<nomachine>
Nafai: Ah, I just grepped all hardcoded references to binaries, that binary was only for MacOS, that explains it, thanks!
<tokudan>
does someone use snapper in nixos? i can't seem to figure out how to use it, i just get dbus errors
mizu_no_oto has joined #nixos
<rod_>
hello , cant get rid of "/nix/store/lwn84z6cqlvdbbbjzsal5958r211adn5-vim-7.4.2367" , haven't any referenc in configuration.nix and not in user-env ... did "nix-env --delete-generations old" and "nix-env --delete-generations old -p /nix/var/nix/profiles/system" plus "nix-store --gc" but it doesnt go away ... any clue , someone ?
<tokudan>
rod_, what does "nix-store --query --requisites /nix/store/lwn84z6cqlvdbbbjzsal5958r211adn5-vim-7.4.2367" show? any clue in there?
<rod_>
LnL, got "chase" packaged ... with boemgc as u suggested ... did pull request for nixos on github
<LnL>
rod_: --referrers or --referrers-closure that will most likely point to a user profile
<berce>
clever: ofcource it would be nicer if mesa didn't depend on those drivers that won't be needed on an arm platform
<LnL>
rod_: compare that with nix-store --gc --print-roots
<rod_>
ok ... i should say that i am using nixos for 4 days now ... and i am figuring all out right now ....
<simpson>
rod_: No worries. What's it say?
<nomachine>
simpson: Can I also use buildFHSUserEnv for the install phase? Those hardcoded references to /bin/* are also plenty in the installer scripts.
<rod_>
ok some entries in both outputs .... there is one in common , that is "/nix/var/nix/profiles/default-7-link -> /nix/store/g3j47aikknszzpscscrbwz3q6lvwfiqj-user-environment"
<rod_>
so my user environment is using that vim version am i right ....
<simpson>
nomachine: Maybe? I'm not sure.
<simpson>
rod_: Looks like it.
eacameron has joined #nixos
<rod_>
i wanna use vim_configurable in .nixpkgs/config.nix to have .vimrc but when i vi(m) i not get the defined .vimrc options
ToxicFrog has quit [(Quit: WeeChat 1.5)]
<tokudan>
I'm using vim_configurable... bit complicated though, i can copy you an example, if you want
<rod_>
it starts that vim-7.4.23.... in config.nix i set name to "v" ... when i do "v" .vimrc gets loaded ..... so how do i get rid of that other vim
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<rod_>
tokudan, thx ... have it working with "v" ... but got that other instance of vim wich doesnt load .vimrc i wanna get rid of
<rod_>
hmmmm....
justan0theruser has joined #nixos
<tokudan>
what does nix-env -q show as user?
<tokudan>
does it show vim?
<berce>
rod_: dit you try uninstalling with nix-env -e vim?
<rod_>
┌─[rod@polyphem] - [~] - [Di Jan 03, 01:20]
<rod_>
└─[$]> chase `which vi` ~
<rod_>
┌─[rod@polyphem] - [~] - [Di Jan 03, 01:27]
<rod_>
└─[$]> chase `which vim` ~
<rod_>
┌─[rod@polyphem] - [~] - [Di Jan 03, 01:28]
eacameron has quit [(Ping timeout: 272 seconds)]
<rod_>
└─[$]> chase `which v` ~
<rod_>
berce, yes did nix-env -e vim
<rod_>
berce , also cleaned configuration.nix of all vims
mizu_no_oto has joined #nixos
<nomachine>
simpson: Got it, I can of course first create an env, then explicitly call it using ${fhsEnv}/bin/fhs-name which will run what I specified as runScript, just the default bash, so I can just supply it with the patched install script
<berce>
rod_: have you had vim in your /nixos/configuration.nix?
justanotheruser has quit [(Ping timeout: 258 seconds)]
<berce>
Then you can still reboot to that old configuration, so it still has a root.
<rod_>
tokudan, nix-env -q doesnt show vim
<toogley>
is there a reason there's a "env" link to /usr/bin/ but not for sh, bash, and python and such?
<toogley>
i mean this breaks compatibility with a lot of unix scripts
<toogley>
which would work otherwise
<rod_>
berce, yes first i had it in but now not , deleted all old generations though
<berce>
"nix-env --delete-generations old" done as both root and normal user?
<simpson>
toogley: NixOS doesn't implement LSB or FHS. /usr/bin/env is a courtesy to make it easier to do Nix shebangs: #!/usr/bin/env nix-shell
<rod_>
hello , cant get rid of "/nix/store/lwn84z6cqlvdbbbjzsal5958r211adn5-vim-7.4.2367" , haven't any referenc in configuration.nix and not in user-env ... did "nix-env --delete-generations old" and "nix-env --delete-generations old -p /nix/var/nix/profiles/system" plus "nix-store --gc" but it doesnt go away ... any clue , someone ?
<rod_>
berce ^
<toogley>
simpson: but if you'd working with nixOS would be so much easier...
<simpson>
toogley: But then it wouldn't be NixOS!
<toogley>
simpson: is there a reason you don't implement that? i'm trying to understand it.
<toogley>
simpson: i mean does it make sth harder ..?
<simpson>
toogley: Yes. Nix is a system that tries to minimize *ambient authority*.
<tokudan>
rod_, you need to decide if you want to use /etc/nixos/configuration.nix (that's my choice) or the nix-env -i method to install your packages
<berce>
rod_: still the question: as normal user AND as root?
<toogley>
simpson: and i guess i can't turn on an option to have /usr/bin/bash and such?
<simpson>
toogley: This means that the power to use certain system-level or "ambient" packages is not available; packages may only use dependencies which they have explicitly been granted.
<simpson>
toogley: Correct, there is no such escape hatch. There *are* several tools like nix-shell to make it easier.
<rod_>
berce, yes as user AND as root ....still
<tokudan>
rod_, it seems to me that you're mixing them, which can lead to confusion
<simpson>
rod_: Your system (root) profiles may not be unpinned by the GC until you reboot onto your new kernel.
<simpson>
rod_: But that's a wild guess.
<toogley>
simpson: ah, okay. so do i understand you correctly: allowing things like /usr/bin/bash makes it easy to break package isolation?
<rod_>
tokudan, yes my strategy is configuration.nix for crucial software ... and nix-env -i for throwaway environments when developing or doing other kind of session
<aszlig>
gchristensen: get_physical_spec() should be correct, the "overriding" part you mentioned is probably because it's merged with the main physical spec in Deployment
<simpson>
toogley: Yep. Because which bash does /usr/bin/bash point to?
<rod_>
tokudan, vim i consider crucial , still .vimrc should be per user ....
<aszlig>
gchristensen: so from within a resource it should be self.depl.get_physical_spec()
<nomachine>
toogley: You could use buildFHSUserEnv and define which packages you would want to be available, and change your shell to it, should work, but see simpson's responses why you might not want to even try and do this
<NixOS_GitHub>
[nixpkgs] LnL7 pushed 4 new commits to master: https://git.io/vMmap
<NixOS_GitHub>
nixpkgs/master 1f896f4 Chris Martin: libbitcoin: init at 2.11.0
<NixOS_GitHub>
nixpkgs/master cb7c3d9 Chris Martin: libbitcoin-client: init at 2.2.0
<NixOS_GitHub>
nixpkgs/master 1d278fe Chris Martin: libbitcoin-explorer: init at 2.2.0
<gchristensen>
aszlig: is there a way to get the physical spec before it has been saved to depl? ie: in my bootstrapping (the create()) I want to write out a configuration.nix which resembles what nixops would replace it with
<tokudan>
rod_, hmm... can't really help you there, as I setup every software in configuration.nix. here's my variant for vim_configurable: rod_, here's my variant of including vim_configurable: https://gist.github.com/tokudan/c92c0acc2766ceb06659363b815efba4
<rod_>
tokudan, as i understood big advantage of nixos over other distros is exactly that you have system package management but aslo per user package managment
<tokudan>
rod_, yeah, but .vimrc can be in every home directory and every user can set that up for himself
<msd>
LnL: if it's not too late, could you check this out please? https://github.com/NixOS/nixpkgs/pull/21600 <- just a few explicit sets for wxGTK for Erlang, and a formatting issue. Just calling on your knowledge of how the Erlang expression works. Thanks!
<aszlig>
gchristensen: why do you want to write out a configuration.nix at all?
<toogley>
nomachine: simpson thanks
<berce>
rod_: I prefer .nixpkgs/config.nix over nix-env -i
<berce>
it's the same as configuration.nix, but on user level
<gchristensen>
aszlig: to call nixos-install... which I want to do because in the kexec, the /nix/store is read-only.
<aszlig>
gchristensen: i mean after all you already have the .drv after eval
m0rphism has quit [(Quit: WeeChat 1.4)]
<rod_>
berce, thats how i got my working "v" vim .... .nixpkgs/config.nix
<LnL>
msd: doesn't erlang use wxmac?
