<krey>
ryantm: where is "sage" looked up? nix-env is having trouble finding the derivation
hakujin has joined #nixos
<ryantm>
krey: It looks like it is looking it up in your user environment. I did `nix-env -iA emacs` and after that I could do `nix-env --set-flag priority 10 emacs`
<ryantm>
krey: Maybe you haven't installed sage into your user environment?
hakujin4 has quit [Ping timeout: 276 seconds]
<krey>
ryantm: yeah, it's in my systemPackages, not in my user environment
<krey>
ryantm: maybe it shouldn't be in my systemPackages, I guess
<krey>
ryantm: it's not really a "system package"
dynamicudpate has quit [Read error: Connection reset by peer]
<ryantm>
krey: Yeah, I'd move that to your user env.
<tilpner>
sphalerite - I haven't verified the values it outputs. I mostly use it to answer "why is my system slow, did it swap out important things again?"
<sphalerite>
fair enough
Turion has joined #nixos
zzamboni has quit [Client Quit]
davidlt_ has joined #nixos
xy2_ has quit [Ping timeout: 256 seconds]
vidbina has joined #nixos
<Guest29>
Quick question: I'm trying to get helm-2.7.0 installed; what is a good way to go about finding a derivation in the nixpkgs repo with the correct version number?
<Turion>
I'd have thought "nix-env -qaP 'helm'", but I can't test, something is weirdly broken
<Turion>
Why am I getting "error: undefined variable ‘placeholder’ at /nix/store/w4y93fsqvj9j6g6i5grm6cdb9nq5..." when running the above mentioned command?
hakujin1 has joined #nixos
<Guest29>
Did you get "error: undefined variable 'placeholder' ..." when you ran that?
zzamboni has joined #nixos
<Turion>
Yes
<tilpner>
That's a sign you need to update to Nix 2.0
<Turion>
I just updated channels and tried to nixos-rebuild switch
<Guest29>
Ah
<Turion>
but that failed because some perl module didn't build
<Turion>
tilpner, ah! so if I install it as user...
<Guest29>
correct me if I'm wrong, but I can do that with "nix-channel --update" and "sudo nixos-rebuild switch"
<tilpner>
That depends on what channel you're on
<Turion>
nixos-unstable
<Guest29>
same
<Turion>
Guest29, yes, that's what I did
<tilpner>
nixos-17_09.nixStable is 1.11.16, but nixos-unstable.nixStable is 2.0
<Guest29>
wait, I'm on nixpkgs-unstable
<tilpner>
Are you running NixOS from that, Guest29?
<Guest29>
yes
hakujin has quit [Ping timeout: 256 seconds]
<tilpner>
You should not do that. Prefer nixos-unstable
<tilpner>
nixpkgs-unstable doesn't receive NixOS specific testing, so it's much more likely to be broken
<Turion>
tilpner, after a successful nix-env -iA nixos.nix, which installed nix-2.0, I'm getting the same error
<clever>
Turion: nix-env cant change the nix version on nixos
zzamboni has quit [Client Quit]
<tilpner>
Your old 1.11.* daemon is still running in the background
<clever>
you must change it via nix.package in configuration.nix, and any attempt at installing it with nix-env will just cause problems
<elvishjerricco>
clever: It can change the version of the evaluator that users run, but not the daemon, right?
zzamboni has joined #nixos
<clever>
elvishjerricco: right
dvim has joined #nixos
<elvishjerricco>
I've always been a little fuzzy on how much happens in the evaluator and how much happens in the daemon
<Turion>
tilpner, clever, oops... is this mentioned somewhere in the docs?
<clever>
elvishjerricco: it also causes other fun problems, somebody in here has 1.11 installed with nix-env, but 2.0 as the daemon, so NIX_REMOTE was no longer setup, which broke the 1.11 nix
<clever>
elvishjerricco: and also, some of the 2.0 features need a 2.0 daemon, and fail with weird errors
<Turion>
Couldn't there be a version check when the evaluator connects to the daemon?
<clever>
elvishjerricco: nearly everything nix-instantiate does is client side, and it generates a tree of .drv files (which it has to forward to the daemon to put in the store)
<clever>
elvishjerricco: then nix-store -r asks the daemon to build those drv files
<elvishjerricco>
So the daemon is 100% responsible for substitution?
<elvishjerricco>
with flags passed by the client?
<clever>
and if the process running has +w to /nix/store, it just bypasses the daemon and forks out the same worker the daemon would have made
<clever>
yeah
<Guest29>
clever, do I update my changing system.stateVersion?
<clever>
Guest29: never
<Turion>
clever, so I can meaningfully install nixos.nix as root?
<elvishjerricco>
That's a mistake I made before I knew what stateVersion was for
<clever>
Turion: that still wont change the daemon that systemd runs
<elvishjerricco>
should probably be better named
davidlt_ has quit [Remote host closed the connection]
<clever>
Turion: you need to set the nix.package entry in configuration.nix to properly update it
zzamboni has quit [Client Quit]
davidlt_ has joined #nixos
<Turion>
clever, but if it can bypass the daemon with the correct privileges?
<Guest29>
then what do I set nix.package to? What exactly is stateVersion for?
<clever>
Turion: it would only work for root
<Turion>
Well, yes, that's what I mean
<elvishjerricco>
Guest29: stateVersion is the version of your state, not the statement of your version :P
<elvishjerricco>
some services need to know what version their state is from
<elvishjerricco>
easiest example being services.postgres
zzamboni has joined #nixos
<Turion>
Ah, then the other users of course don't have nix-2.0 in their path
<clever>
postgresql has 2 different incompatible changes it has gone thru
<clever>
Turion: all users inherit roots nix-env packages, but systemd doesnt
<elvishjerricco>
Guest29: Right. The name is misleading to some people IMO.
<jluttine>
hmm.. how to use github pull requests as patches in a package derivation?
<clever>
jluttine: if you add .patch to a URL for a commit, github gives it to you in patch form, but applying that patch to nixpkgs isnt so simple
<clever>
you would need to import an unpatched nixpkgs, then use a derivation to apply the patch to a copy of nixpkgs, then import-from-derivation that result
<elvishjerricco>
clever: I think jluttine just wants to patch a specific derivation, not nixpkgs
<Turion>
jluttine, a patch of the source or of nixpkgs?
<Turion>
yes, my thought
zzamboni has quit [Client Quit]
<jluttine>
yes, a patch for a particular package from its upstream github repo
<jluttine>
so not a patch for nixpkgs itself
zzamboni has joined #nixos
<jluttine>
i'm modifying a package derivation and would like to use a few patches from the package's upstream pull requests
<jluttine>
because it can take anything from 1 day to 1 year until they are merged
asuryawanshi has quit [Remote host closed the connection]
<bkchr[m]>
jtojnar: Could you send me your patch for `gst-python` again? Your link is not valid anymore :/
<elvishjerricco>
jluttine: Oh I just realized: PRs are mutable
Tucky has joined #nixos
ertes-w has joined #nixos
<elvishjerricco>
So `github.com/.../pulls/1.patch` will have potentially mutating contents
<{^_^}>
[nixpkgs] @andrew-d closed pull request #29556 → nixos/nginx: use the directory given in the ACME configuration → https://git.io/v5Nbm
<Guest29>
elvishjerricoo, how do I update my channel for that to work? A `sudo nixos-rebuild switch` tells me nixStable2 is a missing attribute. I'm on nixos-unstabl
<clever>
Guest29: when did you last do a nix-channel --update as root?
