<kiloreux>
I removed the /srv from LD_LIBRARY_PATH and it won't work giving me that "ImportError: /lib/x86_64-linux-gnu/libz.so.1: version `ZLIB_1.2.9' not found"
<clever>
your /lib/x86_64-linux-gnu also has to be removed
<clever>
libraries in /nix/store should only ever be loading other libraries in /nix/store/
<clever>
and normally have the search path pre-set
<clever>
but LD_LIBRARY_PATH has a higher priority, so it can break things
eacameron has joined #nixos
magnetophon has joined #nixos
<kiloreux>
That's enlightening I completely deleted that variable but still getting this error. Not sure how can I make the python runtime call only /nix/store libraries in this case
<clever>
"which python", is it using the nix build of python?
<kiloreux>
It's not but rather "/srv/current/stack/bin/python"
<kiloreux>
how can i install the nix build of python
<clever>
ah, that one is using the ld.so from your host
<clever>
which defaults to looking in /lib
<clever>
nix-shell -p python
<kiloreux>
this will install python for me ?
<clever>
that will spawn a shell that has the nix version of python in $PATH
<kiloreux>
(it's actually downloading now), is there i can have it installed in a fix place like other nix packages ?
ryanartecona has joined #nixos
<clever>
'nix-env -iA nixpkgs.python' would place it at ~/.nix-profile/bin/python
<kiloreux>
And i will still have to run it with nix-shell -p python when running it ?
ryanartecona has quit [(Client Quit)]
<clever>
either nix-env or nix-shell
<clever>
nix-env will persist, nix-shell wont
<kiloreux>
Ohh I see. So it's one of them, and running the nix version of python, it will come with all nix variables by default ?
phinxy has joined #nixos
<clever>
that version wont look in /lib by default
<clever>
so its going to be more likely to ignore the incompatible libz.so
nh2 has quit [(Ping timeout: 255 seconds)]
joehh has joined #nixos
nh2 has joined #nixos
<kiloreux>
clever, that solved it for me. Thank you so much :D One last question, how can I run different version of python for nix
<kiloreux>
should I rename the nix version ?
<clever>
nix-shell ensures it will only be available in that shell, so nothing else can use it
<clever>
which prevents that problem entirely
<kiloreux>
I can't use nix-shell since it needs to be executed everytime and we run this in a production settings.
<kiloreux>
So it's unpractical for us to do that :(
<clever>
then write a nix expression that will generate a suitable script
<clever>
for example, make a $out/bin/foo that starts with #!/usr/bin/env python
<clever>
and then put python into the buildInputs
<clever>
nix should automaticaly patch it to /nix/store/<hash>-python/bin/python
<clever>
if you gist your current script, i can modify it
<kiloreux>
The current script is just an opencv script with #!/usr/bin/env python at its beginning
takle has quit [(Remote host closed the connection)]
<clever>
mkdir -pv $out/bin/
<clever>
cp -vi ${./input.python} $out/bin/output
<clever>
patchShebangs $out/bin/
<clever>
''
<clever>
then run nix-build on that file
<clever>
does it build?
<kiloreux>
Perfectly clever . Thank you so much for your help.
<clever>
you can also add the wrapProgram i gave previously to this
<clever>
and it will add a bash script around it to modify env variables for you
<kiloreux>
Alright, I am keeping a log of this anyway and I will get back to it when I get stuck. Your help is greatly appreciated <3 Last time and this time.
<ben>
i installed the go compiler in nixos, naturally it generates binaries with nixos-specific dynamic loader paths. is there a way to get it to not do that for "crosscompiling" to normal linux?
<ben>
patchelf seems helpless(?)
<ben>
ok well i edited the binary with vim, nevermind
sheenobu has quit [(Quit: My Mac has gone to sleep. ZZZzzz…)]
riclima has joined #nixos
Guest89 has joined #nixos
justelex_ has joined #nixos
Supersonic112 has quit [(Disconnected by services)]
Supersonic112_ has joined #nixos
Supersonic112_ is now known as Supersonic112
justelex has quit [(Ping timeout: 268 seconds)]
<Infinisil>
ben: Well if it works... Can't you just use the go compiler directly? If nix isn't involved it certains doesn't do anything to the built thing
strobelight has quit [(Quit: gtg)]
<contrapumpkin>
ben: pachelf should work. Why helpless?
<Infinisil>
s/certains/certainly
<ben>
Infinisil: nix installed and presumably patched the go compiler already
<hodapp>
I was running "nix-shell -A foo", but now it turns out that I need something more like (foo.override { enableBar = true; }) and am a bit lost as to how
<Infinisil>
hodapp: Does something like nix-shell -A foo --arg enableBar true work?
<hodapp>
oh, I'll try
joehh has quit [(Ping timeout: 240 seconds)]
montag451_ has quit [(Ping timeout: 255 seconds)]
eacameron has quit [(Ping timeout: 248 seconds)]
<Infinisil>
hodapp: Hmm, actually the manpage says this argument doesn't apply to nix-shell, something like nix-shell -E 'with import path/to/nixpkgs {}; callPackage path/to/foo { enableBar = true; }' should probably work
<Infinisil>
s/argument/option
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] ttuegel closed pull request #27522: Fixes for various bugs introduced by stdenv update in staging (staging...bugfix/staging/stdenv) https://git.io/v7fsl
NixOS_GitHub has left #nixos []
lverns has joined #nixos
<yegortimoshenko>
can i rely on /usr/bin/env being present?
simukis has quit [(Ping timeout: 268 seconds)]
<gchristensen>
yes
<yegortimoshenko>
great
eacameron has joined #nixos
mrcheeks has quit [(Quit: ERC (IRC client for Emacs 26.0.50))]
eacameron has quit [(Ping timeout: 240 seconds)]
<pie_>
some fellow on nixos "puppet does all of this plus like, three hundred more things, probably with about th e same level of total competence"
<pie_>
how true is this? ive never used puppet
<dash>
pie_: puppet is a poor imitation of nixos
<dash>
puppet does things by mutating system state until it converges with the desired outcome
<dash>
puppet provides no facility for rollbacks
<pie_>
ah
<lverns>
I think the big thing that nixos can provide that puppet can't is that if I don't tell nixos to install a package, then it won't be there
<Infinisil>
`man make`
<Infinisil>
No manual entry for make
<lverns>
but with puppet, it could get installed manually by me or someone else
<Infinisil>
Why don't I have the manpage for make..
<gchristensen>
Infinisil: do you have make installed?
<Infinisil>
gchristensen: Of course
<Infinisil>
gchristensen: Do you have a manpage?
<gchristensen>
I don't have make installed
<Infinisil>
gchristensen: You don't have make installod :O
<gchristensen>
*shrug* never needed it
<Infinisil>
Well, build environments and such, right
<gchristensen>
nix-shell :)
<dash>
Infinisil: make is usually present in build environments :)
<gchristensen>
anyway, a nix-shell -p make, indeed, seems to not have `man make`
<Infinisil>
gchristensen: Gosh damnit
<gchristensen>
indeed
<Infinisil>
And also damnit bash, I'm currently trying to do some abstraction in setup.sh, but bash makes this pretty hard..
<hodapp>
hrmph. if I run this build (opencv updated to 3.3.0-rc) normally, something in it causes xfeatures2d to be disabled. If I step through it in nix-shell (so far, unpackPhase and eval $postUnpack and then cmakeConfigurePhase), xfeatures2d is properly enabled
<dash>
abstraction in sh is an impossible dream
<yegortimoshenko>
i want to wrap a perl script that doesn't have shebang, with .pl extension, and make a derivation with it so that it is installed in $PATH. how do i approach this? does makeWrapper do that?
<hodapp>
not sure what steps I'm missing in this build
jgertm has quit [(Ping timeout: 260 seconds)]
<Infinisil>
hodapp: Calling `genericBuild` in nix-shell does all the steps as normal
joehh has joined #nixos
hyphon81 has quit [(Remote host closed the connection)]
<Infinisil>
Now I know why changes to setup.sh are so rare.. because changing it rebuilds every single thing..