<aszlig>
gchristensen: if /nix/store is read-only everything is fine, no?
<gchristensen>
one moment
<aszlig>
gchristensen: i mean, you already have networking support within the kexec'd instance, so you only need to copy the closure to the new partitioning scheme, right?
<rod_>
berce, the (configuration|config).nix files are nicely declarative , whereas nix-env -i is like a apt-get install
<msd>
LnL: wxGTK is stated within the Erlang interpreter expression
<gchristensen>
aszlig: I suppose so
<berce>
rod_: if you're really desparate, you can just backup config.nix and confituration.nix and start from scratch again :-)
<gchristensen>
aszlig: how can I tell nixops to put it under /mnt?
<rod_>
ok ... i hack around till i get it .... thanx for all help ....
<berce>
good luck!
<gchristensen>
or: something else
<rod_>
berce, last resort .... :) , but dont intend to ... but yea with nixos its much less of a hassle then with debian
<aszlig>
so maybe add a function argument like { withNixpart ? true }
aarvar has joined #nixos
<msd>
LnL: wait, you're right. However, I'd still argue for the change: Linux's wxGTK version is 2.8, which is below the default 3.0.
brh has joined #nixos
<tokudan>
does someone use snapper in nixos? i can't seem to figure out how to use it, i just get dbus errors
bennofs has quit [(Ping timeout: 260 seconds)]
<gchristensen>
aszlig: I wouldn't mind using nixpart
msd has quit [()]
<LnL>
msd: yeah indeed, I'll test tomorrow
<gchristensen>
aszlig: the whole implementation process is a bit exhausting, though :9
<aszlig>
gchristensen: i know... it took me one month for the hetzner implementation
<gchristensen>
ow!
brh has quit [(Ping timeout: 256 seconds)]
<gchristensen>
well since you have so much practice at it ...!
<nomachine>
I should patch a package to write its log files to /var/lib/pkgname, right? Or what is best practice?
<aszlig>
gchristensen: but as said, most of the stuff should already be in place nowadays but probably needs to be more generalized for other backends
kampfschlaefer has quit [(Ping timeout: 258 seconds)]
<LnL>
nomachine: if it's a service it should just log to stdout so it ends up in journald
mudri` has quit [(Ping timeout: 264 seconds)]
awoidh has quit [(Ping timeout: 260 seconds)]
<nomachine>
LnL: Its a log of the installation (ok, so I will try and redirect that to stdout) and some small database (that should be in /var right?) I am just wondering about this, because would that not potentially badly poison your data if you have multiple version of some application?
ryanartecona has joined #nixos
* hodapp
realizes, slowly, how many otherwise unrelated tools "nix-shell --pure -p foo bar baz..." can replace.
<gchristensen>
yeah ... :)
aarvar has quit [(Ping timeout: 256 seconds)]
justinkaz has joined #nixos
zraexy has joined #nixos
proteusguy has quit [(Ping timeout: 258 seconds)]
m0rphism has joined #nixos
proteusguy has joined #nixos
slack1256 has joined #nixos
slack1256 has quit [(Remote host closed the connection)]
aarvar has joined #nixos
<justinkaz>
how do I stop the httpd service to override one of the config files
<justinkaz>
I've tried /run/httpd -k stop and receive this error AH00543: httpd: bad user name daemon
<justinkaz>
just for a test I created a directory entry in extraConfig for "/" and it';s not processing anything
aarvar has joined #nixos
<justinkaz>
the good news, I can tell it received my extra config options because the log reports mod_rewrite (my test config line) is already enabled so it skips that line
<justinkaz>
So I have to port all my web service specific overides into the config for now... till I figure out what else is going on but so far it's working
tokudan has quit [(Quit: Leaving)]
<makefu>
simpson: seems the grammar doc is also outdated and has not been updated for 7 years. i will use the link to the nix manual.
mizu_no_oto has joined #nixos
AllanEsp1nosa has quit [(Ping timeout: 246 seconds)]
mbrgm has quit [(Ping timeout: 246 seconds)]
mbrgm has joined #nixos
sigmundv__ has quit [(Ping timeout: 268 seconds)]
eacameron has joined #nixos
eacameron has quit [(Ping timeout: 246 seconds)]
filterfish_ has quit [(Remote host closed the connection)]
filterfish has joined #nixos
matthewbauer has joined #nixos
<mpickering>
in the ghcHEAD derivation you have to set the "version" attribute manually
<mpickering>
however, this information is dependent on the commit
<mpickering>
is there a way to remove this duplication by setting the "version" with a bash script or something rather than a string?
brh has joined #nixos
filterfish has quit [(Remote host closed the connection)]
edapa has joined #nixos
brh has quit [(Ping timeout: 265 seconds)]
derjohn_mob has quit [(Ping timeout: 258 seconds)]
filterfish has joined #nixos
<edapa>
I'm having trouble importing modules from xmonad-contrib to my xmonad.hs
<edapa>
Does anyone know the right way to solve this. I have tried to package xmonad-contrib with my ghc, and also to add it to the list of extraPackages in the xmonad windowManager set
<hyper_ch>
why do I get this: error: unable to download ‘file://root/.nixos/Oracle_VM_VirtualBox_Extension_Pack-5.1.10-112026.vbox-extpack’: Couldn't read a file:// file (37)
<hyper_ch>
when I run nix-prefetch-url file://root/.nixos/Oracle_VM_VirtualBox_Extension_Pack-5.1.10-112026.vbox-extpack
paperd has quit [(Ping timeout: 272 seconds)]
indi_ has quit [(Remote host closed the connection)]
<Dezgeg>
maybe file:///root etc.
indi_ has joined #nixos
<hyper_ch>
yeah thx
aarvar has quit [(Ping timeout: 248 seconds)]
indi_ has quit [(Ping timeout: 246 seconds)]
amarsman has joined #nixos
MercurialAlchemi has joined #nixos
aarvar has joined #nixos
indi_ has joined #nixos
pi3r has joined #nixos
nanobapt has joined #nixos
indi_ has quit [(Remote host closed the connection)]
indi_ has joined #nixos
indi_ has quit [(Remote host closed the connection)]
indi_ has joined #nixos
brh has joined #nixos
nanobapt has quit [(Ping timeout: 264 seconds)]
brh has quit [(Ping timeout: 256 seconds)]
mizu_no_oto has joined #nixos
mizu_no_oto has quit [(Ping timeout: 265 seconds)]
filterfish has joined #nixos
filterfish_ has quit [(Read error: Connection reset by peer)]
pi3r has quit [(Read error: Connection reset by peer)]
<LnL>
gchristensen: I think we also need more documentation on how to build the manuals, only nixpkgs has instructions and those are very basic
<avn>
btw I also have Filesystem related question --
<LnL>
ah, but doesn't mount -t bind also work?
<rly>
How can I read the version of attribute nixos.foobar which has an attribute version?
<peter-holm>
I doubt. It is a local path I want to bind mount. This /dev/disk/by-label is the exact same thing as /dev/disk/by-uuid or similary to /dev/sdc39 ... It is a path to a device
<LnL>
rly: nix-instantiate '<nixpkgs>' --eval -A nix.version
tvon has quit [(Remote host closed the connection)]
<avn>
peter-holm: you use device path for "source dir"
<rly>
LnL: can I also get that via nix-repl?
<LnL>
sure, nix-repl '<nixpkgs>' -> nix.version
<LnL>
I think it would be nice if all derivations had a version attribute
<peter-holm>
YI have tested, But I am alll to neww to the syntax. And there are several ways to declare the same thinf. I am to new to convert between thoose difreernet declarations.
fritschy has joined #nixos
<peter-holm>
Sory speeling errors. I have 'fat finges' and shold need have my glasses .
<LnL>
it looks like your keyboard repeat is set a little to fast :p
<peter-holm>
Is it ok to paste a few lines of code.?
<LnL>
general irc conventions is to use a paste service if it's more then about ~3 lines
<peter-holm>
its about five lines.
<peter-holm>
Well - I will continue search google.
<LnL>
just put it in a gist, pastbin or something like that
peter-holm has quit [(Ping timeout: 272 seconds)]
<avn>
little offtopuc -- folks, do you know where is "print" hotkey in urxvt, and how to turn it off permanently?
goibhniu has joined #nixos
ixxie has quit [(Remote host closed the connection)]
derjohn_mob has quit [(Ping timeout: 258 seconds)]
<FaustXVI>
Hi there
<ixxie_>
gnome tweak tool doesn't seem to manage disabling middle mouse paste and I am wondering whether it's a nix issue
<FaustXVI>
happy new year to all :)
<ixxie_>
heya FaustXVI
<ixxie_>
Happy new year
ixxie has quit [(Ping timeout: 260 seconds)]
uralbash has quit [(Ping timeout: 246 seconds)]
<FaustXVI>
I'm trying to work on kotlin (https://github.com/JetBrains/kotlin) and one of the requirements is "JDK 1.6, 1.7 and 1.8" and I must admit I have no idea on how to make a nix derivation for that. Could any body help me ?
