<{^_^}>
[nixpkgs] @joachifm opened pull request #48440 → nixos/hardened: assert incompatibility with nix.useSandbox → https://git.io/fx0OA
carlosdagos has joined #nixos
drakonis has quit [Ping timeout: 272 seconds]
drakonis has joined #nixos
<adamantium>
Hi, i want to install texlive, which package should i go for?
<adamantium>
I'm a little confused by 'nix search texlive' , it shows stuff related to "scheme". I am looking for the program to work with latex files and need to export to pdf
<adamantium>
(i already have pandoc)
<drakonis_>
there's a package set related to texlive
<carlosdagos>
I had success with pandoc and `texlive.combined.scheme-small`
<drakonis_>
look up the repository for that
<drakonis_>
or that
<carlosdagos>
:)
<adamantium>
huh, i guess i'm confused between so many options
<adamantium>
i thought it was even the wrong thing.
<samueldr>
>> Figure 6 displays the TEX Live scheme menu; from here, you choose a “scheme”, which is an overall set of package collections. The default full scheme installs everything available. This is recommended, but you can also choose the basic scheme for a small system, minimal for testing purposes, and medium or teTeX to get something in between. There are also various specialized and country-specific schemes.
<jasongrossman>
Oh, THAT meaning of "scheme"! That had occurred to me but I thought it was unlikely. Anyway, I'm very glad to see it's someone else's documentation problem and not Nix's.
<jasongrossman>
samueldr++
<{^_^}>
samueldr's karma got increased to 35
Rusty1 has joined #nixos
<samueldr>
:)
<samueldr>
words are hard
<samueldr>
we should communicate idea and emotions only through GUIDs
andersk has quit [Remote host closed the connection]
andersk has joined #nixos
<{^_^}>
[nixpkgs] @matthewbauer pushed commit from @ambrop72 to release-18.09 « nixos/xserver: Implement configuration of NVIDIA Optimus via PRIME »: https://git.io/fx0sN
<drakonis_>
maybe nix in even more languages
<drakonis_>
nix in python...
<drakonis_>
in pure C.......
lassulus has quit [Ping timeout: 252 seconds]
lassulus_ is now known as lassulus
<samueldr>
implementing nix in nix
<{^_^}>
[nixpkgs] @thoughtpolice merged pull request #48336 → tests/prometheus-exporters: add tests for new Tor exporter → https://git.io/fxRZi
<{^_^}>
[nixpkgs] @thoughtpolice pushed commit from @delroth to master « tests/prometheus-exporters: add new Tor exporter »: https://git.io/fx0Gv
<worldofpeace>
^ paradox
<drakonis_>
i'd do it
fiatjaf has joined #nixos
<fiatjaf>
"fetchurl requires a hash for fixed-output derivation"
<fiatjaf>
how can I know the hash?
<gchristensen>
,tofu fiatjaf
<{^_^}>
fiatjaf: To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000) then replace it with the correct hash Nix expected.
<fiatjaf>
:D
<drakonis_>
tofu :V
<fiatjaf>
it's great that you have a bot like that
<fiatjaf>
,tofu fiatjaf
<{^_^}>
fiatjaf: To get a sha256 hash of a new source, you can use the Trust On First Use model: use probably-wrong hash (for example: 0000000000000000000000000000000000000000000000000000) then replace it with the correct hash Nix expected.
<dhess>
I've somehow created a Hydra job that seems to do the build during the job evaluation :\
<dhess>
so all I see is an evaluation that takes hours and finally produces one big error log, and no actual build is run
garbas has quit [Quit: WeeChat 2.2]
alex-v has quit [Remote host closed the connection]
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<elvishjerricco>
dhess: Does that imply you've accidentally done IFD?
<elvishjerricco>
I thought Hydra didn't allow IFD by default
<dhess>
elvishjerricco: I honestly don't know what's going on
<dhess>
I've never seen this happen before. Do you happen to remember the trick to run a a hydra evaluation from the command line?
<dhess>
I've got it written down here somewhere
<dhess>
yeah I'll be damned
<dhess>
The evaluation step built GHC on aarch64 and then tried to build my packages
<dhess>
and it died the same place that `nix-build` did from the command line running the same job by hand
<dhess>
that is so weird
<dhess>
elvishjerricco: so, dunno if you've seen my update to the GHC aarch64 issue, but it looks like -jN is still happening, at least for one build phase, for haskellPackages even when enableParallelBuilds is disabled. Looks like the configure phase maybe? Or wherever it is that a cabal project builds Setup.hs
<dhess>
at the same time it doesn't seem to trigger the bug.
<dhess>
so, I think you were effectively right that GHC's build system doesn't use -j and we can enable parallel builds for GHC aarch64
<dhess>
which means it will compile in just a few hours on these Packet.net machines
<elvishjerricco>
elvishjerricco: Ah, right. I forgot that it does a `ghc --make` to build... ghc-cabal, maybe? Some tool internal to the build process that uses a lot of libs shipped in the source tree. I guess that `ghc --make` may be using `-j` as well.
<dhess>
I only ever saw the -j flag leaking through when it was compiling Setup.hs or Setup.lhs, and it doesn't appear to trigger this bug
b has quit [Quit: Lost terminal]
<dhess>
anyway you can get pretty far into haskellPackages. It dies in tls's tests; could be related to an issue that peti filed on tls-1.4.1 on CI servers where something times out, or it could be a genuine aarch64 issue
zopsi has quit [Quit: Oops]
<dhess>
but I think I'll make another PR to set enableParallelBuilds on the GHC builds. That should get Hydra going and we can start to chip away at the other issues, like whether -j is leaking through or what's going on with TLS.
halfbit has quit [Ping timeout: 250 seconds]
zopsi has joined #nixos
jasongrossman has quit [Ping timeout: 252 seconds]
<colemickens>
So I'm packaging Sway 1.0-HEAD for nixpkgs, and I've run into an issue in the last two days with building.
<colemickens>
The project uses `#include <drm_fourcc.h>` without the leading `libdrm/` prefix.
<colemickens>
But, I can load a nix-shell with libdrm and pkg-config, and use pkg-config and see that `libdrm/` is a directory that should be searched.
<colemickens>
Am I doing something wrong? `libdrm` is part of buildInputs...
<colemickens>
nix-shell -f /etc/nixpkgs-sway -A wlroots
<colemickens>
pkg-config --cflags libdrm
<colemickens>
shows -I.../include/libdrm
<colemickens>
so shouldn't meson/gcc be able to find drm_fourcc.h ?
fragamus has joined #nixos
jasongrossman has quit [Quit: ERC (IRC client for Emacs 26.1)]
paulyoung85_ has quit [Quit: Page closed]
Mateon1 has quit [Read error: Connection reset by peer]
Mateon2 has joined #nixos
Mateon2 is now known as Mateon1
hamishmack has quit [Ping timeout: 268 seconds]
rprije has joined #nixos
slack1256 has joined #nixos
Rusty1 has quit [Quit: Konversation terminated!]
halfbit has quit [Ping timeout: 276 seconds]
sevanspowell has quit [Ping timeout: 256 seconds]
andersk has quit [Remote host closed the connection]
shabius has quit [Remote host closed the connection]
shabius has joined #nixos
shabius has quit [Remote host closed the connection]
mayhewluke has quit [Ping timeout: 244 seconds]
mayhewluke has joined #nixos
mokasin has joined #nixos
<mokasin>
Hi. I've set users.mutableUsers=False and this has apparently deleted all my passwords. I only can login to my user account using public keys. What do I do?
<jackdk>
you can set `users.extraUsers.someuser.hashedPassword = "...";`?
<mokasin>
I cannot modify my system any more, I can't log into root -.-
<mokasin>
I only can log into an unprivileged user
<jackdk>
no sudo?
