matthewbauer has quit [(Ping timeout: 260 seconds)]
newhoggy2 has quit [(Ping timeout: 250 seconds)]
Guest40686 has joined #nixos
Guest40686 is now known as christine_
FRidh has joined #nixos
<hlavaty>
FRidh: I dont think that works, e.g. see http://paste.lisp.org/+759J/1 where I replace glibc with one from older nixpkgs but the linked symbols are still too new
<NixOS_GitHub>
[nixops] domenkozar pushed 1 new commit to master: https://git.io/v1Bdh
<NixOS_GitHub>
nixops/master b104025 Domen Kožar: EC2: disable strict host checking for SSH...
Itkovian has joined #nixos
jazzencat has joined #nixos
<rly>
Is the lzop library also packaged in NixOS?
<rly>
As opposed to the binary.
<jazzencat>
Which is the recommended filesystem type for NixOS? Ext4 or btrfs?
<rly>
jazzencat: depends on the usecase.
<hlavaty>
rly: checkout nixpkgs and grep; or nix-env -q iirc
<rly>
jazzencat: I use ZFS and ext4.
<rly>
hlavaty: but the ext4 was more out of laziness.
<hlavaty>
jazzencat: ext4 works
* goibhniu
reckons XFS is worth considering too
<clever>
rly: and at the other end of lazy, one of my build slaves is now running from squashfs
<rly>
clever: what is the benefit of that?
<clever>
rly: the entire buildslave is inside a 40mb squashfs held in ram
<clever>
rly: no state remains on the hdd, and improper shutdowns cant corrupt anything
<rly>
clever: ok, that is nice.
<hlavaty>
clever: do you have your config public somewhere?
<clever>
and i'm doing all of the testing against a raspberry pi, but it should also work against x86
<clever>
[root@router:/tftproot/try2]# time nix-build not-os/release.nix -A rpi_image -o ../9080d9b6 --show-trace -I nixpkgs=/tftproot/try2/nixpkgs/ && ls -ltrh ../9080d9b6/
<jazzencat>
I'm in the learning stage, I want to use this system for my day-to-day laptop, but eventually it might get turned into a personal NAT or file server when the battery gives out or I upgrade or something.
<clever>
with just this one command, i can update all of the rpi firmware, and the not-os image it boots into
<clever>
at the moment, i am fixing DNS, ntp cant resolve 192.168.2.1
<hlavaty>
clever: amazing
<hlavaty>
i wish nixos just worked on rpi2 and rpi3
<clever>
hlavaty: the main issue i had on the rpi3, is that the kernel in nixpkgs doesnt like the start.elf i'm using
<clever>
the rest of nixos should work, but i havent gotten around to doing a proper install yet
<hlavaty>
my issue is that whenever i manage to upgrade, /boot is updated with wrong stuff and i have to copy over working files, which is really annoying
<clever>
ah
<hlavaty>
on rpi3
<clever>
my rpi3 only has bootcode.bin and a swap partition, nothing else
<clever>
it gets everything else over tftp
<jazzencat>
Are there any performance issues with btrfs?
<rly>
jazzencat: I'd be more concerned with losing all your data.
<rly>
jazzencat: many people have other experiences.
<rly>
I think brtfs should never have been created myself.
<rly>
Just a waste of talent and resources.
<clever>
rly, jazzencat: when i was messing with hydra stuff on btrfs (it wanted to create 12,000 files), it took over an hour, and ended with the filesystem going read-only until i rebooted
<clever>
then it took another hour for nix-collect-garbage to get rid of the test files
<rly>
clever: there are probably knobs, though.
<rly>
I consider data loss to be a cause for ignoring a project for at least 5 years.
<clever>
i was on a magnetic hdd at the time, and performance didnt really improve when i switched to zfs
<clever>
i now run on a pair of SSD's in a zfs mirror, so data loss isnt as likely
<clever>
6 Dec 10:19:16 ntpdate[198]: step time server 192.168.2.1 offset 1481019508.187349 sec
<jazzencat>
From a data recovery standpoint which is preferrable?
<clever>
and sucess!, it needed a /etc/services file to "resolve" an ip to an ip, lol
newhoggy2 has joined #nixos
<clever>
i think i stripped it down a bit too much
<goibhniu>
jazzencat: I think it's fair to say that ext4 and xfs are more reliable than btrfs
arjen-jonathan has quit [(Ping timeout: 248 seconds)]
<clever>
rly: another quirk of how i setup this build slave stuff, that may improve speed a lot, the nix store is union'd with a tmpfs, so all buildInputs go to ram first (and swap second)
<clever>
rly: so it will just skip the hdd half the time, if you have enough ram
<kmicu>
jazzencat: btrfs works great for me and even for hardcore Nix users like vcunat or wak-work, and it works great in production on Facebook’s servers :)
<jazzencat>
Okay. My openSUSE install used it as well, but I got tired of arguing with the system when an update broke it.
bin7me has joined #nixos
<kmicu>
NixOS updates are tested, Nix store is immutable so you can always rollback. Pure joy.
jensens has quit [(Ping timeout: 258 seconds)]
<rly>
I tried a NixOS update recently and it didn't work. So, I rolled back.
<rly>
A great lack of excitement followed.
<jazzencat>
How big does my swap need to be?
<rly>
jazzencat: depends on the type of user you are.
arjen-jonathan has joined #nixos
<rly>
jazzencat: The size of your RAM is a value some people use.
<clever>
if your as insane as i am (over 800 tabs in chrome right now), 32gig of swap and 16gig of ram
newhoggy2 has quit [(Remote host closed the connection)]
<rly>
clever: 800? How do you manage that?
<clever>
rly: poorly :P
<rly>
clever: I think 50 is already a lot.
<rly>
clever: at least, I clean up when I hit 50.
<Havvy>
I've never had more than 90 tabs in Firefox.
newhoggy2 has joined #nixos
<Havvy>
I used to always hover around 30.
<rly>
I do have multiple windows open ;)
<clever>
rly: it takes chromium an hour to resume
<rly>
But I shouldn't do that.
<Havvy>
Okay, I actually have 38 open right now.
<rly>
I eliminated the "too many terminal windows"-problem via tmux.
<Havvy>
So it's not just "used too"...
<ronny>
clever: whats the median agne/untouched time of your tabs?
<clever>
ronny: ive found some tabs that date back to 2 or 3 months ago before
<spacekitteh>
todo: add "device description" as config option :B
newhoggy2 has joined #nixos
iyzsong has joined #nixos
wangoe has joined #nixos
<michalrus>
So I’ve got this weird problem. When trying to do `systemctl suspend` from the only physically logged-in regular user (or reboot, or poweroff, or hibernate, or halt…), I’m getting: Interactive authentication required.
<michalrus>
Could someone try `systemctl suspend` on the desktop?
<michalrus>
Because of that, reboot/poweroff buttons in GNOME don’t work.
<niksnut>
works for me (in KDE 4)
<michalrus>
niksnut: on 16.09?
newhoggy2 has quit [(Ping timeout: 248 seconds)]
<niksnut>
yes
* michalrus
sighs
<michalrus>
A pity there’s not anything helpful in journalctl, either.
amir has quit [(Ping timeout: 258 seconds)]
Seichi has quit [(Remote host closed the connection)]
newhoggy2 has joined #nixos
Seichi has joined #nixos
amir has joined #nixos
* avn
always used `sudo reboot` etc
<niksnut>
maybe loginctl shows something useful?
<michalrus>
avn: `reboot` (really `systemctl reboot`) should work when there’s only one user logged in, in a physical seat. :)
signal4 has joined #nixos
<michalrus>
niksnut: no… according to its `list-user` and `-session`, there’s only one session and one user.
<michalrus>
And one seat.
<signal4>
o/
newhoggy2 has quit [(Remote host closed the connection)]
newhoggy2 has quit [(Remote host closed the connection)]
[0x4A6F] has joined #nixos
<clever>
domenkozar: resize2fs refuses to resize an FS until fsck has flagged it as clean
<domenkozar>
can we force that without wasting time on fsck?
<clever>
domenkozar: it wont risk the chance that some corruption went unnoticed, and horribly screws things up
echo-area has quit [(Remote host closed the connection)]
<clever>
domenkozar: there is a force option in resize2fs, -f Forces resize2fs to proceed with the filesystem resize operation, overriding some safety checks which resize2fs normally enforces.