<manveru>
ixxie_: that's something you'd set in your xorg config, not gnome, afaict
<ronny>
FaustXVI: i believe you can just use the openjdk thats already in nix
takle has quit [(Remote host closed the connection)]
jgertm_ has joined #nixos
georges-duperon has joined #nixos
thc202 has joined #nixos
<ixxie_>
manveru: I see.... and luckily there seems to be an option entering it into a module
<ixxie_>
but isn't it partially autogenerated?
Wizek has joined #nixos
Wizek_ has joined #nixos
<Dezgeg>
don't we already have kotlin for that matter?
<FaustXVI>
Dezgeg: yes we do, but I want to work on it, the language itself :)
<manveru>
just add your config to the section you want
takle has joined #nixos
<ixxie_>
manveru: awesome theres a specific option for button mapping
<NixOS_GitHub>
[nixpkgs] peti closed pull request #21595: ghc: when building with 8.0.2 or later, enable parallel builds (master...ghc-8.0.2-parallel-builds) https://git.io/vMmGD
<ixxie>
the root of that repo also has a set of scripts I am working on to manage the whole thing, and some of those nix files are templated and managed by those scripts to allow, for example, user management
ThatDocsLady has joined #nixos
<trqx>
nice, thank you ixxie & manveru
indi_ has quit [(Remote host closed the connection)]
<ixxie>
(not that nix doesn't support user management, I just wanted to manage users in the configuration files systematically
mudri` has joined #nixos
ThatDocsLady has quit [(Client Quit)]
<trqx>
ok
<trqx>
regarding a permanent liveusb installation, anyone tried it? will I notice a huge performance gap compared to any other linux distro?
<trqx>
this might have been reported already but certificate does not work for www.nixos.org, only nixos.org and search engines send results for www.nixos.org/stuff
<rly>
LnL: how do I add bash completion support to a package?
systemfault has joined #nixos
indi_ has quit [(Remote host closed the connection)]
<rly>
LnL: i.e., a package wants to install into /etc/bash<whatever>, but on NixOS that's not going to work.
<rly>
LnL: I think I already found it.
indi_ has joined #nixos
<mbrgm>
what could be a reason for a package not appearing when I'm doing `nix-env -qa -I nixpkgs=/path/to/my/nixpkgs`, although that pkg is in the top-level/all-packages.nix?
brh has joined #nixos
sdothum has joined #nixos
brh has quit [(Ping timeout: 260 seconds)]
Intensity has quit [(Remote host closed the connection)]
ublubu has quit [(Ping timeout: 258 seconds)]
ublubu has joined #nixos
indi_ has quit [(Remote host closed the connection)]
<ixxie>
rly: what is your solution? I will also be needing to look into this soon
<LnL>
rly: I was in a meeting, it's something like $out/share/bash_completion
<peter-holm>
error: The option value `fileSystems./workspace/peter.options' in `/etc/nixos/configuration.nix' is not a list of strings.which fails with error
zagy has quit [(Quit: Leaving.)]
<peter-holm>
Have tryiing to get this to work a half day and a whole nigt ( now It is morning here in sweden).
<page>
peter-holm: options = [ "bind" ]
phanimah1sh has joined #nixos
phanimahesh has quit [(Ping timeout: 248 seconds)]
<NixOS_GitHub>
nixpkgs/master 2a4c831 Tuomas Tynkkynen: linux_testing: 4.10-rc1 -> 4.10-rc2
indi_ has quit [(Remote host closed the connection)]
<gchristensen>
domenkozar: this person seems really sketchy
indi_ has joined #nixos
<gchristensen>
this is the second or third time they've said they have a bit of time, add me to the organization
ertes has quit [(Ping timeout: 248 seconds)]
ertesx is now known as ertes
<gchristensen>
it reminds me a lot of https://github.com/NixOS/nixpkgs/issues/20836's "I would also like to perform security review of nixpkgs and would possibly pay some security researcher to inject code into nixpkgs/nixos on purpose, to see how bad security is"
<LnL>
and it's not very likely that something strange gets introduced in the core system without somebody noticing that hydra is rebuilding everything
<gchristensen>
yeah ... I think the state of affairs is a lot better than they think
guillaum1 has joined #nixos
<LnL>
huh, he wants commit access and he's not even a contributer yet?
<gchristensen>
LnL: to _anything_!
<gchristensen>
https://github.com/itsN1X scroll through his github "Contribution activity" for the last 3 years
modulistic has quit [(Remote host closed the connection)]
msd has joined #nixos
<LnL>
lol
<gchristensen>
oh maybe a few that I hadn't seen before
<msd>
Getting hangs on waiting for the IP address from NixOps when following the default tutorial. Any ideas?
<msd>
Using all default settings, turned firewall off
<msd>
VBox has extensions and is headless
<LnL>
msd: might be cause of the vboxnet0
<msd>
vboxnet0?
<LnL>
VBoxManage list hostonlyifs
<msd>
LnL: using this on a NixOS machine
<msd>
LnL: everything seems to be alright when running that
<LnL>
nixops uses that network to connect to the nodes, but it expects it to have the "default" ip range and a dhcp server
<msd>
DHCP disabled
<msd>
Ah
<LnL>
list dhcpservers
<msd>
IP: 192.168.56.100
<LnL>
oh that looks fine
<LnL>
not sure then
<msd>
lowerIPAddress: 192.168.56.101
<msd>
upperIPAddress: 192.168.56.254
indi_ has quit [(Remote host closed the connection)]
<msd>
Oh LnL is DHCP enabled by default?
<LnL>
yes it should be configured correctly by default
<msd>
Strange
<LnL>
but if you remove that hostonly network nixops will get confused
<msd>
Right right
<guillaum1>
Hi there. I just discovered nixos and I'm tempted to test it. Could you just confirm me that I did not missunderstood it. I'm working on a complex C++ codebase which depends on a bunch of packages, with specific version and specific build configurations. I can just list my package and its specific configuration and run nix-shell and get an environnement which the right package built with the right setup?
<guillaum1>
Will it rebuild also all the dependencies?
<msd>
Maybe...this requires a restart?
<gchristensen>
guillaum1: it should :) the one caveat is we may not package the specific versions you require, but you could create a nix-shell which does
<LnL>
yep, you might have to write some expressions yourself
msd has quit [(Remote host closed the connection)]
<LnL>
but nix-shell is ideal for this kind of stuff IMHO
msd has joined #nixos
msd has quit [(Remote host closed the connection)]
<guillaum1>
cool, thank you. And does that means that if I'm rebuilding a specific gcc version, because most libs depends on the gcc c++ std library, it will recompile most of the lib, turning my system into a source based one? ;)
Hunner has joined #nixos
<gchristensen>
guillaum1: you could use a specific version of GCC _just_ in that one nix-shell :) but yes, if you changed the whole system's gcc, you'd be compiling everything
<LnL>
we have a bunch of versions in nixpkgs, but if is's not in there you'll have to compile it yourself
Hunner has joined #nixos
ericsagn1 has joined #nixos
bfrog has joined #nixos
abcrawf has joined #nixos
<trqx>
trying to add a user to a nix installation via configuration.nix, I get "This user is currently not available" even after reboot
charlesofarrell_ has joined #nixos
<LnL>
did you run nixos-rebuild switch?
<trqx>
it seems there is no shell but setting a shell is not in the manual
<trqx>
LnL: yes
codedmart has joined #nixos
<gchristensen>
can you share that part of your configuration.nix?
<trqx>
sure
<guillaum1>
Nice, I'm really exited to test all of this. And I will for sure contribute a bunch of package because actually none of my dependencies are in the list of nixpkg.
<gchristensen>
LOL! well, we're happy to help with that :) you may find my search engine useful: https://search.nix.gsc.io/
<ixxie>
I was thinking of doing something along those lines for some R simulation code, but I wanna do it with a database.
<fre>
I have R code in there
<gchristensen>
fre: the second hardest job I've ever had included trying to get a team of scientists to have reproducible research
<niksnut>
heh
<ixxie>
fre: the thing is, I wanted to cache runs to prevent people from needlessly repeating runs, so I thought the model would be simulated and stored in a db with all parameter values AND the seed
FRidh has joined #nixos
<ixxie>
so that a highly dimension statespace can be explored ad-hoc, building a juicey database
<fre>
I just have some nix-shell --run zymake zymakefile type rules in a Makefile
<LnL>
niksnut: anything special new in 1.11.5?
<ixxie>
<3 rWrapper
<fre>
the database is in the disk, managed by zymake
<ixxie>
every package should handle it's plugins like that :)
<gchristensen>
niksnut: those 9p changes are incredible.