<mokasin>
nope
<mokasin>
it asks for a password that I apparently do not have, my old one is invalid
<jackdk>
boot an older build and see if can get in that way? otherwise I'd say you might need to boot a livecd, chroot in, fix configuration.nix and rebuild
<mokasin>
Yeah, I feared so.
<jackdk>
check your kernel for privilege escalation exploits?
<mokasin>
What are the passwords if not specified in the configuration.nix?
hyper_ch has quit [Read error: Connection reset by peer]
<mokasin>
Now I have to figure out, how to select a grub entry on a VPS -.-. Darn
hyper_ch has joined #nixos
<jackdk>
if you still have read access to configuration.nix you can pull that out and build a new vps so you don't have downtime
<mokasin>
Yes, sadly that seems to my only option.
<slack1256>
^ this
<mokasin>
It makes sense that mutableUsers=False resets passwords, but there should be a warning, if you do not set passwords in the config file -.-
<mokasin>
I was setting the option a bit blue eyed and fucked up
<mokasin>
When chrooted to a NixOS, how can I rebuild the configuration? I get errors because of bad heap size and that system has not been booted with systemd.
<mokasin>
I somehow need to set mutableUsers = true, otherwise I'm locked out of my system
<mokasin>
Hmm, and the issue seems to be, I have no inet access in the rescue console. Damn
<mokasin>
At least not in the chrooted nixos *scratch*
<{^_^}>
[nixpkgs] @dywedir pushed commit from @magnetophon to master « spectmorph: init at 0.4.1 »: https://git.io/fx0BM
viric_ is now known as viric
siers has joined #nixos
<mokasin>
I guess nixos-install is what I was looking for
Maxdamantus has quit [Ping timeout: 244 seconds]
Maxdamantus has joined #nixos
Niall has quit [Remote host closed the connection]
Niall has joined #nixos
orivej has quit [Ping timeout: 252 seconds]
mokasin has quit [Remote host closed the connection]
Tucky has joined #nixos
Thra11 has joined #nixos
Ariakenom has joined #nixos
<manveru>
So... Tried upgrading to 18.09 with my notebook yesterday. But it failed to fetch a bunch of patches and didn't use much of the cache. Any issue open about this already?
<srhb>
manveru: That usually occurs if you're not using the cache for some reason. The patches should have their urls fixed though.
Thra11 has quit [Ping timeout: 252 seconds]
<Myrl-saki>
Wait. WAIT.
<Myrl-saki>
Um, how does `nix-collect-garbage --delete-older-than` work
<Myrl-saki>
The nix-store sets the timestamp to 0, so..
<manveru>
I had hash mismatches as well, I'm on 2G at the moment so I can't give you much details. Will try to write up something when I get to the airport. Just worrying why hydra didn't catch that
<manveru>
Myrl-saki: /nix/var can have timestamps
<Myrl-saki>
manveru: Yes.
<manveru>
And the GC roots are there
civodul has joined #nixos
<Myrl-saki>
manveru: Does that mean nix-collect-garbage just checks old symlinsk?
<Myrl-saki>
symlinks*
<Myrl-saki>
s/symlinks/GC roots/
alex`` has joined #nixos
<manveru>
How else would it work?
<srhb>
manveru: If the artifacts are already cached on Hydra, but changed upstream, Hydra doesn't care.
<srhb>
manveru: You're (most likely) seeing the hash mismatches exactly because you're not using cache.nixos.org (in all cases) -- possibly due to network issues, possibly due to local overrides, substituter settings, ...
<manveru>
Ok
<manveru>
Something wrong with my cache config then
<{^_^}>
[nixpkgs] @ryan4729 opened pull request #48450 → protobuf3: support cross compilation → https://git.io/fx00j
<srhb>
Myrl-saki: Specifically it looks at the timestamps on the profile symlinks iirc. :)
<Myrl-saki>
Ah, thanks. :P Wouldn't it be fairer to call it nixos-collect-garbage or similar then?
goibhniu has joined #nixos
thc202 has joined #nixos
johanot has joined #nixos
alex`` has quit [Ping timeout: 260 seconds]
alex`` has joined #nixos
fragamus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Synthetica has joined #nixos
Phillemann has joined #nixos
<Phillemann>
I'd like to have the path to the fontconfig library. The one I get when adding fontconfig as a buildInputs.
<Phillemann>
Not sure how to get that.
<Phillemann>
nix-build only gets me a /bin directory
Itkovian has joined #nixos
<sphalerite>
Phillemann: nix-build '<nixpkgs>' -A fontconfig.all
<Phillemann>
Ah, interesting
<Phillemann>
Yes, that works. Thanks!
<sphalerite>
or .lib if you only want the lib output
<sphalerite>
`nix eval nixpkgs.fontconfig.outputs` can tell you what's available :)
<Ralith>
Phillemann: try fontconfig.dev?
<Phillemann>
Ralith: Ah, that's even more precise, great.
<Phillemann>
I'm trying to manually patch an executable, and I'm juts missing libstdc++.so.6 now.
<Phillemann>
That's somehow part of the gcc package.
<{^_^}>
[nixpkgs] @joachifm opened pull request #48452 → A few python package updates → https://git.io/fx0aq
Khetzal has joined #nixos
<sphalerite>
why is libgit2 an older version than libgit2_0_27? Shouldn't we only have version-specific things for *old* versions of libraries that are still needed by other things?
zgrep has quit [Quit: This is me *really* leaving.]
<sphalerite>
but since I've started doing nix stuff for work, I've decided to try not to do much on my own time anymore. I don't want to stop enjoying it :p
<sphalerite>
thank you :D
<rawtaz>
i thought you were studying, what are you working with?
orivej has joined #nixos
<siers>
sphalerite, congratulations! :)
<sphalerite>
rawtaz: graduated in June, just started working for Mayflower on Monday. Currently my main task is doing nix stuff — packaging stuff and such
<sphalerite>
siers: ty :)
__Sander__ has joined #nixos
<rawtaz>
wow cooll. is the entire purpose of the company to provide nix support?
<sphalerite>
haha no, there's general software dev and agile consulting (focused on the German-speaking market) https://mayflower.de/ and more recently started doing nix consulting (with more english) as well https://nixos.mayflower.consulting/
<steveeJ>
is anyone else hitting issues with fprintd since the upgrade to 18.09? here it can't find the enrolled fingerprints anymore
<tazjin>
did something change with how/when pkg-config's paths are constructed in 18.09? It used to just work in normal userspace for me (for linking openssl in particular), now it only works inside of nix-shell or nix-build
<{^_^}>
[nixpkgs] @peti pushed 3 commits to haskell-updates: https://git.io/fx0iM
sigmundv has joined #nixos
<tobiasBora>
Hello,
<tobiasBora>
I'd like to know, is it possible to "pin" some packages, so that even if there is a newer version the system version is used instead? Indeed, I don't mind to update the full set of LaTeX packages...
<sphalerite>
or you can put it in a fixed nix expression which uses builtins.fetchurl or similar to get the nixpkgs version and you always install that expression
<sphalerite>
I prefer to stick to using overlays only when I need to replace a dependency, in most other cases it's an unnecessary layer of indirection/complexity/confusingness ;)
<sphalerite>
then use pinnedPkgs.texlive.combined-full or whatever
<tobiasBora>
sphalerite: and that's why I want to "pin" texlive, because when I want to have a newer package, I run "nix-channel update && nixos-rebuild switch", it updates everything
<tobiasBora>
sphalerite: great thank you!
<tobiasBora>
Hum... I've the feeling that nix is rebuilding sagemath... Is it normal?