<Infinisil>
I don't even know how I can test this properly other than waiting a whole lot
Rotaerk has quit [(Quit: Leaving)]
Rotaerk has joined #nixos
Wizek_ has quit [(Ping timeout: 248 seconds)]
lverns has left #nixos []
dtzWill has quit [(Ping timeout: 255 seconds)]
<Infinisil>
Going to sleep, see ya
joehh has quit [(Ping timeout: 260 seconds)]
ebzzry has joined #nixos
Infinisil has quit [(Quit: Zzzzz)]
joehh has joined #nixos
Fare has joined #nixos
sary has joined #nixos
alx741_ has quit [(Ping timeout: 276 seconds)]
joehh has quit [(Ping timeout: 240 seconds)]
alx741_ has joined #nixos
MP2E has quit [(Quit: reboot)]
MP2E has joined #nixos
sary has quit [(Quit: Changing server)]
jgertm has joined #nixos
losthaskeller has joined #nixos
<losthaskeller>
Hi people, any Haskellers around? I'm having trouble building a package
<losthaskeller>
My nix file starts like this: { mkDerivation, assert, base, ..... }: ....
<losthaskeller>
assert is a Haskell package, but it seems like it's a nix keyword too, and so the parser complains: error: syntax error, unexpected ASSERT, expecting '}'
<losthaskeller>
Any thoughts on what I can do?
sary has joined #nixos
justan0theruser has joined #nixos
joehh has joined #nixos
silver_hook has joined #nixos
silver_hook has quit [(Changing host)]
silver_hook has joined #nixos
justanotheruser has quit [(Ping timeout: 255 seconds)]
dtzWill has joined #nixos
silver_hook has quit [(Ping timeout: 240 seconds)]
iyzsong has joined #nixos
Wizek__ has quit [(Quit: Connection closed for inactivity)]
mkoenig has quit [(Remote host closed the connection)]
Fare has quit [(Ping timeout: 260 seconds)]
tmaekawa has quit [(Quit: tmaekawa)]
tmaekawa has joined #nixos
joehh has quit [(Ping timeout: 240 seconds)]
iyzsong has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
alex-v has left #nixos ["ERC (IRC client for Emacs 25.2.1)"]
kiloreux has quit [(Ping timeout: 255 seconds)]
sary has quit [(Remote host closed the connection)]
kiloreux has joined #nixos
proteusguy has joined #nixos
proteusguy has quit [(Client Quit)]
b has quit [(Quit: Lost terminal)]
simukis has joined #nixos
michas has joined #nixos
jgertm has quit [(Ping timeout: 240 seconds)]
endformationage has quit [(Quit: WeeChat 1.7)]
wkennington has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] calbrecht opened pull request #27577: module auditd: unbreak ordering cycle on basic.target/start #11864 (master...upstream-auditd-11864) https://git.io/v7Inc
NixOS_GitHub has left #nixos []
tmaekawa1 has joined #nixos
tmaekawa has quit [(Read error: Connection reset by peer)]
tmaekawa1 is now known as tmaekawa
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Ma27 opened pull request #27578: programs.thefuck: support shells that don't use `/etc/profile` (master...bugfix/thefuck/support-for-non-posix-compliant-shells) https://git.io/v7In9
<NixOS_GitHub>
nixpkgs/master 5328102 Frederik Rietdijk: Merge pull request #27572 from winniequinn/vscode-1.14.2...
NixOS_GitHub has left #nixos []
<rvolosatovs>
I need to reference a package defined in an overlay in the `configuration.nix`, but can't seem to find a way to do that - I tried to set `nixpkgs-overlays` in `nix.nixPath` and importing and merging the overlay in `nixpkgs.overlays`(I cannot pass a path to this attr) with `super`, but none seem to work
lukego has joined #nixos
Filystyn has joined #nixos
Filystyn has quit [(Changing host)]
Filystyn has joined #nixos
<lukego>
I'm using fetchTarball to grab a specific version of nixpkgs. I would like to make this cache indefinitely. Is there a way to do that? Currently it seems to be re-fetching the tarball fairly often but I know it is static (specific git commit id)
<LnL>
rvolosatovs: for configuration.nix you have to define overlays with nixpkgs.overlays
<lukego>
Maybe I can just as well do `import fetchFromGitHub { ... }` ?
<LnL>
yes, you need a fixed output drv
<lukego>
I tried passing fetchTarball a {url=...;sha256=...} set but it didn't take
<FRidh>
lukego: you need nixUnstable for that
<lukego>
ok. I'll see if fetchFromGithub works as a replacement.
<FRidh>
you could make a similar function yourself, but then it will be importing from a derivation, and you won't be able to use it on a Hydra
<LnL>
nix 1.11 doesn't have builtins.fetchurl?
<lukego>
Hydra can't import from a derivation?
<lukego>
Maybe then when I run under Hydra I would need to pass in an overriding value, which I could arrange to be exactly the same
<lukego>
like `{ pkgs ? import fetchFromGitHub { ... } }:` where Hydra would pass in `pkgs` and I be careful to make sure it's the same one
sooheon has quit [(Remote host closed the connection)]
<rvolosatovs>
LnL: that's what I tried to do, could a construct like this work(this obviously doesn't, but hopefully you get the idea)? what should I assign to? `nixpkgs.overlays = [ (self: super: { self = super // (import ./overlays/default.nix); })];`
<rvolosatovs>
`./overlays/default.nix` is similar to `top-level/all-packages.nix`, but calls packages from my own fork of `nixpkgs`.
<FRidh>
I wasn't entirely correct when I said Hydra doesn't support IFD. It can, its just that the evaluator does the work. Network access is however limited.
<FRidh>
lukego: any reason for not using `nixUnstable`?
nix-gsc-io`bot has quit [(Client Quit)]
<lukego>
FRidh: Bootstrapping. I'd like people to be able to run this expression with the version of nix that they already have installed.
<FRidh>
understandable
<rvolosatovs>
https://gist.github.com/rvolosatovs/a38266c8a873d327ae43ab6fba57cd78 why do I get "called without required argument ‘mopidy-local-images’" at evaluation of `mopidy-iris`(`mopidy-local-images` does not exist in the channel, but is a dependency of `mopidy-iris`)? how do I pass the dependency to the derivation? (that's part of `configuration.nix`)
<rvolosatovs>
ahaa, I see. thanks for clarification!
tmaekawa has quit [(Ping timeout: 260 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] FRidh pushed 1 new commit to staging: https://git.io/v7Il6
<NixOS_GitHub>
nixpkgs/staging 29f91c1 Frederik Rietdijk: Merge remote-tracking branch 'upstream/master' into HEAD
NixOS_GitHub has left #nixos []
<LnL>
is there a reason you're using callPackage instead of just inheriting the packages from the other nixpkgs set?
rigelk has quit [(Killed (niven.freenode.net (Nickname regained by services)))]
rigelk_ has quit [(Changing host)]
rigelk_ has joined #nixos
rigelk_ is now known as rigelk
rigelk_ has joined #nixos
rigelk_ has quit [(Client Quit)]
lukego has quit [(Ping timeout: 260 seconds)]
<rvolosatovs>
LnL: I followed examples I found in the manual and in mozilla overlay repo, not sure I know how to do that. (it's only my 2nd day actually writing some `nix` and first-time experience with functional languages :P)
<rvolosatovs>
I want to use the newest version of a package available, either in my fork or in the channel, with my defined `attrs` or packages only available in my fork, while allowing the automatic upgrades and easy work in the fork
<rvolosatovs>
some packages need to be referenced in `configuration.nix`
<rvolosatovs>
that's what I want to do. Currently I have my fork in `/nix/nixpkgs`, where I pull from 17.03 channel each time I want to upgrade
<deltasquared>
oh yeah, I forgot about the extreme modularity in node culture in particular
<joepie91>
you can technically still have modular frameworks but at that point the line in the sand between 'framework' and 'library' blurs considerably
<joepie91>
the things that most people call 'frameworks' are just the monolithic things
<deltasquared>
I'll have to ask #archlinux if I can borrow phrik
<joepie91>
deltasquared: also, bit of advice concerning dependency use; in JS (both Node and bundler-bundled browser JS), you want to explicitly *avoid* trying to define all your dependencies upfront; just add them as you go along, and you find yourself having a problem that a dependency can solve... otherwise you'll get stuck in analysis paralysis
<joepie91>
deciding on a stack upfront makes sense in a monolithic environment but in a modular ecosystem it'll just cause you grief
<joepie91>
:p
<gchristensen>
hello zarel
CMCDragonkai has quit [(Quit: WeeChat 1.7)]
<joepie91>
(this is one of the more difficult things for people new to JS but with experience in another language, usually)
<deltasquared>
joepie91: oh dw I'm known for lack of up-front construction anyways.