<fre>
yes it's really cool :-)
<ixxie>
cool
<fre>
niksnut: grazzie, I'll use this paper.
<LnL>
gchristensen: are they in master or unstable?
<ixxie>
oh yeah, and the last thing I wanted is a complete web interface for that, fre xD to really make the model accessible to anybody
<fre>
I can reccomend you the sumatra system tho ixxie
<gchristensen>
LnL: eelco just pushed them 1 minute ago
<FRidh>
ixxie: I wrote Nix expressions for some of my analyses and simulations. Having the "caching" in the store was quite neat, and having it entirely reproducible. In the end I quit working with this because it required wrapping a lot of my Python functions as Nix functions
<gchristensen>
LnL: to stable
<gchristensen>
16.09
<LnL>
ooh even better!
<ixxie>
that last one made me quit academia and enter the startup world >.<
<fre>
FRidh: that sounds very cool, do you have some example code?
<gchristensen>
LnL: they've been in master for a few weeks now
<ixxie>
FRidh: so every run gets a different entry in the store?
<FRidh>
I could upload that code. It won't run without the data though.
<fre>
Sure. I'm extremely interested. I work in a team where some people research exp.reproducibility full time
<FRidh>
ixxie: well, no, it depends on your input. For example, if I change a parameter in my simulation (goal was to use json files) it might result in a new store path, but not if it had already been computed before.
<fre>
they would love to see that
brh has joined #nixos
<ixxie>
FRidh: yeah thats what I thought
<FRidh>
I think it can be worth using Nix for this if you will use this simulation or whatever it is more often, but otherwise it just seems too much work.
<fre>
FRidh what was your personal incentive to do it?
<gchristensen>
my guess is b/c nix is an addiction
Lowl3v3l has joined #nixos
<FRidh>
fre: An interest in reproducibilty and getting annoyed by so much academic work appearing far from reproducible.
<FRidh>
ha
<fre>
FRidh Oh. I want to see some code. :-) probably because I'm addicted.
indi_ has quit [(Remote host closed the connection)]
<ixxie>
FRidh: if you make a framework for any code in a language you could quite easily do it for any project
<ixxie>
I imagine with R wrapper its easy to make R code reproducable
<FRidh>
fre: simulations and tools are getting more complicated, and papers just don't include enough information.
<ixxie>
just curious, what kinda research are yall doing?
<FRidh>
ixxie: I simulate the audible sound of aircraft in urban environments and specifically looked at how atmospheric turbulence affects the audible sound.
<ixxie>
kewl, FRidh - so wave equations and fluid dynamics?
<FRidh>
ixxie: well, no fluid dynamics, that goes a bit too deep :) But indeed wave equation. Lots of signal processing, statistical methods for describing atmospheric turbulence, and finally human perception. In the end its a lot of models and custom code put together. And how about yourself?
<fre>
FRidh: cool, yes i'm familiar with the issue of reproducibility.. I work in HPC and we have none. :-)
<gchristensen>
fre: does your team use nixos elsewhere? some other HPC people use nixos on their clusters
<fre>
ixxie: I'm working in parallel job scheduling in HPC, I use machine learning techniques to learn to allocate jobs
<FRidh>
fre: haha, sounds familiar
<fre>
gchristensen: yes we use nix on our clusters too
<fre>
gchristensen: its in the University of Grenoble/ Inria
<FRidh>
nix, or nixos?
<fre>
nix
<fre>
is there an irc channel for just nix?
<gchristensen>
this is it :)
<FRidh>
if I am correct there's a torque module for nixos
<fre>
ok :)
<gchristensen>
would you / your team like to write something up about how much time / money / frustration nix saves you by promising reproducibility and package isolation etc. etc. ? :P
<gchristensen>
fre ^
<fre>
Not me personally but our guys surely would love to
<FRidh>
fre: How do you use Nix on the cluster? On our cluster the main disks are rather small, so I had to install nix as a user on the Lustre filesystem, but those access times...
<fre>
I don't know if they have time for that
<ixxie>
cool stuff FRidh, fre
<gchristensen>
fre: I'll send your team Nix stickers if you do :d
<ixxie>
I did some ecological simulations, I am interested in analyzing fragility / noise response in complex systems; but then I got more interested in methodological issues, which led to a realization that I wanna have a better technology stack
<ixxie>
now I am trying to learn more coding and develop some kind of startup
<ixxie>
hopefully with a Nix heart :)
<gchristensen>
:)
indi_ has joined #nixos
<fre>
oooh that's really cool ixxie
<FRidh>
ixxie: Sounds familiar. I think there's so much potential for improvement in academia on this matter, it just seems there's hardly any funding for it.
<fre>
scheduling in hpc is really sensitive to noise in workload
michaelpj has joined #nixos
<ixxie>
FRidh: I want to continue my research in the private sector; my dream is to create an organization that is something between business, NGO and academia
mguentner2 is now known as mguentner
<fre>
ixxie: you can work for the gates foundation then. :)
<fre>
ixxie: it's controversial but that's the basic idea.
<NixOS_GitHub>
nixpkgs/master 85ff5a2 Stefan Goetz: youtube-dl: 2016.12.22 -> 2017.01.02 (#21607)
phanimahesh has joined #nixos
nomachine has joined #nixos
phanimah1sh has quit [(Ping timeout: 258 seconds)]
greymalkin has quit [(Ping timeout: 260 seconds)]
<kmicu>
mbrgm: nix-env requires ‘-f’ flag so use ‘nix-env -qa -I nixpkgs=/path/to/my/nixpkgs -f '<nixpkgs>'’ or ‘nix-env -qa -f /path/to/my/nixpkgs’.
<nomachine>
I am trying to use buildFHSUserEnv run an install script during the installPhase, but I am getting an error I do not know how to solve: in `block in make_fcall': No such file or directory (Errno::ENOENT)
msd has joined #nixos
<mbrgm>
kmicu: ahhh! thank you! I was so focused on `-I` :-/
<nomachine>
I run it like this: ${installEnv}/bin/nomachine-install -c './nxserver --install fedora' (runScript of buildFHSUserEnv is bash)
greymalkin has joined #nixos
<nomachine>
So I assume ./nxserver cannot be found, however it does exist in the $PWD at that point. Does the script nxserver need to be available within installEnv for it to work?
<kmicu>
(#xmonad has ‘lambdabot’, #guix has ‘sneek’, probably written in Guile. They both deliver messages successfully and help impatient IRC users seeking help.)
* LnL
wants a nix-instantiate bot
<kmicu>
! 2+2
<kmicu>
nixbot: 4
<LnL>
what
<kmicu>
Proof of concept. I’m sitting inside the bot.
<gchristensen>
niksnut: can you send me some usage statistics about the binary cache?
mkoenig has joined #nixos
cpennington has quit [(Ping timeout: 256 seconds)]
<gchristensen>
kmicu: we don't even have operators for 2 + 2 btw :P
takle has joined #nixos
Baughn has quit [(Remote host closed the connection)]
<rly>
gchristensen: why doesn't the TravisCI build script act a bit more intelligently then? I.e. it could just display the last 4MB, for example (as a completely trivial improvement).
<gchristensen>
rly: I invite you to try and make improvements, many improvements have been attempted.
<gchristensen>
one issue is that requires not emitting anything until the very end of the build, which I think is also not possible since if your build is quiet it kills it assuming it is broken.
<rly>
gchristensen: yes, but that can also easily be fixed.
derjohn_mob has quit [(Ping timeout: 258 seconds)]
<rly>
gchristensen: are there any other limitations besides size and frequency of output?
<gchristensen>
probably, I recommend reading up on travisci's docs
<LnL>
also time with a mass rebuild, but that's not something we can work around AFAIK
cpennington has joined #nixos
brh has joined #nixos
<msd>
LnL: "Failed to create domain from /tmp/nixops-tmpozTNU8/webservice-domain.xml" <- any ideas what the failure is? This seems like something to do with the IP mapping schema; however, I'm not too sure what the problem could be. Image deployed using QEMU with default settings
<kmicu>
gchristensen: ah, that chapter is not updated since 1.6 with ‘Nix now has proper integer arithmetic operators. For instance, you can write x + y instead of builtins.add x y, or x < y instead of builtins.lessThan x y. The comparison operators also work on strings.’.
<gchristensen>
ahh right on
<rly>
LnL: why is that relevant?
<gchristensen>
sometimes they fail because they just take too long
<gchristensen>
or they fail because they fill up the disk
<rly>
LnL: ah, I I thought you meant "time the PR".
<rly>
s/I I/I
brh has quit [(Ping timeout: 248 seconds)]
m0rphism has quit [(Ping timeout: 272 seconds)]
<LnL>
each pr will try to rebuild the world and take forever or fill the disk of the travis builder
<rly>
LnL: so, when does TravisCI do something useful then?