<tobiasBora>
It would be great if it would be possible to have something like "unstable", except that the packages moved to unstable are always the one built by Hydra
<sphalerite>
tobiasBora: nixpkgs-unstable advances quite quickly, before everything has been built. I'd suggest nixos-unstable if you want everything to be in the binary cache
<tobiasBora>
sphalerite: nixpkgs-unstable and nixos-unstable are not the same??
<Profpatsch>
gchristensen: Do we have a place where we document build-support/?
YellowOnion_ has quit [Ping timeout: 244 seconds]
<gchristensen>
not that I know of
<Profpatsch>
Then we should make like you made with lib/ and start that. :P
<gchristensen>
:)
<sphalerite>
yeah build-support has some amazing stuff
<Profpatsch>
gchristensen: What did you use to write the XML for the lib/ documentation?
<tobiasBora>
gchristensen: you may answer my question then, I'm installing sagemath, and in the last 2 hours I just saw "doctest". Is it possible to disable these tests?
<Profpatsch>
Because I’m going crazy when I have to write long <tagtagtag> stuff by hand.
<gchristensen>
must be doctests for sagemath itself, I don't know anything about sagemath
<gchristensen>
Profpatsch: I have some YASnippets that I've used and find helpful
<gchristensen>
I could do a screencast about how I do it
<Profpatsch>
Alternatively, I could start by writing something on nixos.wiki and then we move it to docbook afterwards.
<tobiasBora>
gchristensen: hum maybe yes, but nix is running these tests right? And as I don't want to compile, just install, I don't really mind to run 2 hours (at least) of tests ^^
<Profpatsch>
gchristensen: Just the yasnippet code is enough (at least for me)
<symphorien>
hydra builds sage, so if you build locally this likely means that the build will fail
<gchristensen>
Profpatsch: note the docs I'm trying to write are in the Reference section of https://www.divio.com/blog/documentation/ -- ie: not writing much expository, and not even use-case docs
<gchristensen>
Profpatsch: just: this is the arguments of the function and the functions' behavior.
<gchristensen>
Profpatsch: I've tried combining all four in to the same document and it is not good. (thanks to joepie91 for setting me straight on that.)
<gchristensen>
Profpatsch: but... I should probably write docs on how I write these docs...
jasongrossman has joined #nixos
<{^_^}>
[nixpkgs] @Mic92 opened pull request #48460 → postfix: add setgid wrapper for postqueue/postdrop → https://git.io/fx0Nc
<tobiasBora>
symphorien: I'm not sure to follow. You manually download sage?? So what's the point of using nix then? ^^
<symphorien>
go on a successful build and click "help": "You can install this package using the Nix package manager from the command-line: nix-env -i /nix/store/jywvwahnav864iry6d60g1qmghb15dw2-sage-wrapper-8.3"
YellowOnion has joined #nixos
rprije has quit [Remote host closed the connection]
<Profpatsch>
gchristensen: tbh I hate snippets. If we need snippets we should generate the resulting code instead.
rprije has joined #nixos
<Profpatsch>
It’s basically copy & paste coding made even easier by the editor.
strobelight has joined #nixos
<gchristensen>
Profpatsch: eh, I don't like them either. my view is the job needs doing, this is the output I'd like to see, and however it is done is great
<gchristensen>
the semantics are very good though, and are how I'm able to implement doctests in less than 50 lines of python
dvim has joined #nixos
<gchristensen>
it isn't like the docs need to be rewritten over and over, and that is why I haven't tried to make much more of a tool to do it: it is pretty much only done once
<tobiasBora>
symphorien: and if I do so it won't re-run all the tests? That's interesting. But it means that I can't use declarative approch right?
<gchristensen>
it can and will change, and the position is dynamically updated on every build of the docs =)
<symphorien>
tobiasBora: I you do it it is alreeady built and tested, you just download the result
<ng0>
can someone give me the gist for your motivation to change TMPDIR in the Linux sandbox builds to /build instead of /tmp? What kind of builds did you encounter that could store their TMPDIR in RPATH as the release of 2.0 states?
<tobiasBora>
sphalerite: Hum, I just got a message "All tests passed!" from my last install :D But then I don't get it, why can I build it, but not Hydra?
<Profpatsch>
Ah, I see.
<gchristensen>
Profpatsch: this serves a very nice check, too!
<Profpatsch>
By doing a grep for $NAME[[:whitespace:]]= I suspect? :)
<gchristensen>
no!
<gchristensen>
what do you take me for, some sort of Bash abuser? ;)
<{^_^}>
[nixpkgs] @lheckemann pushed commit from ryan4729 to staging « protobuf3: support cross compilation »: https://git.io/fx0Ah
<Profpatsch>
gchristensen: No, but some sort of nix abuser. :pP
<Profpatsch>
But yeah, that’s nice.
<gchristensen>
:)
<tobiasBora>
symphorien: Hum, I just got a message "All tests passed!" from my last install :D But then I don't get it, why can I build it, but not Hydra?
<symphorien>
timeout maybe ?
<tobiasBora>
symphorien: Hum, in fact it seems that the new install fails: Error loading server extension sagenb_export.nbextension
<symphorien>
look at the log on the web interface of hydra
<tobiasBora>
No module named sagenb_export.nbextension
hyper_ch2_ has joined #nixos
hyper_ch2 has quit [Ping timeout: 256 seconds]
ryantrinkle has quit [Ping timeout: 246 seconds]
mayhewluke has quit [Ping timeout: 246 seconds]
strobelight has quit [Remote host closed the connection]
<ma27>
does anybody know how to connect to a wpa2 enterprise network using `networking.wireless.networks` or is it preferred to use something like `wpa_gui` then?
<sphalerite>
the description of the option explains how :)
<sphalerite>
I wouldn't use it myself though, because plaintext passwords in the store aren't fun
<sphalerite>
I just use wpa_supplicant and configure it statefully by editing /etc/wpa_supplicant.conf myself
dvim has quit [Read error: Connection reset by peer]
<ma27>
sphalerite: first of all thank you, I only had a look at the manual (it's never mentioned there, I'll file a PR for that later)
<sphalerite>
:)
<ma27>
but yeah, it doesn't seem to be a good idea to specify credentials for wpa2 enterprise in the store, I'll see if I find a workaround, otherwise we'd have to wait until Nix can encrypt store paths ...
rprije has quit [Remote host closed the connection]
rprije has joined #nixos
dvim has joined #nixos
rprije has quit [Remote host closed the connection]
<ma27>
sphalerite: so AFAICS it's possible to specify a hashed password. I'll see if that works and document this in the NixOS manual %)
<gchristensen>
ma27++
<{^_^}>
ma27's karma got increased to 1
<sphalerite>
ma27: that probably depends on the auth method in use
ThatDocsLady has quit [Remote host closed the connection]
ThatDocsLady has joined #nixos
<ben>
Is there a way to install manpages with nix-env yet?
<sphalerite>
ben: yes. It's a great fuss though :/
<srhb>
ma27: That won't really help though
IRCsum has joined #nixos
<nschoe>
Hi everyone! I'm wondering if any of you had already managed to package ROS nodule "joy" (or "joystick_drivers")?
<Zajcev_>
guys, really silly question excuse me but ... when I have only services.openssh = enable in nixos container and I'm sshing to that container, where am I defacto connecting to?
<Zajcev_>
I can't get containers in my brain ...
