<newdan>
Ralith: Thanks for the tip! I see a way in systemd to specify a /run folder but that's mounted as tmpfs, ideally the deployed code would remain after a system reboot
phreedom has quit [(Ping timeout: 246 seconds)]
paroneayea has quit [(Read error: Connection reset by peer)]
thc202 has quit [(Ping timeout: 258 seconds)]
<Ralith>
newdan: I was thinking of systemd-tmpfiles
paroneayea has joined #nixos
<Ralith>
setting a unit's RuntimeDirectory might also make sense if you need a custom unit already
<Ralith>
newdan: see nixos option systemd.tmpfiles.rules
<Ralith>
note "You should not use this option to create files required by systemd services"
<Ralith>
so perhaps arrange a systemd unit
<Ralith>
with suitable RuntimeDirectory
<newdan>
Ralith: it seems like RuntimeDirectory is intended for things that only need to exist while the service is running (I'm guessing stuff like pidfiles). I'm not a FS expert by any stretch but I get the feelilng it isn't where I want to put my app code
<Ralith>
newdan: another thing I've seen a lot is idempotent setup scripts stuffed into ExecStartPre
<Fare>
when I raco pkg install sha, I get raco setup: docs failure: query-exec: unable to open the database file
<NixOS_GitHub>
[nixpkgs] volth opened pull request #21747: xfce.mousepad: use keyfile instead of gconf (master...define-MOUSEPAD_SETTINGS_KEYFILE_BACKEND) https://git.io/vMCX5
sellout- has quit [(Ping timeout: 240 seconds)]
alx741_ has quit [(Quit: alx741_)]
sellout- has joined #nixos
sellout- has quit [(Max SendQ exceeded)]
alx741 has joined #nixos
sellout- has joined #nixos
Wizek_ has quit [(Ping timeout: 245 seconds)]
ryantrinkle has quit [(Ping timeout: 255 seconds)]
ryantrinkle has joined #nixos
mrkgnao has joined #nixos
mkoenig has quit [(Ping timeout: 245 seconds)]
mkoenig has joined #nixos
<mrkgnao>
how do I make the X server create a log?
<mrkgnao>
/var/log/X{org}.0.log doesn't exist
mizu_no_oto has quit [(Quit: ["Textual IRC Client: www.textualapp.com"])]
[0x4A6F] has quit [(Ping timeout: 260 seconds)]
[0x4A6F] has joined #nixos
<hodapp>
foiled again, my old NixOS installation on a Raspberry Pi doesn't want to be used as a print server because ImageMagick is not supported(?)
<hodapp>
but it's an install I did early on in 2016 and I don't remember it working amazingly there
justanotheruser has joined #nixos
<mrkgnao>
does anyone know how to make emacs not have non-jagged fonts? I run into this every time I do a new install of any Linux, solve it, and then proceed to forget about it
<mrkgnao>
there's some Xresources trick
OfficialLeibniz has quit [(Ping timeout: 256 seconds)]
mguentner has quit [(Quit: WeeChat 1.6)]
<hodapp>
huh, I have Emacs on both my NixOS boxes and fonts aren't jagged
<hodapp>
it's possible something in my configuration gets it that way, I don't know
<ToxicFrog>
I'm not convinced that getting sshfs into the unit closure would help, the fact that it's looking in /no-such-path implies that $PATH isn't set at all
<ToxicFrog>
(and for the record, yes, sshfs is in systemPackages and running it manually works just fine)
mkoenig has quit [(Ping timeout: 245 seconds)]
mkoenig has joined #nixos
<ToxicFrog>
Hmmmm
Wizek has quit [(Ping timeout: 245 seconds)]
<ToxicFrog>
So if I run the same `mount` command that systemd runs by hand, PATH gets passed to mount(8) just fine
<ToxicFrog>
But when it forks and calls mount.fuse(8), PATH doesn't get propagated.
Fare has quit [(Quit: Leaving)]
<ToxicFrog>
Hah. mount(8) is specifically coded not to permit PATH to propagate to its children.
<ToxicFrog>
Presumably because otherwise you could do something like: PATH=/home/blackhat/bin:$PATH mount -t fuse.evil ...
<ToxicFrog>
And have /home/blackhat/bin/mount.fuse.evil execute as root.
<ToxicFrog>
So any FUSE module depends on /bin/sh having a sensible default path compiled in, which in Nix it really doesn't.
ublubu has quit [(Ping timeout: 272 seconds)]
[0x4A6F] has quit [(Remote host closed the connection)]
[0x4A6F] has joined #nixos
<ToxicFrog>
Filed a bug.
<ToxicFrog>
It seems like this would be easily fixed by compiling a sensible default path into bash, so I'm going to try that now.
mguentner2 has joined #nixos
roconnor__ has quit [(Ping timeout: 240 seconds)]
mguentner has quit [(Ping timeout: 245 seconds)]
ublubu has joined #nixos
<hodapp>
systemd.mounts? I need to read about this
<mrkgnao>
hodapp: I figured it out, apparently you need hintfull in Xresources instead of hintslight like Arch+infinality requires
<hodapp>
huh
* mrkgnao
feels a thorn leave his side as the world looks beautiful again.
<mrkgnao>
although it's great how nixos has excellent font rendering by default
<mrkgnao>
except for this weird thing ... iosevka bold/heavy weights seem to be missing the 't' glyph somehow and falling back on some shitty unhinted font
<mrkgnao>
this is the weirdest font rendering bug I've seen, ever
<hodapp>
weird
<mrkgnao>
can I force a build from source?
sdothum has quit [(Quit: ZNC - 1.6.0 - http://znc.in)]
<ToxicFrog>
Ok, that didn't work out so well
<ToxicFrog>
Apparently modifying bash makes it want to rebuild the entire system
<ToxicFrog>
from source
<ToxicFrog>
hodapp: systemd.mounts autogenerates systemd mount units, which are basically eqv to fstab entries except they have more tuning knobs and automatically depend on parent mount points and the like
<ToxicFrog>
systemd.automounts generates systemd automount units, which let you automatically activate mount units when the mountpoint is accessed and automatically unmount them after some amount of time idle
<ToxicFrog>
Before installing Nix on this system, I had automount units for my sshfs mounts, so they'd automatically mount when accessed and then umount when idle.
[0x4A6F] has quit [(Ping timeout: 240 seconds)]
[0x4A6F] has joined #nixos
mkoenig has quit [(Ping timeout: 260 seconds)]
roconnor__ has joined #nixos
mkoenig has joined #nixos
<aszlig>
ToxicFrog: the problem shouldn't be in mount(8) itself
<aszlig>
ToxicFrog: the problem is with mount.sshfs, which should use the absolute store path of "sshfs" instead of relying on PATH
ertes has joined #nixos
athan has quit [(Ping timeout: 240 seconds)]
sdothum has joined #nixos
pie_ has joined #nixos
pie_ has quit [(Changing host)]
pie_ has joined #nixos
mkoenig_ has joined #nixos
<hodapp>
ToxicFrog: any good docs on this?
sigmundv has quit [(Quit: Leaving)]
mkoenig has quit [(Ping timeout: 240 seconds)]
sdothum has quit [(Quit: ZNC - 1.6.0 - http://znc.in)]
sigmundv has joined #nixos
johnsonav has quit [(Ping timeout: 255 seconds)]
sigmundv has quit [(Remote host closed the connection)]
<pie_>
how can i update my system? i seem to be stuck on some old stuff for some reason, for example i still appear to be running firefox 47
<pie_>
ive run nixos-rebuild switch --upgrade
zraexy has joined #nixos
danharaj has quit [(Ping timeout: 246 seconds)]
<pie_>
i have a 16.03 channel in my channel list for some reason
<pie_>
i see the latest is 16.09
ryantrinkle has quit [(Remote host closed the connection)]
<pie_>
nix-env uses a rediculous amount of ram btw
<mrkgnao>
ansible2 can't find python2 :(
<mrkgnao>
nix-shell -p ansible works, but not ansible2
<mrkgnao>
for now, I'm just setting ansible_python_interpreter to $(which python) in /etc/ansible/hosts
<mrkgnao>
but this is obviously suboptimal, what's wrong with the path rewriting? o.O
sboosali has quit [(Ping timeout: 240 seconds)]
siarheit_ has joined #nixos
glines has quit [(Ping timeout: 245 seconds)]
mkoenig_ has quit [(Ping timeout: 255 seconds)]
mkoenig has joined #nixos
mrkgnao has quit [(Quit: WeeChat 1.5)]
siarheit_ has quit [(Ping timeout: 248 seconds)]
roconnor__ has quit [(Ping timeout: 245 seconds)]
siarheit_ has joined #nixos
FRidh has joined #nixos
uralbash has joined #nixos
mkoenig has quit [(Ping timeout: 255 seconds)]
mkoenig has joined #nixos
<srhb>
mrkgnao: Which version of nixpkgs?
mkoenig has quit [(Ping timeout: 255 seconds)]
mkoenig has joined #nixos
nanobapt has joined #nixos
mkoenig has quit [(Read error: Connection reset by peer)]
<pie_>
help :C
<pie_>
collision between ‘/nix/store/qblj6pk8067czq8ya6bfllfvy0mndqjh-python3-3.5.2/bin/2to3’ and ‘/nix/store/7zqg0icnd3ymbzpgwsh4qdv064ccgwc4-python-2.7.13/bin/2to3’; use ‘nix-env --set-flag priority NUMBER PKGNAME’ to change the priority of one of the conflicting packages
<pie_>
i dont want to try shoving a square in a triangular hole
<pie_>
this is me updating with nix-env -u
<srhb>
pie_: Why do you need to have that package installed at all?
<srhb>
pie_: In your user environment, that is.
<pie_>
i install everything in my user environment :C
<pie_>
i am the stereotypical bad user
nanobapt has quit [(Ping timeout: 245 seconds)]
<srhb>
pie_: That invariably leads to this situation, I'm afraid.
<pie_>
maybe i dont need it
<pie_>
i guess i could just try installing it again later
<srhb>
Using development environments (nix-shell) will save you a lot of pain.
kampfschlaefer has joined #nixos
<srhb>
But yeah, you can just get rid of it for no.
<srhb>
now*
<pie_>
yeah i need to figure that out at one point
<pie_>
ok how to i get rid of it
<srhb>
nix-env -e packagename
<srhb>
You can see what it's called with nix-env -q
matthewbauer has quit [(Remote host closed the connection)]
matthewbauer has joined #nixos
mkoenig has joined #nixos
<pie_>
ok i think removing python 2 worked thanks
<pie_>
err too early :)
<pie_>
uh...