<LnL>
when master is stable
<rly>
LnL: when binary cache works, you mean?
<gchristensen>
when the PR is small enough and master is stable / thoroughly built and travis only has to build a few things.
<clever>
gchristensen: one anoyance with the travis stuff, any mass-rebuild related products cant be shared between PR's
<gchristensen>
many annoyances
<clever>
gchristensen: and dont contribute back to the cache
<clever>
replacing travis with hydra would solve a number of problems
<gchristensen>
yeah :)
<LnL>
ideally these rebuilds only happen on staging, but that's not always preferrable
<rly>
clever: and the reason we don't do that is because we are all evil and are going to haxor it?
<clever>
rly: last i heard, somebody said it would take another year of work to get that hydra stuff going
<gchristensen>
well because travis isn't really very fit for the task in basically any way.
<rly>
gchristensen: I mean the reason for not doing it.
<gchristensen>
what is "it"?
<rly>
gchristensen: clever understood.
<gchristensen>
but either way my response is "time and money"
<gchristensen>
please don't be an ass, rly
<rly>
gchristensen: that's not what I meant.
<rly>
gchristensen: I could have communicated better.
<gchristensen>
you still can
eacameron has joined #nixos
<rly>
gchristensen: what is the reason for not currently using Hydra to achieve the goal which we are supposed to achieve by using TravisCI?
<NixOS_GitHub>
nixpkgs/rocksdb addcb0d Domen Kožar: generic-stack-builder: set DYLD_LIBRARY_PATH to fix OSX dynamic linking
<gchristensen>
a big one is Hydra doesn't support building arbitrary things yet, so any sort of dynamic jobset or evaluation which gets created based on a changing input
bennofs has joined #nixos
<gchristensen>
ie: when a PR is in, we need to make an evaluation for it. hydra doesn't have support for this
offline has quit [(Quit: Quit)]
jeaye has quit [(Ping timeout: 258 seconds)]
<rly>
gchristensen: so, cgroups support basically and adding meta-data for jobs to set memory limits, etc.
<rly>
gchristensen: is it even realistic to modify Hydra to do this?
jeaye has joined #nixos
<gchristensen>
this isn't a security issue or cgroup issue, it is literally hydra cannot be configured to build
<gchristensen>
hydra can be changed to support this behavior, hydra isn't really that complicated I think, people just get spooked by perl
<clever>
gchristensen: biggest road-block i had when editing the perl, is that the build scripts nuke anything not tracked byt
<clever>
gchristensen: so when i added new perl files, git didnt know of them, and they got nuked
<gchristensen>
heh, yeah
dash has quit [(Ping timeout: 258 seconds)]
mkoenig has quit [(Read error: Connection reset by peer)]
Rotaerk has quit [(Quit: Leaving)]
dash has joined #nixos
<gchristensen>
a concern I have is about capacity, building arbitrary numbers of PRs with arbitrary amounts of "mass-rebuildyness" is a substantial amount of additional load on the current infrastructure.
siel has quit [(Ping timeout: 258 seconds)]
Myrl-saki has joined #nixos
<LnL>
yeah, I think everything currently gets uploaded to the cache
<clever>
gchristensen: priority could be set on the PR stuff to run less often
<LnL>
that probably won't scale to do that with PRs
<clever>
and expiry on the cache would be more important if that happens
<clever>
yeah
rod has quit [(Remote host closed the connection)]
zagy has quit [(Ping timeout: 260 seconds)]
zagy has joined #nixos
mkoenig has joined #nixos
zagy has quit [(Client Quit)]
cpennington has quit [(Ping timeout: 248 seconds)]
earldouglas has left #nixos []
thc202 has quit [(Ping timeout: 248 seconds)]
mbrgm has quit [(Ping timeout: 265 seconds)]
mbrgm has joined #nixos
peter-holm has quit [(Ping timeout: 265 seconds)]
<msd>
Does systool exist in NixOS?
cpennington has joined #nixos
m0rphism has joined #nixos
eacameron has quit [(Remote host closed the connection)]
<bennofs>
msd: nixpkgs.libsysfs is the package
eacameron has joined #nixos
alx741 has quit [(Ping timeout: 248 seconds)]
<msd>
thanks bennofs
<FRidh>
gchristensen: why even bother with using hydra for this. Hydra you like to have in order to spread work among machines. But at the same time, we don't want mass rebuilds to be build. So, why not have just a single more powerful machine for building PR's? And have it fail after x time.
<gchristensen>
FRidh: well I think running the VM tests can be helpful, and having a cluster of machines can be helpful too, especially for big rebuilds
<gchristensen>
FRidh: I mean. ... if all we want is like Jenkins or Go.CD running on big fat machines we can do that very easily
msd has quit [(Remote host closed the connection)]
<msd_>
domenkozar: have you run into issues with KVM's nesting and QEMU's ability to launch VMs? Seems like nesting not being activated is what's blocking my QEMU VMs from being launched (through NixOps)
<gchristensen>
LnL: :boom:
<msd_>
LnL: lol
<msd_>
domenkozar: "error: Failed to create domain from /tmp/nixops-tmpkJY_Mx/example-domain.xml ; error: invalid argument: could not find capabilities for arch=x86_64 domaintype=kvm "
<LnL>
and I even have a PR open that removes it from the darwin stdenv
<msd_>
domenkozar: ah! I get it. Just caught it. For some reason, hvm is running on arch i686, but my host is x86_64
<Mic92>
I would have expected, that he had tested the package at least before
<msd_>
domenkozar: so it looks like somehow QEMU is trying to run in 32 bit mode
_c_ has quit [(Ping timeout: 248 seconds)]
<domenkozar>
usualy that means wrong executable for qemu
<Mic92>
jaym: useage is the following btw: systemPackges = [ rustNightlyBin.rustc rustNightlyBin.cargo ];
<Mic92>
*usage *systemPackages
indi_ has joined #nixos
sellout- has joined #nixos
_c_ has joined #nixos
<msd_>
domenkozar: AH! I actually found it. dmidecode's trying to be pulled by libvirtd
itsN1X has joined #nixos
mizu_no_oto has quit [(Quit: Computer has gone to sleep.)]
bfrog_ has quit [(Ping timeout: 240 seconds)]
thc202 has quit [(Ping timeout: 272 seconds)]
fritschy has quit [(Quit: Leaving.)]
thc202 has joined #nixos
cjhowe has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
cjhowe has joined #nixos
jensens has joined #nixos
Mercuria1Alchemi has joined #nixos
ryantrinkle has joined #nixos
johnsonav has quit [(Ping timeout: 258 seconds)]
danharaj has joined #nixos
<jaym>
Mic92:
<jaym>
output path ‘/nix/store/c3p0z1jbsd0v3vgks2pwx78w4wia3m2y-rust-nightly-x86_64-unknown-linux-gnu.tar.gz’ has sha256 hash ‘05bppmc6hqgv2g4x76rj95xf40x2aikqmcnql5li27rqwliyxznj’ when ‘1afsqaavhwiaxm38zr08cbq0985c7lrb1jzdcmq0jh6y8rb8ikff’ was expected
<jaym>
cannot build derivation ‘/nix/store/3sgkzz8gbrs5s63z9dqs2djp65fhbj20-cargo-nightly-2016-12-29.drv’: 1 dependencies couldn't be built
<jaym>
error: build of ‘/nix/store/3sgkzz8gbrs5s63z9dqs2djp65fhbj20-cargo-nightly-2016-12-29.drv’ failed
<itsN1X>
Where do I find the documentation channel?
<gchristensen>
itsN1X: this is the channel for everything about nix/nixos/nixpkgs, including talking about documentation :)
<LnL>
gchristensen: programs.zsh.enable = true; should enable completion and everything for you
<itsN1X>
Oh that's nice then. How can I get into contribution?
<eacameron>
domenkozar: Re: AWS lambda: Yes that's how it seemed but I was curious if there was some sort of intersection. I don't know if/how AWS Lambda handles OS-level dependencies.
<gchristensen>
itsN1X: you'd like to help with documentation?
AndroUserN1X has joined #nixos
itsN1X has quit [(Remote host closed the connection)]
<Mic92>
I think the entry level to contribute documentation for nixos is currently too high
<ronny>
Mic92: that one looks scarry at first glance
<Mic92>
domenkozar: I agree sphinx is the better option
teknico has left #nixos []
<Mic92>
domenkozar: Ideally it should be also possible without git knowledge to write documentation. We are developers, we know git. But actually git has a very inconsistent UX. Github could fix this partially.
bennofs has quit [(Ping timeout: 248 seconds)]
<domenkozar>
year 2017: Nix converts to an useful tool for docs and gets docs contributions
<domenkozar>
ideal timing for my brithday on 6 feb
<domenkozar>
:P
roconnor has joined #nixos
<gchristensen>
unfortunately I agree with edolstra :(
Lowl3v3l has quit [(Remote host closed the connection)]
<domenkozar>
gchristensen: all of it?
