Before I merge anything I always wait for acks.
<Profpatsch>
Even for simple packages, there’s always somethign wrong someone else notices.
fhoffmeyer has quit [(Ping timeout: 240 seconds)]
<nekroze>
LnL: Changing the order of those paths does the same thing, su not found in path
<Profpatsch>
Just my thoughts, maybe it’s going to turn out awesome.
wangoe has joined #nixos
<ToxicFrog>
How long will 16.09 keep getting updates now that 17.03 is live?
<clever>
stepcut: check dmesg, that may tell you about oom issues
<bennofs>
ToxicFrog: i think gchristensen said about a month of security updates
<gchristensen>
17.03 is not live
<gchristensen>
17.03 is beta
<ToxicFrog>
Oh, it's not? It shows up in nixos.org/channels
<ToxicFrog>
Aah
<gchristensen>
also, I did commit to 1mo of security updates, and after that it is "Community Supported" ie: dead ;)
freusque has quit [(Ping timeout: 260 seconds)]
shaugh has quit [(Quit: Page closed)]
freusque has joined #nixos
RchrdB has quit [(Ping timeout: 240 seconds)]
<stepcut>
clever: alas, no OOM message in dmesg, but even if there was, I am not sure what I would do about it. I'm curious how anyone could have run this since I am on an RPi3 -- which has the most amount of RAM you can get on an RPi. (1GB)
griff_ has quit [(Quit: griff_)]
wangoe has quit [(Ping timeout: 240 seconds)]
nevermind has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<Yaniel>
hmm running configurePhase complains about nix_qt5 subdirectories being missing
<clever>
stepcut: using nix build slaves, you can do the ram-heavy eval on another box
<clever>
[clever@amd-nixos:~/apps/nixpkgs]$ nix-build nixos/default.nix -I nixos-config=nixos/modules/installer/cd-dvd/sd-image-armv7l-multiplatform.nix --argstr system "armv7l-linux" -k -A config.system.build.sdImage
<Yaniel>
qmake: File exists; no configure script, doing nothing
<Yaniel>
the heck?
<clever>
stepcut: with that config and command, i can initiate an arm build on an x86 machine
<clever>
stepcut: and it will ssh into the pi to do the arm builds, but use the x86 ram for the nix expressions, where you are currently having the problems
<stepcut>
clever: aha
<stepcut>
clever: clever!
<Dezgeg>
stepcut: maybe temporarily moving the .git directory out of nixpkgs helps a bit
<clever>
stepcut: nix.buildMachines in configuration.nix is how you configure that on nixos
<clever>
and what Dezgeg said can help also
<Yaniel>
also when I exit nix-shell it keeps the `__nix_qt5__` dir in place and the next time I try to run nix-shell it fails because of that
<clever>
building path(s) ‘/nix/store/n5n74pijd3x9jpcz0qrni9cdl7vgn1is-sd-image-armv7l-linux.img’
<nekroze>
quit
<clever>
/nix/store/gjhsysgkl234r95ly2kn1ypmfp704v4v-stdenv/setup: line 500: export: write error: No space left on device
nekroze has quit [(Quit: leaving)]
<clever>
Dezgeg: looks like my build compiled everything, but is now running out of space making the disk image
<stepcut>
I am hoping that some day I'll be able to use the odroid-c2 as a armv7 build machine, but it is proving difficult at the moment. There is hope that mainline Linux 2.11 will support enough of the odroid-c2 to be useful though
<teh[m]>
Profpatsch: I think bike-shedding is a valid concern, but there's no evidence of that yet. There's plenty of evidence of large, positive and not super controversial changes dying of bitrot though
<wrl>
stepcut: why not just cross-compile?
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 1 new commit to master: https://git.io/vygu2