<joepie91>
hehe, alright
<deltasquared>
I go in with a high level view sure but not specific libs
<joepie91>
then you'll probably be fine :)
<deltasquared>
partly due to NIH syndrome but ┐(ツ)┌
<zarel>
deltasquared: what should w- do anyway?
<deltasquared>
zarel: it was supposed to be a cut-off "welcome"
<joepie91>
that's a much better shrug emoji than the one I usually see...
* joepie91
takes it
<deltasquared>
joepie91: I stole it from someone's xcompose
schoppenhauer has quit [(Ping timeout: 268 seconds)]
<deltasquared>
who probably stole it from someone else...
<gchristensen>
undoubtedly :)
<zarel>
in the end it was invented at silicon graphics
Bane^ has joined #nixos
<deltasquared>
zarel: well that's a new one
<joepie91>
nah, xerox
<joepie91>
:p
<mpickering>
I am trying to package a project which uses cmake, if I bring all the dependencies into scope and run "cmake ..." as is already in a makefile then it works fine but it seems I have to let nix called cmake because of the setup hook in the cmake derivation
<deltasquared>
bell labs!
<deltasquared>
hmm, I'm not that experienced with building things within nixos
<gchristensen>
mpickering: you're trying to do this inside a nix-shell or something?
schoppenhauer has joined #nixos
<mpickering>
I try and set cmakeFlags to be the right thing so that the invocations are the same but can't get things to work out properly
<deltasquared>
I haven't really had a chance to try yet because my host system would contaminate the default paths... >_>
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] rycee pushed 1 new commit to master: https://git.io/v7I0Q
<NixOS_GitHub>
nixpkgs/master 9e90d9e Robert Helgesson: gpsbabel: 1.5.3 -> 1.5.4...
NixOS_GitHub has left #nixos []
<mpickering>
gchristensen: I am trying to make a normal package, I was just trying it in nix-shell to see if I could do the build at all
<gchristensen>
mpickering: in this case, it might be easier to just use nix-build :P
mudri has joined #nixos
<mpickering>
that's what I am trying to do
<deltasquared>
joepie91: "don't rely on external state" well fucking said. I have a loathing for "trapdoors" and global state of most kinds
<zarel>
I'm reading about having ssh in initrd in order to unlock the encrypted root remotely but I can't find how it works. Should I log in remotely and manually unlock the volume? and then the stage 1 resumes his things when I logout?
<deltasquared>
zarel: well that's certainly one way of doing it. I assume you're trying to run sshd in the initramfs
<deltasquared>
make sure you've got all the network bits in there...
<joepie91>
deltasquared: it's one of the core things I teach people
<joepie91>
"global state is bad and you do not need it, the end"
<deltasquared>
joepie91: alas, most languages don't prevent arbitary modules calling exit() or using the FS
<deltasquared>
I want capabilities damnit
<deltasquared>
as in, object capabilities
<clever>
deltasquared: lua can often do that
<deltasquared>
clever: I think I saw that in a minetest mod for it's equivalent of redstone. a few keywords disabled even
<gchristensen>
mpickering: yeah but you're using nix-shell, I'd suggest skipping it for now
<clever>
deltasquared: world of warcraft also uses a slightly modified lua engine for all of its mods
<joepie91>
deltasquared: keep in mind that JS in general will not prevent you from doing anything; it's extremely easy to build abstractions (and to therefore produce reliable and maintainable code) if you make an effort, but it's also extremely easy to cause a mess if you don't :P
<mpickering>
I'm not using nix-shell
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] joachifm closed pull request #26632: coturn: allow use of ports < 1024 (master...nixpkgs) https://git.io/vHxn4
NixOS_GitHub has left #nixos []
<clever>
deltasquared: things initiated by the game have a special priv, that lets them call special functions (that trigger ingame actions)
bennofs has joined #nixos
<clever>
deltasquared: but if any mod based code is in the stack, that permission is lost
<joepie91>
deltasquared: so however strange that may sound, I kind of view JS as a powertool; it will let you do a lot of cool things, but paying attention to what you're doing is not optional
<clever>
deltasquared: so its imposible to automate any ingame action
<joepie91>
unless you want to lose a finger
<deltasquared>
better than a foot I suppose... ;)
<joepie91>
deltasquared: at least it doesn't have random pieces falling off and chopping off your foot, like certain other languages...
<joepie91>
*cough PHP cough*
<clever>
joepie91: " <?php" or "?>\n" in a library!
<joepie91>
so every once in a while, I make a comment about PHP being poorly designed and dangerous to use, and some PHP fan starts getting angry at me and saying that it's bullshit etc. etc.
<joepie91>
my favourite challenge to give them is "decode a JSON value and show an error if the JSON was malformed"
<joepie91>
almost nobody can correctly implement that in PHP, self-proclaimed PHP experts included.
<joepie91>
and it's a really good example of how everything is so poorly designed that there are a ton of non-obvious footguns
<clever>
decode_json with a catch block i'm guessing
<clever>
but id have to give it a test
<joepie91>
clever: nope!
<joepie91>
json_decode does not throw errors
<deltasquared>
joepie91: I saw many pitfalls with unicode handling
<clever>
returns false?
<joepie91>
clever: you're allowed to use the manual, btw - and even with that allowance, most people still get it wrong :)
<deltasquared>
... what do you mean it doesn't throw
<joepie91>
deltasquared: it predates PHP exceptions
<joepie91>
therefore it does not throw
<deltasquared>
...
<deltasquared>
so how did it report errors
<joepie91>
part of the stdlib produces exceptions, part of it doesn't
<deltasquared>
if it says "none" I'm going to strangle someone
<zarel>
deltasquared, clever: I read that it runs a dropbear instance and everything about the host keys but I don't know how the stage 1 behaves when something like cryptsetup-askpass is waiting for an input somewhere that is not the SSH shell I open
<gchristensen>
mpickering: can you paste your code & your error? :)
<joepie91>
clever: see, and that's the fun footgun; that bit of text usually leads people to check whether the JSON was valid by checking for a NULL return value
<joepie91>
except that is wrong
<joepie91>
because there are actually TWO scenarios where it can return NULL
<joepie91>
one is "the JSON was invalid"
<joepie91>
the other is "the JSON was valid but contained a literal null"
<clever>
zarel: i'm guessing you just manualy run cryptsetup open
<joepie91>
so the *correct* way to check whether json_decode failed, is to call json_last_error and look at the error code
<clever>
joepie91: yeah
<clever>
ahh
<joepie91>
naturally, this was only added in PHP 5.3
<deltasquared>
joepie91: eww, that sounds like C's strtoul()
<joepie91>
ie. later than json_decode
<deltasquared>
did I mention I despise errno
<clever>
we need an Either type!
<joepie91>
so in older PHP versions it was not possible to distinguish between a valid null and a decoding failure
<joepie91>
which means that all third-party docs on the matter tell you to check for null
<joepie91>
ta-da
<ToxicFrog>
deltasquared: an awful lot of the PHP stdlib is in fact the thinnest possible wrapper around C functions, so
<joepie91>
anyway, these kind of footguns are super common in PHP, virtually nobody knows them all
<joepie91>
it's extremely, EXTREMELY hard to write correct PHP
<joepie91>
to the point of non-viable
<ToxicFrog>
(this also explains some, but not all, of the wild inconsistencies in naming and argument order conventions)
<srhb>
PHP: A fractal of bad design is a hilarious read
<joepie91>
it's just easy to write PHP that *looks* like it works
<srhb>
(I feel much the same about most languages though)
<joepie91>
and people confuse that with "correct"
<gchristensen>
I dunno, it pays my salary nicely
<srhb>
It's like misanthropy but towards programming languages.
<zarel>
joepie91: gosh. I also often argue about php, and usually redirect them to phpsadness
<ToxicFrog>
srhb: all languages are terrible, but some are more terrible than others
<srhb>
ToxicFrog: Exactly my sentiment :-P
<joepie91>
zarel: unfortunately I have 9 years of PHP experience to source from...