* gchristensen
runs away before he has to defend that statement
<gchristensen>
back in 30min ;)
jgertm_ has quit [(Ping timeout: 248 seconds)]
<Mic92>
there should a docbook template, which does generate a giant one-pager.
Lowl3v3l has joined #nixos
<domenkozar>
there was one
<domenkozar>
we split it up 2 years ago
<ronny>
Mic92: im wondering if it would e possible to use the same thigns that manage /etc to build /home/...
<domenkozar>
you might be able to dig it up and make a one-page manual
<Mic92>
domenkozar: I mean the opposite
msd_ has quit [(Remote host closed the connection)]
georges-duperon has quit [(Ping timeout: 260 seconds)]
AndroUserN1X has quit [(Ping timeout: 240 seconds)]
<Mic92>
and a sidebar navigation instead of one chapter list at the top
_rht has joined #nixos
<Mic92>
But everytime I touch xslt I regret it immedately.
bfrog_ has joined #nixos
<ronny>
domenkozar: that one looks a lot better than the last thing, but why is it not possible to use nix's own tools for that? as far as i can tell this just replicates the system switch
<NixOS_GitHub>
[nixpkgs] LnL7 closed pull request #21596: stdenv: bootstrap cmake and python on darwin (master...darwin-stdenv) https://git.io/vMm8B
<NixOS_GitHub>
nixpkgs/master 7c29295 Michael Raskin: gcl_2_6_13_pre: init at 2.6.13pre50. GCL hasn't had releases for a long time, but there are relatively regular prereleases
<NixOS_GitHub>
nixpkgs/master 95f3473 Michael Raskin: ecl_16_1_2: init at 16.1.2; reinstating for Maxima
<NixOS_GitHub>
nixpkgs/master efd2b27 Michael Raskin: sbcl_1_3_12: init at 1.3.12; reinstating the previous version for Maxima
takle has quit [(Remote host closed the connection)]
georges-duperon has joined #nixos
AndroUserN1X has joined #nixos
AndroUserN1X has quit [(Read error: Connection reset by peer)]
AndroUserN1X has joined #nixos
sellout- has quit [(Quit: Leaving.)]
reinzelmann has quit [(Ping timeout: 246 seconds)]
civodul has quit [(Quit: ERC (IRC client for Emacs 25.1.1))]
<NixOS_GitHub>
nixpkgs/master 1fe136e Vladimír Čunát: Make all meta.maintainers attributes lists...
amarsman has joined #nixos
reinzelmann has quit [(Ping timeout: 256 seconds)]
georges-duperon has quit [(Ping timeout: 240 seconds)]
indi_ has quit [(Remote host closed the connection)]
pstn has joined #nixos
ambro718 has joined #nixos
<itsN1X>
kmicu,
systemfault has quit [(Quit: Bye!)]
rly has joined #nixos
<ronny>
domenkozar: is there yet an easy way to get mutable virtualenvs on nixos that dont die on a gc?
<domenkozar>
I'm not doing Python these days so no idea, sorry
<ronny>
else i'd like to create one
<ronny>
i suppose FRidh is doing things on that now?
<simpson>
ronny: I am interested but have nothing to contribute.
<LnL>
ronny: I have some custom expressions that I use for work, but it I have to recreate them after a reboot
<LnL>
a gc I mean
<ronny>
LnL: main reason is that virtualenv calls dont add gc roots to the used pythons
<LnL>
my setup actually uses a nix buildEnv for most dependencies
c0bw3b has joined #nixos
<LnL>
except for some special dependencies
<ronny>
LnL: still no gc root added
<ronny>
buildEnv !0 gc root
<LnL>
nix-instantiate --indirect --add-root $PWD/result.drv -A env
<ronny>
a rdv file is also no gc root
zagy has joined #nixos
<ronny>
*drv
habibi has joined #nixos
<LnL>
that caches everything for me
<itsN1X>
kmicu: links?
<ronny>
LnL: a file that lists some hashes is not a gc root
<LnL>
ronny: --add-root makes it a gc root
<LnL>
and since it's a drv all the buildInputs it needs will stay allive
ertes has quit [(Ping timeout: 264 seconds)]
<ronny>
LnL: so why do you need a recreate after gc?
dgn has quit [(Ping timeout: 248 seconds)]
<LnL>
ronny: because some of the stuff in the virtualenv is impure
zagy1 has joined #nixos
<ronny>
LnL: huh?
zagy has quit [(Ping timeout: 245 seconds)]
<ronny>
LnL: why would that matter, a virtualenv is supposed to be outside of the nix store so you can modify it to begin with
<LnL>
ronny: it refers to some nix stuff
<LnL>
that might disappear
alx741 has quit [(Ping timeout: 240 seconds)]
<ronny>
LnL: why would it disappear if a gc root needs it?
msd has joined #nixos
AndroUserN1X has joined #nixos
itsN1X has quit [(Read error: Connection reset by peer)]
<ronny>
LnL: because that would be like a major bug in the gc
reinzelmann has joined #nixos
<LnL>
ronny: I can give you an example in 30min
danharaj has quit [(Quit: Leaving.)]
danharaj has joined #nixos
<johnw>
can I turn {a = 1; b = 2;} into [a b]?
<johnw>
ah, attrNames
<LnL>
attrNames
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
<ronny>
LnL: k, thx
<peti>
ocharles_: Yes, that's a good solution. Those interfaces are not going to change in the forseeable future.
<ocharles_>
Cool, thanks!
tsurai has joined #nixos
Shou has quit [(Ping timeout: 256 seconds)]
sellout- has joined #nixos
michaelpj has quit [(Ping timeout: 256 seconds)]
<tsurai>
hey, I'm trying to install nixos from an usb stick on a thinkpad x220 but get an error in stage1 that it cant mount /mnt-root when booting from the stick
reinzelmann has quit [(Ping timeout: 245 seconds)]
<johnw>
eacameron: no, I haven't, but snapshot sending/receiving works over ssh; it's just a data stream like any other
<Mic92>
I only use zfs snapshots for mirroring to a new location
<eacameron>
johnw: Are there any compatibility considerations with ZFS? I've never used it.
<johnw>
a filesystem at a particular rev/version is generally compatible with other ZFS implementations that advertise support for the same rev/version
<Mic92>
for backups in combination with auto snapshots the backup performance is slow.
<johnw>
i've read/written to/from OS X-built ZFS filesystems in NixOS
<johnw>
Mic92: are you doing it incrementally? those are *fast*
<Mic92>
johnw: but If I use autosnapshot all intermediate snapshots have to replicated to my backup. This is much slower then just using rsync for the latest version
<johnw>
you shouldn't need all intermediates
<johnw>
just at a minimum the two endpoints of the transfer
<Mic92>
johnw: you have to, otherwise incremental snapshots do not work
<johnw>
if both have "base", and you're sending "foo", it doesn't matter how many snapshots there have been between base and foo
<johnw>
Mic92: I've done it; maybe we're not talking about the same thing
<johnw>
zfs send -I will send whatever has changed since "base", in this example
<LnL>
johnw: talking about zfs, does the zfs-mount service run properly at boot for you?
<Mic92>
johnw: I see all my autosnapshots also in the destination
georges-duperon has joined #nixos
<johnw>
LnL: no, I added this to postBootCommands: ${pkgs.linuxPackages.zfs}/bin/zfs mount -a
<LnL>
johnw: ah, I'll probably look into that sometime then
_c_ has quit [(Ping timeout: 248 seconds)]
reinzelmann has quit [(Quit: Leaving)]
<Mic92>
LnL: If you mount other filesystems on top of zfs, I would add all mounts manually to hardware-configuration.nix. Otherwise there there are race condition between systemd mount generator and zfs mount
<Mic92>
It is also actually faster. Because the zfs kernel module call out the userspace mount executable due license issue with vfs_mount in the kernel.
<eacameron>
So with ZFS: would the recommended approach be to mount a ZFS volume and configure your apps and DBs to store all their state on that volume?
<LnL>
well I don't really want to add a new entry everytime I create a new fs
<Mic92>
zfs assumption about mouting are acient. We have ude and hotplugging these days.
<ronny>
FRidh: i noticed that the metadata for python packages is wrong, listing the position of the library instead of the position of the package definition
<Mic92>
ancient
<cheshircat>
hey, I run a dualboot nixos and windows 10, and a recent windows update seemed to have messed with my EFI startup partition because now I can only boot into windows
<Mic92>
+udev
<Mic92>
eacameron: I found zfs as a root filesystem more flexible.
tsurai has quit [(Quit: Leaving)]
<cheshircat>
I’ve tried booting from a usb, but it wont show up in my boot menu
<LnL>
I do have /nix/store on a separate fs with a lower quota so I cant fill the other filesystems to 100%
jarlg has quit [(Ping timeout: 256 seconds)]
<LnL>
you can't run gc when /nix/var is full for example
<eacameron>
Mic92: Hmm...I've got some production servers that don't really have a good backup story and I'm guessing switching to zfs as root fs would be pretty disruptive...but maybe I'll try it locally.