<Guest29>
awhile...
<Guest29>
that might be the problem, always do it as my user
<Guest29>
elvishjerricco, makes sense, just didn't think about it. Thanks
<elvishjerricco>
You'll have to update the channel. `sudo nixos-rebuild --upgrade switch`
<elvishjerricco>
convenient shorthand
<Guest29>
replaces `sudo nixos-rebuild switch` with `sudo nixos-rebuild --upgrade switch` in my fish history. Future me will thank me
orivej has joined #nixos
<Turion>
clever, ah ok :)
<jluttine>
elvishjerricco: ok, thanks!
<Turion>
How can I ever update system.stateVersion?
<Turion>
Or can it ever go out of date?
<clever>
Turion: that would break the very thing its meant to fix
<clever>
Turion: it must never be changed
<bkchr[m]>
jtojnar: ty, will build it :)
<Turion>
clever, and in 50 years?
<clever>
Turion: its meant to track what version you originally installed, so non-compatible changes in nixos can be worked around
<Turion>
Yes sure, but that means that all future releases somehow have to deal with how older releases managed state. Won't that be a burden at some point?
<clever>
Turion: possibly
<elvishjerricco>
Turion: No more a burden than maintaining a full history of migrations
zzamboni has joined #nixos
<Turion>
elvishjerricco, well ok, if I want to rollback indefinitely, yes. What if I'm ok with e.g. updating my postgresql databases at some point and discard all older generations?
<clever>
yeah, if you deprecate a certain value of it, you would need to make something to migrate the state, and then you have the same issue
<clever>
Turion: the first incompatible change postgresql had, i think was the on-disk format of the db, you have to export, upgrade, wipe, and import
<clever>
Turion: the 2nd one, is that the admin role that has all control got renamed, so if you know how to use it, you can just create a new one
pkill9 has joined #nixos
<Turion>
clever, the db is really just the table of installed packages (via nix-env)?
<clever>
Turion: postgress is not used by nix at all
<clever>
so if your not using postgress, then that specific problem wont happen for you when changing the stateVersion
<clever>
but other services have similar issues that need to be worked around
<bkchr[m]>
jtojnar: yeah building works for me. What else do I need to do for this pr?
aarvar has quit [Ping timeout: 260 seconds]
<Turion>
Ah, ok... and I guess there is no comprehensive list of services that make use of this variable?
<clever>
Turion: grep nixpkgs for stateVersion
<clever>
thats about the only list i know of
<Turion>
Ok
John882 has joined #nixos
zzamboni has quit [Ping timeout: 256 seconds]
John882 has left #nixos [#nixos]
zzamboni has joined #nixos
<tilpner>
Looks like caddy, mxsql, postgresl, ipfs, matrix-synapse, and radicale
<jtojnar>
bkchr[m]: can you try if the absolute directories fixed webkitgtk build? and if yes, whether gnome3.gnome-music (or other app using gst-python) works?
<clever>
tilpner: the sshd work-around is gone, and it your install was from 2015, the ssh hostkeys will just change on their own and set off mitm alerts
<tilpner>
:/
<Turion>
Hmm
<Turion>
So 18.03 is not officially released yet, but I can already use it as a channel
<clever>
Turion: its in more of a beta phase right now
<Turion>
Ok, then I'll gladly beta test now :)
zzamboni has quit [Remote host closed the connection]
<fearlessKim[m]>
I am a bit annoyed by this makeWrapper error, it would be cool if someone had an advice on how to fix (I can do the fix) https://github.com/NixOS/nixpkgs/issues/37968
<Turion>
Why do I get "warning: unknown setting 'signed-binary-caches" three times when running nixos-rebuild switch? (already in the "building nix" phase)
periklis has quit [Ping timeout: 256 seconds]
zzamboni has joined #nixos
<clever>
Turion: because your using nix 1.11 config with nix 2.0
<clever>
Turion: after the rebuild finishes, your nix config will be updated
<Turion>
Ah ok
zzamboni has quit [Client Quit]
zzamboni has joined #nixos
zzamboni has quit [Client Quit]
zzamboni has joined #nixos
zzamboni has quit [Client Quit]
goibhniu has joined #nixos
obadz has quit [Quit: WeeChat 2.0]
zzamboni has joined #nixos
zzamboni has quit [Client Quit]
<Turion>
How do I permanently change the niceness of the nix daemon?
periklis has joined #nixos
civodul has joined #nixos
zzamboni has joined #nixos
obadz has joined #nixos
obadz has left #nixos [#nixos]
obadz has joined #nixos
Turion has left #nixos [#nixos]
<LnL>
about the 1.11 -> 2.0 upgrade is there some way we could do that more atomically?
hakujin2 has joined #nixos
<LnL>
yesterday I helped somebody with that because he used an url import in his config
jluttine has quit [Remote host closed the connection]
xy2_ has joined #nixos
jluttine has joined #nixos
<{^_^}>
[nixpkgs] @andrestylianos opened pull request #38380 → nixos/doc: fix instructions for nox-review usage → https://git.io/vxPXS
<{^_^}>
[nixpkgs] @xeji opened pull request #38381 → mailutils: disable parallel building → https://git.io/vxPX9
__monty__ has joined #nixos
hakujin1 has joined #nixos
<__monty__>
Is there a way to get nix installed programs to see libraries that are installed on the host? I'd like to get avahi local domain resolution working for firefox for example but I'm not on nixos so I can't just enable the service.
alex`` has joined #nixos
hakujin has quit [Ping timeout: 276 seconds]
<{^_^}>
[nixpkgs] @adisbladis pushed commit from Yannik Sander to master « krunner-pass: init at version 1.3.0 »: https://git.io/vxP1s
<makefu>
__monty__: this is against the nix philosophy. however you could enable avahi and if the bindings are compiled into firefox name resolution should "just work"(tm)
<__monty__>
makefu: But it doesn't.
<{^_^}>
[nixpkgs] @adisbladis merged pull request #37494 → krunner-pass: init at version v1.3.0 → https://git.io/vxnFn
<makefu>
does avahi actually work for you on the host system?
<__monty__>
Nix-installed firefox can't resolve the .local names. Pacman installed firefox or downloaded firefox can.
<makefu>
okay, let me check what the default build recipe is for firefox
<makefu>
mh there is nothing which indicates that there are extra libs in use
raynold has quit [Quit: Connection closed for inactivity]
<makefu>
tbh i never used avahi together with firefox
<__monty__>
Opera and ping show the same problem. Can't resolve .local domains when installed through nix.
<__monty__>
Note that I'm not on nixos, not claiming anything about nixos since it might just work there.
<makefu>
__monty__: i've just tested it and it seems that avahi works correctly on my machine but it is not used in firefox for resolution
<__monty__>
I don't think firefox touches avahi, it uses mDNS.