<pie_>
collision between ‘/nix/store/dci0fxi2i302agir9b1n8lq532q978q7-binutils-2.27/bin/addr2line’ and ‘/nix/store/4kkdfxqghb7ax6hfsmkfhz97zfwhwcp4-binutils-2.27-dev/bin/addr2line’; use ‘nix-env --set-flag priority NUMBER PKGNAME’ to change the priority of one of the conflicting packages
<srhb>
pie_: I think maybe you should paste your user env. It looks fishy.
<pie_>
i only have a binutils 2.26 in -q
<pie_>
well i just swapped out my 16.03 channel to 16.09 if that sounds like it would break something
<srhb>
pie_: You also have it installed globally, presumably.
<srhb>
pie_: If you upgraded your system as well, that should be fine. Well, really, either should be fine.
<pie_>
well its probably that i tried to install gcc in my env at some point because i couldnt figure out an easy way to use it or somethign :P
<srhb>
pie_: Do you need it? The easy solution out of all this is to delete, delete, delete. :)
<pie_>
yup
<pie_>
worst case i just offload the problem to later :PPP
<srhb>
Indeed.
<srhb>
Where perhaps you'll be motivated to do The Right Thing. ;-)
<pie_>
oh im motivated xD i just have other things to do before getting around to learning nix :c
<srhb>
Yeah.. :)
<srhb>
Well don't sweat it, you'll rarely break something.
<pie_>
thanks looks like upgrading my packages finished \o/
<pie_>
yeah nix is nice like that :P
<srhb>
Indeed.
<pie_>
i just get to feel dirtier and dirtier heh
<pie_>
ok firefox is now up to date
<pie_>
TIL i thought i was updating my stuff but apparently i wasnt...
matthewbauer has quit [(Ping timeout: 240 seconds)]
<pie_>
is this because i installed stuff to my environment?
<pie_>
or do i always need to run nix-env -u to update
<pie_>
(i thought nixos-rebuild switch --upgrade did it)
<srhb>
pie_: nixos-rebuild switch --upgrade updates the global channel and builds the new, upgraded system
<srhb>
nix-env -u updates your user environment to the current channel
<pie_>
right.
<pie_>
ok thanks
<srhb>
If you have different channels for your user and system, you'll need to upgrade your user channel as well
<srhb>
ie. nix-channel --update
<pie_>
ok
<pie_>
i see
<srhb>
When using channels you might want to NOT have a user channel at all.
<srhb>
It avoids some confusion, and the system channel will just be used.
mkoenig has quit [(Ping timeout: 240 seconds)]
mkoenig has joined #nixos
<pie_>
on an unrelated note, any comments on nixops vs cfengine?
<pie_>
or disnix or what is it
hyper_ch has left #nixos ["Konversation terminated!"]
hyper_ch has joined #nixos
<pie_>
also why is it that nix-env can consume rediculous amounts of memory?
<pie_>
not much of a point to having a "lightweight" e
<pie_>
xecutable if it does that..
bb has joined #nixos
nanobapt has joined #nixos
mkoenig has quit [(Quit: Lost terminal)]
alx741 has joined #nixos
<srhb>
pie_: The only thing similar to nixops I've ever used is ansible.
alx741_ has quit [(Ping timeout: 240 seconds)]
<srhb>
pie_: And it sucks so much I can't even express it in words.
meocs has joined #nixos
<pie_>
heh
<pie_>
noted.
<pie_>
i wish my room was stateless
<pie_>
people keep moving my stuff around and i just remembered i have a todolist i need to look at....if i could find it
knupe has joined #nixos
<siarheit_>
withothers' help i've got basic idea of overriding CFLAGS for a single package: http://dpaste.com/24B5Z74.txt . Is there an easy way to do it for all installed or to be installed packages?
<siarheit_>
it's also not clear if i can do this change only for a specified package + it's build-time depends
rly has joined #nixos
mudri has joined #nixos
[0x4A6F] has quit [(Ping timeout: 240 seconds)]
[0x4A6F] has joined #nixos
<rly>
How do I make my media keys (Fn + F3) work to increase the sound volume with Plasma 5?
<rly>
In KDE4 this all worked out of the box.
bb has quit [(Ping timeout: 260 seconds)]
maxxk has joined #nixos
matthewbauer has joined #nixos
[0x4A6F] has quit [(Ping timeout: 260 seconds)]
justbeingglad has joined #nixos
justbeingglad has left #nixos []
[0x4A6F] has joined #nixos
<NixOS_GitHub>
[nixpkgs] the-kenny pushed 1 new commit to master: https://git.io/vMC5r
<NixOS_GitHub>
nixpkgs/master 3dc8b5d Jay Mundrawala: rustUnstableBin.rustc: Run patchelf on bin/rustdoc (#21746)...
matthewbauer has quit [(Ping timeout: 240 seconds)]
<rly>
OK, so plasma-pa is just a piece of shit.
<rly>
If you want working media keys, install kde4.kmix.
<rly>
#whyevenshipit
pxc has quit [(Quit: WeeChat 1.5)]
<MichaelRaskin>
pie_: nix-env -i consumes a ton of memory because it evaluates all the package names in nixpkgs, nix-env -iA takes attribute paths and consumes less memory
hotfuzz_ has joined #nixos
hotfuzz has quit [(Ping timeout: 258 seconds)]
<pie_>
MichaelRaskin, nix-env -u was also using like 40% of my ram (i specifically closed stuff so that it would have enough)
<rly>
MichaelRaskin: nix-env -i could be working more intelligently.
<pie_>
so thats like 1.5 gigs of ram or somethign
<pie_>
i mean what if someone tried to use nixos on a less powerful machine?
<rly>
MichaelRaskin: there is no heuristic now that an attribute name could be related to the name attribute.
<pie_>
with like only a gig of ram or something
<MichaelRaskin>
pie_: NixOS without nix-env would be fine…
<rly>
pie_: the idea is that closures are copied there, AFAIK.
<MichaelRaskin>
Although a gig may be not enough to resolve a complicated config, I don't remember
<pie_>
re: i was easily using over a gig ;_;
<rly>
pie_: are we talking about a desktop system?
<pie_>
rly, i guess yeah, but for the sake of argument we could also say no?
<MichaelRaskin>
nix-copy-closure from outside and manual start of the activation script would work fine
maxxk has quit [(Ping timeout: 240 seconds)]
<rly>
pie_: for enterprise deployments you can also just use container technologies.
<rly>
pie_: playing around with NixOS on your Pi is a bit for the more expert users/hobbyists.
<MichaelRaskin>
With Pi, instantiate on desktop, then copy the derivations
<pie_>
i mean it just doesnt sound very nice for a linux os to require over a gig of ram at minimum :P
<rly>
pie_: I run Nix based systems with << 1GB of memory.
<rly>
pie_: Nix is a configuration management tool. The output of which does not have to be a traditional Linux system.
<rly>
pie_: e.g. Sony could build it to build a PS4 system.
<rly>
pie_: a hardware company could even use it to build the final chip design.
<rly>
pie_: support for these things is close to zero (probably you would have to hire someone to pull it off), but in principle that's possible.
<pie_>
well sure but its still used for linux
<rly>
pie_: I target Docker for example.
<pie_>
*most of the time
<rly>
pie_: my laptop has 16GB of RAM.
<pie_>
good for you lol
<rly>
pie_: that has been like that for 3 years now, or so.
<pie_>
mine maxes out at 4 :/ i was hoping i could get some upgrades but not supported
<rly>
pie_: I receently wished for 32GB.
<rly>
pie_: but I will probably fill in that usecase by building some infrastructure.
<rly>
pie_: NixOS also requires a SSD, IMHO.
<pie_>
i was using it fine without one
<rly>
The software isn't really optimized like PhotoShop is.
<rly>
pie_: I have more needs than most users, as you can imagine.
wkennington has quit [(Read error: Connection reset by peer)]
<rly>
pie_: I also use it on a machine without an SSD, but the waiting time annoys me.
<simpson>
rly: NixOS works fine without an SSD.
zraexy has quit [(Ping timeout: 240 seconds)]
<pie_>
well in that case its not fair to expect most users to have huge amounts of resources just because you do :P
<rly>
simpson: creating 20000 symlinks takes 10 seconds or so.
<rly>
simpson: on my current system I create 40K symlinks.
<simpson>
rly: Oh, I remember this conversation. Yeah, I dunno why you run into weird problems. I suspect that your setup is quirky somehow.
<rly>
simpson: it's not the setup; it's the scalability which is absent.
<rly>
simpson: we already discussed the required optimization to make it work on HDDs for people without patience.
<simpson>
rly: Patches are still welcome.
<simpson>
But, uh, I just don't have these problems on my NixOS fleet with non-SSDs.
<rly>
simpson: and your fleet installs how many packages interactively?
<rly>
simpson: zero?
<rly>
simpson: thanks
<rly>
simpson: on your work station, I am going to guess that you do have a SSD.
<siarheit_>
what filesystem you are using?
<rly>
siarheit_: ext4
<simpson>
rly: I have a not-quite-outdated old laptop, same as last time, and I'm administering a Hydra, same as last time.
<rly>
simpson: how many symlinks do you create?
<rly>
If it's 10 times less, it's not noticable obviously.
<rly>
Anyway, I think it's pointless to discuss facts or to say that a particular system is quirky. A good application would even be able to detect broken hardware at run-timme.
<rly>
run-tim
<rly>
run-time*
<rly>
A production system needs to do all these things, but I guess my definition of a production system goes a bit beyond most people's ideas of a production system.
<simpson>
It could take a while since it's not on an SSD~
<rly>
simpson: not agreeing with reality is just foolish; building a system free from failure (and being slow is failure too) is difficult and saying people have "quirky systems" is just plain denial.
<simpson>
rly: I'm an SRE. Do whatever it takes to make your services reliable. Just don't expect much sympathy if your attitude tends towards trash talk instead of contributing the features that you want.
<rly>
simpson: it's not trash talk; it's reality telling people that their castles are actually more like small cottages.
<rly>
simpson: my systems run with 100% reliability, btw.
<rly>
simpson: I could show you the charts.
jgertm has joined #nixos
<simpson>
rly: In reality, there exist multiple production NixOS systems running in many different environments. I'm not sure why you feel the need to say that something that is clearly production-ready is not production-ready.
<siarheit_>
must be a huge system
<simpson>
rly: When you say "reliability", what kinds of SLIs are you measuring? Availability, alacrity, latency, durability, etc.?
<rly>
siarheit_: the size of a system has nothing to do with reliability when properly engineered.