<clever>
but you would have to modify whatever was calling resize2fs
<signal4>
clever: lib/libsystemd.so and aliases match
<clever>
signal4: now run strings on that so, and pipe it thru the same grep (omiting the lib path), that should tell us which thing its referencing
Ivanych has joined #nixos
<domenkozar>
clever: hmm :)
<domenkozar>
I'll time it first to see how long it even takes.
<clever>
domenkozar: tune2fs may also work, -C i think
<clever>
domenkozar: i think the kernel will increment the mount-count every time you mount it r/w, and fsck resets it to zero upon finishing
<clever>
domenkozar: so once the 68-98 block has ran, mount-count should be +1, so the -C 0 on 65 wont help any (it will be 1, not 0, and resize2fs will want a check)
redmq has joined #nixos
<Profpatsch>
FRidh: Sure, pulseaudio streaming works wonderfully.
<Profpatsch>
Any bugs?
<Profpatsch>
musicmatze: I don’t quite understand; what kind of load balancing?
<clever>
signal4: ok, so you have a single library in the lib output, that depends on $out
<musicmatze>
if you have to write a generic load balancing algorithm, why do you want it to be pure?
<sphalerite>
Why does nix-shell -p 'haskellPackage.ghcWithPackages (pkgs: ...)' not use the same ghc that I already have through nix-env -iA nixos.ghc ? (nix-shell gets dg7ak1hvlj66vgn4fwvddnnr4pfncd04, while nix-env gets 2gibm58q01bigb6xxbnyw9fv871qaxc9, both 8.0.1)
<clever>
signal4: and a bunch of libraries in $out that depend on $lib, there is the cycle
<sphalerite>
clever: while nix-shell downloaded dg7ak1hvlj66vgn4fwvddnnr4pfncd04
<clever>
sphalerite: look at the contents of the ghc script itself
<clever>
signal4: $lib doesnt make any reference to $out on my systemd
redmq has quit [(Ping timeout: 260 seconds)]
<clever>
signal4: so now you need to figure out why /nix/store/yw47gp1zd8y143d3wvsank97gbq55lg8-systemd-231/lib/systemd/catalog/ is present in the $lib output
<spacekitteh>
how can i change my default compiler to be clang?
<ronny>
spacekitteh: wouldnt you typically get noncached packages anyway if you would reconfigure the compiler flags? (also sorry for the delay)
<spacekitteh>
ronny: sure
<ronny>
hmm
<Profpatsch>
musicmatze: Well, the algorithm can be pure trivially, you just have to determine the inputs and put them in at the top. :)
<LnL>
clever: ...
<clever>
LnL: ?
<Profpatsch>
So I’d guess a load balancer can be a loop which gets the current system state, inputs it into a pure algorithm, which gives back which jobs should run on what agent.
<Profpatsch>
And then the output is used again for the next round.
<Profpatsch>
Err, the new system state.
<Profpatsch>
But I’d look at the Controller-Observer pattern from Organic Computing for that specific problem.
TheAppleMan has quit [(Ping timeout: 246 seconds)]
<Profpatsch>
It’s probably a good generalization of load balancing and can be specialized.
<Profpatsch>
Whether it’s done purely is then a choice of implementation.
<signal4>
clever: huh indeed the default systemd-231-lib doesn't have any references to systemd-231
<clever>
signal4: only thing left now, is to figure out which of your tweaks to the build is causing it, try just turning them off one by one?
<LnL>
clever: your message about your ram usage
<clever>
LnL: ah
<LnL>
my znc was a little slow with replaying the messages
TheAppleMan has joined #nixos
TheAppleMan has quit [(Changing host)]
TheAppleMan has joined #nixos
<clever>
LnL: ive done worse, one of my servers has 64gig of swap, which fills up about 2 or 3 times a year
<clever>
LnL: that has no effect on performance, until it runs out, the cause is a memory leak, so the data in swap is never referenced again
<musicmatze>
Profpatsch: shure it _can_ be - but why _should_ it be? Are there advantages from writinng the algorithm in a pure way?
<musicmatze>
... are there arguments for this? or is it just "because it feels better"-like?
christine_ has quit [(Ping timeout: 258 seconds)]
<clever>
LnL: and at the other extreme of memory usage, i now have a nix build slave, running on a 40mb ramdisk
matthewbauer has joined #nixos
<musicmatze>
Profpatsch: ah, sorry, did not see your reply above. Thanks, I will have a look at controller-observer from OC
<LnL>
clever: that's annoying, I also thought I had a memory leak. but it was just journald eating all the memory on a small instance :p
gilligan_ has joined #nixos
<Profpatsch>
musicmatze: Err, not inherently, no.
<gilligan_>
o/
<Profpatsch>
Depends on how your language handles that.
<Profpatsch>
Like I said, implementation decision.
<musicmatze>
k
<gilligan_>
hi Profpatsch
<clever>
LnL: its an extremely slow leak, so its hard to even notice until you have months of uptime
<musicmatze>
thx
<Profpatsch>
I’d always go for most purity if possible.
<Profpatsch>
Just becauese it’s easier to reason about, since everything’s more local.
<musicmatze>
yeah, but as you said: design decision...
<Profpatsch>
That’s the only way I got it running, since yes, you *can* start pa as root, but then half of the modules simply stop working.
<Profpatsch>
For example avahi zeroconf.
<FRidh>
yes, I don't even want to run it as root
<FRidh>
but that's good, I'll have a look at it again later then
viric___ has quit [(Ping timeout: 260 seconds)]
<Profpatsch>
If you find a better way to get the user up, for example with a socket-driven start, that would be nice.
<Profpatsch>
But I haven’t found it.
aluminous has joined #nixos
goibhniu has quit [(Ping timeout: 246 seconds)]
mudri has quit [(Ping timeout: 256 seconds)]
lula_ has joined #nixos
<lula_>
I have declard a derivation with stdenv.mkDerivation, but when I try to add it to imports = [] in my nixos configuration I get an infinite recurstion while evaluating the module argument `stdenv'. any ideas what I might be doing wrong?
<clever>
lula_: derivations cant go into imports, only nixos modules
<lula_>
I see, so how do I reference the derivation from my nixos config?
<gchristensen>
good morning, #nixos
<clever>
lula_: what do you want to do with that derivation?
<lula_>
install some scripts
<lula_>
also: learn ;)
<clever>
lula_: then it has to go into the systemPackages list, environment.systemPackages = [ (pkgs.callPackage ./foo.nix {}) ];
uxv has joined #nixos
<NixOS_GitHub>
[nixpkgs] domenkozar pushed 1 new commit to release-16.09: https://git.io/v1RTo
<NixOS_GitHub>
nixpkgs/release-16.09 331a7bc Tom Hunger: digital-ocean: Init at 1.10.1...
<clever>
if the clock advances by more then 10 hours, nix thinks the build has gone 10 hours without any output, and kills it
<clever>
in my case, the clock advanced by ~46 years
<gchristensen>
fun, clever
<domenkozar>
:D
<clever>
gchristensen: ive had an even worse bug in my irc server before, if the clock went backwards even 1 second, all of the math overflowed, and every single user ping timeout'd
<domenkozar>
that's what you get with singletons
<domenkozar>
clever: wait until we have cross planetary times
<domenkozar>
luckily we'll all die by then :-P
<clever>
domenkozar: for measuring elapsed time, the uptime is usualy better
<clever>
uptime cant go backwards, and ntp cant make uptime skip ahead
<domenkozar>
yeah you need some proper monotonic clock
<ronny>
hmm
<ronny>
always use monotonic clocks ^^
<ronny>
"dancing clocks" are going to stumble over you bacwards ^^
<clever>
ronny: clocks, and overly pure envionments also caused a couple build failures on my end
matthewbauer has joined #nixos
<clever>
2 or 3 packages refused to compile, because it was 1970
<ronny>
clever: i proposed pickign a null time thats zip compatible ^^
<ronny>
python wheels are unhappy zip files for example ^^
<clever>
ronny: at least one of the failed builds, where caused by configure.ac being newer then configure, so it wanted autoconf, which wasnt in buildInputs
<ronny>
hmm
<ronny>
autotools absolutely needs to die ^^
<clever>
so technicaly, the build dependencies depends on the time
<clever>
if the time is prior to the source tar's creation, it needs more tools
<clever>
but that will only happen on systems that are not configured properly, until we get time travel :P
<ronny>
hmm, i fail to see why build tools have to predate files, just use the dates/times of the tarballs?