<makefu>
it might be the case that it never worked, can you do 'ldd /path/to/pacman-installed/firefox'
<__monty__>
"not a dynamic executable"
<makefu>
try `cat /pacman-installed/firefox` and follow the rabbit hole
<makefu>
everything looks normal, maybe the nix build is missing some magic mdns build option but i cannot see it on the spot. maybe somebody else can confirm that the local resolution actually works in nixos
<tilpner>
And are you sure it's using /home/grahamc/.nix-defexpr/channels instead of nixpkgs=/nix/var/nix/profiles/per-user/root/channels/nixos/nixpkg?
<gchristensen>
nixpkgs= _is_ defined, but when I do `nix build nixpkgs.git` it builds the one from my user profile's nixpkgs channel
Neo-- has joined #nixos
<tilpner>
If you run NIX_PATH=/home/grahamc/.nix-defexpr/channels nix build nixpkgs.git, does it still work?
<gchristensen>
"No, it's not a bug. Rather, the nix command does not have a replacement for channels yet, so it's currently left undefined where it gets its inputs from. (This is one of the main reasons why the nix CLI is marked experimental.) So currently it only uses prefixed NIX_PATH elements like nixpkgs=...."
<{^_^}>
[nixpkgs] @dtzWill opened pull request #38385 → mesa: 17.3.6 -> 17.3.7, use llvm6 → https://git.io/vxP5M
<tilpner>
You didn't reply to "But do you agree that your last attempt from the gist contradicts edolstras explanation?". If you think it does, we should comment on one of the issues I linked
<gchristensen>
my last attempt being "nix build -f '<nixpkgs>' git"?
newhoggy has quit [Remote host closed the connection]
oida has joined #nixos
joepie91___ has joined #nixos
joepie91___ has quit [Changing host]
joepie91___ has joined #nixos
<chisui>
__monty__: I always thought you could use where and which interchangably
newhoggy has joined #nixos
<chisui>
fendor: ok, you need to set JAVA_HOME to be able to run maven. What is the error message you are getting
<__monty__>
chisui: where isn't installed on my system, have never not had which installed.
<sphalerite>
though: maybe at least the larger texlive-combined packages should copy the packages rather than symlinking them, to avoid needing hundreds upon hundreds of tiny derivations when a cache is available
<fendor>
there are multiple pom.xml, but this is the top-level xmlk
chreekat has joined #nixos
<brodul>
Have somebody packaged the VMWare player?
<fendor>
*pom.xml
ma27 has quit [Ping timeout: 240 seconds]
jrolfs_ has joined #nixos
jrolfs has quit [Ping timeout: 256 seconds]
ma27 has joined #nixos
<chisui>
fendor: actually the project "Assignment 1 :: JPA" fails to build so I would need that one too. But unless there is some strange stuff there that overrides test behavior it looks good
<WilliamHamilton>
but the bounds error persist. Am I calling doJailbreak the wrong way?
<fendor>
installed surefireplugin for intellij and reimporting the project
<WilliamHamilton>
(even if the command is doing something and not reporting errors?)
coot has joined #nixos
<fendor>
still fails :(
jophish has quit [Quit: WeeChat 2.0]
<chisui>
fendor: the problem isn't the surefire plugin itself. That will be downloaded by maven dynamically and it works. The problem is that creating forks in java is essentially impossible so the surefireplugin generates a jar and tries to execute it. the log essentially says that the surefire plugin couldn't create this "fork". Unfortunatly it doesn't show the actual error message the command printed.
<chisui>
You can see the command that was tried in the log
freeman42x]NixOS has quit [Ping timeout: 256 seconds]
<chisui>
but without the generated jar that doesn't help
MichaelRaskin has joined #nixos
<chisui>
so you need to run another `mvn package` that fails to generate the jar
Mercuria1Alchemi has quit [Ping timeout: 246 seconds]
newhoggy has quit [Remote host closed the connection]
erasmas has joined #nixos
<fendor>
chisui, mvn package fais but still no jar
newhoggy has joined #nixos
<WilliamHamilton>
pbogdan: thanks, that activates the jailbreak function; I forgot that what I wrote was interpreted as a list, and if I want application I have to signal it clearly with parentheses.
<chisui>
fendor: does `ass1-jpa/target/surefire/surefirebooter5820127360136840870.jar` exist?
<fendor>
no
blankhart has joined #nixos
asuryawanshi has quit [Ping timeout: 276 seconds]
newhoggy has quit [Ping timeout: 264 seconds]
kmicklas has joined #nixos
shoogz has quit [Remote host closed the connection]
blankhart has quit [Read error: Connection reset by peer]
blankhart has joined #nixos
<chisui>
fendor: The only way I know to get further in this case would be to run maven in debug mode and step through the surefire plugin code.
szicari has joined #nixos
<fendor>
chisui, so executing it with the -X flag?
akscram has quit [Ping timeout: 268 seconds]
<chisui>
add a breakpoint to the `org.apache.maven.plugin.surefire.booterclient.ForkStarter.fork` method and run `mvnDebug package` and attach intellij to the jvm
akscram has joined #nixos
<chisui>
fendor: but you should only do that I you want to sink a lot of time into that. You might just experiment with the JAVA_HOME variable or stuff like that
mirthlos has joined #nixos
<fendor>
chisui, yeah since tomorrow is the deadline for this task i'd prefer to to have a workaround
<dash>
i'm trying to cope with some nodejs packages with uncivilized behavior, namely, relying on external binaries without providing a way to override them. such as https://github.com/imagemin/pngquant-bin
<dash>
at `npm install` time, when it can't run its vendored executable, it tries to compile one
<dash>
any suggestions for coping with this in a node2nix/yarn2nix context?
<dash>
joepie91: unix was a mistake
<dash>
joepie91: yeah, i figured it would be something like that
zopsi has quit [Ping timeout: 276 seconds]
Arcaelyx_ has joined #nixos
Forkk has quit [Remote host closed the connection]
awygle has quit [Remote host closed the connection]
ahmedtd has quit [Quit: No Ping reply in 180 seconds.]
yrashk has quit [Ping timeout: 276 seconds]
oscarduignan has quit [Ping timeout: 276 seconds]
mduggie has quit [Ping timeout: 276 seconds]
lostman has quit [Ping timeout: 276 seconds]
Kim_ has quit [Ping timeout: 276 seconds]
y0no has quit [Ping timeout: 276 seconds]
jasom has quit [Ping timeout: 276 seconds]
awygle has joined #nixos
Forkk has joined #nixos
ahmedtd has joined #nixos
mduggie has joined #nixos
zopsi has joined #nixos
Kim_ has joined #nixos
Arcaelyx has quit [Ping timeout: 276 seconds]
<zybell_>
joepie91:That is the Block Embeddin and Extension Protocol(?) something to crash through firewalls by pretending to be a webbrowser. Of course that has security problems.
jasom has joined #nixos
lostman has joined #nixos
oscarduignan has joined #nixos
logzet has quit [Remote host closed the connection]
<{^_^}>
→ 7f3e2f62 by @hedning: gnome3.gnome-tweaks: make tweaks aware of gnome-shell-extensions
<{^_^}>
→ 43ce9b93 by @hedning: gnome3.gnome-shell-extensions: let apps-menu know about gnome-menus
<zybell_>
nixos IS unix,or do you mean a few aspects of unix(which)? Of course you can set anything suid. But which sane packager would do it if no need/no provisions for secure op?