<joepie91>
ToxicFrog: definitely not all. something something name bucketing
<clever>
joepie91: what about the speed difference between "foo bar baz" and 'foo bar baz' ? lol
adeohluwa has joined #nixos
<joepie91>
clever: that's actually the one vaguely justifiable one
<joepie91>
because one supports interpolation and the other does not
stanibanani has quit [(Ping timeout: 260 seconds)]
<joepie91>
so in certain runtime designs that can make an explainable perf diffference
<clever>
but its not doing any pre-processing at parse time
<ToxicFrog>
gchristensen: it used to pay mine, but that doesn't mean I forgive it~
<zarel>
ok, I'll try the encrypted root remotely unlocked in a VM and tell you how it goes
<joepie91>
it depends on the runtime design :P
<gchristensen>
maybe we need a #nixos-offtopic :P
<clever>
so even with zero use of interpolation, it still does 2 concat operations
<joepie91>
if you have a different execution path for potentially-interpolated strings...
<clever>
and the more words you have, the worse it gets
<joepie91>
clever: keep in mind that until recently PHP didn't even have a proper AST parser
<gchristensen>
it is hard to see the people needing help
<clever>
oh god, lol
<ToxicFrog>
Speaking of terrible languages, I'm sad that Abandon Matlab is (a) gone and (b) not on web.archive.org
<joepie91>
gchristensen: I think "let's cut down on the offtopic when questions are asked" is probably more useful :P
<deltasquared>
I concur, let's let the guy talk about the problem for a bit
<mpickering>
In that gist are, the default.nix, foo.nix which I am running with nix-build foo.nix, the script I am trying to port and the cmake error
<mpickering>
perhaps I am just approaching this packing completely wrong and should make a different derivation rather than try and do all this in the preconfigure phase
<mpickering>
but I wanted to stay close to the original build system
montag451_ has quit [(Quit: Leaving)]
<joepie91>
ToxicFrog: (not expecting a response, because support questions, but this is the thing I was referring to earlier with regards to 'name bucketing': http://news.php.net/php.internals/70691 )
<mpickering>
as I am a beginner
<gchristensen>
oh man, bazel adds so much craziness
<clever>
mpickering: without an unpackPhase, postUnpack and src are broken
<clever>
mpickering: so thats going to try to patch the source it hasnt unpacked
<mpickering>
seems to work fine?
<mpickering>
The source is definitely unpacked
<mpickering>
I have been checking with -K, the error comes from later at the cmake --build step
<mpickering>
which is why I think the CMake configuration is wrong somehow
<mpickering>
There are also some warnings which appear warning about unused configuration flags which don't appear when running in a shell
<mpickering>
which makes me think I have got a path wrong somewhere or it's not finding the right cmake files
<hodapp>
weird... still trying to do this OpenCV 3.3.0-rc build, and if I try to do it inside 'nix-shell', a certain feature (xfeatures2d) is "unavailable", but if I run nix-shell -A and then genericBuild, it appears to work fine
mudri has quit [(Quit: WeeChat 1.9)]
<mpickering>
AH I think I might know why.. lemme check
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] ehmry pushed 1 new commit to master: https://git.io/v7IEh
<NixOS_GitHub>
nixpkgs/master aad2a21 Jon Banafato: nim: 0.16.0 -> 0.17.0
NixOS_GitHub has left #nixos []
<mpickering>
yes here we go lol, I put the clangdir in the wrong path
<mpickering>
or rather, I did the copying wrong so that the path ended up being clang/HASH-clang-HASH/files
<mpickering>
what a stupid mistakes
jensbin has quit [(Quit: Leaving)]
<clever>
mpickering: i always cp -vi out of habbit, because i never get around to creating the alias
<mpickering>
I copied both dirs the same way and one was copied like this and the other wasn't
<hodapp>
hmm. if I just do 'nix-shell' I see "Couldn't download files from the Internet. Please check the Internet access on this host." at a certain step (seems to be from CMake), but with nix-shell -A & genericBuild, this doesn't happen
<clever>
if clang already exists, the clangdir will be put INSIDE the existing clang, with its original name
<qknight>
hey. how can i reference the value of a mkOption of another mkOption in the same scope?
<clever>
but if clang doesnt exist, it will be renamed, and take the place of clang
<clever>
qknight: via config like normal
<deltasquared>
hodapp: you haven't done anything funny with your /etc/nsswitch.conf have you
<mpickering>
clever: Amazing, thanks for taking the time to help me and gchristensen
<hodapp>
deltasquared: not that I know of
<gchristensen>
clever: nice work :D thank you for helping mpickering
<deltasquared>
clever: I swear, your nick and the colour it's appearing as, you remind me of a certain... individual
<clever>
deltasquared: who? heh
<deltasquared>
clever: uhhmmm
<mpickering>
yes and now it buids
<deltasquared>
this is awkward
<mpickering>
*builds
<clever>
i cant see the colors on this end
<deltasquared>
clever: it's purple/violet ish
<mpickering>
I might actually finish packaging it now this weekend
<gchristensen>
deltasquared: it would make sense, nick color is determined by the nick itself :)
<clever>
deltasquared: only thing that comes to mind for purple is willy wonka, lol
<deltasquared>
clever: you haven't seen the same TV shows evidently... hold on.
<hodapp>
weirdly, it just got through "-- xfeatures2d/boostdesc: Download: boostdesc_binboost_256.i" and several others... then failed at "xfeatures2d/boostdesc: Download failed: 6;"Couldn't resolve host name"" on the next file it was trying to get
ng0 has left #nixos []
<clever>
deltasquared: the color is probably just based on a hash of the nick
<deltasquared>
oh, I'm aware
<clever>
i got bored once and read the source of an !8ball script on irc, and it was a basic crc over the entire question, including non-printable characters
<clever>
so you can just pad it with ctrl+o's until you get the right answer
<hodapp>
I guess I should ignore the question in my head of "why is this CMake build getting shit off the Internet anyway?"
<deltasquared>
hodapp: you could ignore it but such build impurity will bring you grief for certain
<deltasquared>
seriously, who even /does/ that
<hodapp>
opencv is such a fucking dumpster-fire
<dash>
hodapp: i had to fix one of those once
<gchristensen>
hodapp: not uncommon :(
<gchristensen>
deltasquared: nix of course won't allow it :)
<deltasquared>
should we define good code as what builds easily in nix envs and which passes clang on -std=c<whatever> -pedantic -Wall -Werror? :P
<deltasquared>
<replace clang with lang compiler of choice>
<dash>
deltasquared: no
<deltasquared>
awww :/
<dash>
deltasquared: if it's written in C++ or C it's almost certainly not good code
<hodapp>
can someone try that build in my nixpkgs fork and rule out that it's a transient network issue? just do pkgs.opencv33rc.override { enableContrib = true; }
<deltasquared>
*mock offense* oh! he badmouthed muh language! the cheek- eh jk
tmaekawa has quit [(Ping timeout: 255 seconds)]
<hodapp>
and see if it fails early on at "-- tiny-dnn: Download: v1.0.0a3.tar.gz" or a big string of error banners after "-- xfeatures2d/boostdesc: Download: boostdesc_bgm.i" just in the configuring stage
Infinisil has joined #nixos
<clever>
hodapp: build started...
<hodapp>
clever: thank you. still configuring?
<clever>
hodapp: fetching deps
<hodapp>
alrighty
<clever>
configure has begun
<clever>
it shouldnt be capable of doing any network now
<hodapp>
wait, what?
<clever>
nix disables all network access during the build
<hodapp>
uhh, do you suppose that is my entire issue?
<clever>
yeah
<clever>
nix will spawn a new container for every build, with no network access, and limited access to the /nix/store/
<hodapp>
and during 'genericBuild', it hasn't disabled this?
<clever>
during the entire derivation, the network is off
<hodapp>
this is making me wonder how this build ever worked at all with OpenCV 3.2
<clever>
fetchFromGitHub is a different derivation, so it can have the network on, and nix will enforce it always giving the same result (via the sha256)
<joepie91>
cute, Atom's github plugin installs a statically-compiled version of Git into its own module directory, which, of course, doesn't work on NixOS...
<hodapp>
it's not clear to me whether this just never worked right in OpenCV 3.2 (I never used xfeatures2d until now and need a feature only in 3.3.0-rc), or whether this was due to me getting some repo revs wrong or something
<clever>
hodapp: the error is deep inside a generic ocv_download function
<hodapp>
my conjecture: CMake sees incorrect hashes and tries to re-download
<Infinisil>
joepie91: deltasquared: I rather think of it as fixing stuff
<joepie91>
I suppose that all things considered, my nonfunctional Git plugin isn't that bad
<joepie91>
lol
<hodapp>
joepie91: as a general rule, OpenCV is worse than everything
<joepie91>
(not that I was using it anyway)
<joepie91>
hodapp: I've found that generally anything starting with "Open" is bad
<joepie91>
OpenCart, OpenSSL, OpenCV...