<clever>
ronny: tools like make compare the mod-times of everything, and if the new .o's are from 1970, and the source is from 2016
<clever>
then the source is always newer, and it can never finish the build
matthewbauer has quit [(Ping timeout: 265 seconds)]
<gchristensen>
that is very annoying :)
<clever>
gchristensen: whats worse, is that ntpdate was unable to resolve 192.168.2.1, because /etc/services was missing
<clever>
the entire dns subsystem of linux lost the ability to identify what a bare IP address was
<gchristensen>
uhhh....
<clever>
i made my build slave a little too minimal
<gchristensen>
it seems silly to me that it can't use bare IPs without /etc/services
<clever>
yeah
p_l has quit [()]
p_l has joined #nixos
<clever>
gchristensen: but i now have a build slave running on a 40mb ramdisk
<clever>
and it is happily building nix packages in a sandbox
MercurialAlchemi has quit [(Ping timeout: 260 seconds)]
<gchristensen>
oh wow
<gchristensen>
that is really amazing
MercurialAlchemi has joined #nixos
<clever>
gchristensen: 100% of its writeable storage, is tmpfs's, with swap for the overflow, so there is nothing left to corrupt upon improper shutdown
<clever>
and all state is wiped clean at reboot
chpatrick has joined #nixos
<chpatrick>
hi
<chpatrick>
does building in a sandbox work currently?
<clever>
gchristensen: and with the hydra option you helped find, i can dynamicaly add that slave to hydra without a rebuild-switch
<clever>
gchristensen: so i just need a bit of software to detect the slaves coming and going, and add/remove them from a 2nd machines file
<chpatrick>
alternatively do you have any advice for using nix to set up an environment for a classical incremental build?
<chpatrick>
ie I want to get an environment from nix where my non-nix incremental build tools just work
<clever>
chpatrick: for a manual or automated build?
<chpatrick>
clever: manual
<chpatrick>
nix works fine when we want to make a big build from scratch
<clever>
chpatrick: nix-shell should be fine for that
<chpatrick>
but when developing I want to rebuild quickly
<clever>
chpatrick: just run nix-shell on your derivation, and use configure+make as normal
<chpatrick>
ok I'll try it
<gchristensen>
clever: maybe some sort of broadcasting on startup
<clever>
gchristensen: i was thinking of udp packets sent at regular intervals, advertising the builder username, and the supported arches
<gchristensen>
oh sure
<clever>
gchristensen: so if it goes offline without warning, the lack of a msg causes an update
<gchristensen>
yup makes sense
<chpatrick>
clever: and how about the sandbox?
<chpatrick>
with nix-shell it seems to find the CUDA from my host system for example
<clever>
chpatrick: the sandbox only works for automated builds
<chpatrick>
clever: is it documented anywhere?
<clever>
chpatrick: but there is nix-shell --pure, which will clean $PATH up more
<gchristensen>
but CUDA might also just be playing dirty tricks that nix-shell won't help fight
<gchristensen>
like if they hard-coded some paths, and you use it in nix-shell, it can still access those paths
<musicmatze>
ronny: sorry, didn't see your msg. It is independent of a checkout, yes, but it is highly integrated with git, as this is our usecase.
<Dezgeg>
cuda is probably tricky as it might need to dynamically link to some libraries that need to match the kernel driver version
<chpatrick>
Dezgeg: yes we have a dirty hack to fix that problem ;)
<chpatrick>
gchristensen: is there any difference between calling cmake myself in nix-shell vs it being called in a derivation?
<chpatrick>
because in stdenv it finds the cuda path correctly and if I call it myself it doesn't
<gchristensen>
clever: ^
tomberek has joined #nixos
mudri has joined #nixos
<chpatrick>
gchristensen: nm we passed it manually
christine_ has joined #nixos
<NixOS_GitHub>
[nixpkgs] joachifm pushed 2 new commits to release-16.09: https://git.io/v1ROu
<spacekitteh>
hmm, crap. i don't think i can work around building polly twice :(
<musicmatze>
our tool is a subcommand to git as "commit" is
<mudri>
I recently fiddled with my partitions, and now NixOS's grub puts me in rescue mode. There, I have to change prefix from “(hd0,msdos10)/boot/grub2” to “(hd0,msdos10)/boot/grub” and run ⟨insmod normal⟩ then ⟨normal⟩. Is there any way to fix this, so I don't enter rescue mode?
<gchristensen>
musicmatze: oh I see. nevermind :)
<musicmatze>
np
<aloiscochard>
did anyone tried to use/install xmind on nix?
__20h___ has quit [(Quit: WeeChat 1.5)]
__20h__ has joined #nixos
jazzencat has quit [(Ping timeout: 258 seconds)]
<chpatrick>
clever: how can I use a regular g++ invocation in nix-shell?
<signal4>
clever: it's not done yet but i'm pretty sure the cause of the problem was me specifying raw CFLAGS and CXXFLAGS. using just NIX_CFLAGS_COMPILE seems to work just fine
<chpatrick>
given that there's no LIBRARY_PATH
<clever>
chpatrick: what arguments are you calling nix-shell with?
<clever>
chpatrick: that one should have g++ available, the stdenv will alter $NIX_LDFLAGS and similar variables, to make it find everything in the buildInputs at compile/link time
<clever>
chpatrick: have a look at the output of "env | grep NIX_"
<chpatrick>
clever: if I have a custom g++ invocation do i need to put in NIX_LDFLAGS in myself?
<clever>
chpatrick: the g++ in $PATH is a shell script that obeys the special $NIX_ ones
<clever>
chpatrick: because packages keep ignoring the old $LDFLAGS variables
<NixOS_GitHub>
[nixpkgs] bzizou opened pull request #20944: pythonPackages.escapism: init at 0.0.1 (master...escapism) https://git.io/v1R34
<signal4>
is NIX_LDFLAGS in the same class of variables as NIX_CFLAGS_COMPILE? does it make sense to override it in stdenv?
<clever>
signal4: i dont think there are any ld flags you would want to tune
roconnor has joined #nixos
<vdemeester>
hum.. is there a trick to compile a static binary on a package in nixpkgs ? O:)
<NixOS_GitHub>
[hydra] edolstra pushed 1 new commit to master: https://git.io/v1R3H
<NixOS_GitHub>
hydra/master d0ad3fd Eelco Dolstra: Revert "Remove the global channel"...
<signal4>
i usually add "--gc-sections" to ldd flags
<signal4>
ld*
<musicmatze>
Profpatsch: /me just started to listen to the mix from Couchsofa from GPN16 - you remember?
<musicmatze>
Looking forward to a super-productive day now :-)*
<gchristensen>
fpletz: which screen do you have on the x250?
<NixOS_GitHub>
[nixpkgs] edwtjo pushed 2 new commits to master: https://git.io/v1RsT
<NixOS_GitHub>
nixpkgs/master 1915170 Ignat Loskutov: idea.datagrip: init at 2016.3
<NixOS_GitHub>
nixpkgs/master 6344039 Edward Tjörnhammar: Merge pull request #20933 from loskutov/datagrip...
<vdemeester>
clever: hum, it's on `tini` (which use cmake… and I am soo not used to that :D)
<clever>
vdemeester: there are sometimes options in the cmakelists.txt file to enable static building
<vdemeester>
yep it seems there is
<vdemeester>
got `ld: cannot find -lc` now :(
<clever>
vdemeester: glibc.static needs to be in the buildInputs
<vdemeester>
oooh glibc.static :D
<vdemeester>
damn I was close :P
matthewbauer has joined #nixos
<vdemeester>
yay looks like it's working
<vdemeester>
thx clever :D
<gchristensen>
fpletz: also: how the heck did you manage to actually bear their website long enough to buy one? my browser redraws the screen about 5 times while loading
<gchristensen>
* each page
Rizy has joined #nixos
matthewbauer has quit [(Ping timeout: 256 seconds)]
<fpletz>
gchristensen: unfortunately the IBM days are over and the quality of thinkpads is steadily decreasing :/ I've been using thinkpads for 10 years now but am thinking about switching to another vendor for my next one… though I'm not sure where to go. I hope the x250 lasts at least one year (have problems with the keyboard and the integrity of the case :/)
aluminous has quit [(Quit: WeeChat 1.6)]
<fpletz>
gchristensen: I don't have any experience with macbooks, I can just say that a collegue ordered the other screen and the resolution is just not up to current standards
<lula_>
clever: I would rather not add it to the systemPackages directly but have it 'imported' so I can define it once and then add it depending on the machine configuration with systemPackages = [ myDerivation ]. What would be required for that?