<nschoe>
I'm trying to make it the same way akru (github.com/airapkgs) did, but for _some_ reason, the joyConfig.cmake file doesn't get generated, so any `catkin` command fails with "missing joyConfig.cmake file". I'm reaching the end of what I can think of. Especially since I really adapted akru's default.nix files so I don't get it.
o1lo01ol1o has joined #nixos
trevorriles has joined #nixos
rprije has quit [Remote host closed the connection]
hiroshi- has joined #nixos
rprije has joined #nixos
johanot has quit [Quit: WeeChat 2.2]
<sphalerite>
ma27: oh yeah. If the hashes password is enough to authenticate you it will of course let anyone who can read your store authenticate as you as well.
lohfu has quit [Quit: WeeChat 1.9]
hiroshi has quit [Ping timeout: 244 seconds]
hiroshi- is now known as hiroshi
lohfu has joined #nixos
<NickHu>
Is fetchFromGitHub buggy? I have two derivations built in the same file, but one fails in the strangest way - it seems like the src passed to it is actually the src for the other
<sphalerite>
NickHu: I highly doubt it. Can you share the file? :)
<gchristensen>
sounds like a sha256 mixup
<NickHu>
sphalerite: Ah, so what had happened was that they accidentally shared the same sha256
<NickHu>
Yes, and then it failed somewhat silently
<sphalerite>
yep
<gchristensen>
it is unfortunate, but there is very little to do. the sha256 is the only important detail Nix can know about when considering a source. if you give it the wrong (or a duplicate, or old...) sha it has no idea it isn't right
<NickHu>
I understand
<gchristensen>
none of the other details (like name or url) matter because those can change
Rusty1 has joined #nixos
halfbit has joined #nixos
rprije has quit [Ping timeout: 252 seconds]
rprije has joined #nixos
asymmetric has joined #nixos
jperras has joined #nixos
<nschoe>
What nixOS package should I include to have <linux/joystick.h>?
rprije has quit [Remote host closed the connection]
<asymmetric>
hey, i'm having a weird problem when i run `nixos-rebuild switch`: it fails with `Failed to start local-fs.target: Unit home.mount is masked.`
<asymmetric>
sometimes the unit chnges: Failed to start local-fs.target: Unit \x2esnapshots.mount is masked.
<asymmetric>
these are btrfs volumes btw (home and snapshots)
<{^_^}>
[nixpkgs] @jhenahan closed pull request #41333 → haskell-modules/configuration-ghc-8.4.x: bugfix - hoopl is still a core library → https://git.io/vhcUA
<{^_^}>
[nixpkgs] @romildo opened pull request #48465 → deepin-wm: init at 1.9.32 → https://git.io/fxEqz
Sigyn has quit [Quit: People always have such a hard time believing that robots could do bad things.]
Sigyn has joined #nixos
jabranham has joined #nixos
asymmetric_ has joined #nixos
asymmetric has quit [Ping timeout: 252 seconds]
<nschoe>
Nobody for the joy node on ROS? :/
<sphalerite>
,locate linux/joystick.h
<{^_^}>
Found in packages: linux.dev, androidndk, linuxHeaders, mentorToolchains.armLinuxGnuEabi
<nschoe>
sphalerite, thanks for your answer :) Turns out, I already had it in my dev env apparently, so the problem is not here :/
<nschoe>
sphalerite, any experience with ROS modules by any chance? ^^
<sphalerite>
nnnnnnope. I don't even know what ROS is
[Leary] has quit [Remote host closed the connection]
Lears has joined #nixos
<gchristensen>
Robot Operating System iirc
pointfourone has quit [Remote host closed the connection]
<nschoe>
that's it, yes.
<nschoe>
sphalerite, well no luck then ^^
mystfox has quit [Max SendQ exceeded]
mystfox has joined #nixos
lrvick has quit [Ping timeout: 276 seconds]
<nschoe>
Where is the nixOS page that explains how to go through the build process step by step? I'd like to inspect the soruc ebeing unpacked, etc. In order to find where my issue comes from :/
<hodapp>
"error: cannot download postman-6.2.5.tar.gz from any mirror" hrmph
<hodapp>
oh, looks like Postman broke their filename conventions
<nschoe>
Stupid question: are the pills kept up-to-date?
<jasongrossman>
nschoe: I think not.
<nschoe>
erf ok.
<nschoe>
I remember some changes last time I read them.
<jasongrossman>
Oh cool!
<nschoe>
Anyway it should still help me.
<jasongrossman>
Maybe they are then.
<nschoe>
jasongrossman, no no what I meant is:
<nschoe>
last time I read them, some commands epxlained on the pills were not working anymore in my nixOS installation, so I wondered if the pills were updated.
<jasongrossman>
That sounds very plausible, sadly.
<jasongrossman>
There ARE people who work hard on the documentation in general ... but not enough of them. :-(
<hodapp>
I'd work on docs if I understood stuff better...... which I have a hard time doing without docs
<jasongrossman>
Or maybe what I should say is that the ratio of people doing work that necessitates changes to the documentation to people changing the documentation is not real good.
<jasongrossman>
hodapp: Ditto!
<hodapp>
I can mudball stuff okay without them and still get by where I need to
<hodapp>
but... that's not really understanding things
<{^_^}>
[docker] @domenkozar pushed to 2.1.4 « Bump to Nix 2.1.4 »: https://git.io/fxEss
<acowley>
I have this nasty situation where I have a library that I can load in python, and in ipython, but jupyter gets a std::bad_cast exception
<nschoe>
Yeah I'd love to contribute too, but honestly any operations I do on nixOS is a miracle when it works.
<{^_^}>
[docker] @domenkozar opened pull request #3 → Bump to Nix 2.1.4 → https://git.io/fxEsW
<{^_^}>
[docker] @domenkozar pushed to 2.1.3 « Bump to Nix 2.1.3 »: https://git.io/fxEsB
<{^_^}>
[docker] @domenkozar opened pull request #4 → Bump to Nix 2.1.3 → https://git.io/fxEsE
<acowley>
I haven't yet tracked down a way to replicate the suggested "fix" in that issue thread
fragamus has joined #nixos
<acowley>
But it's made more difficult by the fact that I don't understand what could be happening there
<acowley>
It doesn't seem that jupyter links libstdc++
<acowley>
One current guess is that my library is linking libstdc++ both statically and dynamically at different points in the build (since ldd on a resulting .so does show libstdc++), but I'm not sure how to prevent that.
<dhess>
Unfortunately it appears that it didn't build any haskellPackages because of the cached failure of previous GHC attempts
silver has quit [Quit: rakede]
<dhess>
gchristensen: elvishjerricco pointed out that GHC's build system doesn't use -j, so it was safe to turn parallel builds back on just for GHC (not haskellPackages, which do build with -j) and then the aarch64 builder was able to finish it before the 10 hour limit :)
<gchristensen>
nice!
patrl has joined #nixos
<gchristensen>
dhess: seems it still has many haskellpackages to build?
<gchristensen>
anyway, I restarted all the aborted and failed builds.
init_6 has quit [Ping timeout: 252 seconds]
aleph- has joined #nixos
lorilan has joined #nixos
<dhess>
gchristensen: oh thanks
<dhess>
yeah it was not looking past the cached failures from previous attempts for some reason
<dhess>
that should do the trick. haskellPackages.tls will fail on tests but lots of other packages should build successfully
<gchristensen>
something something hard things in computer science
<dhess>
tls is a blocker at the moment though for a lot of useful stuff
Tucky has quit [Remote host closed the connection]
<dhess>
the other thing that's weird is that it only seems to be building the aarch64 part of the job, not the x86_64 part
<dhess>
during evaluation, I mean
<Taneb>
Is there a way to convert a single user nix installation into a multi user one?
<kai_w>
mpickering: I don't quite understand - is there an example that shows this? sorry was away for a bit.
<lorilan>
hi, I'm trying to install the build tool pants. I end up with an error so I opened an issue : https://github.com/NixOS/nixpkgs/issues/48300 . Today I investgated with a colleague and It seems that the error originates from the fact that the derivation download the sources from here : https://pypi.org/project/pantsbuild.pants/1.7.0/#files which contains a `native_engine.so` for mac (after searching the magic number). Now I'm trying to fix the derivation.