<joepie91>
not sure why that is
<Infinisil>
OpenGL?
<joepie91>
the jury's out on that one
<hodapp>
it's RMS raining his wrath down on people who used the term "Open"
<joepie91>
:p
<deltasquared>
Infinisil: if openGL didn't use thread-locals I'd be happy :P
<hodapp>
thread-locals vs. steaming dumpster-fire....
<hodapp>
hmmm....
<hodapp>
yeah, gonna call it for OpenCV on that one
<Infinisil>
deltasquared: I'm not too experienced with openGL, but aren't thread-locals just something GPU's have?
<clever>
Infinisil: things can also be thread-local to the main cpu
<deltasquared>
Infinisil: I'm referring to the "context"
<clever>
Infinisil: which means you must always call the functions in the same thread on the OS side
<deltasquared>
there is one gl context per thread, manipulated by free functions
stranger__ has quit [(Quit: Connection closed for inactivity)]
<Infinisil>
Ah i see
<deltasquared>
*shudders*
zarel has quit [(Ping timeout: 276 seconds)]
<hodapp>
okay... so bootdescFiles is already using the hash the repo says to use
<clever>
Infinisil: some things like haskell and some event libraries will randomly switch threads for performance reasons
<hodapp>
likewise vggFiles
<dash>
joepie91: it used to be a good topic for jokes, Sun's "OpenDesktop" was sometimes referred to as "Open Deathtrap"
<Infinisil>
Some time ago I wanted to start using OpenGL on macOS with Xcode. After about a week of frustration I gave up, the tools were horrible
anelson has quit [(Ping timeout: 276 seconds)]
<Infinisil>
Then some time later Apple released the first version of Metal
<hodapp>
blugh. so... CMake is still just not finding the files, and I need to find out why
<Infinisil>
And holy damn, the integration, tool and language support are just sweet
<deltasquared>
Infinisil: the apple doth always taste sweet, but it be poison
<Infinisil>
Using Metal with Xcode and Swift is truly enjoyable
<Infinisil>
deltasquared: I really don't like Apple's general direction anymore, especially with them not supporting Vulkan.. But they done very well with Swift
<clever>
hodapp: and download_boostdesc.cmake appears to have the expected hash of each file it wants
<deltasquared>
clever: would it not be possible to simply include a patch as part of the "standard" openCV packaging for nix here? a patch to disable the hashing (and let something else do it)
<clever>
deltasquared: yeah
<deltasquared>
disclaimer: understandably, I'm still new to this
<hodapp>
clever: haven't checked hashes, but the same commit number is already in place in the derivation
<clever>
hodapp: id patch this to just not call download_boost_descriptors
<clever>
then see what happens
<deltasquared>
inb4 explosions
iyzsong has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
<clever>
/tmp/nix-build-opencv-3.3.0-rc.drv-0/opencv-3.3.0-rc-src/modules/stitching/include/opencv2/stitching/detail/matchers.hpp:52:42: fatal error: opencv2/xfeatures2d/cuda.hpp: No such file or directory
<hodapp>
but I can't understand *how* I got the error if I can't get the build to get past downloading this stuff...
<deltasquared>
hodapp: maybe it's compiling something to be used in the build process itself, like the kernel does
<hodapp>
clever: yeah, that seems to happen because for some idiotic reason someone includes cuda.hpp simply if xfeatures2d is enabled, and ignores whether or not cuda support was actually requested
<hodapp>
deltasquared: more likely, it's just an idiotic include that shouldn't be there
<deltasquared>
this is starting to sound like something I wouldn't want anywhere near any of my systems
<deltasquared>
if only because of the hacks in it
zarel has joined #nixos
<hodapp>
the problem is that OpenCV is one of the only libraries that has a bunch of algorithms that really aren't worth the time screwing around with coding myself
<hodapp>
now I do recall that if I run with genericBuild, I don't see that error about a missing CUDA header
<clever>
i recently discovered that a "CSV" export from MS access, includes fields like "1234\"
<deltasquared>
clever: that doesn't sound bad...? was that not supposed to be there
* deltasquared
prods gchristensen
<deltasquared>
are you still sane
<hodapp>
ugh. I am going to try commenting out "ocv_module_disable" for xfeatures2d
<gchristensen>
deltasquared: I write 1,000 line shell scripts. I've already lost it.
<deltasquared>
hodapp: I think you're in for an afternoon of this, I recommend having alcohol ready
<clever>
deltasquared: it escaped the end of field quote, causing my csv parser to eat half of the next field and go entirely out of sync
<hodapp>
it disables the entire module if ocv_download reports any nonzero status
<deltasquared>
clever: actually I'm pretty sure there's no "standard" to say that CSV files have backslash quoting behaviour, so I wouldn't call it their fault
<hodapp>
and I don't know if ocv_download is even properly reporting that the file was already there, or just trying to download it anyway and I sorta don't feel like learning enough CMake to see
<Infinisil>
nixos-unstable is still not updated..
<deltasquared>
Infinisil: error during the retrieval?
<mpickering>
can I have a mkDerivation with multiple srcs?
<hodapp>
deltasquared: let's be honest though; if there actually was a formal standard for CSV, they'd still ignore it.
<deltasquared>
hodapp: oh, sure. it just means we can't bash them as much.
<clever>
deltasquared: the php csv read function has no way to turn that escape feature off, making it imposible to read the file as-is
<dash>
deltasquared: haha if I rewrite anything it won't be in C
<hodapp>
dash: but C has a lot of benefits, like the fact that you will *always* have things to complain about and blame on developers whenever you need
<hodapp>
and... uhhhh.... all those other benefits which are so obvious I won't even list them
<joepie91>
lol
<Infinisil>
Well said
<deltasquared>
I'd probably settle for something like C++ with an extra semantic pass on top to add rust's lifetime checking rules in addition
<dash>
i can complain when working in any programming language
<dash>
deltasquared: why settle when there's rust
<deltasquared>
I have a particular few about qt if that helps
<hodapp>
I'd settle for something like C++... just... without most of what makes it C++
<joepie91>
dash: ah, you're a polyglot complainer!
koserge has quit [(Ping timeout: 255 seconds)]
<joepie91>
:p
<joepie91>
(so am I, though)
<deltasquared>
polyglot, the best kind
<deltasquared>
polyglots can be the best kind of anything ( ͡° ͜
<deltasquared>
oh, fuck.
<deltasquared>
erm
<joepie91>
lol
<deltasquared>
oops
* deltasquared
dives for table
<dash>
hodapp: let's take out all the compatibility with/resemblance to C and make a language called ++
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 1 new commit to release-17.03: https://git.io/v7IgF
<NixOS_GitHub>
nixpkgs/release-17.03 e982b64 Sarah Brofeldt: spotify: 1.0.57.474.gca9c9538-30 -> 1.0.49.125.g72ee7853-111 (downgrade)...
NixOS_GitHub has left #nixos []
<joepie91>
dash: so... Java?
<dash>
too much like C
<joepie91>
:P
<deltasquared>
joepie91: too many braces
<joepie91>
"it has braces"
<joepie91>
well there we go
<joepie91>
lol
<dash>
yep ;)
<Infinisil>
While we're already bashing programming languages, screw you Go
<dash>
yes
<joepie91>
insert generics complaint here
<dash>
C++ and C were designed like they are for semi-good reasons
<dash>
go has no excuse for being like that, at all
<joepie91>
also, who thought it was a good idea to make literally every number in JS a double-precision float
<dash>
joepie91: eich was in a hurry
<joepie91>
this has some fun crypto implications...
<deltasquared>
crypto? in JS? wat
<dash>
it's not his fault that the web got popular
<dash>
deltasquared: eh, you can use babel to get the features not in browsers yet
<Infinisil>
Regarding Go, I really don't like the trend of "<foo> written in Go", it's happening on a daily basis and I'm starting to lose faith in humanity
<joepie91>
deltasquared: equivalent table for Node.js: http://node.green/
<dash>
Infinisil: why did you have faith in humanity
<dash>
deltasquared: hopefully not much more "ew" than web development in general
<deltasquared>
I think html and JS could work out fine, it's the security model that has me worried.