Sonarpulse-Work has joined #nixos
<eacameron>
cheshircat: Sometimes BIOS doesn't look a USB by default. You may have to configure your BIOS to do that.
<FRidh>
ronny: you should be able to pass mkDerivation parameters, so just pass dontStrip
rly has quit [(Ping timeout: 272 seconds)]
<cheshircat>
OK, thank you eacameron
<LnL>
ronny: ping
zraexy has joined #nixos
<ronny>
LnL: pong
<LnL>
a bit of context, we have an internal pypi server at work so I don't use any of the python stuff from nixpkgs, my pythonPackages set is basically a set of derivations that result in a python wheel
<LnL>
the pure part is the buildEnv that unpacks a list of wheels into a site-packages directory, this gets added to my PYTHONPATH in nix-shell
<LnL>
and the other part is a virtualenv where a bunch of wheels get installed when you enter the nix-shell
<ronny>
i see, then what you describe is a total hack and im not surprised it falls appart
<LnL>
exactly
<LnL>
so I install as little as possible into the actual virtualenv
<LnL>
that part sometimes breaks, but it's a maximum of 3 dependencies so it's not to bad
jarlg has joined #nixos
mudri` has joined #nixos
pstn has joined #nixos
<ronny>
LnL: well, in that case its completely useless to me and what i want to do ^^
<LnL>
I could probably get it working properly, but it's just for me
<LnL>
nobody uses nix at work so I have not spent the time to make it better
<Mic92>
eacameron: I run my containers also for long time on a non-root zfs. But i guess you should get a backup first before switching a filesystem. Btw. zfs is also great as a backup target. Just rsync the data to your storage and do a snapshot.
<savanni>
Hey, y'all. Sorry i"ve been so absent. But I have one really quick quesiton.
eacameron has quit [(Read error: Connection reset by peer)]
eacameron has joined #nixos
<savanni>
I now that there is a particular package path `nixpkgs.haskell.compiler.ghc7103`. How do I use nix-env to list everything in `nixpkgs.haskell.compiler`?
habibi has quit [(Quit: My MacBook Air has gone to sleep. ZZZzzz…)]
<LnL>
savanni: nix-env -f '<nixpkgs>' -qaP -A haskell.packages.ghc7103
<savanni>
*AH*. It's the -f that I was missing.
<viric>
wasn't there once a "man proc" and "man ld.so"? Where are they?
<LnL>
savanni: you can also use nixpkgs.haskell... or nixos.haskell... instead, but that depends on if you are on nixos or not
<savanni>
I'm on nixos, but for whatever reason that wasn't working...
<savanni>
`nix-env -A nixos.haskell.compiler -qaP` -- this always said that `nixos.haskell.compiler` matched no derivations.
<LnL>
savanni: haskell.packages, not haskell.compiler
Elon has quit [(Remote host closed the connection)]
<viric>
ah, in manpages. Got it
<viric>
(damn ~/.nix-profile/man things)
cpennington has quit [(Ping timeout: 248 seconds)]
<savanni>
Well... nixos.haskell.compiler gives me no error, but no result.
<savanni>
Either way, though, I found the information I needed.
<stepcut>
when using nixops to administer a machine, is there a way to get nix-env to show all the packages that are installed via environment.systemPackages?
<stepcut>
oh.. I think i found it in the manual, nix-store -q --references /var/run/current-system/sw
peter-holm has joined #nixos
<hyper_ch>
hmmm, is something wrong with Chromium=
digitalmentat has quit [(Ping timeout: 245 seconds)]
<NixOS_GitHub>
[nixpkgs] chris-martin opened pull request #21622: Add some more details about the nix.useSandbox option (master...pr/useSandbox-doc) https://github.com/NixOS/nixpkgs/pull/21622
danielrf has quit [(Read error: Connection reset by peer)]
<eacameron>
Mic92: johnw: is there a good way to port an entire machine to use ZFS as the root FS?
Happy_Ente has joined #nixos
<johnw>
eacameron: no, that I don't know anything about
<johnw>
I've never used ZFS as a boot volume before, except when I was running an OpenIndiana box, and then no migration was needed
gchristensen has joined #nixos
tv has joined #nixos
<Mic92>
eacameron: you will need intermediate storage.
jp302 has joined #nixos
baldo[m] has joined #nixos
rycee[m] has joined #nixos
M41px[m] has joined #nixos
av[m] has joined #nixos
Kiko[m] has joined #nixos
M-Rym has joined #nixos
cinderfish[m] has joined #nixos
cstrahan has quit [(Ping timeout: 248 seconds)]
unlmtd[m] has joined #nixos
nhooyr has joined #nixos
SShrike has joined #nixos
MrAxilus[m] has joined #nixos
musicmatze[m] has joined #nixos
schoppenhauer has joined #nixos
eddyb has joined #nixos
grill_ has quit [(Ping timeout: 256 seconds)]
grill_ has joined #nixos
msd has quit [(Remote host closed the connection)]
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
msd has joined #nixos
msd has quit [(Client Quit)]
ragga has joined #nixos
<ragga>
first day of using nixos and i can't wrap my head around splitting up my config.
<ragga>
when using two files where one is device specific and the other one has "common" expressions, how can i used stuff that is defined in the common in my device specific one?
<ragga>
do i need to declare options and configs in the common one?
<LnL>
use the imports, just like the hardware-configuration
<ragga>
i always get error: undefined variable ‘commonPkgs’ when issuing a rebuild
<LnL>
and no {} is a valid module
<LnL>
what's commonPkgs in you configuration?
<ragga>
a list of packages that i'd like to concat in the device.nix for using environment.systemPkgs
cstrahan has joined #nixos
<ragga>
device.nix imports common.nix where commonPkgs = with pkgs; [] is defnied
ocharles_ has joined #nixos
<LnL>
I assume you used commonPkgs somewhere instead of { config, ... }: { foo = config.commonPkgs; }
takle has quit [(Remote host closed the connection)]
<peter-holm>
How do I mount a btrfs subvolume within configuration.nix ?
<eacameron>
Mic92: johnw: Can you recommend any cloud-based service (or similar) that will host your backups for you (ideally supporting ZFS snapshots as well)?
<peter-holm>
Thankyou mguentner
<peter-holm>
I used afterburst vps in the pasd.
Itkovian has joined #nixos
<peter-holm>
eacameron, In my opinion (which cab be wrong), an vps is far better than cloud-based storage. if you want to store a little larger files.
<eacameron>
peter-holm: Interesting...why use a whole VPS when all I need is a disk?
Shou has quit [(Quit: 🛏)]
<peter-holm>
Because cloud-storage ar often designed for the office.-guy who want to store bnotes and spreadsheats. There are not designed to handle large files. Nor are large files allowed on most free cloudbased environments.
ryanartecona has quit [(Quit: ryanartecona)]
<peter-holm>
If you go for the cloud, you better read up on how large files are allowed - and then split the images.
<eacameron>
peter-holm: What about things like S3 or Google Cloud storage (instead of Dropbox, Google Drive, etc.)
AllanEsp1nosa has joined #nixos
sboosali has joined #nixos
<peter-holm>
I am not familjar with them for store bif files . If you figure out, please tell me :-) . But I assume that what I said about the 'office-guy' is not very far away from the true about thoose services.
<Mic92>
eacameron: I never stored backups as zfs snapshots at an online service, because I usally want to encrypt my backups before uploading them to NSA. I currently use attic-backup and store the stuff on my own infrastructure. Apart from that I setuped once S3-Backups with a policy to archive to glacier for a customer. Then there is also backplace, rsync.net, online.net's c14... But I used none of them so fwar.
<Mic92>
*far
<sboosali>
hi, how can I override an Emacs package to use a local file? my emacs.nix: http://pastebin.com/STg8Jpv6
<Mic92>
*backblaze
<Mic92>
I mean borgbackup (fork of attic)
<eacameron>
Mic92: I hear the NSA prefers your files encrypted...helps them know its worth looking at. ;)
<Mic92>
eacameron: well, all my machines except my router have their rootfs encrypted.
<Mic92>
aes-ni makes it pretty cheap these days.
<eacameron>
Mic92: aes-ni?
[0x4A6F] has joined #nixos
<Mic92>
cpu based hardware acceleration for aes, comes by default in modern cpus
thc202 has joined #nixos
Sonarpulse-Work has quit [(Ping timeout: 256 seconds)]
zraexy has joined #nixos
nanobapt has joined #nixos
takle has joined #nixos
<eacameron>
Mic92: Ah. Sweet.