<siarheit_>
rly, i'm an also sre :)
<rly>
siarheit_: do you also design new services?
newbie has joined #nixos
<simpson>
rly: (Just so we're clear: Most organizations cannot afford 100%. I've worked for a huge-class organization and seen 99.999% systems. I am confident that your circumstances are exceptional.)
<siarheit_>
occasionally, yes
civodul has joined #nixos
<simpson>
siarheit_: Oh cool. I started at Google; I'm currently serving at Acquia.
<siarheit_>
nice :)
mrkgnao has joined #nixos
<mrkgnao>
how do I override rust to use the nightly?
<newbie>
I suppose this is not nix-related, but is there a way to completely disable notification sound (I could live with if there is the side-effect that all notification-messages would be disabled).
<rly>
newbie: in KDE?
<newbie>
newbie in nix, now using fvwm as default wm.
<mrkgnao>
what's your notif provider? dunst/twmn/... have config files you can tweak
tvon has quit [(Quit: System is sleeping...)]
tvon has joined #nixos
matthewbauer has joined #nixos
tvon has quit [(Client Quit)]
tvon has joined #nixos
tvon has quit [(Client Quit)]
<rly>
simpson: it also depends a bit on how you count your 100%. If the rest of the company's services are down, should yours still be up, even if it is useless in that context?
<newbie>
So ps -e | grep notif is where I should start, ty,.
tvon has joined #nixos
<simpson>
rly: This is why you define per-service SLAs.
tvon has quit [(Client Quit)]
<newbie>
I should goodle how to disable the ntification provider. For now i will jsut try to kil lthe process.
<newbie>
*kill
Lowl3v3l has joined #nixos
ilja_kuklic has joined #nixos
<rly>
simpson: sure.
tvon has joined #nixos
tvon has quit [(Client Quit)]
matthewbauer has quit [(Ping timeout: 240 seconds)]
tvon has joined #nixos
<c74d>
mrkgnao: AIUI, nightly releases (of anything) work poorly (conceptually) with the design of Nix packages, but that restriction has been hacked around for Rust: <https://github.com/solson/rust-nightly-nix>. (I've not used that packaging personally; Sonarpulse and scott would know more about it.)
tvon has quit [(Client Quit)]
<newbie>
mkgman kill -9 let me get rid of knotifty4 and xfcef-notifyd. But i can't get rid of the snotify_mark
<newbie>
* fsnotify_mark. But I suppose it is not the same thing (fs).
c0bw3b has joined #nixos
tvon has joined #nixos
tvon has quit [(Client Quit)]
<newbie>
Could i completely disable knotify and xfce4-notify in configuration.nix.?
<c74d>
newbie: fsnotify notifies software of changes in the filesystem, and isn't user-facing (although I'm not sure fsnotify_mark is for fsnotify)
tvon has joined #nixos
tvon has quit [(Client Quit)]
thc202 has joined #nixos
<mrkgnao>
c74d: I can't figure out how to put that in my configuration.nix
<srhb>
How do you guys install spacemacs on Nix? Using git out of band from nix?
<mrkgnao>
I mean, I can get it to "compile", but nix-env -qa doesn't mention the new package
<mrkgnao>
srhb: that's what I'm doing for now, but I understand that there is a not-very-high-priority to get spacemacs to work better with Nix
<mrkgnao>
s/ty/ty project/
<c74d>
mrkgnao: I think it should be stuck somewhere such that it can be used with `nix-shell -p`
<srhb>
mrkgnao: I just found that issue, yeah. Thanks. :)
<mrkgnao>
nix-shell -p can't find rustNightly :(
<srhb>
mrkgnao: What exactly are you trying and with which channel?
<c74d>
mrkgnao: I don't think Sonarpulse or scott are available at this time, but its possible (albeit unlikely) that there would be someone available to help you in #rust-offtopic on irc.mozilla.org:6697
<newbie>
mrkgnao Just a guess from a newbie, but what if If you put it in pks in configuration.nix, it should be enabled for all ussers.?
<mrkgnao>
about this rust thing?
<srhb>
mrkgnao: Yes.
<mrkgnao>
c74d: I think this is more a general nix question about package overrides though
<srhb>
mrkgnao: Because my nix-shell has no problem finding it, but you're not giving us enough to go on.
ambro718 has joined #nixos
<srhb>
mrkgnao: For instance, this works with a checkout of nixos-unstable: nix-shell -I nixpkgs=/home/sarah/src/nixpkgs -p rustNightly.cargo
<srhb>
So I'm wondering what you're doing, which channel you're using, etc.
<mrkgnao>
Um. I put the default.nix from the repo in /etc/nixos/rust-nightly.nix, and then put rustNightly = pkgs.callPackage ./rust.nix {}; in nixos.pkgs.packageOverrides in configuration.nix
<mrkgnao>
which does nothing, but also doesn't cause a rebuild to fail
<mrkgnao>
but I realised that I can use my local checkout of unstable as you showed, srhb
<srhb>
You'd probably want that in your config.nix, not configuration.nix
<srhb>
If you use the method from that repo, I mean.
<mrkgnao>
how are you doing it?
<srhb>
mrkgnao: Otherwise it would be nixpkgs.config.packageOverrides = ...
<srhb>
I'm not.
<mrkgnao>
I mean, I tried pointing nix-shell at my local nixpkgs checkout and it can't find rustNightly
<srhb>
mrkgnao: What command are you using, what branch is checked out at which version, and what's the error?
[0x4A6F] has quit [(Ping timeout: 255 seconds)]
<mrkgnao>
I have a local checkout of nixpkgs master from GitHub
<srhb>
mrkgnao: And which nix-shell command are you trying? And what's the error?
[0x4A6F] has joined #nixos
<mrkgnao>
undefined variable 'rustNightly' at <location>
<mrkgnao>
I did nix-shell -I . -p rustNightly.cargo
<srhb>
mrkgnao: Did you mean -I nixpkgs=.
<srhb>
(This is why you give the full command you tried) :-)
* mrkgnao
thinks he has learned a lesson, but this isn't the first time
<srhb>
;-)
<MichaelRaskin>
-I .. could also work, I guess
<srhb>
Yeah
<MichaelRaskin>
if . is called nixpkgs
<mrkgnao>
in my defense, -I means something like "search path" to my brain because of xrdb
mudri has quit [(Ping timeout: 240 seconds)]
Lowl3v3l has quit [(Ping timeout: 245 seconds)]
<srhb>
mrkgnao: That's not wrong, indeed, it searches for nixpkgs in that path
<srhb>
mrkgnao: But you're already in it.
<srhb>
mrkgnao: (Which is why -I .. would work)
<srhb>
Assuming it's actually named nixpkgs :)
jgertm has quit [(Ping timeout: 248 seconds)]
<mrkgnao>
with an implied nixpkgs=nixpkgs
<mrkgnao>
Yeah.
<srhb>
Right. :)
Lowl3v3l has joined #nixos
newbie has quit [(Quit: Konversation terminated!)]
latk has joined #nixos
<FRidh>
Does Nix do any scheduling regarding amount of concurrent jobs and how the cores are divided? I'm aware of --max-jobs and --cores. Can it lower the amount of concurrent jobs when a certain job has enableParallelBuilding set to true?
lukego has joined #nixos
<lukego>
Hoi! is there any HOWTO for installing NixOS on a Chromebook?
Ivanych has quit [(Quit: Leaving.)]
Ivanych has joined #nixos
Ivanych has quit [(Client Quit)]
Ivanych has joined #nixos
pi3r has joined #nixos
ebzzry has joined #nixos
<mrkgnao>
lukego: I'm running it on one now ... if you have any other Linux installed already then it's no different from installing on any laptop
<mrkgnao>
although I think going from Chrome OS to a Linux requires a bit of fiddling
<FRidh>
It's working, simulations and analyses in Nix :) It does import from derivations several time and required some work-arounds like using writeTextFile instead of toFile (which was really confusing).
<gchristensen>
oh wow, FRidh! that is amazing!
phanimahesh has joined #nixos
pstn has joined #nixos
<gchristensen>
FRidh: is it easy to setup new experiments?
<FRidh>
gchristensen: Much easier now. I have one json config file where I describe my test config, and another file with specific settigns for the simulations
<FRidh>
with the first it samples some of the measurement data and then runs simulations based on that
rmrfroot has quit [(Ping timeout: 255 seconds)]
<gchristensen>
vozz: I _think_ you could do it with `with (pkgs.makeRustPlatform pkgs.rustNightly);` instead of `with pkgs.rustPlatform;`.
jgertm has joined #nixos
<FRidh>
Now I don't have to care anymore about where I put temporary files and such.
<gchristensen>
that is fantatsic
phanimah1sh has quit [(Ping timeout: 272 seconds)]
ebzzry has quit [(Ping timeout: 240 seconds)]
<gchristensen>
I tried for weeks to package https://github.com/kaldi-asr/kaldi and use it with nix, but it is done with thousands of bash scripts
<gchristensen>
and the expected use is you put exit 0 in the middle and when the first half runs okay you comment it out and remove the exit 0
<FRidh>
do you mean the packaging needs to consider these bash scripts, or when you actually want to use it you need these scripts?
<FRidh>
ohh yes, I see the tools folder now
pstn has quit [(Ping timeout: 240 seconds)]
<mrkgnao>
FRidh: do you do something akin to scientific computing or data science?
<FRidh>
mrkgnao: I simulate audible aircraft sound based on sound recordings of aircraft and other data.
<mrkgnao>
oh.
<mrkgnao>
wow.
<mrkgnao>
that is amazing.
petar has quit [(Remote host closed the connection)]
<mrkgnao>
so, wait. does that mean something along the lines of converting images of aircraft to sound based on predicted make, distance from the POV, weather, etc?
petar has joined #nixos
proteusguy has joined #nixos
<FRidh>
mrkgnao: No, what I do is I have recordings of aircraft sound, I 'backpropagate' the sound to nearby the aircraft, as if you're flying along with the aircraft. Then I analyze its spectral content and such and store That. Then I simulate the emission, combine it with other models, and propagate the sound to the receiver, e.g. you walking outside
Wizek has joined #nixos
<mrkgnao>
so you predict the position from the sound based on something like an NN trained from recordings?
<FRidh>
Along with sound recordings we have also the position of aircraft based on several sources, including some 'cockpit' data.