<joepie91>
I'm probably weird but I like fixing bugs, so long as the environment they exist in is designed in a maintainable fashion
<joepie91>
deltasquared: what aspect of it?
<deltasquared>
joepie91: you mean, if fixing them is as easy as it should be?
Fare has joined #nixos
<joepie91>
deltasquared: more like "if fixing them involves mostly tracking down the cause and not so much juggling flaming swords to have them all arrange in a neat line"
<deltasquared>
joepie91: the "loading arbitary scripts from external servers" part, sandbox or no
<dash>
joepie91: definitely not; i'd rather fix bugs than add features, any day
<joepie91>
deltasquared: right.
<Infinisil>
I'd rather RIIR than anything
<joepie91>
deltasquared: realistically, that's not so different from almost any other environment though
<joepie91>
other than that it's a bit more restricted
<joepie91>
it's the combination of that security model and the assumptions people make as a result of that
<joepie91>
eg. "it's fine to shove in 20 tracker scripts from third parties"
<dash>
Infinisil: I sympathize but on the other hand there's also a big stack of new stuff that needs to be written :)
<joepie91>
mobile apps suffer from a similar issue
<joepie91>
it's just less visible :)
<dash>
the web can be destroyed in our lifetime, if we think big and work together
<joepie91>
deltasquared: just for the fun of it, install noroot firewall on a random Android device, use a few apps for a while, and look at how many connections are being made everywhere
<Infinisil>
Can anybody maybe give me a pointer on how to even reproduce this? Run the tests hydra does?
<Infinisil>
And what's the difference of nipkgs-unstable and nixos-unstable? nixpkgs builds all packages and nixos builds a nixos system?
ryanartecona has joined #nixos
Filystyn has quit [(Ping timeout: 240 seconds)]
<hodapp>
dash: ehhh, I thought Go was designed the way it was because it was for people who either didn't understand what a strong type system was for, or did and didn't want it :P
<hodapp>
but still wanted static types
<dash>
hodapp: meh, the dumb type system doesn't bother me much
<deltasquared>
C++ can have strong types! just throw structs everywhere
<symphorien>
Infinisil: I think that nixpkgs-unstable includes darwin builds
<dash>
hodapp: the bad concurrency model is my main complaint
<deltasquared>
where am I gonna put my network cable patch box now
<deltasquared>
dash: what, goroutines?
<Infinisil>
dash: I heard it's one of the best things of go?
<dash>
mutable shared-by-default state is no way to build a modern language
<dash>
deltasquared: doesn't your house have a wiring closet
<deltasquared>
dash: this network is only on the scale of my desk. and no, not for network cables in any case
<hodapp>
dash: welllll... it is sorta the basis of a lot of OOP still, somehow
<deltasquared>
hell the firewall profile is even called "desknet"
<hodapp>
dash: so they're clearly just doing it because it's popular. or something.
<joepie91>
deltasquared: double-sided tape under desk
<dash>
hodapp: ???
<hodapp>
dash: MUTABLE STATE ALL THE THINGS
<hodapp>
all the abstractions should be like miniature computers!!!1111
<dash>
hodapp: yeah that's great, the problem is sharing it between tasks/threads
* hodapp
throws something at Alan Kay
<hodapp>
dash: nah, that's fine, we'll just solve that with more objects
<dash>
nah, Kay was right, the modern expression of that idea is Erlang
<dash>
(and soon, Monte :)
<hodapp>
Kay just never really seemed to comprehend why Erlang purposely made local function calls different both semantically and syntactically from messages
stanibanani has joined #nixos
<srhb>
Infinisil: I don't think there's any active problem with nixos-unstable not updating, its tests are just not passing currently
<srhb>
Infinisil: So fixing it involves fixing those breaking tests (or more likely the broken software)
<dash>
hodapp: yeah, the distinction is slightly flatter in monte
<hodapp>
and it was in large part from his vision that we saw 1,500 iterations of "oh my god, since objects just use messages, and since networks just pass messages... LET'S... USE... OBJECTS... OVER... THE... NETWORK! Just like they're local!"
<Infinisil>
srhb: Do you know how I can reproduce this failed test locally?
<Infinisil>
Like how does hydra run these tests?
<hodapp>
and it took awhile before people actually realized that you can't just pretend the network's not there
<srhb>
Infinisil: nix-build nixos/tests/ipv6.nix I think
<dash>
hodapp: it's not a fundamentally bad idea!
<joepie91>
[16:02] <hodapp> and it took awhile before people actually realized that you can't just pretend the network's not there
mkoenig has joined #nixos
<joepie91>
there are plenty of people who still do not realize that...
<joepie91>
(see also: microservices)
<Infinisil>
srhb: Ahhh the tests directory, I'll have a look
<dash>
but yes, you have to notice the message sends across the network behave different
<deltasquared>
hodapp: CORBA was always doomed to failure this way
<deltasquared>
at least cap'n proto allows you to "promise chain" without multiple round trips.
<dash>
joepie91: next year, i expect megaservices to be trendy
<hodapp>
dash: yeah, it sort of is a fundamentally bad idea, because it's a fundamental misunderstanding of how networks actually behave
<srhb>
Infinisil: Repeat (on master) for each test in that page until everything is green :-)
<srhb>
Then the channel should update.
<dash>
hodapp: as deltasquared says, capnproto does this right
<deltasquared>
we should make everyone test their code on networks with 1s ping
<hodapp>
dash: which is sort of why in Armstrong's thesis Erlang was based very strongly around embracing the fact that things will fail and stuff in the network will go down
<srhb>
If the database doesn't run full again or something.
<clever>
deltasquared: nixos tests use vde_switch to link all of the qemu instances together
<hodapp>
joepie91: ehh, I don't have a particular issue with microservices, in that it seems to be a decent way to extend modularity to other things
<clever>
deltasquared: and vde_switch has some modes, where it can inject latency, bandwidth limits, and even packet corruption into the links
<Infinisil>
deltasquared: I occasionally have a ping about equal to my birth year :D
<dash>
microservices make sense for companies with microteams
<joepie91>
hodapp: it's not, though, because it's adding a network boundary for things that typically do not need a network boundary; in effect you're turning everything into a distributed system *by default*
<joepie91>
hodapp: modularity is fundamentally a problem to solve on the codebase level; you only move to the process level if there's some sort of technical constraint that makes it impossible to solve on the codebase level
<srhb>
NixOS tests are so cool.
<joepie91>
(eg. different, incompatible stacks for different tasks in the same application)
<deltasquared>
clever: oh yeah, VDE looks neat. never played with it personally
<dash>
joepie91: mm, why so?
<Infinisil>
srhb: Does nix-build <test>.nix run the test already?
<Infinisil>
Ah yes it does
<joepie91>
dash: because distributed systems are hard and network boundaries are expensive, and you want to avoid them wherever your requirements and constraints allow you to do so.
<dash>
joepie91: are you familiar with Conway's Law?
<dash>
hodapp: is basho still around?
deltasquared has quit [(Quit: segmentation fault)]
<hodapp>
dash: fairly sure I know 2-3 employees there
<joepie91>
dash: yes
<Infinisil>
Well, the ipv6 test passed
<dash>
joepie91: too many people adopting microservices don't I reckon :(
<gchristensen>
what provides libc++.so.1?
b has joined #nixos
<joepie91>
dash: honestly, the #1 problem with people adopting microservices is that they do not understand the distinction between "codebase modularity" and "process modularity", or that there even are these two separate concepts and they're not one and the same thing
<clever>
sometimes, zfs is just really slow, for no obvious reason
<hodapp>
I'll read that article though
<dash>
clever: magic takes time i guess
<joepie91>
dash: you can have a modular codebase that runs in a monolithic process, for example; and that gives you nearly every 'modular system' advantage (though not every single one)
<clever>
Jul 23 11:10:30 nas nix-gc-start[6246]: deleting ‘/nix/store/.tmp-link-2816-1576729590’
<clever>
Jul 23 11:10:30 nas nix-gc-start[6246]: error: getting status of ‘/nix/store/.tmp-link-2816-1576729590’: No such file or directory
<clever>
dash: and i think i just found a concurrency issue in the GC process
<Infinisil>
clever: What's zfs slow with?