<fpletz>
aloiscochard: as said, not sure yet, but I would try dell or hp (even though I do not really like both for their other professional hardware products), but at least they've been in the laptop for quite some time
<clever>
lula_: create a nixos module that puts it into systemPackages?
<NixOS_GitHub>
[nixpkgs] bzizou opened pull request #20945: pythonPackages.pamela: init at 0.3.0 (master...pamela) https://git.io/v1RZa
<chpatrick>
can I callPackage something from nixpkgs from outside?
<clever>
lula_: the syntax is identical to configuration.nix, thats already a nixos module
<fpletz>
aloiscochard: I've also thought about switching to chromebooks because they run coreboot by default, but the general build quality is not that impressive
* gchristensen
may end up buying another (older) macbook pro this time around
<fpletz>
just a week ago I had to temporarily use my old x230… man, what a great keyboard in comparison… I've violated that thing so much, but the case is still intact
<domenkozar>
obadz: here?
signal4 has left #nixos ["Leaving"]
<FRidh>
I'm using multiple outputs, but only custom ones, so no $out or $bin. Anyone knows how to deal with Error: _assignFirst found no valid variant!
<chpatrick>
warning: dumping very large path (> 256 MiB); this may run out of memory
<chpatrick>
error: while evaluating the attribute ‘src’ of the derivation ‘serkezet-0.1.0.0’ at /home/chpatrick/benaco/benaco/experiments/serkezet/nix-channel/nixpkgs/pkgs/development/haskell-modules/generic-builder.nix:145:3:
<chpatrick>
Nix daemon out of memory
<chpatrick>
:(
ryantrinkle has quit [(Ping timeout: 260 seconds)]
gilligan_ has quit [(Ping timeout: 245 seconds)]
systemfault has joined #nixos
<niksnut>
chpatrick: you probably want do do something like: src = if lib.inNixShell then null else ./.;
ebzzry has quit [(Ping timeout: 250 seconds)]
<niksnut>
*to do
jazzencat has joined #nixos
lula_ has joined #nixos
Rizy has quit [(Quit: Rizy)]
<lula_>
my current nixos config causes an issue: value is a function while a set was expected, at /nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/lib/modules.nix:128:29. Sadly, there is not much info on what is causing this. is there a way to get more useful output to debug the issue?
<Profpatsch>
We’ve got a Haskell Workshop in Augsburg, Bavaria, on 17.12.
<lula_>
regnat_: it is rased in /nix/var/nix/profiles/per-user/root/channels/nixos/nixpkgs/nixos/lib/eval-config.nix, but that is not much help sadly
<regnat_>
No indeed. But sometimes the stackstrace reveals the problematic bits in the configuration
wangoe has left #nixos []
wangoe has joined #nixos
suvash_away has joined #nixos
matthewbauer has joined #nixos
<kmicu>
Hi lula_ could you pastebin the stack trace?
<monad_cat>
Hey, I'm trying to build go-ipfs and get a message telling me that:
<monad_cat>
Package ‘go1.7-ipfs-i20160112--7070b4d’ in ‘/nix/store/qikh39zk0nb60qgzi2914f044mwmdc52-nixos-16.09.1202.2a97055/nixos/pkgs/applications/networking/ipfs/default.nix:18’ is marked as broken, refusing to evaluate.
<monad_cat>
anyone knows what's up?
<kmicu>
Hi monad_cat what’s your Nixpkgs’ version? You can check with ‘nix-instantiate --eval '<nixpkgs>' -A lib.nixpkgsVersion’?
matthewbauer has quit [(Ping timeout: 268 seconds)]
amarsman has quit [(Ping timeout: 260 seconds)]
jazzencat has quit [(Remote host closed the connection)]
<clever>
lula_: can you gist your configuration.nix and the files listed under imports?
cpennington has quit [(Ping timeout: 260 seconds)]
angerman has quit [(Quit: Gone)]
<aloiscochard>
fpletz: interesting, thank you
<gchristensen>
fpletz: I would be interested in that smartcard if you still have it
<kmicu>
monad_cat: ipfs is marked as broken on 16.09 revision 2a97055 (it’s not on master).
the_ktosiek is now known as ktosiek
<monad_cat>
is there a way to tell nixos to get a package from a different version of nixpkgs? I've just cloned nixpkgs from master
<kmicu>
‘nix* -I nixpkgs=/path/to/clone …’
<rly>
monad_cat: you can also set the thing kmicu refers to in your configuration.nix.
<kmicu>
(nix-env requires -f flag)
<clever>
monad_cat: you can just make a packageOverride with foo = pkgs.callPackage /home/clever/nixpkgs/foo/bar/default.nix {}; to get just 1 package, or kmicu's command to get all
<rly>
monad_cat: that's what I use and yes, you can mix unstable dev stuff with stable.
<rly>
monad_cat: it's beatiful.
<rly>
monad_cat: it's beautiful.
roundhouse has joined #nixos
<monad_cat>
nice, thanks for your help
<monad_cat>
clever: should I just put the pkgs.callPachage /home... thing in my systemPackages list?
<kmicu>
lula_: as always thanks to amazing Nix’s stack traces it’s impossible to know anything. Do you know from what version did you update? Did you jump e.g. from 15.* to 16.*?
<clever>
monad_cat: that can also work, as long as you wrap it with ( and )
<roundhouse>
hi, I have a problem when reading files with haskell on nixos. I get " hGetContents: invalid argument (invalid byte sequence)" errors. The friendly folks from #nixos pointed me to the LC_ALL, LC_LOCALE, etc. variables, which are empty in my shell. How can I set them in Nixos?
<roundhouse>
*#haskell
<roundhouse>
I'm on Nixos unstable btw.
iyzsong has quit [(Quit: night.)]
<kmicu>
Hi roundhouse you can always fallback to generic ‘environment.variables.MY_VAR = "sth";’ in configuration.nix.
cpennington has joined #nixos
<NixOS_GitHub>
[nixops] domenkozar closed pull request #446: Virtualbox: boot.loader.grub.timeout was renamed to boot.loader.timeout (master...patch-1) https://git.io/vovpN
systemfault has quit [(Quit: Bye!)]
<kmicu>
roundhouse: before you do that could you check the output of ‘locale’ command?
<NixOS_GitHub>
[nixpkgs] joachifm pushed 3 new commits to master: https://git.io/v1R4i
<NixOS_GitHub>
nixpkgs/master 1a427cf Joachim Fasting: privoxy: enable pie
<NixOS_GitHub>
nixpkgs/master 148ee2c Joachim Fasting: privoxy: homepage now reachable via https
<NixOS_GitHub>
nixpkgs/master db0bd30 Joachim Fasting: ion3: specify meta.{platforms,license}
<kmicu>
(Your question reminds me common Hakyll problem.)