<lorilan>
I'm wondering If I can use the wheel file instead of the sources but the python version is cp27. I'm afraid it's not supported by the nix python packages is it ?
<zduch4c>
hello, how can I make dhcpcd or whatever controls iPhone ethernet tethering connect automatically if an iPhone is detected? under NetworkManager it works as intended, plugging it in makes it instantly work, but when I use wpa_supplicant & dhcpcd it doesn't
ryanofsky has joined #nixos
zduch4c has quit [Quit: ERC (IRC client for Emacs 26.1)]
zduch4c has joined #nixos
aleph- has quit [Ping timeout: 268 seconds]
zduch4c has left #nixos ["ERC (IRC client for Emacs 26.1)"]
<exarkun1>
Can I refer to node addresses in nixops configuration somehow? For example, to write a configuration file on one node that points it at a second node?
<clever>
exarkun1: gchristensen's example relies on nixops updating /etc/hosts, and your app supporting /etc/hosts (some apps ive delt with dont!!)
<gchristensen>
oh true
<clever>
the other option is nodes.backend1.config.networking.publicIPv4
<gchristensen>
or any other attribute you'd like to use
justbeingglad has joined #nixos
patrl has quit [Ping timeout: 250 seconds]
<exarkun1>
referring to ex-logical-multi.nix and /etc/hosts, is the implication that nixops always automatically puts entries in /etc/hosts for the different nodes in the configuration? So I can always just refer to nodes by the name I refer to them in my nix configuration?
<exarkun1>
(oh if I scroll up ever so slightly from there, that's just what it says)
<{^_^}>
[nixpkgs] @grahamc pushed 2 commits to release-18.09: https://git.io/fxE6O
aleph- has joined #nixos
<fiatjaf>
oh
<fiatjaf>
now I see
<fiatjaf>
so, nix will build everything from scratch, including the compiler, every time I call nix-build?
<clever>
it will cache things in /nix/store/
<clever>
and if the inputs havent changed, it will reuse them
<clever>
and if cache.nixos.org has a pre-built copy, it will download that
<fiatjaf>
that's for each package individually, right? so I just change something in the source code for my own package it won't try to rebuild everything
<{^_^}>
[nixpkgs] @ryantm pushed 2 commits to haskell-updates: https://git.io/fxEXG
<elvishjerricco>
dhess: Congrats :)
<dhess>
elvishjerricco: it's all you buddy :)
<elvishjerricco>
dhess: So is that on master now? Everyone will be able to benefit from the cache soon?
<dhess>
yes
<elvishjerricco>
fantastic news. Probably worth backporting to 18.09
<dhess>
again, haskellPackages.tls is broken so lots of stuff won't get very far, but hopefully now we can point the tls maintainers at it and get a fix. (Or possibly it's related to building on a CI server and just times out, unrelated to aarch64)
<dhess>
I will take a look at that later if someone doesn't beat me to it
aleph- has quit [Ping timeout: 252 seconds]
arq1 has joined #nixos
<elvishjerricco>
There may be 96 cores on that thing, but they're not the fastest cores, are they? :P I'm watching the build log of tls and it moves quite slowly: https://hydra.nixos.org/build/82810583/nixlog/1/tail
<elvishjerricco>
Or tls just makes GHC take forever in general
<dhess>
They are not
<dhess>
the community builder has much faster single-thread performance
<dhess>
but only 64 cores
rprije has quit [Remote host closed the connection]
<elvishjerricco>
dhess: What is the community builder used for?
rprije has joined #nixos
<dhess>
unofficial builds
<dhess>
e.g., I use it as a remote builder for my own work
<elvishjerricco>
Ah. So it wouldn't be suited to try to direct haskellPackages builds to it.
<dhess>
well it's how I got the aarch64 GHC/haskellPackages stuff vetted, so it works great... it's just not something you want to be relying on for production builds as there are a number of trustedUsers on that box
<dhess>
maybe if there's enough demand for haskellPackages.aarch64-linux we can get one of these faster Packet.net aarch64 boxes in the Hydra pool :)
<samueldr>
only the 96 cores machines are available
<dhess>
samueldr: available for what?
<samueldr>
general availability; the other community builder is AFAIUI not available; don't know the exact details, probably something about the hardware being scarce
<dhess>
ahh ok, interesting.
<{^_^}>
[nixpkgs] @andir pushed commit from @florianjacob to release-18.03 « gitea: 1.5.1 -> 1.5.2 »: https://git.io/fxEMs
<elvishjerricco>
TIL you can use `--store` with `nix log` to just see the log of a derivation from the cache. Don't even have to build or copy the build result or anything.
<{^_^}>
[nixpkgs] @Ma27 opened pull request #48476 → nixos/manual: document WPA2 enterprise configuration when declaratively configuring wifi networks → https://git.io/fxEDX
<samueldr>
hmm, without --store it works here, I guess elvishjerricco it's needed if it's not for the cache?
<samueldr>
sorry, from the configured caches*
<elvishjerricco>
samueldr: You're right. I just assumed I needed --store and never tried it without :P
<symphorien>
gnome has this option, apparently, but not xfce, so it may depend on your DE
freeman42x has joined #nixos
<tnks>
symphorien: I'm just running plain i3 over X.
<tnks>
maybe it's worth losing a pointer every so often... to keep my slim windows workflow
oida has joined #nixos
<__monty__>
fiatjaf: I was talking about keyboard layout. Do you use the most common en_US qwerty layout? Some might define backspace differently and terminals and backspace have a, let's say "interesting", interaction (other control characters too). It's not a nix-shell property, nix-shell just runs bash and backspace works fine for me. I'm just taking a guess at what might be the problem.
rprije has quit [Remote host closed the connection]
rprije has joined #nixos
smolboye has quit [Quit: WeeChat 2.2]
smolboye has joined #nixos
<fiatjaf>
clever, I don't have a Setup.hs
<fiatjaf>
should I?
<fiatjaf>
cabal fails, it says: "Warning: The package list for 'hackage.haskell.org' does not exist."
<fiatjaf>
then it says all my dependencies are missing
<fiatjaf>
oh, I'm using the global cabal, there isn't a nix-based cabal
halfbit has joined #nixos
<fiatjaf>
I don't understand why. the Gabriel439 tutorial is very confusing.
<jabranham>
how do people usually manage multiple machines with /slightly/ different configuration.nix files? Keep files like common.nix, machine1.nix, and machine2.nix (which import common.nix) and then symlink /etc/nixos/configuration.nix to machine1.nix on that machine and machine2.nix on the other?
<{^_^}>
[nixpkgs] @rycee pushed commit from @joakim1999 to master « eclipses: add Eclipse IDE for Java Developers 4.9 »: https://git.io/fxE7l
<{^_^}>
[nixpkgs] @rycee closed pull request #48395 → eclipse: add Eclipse IDE for Java Developers 4.7 and 4.8 → https://git.io/fxRih
<elvishjerricco>
jabranham: You could have a common.nix file with all the common configuration, then separate configuration.nix files for each machine with `imports = [./common.nix];` Then you'd either symlink /etc/nixos/configuration.nix to the appropriate one, or just set `nixos-config` in the NIX_PATH or with `-I`
<srhb>
Ericson2314: So presumably cabal2nix thinks it won't build for some reason.
<jabranham>
elvishjerricco: thanks. Your first suggestion is what I was thinking of.
<gchristensen>
ikwildrpepper: thank you for nixops deploy's --dry-activate command <3 <3 <3
<ikwildrpepper>
gchristensen: why thank me?
<gchristensen>
because you wrote it!
<ikwildrpepper>
did I make that? :D
<ikwildrpepper>
I like to forget things
<gchristensen>
you did: github.com/nixos/nixops/commit/d72c2233d2133d828a895601bfd53fa4cca3d9b8
<gchristensen>
it is so cool how it copies all the closures over and lists the services which will restart, meaning I can ignore the dry deploy step, wait until it is finished and then review -- and not be stressed out watching the deploy.