<vozz>
gchristensen: Hmm that gives me `cannot coerce a set to a string`
<mrkgnao>
interesting. I'm assuming there are applications that people care about? (not sure if that came out right)
<gchristensen>
vozz: :/ I don't know, then
<vozz>
okay, thanks anyway :P
<gchristensen>
mrkgnao: perhaps a better way to ask that is "what is the application of this research?" :P
<FRidh>
mrkgnao: haha, yes, its not entirely just to keep me off the street. The reason we're looking at this is to study the impact of aircraft sound/noise on humans, e.g. sleep disturbance, annoyance, and so on.
<gchristensen>
vozz: no, that should work, can yo paste your code / error?
<vozz>
error: cannot coerce a function to a string, at /home/ollie/nixpkgs/alacritty/default.nix:7:3
<gchristensen>
vozz: I'm trying to build :)
<gchristensen>
I have to download a lot of stuff though for rustnightly
<vozz>
cool, thank you for the help :)
mudri has quit [(Ping timeout: 240 seconds)]
<FRidh>
gchristensen: that egs directory contains examples. Is it is a problem if those aren't right away usable? Although I suppose using such shell scripts is the recommended way to use kaldi
<gchristensen>
exactly
<gchristensen>
vozz: hrm. are you on 16.09? unstable? nixos or nixpkgs?
petar has quit [(Remote host closed the connection)]
Lowl3v3l has quit [(Ping timeout: 248 seconds)]
<vozz>
gchristensen: I'm on nixos unstable
<gchristensen>
ok
<vozz>
perhaps I've not been updating properly
<gchristensen>
just asking
<gchristensen>
how are you updating?
<vozz>
nixos-rebuild switch --upgrade
petar has joined #nixos
<gchristensen>
that should do
<gchristensen>
brr... -14C here
<vozz>
wow, where you at?
<pie_>
gchristensen, well i know what you shouldnt
Lowl3v3l has joined #nixos
<pie_>
dont freeze,we will be sad :C
<pie_>
vozz, some dude was talking about how its -30C, how does that even exist
<vozz>
didnt know antartica had internet connection
<ToxicFrog>
pie_: -30 with another 10-15 degrees of windchill is pretty typical midwinter weather here
pi3r has quit [(Ping timeout: 245 seconds)]
<ToxicFrog>
Right now it's a comparatively toasty -12
* pie_
flails in horror
<FRidh>
gchristensen: where are you situated?
<pie_>
ToxicFrog, you people live in space
<pie_>
well at least space doesnt have windchill
<vozz>
cant help but wonder what your heating bills are like
<pie_>
well id hope they just have good insulation
<gchristensen>
vozz, FRidh: western Massachusetts, USA :) now I have to take the dog out, so, wish me luck.
<vozz>
you're closer to the equator than me, climate is a weird thing
<MichaelRaskin>
vozz: you may be closer to a large water mass, though
<ToxicFrog>
vozz: not as bad as you might think, between good insulation and gas heating
<MichaelRaskin>
Although not by that much, I guess
<ToxicFrog>
Of course, it can get up to +40° in the summer, so we also get slammed with AC bills
<ToxicFrog>
"temperate climate" my ass
<clever>
living room: 17.56c(63.61f), outdoor: -12.25c(9.95f), server: 21.38c(70.47f)
Wizek_ has joined #nixos
<ToxicFrog>
Yeah, that sounds pretty similar to what it is here
<ToxicFrog>
(and on a day-to-day basis the temperature can easily swing 20° in either direction)
[0x4A6F] has quit [(Ping timeout: 240 seconds)]
<MichaelRaskin>
Sounds quite continental.
MoreTea has joined #nixos
<MoreTea>
hi :)
mudri has joined #nixos
<vozz>
hey MoreTea :)
<MoreTea>
I'm trying to make httpd listen ONLY to port 82.
<MoreTea>
No luck so far
<NixOS_GitHub>
[nixpkgs] 7c6f434c pushed 1 new commit to master: https://git.io/vMCpw
<NixOS_GitHub>
nixpkgs/master 9653be4 Michael Raskin: firejail: 0.9.44.2 -> 0.9.44.4
<MoreTea>
(context: I want to run tinytinyrss over ssl, behind HTTP basic auth)
[0x4A6F] has joined #nixos
<MichaelRaskin>
Let me guess: you set a port list by just assigning it without mkForce?
matthewbauer has joined #nixos
<MoreTea>
ah, yes
<MichaelRaskin>
And the merger logic concatenates all the assignments
<MoreTea>
I figured that I should set `listen = { port = 82; };` in both services.httpd and services.httpd.virtualHost.xxx
<gchristensen>
ToxicFrog: CA?
<NixOS_GitHub>
[nixpkgs] 7c6f434c pushed 1 new commit to release-16.09: https://git.io/vMCp6
<NixOS_GitHub>
nixpkgs/release-16.09 145d33c Michael Raskin: firejail: 0.9.44.2 -> 0.9.44.4...
<MoreTea>
Maybe I should just make a PR to enable setting extra vhost options for ttrss
<ToxicFrog>
gchristensen: yeah. Ontario.
<MoreTea>
Might be easier than working around with these hacks, hehe
<gchristensen>
:)
<clever>
ToxicFrog: and i'm over in NB
rly has quit [(Ping timeout: 258 seconds)]
petar has quit [(Remote host closed the connection)]
jgertm has quit [(Ping timeout: 240 seconds)]
<gchristensen>
vozz: I'm not going to be able to help you for basically ever, because I have to compile multiple rusts.
<gchristensen>
vozz: but! it evaluates for me, so I don't know why you have an error
<vozz>
That's odd... my rust compiled pretty quickly, hadn't used it before
<vozz>
that is bizarre
petar has joined #nixos
<vozz>
what command did you use to build it?
matthewbauer has quit [(Ping timeout: 240 seconds)]
<gchristensen>
nix-build ./alacritty.nix
<vozz>
hm same :/
<MichaelRaskin>
gchristensen: is it reasonable to tag NixOS/security-notifications if I have committed an update that will be mentioned in the roundup tomorrow?
<gchristensen>
MichaelRaskin: in a commit?
<MichaelRaskin>
In a comment
<MichaelRaskin>
To the commit
<gchristensen>
that would be fine, sure
<vozz>
Ah I had a package override in my .nixpkgs/config.nix that was interfering...
<vozz>
seems to be building now! thanks :)
<gchristensen>
ah! perfect!
<gchristensen>
vozz: will you contribute alacritty upstream?
Itkovian has joined #nixos
<vozz>
I'm gonna try, you have to test it in a lot of ways don't you?
matthewbauer has joined #nixos
<gchristensen>
before contributing to nixpkgs?
<vozz>
Yeah, I tried contributing something a bit ago and it got rejected, can't remember what I did wrong
<gchristensen>
it is pretty easy to contribute to nixpkgs I think ...
<MichaelRaskin>
Well, if you don't do any magic w.r.t. to upstream _and_ it launches fine and it is justan addition — that is actually OK
<gchristensen>
vozz: tag me (@grahamc) when you send the PR
<vozz>
Will do :)
<MichaelRaskin>
I wonder how to build librsvg correctly — it has _some_ Rust and some Rust dependencies, but it is not majority-Rust
<sphalerite>
Can I install adb from androidsdk without needing the huge android system images?
matthewbauer has quit [(Ping timeout: 248 seconds)]
<sphalerite>
ah, got it. androidEnv.platformTools
<vozz>
I think I was doing several python packages or something, and maybe I had to split them into several pull requests or something, i really cant remember
markus1 has quit [(Ping timeout: 265 seconds)]
<vozz>
Ive contributed stuff before that, but that was ages ago and it seemed easier then :P
<MoreTea>
does anyone run ttrss with the setup provided in nixpkgs? nginx + phpfpm?
<gchristensen>
ahh ... yeah, we are a bit picky about having one package per commit, and preferably one package per PR
systemfault has joined #nixos
ThatDocsLady has joined #nixos
jgertm has joined #nixos
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<LnL>
what... my vim setup suddenly does syntax checking for nix
<gchristensen>
~magic~!
<domenkozar>
hydra-evaluator.cc: In member function 'void Evaluator::readJobsets()':
<domenkozar>
hydra-evaluator.cc:64:39: error: 'Evaluator::Jobsets {aka class std::map<std::pair<std::__cxx11::basic_string<char>, std::__cxx11::basic_string<char> >, Evaluator::Jobset>}' has no member named 'try_emplace'
<domenkozar>
auto res = state->jobsets.try_emplace(name, Jobset{name});
<domenkozar>
anyone seen this?
<domenkozar>
latests hydra + nixpkgs master
sdothum has joined #nixos
<NixOS_GitHub>
[nixpkgs] vcunat pushed 1 new commit to staging: https://git.io/vMChM
<NixOS_GitHub>
nixpkgs/staging 3a3b026 Vladimír Čunát: subsurface: disable parallel make, as it fails on Hydra...
Itkovian has joined #nixos
<gchristensen>
domenkozar: I can try reproducing if you'd like.
<domenkozar>
nah, need to figure out what's missing
<gchristensen>
ok
truh has joined #nixos
<gchristensen>
there is a very good docs PR to hydra, btw, and it would be a shame to allow it to languish
<gchristensen>
don't know, much less positive this way.
<MichaelRaskin>
gchristensen: I would offer you to punt such PRs to me, but some of them are full-NixOS-related
<gchristensen>
yeah
<gchristensen>
I could just block the one URL ...
<gchristensen>
or cover that part of the screen :)
<MichaelRaskin>
Hm, I actually use Squid URL rewriting, although not for GitHub avatars
<MichaelRaskin>
If you want any assistance setting that up…
<gchristensen>
domenkozar: can you restart all-failed on 16.09?
seagreen has joined #nixos
phreedom has joined #nixos
pi3r has joined #nixos
<MichaelRaskin>
gchristensen: also, if NixPkgs were on GitLab, on some zoom level tha avatars are hidden because of mobile-optimized CSS kicking in
<gchristensen>
MichaelRaskin: interesting! I'll think about that squid option
<gchristensen>
interesting denial of service attack though
[0x4A6F] has quit [(Ping timeout: 245 seconds)]
<MichaelRaskin>
Survival of the fittest.
matthewbauer has joined #nixos
[0x4A6F] has joined #nixos
hotfuzz_ is now known as hotfuzz
<MichaelRaskin>
You just don't have enough tolerance towards those unlike yourself!
matthewbauer has quit [(Ping timeout: 248 seconds)]
<gchristensen>
;)
jgertm has quit [(Ping timeout: 240 seconds)]
<hodapp>
gah. I tried to enable Avahi on my laptop last night, and then my desktop. Both times, nixos-rebuild switch complained that it couldn't start avahi.