<NixOS_GitHub>
[nixops] domenkozar pushed 1 new commit to master: https://git.io/v1R4M
<NixOS_GitHub>
nixops/master 5e5db68 Domen Kožar: Fixes #556
<roundhouse>
LANG=en_US.UTF-8
<roundhouse>
LC_CTYPE="en_US.UTF-8"
<roundhouse>
LC_NUMERIC="en_US.UTF-8"
<roundhouse>
LC_TIME="en_US.UTF-8"
<roundhouse>
LC_COLLATE="en_US.UTF-8"
<roundhouse>
LC_MONETARY="en_US.UTF-8"
<roundhouse>
LC_MESSAGES="en_US.UTF-8"
<roundhouse>
LC_PAPER="en_US.UTF-8"
<roundhouse>
LC_NAME="en_US.UTF-8"
<roundhouse>
LC_ADDRESS="en_US.UTF-8"
<roundhouse>
LC_TELEPHONE="en_US.UTF-8"
<roundhouse>
LC_MEASUREMENT="en_US.UTF-8"
<roundhouse>
sry, bad formatting, that is the output of locale
<roundhouse>
it seems as if LC_ALL should also be set
<clever>
error: attempt to call something which is not a function but a set, at /nix/store/442fh8j84z1ixqa1sj71kdl3hlbjy3dx-git-export/pkgs/top-level/default.nix:52:9
cpennington has quit [(Ping timeout: 268 seconds)]
<clever>
i think somebody broke config.platform on unstable, but it appears to already be fixed on master
MercurialAlchemi has quit [(Ping timeout: 250 seconds)]
<monad_cat>
one more thing - can I set the custom nixpkgs location in my configuration.nix?
<clever>
that makes no sense, unstable updated 4 hours ago, but the change was commited in march
<clever>
monad_cat: nixpkgs is what loads configuration.nix, so whatever you set within there will only take effect after the build is done (when its too late to matter, but it will effect the NEXT rebuild)
<monad_cat>
sure, that would still work for me though, this means that I just have to remember about the cli param once
<clever>
monad_cat: i think its nix.nixPath, check the man page for configuration.nix
<Profpatsch>
roundhouse: LOL, I just wanted to ask about a very similar problem.
<Profpatsch>
roundhouse: Even if I set LANG & LCALL to en_US.UTF-8 here, it still errors out with [ERROR] jbovlaste.xml: hGetContents: invalid argument (invalid byte sequence)
<peti>
roundhouse: Do you have $LOCALE_ARCHIVE or $LOCPATH set, too?
<Profpatsch>
I get errors like these: locale: Cannot set LC_CTYPE to default locale: No such file or directory
<roundhouse>
so here is the strange thing. If I use locale, then the output is as above. However "echo $LC_CTYPE" (e.g.) gives me an empty output in my shell
<roundhouse>
Profpatsch: Btw, a work around is to either set the locale of the handle explicitly or to read as ByteString and then convert to Text. But that doesn't really solve the actual problem
systemfault has joined #nixos
matthewbauer has joined #nixos
<Profpatsch>
Oh, I just saw that haskell mkDerivation sets those variables
<Profpatsch>
At least in ENV
<avn>
joachifm: ohh, ty ;)
<marvin3k>
hi. question concerning multi-user installation of nix: does the nix-daemon have a config file somewhere? i would like to influence NIX_PATH to add ssh-config-file for fetchgitPrivate to work
<roundhouse>
I use stack with nix btw (but I guess this belongs to #haskell)
cpennington has quit [(Remote host closed the connection)]
<Profpatsch>
roundhouse: Are you getting the error in a derivation build, or when using it interactively?
<marvin3k>
i tried, for example, to set a "Environment=NIX_PATH..." in the systemd-service file. to no avail
systemfault has quit [(Client Quit)]
hlavaty has quit [(Ping timeout: 244 seconds)]
<roundhouse>
Profpatsch: Interactively in the sense of ghci? Then no.
<clever>
marvin3k: nix-daemon comes into play after $NIX_PATH has done its job
cpennington has joined #nixos
matthewbauer has quit [(Ping timeout: 258 seconds)]
<clever>
marvin3k: when you run nix-build, it will internaly call nix-instantiate as your user (so your NIX_PATH matters), and that will talk with the daemon to write files to /nix/store
<clever>
marvin3k: then nix-build will run nix-store -r on the drv files, to build everything
<Profpatsch>
roundhouse: Interactively as inside a nix-shell
<Profpatsch>
peti: Always? Or just with haskel packages?
<clever>
marvin3k: so only you can revoke the keys when your done, or just not put the agent in the include path when you dont need it
<Profpatsch>
peti: Right.
<Profpatsch>
But not when I’m in a build environment
<peti>
Profpatsch: Without it, your locale settings have no effect.
<Profpatsch>
I can do inherit (myhaskellpackage) LOCALE_ARCHIVE LANG;
<Profpatsch>
Or I can change the handler to read as utf-8
<roundhouse>
Profpatsch: I guess that happens automatically with stack + nix, but I don't have an extra nix-shell for haskell
<Profpatsch>
I think the second one is nicer.
<avn>
joachifm: fix which I mentioned in issue works for me, when full build complete, I make PR ;)
<Profpatsch>
roundhouse: Depends on how you use that.
<marvin3k>
clever: hm, ok. will try this out... thanks again (you helped me before ;-)
<Profpatsch>
roundhouse: I personally don’t use stack with nix, only cabal and cabal new-build
roconnor has joined #nixos
<domenkozar>
are .hi files needed by haskell at runtime?
<roundhouse>
Profpatsch: I have nix: enable: true, packages = [...] in my gloabel config yaml
<Profpatsch>
And the error appears in?
matthewbauer has quit [(Ping timeout: 256 seconds)]
<joachifm>
avn: cool
<marvin3k>
clever: so the actual build (make, gcc...) is still performed by the user, and just the resulting files+store location are passed to the daemon?
<clever>
marvin3k: nix-daemon also manages the builds
<clever>
marvin3k: nix-instantiate will generate the contents for .drv files, which are stored at a path based on the hash of their contents
<clever>
marvin3k: the rest of the derivations, nix-daemon will fork of a child to do the build, which runs as a member of the nixbld group
<clever>
marvin3k: and the $out path is based on a hash of the script nix-daemon runs to produce it, so if a malicious user builds a tainted copy of sudo, it will be at a different path, and nobody will use it
<clever>
and nix-daemon enforces the rules, so the bad sudo winds up at a different path
<marvin3k>
yes
<marvin3k>
so (theoretically) nix-instantiate could implement a git+ssh-client (in nix langauge), perform the clone as my local user, create the drv and pass it to the daemon. but it cannot call tools on its own
<clever>
marvin3k: but then the output path of that will depend on the implementation, and everything after that will rebuild without warning
<Profpatsch>
huh, I could have sworn it had to work now
<Profpatsch>
h <- IO.openFile f IO.ReadMode
<Profpatsch>
IO.hSetEncoding h IO.utf8
<Profpatsch>
TIO.hGetContents h
<clever>
marvin3k: fixed-output derivations are special, the path depends on the contents, rather then the directions
derjohn_mobi has quit [(Ping timeout: 250 seconds)]
<marvin3k>
and putting the ssh-key into the store makes it readable to the whole world (my computer, but still). setting GIT_SSH for the nix-daemon process would be another, less invasive hack?
FrozenCow has joined #nixos
matthewbauer has joined #nixos
<marvin3k>
clever: yes, ok, the hash would change...
<roundhouse>
Profpatsch: does this not work for you? It works fine for me
<clever>
marvin3k: nix will use the env variables from the nix-store command, not the nix-daemon command
<Profpatsch>
peti: Should explicitely setting handle encoding like above fix the error independent of locale set?
jensens has joined #nixos
<clever>
marvin3k: nix will also block out nearly all env variables, only a fixed-output derivation can get some, but you need to use an agent, or the key is still readable to every builder
<marvin3k>
clever: but the nix-daemon is the one calling git, which in turn calls ssh? (in the fetchgitPrivate case)
<Profpatsch>
ooooohhh
<clever>
marvin3k: nix-store forwards the env variables to nix-daemon
<Profpatsch>
It comes up for a different file now.
<Profpatsch>
Hakyll tries to read template files and those are UTF-8 as well.
<Profpatsch>
So correctly setting the locale is the right way to go here.
<roundhouse>
Profpatsch: Do you have a way that works globally?
fritschy has quit [(Quit: Leaving.)]
<Profpatsch>
afaik there’s only two ways: 1 Each library has to explicitely define what encoding inputs need to be
<Profpatsch>
2 The locale is set to UTF-8 and you have to hope everything is UTF-8
<Profpatsch>
3 You don’t set the locale explicitely and pray
<marvin3k>
clever: thanks for the words! will go and see what i can get
<peti>
Profpatsch: I think it would.