<{^_^}>
→ 774253e5 by @ElvishJerricco: Fix GHCJS 7.10
<{^_^}>
→ 97e891ae by @peti: Merge pull request #38314 from ElvishJerricco/fix-ghcjs-7.10
szicari has quit [Client Quit]
szicari has joined #nixos
lopsided98 has quit [Remote host closed the connection]
ryantrinkle has joined #nixos
digitus has joined #nixos
<joepie91>
is there some way to make Nix builds (in particular nixos-rebuild) run with an idle scheduling class for I/O?
<joepie91>
right now, every time I do a rebuild, everything starts lagging :)
<clever>
joepie91: nixos-rebuild and any build started by root will generally bypass nix-daemon, so you can just ionice the command directly
<joepie91>
clever: that didn't seem to work here, as far as I can tell it handed off the job to nix-daemon
<joepie91>
I certainly see processes running at be/0 in iotop
<joepie91>
despite the rebuild running in the idle class
<clever>
joepie91: what about the process tree?
<joepie91>
not sure what you're asking
<clever>
are the heavy loads actually a child of nixos-rebuild?
szicari has quit [Quit: szicari]
<joepie91>
no idea, can't check now because my rebuild finished :p
<joepie91>
(incidentally, wouldn't it be reasonable to have rebuilds be run in the idle class by default?)
<clever>
the units in iotop are really throwing me off
hakujin3 has joined #nixos
<clever>
i keep seeing things like 1200 b/sec sorted above 50 kb/sec!
<joepie91>
yeah, iotop sorting is a little weird
chreekat has quit [Quit: quitting]
xcmw has joined #nixos
<{^_^}>
[nixpkgs] @jerith666 opened pull request #38397 → Update OpenJDK 10 and Eclipse to GTK 3 → https://git.io/vxX44
eyJhb has quit [Ping timeout: 256 seconds]
jrolfs_ has quit [Ping timeout: 240 seconds]
hakujin2 has quit [Ping timeout: 256 seconds]
eyJhb has joined #nixos
<ottidmes>
joepie91: I always substract some core count from maxJobs, that seems to help a lot
vidbina has quit [Ping timeout: 264 seconds]
Lears has quit [Ping timeout: 240 seconds]
goibhniu has quit [Ping timeout: 256 seconds]
<joepie91>
ottidmes: the problem here is I/O congestion, though, not CPU congestion
<joepie91>
limiting cores won't fix that :)
<joepie91>
and honestly I'm a little surprised that nixos-rebuild doesn't idle-class itself by default, because this seems like a textbook example of a job you'd want to run idle-class
ryanartecona has quit [Quit: ryanartecona]
ryanartecona has joined #nixos
ryanartecona has quit [Client Quit]
<bkchr[m]>
Pascal Bach: online?
jrolfs_ has joined #nixos
ryanartecona has joined #nixos
ryanartecona has quit [Client Quit]
ryanartecona has joined #nixos
<ottidmes>
joepie91: In my experience its always been the CPU that caused the lagging, so I thought I just mention it just in case.
<ottidmes>
joepie91: HDD as in regular HDD not SSD? If so, that explains :P I never really experienced an I/O issue when building something with Nix while on a SSD
<joepie91>
honestly I've been getting increasingly annoyed by the fact that nearly all modern software seems to assume that everybody is running an SSD
<ottidmes>
joepie91: I can imagine, I cannot stand to work on any laptop/desktop that still uses one as their OS drive, my low power laptop upgraded with a SSD felt faster than my otherwise superior desktop with a HHD at the time, and that was when HDD were still the norm, should only have gotten worse
<fyuuri>
Does cisco vpn work under nixos? (I want to access the university VPN)
<benley>
fyuuri: I haven't tried using it, but I think it can be done. NetworkManager includes a Cisco AnyConnect client, and I know at least some of its other vpn client types work.
<benley>
fyuuri: oh I also see a cisco "vpnc" client in networkmanager
<fyuuri>
Ah. That is nice. Is there a howto for setting up the network manager vpn?
ryanartecona has quit [Quit: ryanartecona]
<benley>
If you're using a desktop like gnome or kde, it ought to include a UI for configuring network manager. I run a standalone one called nm-applet (part of gnome, probably?)
<benley>
other than getting a desktop setup, the rest of vpn client setup through NM ought to be similar to any other distro
<benley>
(in theory)
<fyuuri>
nice. I used to use the nm-applet under arch linux. Is it enogh to install that?
<fyuuri>
There are some options regaring nm manager in the configuration.nix
<fyuuri>
Ill just test installing nm-applet and come back here if I run into problems :)
<fyuuri>
thx!
<benley>
you'll want to set networking.networkmanager.enable = true; at least
<benley>
and install nm-applet
<fyuuri>
ok! thx!!
<benley>
I also find it useful to set networking.networkmanager.unmanaged = ["interface-name:docker*" "interface-name:veth*"]; to make it ignore docker and VirtualBox interfaces
orivej has quit [Ping timeout: 256 seconds]
<fyuuri>
thank you!
hellrazo1 has quit [Quit: WeeChat 2.1]
ryanartecona has joined #nixos
fendor has joined #nixos
<infinisil>
fyuuri: the standalone client vpnc doesn't seem to work well for me. While it does connect just fine and I can ping ip's, DNS is borked up
<benley>
mmm networkmanager does seem to insist on doing elaborate and confusing things with resolvconf
<infinisil>
That's also why vpnc doesn't work, also does stuff with resolvconf
<benley>
infinisil: have you tried setting network.networkmanager.useDnsmasq = true; ?
<infinisil>
I'm not using networkmanager
<benley>
ahhh ok
<fyuuri>
Ill try if nm works for me
<fyuuri>
I really need to get this VPN working asap^^
<infinisil>
I heard somebody else say that cisco vpn works with networkmanager
<benley>
I don't quite remember the details of why that useDnsmasq mode got things working well for me but it definitely did. Something to do with split-tunnel DNS.
<fyuuri>
that would be great
<benley>
*split-horizon dns
<infinisil>
fyuuri: I ended up just booting into my Mac to get it to work lol, was really urgent
coot has quit [Quit: coot]
endformationage has quit [Ping timeout: 268 seconds]
<fyuuri>
I really want to access the network with nixos ^^
<benley>
"networkmanagerapplet" is the name
<fyuuri>
thx!
<benley>
the attribute in nixpkgs, at least
coot_ has joined #nixos
<tnks>
I'm beginning to second-guess which channel I follow with respect to balancing having latest-curation, but with a hint of stability.
<benley>
which have you been using?
<tnks>
nixpkgs-unstable
<tnks>
I think that's the popular "recommendation" that's in the docs.
<tnks>
but I don't really understand how branches from the normal nixpkgs repository feed into nixpkgs-unstable and also nixos-unstable.
<benley>
If you're running nixos, you probably don't want to use nixpkgs-unstable. nixos-unstable includes a bunch of tests that keep it from breaking your whole machine :-P
<tnks>
in this case, I'm not running nixos, just nix to do package management for Docker images.
<benley>
both of the -unstable channels come from the master branch - they're the most recent commit where the test suite for that channel has passed on hydra.