<hodapp>
trying to find the exact error; I tried restarting dbus after one page recommended that, which just killed my entire X server, so I lost whatever I had up
<gchristensen>
domenkozar: I'm hopefully open sourcing some test tooling that I built for some work projects using nix / make-test.nix, which does, I think, very interesting things :)
m3tti has quit [(Remote host closed the connection)]
<gchristensen>
similar to your snabb work, demonstrating business cases for nix
<exi>
nix-shell -A rustNightly.cargo pulls in rustc-nightly and rustc-beta
<zarel>
hi folks
<exi>
cannot figure out why because i already have rustc-nightly in my store
<MichaelRaskin>
Maybe a different hash?
<exi>
and it seems absurd that i have to compile 2 versions of rustc just to have cargo work
<gchristensen>
exi: building rustc-nightly requires rustc-beta, and rustc-nightly will need to rebuild if anything else in its dependencies has been rebuilt
<exi>
gchristensen: but i did not change anything (yet) and build rustc-nightly already
<MichaelRaskin>
-A sets up an environment to _build_ cargo, not to use it
<gchristensen>
exi: did you update nixpkgs?
<exi>
not since starting this
<hodapp>
oh dear, this poor Raspberry Pi is going to catch on fire before it finishes printing this page
<gchristensen>
hodapp: that is funny! :D
<exi>
and nix-shell -p rustNightly.rustc does not rebuild anything
<hodapp>
I could have driven to Staples in this time and used their printers
<MichaelRaskin>
Thermoprinting gets a new meaning
<gchristensen>
well, MichaelRaskin has clearly explained the difference, exi
pstn has joined #nixos
<hodapp>
maaaaaybe I should just find a shitbox someplace and put NixOS on that instead
<exi>
gchristensen: yes, that helped. i'm still not sure why it pulls in rustc-beta because afaik rustNightly.cargo depends on rustNightly.rustc which is already built
<exi>
is there any way to get debug output that tells my why a particular dependency got pulled in?
<MichaelRaskin>
You could nix-instantiate, and then nix-store -q --graph
<MichaelRaskin>
dot would be able to convert it to a reasonable svg
alx741_ has joined #nixos
<exi>
on it, just takes a while to layout
<MichaelRaskin>
Yes, large graphs + dot = large waits
<MichaelRaskin>
And then large svg + browser = another wait
alx741 has quit [(Ping timeout: 240 seconds)]
<exi>
ok, according to the graph, rustNightly.cargo has a direct edge to rustc-beta
<exi>
i would like to know which part of the expression pulls that in
<exi>
ah, sorry
<exi>
i was wrong
Lowl3v3l has quit [(Ping timeout: 240 seconds)]
<exi>
cargo-night-<hash>-fetch.drv depends on rustc-beta
<exi>
that seems like an error in the expression, that should also depend on rustc-nightly
<exi>
to not pull in both
<MichaelRaskin>
Well, drvs are actually reasonably readable
<MichaelRaskin>
You could open the drv and try figuring out where it mentions -beta
<exi>
i'll see if i can change this to use the right rustc version
bennofs1 has joined #nixos
<exi>
MichaelRaskin: graphing helped a lot already, thank you
<exi>
i'll report back if i get stuck again
<MichaelRaskin>
indent on drv's also sometimes gives something useful
<MichaelRaskin>
We-ell
<gchristensen>
I've never looked at a drv very close... neat
<MichaelRaskin>
rustNightly needs a full rustPlatform, so it uses beta platform, and cargo gets built with beta platform and cargo-fetch as a part of it
<MichaelRaskin>
gchristensen: there are multiple reasons why I hope removal of drvs doesn't takke off
<MichaelRaskin>
They are just incredibly convenient for tracing and debugging issues
<gchristensen>
oh is that a goal?
<gchristensen>
of someone?
<MichaelRaskin>
Well, writing a lot of drvs to disk does take time
<gchristensen>
yeah
<MichaelRaskin>
(writing the build ouputs takes more, of course)
<gchristensen>
practically speaking, what is the % of time writing drvs vs. ... I dunno, nix not being really optimized? or like ... compiling?
mrkgnao has quit [(Quit: WeeChat 1.5)]
<MichaelRaskin>
In GSoC thread someone said something about describing the drv removal in a GSoC-fitting way
<exi>
MichaelRaskin: yes, but if cargo-nightly already depends on rustc-nightly, why not use rustc-nightly for the fetch instread of the beta?
<exi>
in my case, thats just annoying because i already have rustc-nightly build and just because of the cargo-nightly-fetch it now wants to build rustc-beta as well which takes hours
<MichaelRaskin>
cargo uses rust, but then it wants cargo-fetch, which it gets from the platform, as far as I get it
<MichaelRaskin>
It's the same problem as «I want gcc to use my glibc and glibc be built with my gcc», but with cargo nightly it doesn't get the same amount of workaround-building efforts
<exi>
MichaelRaskin: is not cargo build on top of rust/with rustc?
<exi>
i don't need cargo for rustc to work
<MichaelRaskin>
Yes, but it is a library with dependencies
ryanartecona has joined #nixos
<exi>
seems weird to use 2 different sets of compilers for stuff that is already build on top of a compiler
[0x4A6F] has quit [(Ping timeout: 240 seconds)]
<MichaelRaskin>
I think cargo from beta has the same relationship with cargo-fetch from stable
<exi>
then how do we get carg-stable-fetch to work? it has to depend on rustc-stable in that case?
<exi>
cargo-stable-fetch*
[0x4A6F] has joined #nixos
<MichaelRaskin>
I think there is some bootstrapping effort spent on the bottom level that noone wnts to spent on higher levels
zarel has quit [(Ping timeout: 240 seconds)]
<exi>
uaahh
<gchristensen>
yes
<exi>
cargo-fetch depends on cargo
<exi>
wut
<gchristensen>
to be frank, nobody wants to spend that effort at the bottom level either, but ...
<exi>
what is this black magic^^
<MichaelRaskin>
Because it is a library with dependencies and it is easy to fetch dependencies using cargo!
<gchristensen>
also -- this will likely improve as more things move to rust
<gchristensen>
exi: this is what happens when you need cargo + rust to build cargo + rust
<MichaelRaskin>
I warned you with a case of glibc and gcc but you wanted to go deeper, now pay the price.
<MichaelRaskin>
As usual, 1 GiB of disk space per sanity point is acceptable
<MichaelRaskin>
gchristensen: so far things get more complicated as something moves to Rust, as now there are mixed-packages not fully fitting buildRustPackage
<gchristensen>
yeah, I saw that with libsrvg
<gchristensen>
rsvg? whichever
<MichaelRaskin>
librsvg
<gchristensen>
we'll have to figure something out
<MichaelRaskin>
How do I just build a package out of rustPlatform
rmrfroot has joined #nixos
pi3r has quit [(Ping timeout: 245 seconds)]
<MichaelRaskin>
Not buildRustPackage, build something as if it were a dependency of a buildRustPackage expression?
rod has quit [(Ping timeout: 240 seconds)]
<MichaelRaskin>
exi: it's turtles all the way down, actually: stable cargo depends on bootstrap rustPlatform
<MichaelRaskin>
And bootstrap is just a binary release
matthewbauer has joined #nixos
markus1 has joined #nixos
polyrod has joined #nixos
pi3r has joined #nixos
zarel has joined #nixos
matthewbauer has quit [(Ping timeout: 248 seconds)]
cjhowe has joined #nixos
sitedyno has quit [(Ping timeout: 246 seconds)]
<vozz>
gchristensen: something finally happened in my build: make: target/snapshot/bin/cargo: Command not found
uralbash has joined #nixos
<exi>
MichaelRaskin: could we not just let rustNightly use the rustStable rustPlatforM?
<exi>
at least we have binary packages for this
<gchristensen>
vozz: just checking, I got: "curl: (6) Couldn't resolve host 'static.rust-lang.org'" :|
<vozz>
nice
Lowl3v3l has joined #nixos
<MichaelRaskin>
Maybe the failure is inside sandbox?
sellout-1 has joined #nixos
<gchristensen>
mine is due to that for sure
<MichaelRaskin>
exi: may be that stable platform os too old.
sellout- has quit [(Ping timeout: 255 seconds)]
<exi>
i'll try to build it like this
<exi>
i started this yak shaving spree because rustNightly.cargo does not build...
<gchristensen>
vozz: ^ :|
<exi>
maybe after this i can confirm that rustStable works as rustPlatform base and send a PR to fix the carbo build errors
<exi>
cargo*
<exi>
or not, depending on how it works out
<vozz>
oh joy :(
<MichaelRaskin>
We should choose an official yak-like sound for such occasions.
<MichaelRaskin>
Rn, maybe
<exi>
ok, to be honest, i started this yak shaving spree because i wanted to package this: https://github.com/jwilm/alacritty which required rust nightly :D
<vozz>
exi: me too!
<exi>
ohi! :D
<gchristensen>
I love that you two are packaging that
<gchristensen>
because I want to try it :)
<vozz>
i like that someone else is packaging it because i'm really stuck now :)
<exi>
on a tanget: why does hydra not build rust nightly for me? :P
<exi>
tangent*
<MichaelRaskin>
Maybe it also failed?
<exi>
could not find it in any jobset but maybe i overlooked it
<MichaelRaskin>
Or maybe there is no recurseIntoAttrs at the correct place
<exi>
would it be acceptable to make alacritty depend on rustNightlyBin ?
<exi>
so i don't have to shake the cargo yak
<exi>
shave*
<MichaelRaskin>
It's defeatism, but maybe it's OK in this case. The yak stands a bit shaken.
sitedyno has joined #nixos
<exi>
on the lowest level we depend on rust binary releases anyway for bootstrapping
<exi>
why build it at all? :D
<exi>
maybe just because it's winter and i need some heating in my apartment:)
<gchristensen>
well you could, but that probably wouldn't be merged to upstream
<gchristensen>
pi3r: "How many different binaries of the same package can coexist?" I reckon the limit is the 160-bit hash of the nix store
[0x4A6F] has quit [(Ping timeout: 240 seconds)]
<gchristensen>
pi3r: "Is there a timeline to remove old binaries from the servers and replace with newer?" right now, I don't think the binary cache does any pruning of old builds
<exi>
error: undefined variable ‘rustNightlyBin’ at (string):1:90
<exi>
ok now it's just being mean
[0x4A6F] has joined #nixos
<gchristensen>
pi3r: I just checked out 6f1b2e7de2e631cddeb446736dd72771246629ef which was committed "Mon Jul 27 13:28:34 2015 -0300" and built vim, and was able to get it entirely from binary cache -- no compiling at all.