<NixOS_GitHub>
[nixpkgs] FRidh pushed 2 new commits to master: https://git.io/v1Raj
<NixOS_GitHub>
nixpkgs/master 736169d Frederik Rietdijk: buildPythonPackage: yet another fix, fixes #20943...
<niksnut>
probably not much that can be done about it though...
<michalrus>
A pity, because logind runs with root privileges. It *has* the information about all processes. Why would it matter that a regular user that wants to reboot doesn’t have it?
civodul has quit [(Quit: ERC (IRC client for Emacs 25.1.1))]
lsix has quit [(Quit: Konversation terminated!)]
proteusguy has joined #nixos
roundhouse has quit [(Ping timeout: 244 seconds)]
agjacome has joined #nixos
agjacome has quit [(Client Quit)]
FRidh has quit [(Remote host closed the connection)]
Seichi has quit [(Remote host closed the connection)]
eacameron has quit [(Quit: Leaving...)]
ckauhaus has quit [(Quit: Leaving.)]
<marvin3k>
@clever: i have ssh-agent running, the socat command you linked is running, the keys are known to ssh-agent and in the git-server, I see the correct output of "/usr/bin/ssh-add -l > /tmp/knownKeys" in a modified fetchgit-ssh script created by fetchgitPrivate. still: "host key verification failed". any idea where things went wrong, where to test?
eacameron has joined #nixos
zagy has joined #nixos
<michalrus>
Hey, is anyone here on GNOME, perhaps?
<stepcut>
I wonder if I can get nixpkgs installed on my phone and use it as a build slave for my RPi. My phone has twice the CPU power and 3x the RAM... seems.. possible
<gchristensen>
Profpatsch: link to your go-git-thinger?
zagy has quit [(Quit: Leaving.)]
zagy has joined #nixos
jgertm_ has quit [(Ping timeout: 250 seconds)]
redmq has joined #nixos
griff_ has joined #nixos
stepcut has quit [(Remote host closed the connection)]
RchrdB has quit [(Quit: Leaving)]
stepcut has joined #nixos
bgamari has quit [(Ping timeout: 246 seconds)]
viric has joined #nixos
lula_ has quit [(Ping timeout: 260 seconds)]
RchrdB has joined #nixos
stepcut has quit [(Ping timeout: 244 seconds)]
glines has quit [(Ping timeout: 258 seconds)]
viric___ has quit [(Ping timeout: 260 seconds)]
<ronny>
hmm
ryanartecona has quit [(Quit: ryanartecona)]
chpatrick has left #nixos []
griff_ has quit [(Quit: griff_)]
<ronny>
michalrus: yes
<michalrus>
ronny: could you please check whether you can set the keyboard repeat rate in System Settings → Keyboard?
<NixOS_GitHub>
[nixpkgs] grahamc pushed 1 new commit to master: https://git.io/v1RhL
<NixOS_GitHub>
nixpkgs/master 14e53ca Graham Christensen: sudolikeaboss: 0.2.1 -> 20161127-2d9afe1
<gchristensen>
thank you, Profpatsch
<gchristensen>
though I found go2nix to work with the checked out revision, Profpatsch
mudri has quit [(Ping timeout: 258 seconds)]
<ronny>
michalrus: thats intended - there is something about barrier free usage, where there is mouse delays and keyboard delays
<ronny>
is anyone working on wayland for nixos?
<michalrus>
ronny: uh… wicked. Do you have any resources about that, perhaps?
<Profpatsch>
gchristensen: It does? It just skipped right to master for me
<gchristensen>
it seemed to
<Profpatsch>
Or that’s what it looked like.
<Profpatsch>
Fuck, Prokrastination.
m0rphism has quit [(Quit: WeeChat 1.4)]
<Profpatsch>
Oh, wrong channel. :)
<maurer>
So, I don't know if it's nixos specific, but I can't seem to unshare -m without root on nixos, and the documentation for unshare() seems to indicate I should be able to
jensens has quit [(Ping timeout: 265 seconds)]
<maurer>
do we have some additional hardening I need to disable if I want to play with unprivileged fs namespaces?
griff_ has joined #nixos
magnetophon has joined #nixos
danharaj has joined #nixos
ryanartecona has joined #nixos
kampfschlaefer has joined #nixos
zagy has quit [(Quit: Leaving.)]
<magnetophon>
I'm trying to find out the cause of https://hydra.nixos.org/build/44408606/log "do not know how to unpack source archive". The upstream author of the pkg asks whether hydra uses a direct download or if we cache it.
<magnetophon>
I think we cache it, right?
<simpson>
magnetophon: Probably a direct download.
<simpson>
As opposed to fetching from a NixOS-managed mirror, right?
<michalrus>
ronny: found it with help of someone @ #gnome-shell. :) Settings → Universal Access → Typing → Repeat keys.
<clever>
magnetophon: looks like it should just use line 519-522
<clever>
magnetophon: and one neat quirk of how the binary cache works, nix-store -r nix-store -r /nix/store/yckqmy6q8sl988x3i5g1sbb310ig5f1v-x42-plugins-20160825.tar.xz
mudri has joined #nixos
<clever>
oops, doubled it up a tad, but that command will give you the exact tar hydra is using
<clever>
magnetophon: cant see anything weird with the derivation, it should just work
<magnetophon>
clever: file /store/path/to/source.xz confirms it's a xz. your command gives me "warning: you did not specify ‘--add-root’; the result might be removed by the garbage collector" and a copy of the path on the next line.
<clever>
yep, thats normal
<magnetophon>
clever: that pkg has been alternating between building fine and givig the above error for moths now. Even with updated versions of the pkg.
<clever>
yeah, i see it on hydra
<clever>
same storepath for the .tar.xz, for both the good and bad builds
<magnetophon>
clever: that's 100% proof that the actual tar.xz didn't change, right?
<clever>
yeah
<magnetophon>
so the problem is on our end, not upstream.
<clever>
something would have to be seriously wrong with the build slave, or the xz program is being broken
<magnetophon>
clever: there seem to be no related issues on github. afaik this only happens for this particular pkg...
c_ has joined #nixos
c_ is now known as Guest87276
<magnetophon>
Then again, I haven't opened an issue on it yet either, so it might be more common. I'll open an issue.
<gchristensen>
nixpkgs was broken, the `unstable` job doesn't evaluate, I think because curlFull depends on shishi which doesn't work on darwin?
mkoenig has joined #nixos
<gchristensen>
oh I think I misread the output, but anyway, nixpkgs is broken due to unstable not evaluating
* clever
heads to bed
<magnetophon>
clever: thanks so far
<magnetophon>
GN
<eacameron>
Sigh...nixops or VirtualBox keeps assigning the same IP to two different deployments (because one might not be running at the time another deployment starts). Then, if I try to deploy the SSH key doesn't match any more. Is there a way to force a deployment to get a new IP?
<eacameron>
Or possibly to put them on different networks?
cpennington has quit [(Ping timeout: 250 seconds)]
zraexy has joined #nixos
JagaJaga has joined #nixos
ilja_kuklic has joined #nixos
ilja_kuklic has quit [(Ping timeout: 250 seconds)]
cpennington has joined #nixos
sigmundv has joined #nixos
ilja_kuklic has joined #nixos
<LnL>
I don't think so, but the vbox dhcp server should remember the leases until you reboot
<mkoenig>
hello everyone. how can I get the path of a globally installed package? $(nix-env -q --out-path hello | cut -d' ' -f3) seems to work with user environment only.
<LnL>
mkoenig: "$(nix-build '<nixpkgs>' -A hello)/bin"
zimbatm has quit [(Remote host closed the connection)]
systemfault has joined #nixos
<mkoenig>
LnL: Thanks!
griff_ has quit [(Quit: griff_)]
<NixOS_GitHub>
[nixpkgs] pSub pushed 1 new commit to master: https://git.io/v10O1
<xwvvvvwx>
but on my local machine it's still at 1.4
<magnetophon>
LnL: what is the reason for addig an alias? Wouldn't that just delay the need for people to adapt their config? (assuming we will not keep the alias forever)
<xwvvvvwx>
though since the wiki is covered in warnings about how it's deprecated I'm not sure how much I should rely on the info there
<c0bw3b>
but still contains useful infos
<xwvvvvwx>
ok
<xwvvvvwx>
thanks for the help! :)
xwvvvvwx has quit [(Quit: WeeChat 1.5)]
<savanni>
I've some questions. There's a package that I need for work that is failing to install. I thought I would use this opportunity to track down the problem and fix it and contribute the fix back. But it actually looks like it's a dependency that's failing... and installing that dependency manually seems to work fine.