<tnks>
I also am most interested in Python machine learning libraries (this is for work).
<benley>
ahhhhh cool, we're starting to use nix that way at work too :-D
<tnks>
okay, then maybe I'm on the right channel.
<tnks>
benley: any thoughts of what to use as a reference point?
<benley>
I've been considering going with the most recent nixos-xx.xx stable channel
<tnks>
I've been using nixpkgs-unstable for about a year, and with a recent attempt to upgrade to a new version of nixpkgs-unstable, a few things broke.
<tnks>
maybe that's a way to go... it's just got less parity with Conda.
<tnks>
Python is a tire fire from a correctness perspective.
<benley>
I figure with a -stable channel the package versions remain stable for at least half a year at a time, including security patches, and when it does come time to upgrade there will be other people working out the kinks too
justanotheruser has quit [Ping timeout: 264 seconds]
<benley>
oy vey, yeah.
<tnks>
sometimes stability is just a farce, though. It's just pinning at some arbitrary point of time.
<tnks>
(especially for Python libraries).
<tnks>
and for this particular ML work, I don't feel I'm in a position to make security much of a concern. I feel Python already lost that war.
<{^_^}>
[nixpkgs] @primeos opened pull request #38401 → nixos/rootston: Use an attribute set and the new keyboard configuration format → https://git.io/vxXwI
<fyuuri>
nm-applet and networkmanager both work fine
<fyuuri>
But the VPN connect seems not to work :(
<benley>
d'oh. Anything interesting in syslog?
<fyuuri>
failed to commit changes to dconf
justanotheruser has joined #nixos
sigmundv has joined #nixos
fendor has quit [Ping timeout: 256 seconds]
<fyuuri>
benley: network manager cannot add the connection because it does not have the previleges on my machine.
fendor has joined #nixos
<fyuuri>
So maybe it works if I can give it the required rights^^
WhittlesJr has quit [Ping timeout: 260 seconds]
astsmtl has quit [Ping timeout: 268 seconds]
<benley>
oh man I cannot remember how the fancy permissions policy stuff works ... "polkit" I think it's called?
<fyuuri>
:( Is that really necessary?
pie___ has quit [Ping timeout: 260 seconds]
<benley>
brb, in a meeting
<fyuuri>
Ah!
<fyuuri>
I just looked in the .nix
<fyuuri>
Is the user needs to be in the networkmanager group it seems.
<fyuuri>
OK. thx for the help again. Ill tell you when I get it working :)
bkchr has quit [Quit: Konversation terminated!]
kreisys has quit [Read error: Connection reset by peer]
<{^_^}>
[nixpkgs] @dtzWill merged pull request #37810 → udisks2: 2.1.6 -> 2.1.8, extend workaround to include musl → https://git.io/vx0LU
<{^_^}>
→ 8f2102de by @mbode: tectonic: fix darwin build
<{^_^}>
→ cc1d4add by @LnL7: Merge pull request #38384 from mbode/tectonic_darwin
kreisys has joined #nixos
hakujin4 has joined #nixos
<kreisys>
hey! it looks like every time I run gc it removes packages that are linked-to from other packages
<kreisys>
is it not allowed to have symbolic links in packages?
astsmtl has joined #nixos
astsmtl has quit [Changing host]
astsmtl has joined #nixos
kiloreux_ has joined #nixos
hakujin3 has quit [Ping timeout: 246 seconds]
<zybell_>
idk but the links have to be registered somwhere,
<zybell_>
it seems from some talk on this channel.
<infinisil>
Yeah, can you describe the problem with a bit more detail kreisys?
<kreisys>
So... I'm writing yet another nodejs builder
kiloreux has quit [Ping timeout: 246 seconds]
<kreisys>
and since the /lib part takes (waaaaay) longer to build then the /bin part of the package, I split them to 2 packages and then created a "meta" package that just links /bin to the bin package and /lib to the lib package
<monotux>
how is the 18.03 release going btw?
<kreisys>
whenever I run the gc it deletes the bin and lib packages and the meta package remains. broken.
<kreisys>
that also breaks nix-env until I explicitly remove all those packages
<lejonet>
kreisys: how do you link em? If you add em like buildInputs, they should become deps
<kreisys>
Symlinks
<infinisil>
monotux: Earliest release date is today, but it might be delayed a bit longer
<lejonet>
O.o nix isn't aware of symlinks
<infinisil>
kreisys: What's your nix file for building?
<zybell_>
does the meta pkg correctly use the bin and lib as build-inputs?
<lejonet>
so it won't treat the dependencies as dependencies if you only symlink them
<infinisil>
No nix should do that
<lejonet>
Well, I guess that depends on how the symlinks are made I guess
<kreisys>
ln -s
<infinisil>
Yeah that should work, what's the nix code?
<zybell_>
Using them as buildInput registers symlinks in nix.
<kreisys>
let me query the closure...
civodul has joined #nixos
<gchristensen>
a dependency doesn't strictly have to be a buildInput for it to be considered a run-time dependency, zybell_
<monotux>
infinisil ah, ok! I've noticed the release branch but it hadn't been updated in a while
<zybell_>
But I think its easiest in his case.
<gchristensen>
buildInputs do not automatically become run-time dependencies
<infinisil>
Indeed ^^ Whatever is contained in $out will become a run-time dependency, and buildInputs don't get added to $out automatically
<infinisil>
I mean that, whatever /nix/store path occurs in $out will become a dependency
<zybell_>
If they are linked?
<kreisys>
But do symlinks count?
<infinisil>
Yes
<Guest8388>
Anyone installed Dropbox and linked it, only to get dropbox status == "up to date" with nothing synced?
fendor has quit [Remote host closed the connection]
<infinisil>
I could help a whole lot more efficiently if you could show the nix code you used to produce the packages
<kreisys>
ok that's strange...
spear2 has joined #nixos
<kreisys>
So the -bin part appears in the nix-store -qR output
<kreisys>
but the other part does not
fendor has joined #nixos
endformationage has quit [Ping timeout: 263 seconds]
<kreisys>
perhaps because the lib part doesn't actually have the -lib suffix? it is named exactly the same as the top level package
roberth has joined #nixos
<kreisys>
except for the hash of course
<kreisys>
could that be a problem?
<lejonet>
Hmm, is there any easy way to mark certain parts of a package to be setuid? (I'm thinking about chromium and vivaldi mainly, they apparently need setuid root for their sandboxes...)
<kreisys>
I thought names are just for us humans
ma27 has quit [Ping timeout: 252 seconds]
endformationage has joined #nixos
<infinisil>
kreisys: I can help you if you tell me how to reproduce the problem
<kreisys>
kk... let me try and whip up a minimal repro
coot_ has quit [Quit: coot_]
rauno has quit [Ping timeout: 260 seconds]
ryanartecona has joined #nixos
periklis has joined #nixos
michas_ has joined #nixos
<periklis>
vcunat: is there actually a date when 18.03 will be officialy on nixos.org available?