<joepie91>
hodapp: way too long for me to listen to :) but yes, dealing with multiple cores does mean a distributed system; however, there's a different set of guarantees and failure scenarios from eg. multiple network systems
<clever>
Infinisil: reading the directory listing for /nix/store/ and /nix/store/.links/
<joepie91>
[on typical CPU architectures]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/v7IVn
<NixOS_GitHub>
nixpkgs/master 22292c0 Jamey Sharp: gnss-sdr: init at 0.0.9...
<joepie91>
you don't talk directly to cores, so you should really think in threads instead; and you have a guarantee that either all threads will be running or none of them will
<hodapp>
joepie91: it's a podcast worth listening to
<joepie91>
there's no "some of them are running" like there is in a networked system
<joepie91>
I'm sure, but I can't skim a podcast :P
<joepie91>
so it's not of terribly much use right now
<joepie91>
and I tend to have trouble focusing on audio-only podcasts anyway
<Infinisil>
clever: Well that's what a cache does, right?
<clever>
Infinisil: i think that perl function tuns the script under the context of the guest vm
<Infinisil>
clever: Yeah but still
<clever>
Infinisil: it already has 1.7gig of in-ram cache
<clever>
it could have bumped that up by 200mb, and gotten bigger benefits
Filystyn has quit [(Read error: Connection reset by peer)]
<clever>
so why does 200mb of on-platter cache help more?
Filystyn has joined #nixos
<Infinisil>
clever: Ahh
<Infinisil>
No idea
<clever>
that is the true mystery of zfs
<clever>
Jul 23 11:42:08 nas nix-gc-start[7180]: deleted or invalidated more than 16581001216 bytes; stopping
<clever>
Jul 23 11:42:08 nas nix-gc-start[7180]: deleting ‘/nix/store/trash’
<Infinisil>
And btw, that bash snippet is what makes the test fail
<clever>
i cant even tell what that bash is doing
<clever>
wait for a file, cat it, then delete it?
<clever>
Infinisil: aha, line 69 runs a script in the guest, in the background
<Infinisil>
clever: I'm not sure myself
<clever>
so it has to use that to block on things inside the script
<clever>
the perl on the host sends keys to the guest
<clever>
and the testreader in the guest reads them
<Infinisil>
from the test run: error: command `for i in $(seq 600); do if [ -e '/tmp/reader.ready' ]; then cat '/tmp/reader.ready' && rm -f '/tmp/reader.ready' && exit 0; fi; sleep 0.1; done; echo timed out after 60 seconds >&2; exit 1' did not succeed (exit code 1)
<clever>
then the host blocks on the guest creating files
<clever>
machine# [ 26.269645] layer3[955]: SUCCESS: Got back '}' as expected.
<clever>
machine# [ 30.265567] layer1[1012]: Waiting for 'f' to be typed
<Infinisil>
clever: The command nix-build '<nixpkgs/nixos/release.nix>' -A tests.keymap.x86_64-linux doesn't work, but nix-build '<nixpkgs/nixos/tests/keymap.nix>' -A qwertz does
<clever>
machine: sending monitor command: sendkey f
<Infinisil>
With the neo test I'm having the IceWM: MappingNotify thing again and it's gonna fail because of that
<Infinisil>
Maybe i'll run it a few times and see if it's consistent
bennofs has quit [(Quit: WeeChat 1.9)]
<Infinisil>
Yup, ran neo test again, succeeded this time
<Infinisil>
gchristensen: Well, hydra now succeeds with all keymap tests.. while that's nice, the problem is still there and will probably reappear on the next one
<Infinisil>
And a whole lot of other tests are failing too now
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] domenkozar created ghc-gold-linking (+1 new commit): https://git.io/v7IKk
<NixOS_GitHub>
nixpkgs/ghc-gold-linking 62781b3 Domen Kožar: ghc: apply a patch to fix common gold linker problem...
NixOS_GitHub has left #nixos []
<Infinisil>
Oh wait no
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] domenkozar opened pull request #27584: ghc: apply a patch to fix common gold linker problem (master...ghc-gold-linking) https://git.io/v7IKL
NixOS_GitHub has left #nixos []
<Infinisil>
keymap tests are still running
<Infinisil>
I think they're gonna fail, exactly the ones that failed last time are still pending
<clever>
the tests that have passed cant be restarted
<Infinisil>
clever: Ohh..
fikse has quit [(Ping timeout: 255 seconds)]
<Infinisil>
Ahh
<clever>
Infinisil: i ran the test twice on my machine, first one failed timed out waiting for the VM to connect
<clever>
second run passed, test script finished in 18.04s
<Infinisil>
clever: On hydra most tests fail because of connection timeout
<FRidh>
gchristensen: is there a way to retrigger the failed jobs of http://hydra.nixos.org/eval/1377845? Many failing jobs weren't retriggered although they should pass now (the fetchurl issue).
<Infinisil>
clever: Which one did you run?
<clever>
so there is a dedicated serial port on the guest, linked to the shell unix socket
<clever>
[clever@amd-nixos:~/nixpkgs]$ nix-build nixos/release.nix -A tests.keymap.neo.x86_64-linux
<Infinisil>
clever: Gonna run this a few times until I get a timeout
<clever>
Infinisil: after running it about a dozen times, it has hung, but the order of the prints i added says the accept blocks, and its not a race
<clever>
which would imply that the nixos guest isnt responding on the serial console
silver_hook has joined #nixos
<Infinisil>
clever: Wait which rev are you on? I can't reproduce the vm connection timeout error
<Infinisil>
I'm on the same rev as the hydra build
<Infinisil>
138eba
<clever>
407acb80fc0890e62597e170cc00d9e873e37e34
<Infinisil>
AH latest master
<Infinisil>
I'll try again with that
Ivanych has joined #nixos
pie_ has quit [(Ping timeout: 260 seconds)]
<Infinisil>
clever: How often does it fail for you? 1 in n?
<clever>
havent kept count
fikse has joined #nixos
<clever>
Infinisil: i think ive only had 2 failures in the last half hour
Ivanych has quit [(Ping timeout: 240 seconds)]
nslqqq has quit [(Ping timeout: 260 seconds)]
zarel has quit [(Ping timeout: 260 seconds)]
<clever>
machine# mke2fs 1.43.4 (31-Jan-2017)
<clever>
machine# Creating filesystem with 131072 4k blocks and 32768 inodes
<Infinisil>
clever: I've run it like 7 times now and haven't had it once
<clever>
Infinisil: this is the output from the middle of a passing run
<clever>
machine# mke2fs 1.43.4 (31-Jan-2017)
<clever>
this is the last thing a hung run has produced
Neo--- has quit [(Ping timeout: 240 seconds)]
nslqqq has joined #nixos
<Infinisil>
clever: I'm just automating it so I can let it run for a bit
<clever>
nixos/modules/virtualisation/qemu-vm.nix
<clever>
Infinisil: i suspect its hanging here, adding more prints
<clever>
399 mke2fs -t ext4 ${cfg.bootDevice}
noobly has joined #nixos
<noobly>
could someone kind and generous package exodus for nixos pls
<mpickering>
I think I'm going to give up trying to package this. Now I finally get to running "bazel build" and there is a very mysterious failure with no error message "Process exited with status 126"
<clever>
+ nix-store --delete warning: you did not specify '‘--add-root’;' the result might be removed by the garbage collector /nix/store/c229x42kja0bkdj3mipvz12bhpcc66ym-vm-test-run-keymap-neo
<clever>
error: getting status of ‘/home/clever/nixpkgs/warning:’: No such file or directory
<mpickering>
Now it seems I have completely hosed my install
<mpickering>
running any command leads to Died at /nix/store/866pyym5zgvx78kr55pwas16xp1m0sjp-nix-1.11.11/lib/perl5/site_perl/5.24.1/x86_64-linux-thread-multi/Nix/Utils.pm line 31.