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
cjhowe has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<pi3r>
gchrstensen: that's impressive (and rather surprising for me)
<pi3r>
gchrstensen: thanks for the search
<gchristensen>
one second pi3r
<gchristensen>
pi3r: the oldest vim I can get is 2015-01-20 00:23
<gchristensen>
git checkout `git rev-list -n 1 --before="2015-01-20 00:23" master`; nix-build . -A vim
jgertm has joined #nixos
<exi>
an idea why i can nix-build -A rustNightlyBin.rustc
<exi>
but i cannot nix-shell --pure -p rustNightlyBin.rustc
<exi>
first one works as expected, second one goes error: undefined variable ‘rustNightlyBin’ at (string):1:90
<pi3r>
gchristensen: I have just linked the reddit question to the current IRC thread. Thx again.
<gchristensen>
:)
<gchristensen>
MichaelRaskin: is there a way to see if `nix-build . -p vim` would be able to get it without building at all? -j0 still fetches
<MichaelRaskin>
nix-store has --dry-run
<MichaelRaskin>
But I think Hydra's cache is correct
<MichaelRaskin>
So you only need to check the result's availability by fetching narinfo
<MichaelRaskin>
That implies that all the dependencies are there
<gchristensen>
cool
cjhowe has joined #nixos
<NixOS_GitHub>
[nixpkgs] the-kenny pushed 2 new commits to master: https://git.io/vMWkO
<NixOS_GitHub>
nixpkgs/master 2a2e8e9 Moritz Ulrich: rustRegistry: 2016-12-28 -> 2017-01-08
<NixOS_GitHub>
nixpkgs/master 7435fef Moritz Ulrich: rustc: Disable fragile tcp tests on Darwin...
uralbash has quit [(Ping timeout: 258 seconds)]
truh has quit [(Ping timeout: 240 seconds)]
<rmrfroot>
gchristensen: do you have chromium installed and want to see something cool?
<gchristensen>
rmrfroot: I have google-chrome installed, probably chromium, and would like to see something cool.
<mbrgm>
hi! how do I escape double single quotes inside double single quotes?
<polyrod>
mbrgm, '''
<mbrgm>
polyrod: th
<mbrgm>
polyrod: thx*
matthewbauer has joined #nixos
m0rphism has quit [(Quit: WeeChat 1.4)]
<gchristensen>
rmrfroot: you have me at the edge of my seat in excitement!
<rmrfroot>
gchristensen: nice! i have been putting together that API spec i was talking about earlier. using Swagger UI you can interactively try out the API in your browser. but since the Swagger UI is not hosted on hydra.nixos.org you will not be allowed to use it due to CORS.
<rmrfroot>
gchristensen: why i asked about chromium/chrome is because you can start an instance with "--disable-web-security" to try it out
<rmrfroot>
it should then show you two categories where you can try out the endpoints
tvon has joined #nixos
matthewbauer has quit [(Ping timeout: 245 seconds)]
<gchristensen>
hrmm the first one is producing no content
<rmrfroot>
gchristensen: they will all do that if CORS is active
<gchristensen>
ahh
<gchristensen>
"Web security may only be disabled if '--user-data-dir' is also specified."
<rmrfroot>
gchristensen: ah, true! forgot about that one
tvon has quit [(Ping timeout: 260 seconds)]
<rmrfroot>
gchristensen: when web security is actually disabled you should see a yellow tooltip in the browser that warns you about it
<gchristensen>
this is amazing!
<flxw>
Hi all. With pkgs.lib.overrideDerivation I'm able to override a package derivation. How would I do the same for a configuration module, e.g. nixos.modules.programs.bash?
<rmrfroot>
gchristensen: using swagger-codegen you can also generate client code for languages like java, javascript, haskell, scala, c++, perl, python, etc etc.
<the-kenny>
Ugh, I hate it when people don't accept any feedback on their PRs, then complain that the bug isn't fixed in <24h :/
glines has joined #nixos
<gchristensen>
the-kenny: :/ which one?
<gchristensen>
rmrfroot: incredible, where do we take this?
<Sonarpulse>
any ideas on what I can do to make it happen?
<Sonarpulse>
now that I am using nixUnstable on nixos
<Sonarpulse>
I worry with schema change I cannot go back
<Sonarpulse>
:/
<rmrfroot>
gchristensen: the easiest way to get value out of this would be to deploy that Swagger UI as a page in the docs. it's basically just a static web page with some javascript that loads the YAML specification.
<rmrfroot>
gchristensen: or we could just supply a link to the file in the docs, and then people who find it useful can use it.
<MichaelRaskin>
the-kenny: sometimes I also prefer sed
pstn has quit [(Ping timeout: 240 seconds)]
<the-kenny>
MichaelRaskin: Hm, maybe I should've explained my intentions more cleanly at the beginning. My biggest complain was that this sed-call will silently fail as soon as the function names change
<MichaelRaskin>
That's true, but in my experience that is rarely actually a problem
<MichaelRaskin>
Especially when we want to kill a fragile test: if it changes the name, there is a good chance it also changes its behaviour
<the-kenny>
but then we should remove the sed line from our build script as well
<gchristensen>
the-kenny: perhaps a bit too critical. acowley has a nice history of contributions: https://github.com/NixOS/nixpkgs/pulls?page=3&q=is%3Apr+author%3Aacowley+is%3Aclosed and while the sed is a bit ugly, I'm not sure I would constitute it a make-it-or-break-it issue on merging the patch. lastly, darwin is a bit of a labor of love right now
<MichaelRaskin>
Which is not time-critical, especially with rust that merits a few complete overhauls a year anyway
<the-kenny>
gchristensen: ah well, I usually don't look at someones history
<gchristensen>
the-kenny: I don't either, I've just seen a number of their contribution / efforts to make nix better on darwin
<gchristensen>
my main point about their history of contributions is re not being paid to work on nixos. they aren't either, and put in a good bit of effort too
<gchristensen>
the-kenny: I don't think you did anything wrong, both of you are trying to make nixpkgs a better place
<vozz>
How's it going exi? Get anywhere? :D
zraexy has joined #nixos
<gchristensen>
rmrfroot: docs are on a different domain, no?
<rmrfroot>
gchristensen: that's true! you could allow usage of hydra.nixos.org from nixos.org, though.
<gchristensen>
rmrfroot: would it be weird to deploy it in the webroot of hydra itself?
<gchristensen>
ie: distribute it with hydra
<rmrfroot>
gchristensen: i think that would be a good idea, actually. that way we know it won't be any version mismatch between the docs and hydra itself.
<gchristensen>
and other deployments get it automatically
<rmrfroot>
also if people install their own hydra they can use it to test *their* install
<gchristensen>
yeah
<rmrfroot>
:D
<NixOS_GitHub>
[nixpkgs] the-kenny closed pull request #21741: rustc: disable fragile tests on darwin (master...rustc-1.14-darwin) https://git.io/vMCWf
bennofs1 has quit [(Ping timeout: 240 seconds)]
reinzelmann has joined #nixos
<gchristensen>
hrmm... neat... hydra is hard to get PRs merged for... but this is really amazing., and I think it is worth a try
rly has quit [(Ping timeout: 240 seconds)]
<FRidh>
Sonarpulse: you can actually nix-build with nixUnstable?
cheshircat has quit [(Quit: WeeChat 1.5)]
cheshircat has joined #nixos
<rmrfroot>
gchristensen: yeah, i can understand why! :) i'm not trying to rush, though. my first goal is to complete the static docs (authenticate and creating resources is missing).
<Sonarpulse>
FRidh: yeah?
<Sonarpulse>
in most cases
<Sonarpulse>
am I lucky here?!
<LnL>
the-kenny: gchristensen: I think he was mainly talking about darwin maintainers/contributers
<FRidh>
yesterday I upgraded to nixUnstable and got stuck with a system where I couldn't build anything.
<FRidh>
obiously also wasn't able to roll back
<LnL>
the-kenny: gchristensen: I'm basically the only person that actively looks at darwin pull requests
<FRidh>
it could instantiate, but then....nothing
<gchristensen>
LnL: LMK if I can build something for you to give you extra capacity
bennofs1 has joined #nixos
<gchristensen>
rmrfroot: sounds good :) do you have a project for hydra?
<LnL>
gchristensen: what do you mean?
<gchristensen>
LnL: sometimes my limit of how much I can merge is based on my capacity to test builds
<FRidh>
Sonarpulse: since then I've tried also with nixos-rebuild build-vm with different version of nixUnstable and didn't get any working system
<gchristensen>
LnL: I have a darwin machine here, and can run tests for you
cheshircat has quit [(Quit: WeeChat 1.5)]
<LnL>
gchristensen: I'ts mainly more free time that I need, but having some more build capacity available would certainly help for some things
cheshircat has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/vMWLi
<pierron_>
peti: One idea I had, which would be necessary for Nixpkgs, to reduce the memory usage anyway, is to get rid of the immediate application of mkDerivation & callPackage.
johnsonav has joined #nixos
<Sonarpulse>
pierron_: ah
tvon has quit [(Remote host closed the connection)]
<Sonarpulse>
I've been thinking similar things would be good for bootstrapping stages
<Sonarpulse>
or rather your overrideWith gives nicely with this
<pierron_>
peti: instead, of `p = callPackage f { args }` where f is `{a,b}: mkDerivation { recipe }` , we would have something like: `p = importWithNixpkgs f`, where f is something like `{ args = { args }; derivation = stdenv.mkDerivation; recipe = { a, b }@args: { buildInputs = []; … }; }`
cjhowe has joined #nixos
tvon has joined #nixos
<pierron_>
Sonarpulse: I tihnk the update operator is enough, in all cases.
<pierron_>
Sonarpulse: the reason why we have override and overrideDerivation, is to undo the function application.
<pierron_>
Sonarpulse: but doing that keeps the function content alive.
<LnL>
gchristensen: "know" about bootstrapping?
<Sonarpulse>
pierron_: you mean __functor__ ?
<Sonarpulse>
what "update operator?
<gchristensen>
LnL: well ... I have no idea how it works, heh
<pierron_>
Sonarpulse: update operator `//`
pi3r has quit [(Ping timeout: 255 seconds)]
<gchristensen>
LnL: so I'd like got get gpgv in there and then I'm stuck, because I have no idea where to start
pierron_ is now known as pierron
ixxie has joined #nixos
<Sonarpulse>
pierron: oh ok
<Sonarpulse>
pierron: did you see my extendStage?
<LnL>
gchristensen: no idea how different the linux bootstrapping is from that of darwin
<pierron>
Sonarpulse: yes, I gave a brief look, I would have to dig futher, good to see the removal of overridePackages
<pierron>
Sonarpulse: I really like the list of stages, but I wonder why making a special case for the base of packages?