<fyuuri>
benley: I can now add connections, but it keeps me asking for a group password which does not exist. (At least to my knowledge; maybe it is somewhere in the vpnc config file)
ryanartecona has quit [Read error: Connection reset by peer]
<benley>
weird - a vpn group password? Not a posix group password (which would be even weirder, but possible)
ryanarte_ has joined #nixos
MP2E has joined #nixos
<fyuuri>
benley: the nm vpn setup from asks for a group password. You can say: "The password is not required". But when I choose this option it asks for the password as soon as I try to connect
hakujin4 has quit [Ping timeout: 276 seconds]
<fyuuri>
The "official" cisco vpn client does not ask for this group password.
justanotheruser has joined #nixos
drakonis has joined #nixos
<lejonet>
What type of VPN?
<maurer>
fyuuri: If you receive the official client from your internal distribution rather than cisco directly, it may have some preconfiguration done
<maurer>
I know this is true of my institution's vpn client
fendor has quit [Remote host closed the connection]
bara has joined #nixos
<maurer>
(e.g. the group password might be baked into a client you download from your organization)
<lejonet>
Shouldn't be too hard to find it you'd think :P
<zybell_>
I assume its a POSIX group passwd because the user is not in the right group and it tries to remedy that by way of newgrp.
<fyuuri>
But I dont even know yet what to decode
<infinisil>
fyuuri: Yeah i also think it needs a group password, also called ipsec secret or group secret
<fyuuri>
infinisil: yes
jperras has quit [Ping timeout: 252 seconds]
<lejonet>
ah, ipsec, fun times
<fyuuri>
This is when I envy windows users...
rauno has joined #nixos
<infinisil>
fyuuri: I mean if nixos had a great cisco vpn module it would be super easy
<fyuuri>
(not really^^)
<infinisil>
But it doesn't :
<infinisil>
(
<infinisil>
yet at least
<lejonet>
infinisil: if you're willing to pay for the therapy... :P
<fyuuri>
But the network manager seems to work.
jperras has joined #nixos
<lejonet>
Was a while since I fiddled with cisco's own VPN, but I remember it being counter-intuitiv and stupidly annoying to setup
<michas_>
Hi, I stubled upon this short blog post: https://theshortlog.com/2016/06/29/Prepare-NixOS-disk-image-for-another-computer/ which seems to use a very direct nix-build to create a disk image using make-disk-image. This seem much more clean than the command needed by the nova-image. Unfortunately it does not work (anymore?) as it is. Can you give me a hint what the author intended to do here?
taktoa has quit [Remote host closed the connection]
lassulus_ is now known as lassulus
aarvar1 has joined #nixos
aarvar is now known as Guest58191
Guest58191 has quit [Ping timeout: 264 seconds]
aarvar1 is now known as aarvar
<LnL>
nix-build -A config.system.build.isoImage -I nixos-config=/path/to/configuration.nix
<infinisil>
"action timed out after -1 seconds" Sounds like a random vm failure
<michas_>
LnL: In the blog post they simply use `nix-build mkimage.nix`. Did this work back then? Why isn't this possible anymore?
<infinisil>
michas_: I think he's using the file there as mkimage.nix
<LnL>
michas_: not sure, what's the error?
<infinisil>
Huh, that's some weird syntax in that file though: `import <nixpkgs nixos="">`, what does that even do?
<fyuuri>
It works!!! :))
<LnL>
huh, wait ts that a thing?
<fyuuri>
Somebody said "IPSec" and that is what made me think.
<fyuuri>
^^
xcmw has joined #nixos
<michas_>
infinisil: LnL: yes, there are syntax errors and weird expression I do not completely understand. like that <nixpkgs nixos="">.
<infinisil>
LnL: I don't think so, really weird
<michas_>
not sure how to fix that.
<judson>
With GrahamCOfBorg, is there a path to a bors/zuul for nixpkgs?
<fyuuri>
You can use the Cisco AnyConnect Protocol which uses full SSL and that worked.
<fyuuri>
There are two cisco options in network manager
<LnL>
yeah, that's not right afaik
<fyuuri>
benley: It worked. See my messages above. What did not work for you? Maybe we can make yours work, too. Thank you so much for your help. You put me on the right track for solving the problem!
<{^_^}>
[nixpkgs] @volth opened pull request #38405 → network-interfaces-scripted: wlanInterfaces have .device, … → https://git.io/vxXHq
<judson>
In the meantime, can I beg for #38077 to get merged before there's a conflict with another Node expression?
<benley>
fyuuri: I don't have a cisco vpn to use - did you mean maurer?
<fyuuri>
If somebody needs help with cisco via networkmanager, tell me.
pikajude has joined #nixos
<Guest8388>
Hm, any obvious culprit for this? "The option value `environment.systemPackages.[definition 39-entry 11]' in `/etc/nixos/configuration.nix' is not of type `package'."
<benley>
fyuuri: <infinisil> fyuuri: I ended up just booting into my Mac to get it to work lol, was really urgent
<benley>
fyuuri: and you're welcome
<vcunat>
`tested` is the main thing driving whether a channel is bumped
<infinisil>
vcunat: I see, yeah it's weird for systemd not to be in there
<vcunat>
(also it waits for everything to finish - that's the only other rule, I think)
<michas_>
LnL: your expression gives me: syntax error, unexpected '=', expecting ')', at mkimage.nix:5:36
<LnL>
just edited it :)
<infinisil>
vcunat: Isn't this "lucifer" machine the culprit? All errors I see are because of that one.
<Dezgeg>
it's also the only machine that runs tests I think =)
<LnL>
great name for the machine that runs all of the flaky tests :)
<vcunat>
there surely are more, but it's limited IIRC
alex`` has quit [Quit: WeeChat 2.1]
<Dezgeg>
AFAICT from the machine status page, that is the only one
<infinisil>
Dezgeg: I'm a bit confused, because I can only see the machine name for failing jobs, succeeding ones don't show it
<vcunat>
infinisil: the `build steps` tab shows machine names
<michas_>
LnL: now it gives me: anonymous function at /nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/nixos/lib/make-disk-image.nix:1:1 called with unexpected argument 'partitioned', at mkimage.nix:29:1
hakujin1 has joined #nixos
<LnL>
guess make-disk-image.nix changed a bit since that was written
<vcunat>
both firefox and systemd tests succeed for me locally (only one attempt done)
kreisys has joined #nixos
<infinisil>
Dezgeg: Ctrl-F for "tests" on https://hydra.nixos.org/machines, pretty much all linux machines run tests according to that
dbe has quit [Ping timeout: 248 seconds]
<michas_>
LnL: but after removing that attribute, it seems to work fine.
<Dezgeg>
they are building parts of the tests
<Dezgeg>
or dependencies, I mean
<zybell_>
Shouldn't a test machine after sending: xxx broken,Error..., send:btw here is the proof http:...?
<Dezgeg>
none of the others is running vm-test-run-*.drv ones
<infinisil>
I'm gonna run the firefox and systemd checks locally too
dvim has quit [Quit: WeeChat 2.1]
<infinisil>
(on d5c26dd063b)
hakujin has quit [Ping timeout: 276 seconds]
<infinisil>
vcunat: These 2 tests succeed for me. I'm gonna go through more failing ones on hydra to see if they are actually failing
lambdamu has quit [Read error: Connection reset by peer]
<infinisil>
Ah right, only the tested jobset is important, still gonna run tests for the failing ones not in that jobset though
bara has quit [Read error: Connection reset by peer]
bara has joined #nixos
<infinisil>
vcunat: Let me know if I can help you somehow!