<{^_^}>
[nixpkgs] @romildo opened pull request #48481 → deepin-metacity: init at 3.22.22 → https://git.io/fxE77
Hamilton has joined #nixos
<Hamilton>
So I'm happy with Ubuntu+Xfce...why should I try nixOS? (someone suggested it)
<gchristensen>
it makes it not terrifying to have `nixops deploy` be run in BuildKite.
<fiatjaf>
Hamilton, you shouldn't
rprije has quit [Remote host closed the connection]
<fiatjaf>
unless you understand how nix works
<dhess>
gchristensen: ohhh is that how you're doing CI for NixOps networks?
rprije has joined #nixos
<gchristensen>
dhess: yeah
<dhess>
good idea
<gchristensen>
fiatjaf: that isn't very helpful. lots of people have switched to NixOS then learned how Nix works
<symphorien>
Hamilton: for example you can upgrade knowing that if something goes wrong you can rollback.
<fiatjaf>
gchristensen, but it is reeeeally not straightforward
<symphorien>
but the learning curve is steep, fiatjaf has a point
<fiatjaf>
perhaps people should experiment with nix first, before going to nixos
endformationage has quit [Ping timeout: 252 seconds]
<fiatjaf>
that's what I'm doing, anyway
<fiatjaf>
or trying to
<mog>
fiatjaf, i see you are taking the cartmanland aproach
<gchristensen>
that is a perfectly fine way to go, but isn't strictly necessary and many people have successfully taken the jump directly with the support of our happy-to-help community
<symphorien>
fiatjaf: well the benefit of nix is far bigger on nixos, in my opinion, and it helps making the learning more "rewarding"
<gchristensen>
Hamilton: if you're happy, maybe you shouldn't switch. but if you'd like to learn something new and try a new approach to how systems can be managed, check it out! you could try it in a VM without being too committal :)
silver has joined #nixos
<mog>
Hamilton, what i like about nixos is my machine is unbrickable. it never get crufy with 100s of random dev dependencies
<fiatjaf>
I think switching directly could be a good approach too, but only if you have nothing to lose
<mog>
i have upgraded in place for 5 versions now without any troubles
<fiatjaf>
if you're not actively using your OS for a lot of stuff and depend on it for your workflow currently
<__monty__>
fiatjaf: About the "cabal there is no list..." stuff. You need to run cabal update at least once. I don't think you need to keep it up to date because of nix but I'm not sure on that one.
<samueldr>
initially jumped in nixos both feet in with only a superficial look at nix, never regretted it, though it took some work to get my usual setup just right; but my setup isn't that standard
<fiatjaf>
__monty__, should (which cabal) return a path like (/nix/store/blablabla-cabal/whatever) ?
<Hamilton>
fiatjaf, symphorien gchristensen mog thanks guys...should watch a couple of YT vids
<fiatjaf>
Hamilton has left and we're still here talking about his point, what a great community!
<Hamilton>
fiatjaf, no I'm here
<fiatjaf>
haha
<samueldr>
??!?
<Hamilton>
:D
<gchristensen>
Hamilton: interesting! what do you like to see in a video about an os?
<__monty__>
symphorien: I would never have bothered with nixos for lack of time. Nix is the gateway drug that has me almost taking the plunge. Just a couple more hands in hair crises/new computer, whichever comes first and I'll switch. I definitely still needed pacman to hold my hand while learning the ropes.
<Hamilton>
hmm..it seems the main thing about nixos is package management?
gamblerx has joined #nixos
<gchristensen>
reproducibility is a key point of it, which is supported by the package manager / build tool Nix
<symphorien>
well the package manager also manages system configuration (files in /etc, that is)
Berra has joined #nixos
<Hamilton>
And is it suitable for daily driver? Do you guys use it on your eveyday machine?
<samueldr>
daily driver, local home server and distant servers
<gamblerx>
hi everyone - having a frustrating issue on x1 carbon with nixos 18.09 (jellyfish). using KDE, after waking up laptop (after it has been idle) the screen flashes on and off persistently and I have to reboot
fragamus has quit [Ping timeout: 276 seconds]
<samueldr>
daily driver both for personal machine and work machines
<gamblerx>
can anyone point me in the direction of a solution?
<gchristensen>
I use it on my work laptop, personal laptop, and ~many servers.
<hodapp>
I use it on my desktop, work laptop, and one server
<gchristensen>
gamblerx: after you reboot try looking in the logs via: `journalctl -e --boot=-1` and see if there is something interesting in there
<gamblerx>
thanks - random thought - i've just remember i haven't configured any swap memory. any chance that could affect it?
<gchristensen>
I think that only means you can't hibernate, but it can go in to a "light sleep" sort of thing.
<gchristensen>
if it is trying to hibernate ... it would cause problems!
<gamblerx>
hmmm... you know that might be the issue actually. i'll try configuring nix with a swapfile
<samueldr>
I doubt it, if it had hibernated it would have (I think) powered off, you would have been faced with the initial steps of the boot process AFAIK
<samueldr>
gamblerx: the x1 carbon is failry recent, right?
<symphorien>
I think systemctl hibernate just means systemctl suspend when there is no swap
<samueldr>
did you configure a specific kernel or using the default nixos configured one?
Thra11 has quit [Ping timeout: 276 seconds]
<gamblerx>
yep fairly recent (6th gen), just used the default nixos kernel
<samueldr>
gamblerx: you can try using the latest stable kernel by adding this to your configuration: boot.kernelPackages = pkgs.linuxPackages_4_18;
<samueldr>
it default to the last LTS, which is 4.14
pointfourone has joined #nixos
<gamblerx>
aha! many thanks :)
pointfourone has left #nixos [#nixos]
<samueldr>
(no guarantees it'll help, but fresh laptop hardware sometimes fare better on the freshest kernels :))
<gamblerx>
thanks, yeah it hadn't occured to me to look at a fresher kernel. just about to reboot with swapfile configured :)
rprije has quit [Remote host closed the connection]
<symphorien>
ah I thought a revision of the package, not nixpkgs
<srhb>
Oh, I might have misunderstood as well :)
<jabranham>
sorry, maybe I wasn't clear. Package X has release 1, which is in nixpkgs. There have been commits to package X's git branch since release that I want.
mayhewluke has quit [Ping timeout: 245 seconds]
<symphorien>
in the alternate case: foo.overrideAttrs(old: { name="foo-myversion"; src = ...; })
<jabranham>
symphorien: I thought overrideAttrs was deprecated/discouraged?
Hamilton has left #nixos ["WeeChat 1.9.1"]
<symphorien>
the ... are to be replaced by for example fetchzip { url = ....; sha256 = ...; }
<symphorien>
overrideDerivation is
<symphorien>
overrideAttrs is fine as far as I know
dvim has quit [Ping timeout: 252 seconds]
<jabranham>
ah, gotcha. Thanks. I'll try this after work today.
<{^_^}>
[nixpkgs] @matthewbauer pushed 4 commits to master: https://git.io/fxENN
<{^_^}>
[nixpkgs] @Lassulus opened pull request #48482 → ttyplot: init at 1.0 → https://git.io/fxEAO
Thra11 has joined #nixos
<kalbasit[m]>
ldlework: the way you define the overlays in your dotfiles makes them available to the dotfiles themselves but unavailable to `nix-env`, `nix-shell` and alike. Do you have a workaround for this?
<dhess>
Anyone around who has haskellPackages commit powers?
<srhb>
dhess: What's up?