<Sonarpulse>
and are you ok with me trying to get the normalize PR in first as is?
<pierron>
Sonarpulse: couldn't the base of all package just be empty?
<Sonarpulse>
pierron: I vacillated on that
<Sonarpulse>
I had on idea that making it always bootstrap tools would be slick
<Sonarpulse>
but yeah empty list might be better
<pierron>
Sonarpulse: it makes sense, you bootstrap from the void :P
<Sonarpulse>
glad you like the list though :)
<Sonarpulse>
yeah I'd be happy to make it "the void"
<Sonarpulse>
another thing I've been thinking
<Sonarpulse>
with cross its dirt simple where previous stages becomes build packages
<Sonarpulse>
with linux and darwin
<Sonarpulse>
there seems to be two patterns
<Sonarpulse>
either "I want all the same package, but this new bootstrapped package"
<pierron>
Sonarpulse: there is another thing which makes me wondering. You added the bootstrap stage early in the pipeline, but from where is the `super` attribute coming form?
<LnL>
gchristensen: looks pretty similar
<Sonarpulse>
or "I want all new bootstrapped packages, but with these packages preserved"
<Sonarpulse>
pierron: which super?
<Sonarpulse>
so many self supers lol
c0bw3b has quit [(K-Lined)]
<pierron>
Sonarpulse: the one use to find the recipe of packages during the bootstrap
danharaj has joined #nixos
mizu_no_oto has joined #nixos
<pierron>
Sonarpulse: I think we should rename self to dependencies, and super to recipes. This would be so much nicer.
<Sonarpulse>
pierron: yeah I'm tempted
<Sonarpulse>
but I'd want to stage that with haskellPackages
<Sonarpulse>
get those firmly in sync
<Sonarpulse>
if so, definitely!
<Sonarpulse>
pierron: you mean stdenvOVerrides?
<Sonarpulse>
I am looking at PR for "super"s
<pierron>
Sonarpulse: also, for the security-update branch, I will either have to split dependencies in to static & dynamic, and find a way to let callPackages figure out the right one :/
<pierron>
the PR for super?
<Sonarpulse>
that is build/run right?!
<Sonarpulse>
static/dynamic?
<pierron>
no
<Sonarpulse>
or you mean linkage?
danharaj has quit [(Read error: Connection reset by peer)]
<pierron>
static/dynamic are the linking methods for dependencies.
<pierron>
if you link statically, you have to rebuild, if you link dynamically, you have to patch it, in case of a dependency update.
<Sonarpulse>
ok
<Sonarpulse>
pierron: did you look at my comments in the runtime-only deps nix issue?
<Sonarpulse>
so here's the thing
<Sonarpulse>
in my cross PR
<Sonarpulse>
I already do the callPackage slicing
<Sonarpulse>
for build/native vs run/foreign
<LnL>
gchristensen: not sure what you mean by getting it into the stdenv
<Sonarpulse>
but the hypothetical runtime-only deps would only be used for dynamic linking
<pierron>
I did, but I did not understood what you meant
<Sonarpulse>
ok, so basic operational idea with security updates
<Sonarpulse>
is sedding paths right?
[0x4A6F] has joined #nixos
<Sonarpulse>
if the dep is accessed runtime we can sed the build, and it will access the other thing at runtime
<pierron>
currently the security-update branch relies on the callPackage function, which is not correct, but a good enough approximation
<Sonarpulse>
and this is most correct when the path isn't used at build time
<Sonarpulse>
the sedding is most correct that is
<Sonarpulse>
because then it's completely forgetting about the old package, right?
<pierron>
Sonarpulse: then once with have a zip-able list of dependencies, which does not depend on any conditional, then we can construct the patches ahead of time.
<Sonarpulse>
that sounds similar
<Sonarpulse>
my idea with run-time only
<Sonarpulse>
is enforce that
<Sonarpulse>
by passing in dummy paths at build time
<Sonarpulse>
you would get headers but *not* shared objects at build time
<Sonarpulse>
and then, security update or no security update, the build would be sedded for the runtime dep
<Sonarpulse>
so if libc, but not libc's headers change
<Sonarpulse>
the only rebuild is a bunch of sedding
ambro718 has quit [(Ping timeout: 240 seconds)]
<Sonarpulse>
and that falls naturally out of the rules
<Sonarpulse>
pierron: make sense?
<pierron>
Sonarpulse: the problem with content of the packages, is that one could write a Nix expression for a package, which only includes a dependency if the version if x.y.z, which would lead to different list sizes, which are not zip-able
<LnL>
gchristensen: oh, you're idea is to have gpgv -> fetchurlBoot -> fetchUrl?
<LnL>
your
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
<LnL>
that's probably easier, afterwards you can see what stuff can't use it yet
[0x4A6F] has quit [(Ping timeout: 240 seconds)]
[0x4A6F] has joined #nixos
<exi>
hmm, my rust foo is very weak
<exi>
attribute is only usable with crates of the `proc-macro` crate type
<exi>
anyone any idea what that means?
<exi>
error: the `#[proc_macro_derive]` attribute is only usable with crates of the `proc-macro` crate type
digitalmentat has joined #nixos
jgertm has quit [(Ping timeout: 240 seconds)]
zraexy has quit [(Ping timeout: 240 seconds)]
<Sonarpulse>
exi: I'd guess
<Sonarpulse>
that your cargo is too old
<Sonarpulse>
and some crates.io metadata is not being routed right
<Sonarpulse>
curious what this is about as proc macro stuff is unstable
<Sonarpulse>
and I didn't think it was used in cargo or rustc yet
<digitalmentat>
I'm trying to debug why a certain package of our Nix system is not being included in the binary cache of our system (even though it gets built!)
roconnor__ has quit [(Ping timeout: 260 seconds)]
<digitalmentat>
what generally causes a package's build product to not be included in a closure?
<exi>
Sonarpulse: thats part of a dependency of a thing i'm trying to package
<exi>
the thing required rust nightly
<Sonarpulse>
exi: oh this isn't buildRustPackage for cargo?
<Sonarpulse>
or cargo does use it transitively?
<exi>
this is using buildRustPackage, but with a rustPlatform derived from rustNightly
johnsonav has quit [(Ping timeout: 255 seconds)]
Willi_Butz has joined #nixos
<Sonarpulse>
exi: weird
<LnL>
talking about rust, I parallel rust build working :)
<pierron>
Is anybody familliar with vim-plug? I am unable to get it list plugins that I want to install, any idea how to investigate?
<Sonarpulse>
get a new rust nightly then?
<exi>
Sonarpulse: ah, the rustNightly is very old
<LnL>
pierron: I used vim-plug before I moved my configuration to nix
<Sonarpulse>
exi: this isn't Cargo to be clear, because cargo itself should not rely on such a thing
<Sonarpulse>
that would be upstream issue
<pierron>
LnL: I am totally new to vim plugins in general
<Sonarpulse>
they should know about asap
<Sonarpulse>
but yeah you need new rustc nghtly and cargo to get this building
<Sonarpulse>
because proc macros cross-cuts both of those
<pierron>
LnL: I have no idea where to look for one, and vim-plug PlugInstall tells me there is nothing to install.
<pierron>
LnL: even with `Plug 'LnL7/vim-nix' ` listed there
<LnL>
pierron: you have to add stuff you your vimrc
<LnL>
oh
<exi>
Sonarpulse: this yak shaving is going great...
<polyrod>
if i change to ghc784 ... it gives me : error: anonymous function .... called without required argument ‘ghc-boot’
ixxie has quit [(Ping timeout: 248 seconds)]
nanobapt has quit [(Ping timeout: 258 seconds)]
<polyrod>
i am trying to avoid cabal or stack and go only with the nix declarative way
<polyrod>
any ideas ?
<siarheit_>
ghc-7.8 is an old beast
<siarheit_>
i'd suggest applying a patch to allow base-4.9 for 8.0.1
truh has joined #nixos
<polyrod>
siarheit_, thats patching the default.nix from datetime-0.3.1 nix-package ?
<siarheit_>
i am nixos noob. don't know how to do it the nix way. i'd assume you need to add a packageOverrides with your one-line patch . this looks relevant: https://nixos.org/wiki/Nix_Modifying_Packages
<polyrod>
its datetime-0.3.1 thats not building because ghc801 doesnt have base >=4.2 && <4.9
sboosali has joined #nixos
<polyrod>
where do i find the haskellPackages on github , since thex are not under nixos/nixpkgs
<polyrod>
think its cabal file constrain it to have base >=4.2 && <4.9
<polyrod>
and i wantet to look at the default.nix for that package , ithink that was built witch cabal2nix
<polyrod>
but the haskellPackages arn't under the default nix-github-repo , and i need a hint where to find haskellPackages//datetime-0.3.1/default.nix
<flxw>
I'm currently trying to wrap my head a little more about nixos fundamentals. If I understand it correctly, then the derivation is the basic unit for building and installing a package (a derivation expression gets instantiated as a lowlevel derivation into the store, from where the latter gets realised). But what is the basic nix expression for configurations called? I see that even /etc gets populated from a nix-path, but how is th
<flxw>
/nix/store/<hash->-etc collected in the first place?
<flxw>
s/-//
<simpson>
flxw: It's all defined top-down from a Nix expression in /etc/nixos/configuration.nix
<flxw>
simpson: yeah, but who is calling the function defined in /etc/nixos/configuration.nix
<peti>
pierron: So how would one replace a dependency of 'f' in that style? If I have 'f', then I can just mess with the 'args' attribute, but I suppose that 'importWithNixpkgs f' won't give me access to the original 'f' -- or does it?
<simpson>
flxw: It's called interactively by nixos-rebuild.
<peti>
polyrod: What makes you say that Haskell packages are not part of Nixpkgs? Everything lives underneath pkgs/development/haskell, no?
<flxw>
simpson: ah, so do I get this right?: After nixos-rebuild, the whole collection of config files has been put into the store. And (e.g. when the system boots, there is a nixos script called which ensures that /etc gets populated with the right store path)?
<peti>
polyrod: Also, haskell.packages.ghc784.datetime-sb compiles just fine on the current Nixpkgs 'master'. I see no errors about ghc-boot.
<simpson>
flxw: /etc in particular is set up immediately after the paths are built in the store.
<flxw>
aha, thank you both very much. I think I need to take a look into nixos-rebuild now. :)
<simpson>
flxw: What are you actually after?