<vcunat>
infinisil: you don't need to retry the keymap tests
<vcunat>
I know well they're flaky
<infinisil>
Yeah, well the tests run fairly quick so it's no problem
endformationage has quit [Ping timeout: 260 seconds]
<{^_^}>
[nixpkgs] @LnL7 pushed commit from @mbode to release-18.03 « tectonic: fix darwin build »: https://git.io/vxXdy
<zybell_>
Should one run known flaky tests? If they are flaky, what is the worth doing them?
saltsa has joined #nixos
saltsa has quit [Client Quit]
endformationage has joined #nixos
kreisys has quit [Read error: Connection reset by peer]
<infinisil>
zybell_: Well if something does actually break we would like to know that, that's why tests are there in the first place
bara has quit [Read error: Connection reset by peer]
<catern>
how can I run a python program which runs executables in subprocesses so that it will run the executables from Nix?
bara has joined #nixos
<symphorien>
Use a wrapper to set the PATH ?
<infinisil>
zybell_: The best thing would be to fix the flakyness, but that's hard
<zybell_>
infinisil:If a test is flaky,you *can not* know if something breaks. That is my point.
hakujin2 has joined #nixos
<infinisil>
You can look at the logs in our case, which often gives a hint whether it's just failing due to randomness or an actual problem
<{^_^}>
[nixpkgs] @layus opened pull request #38409 → cups-filters: remove myself from maintainers → https://git.io/vxXFD
Guest8388 has quit [Quit: Lost terminal]
kreisys has joined #nixos
hakujin1 has quit [Ping timeout: 276 seconds]
doyougnu has joined #nixos
jb55 has joined #nixos
zybell_ has quit [Ping timeout: 240 seconds]
thisisonlyatest has joined #nixos
thisisonlyatest has quit [Client Quit]
<Dezgeg>
there are spurious false positives but no false negatives
ertes-w has quit [Read error: Connection reset by peer]
coot_ has joined #nixos
<michas_>
Have you ever seen that issue on a current nixos? http://yellowgrass.org/issue/NixOS/149 (the nix-expression built above seems to create that.)
ertes-w has joined #nixos
acarrico has joined #nixos
<vcunat>
I'm afraid we just don't see the false negatives :-)
phreedom has quit [Remote host closed the connection]
<gchristensen>
vcunat: want to wait for more time, or merge & bacport that PR?
<vcunat>
it feels coupled to nix-2.0 -by-default, but even that ended up as a more gradual change
asuryawa_ has quit [Remote host closed the connection]
bara has quit [Read error: Connection reset by peer]
asuryawanshi has joined #nixos
bara has joined #nixos
bara has quit [Read error: Connection reset by peer]
bara has joined #nixos
jb55 has quit [Ping timeout: 268 seconds]
rihards has quit [Quit: rihards]
<joepie91>
I remain amazed by just how often copypasting breaks in chromium
<joepie91>
cc judson
jrolfs has joined #nixos
xcmw has quit [Ping timeout: 264 seconds]
<joepie91>
like, I've seen really bizarre bugs in this area before with eg. synergy + chromium resulting in chromium hanging whenever it tries to access the clipboard (and ONLY chromium), and just outright not copying stuff when you hit copy, etc.
<joepie91>
across distros, across setups...
xcmw has joined #nixos
Neo-- has joined #nixos
<foldingcookie>
chromium substantially implements its own toolkit, and the X11 clipboard is astoundingly complex
jrolfs_ has quit [Ping timeout: 264 seconds]
<foldingcookie>
plus it probably tries to support a lot of different audiovisual and rich-text targets
noel has joined #nixos
<MichaelRaskin>
Tried to run NixOS tests for Firefox and Chromium locally (for upstream release-18.03 branch head)
<MichaelRaskin>
Of course they both pass
noel is now known as Guest3510
scribbler has joined #nixos
chessai has quit [Ping timeout: 260 seconds]
<{^_^}>
[nix] @erictapen opened pull request #2040 → document that writeTextFile from Nixpkgs is an alternative to toFile → https://git.io/vxXpx
Guest3510 has quit [Quit: Page closed]
bara has quit [Read error: Connection reset by peer]
<infinisil>
Yay I found the cause of a ever-since failing test
justanotheruser has joined #nixos
bara has joined #nixos
jensens has joined #nixos
<infinisil>
When I have such a fix, should I make 1 PR and mention it should be backported, or make 2 PR's for each branch?
<tnks>
I'm looking at some of the Python builder stuff. Does anyone know how to build with an earlier version of gcc?
bpa has quit [Remote host closed the connection]
coot_ has quit [Quit: coot_]
<infinisil>
tnks: In nixpkgs/pkgs/top-level/all-packages.nix you can see 5 gcc version (4.8, 4.9, 5, 6, 7)
<infinisil>
tnks: You should be able to override gcc for the thing you want to use it for, but how that exactly works depends on the thing.
newhoggy has quit [Remote host closed the connection]
newhoggy has joined #nixos
hakujin2 has quit [Ping timeout: 256 seconds]
RedNifre has joined #nixos
civodul has quit [Quit: ERC (IRC client for Emacs 25.3.1)]
simukis has quit [Quit: simukis]
hiratara has quit [Ping timeout: 260 seconds]
scribbler has quit [Quit: scribbler]
lord| has joined #nixos
vidbina has quit [Ping timeout: 260 seconds]
newhoggy has quit [Ping timeout: 248 seconds]
hiratara has joined #nixos
<RedNifre>
Would you recommend NixOS as a daily desktop OS or is it better to only use it on a dedicated development machine?
<tilpner>
We're obviously biased, but it can work well as a daily-use OS too
<RedNifre>
How ... durable is it? Can I install it once and update it for years without problems or will it decay over time like so many OSes do, leading to an eventual re-install?
jperras has joined #nixos
endformationage has quit [Ping timeout: 256 seconds]
leotaku has quit [Ping timeout: 240 seconds]
<joepie91>
RedNifre: NixOS is my daily driver across all my systems; however, be prepared to spend a few weeks upfront re-discovering how to deal with your OS, since it will be considerably different from 'traditional' distros (this will probably eventually stop being a problem, but right now usability/documentation still need some work)
<RedNifre>
Could I install a headless version on some spare machine, connect to it over ssh and use it to develop some Haskell in vim or would that be a bad idea.
<joepie91>
RedNifre: if you're willing to spend the few weeks upfront, it's absolutely worth it :)
hamishmack has quit [Quit: hamishmack]
<tilpner>
RedNifre - I started with 16.09 and haven't had to reinstall once
<joepie91>
RedNifre: and no, it won't decay so long as you don't try to 'cheat' the immutable/functional/pure model of it
<joepie91>
(that's actually one of the big benefits of that model)
<RedNifre>
Does it work well on old, weak nettops?