<eacameron>
Mic92: johnw: how do you guys update ZFS version after you're running on it?
takle has quit [(Read error: Connection reset by peer)]
<Mic92>
eacameron: cool, I was not aware for that.
takle has joined #nixos
<mguentner>
eacameron: thanks for that link
<eacameron>
Mic92: mguentner: You bet. :)
<Mic92>
eacameron: I use 0.6.5.8 and 0.7.0-rc2
<peter-holm>
If one count the costs, the better choice is a good harddrive and that-old-pc-you-dont-use-anymore.
<gchristensen>
depends how much your electricity costs
<Mic92>
peter-holm: well, and the restore time can be a problem...
jensens has quit [(Ping timeout: 246 seconds)]
Sonarpulse-Work has joined #nixos
<peter-holm>
hm. i was just looking at pricing for rsync.net, 8 cent pr gigabyte. If I long.-term store 999 Gb, then I have to pay 7982 dollar per month.
zraexy has quit [(Ping timeout: 248 seconds)]
<eacameron>
peter-holm: attic/borg is 3 cents/gig/month
<peter-holm>
thats much better.
<eacameron>
peter-holm: Honestly I'd probably only keep the most recent month or so on something as robust as rsync.net
<eacameron>
peter-holm: after that it's going to the old pc that one you don't use any more ;)
ragga has quit [(Ping timeout: 260 seconds)]
<Mic92>
peter-holm: it is actually $79.92 not $7982.
civodul has joined #nixos
sboosali has quit [(Remote host closed the connection)]
<eacameron>
Has anyone mounted a local drive on a remote NixOS system (i.e. in VirtualBox, etc)?
<gchristensen>
can you be more clear about specifically what you're looking to do?
sigmundv__ has joined #nixos
simendsjo has quit [(Remote host closed the connection)]
<eacameron>
gchristensen: I want to use nixops to deploy a VirtualBox machine that has a mount to my local templates so I can update them and see them live in the web app. Essentially replace vagrant with nixops.
wak-work has joined #nixos
aarvar has joined #nixos
ryantrinkle has quit [(Ping timeout: 272 seconds)]
<LnL>
eacameron: I think you can specify shares with nixops an you can mount them with the fileSystem module (fsType = "vboxsh"
<LnL>
)
<LnL>
no the type is vboxsf
<eacameron>
LnL: Ah that would be sweet to get working. Thanks I'll tinker with that.
mkoenig has quit [(Read error: Connection reset by peer)]
zagy has joined #nixos
zagy has quit [(Client Quit)]
<Ralith>
eacameron: sincerely, that's exactly why it's important to encrypt all your worthless stuff
<eacameron>
Ralith: Heh...you're likely right.
<copumpkin>
niksnut, shlevy: I might have regression in 1.11.5
reinzelmann has quit [(Quit: Leaving)]
<gchristensen>
oh hi copumpkin!
<copumpkin>
niksnut: is there a good way to install a specific version of Nix (like 1.11.4) until I get a fix in?
<copumpkin>
hi gchristensen :)
<martijn923>
I always forget this and it's not easy to Google for: how do you escape variable interpolation in strings?
<gchristensen>
how were your holidays?
mkoenig has joined #nixos
<copumpkin>
martijn923: ''${ or something
<copumpkin>
gchristensen: not very relaxing, but fun :)
<copumpkin>
ended up getting no time to code :(
<gchristensen>
aye... I know the feeling
aarvar has quit [(Ping timeout: 260 seconds)]
<martijn923>
copumpkin: thanks
MichaelRaskin has joined #nixos
mizu_no_oto has joined #nixos
ptrrr has joined #nixos
<copumpkin>
I think 1.11.5 is the first version with the fully refactored c++ tools, right?
<copumpkin>
instead of perl scripts?
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
Lowl3v3l has quit [(Remote host closed the connection)]
<MichaelRaskin>
I think so.
<gchristensen>
I think everyone except those using nix-on-darwin
<MichaelRaskin>
(there are so many lurkers that it is hard to estimate)
<MichaelRaskin>
gchristensen: I feel discriminated
<gchristensen>
some of us deploy nixos to servers, laptops, desktops, a few people here use it in HPC and at least one person deploys it to over 50k servers a month
<Devices>
oh wow
<Devices>
didn't know it was used so much
<MichaelRaskin>
There _are_ people who use Nix-on-whatever-Linux
<dash>
i use nix on ubuntu
<Devices>
reading the site now and it says Nix and Hydra are PhD projects
<gchristensen>
true, good point, sorry :) some people use nix-on-not-nixos
<dash>
Devices: yes but don't let that fool you, they're actually really good
<Devices>
how does Nix work on other distros?
<gchristensen>
Devices: linux is the project of some dude in helsinki because he was cheap
<dash>
Devices: real good!
<MichaelRaskin>
And I guess my system contains slightly less NixOS parts than duct-taped around parts.
<Devices>
can you just replace say apt or dnf?
FareTower has joined #nixos
jmiven has joined #nixos
<MichaelRaskin>
Not replace
<gchristensen>
Devices: for user packages, yes, for system packages, then you might a well use nixos.
<LnL>
gchristensen: what question are you referring to?
<dash>
Devices: no it doesn't replace them, if you use nix on non-nixos, it puts stuff only in /nix and makes symlinks in your homedir
<MichaelRaskin>
If you have a distro, it assumes that the boot-relevant core is managed by the native package manager
<gchristensen>
Devices: I like that I can do `nix-shell -p rustUnstable.rustc` and have a shell with unstable rust ... and then run `nix-shell -p rustc` and have another shell with stable rust, and I can have them both open at the same time, using both of them at the same time, without having to fight my distro
<gchristensen>
no "version switching" or whatever
<Devices>
oh coolio
fare__ has quit [(Ping timeout: 248 seconds)]
<MichaelRaskin>
At some point I wondered how many glibcs I had installed. I had three, they were all used, and this created no hassle
<gchristensen>
Devices: I can get any package I want and use it locally without installing it "globally" or having /etc mucked with, or screwing up any other programs I have installed.
<gchristensen>
it doesn't matter the version or compatibility, you can load anything in to a nix-shell and it won't hurt your system.
<gchristensen>
I've never felt I had to reinstall because a major upgrade was likely to be broken
<MichaelRaskin>
Also, rollbacks for everything, if you don't actively do a GC
<guillaum1>
Something not clear for me, if I' adding the channel nixpkgs-unstable, should I remove the channel nixos-16.09 ?
<MichaelRaskin>
That allows you to pay sanity damage with HDD space gigabytes
<Devices>
yeah the rollback feature looks pretty cool
<gchristensen>
guillaum1: :danger will robinson: you don't want to use nixpkgs-unstable when running nixos
<peter-holm>
Hi, i am trying to mount a btrfs subvolume.. mguentner gave me an an exampel how that should be done. But this part 'options = [ "subvol = root/sysroot/nixos-mountpoints/users/home" ];' gives 'error: syntax error, unexpected '=', expecting ';', at /etc/nixos/configuration.nix:42:22'
<Devices>
do you happen to know how it compares to zypper and btrfs on openSUSE?
<Devices>
nixOS allows you to choose at boot, does openSUSE do that too?
<gchristensen>
Devices: our roll-backs, which you can choose at boot, are also your configuration
mbrgm has quit [(Ping timeout: 246 seconds)]
<gchristensen>
so if you deploy a broken config that breaks everything, you can roll back everythin g-- packages and config -- by rebooting
<gchristensen>
I doubt zypper does
<Devices>
I got the idea it did that too, but don't know specifics
<guillaum1>
gchristensen: yes, thank you.
<Devices>
it makes a snapshot every update and you can roll back, but dunno how really
<gchristensen>
guillaum1: I'm still working on this, I want it to become part of the normal docs
justanotheruser has quit [(Ping timeout: 256 seconds)]
mbrgm has joined #nixos
<MichaelRaskin>
Devices: what about rolling back a botched update, removing a single package, and updating forward again, with almost all the work succesfully reused?
<guillaum1>
gchristensen: the why is interesting, and I'll follow this advice for now, but it does not really answer my question, are we supposed to activate many channels (such as 16.. and unstable) or we are supposed to activate only one ?
<Devices>
it seems like it can from a quick glance MichaelRaskin
<gchristensen>
seems weird to require a filesystem feature to support this
justanotheruser has joined #nixos
<MichaelRaskin>
I am not sure where Snapper reuses work from botched updates
<MichaelRaskin>
Nix installs everything in an own path, so if anything is the same between two configurations, it will get reused (even if the resulting full configuration is unusable for unrelated reasons)
<lassulus>
peter-holm: maybe upload your config to a pastebin? its really hard to get your question inline
<gchristensen>
Devices: you should try it in a VM :)
<Devices>
yeah I should
<gchristensen>
or go whole-hog and install it on your system ... but a VM is easy
<Devices>
I'm running fedora atm, but nixOS looks kinda cool