<LnL>
flxw: so there's a drv for all the files and then afterwards nixos-rebuild runs all the activation scripts, on of which links everything to /etc
<NixOS_GitHub>
[nixpkgs] peti pushed 4 new commits to master: https://git.io/vMWCH
<NixOS_GitHub>
nixpkgs/master 9729a8d Peter Simons: haskell-esqueleto: keep version 2.4.x around so that we can still build git-annex
<NixOS_GitHub>
nixpkgs/master 667c4e4 Peter Simons: hackage-packages.nix: automatic Haskell package set update...
<NixOS_GitHub>
nixpkgs/master dd0d717 Peter Simons: git-annex: fix build
jgertm has joined #nixos
<flxw>
simpson: I'm trying to understand how this great idea of semi-purely functional configuration thingy works. For how nix works, I was able to track it down to derivationStrict, which builds the derivation from the derivation expression. But for the configuration itself, I'm still missing the fundamental steps. (for derivations the two steps are instantiation and building the derivation). So I'm trying to get a better picture of nixo
<flxw>
conceptually.
<flxw>
the two pointers start clearing things up slowly, but surely (or very slowly, that is. haha).
<LnL>
flxw: you might be interested in the nix paper the
<LnL>
or is it just the nixos-rebuild stuff you are looking at?
<simpson>
flxw: Fundamentally, NixOS doesn't do anything special besides symlink certain things from the store into the root filesystem, just like nix-env.
<flxw>
simpson: yes, but this is kind of an understatement I think. ;) Everything in unix is just files, directories, and links. You got compilers, and scripts, they built stuff and install it. ;-))
<LnL>
it's very similar to nixos but it's a lot smaller so it might be easier to naviate
<flxw>
ah, cool.
<flxw>
btw, can I override the configuration modules of nixos in the same way one can overrideDerivation nix packages?
<simpson>
flxw: Well, I don't know what you're reading, and it sounds like you're mostly doing exploratory work. If you're looking for some specific bit of magic, I have no idea where to start. Maybe the manual? http://nixos.org/nixos/manual/
<toogley>
simpson: hm, okay. thanks
<simpson>
flxw: Some configuration fields are catenable, so that "overrides" by specifying the field in multiple modules will actually cause things to be combined instead.
<LnL>
flxw: you can't override modules, but you can use imports = [ ]; to include extra modules
<simpson>
An example is networking.allowedTCPPorts, which is a list of ints. If you specify it in two modules, then the result will be the two lists glued together.
<LnL>
yeah, if you're talking about the configuration then you can use mkForce [ "foo" ];
<LnL>
that will override any previous defined values instead of adding them
Wizek has joined #nixos
<flxw>
okay, a concrete issue: There is nixos/modules/programs/bash/inputrc. In the long run (after I understood enough of it to do it), I'd provide a patch to tune the file in or or two details. But for now, I'd just like to just locally patch the file so that it dows what I want.
Chattered has joined #nixos
<flxw>
bash.nix puts it into /etc by environment.etc."inputrc".source = ...
<flxw>
is there a way to tinker with the file after bash.nix declared it to put it into the store?
<LnL>
if you want to override that, environment.etc."inputrc".source = mkForce (writeText ...); should do what you want
m3tti has joined #nixos
<simpson>
LnL: You beat me to it.
<LnL>
:D
jensbin has quit [(Quit: leaving)]
<flxw>
hmm, yes. this would work, I see now. But if I only need to change one or two lines of it? I mean, I'm qutie happy with 99% of the file.
<m3tti>
hi there how can i contribute packages to nix?
<m3tti>
is it hard to build custom pakcages?
<Chattered>
m3tti: For simple "./configure; make; make install" scripts, it's pretty easy.
<m3tti>
i'm trying to get oolite to work
<simpson>
flxw: You want to only patch parts of it? Nix doesn't make that super-easy.
<m3tti>
it's based on gnustep
<Chattered>
Ohhh...I love oolite.
<Chattered>
That would be cool to have in nixpkgs.
<simpson>
flxw: Maybe you should alter the NixOS module instead.
<peti>
Yo Haskell hackers, http://hydra.nixos.org/jobset/nixpkgs/haskell-updates now a copy of "release-16.09" with GHC 8.0.2 and all other new features from 'master' ported into it. Assuming those builds succeed without revealing any major disaster, I'll push those updates to 16.09 ASAP. Users of "haskellPackages" will notice a few package updates (within LTS 7.x), but I assume none of them will be major enough to
<LnL>
flxw: you'd have to look at whatever made the derivation that's assigned to the source
<peti>
cause trouble for existing setups.
<Chattered>
m3tti: Is it closedsource?
<m3tti>
Chattered: i've trouble to get it working over here? What have you done to get it working with nix?
<m3tti>
oolite is opensource
<m3tti>
like pioneer
<Chattered>
m3tti: I last played it on Arch. Hopefully, it should be straightforward to package.
m3tti has quit [(Read error: Connection reset by peer)]
<flxw>
okay, thank you both very much. I'm glad I didn't overlook a simple solution for it. :) I gues the proper way would be to add config options to the bash module. I saw a nice trick with optionalString and ++ in a few places, so it shouldn't be to hard to fill inputrc with the right content during configuration time.
<LnL>
peti: nice! <3
<LnL>
I don't know where to get the fancy emoji that kmicu uses
<gchristensen>
LnL: textfac.es
<domenkozar>
LnL: done
nanobapt has joined #nixos
<LnL>
don't see the one I'm looking for
pstn has quit [(Ping timeout: 258 seconds)]
mkoenig has joined #nixos
<flxw>
simpson: yes, I'll do that. It seems to be a good idea to clone the nixpkgs in genral. it would make such things pretty easy. (and with rebasing I could stay uptodate on the stable branch).
<FRidh>
I'm not really familiar with gpg. What is that fingerprint for? I would imagine you validate the signature against the archive?
<gchristensen>
the fingerprints are used to look up the public keys, which are downloaded, added to a keyring, and then the signature is compared to the src and verified it is a valid signature by a key in the keyring
zraexy has quit [(Ping timeout: 258 seconds)]
<FRidh>
What is the point of a sha256 on such fingerprint? If the fingerprint is of it won't be able to validate anyway, right?
<gchristensen>
well we need network access
<gchristensen>
to get the public keys
<viric>
gchristensen: what is the keyset sha256?
<gchristensen>
viric: it is the sha256 of all the fetched public keys (it could be a dir of .key files, or it could be a pgp keyring)
<viric>
gchristensen: but the keys in public servers change a lot (who signed them, due dates, etc.)
<viric>
give nix a keyring the user has set up with his own trusts
<viric>
do not download keys
ebzzry has joined #nixos
hiratara has quit [(Remote host closed the connection)]
<gchristensen>
I have considered how public keys change a lot, I'm not sure how to handle it. I'm fairly sure though that if we require every user to trust every package we have that they use isn't going to be a widely used feature, and that as nixpkgs we are a source of truth and I think it is okay to add extra validations here... getting emacs on ubuntu you don't trust the emacs devs, you trust the packager
<gchristensen>
not sure how to cleanly explain that
hiratara has joined #nixos
<gchristensen>
I think baking in some GPG verification would be smart. how we do that, I don't really care, as long as it isn't too hard to use
<FRidh>
So there needs to be a Nixpkgs keyring, and the possibility for a user to use their own.
ublubu has quit [(Ping timeout: 240 seconds)]
<gchristensen>
that makes sense to me
<NixOS_GitHub>
[nixpkgs] bjornfor pushed 1 new commit to release-16.09: https://git.io/vMWEd
<FRidh>
fingerprints are then the keys to the keyring. So fetching public keys should be done there, not in package expressions. There you would just state the fingerprint.
<gchristensen>
oh so that is the other problem
<pierron>
peti: exactly one can just replace the args, but importWithNixpkgs will just provide the default set of "args" based on the argument of the function.
<gchristensen>
gpgv2 will accept a signature from anybody in the keyring
<pierron>
peti: to obtain the derivation, you would have to call the derivation function on the result of the function.
<gchristensen>
so a kernel signed by the zlib people will be just as valid as a zlib package signed by the linus
<pierron>
peti: the derivation is one "view" that we can have of packages, similar to listing packages names could be another one.
<FRidh>
yes, you have to decide whether you like the zlib people to sign the kernl
<gchristensen>
it means the keyring for zlib needs to be different from the keyring for the kernel
<gchristensen>
(and by keyring I mean the file `pubring.kbx` produced by gnupg)
<joko>
Hello, is anyone using sway on Nixos? I've tried installing it, but it complains about missing dri files
<gchristensen>
or multiple-outputs (is that too many outputs?) keyring.emacs keyring.kernel
<FRidh>
well I guess I need to read a bit more about gpg first :) Anyway, I'm done for today, so bye!
<gchristensen>
could be :)
systemfault has quit [(Quit: Bye!)]
<gchristensen>
bye!
<gchristensen>
we can't just have a massive keyring with all the keys and then say use this keyring, but only trust X,Y,Z keys in it
<LnL>
I was thinking an attrset of deriivations for each set of keys
FRidh has quit [(Remote host closed the connection)]
_c_ has joined #nixos
sellout- has quit [(Quit: Leaving.)]
<gchristensen>
aye
<LnL>
not very familiar with gpg either, but but I assume you don't need a combined one
<gchristensen>
also, creating a keychain isn't deterministic -- adding the same key to an empty keychain produces different results each time, so the fetching of the keys and the production of the .kbx have to be different stages
<gchristensen>
where `keychain` could be configurable, by, say, nixpkgs.config.keychain = ...
<gchristensen>
heh, signature could probably even be `builtins.fetchurl` since its hash doesn't matter too much
<viric>
gchristensen: better. I haven't thought how to implement something like that
meocs has quit [(Max SendQ exceeded)]
ublubu has joined #nixos
<gchristensen>
viric: I'll experiment :)
<gchristensen>
domenkozar: not release manager??!
danharaj has joined #nixos
ebzzry has quit [(Ping timeout: 245 seconds)]
mkoenig has quit [(Ping timeout: 240 seconds)]
mkoenig has joined #nixos
ebzzry has joined #nixos
<exi>
any fast hints on how to create a "view"/derivation that consists of all the files of two derivations in one location?
alx741 has quit [(Quit: alx741)]
<exi>
the new rust nightly bootstrapping has a lot of assumptions about rust and cargo being in the same location and it would be easier to just give that to them then to patch the bootstrapping
alx741 has joined #nixos
ThatDocsLady has quit [(Ping timeout: 248 seconds)]
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/vMWg8
<NixOS_GitHub>
nixpkgs/master b11264a Peter Hoeg: rubber: 1.3 -> 1.4