Itkovian has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<dhess>
gchristensen: quite a few of those will need to be rebuilt. There's a bug in cryptonite on aarch64 and I suspect a reasonable number of those will depend on it. But I've got a fix for that coming in a few minutes
<dhess>
srhb: so there's the "re-enable tests on cryptonite" patch and then there's the "fix cryptonite on aarch64" patch. I can do them as one PR or I can do them separately. Preference?
<dhess>
they're 2 separate commits BTW
<dhess>
dependent but separate. First one re-enabling tests as the only way to be sure the aarch64 patch actually works is to test it :)
dvim has joined #nixos
<infinisil>
I'd go for one PR "cryptonite improvements"
<dhess>
that was my preference as well
drakonis has joined #nixos
philippD has joined #nixos
endformationage has joined #nixos
<fiatjaf>
so, every time I change something in my source code and tell nix to rebuild it, I get a new directory on /nix
nDuff has joined #nixos
<fiatjaf>
isn't that going to take my entire disk space?
<__monty__>
fiatjaf: That's not really how it works.
<nDuff>
Can I reasonably use Nix to *generate* Nix expressions, and then evaluate them once the derivation containing them exists? Is there existing code I can leverage towards that end? (It's not so much the code generation that's a head scratcher as the late loading; tools I see "in the wild" like pypi2nix or the hackage automation/plumbing built into nixpkgs appear to be doing the code generation *before* any evaluation takes place).
Ariakenom has quit [Read error: Connection reset by peer]
<__monty__>
fiatjaf: Well, only the things that changed would add to the store. So a change in your project doesn't mean all your dependencies are replicated. It also depends on how you're building your project. For example I use cabal with nix. Cabal doesn't touch the store while developing so it's impossible to fill up the store that way.
simukis has quit [Quit: simukis]
<mightybyte>
philippD: One thing I had been doing was mucking with "root = builtins.filterSource..." instead of ./.
<colemickens>
When building the system configuration, where/how does it come up with the git version suffix?
<colemickens>
It doesn't work with git worktrees and I'd like to fix it.
<gchristensen>
not sure it is fixable to handle worktrees though...
<fiatjaf>
__monty__, I was thinking about just my project anyway, not all dependencies
<colemickens>
gchristensen: great! thanks
patrl has joined #nixos
<philippD>
mightybyte: could you also share your cabal file?
<fiatjaf>
but now I've started using cabal inside nix-shell, that should be saner
jbgi has quit [Ping timeout: 252 seconds]
<fiatjaf>
however, my nix-shell doesn't work well: when I press backspace I get a new space printed!
<mightybyte>
philippD: No. Is there something in particular I should look for?
<__monty__>
fiatjaf: Are you really resource constrained. Normally the build products don't amount to more than a couple MB.
<mightybyte>
BTW, I've made some subsequent changes to the default.nix file...upgrading to a different revision of nixpkgs, etc and am not seeing the problem after making those changes.
<mightybyte>
So at the moment it's not a blocker, but I was wanting to understand more about what might have been going on.
<mightybyte>
(I can still duplicate the problem with the old default.nix that I pasted above)
<fiatjaf>
__monty__, yes, I am. well, a couple MB is a lot if I rebuild my project 100 every day
<__monty__>
fiatjaf: It'll only keep artifacts around if one of the dependencies changes. Otherwise it overwrites. Do use the cabal new-* family of commands though.
<__monty__>
fiatjaf: Are you on windows perchance?
<fiatjaf>
__monty__, I'm on ubuntu
<philippD>
mightybyte: you could dig through the issues on the nixpkgs repo if something relating to `developPackage` or `callCabal2nix` was fixed. maybe this issue?
hamishmack has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
jperras has quit [Quit: WeeChat 2.2]
<{^_^}>
[nixpkgs] @joachifm closed pull request #48207 → profiles/hardened: Do not limit user namespaces if Nix's sandbox builds are enabled → https://git.io/fxl8G
<dhess>
srhb: if you're wondering why that ofBorg build was so fast, it's because I'm using the aarch64 community builder myself :)
patrl has quit [Ping timeout: 252 seconds]
<dhess>
so don't be alarmed
halfbit has quit [Ping timeout: 250 seconds]
ryjm has joined #nixos
<srhb>
dhess: *scream* I hate fast builds! Or something. ;)
<dhess>
hehehe
<dhess>
I thought it might look odd that the aarch64 build finished well before all of the others
<{^_^}>
[nixpkgs] @srhb opened pull request #48486 → Backport 18.09: haskellPackages.cryptonite: fix for aarch64 and re-enable tests → https://git.io/fxukZ
rprije has quit [Read error: Connection reset by peer]
rprije has joined #nixos
patrl has joined #nixos
nbardiuk has quit [Read error: Connection reset by peer]
<dhess>
srhb: running tests on cryptonite makes sense on 18.09 but the aarch64 fix won't have any effect until 4babe7f799e826a040869728a0be56c68071c0ea and a1600d0aa8f9eee9f6d6cd067d6cbdb1fdd51d46 are also backported, FYI
<srhb>
dhess: sadface
<srhb>
dhess: The patch will be harmless though, right? So I guess we can just merge and it'll work once/if those get in
civodul has quit [Quit: ERC (IRC client for Emacs 26.1)]
<dhess>
yes it will be harmless
justbeingglad has left #nixos [#nixos]
<dhess>
elvishjerricco expressed support for backporting the GHC aarch64 fixes as well, so there is some interest in that
<dhess>
It would be cool if GitHub Issues supported a "sidebar" , so that things like ofBorg updates could be seen but not injected into the comment stream.
<srhb>
dhess: The checks feature does this..
sigmundv has joined #nixos
<{^_^}>
[nixpkgs] @joachifm merged pull request #48452 → A few python package updates → https://git.io/fx0aq
<dhess>
and that cryptonite fix will get us to a bunch more
drakonis has joined #nixos
erasmas has quit [Quit: leaving]
<nDuff>
samueldr, ...as someone actively using aarch64 -- I have a friend with a business he's thinking about starting that would involve a bunch of displays (showing which customer has booked a station and how much time is left). Current plan was to use Pi Zero W hardware for them, but obviously that's an older instruction set that isn't as well-supported. As someone actively working on NixOS-on-ARM, is it in your view worth the extra cost (& power requirements
<nDuff>
-- could probably run Pi Zeros off the displays' USB out; more likely to need a proper power supply for a 3) for hardware using aarch64?
brejoc has quit [Ping timeout: 252 seconds]
<drakonis>
so, i'm going to bootstrap nixos on my desktop without a flash drive, yay or nay?
<samueldr>
sorry, had to move things around in the kitchen
<dhess>
well it's progress... I guess.
<samueldr>
nDuff: no ideas if AArch64 itself makes it more power-hungry, I would hazard a guess that it's not a major factor
<nDuff>
Not sure that it's an AArch64 thing, but the Pi Zero definitely requires considerably less amperage than the 3
Thra11 has quit [Ping timeout: 244 seconds]
jackdk has joined #nixos
<samueldr>
yes!
<nDuff>
which is my point, that there are practical tradeoffs made if we go with the 3 hardware, so I was wondering if in your view the benefits to the 64-bit architecture are worth it. :)
<samueldr>
if you intend to use NixOS: yes double triple yes
snajpa has joined #nixos
<samueldr>
simply because there is a binary cache
<samueldr>
(I don't have much knowledge about the cross-compilation advances, but last time I tried to cross-compile things there was still a lot to do)
<samueldr>
for your use case, if cross-compilation works, then there's not much difference I guess
<drakonis>
there's a package called bees
<drakonis>
i want to use bee puns
<nDuff>
Mmm. Almost 20 years ago making software cross-compile for an embedded Linux distro was my day job; I'm not so sure I want to make it my hobby today, so that's a pretty fair point.