<eacameron>
LnL: I think another workaround for Vbox and nixops would be to basically tell nixops not to write no .ssh/known_hosts
<savanni>
And the package that is failing is icu_54_1 / icu4c-57.1. So... why would it fail to install as a depnedency but succeed when I simply install it on the command line?
<savanni>
(I *love* that cheatsheet that ya'll just posted.. it has answers to questions I never figured out how to ask)
<c0bw3b>
:)
<c0bw3b>
what is the package you are trying to install?
autofly has quit [(Ping timeout: 245 seconds)]
asymmetric has joined #nixos
<savanni>
zoom-us is the end goal.
<savanni>
I actually wanted to solve this without bugging you all until time for a PR :(
civodul has joined #nixos
<savanni>
oh, yeah. Could you try installing zoom-us to see if you get a failure?
<savanni>
I did a channel update this morning, but now I'm installing (I think) from a local clone of nixpkgs. If I got the command right.
<c0bw3b>
i think it's just the version of the icu dependency that is wrong
<c0bw3b>
you need 54.1 but the central repo version is 57.1
<c0bw3b>
or the other way around.. :p
<savanni>
I thought that, too... but this:
<savanni>
nixos.icu_54_1 icu4c-57.1
<savanni>
So I thought that alias was intentional.
autofly has joined #nixos
derjohn_mobi has quit [(Ping timeout: 268 seconds)]
david4 has joined #nixos
david4 is now known as xwvvvvwx
<savanni>
There's also something more subtle going on. Like I said... `nix-env -i icu4c-57.1` installs just fine. But `icu_54_1` as included from `zoom-us` doesn't. I don't see anything in the zoom file that implies that patches will be applied.
<c0bw3b>
you should try to override the relevant part of the zoom-us nix definition to force it to build against 57.1
<savanni>
According to the output, it's trying to build 57.1.
<c0bw3b>
but it's dling the 54.1 source
viric has quit [(Ping timeout: 246 seconds)]
<c0bw3b>
"unpacking source archive /nix/store/w66p9wkr42ns1kizd698sh64sd6x4awy-icu4c-54_1-src.tgz" in yourr gist
mkoenig has quit [(Ping timeout: 260 seconds)]
<savanni>
hmm... okay, I'm modifying the nixpkgs source for icu4c. It says 57.1 in it, but I'll see what the output says. And whether I can successfully disable the failing patches.
<savanni>
Like, not totalyl sure I have this process right.
<c0bw3b>
fair warning : not sure i'm solving your issue here :p
<savanni>
I'm modifying it directly in the zoom-us/default.nix file. As soon as I remove that 54_1, I get an error that `icu` isn't actually defined.
jedai42 has joined #nixos
<savanni>
oh wait...
<savanni>
I know why.
<c0bw3b>
it's also referenced i the libPath var
Guest46308 has quit [(Quit: Lost terminal)]
<savanni>
I was updating the libPath var without updating the input parameters. Requirements. Whatever that list at the top of the file is.
<dtzWill>
err wait but why are those patches failing to apply :(.
<dtzWill>
OH
<dtzWill>
lol
<dtzWill>
well
<savanni>
And that gives me a working zoom!
<dtzWill>
fwiw icu_54_1 fails to build for me too
jedai has quit [(Ping timeout: 245 seconds)]
ryanartecona has joined #nixos
<dtzWill>
with patches failing, etc, okay so there's sanity there
<gchristensen>
agh
<savanni>
So, I think that "54_1" was passing for me because what I was actually issuing was...
<dtzWill>
(on recent-ish nixpkgs/master, haven't checked upstream exactly)
<dtzWill>
haha, hooray savanni ٩(^ᴗ^)۶
<gchristensen>
savanni, dtzWill: you need to override patches = [] on icu_54_1 for it to build, but it is riddled with CVEs, so proceed with caution. zoom_us doesn't work with a newer ICU
<savanni>
`nix-env $f NIXPKGS -i icu4c-57.1`. And the package that it was choosing (because of ordering) was nixpkgs.icu
<savanni>
It doesn't work with a new ICU? I started zoom, but haven't tried creating a meeting yet.
<gchristensen>
I don't think so
<dtzWill>
gchristensen: np understood, I just.. assumed it built (icu_54_1) so was mentally debugging my way back to "well why is he getting patch failures" and well anyway, ty that makes sense
<gchristensen>
you replaced icu_54_1 with icu in zoom-us and it worked, savanni?
<gchristensen>
if it does that is great news and I'll send a patch to master / release ASAP
zraexy has joined #nixos
<savanni>
Yes. Video isn't working, though. Claims to be, and the webcam light is on, but nothing there.
<gchristensen>
well
<gchristensen>
who knows if that is icu or (insert problems with webcams on linux)
<c0bw3b>
^^
<gchristensen>
<3
<gchristensen>
thank you for the update, savanni, I'll test that and push a patch
<dtzWill>
haha re:webcams/linux
<savanni>
Insert problems with webcams on Linux.
<savanni>
This might prevent me from retiring my old machine.
<ronny>
michalrus: sorry for the confusion, i use german i18n, and misstranslated
<savanni>
Chrome won't actually talk to my USB webcam but does open the internal one. Zoom talks to either of them but doesn't receive data.
<savanni>
But I do see an error in the console: `pmap: argument missing`. Repeatedly.
redmq has quit [(Ping timeout: 260 seconds)]
redmq has joined #nixos
<gchristensen>
what if you tried using overrideDerivation to set patches to [] on icu_54_1, and use icu_54_1 on zoom-us? ( dtzWill can you help with that if she needs the help?)
<dtzWill>
gchristensen: sure thing
<dtzWill>
also we should set 'name' while we're at it, for sanity lol
<gchristensen>
thank you :)
<gchristensen>
oh gosh yes
<savanni>
Okay, so I'll modify zoom-us/default.nix...
<savanni>
And I'll use the overrideDerivation function to change both patches and name?
FrozenCow has quit [(Remote host closed the connection)]
<dtzWill>
savanni: shouldn't need to change default.nix, at least in the proposed change
<dtzWill>
savanni: i can be more specific (happily) but you'll want to modify the 54.nix to set name and patches where it already does overrideDerivation
<savanni>
Am I changing the icu/54.1.nix?
<savanni>
Oh, that makes sense... okay.
<dtzWill>
yeppers
<savanni>
oh wow. Build is running.
<dtzWill>
^_^
<dtzWill>
great!
zraexy has quit [(Ping timeout: 265 seconds)]
<c0bw3b>
how do i launch nix-repl to be able to use the strings manipulation functions defined in lib?
alx741 has joined #nixos
<savanni>
Builds and launches just fine, but doesn't actually share video or my screen. Which *might* just mean that it won't transmit anything unless someone else is in the room. Dunno about sound.
newhoggy2 has quit [(Remote host closed the connection)]
<savanni>
Fork, create a new branch, check my changes into the new branch, and submit a PR back to nixos/nixpkgs?
viric has joined #nixos
<dtzWill>
savanni: yep!
mudri has joined #nixos
<NixOS_GitHub>
[nixpkgs] joelmo closed pull request #16579: nixpkgs doc: add and clean some paragraphs (master...doc-guide) https://git.io/vodQL
<NixOS_GitHub>
[nixpkgs] joelmo opened pull request #20953: portaudio: remove old version portaudioSVN (master...master) https://git.io/v10ib
viric___ has quit [(Ping timeout: 268 seconds)]
asymmetric has quit [(Ping timeout: 244 seconds)]
cpennington has quit [(Remote host closed the connection)]
sigmundv has joined #nixos
Seichi has joined #nixos
<savanni>
This is going to take a bit. Familiarizing myself with the commit guidelines.
ryanartecona has quit [(Quit: ryanartecona)]
jsgrant has joined #nixos
wkennington has joined #nixos
ryanartecona has joined #nixos
<dtzWill>
savanni: no worries, thanks for going through the effort! ♥
<savanni>
I keep sticking with nixos because as frustrating as desktop use is.. I keep seeing the hints of how the underpinnings are so much better than I've used.
taktoa has quit [(Ping timeout: 265 seconds)]
pie_ has joined #nixos
Shoue has quit [(Ping timeout: 260 seconds)]
<pie_>
Profpatsch, woo 33c3 in _soon_
hiratara has quit [(Ping timeout: 245 seconds)]
jgertm_ has joined #nixos
hiratara has joined #nixos
taktoa has joined #nixos
newhoggy2 has joined #nixos
Shoue has joined #nixos
<NixOS_GitHub>
[nixpkgs] savannidgerinel opened pull request #20955: icu4c: fix #20954: icu_54_1 package does not compile (master...bugfix-icu-version) https://git.io/v10Ds
<savanni>
I'm still trying to do some of the steps in that PR form.