<MichaelRaskin>
It is hard to create a situation where reinstall helps
<MichaelRaskin>
Ouch
<joepie91>
your system always remains 'clean' because you don't have a single global namespace where everything is thrown into; instead it 'rebuilds' your environment anew every time (using symlinks, and only building/fetching the stuff that has changed)
<MichaelRaskin>
Nix loves RAM
<joepie91>
RedNifre: expect to need at least 1GB of RAM for comfortable operation right now
<maurer>
RedNifre: Old weak nettops will only work well if you are using "common" software from stable stuff - otherwise it will have to build locally
<MichaelRaskin>
Nix also loves HDD
<joepie91>
ideally more
<tilpner>
RedNifre - Not very. How old are you talking about?
chisui has quit [Ping timeout: 260 seconds]
endformationage has joined #nixos
<joepie91>
RedNifre: (if you have a very rigorous "cleaning up old revisions of the system" policy, then you can run NixOS on something like a 80GB HDD without too many issues)
<tilpner>
RedNifre - NixOS works well on weak machines if yo u manage them from a more powerful machine, but managing it on-system can be tedious
<MichaelRaskin>
I would say that nettop is OK as long as everything is actually evaluated on a larger machine
<MichaelRaskin>
It is not that hard to run out of RAM during evaluation of the system config
<RedNifre>
I have an old Zotac Mag Mini here with 2 GB RAM
<RedNifre>
How does managing from a more powerful machine work?
<joepie91>
RedNifre: should work so long as you don't try to install too many complex things on it, but you'll want to keep a close eye on your disk usage in particular
<joepie91>
NixOS is - at least currently - not the best option in terms of low resource demands :)
<RedNifre>
Disk space is 160 GB, I'd probably only use it to try out Haskell development with Nix on it.
bara has quit [Read error: Connection reset by peer]
<joepie91>
then you should be fine, unless there's some Haskell-specific caveat I'm not aware of
<tilpner>
RedNifre - I manage a weak netbook from my stronger laptop by building the entire system locally and automatically pushing it via ssh
<joepie91>
(which there well may be, I don't do Haskell)
<tilpner>
RedNifre - So the netbook never has to evaluate anything or compile anything
bara has joined #nixos
bara has quit [Read error: Connection reset by peer]
<RedNifre>
Yeah, but the point is not to find a low resource OS for my old machine, it's to try to find a way to use NixOS without too much disruption :)
bara has joined #nixos
<vcunat>
I haven't needed a reinstall during roughly five years of usage. Then I mostly abandoned the old notebook (and copied my ~/ to a new machine).
bara has quit [Read error: Connection reset by peer]
<joepie91>
so far, my NixOS installations have outlived my hardware iterations :P
<joepie91>
then again, I've only been using it for like 2 years now?
bara has joined #nixos
bara has quit [Read error: Connection reset by peer]
bara has joined #nixos
<MichaelRaskin>
RedNfire: just install Nix on a non-NixOS Linux distribution
<RedNifre>
Well, given that I'm not using my old nettop for anything right now I'll just try it out tomorrow and see what happens. If it doesn't work well I can still switch.
<MichaelRaskin>
2GiB RAM is painful
<joepie91>
in principle, you should be able to upgrade NixOS in perpetuity without reinstallation, without running into issues
<RedNifre>
MichaelRaskin what are the pros and cons of using NixOS + nix vs just nix somewhere else?
davidlt has quit [Ping timeout: 264 seconds]
leotaku has joined #nixos
<MichaelRaskin>
Well, a point for Nix-only: you do not change anything in your driver setup (also networking)
<MichaelRaskin>
A point against: OpenGL means that libGL.so is also a driver…
<symphorien>
NixOS makes some things where proprietary blobs are unavoidable like printing painful
<MichaelRaskin>
Then again, depends on the blob in question, some are way worse than others
<joepie91>
I've personally found the value proposition for NixOS much more compelling than Nix-on-another-Linux; you end up having to learn almost the same amount of stuff, but on NixOS you get far more guarantees and convenience out of it (with the exception of statically compiled binaries needing extra work to run)
<infinisil>
Why is low RAM a problem with NixOS?
<infinisil>
All my machines have >=8GB, which seems to be plenty
<MichaelRaskin>
Because NixOS modules system is a complicated piece of Nix code. We are discussing NixOS on 2GiB and predicting pain.
ryanarte_ has quit [Quit: ryanarte_]
<MichaelRaskin>
At 8GiB that should be fine, yes.
<judson>
joepie91, MichaelRaskin: so that sounds like still busted?
<vcunat>
2 GiB feels painful even for regular web browsing nowadays
<joepie91>
infinisil: Nix itself already tends to consume *considerable* amounts of RAM when evaluating your system, add to that various builders and you run into RAM issues real quick
<judson>
I'm trying to figure out, then, how to roll back to a working Chromium.
<infinisil>
Hmm I guess I just never noticed
<joepie91>
if you have 8GB then you probably wouldn't indeed :P
<MichaelRaskin>
Chromium? I think it works on all channels…
<judson>
On 16G, I used to notice when I tried to `nix-env -i <packagename>`
<vcunat>
8 GiB and --cores 16 would get you into trouble with some packages
<{^_^}>
[nixpkgs] @fpletz pushed to release-18.03 « nixos/version: fix nixops pre 1.6 compatibility »: https://git.io/vx1fR
<judson>
Foolish I, not knowing to -A my --installs
<judson>
MichaelRaskin, I'm here to testify that Chromium out of 17.09 stable crashed tabs on Copy
<MichaelRaskin>
Well, _some_ packages would strain even my 32 GiB system if built in parallel.
<MichaelRaskin>
Aah, 17.09
kerrhau has joined #nixos
kerrhau has quit [Changing host]
kerrhau has joined #nixos
<judson>
Well 18.03 isn't released...
<MichaelRaskin>
I just use master
jtojnar_ has joined #nixos
<MichaelRaskin>
I will recheck if Chromium test passes on 17.09 — I think at some point it passed, and note that it does use copy-paste…
<judson>
The linked issue includes reports to that effect.
newhoggy has joined #nixos
<judson>
This update is only a few days old, though.
<MichaelRaskin>
Maybe I can grab a new version from Hydra and recheck then
<pbogdan>
MichaelRaskin: last time I checked it used to crash in the test on 17.09 but the clipboard content is somehow present so the test passes regardless
<MichaelRaskin>
Oh. That's interesting. Theoretically, X clipboard is not supposed to work like that, of course…
Neo-- has quit [Remote host closed the connection]
Neo--- has joined #nixos
<infinisil>
Great, love it when a derivation doesn't build but with no log of what went wrong, at all
<infinisil>
I just have exit code != 0
<MichaelRaskin>
Of course, NixOS tests have enough overhead for adding the crash check to be annoying…
newhoggy has quit [Remote host closed the connection]
newhoggy has joined #nixos
<kini>
layus: good point on the ML just now
c0ffee152 has joined #nixos
jensens has quit [Ping timeout: 268 seconds]
<kini>
I was also surprised to see a package I maintained had been updated. I didn't really mind since I was planning to update it myself anyway, but it did create a conflict when I rebased my update branch (I was planning to rework the nix expression along with the update)
<kini>
it would be nice if the maintainers were at least @'d on the PRs by the bot
<mpcsh>
hey y'all, I'm back on NixOS after a hiatus, and I came back to some fucky font issues. Basically, many fonts match "FreeMono", even though I didn't explicitly install that package. So my question is: is there a way to see "why" the freefont package was installed? That is, what depends on it? And is there a way to disable it?