<samueldr>
from an outsider, but an enthusiast, of all ARM things, I would hazard a guess that ARMv8 is simply the present; not the future. ARMv7 and previous are probably going to get less mainstream
<nDuff>
drakonis, ...heh. That's my PR (adding bees). Can't help you on the puns, though.
<samueldr>
and how linux (the kernel) handles AArch64 is great, when boards are mainlined, it just works
<samueldr>
drakonis: make a backup, and *then* do silly things :)
<drakonis>
i don't have the means to back anything up right now :V
<nDuff>
drakonis, ...what's your filesystem? Installing onto a new btrfs subvolume could make things pretty easy.
<drakonis>
i have ext4 on lvm
<samueldr>
I'd say, for *any* distro, any silly shenanigans is risky without backups
<drakonis>
i could probably shrink ext4 to free up space then move everything into it
<drakonis>
into the new partition
<nDuff>
Huh. I'd probably do an in-place ext4->btrfs migration *before* the OS change, though advising that without backups being done prior is a little iffy.
<nDuff>
...because once you're on btrfs, you don't even need a new partition
<nDuff>
different operating systems on different subvols, and there you are.
<drakonis>
isn't btrfs known to hose things?
* samueldr
wonders why Rusty1 sent pine(?) hardware por^W picture
<drakonis>
isn't that mining hardware porn?
<samueldr>
there's the pine logo
<samueldr>
guessing it's a spine forest(?)
<samueldr>
sopine*
<drakonis>
i see a pinecone there
<nDuff>
drakonis, ...on recent kernels? Not that I know of, and bees (which, obvs, I'm using in production) stresses it pretty hard.
<drakonis>
there's a thing to do filesystem conversion on fedora
<drakonis>
should be exciting and risky
aleph- has joined #nixos
<drakonis>
fstransform
<samueldr>
yeah, definitely that cluster thing that was availably for only a short duration :/
<Rusty1>
samueldr: the Arm64 disco brought to mind a recent LUG meeting featuring the Pine64 stuff, so eye candy
<drakonis>
i'll buy a flash drive next time i get out
jperras has joined #nixos
<nDuff>
fstransform is a much more generic tool -- for ext4->btrfs there's built-in btrfs support
<samueldr>
aww, Rusty1 I kinda hoped (for your own sake) it was yours :)
<drakonis>
oh neat but i'm on lvm
<nDuff>
you can actually make a filesystem that's *both* an ext4 filesystem and a valid btrfs filesystem
<drakonis>
how do i fix this
<nDuff>
and roll it back to ext4-only (until you delete the btrfs snapshot representing its initial state).
<drakonis>
mind you that i don't have a second hard drive
<nDuff>
drakonis, ...even if that were the case here, you can always kexec into an initrd that has tools in it and modify your former root fs from there.
<drakonis>
i have no experience with those situations
<samueldr>
yeah Rusty1, saw those go by on twitter; thought they weren't available still
<samueldr>
neat they seem to be available
<dhess>
Rusty1: just curious, what's the use case for something like that? I would assume a Jetson TX2 has more actual horsepower/flexibility if you need something like an aarch64 build server.
<samueldr>
dhess: AFAIK the initial design is like a "here look what the sopine can do"
<Rusty1>
owner works for a graphics intensive app company, not sure what his plan was - they bought it for him to play with - I suggested NixOs would make deployment "easier"
<samueldr>
dhess: though I'm curious for AArch64 hardware: where can I find reliable numbers for CPU workload
<samueldr>
like, I'm pretty sure anything GPU the nvidia hardware will be better
<dhess>
don't get me wrong: from an engineering perspective it's cool. I just don't see an obvious product fit.
<dhess>
testing sopines maybe :)
<dhess>
samueldr: what do you mean by CPU workload?
<samueldr>
it was initially supposed to be a small run (maybe it's still that small run) so maybe they didn't see much either
<samueldr>
dhess: nothing specific, just raw cpu power
<samueldr>
do you have a resource to compare the different SOCs?
<dhess>
oh I see. Well that's always tricky with benchmarks because it completely depends on what you're running. This is especially true when it comes to embedded applications since an SoC might be custom-built for your application and completely blow away a generic design with a much faster CPU.
<samueldr>
you said "the TX2 has has more actual horsepower", how would you back that without having the hardware?
<dhess>
samueldr: that was just a guess based on the cores in the TX2 CPU and what I know about the pine64 stuff
<samueldr>
lol, I was looking at the chart the other way around
MinceR has joined #nixos
<samueldr>
thanks both of you
<dhess>
also keep in mind that my evaluation was based on the fact that the TX2 has like 8GB of RAM, which is going to be pretty important when you're running builds
<samueldr>
yes!
<samueldr>
I have to figure out a good test
<samueldr>
for e.g. compilation
<samueldr>
not RAM bound
<samueldr>
not I/O bound
mayhewluke has quit [Ping timeout: 246 seconds]
<samueldr>
so something that likely fits within 512MB max including a kernel
<samueldr>
and make "sammark" out of it :)
<dhess>
even if the total of the 7 sopines is approximately the same RAM, it's divvy'ed up into the 7 modules so not available to any one build. Hence why I mentioned the flexibility of the TX2 as another benefit vs. the sopine board
<samueldr>
14 is approximately 8? :)
<dhess>
oh I didn't see that it had 2GB per board
<samueldr>
though yeah, OS overhead and such makes it much closer
<dhess>
still, if you need 3 you're screwed on the sopine :)
<Rusty1>
makes PXE pretty useless as a deployment tool
<samueldr>
the TX2 will run only one OS, the clusterboard will run 7 :)
mayhewluke has joined #nixos
__monty__ has quit [Quit: leaving]
<samueldr>
(though, there's a tipping point in hardware price where getting a boatload of cheaper boards could become more cost effective)
<samueldr>
(and in workload!)
<dhess>
I suppose so, but then "containers"
<drakonis>
containers...
<drakonis>
how's that anyways
<dhess>
what do you mean?
<drakonis>
how's nixos container support
<drakonis>
ye olde containers that everyone else uses
<dhess>
oh I think it works pretty well with Docker containers. I don't use them myself (yet)
<dhess>
I did try the native container stuff and it seemed not fully baked. (This was declarative containers, not imperative)
<dhess>
for example: declare a container. Deploy with NixOps. Container is running, yay!
<dhess>
Remove declarative container. Deploy with NixOps. Container is still running, boo.
<drakonis>
yay, but when you need docker style containers
<dhess>
and no easy way to garbage collect it, as far as I could tell.
<samueldr>
ouch, the TX2 devkit is 2500$USD :/
carlosdagos has joined #nixos
<dhess>
yeah I think the Docker stuff works excellently and Nix is a great way to build them, as well
<dhess>
samueldr: nah, that's the new one that replaces it
<dhess>
TX2 devkit is around $350 I think last I checked
<dhess>
expesive but a pretty good deal for an aarch64 compute server. There's nothing else even close in that price range.
sigmundv has joined #nixos
<samueldr>
ah, the canadian links are wrong I think or the page misleadiing
<samueldr>
470$USD for the TX2
<dhess>
ok I must have been thinking of the TX1. Anyway, still not bad for what you get, relative to other systems with 8GB of RAM
<dhess>
last I looked anyway
<samueldr>
yeah, not debating whether it's good
<samueldr>
it's not in the "having fun" price range :)
<dhess>
however I cannot get NixOS to boot on my TX2, so don't rush out to buy one unless you want to do some bring-up
<dhess>
no it's not :)
<dhess>
The TX1 is also a pretty good compile server and NixOS does work on that board. You can even attach an NVMe disk with a PCIe adapter.
<dhess>
or SSD using the SATA connector
<samueldr>
hmm, still looking, and I might have been looking at the wrong thing for the TX2... 599$USD :/