<savanni>
I'm stuck on the step of testing everything that depends on it. I can't seem to figure out how to make nix-shell depend on my local filesystem instead of nixpkgs.
<savanni>
I thought that I could specify `nix-shell 'local-path/default.nix' -p icu_54_1`, but no luck.
<savanni>
It just goes directly to the nixpkgs channel.
<dtzWill>
savanni: i think (?) that the buildbot does that for you these days, I've also not had success sorting how to make nox work with my PR's
<savanni>
Oh. Well... so long as it's okay that I've submitted a PR that doesn't have the checklist done..
<dtzWill>
but maybe someone has a better answer for you O:)
<dtzWill>
savanni: yes, I think it's mostly to make it clear what has (and what has not) been tried
magnetophon has quit [(Ping timeout: 248 seconds)]
stepcut has quit [(Remote host closed the connection)]
Itkovian has quit [(Quit: My MacBook has gone to sleep. ZZZzzz…)]
hiratara has joined #nixos
viric___ has joined #nixos
mkoenig has quit [(Ping timeout: 250 seconds)]
viric has quit [(Ping timeout: 245 seconds)]
mkoenig has joined #nixos
<NixOS_GitHub>
[nixpkgs] montag451 opened pull request #20956: containers: fix broken /etc/hosts entries when localAddress contains … (master...containers-hosts) https://git.io/v10FE
AllanEspinosa has joined #nixos
<NixOS_GitHub>
[nixpkgs] abbradar pushed 1 new commit to master: https://git.io/v10Ns
<NixOS_GitHub>
nixpkgs/master c182733 Nikolay Amiantov: skype: return to stock qt4...
AllanEspinosa has quit [(Ping timeout: 246 seconds)]
sigmundv has quit [(Ping timeout: 245 seconds)]
thc202 has quit [(Ping timeout: 250 seconds)]
AllanEspinosa has joined #nixos
calvertvl has quit [(Quit: Page closed)]
wink has quit [(Ping timeout: 245 seconds)]
RchrdB has quit [(Quit: Leaving)]
m0rphism has joined #nixos
<c0bw3b>
running nixos-rebuild test on a 16.09 small running a 4.8.11-grsec kernel
<c0bw3b>
and i have two warnings related to building kernel 4.8.12 :
<c0bw3b>
depmod: WARNING: could not open /nix/store/wljrqyai3jwvyw0hv5dya3dxlskal9f2-kernel-modules-shrunk/lib/modules/4.8.12-grsec/modules.order: No such file or directory
<c0bw3b>
depmod: WARNING: could not open /nix/store/wljrqyai3jwvyw0hv5dya3dxlskal9f2-kernel-modules-shrunk/lib/modules/4.8.12-grsec/modules.builtin: No such file or directory
<c0bw3b>
should i woorry ? ^^
anelson has quit [(Remote host closed the connection)]
anelson has joined #nixos
<joachifm>
c0bw3b: no
<NixOS_GitHub>
[nixpkgs] xwvvvvwx opened pull request #20957: nixos: docs: note that channels are per user (master...docs/channels_per_user) https://git.io/v10he
<c0bw3b>
well after a build switch and a reboot i'm running 4.8.12-grsec with no issues :)
<obadz>
domenkozar: occasionaly
<domenkozar>
obadz: :D
<dtzWill>
c0bw3b: yeah it's kinda amazing how easy Nix makes that :D
<domenkozar>
obadz: I figured out about why new make-disk-image is slow
<domenkozar>
it has to evaluate nixos config again
<dtzWill>
i've never bothered before previously but after a friend had it and ran into a bug building something of mine, I thought I'd try it and.... easy-peasy
edwtjo has quit [(Ping timeout: 260 seconds)]
<dtzWill>
minor problem with chrome/flash or something and priv'd user namespaces aren't supported I think which broke some random workflow I had but... well in light of recent CVE's and such I'm glad I nuked those anyway :)
mkoenig has quit [(Read error: Connection reset by peer)]
<obadz>
domenkozar: you know why it does that? --closure shouldn't require any evals..
<domenkozar>
obadz: no idea, it does say it's building configuration
<domenkozar>
maybe it's just misleading?
<obadz>
domenkozar: in fact, not even clear that <nixpkgs> is available inside qemu
edwtjo has joined #nixos
<domenkozar>
obadz: I'll dig into that
<domenkozar>
obadz: I still didn't understand your comment that it's faster because it copes files on the host
<obadz>
domenkozar: where did I say that it was faster?
<obadz>
domenkozar: so both ways see the closure of the outermost OS
<obadz>
via union-fs or whatever
<obadz>
in both cases, these closures are invalid (in that if you ever tried to invoke some nix command in there, it would delete everything because the db isn't in the right state)
<obadz>
in old case, you ignore that and just rsync over the data into the ext2 partition and fix the issue there
<obadz>
in new case, you fix the issue in the unionfs (possibly requiring hashing all the contents of the nix store, or maybe even rewriting it? I'm not sure), then you nix-env the closure into the ext2 partition
Rizy has joined #nixos
<domenkozar>
obadz: but that's not the only bottleneck
<domenkozar>
I've made fixing store validy optional since I don't need it
<domenkozar>
still 33m
<obadz>
having said all that, I'm not sure I understand my own comment either, so maybe just ignore it.
<domenkozar>
comparedto 8m
<domenkozar>
compared to 8m on 16.03*
<obadz>
but you agree that locally it runs fast?
<domenkozar>
that's you have hardware virtualization
<domenkozar>
that's because*
<domenkozar>
on EC2 you don't have it
<obadz>
right
<domenkozar>
obadz: I think you did good work
<domenkozar>
just want to figure out how to make it fast :)
<domenkozar>
or at least why it's slow
<domenkozar>
to begin with :D
<obadz>
maybe run old and new with date printed between every shell command so we know for sure where time is spent?
<domenkozar>
I did that
<obadz>
nix-env?
<domenkozar>
nixos-install takes up 33m now
<domenkozar>
out of 32min
<obadz>
ah
<domenkozar>
err, otherway around :D
jmiven has joined #nixos
<obadz>
can you put date prints inside nixos-install?
<domenkozar>
sure
<domenkozar>
I'll dig into it tomorrow
<obadz>
if it's the nix-env, then we can't drill any further but then not clear why it's different between the two
<obadz>
but it could be something else?
<obadz>
if you think my change isn't worth the brain damage, you can also revert it…
<obadz>
I won't take it personally :)
<domenkozar>
I think it's great
<domenkozar>
maybe we should instead invest really into VM-less make-disk-image
<domenkozar>
then we don't have to tweak these things at all
goibhniu has quit [(Ping timeout: 245 seconds)]
<domenkozar>
but we'd have to be able to run chroot
<domenkozar>
which is hard without mounts :)
c0bw3b has left #nixos []
<obadz>
domenkozar: the other possibility is that cp -prd is just meaningfully faster than nix-env --set closure when there's no hardware virt
<obadz>
but then the question is still why?
pie_ has quit [(Changing host)]
pie_ has joined #nixos
<obadz>
and if that's the case, then we should modify nixos-install to take advantage of that :-D
<gchristensen>
domenkozar: are you making many AMIs for snabb?
<domenkozar>
no, trying to deploy 100 EC2 nodes in as little time as possible for a fast feedback loop
<gchristensen>
cool, I've done the same :) may I ask what for