<clever>
mpickering: what about nix-store -r /nix/store/kw31dip97zpfjzz24q2j9bzsgcav6mgc-nix-1.11.1
<nh2>
clever domenkozar: btw for a project I'm putting together a coupleof `doJailbreak` and `dontCheck` fixes I need to make a project work on top of nixpkgs master with ghc 8.2, can share those once it builds
zeus_ has joined #nixos
fikse has quit [(Ping timeout: 255 seconds)]
mudri has joined #nixos
nslqqq has quit [(Ping timeout: 260 seconds)]
* Infinisil
is back
<nh2>
what's the equivalent of `ghcWithPackages` that uses ghc 8.2 or HEAD?
zeus_ has quit [()]
zeus_ has joined #nixos
zeus_ has quit [(Remote host closed the connection)]
<bhipple[m]>
I'm considering putting an `include creds.nix` and gpg encrypting that file for source control, but was curious if there's an established idiomatic way within NixOS
<Infinisil>
bhipple[m]: I think the best you can currently do is just not putting passwords in any nix file and using the stateful /var/<foo> instead
<Infinisil>
bhipple[m]: Everything in /nix/store is world-readable
<clever>
there are some things like hydra where the code just doesnt allow reading the secrets from another location
<bhipple[m]>
Ah yeah that's good to keep in mind. In this particular case I'm not too worried about attackers on localhost, but good to practice clean security hygiene regardless
<clever>
and now that i think of it, hydra can also offer up its own nix store in full.....
<pikajude>
has anyone tried to build a GHC env with ghc821
<pikajude>
i get a "dependency foo doesn't exist" message for every package in base
<bhipple[m]>
so you're just suggesting in the password string I read a file somewhere on the filesystem?
<clever>
bhipple[m]: i mostly do a passwords.nix containing { foo = "bar"; ... } and then near the top of a module, let passwords = import ./password.nix; in ...
<bhipple[m]>
But that will hash it and put it in /nix/store, right?
<clever>
bhipple[m]: yeah
darlan has joined #nixos
darlan has quit [(Client Quit)]
<bhipple[m]>
Is there a way to do something like:
<bhipple[m]>
password = readFile "/etc/impure-password.txt" within a nix expression ?
xadi has quit [(Ping timeout: 240 seconds)]
<clever>
bhipple[m]: there is a builtins.readFile, but that happens at eval time, so its not impure
<clever>
bhipple[m]: that password will wind up in at least a .drv file in /nix/store/
pie_ has quit [(Ping timeout: 248 seconds)]
<bhipple[m]>
I've been reading up on https://github.com/NixOS/nix/issues/8, which would be awesome to have, but it looks like it's still in the design phase
pie_ has joined #nixos
<Infinisil>
> Issue #8 (!) > Still in design phase -.-
<bhipple[m]>
Discussion is still alive after 5 years of being open :)
<Infinisil>
bhipple[m]: It is a pretty important thing
<bhipple[m]>
Yeah, this would be huge for enterprise environments and any business work beyond a 2-3 man startup where they don't want to put their postgresql etc. passwords in plain text, might not want to expose exactly which version of openssl / nginx / etc. each application is running with (for intruders to easily scan for unpatched vulnerabilities), and might not want to expose all of their IP to everyone on the system
<bhipple[m]>
clearly looks like it'd be a lot of work to implement though
<Infinisil>
bhipple[m]: Keeping versions secret is a good point
<bhipple[m]>
Yeah I know a lot of companies turn off filenames and line numbers in stack trace log output, since that information leakage can expose what version the app is running to an intruder
<clever>
bhipple[m]: some of my php libraries dump every function argument in the backtrace, including the raw query, and the mysql password
<Infinisil>
clever: Regarding the tests that are failing, my script works now, ran it 10 times, but the checks always succeeded
<Infinisil>
oh no wait
<clever>
Infinisil: i have yet to get it to fail
reinzelmann has joined #nixos
<Infinisil>
Ah yup, succeeds all the time
<clever>
Infinisil: and the instant i say that, i get a new failure
<clever>
sub connectmachine# [ 19.309685] xsession[849]: IceWM: MappingNotify
<clever>
pbogdan_: thats a lot of cpu, but i dont even notice it in the gui, lol
<pbogdan_>
clever: I used that to track some past failures in xkb tests down to waitForX not being reliable (there is comment in the source to that effect)
<clever>
though top is unually slow to refresh
<clever>
ah, everything except the ssh to irc is laggy as hell, lol
<pbogdan_>
also in context of the recent hydra failures it seems to be consistently packet-t2-2 and lucifer machines
<pbogdan_>
at least that was the case last time I looked
tmaekawa has joined #nixos
<clever>
pbogdan_: yeah, i did notice 3 of the failures where on lucifer, but i was able to recreate the problems on my local machine 3 or 4 times
pie__ has joined #nixos
<clever>
heh, nix-build hasnt even started yet, and i can hear the cpu fans ramping up
pie_ has quit [(Read error: Connection reset by peer)]
<pbogdan_>
yeah, I could only conclude (guess rather) they were under heavy load, some earlier failures were due to waitForX before they started timing out before even getting to that point
<clever>
pbogdan_: the problem i reproduced on my end, isnt just a simple timeout
<clever>
pbogdan_: the mke2fs is hanging on boot, triggering the 5 minute timeout
<Infinisil>
clever: Have you ran strace on it?
<Infinisil>
already?
<clever>
since adding the strace, the problem has not happened
gm152 has joined #nixos
<deltasquared>
oh nice, a heisenbug.
<deltasquared>
inb4 the bug is a race condition that strace slows down and/or serialises
<Infinisil>
I'll run it on repeat
gm152 has quit [(Client Quit)]
<Infinisil>
deltasquared: The thing just happens so rarely
<deltasquared>
yep, sounds like a race.
b has quit [(Quit: leaving)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] Ericson2314 opened pull request #1483: WIP nix-shell: Use bashInteractive from <nixpkgs> (1.11-maintenance...nix-shell-bashInteractive) https://git.io/v7IHU
NixOS_GitHub has left #nixos []
<clever>
the real question though, is the race in mke2fs?, the guest linux?, qemu?, the host linux?
<deltasquared>
erm
<clever>
i have previously had a problem interaction between the host zfs and qemu, where the guest /nix/store just randomly swapped entire directories
<clever>
leading to files just not existing
<deltasquared>
how was the VM accessing that FS?
<clever>
9plan over the virtio interface
<deltasquared>
(for the record virtio-9p is dreadfully slow in my experience)
<clever>
which allows sharing host directories directly to the guest
<clever>
slow, but better then making a 2gig disk image for every vm
<deltasquared>
true I guess.
<clever>
and the current disk image routines use virtio-9p anyways, to copy things in
xadi has quit [(Ping timeout: 240 seconds)]
<deltasquared>
one wonders if a local nfs over virtio-net would access faster than 9p
<deltasquared>
still no page cache sharing of course, but sometimes that's desirable
<clever>
i also discovered a rather nasty nfs bug on 32bit clients
<clever>
if your nfs server uses 64bit inodes, then a 32bit client will fail in the weirdest ways
<clever>
getdents()'s will return EOVERFLOW, because the 64bit inode doesnt fit in the 32bit struct
<clever>
and 80% of software silently ignores it, treating the directory as empty
<clever>
there is a special glibc thing you can compile with, that will silently switch over to getdents64() on a 32bit os
<clever>
but it can potentialy break things
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 2 new commits to master: https://git.io/v7IHa
<deltasquared>
guess no-one made the reproducible cross compiler yet eh
<clever>
the cause, is that which compiler you use impacts the hash
<clever>
and a good reason for that, is that some versions of the compiler may be glitched
<clever>
you want to rebuild things when the compiler changes
<deltasquared>
surely you could arrange to use the same version of compiler though, regardless of cross
<clever>
its based on the storepath of the compiler
<clever>
and the storepath of the compiler, depends on the options it was build with, the platform it runs on, and the storepath of every one of its build inputs
<deltasquared>
hmm, there's no equivalent of a "provides" here it seems that could be used to interchange the two
<clever>
nope
<clever>
the purity in nix doesnt allow it
<clever>
even something as simple as write this string to a file, depends on which platform you run it on
<clever>
so you can have an arm, 32bitx86, and 64bit x86 build of a simple "echo foo > $out/bar.txt"
<clever>
the version of glibc and bash will also impact that build
<clever>
and which gcc the glibc was built from
<deltasquared>
in a sense, that compiler thing kinda shoots down the idea of a reproducible compiler in that it shouldn't matter if it produces identical output...
<deltasquared>
you'd have to make the build system assume from the ground up the compiler is a reproducible one so it could be ignored for hash calculation
<deltasquared>
and then only compare hashes of what it generates.
<clever>
deltasquared: at the core of nix, every derivation must be made by calling builtins.derivation, and passing it a set containing, system, builder, args, and name
<clever>
deltasquared: all attributes will then be forced down to a string, and the entire set is hashed
<clever>
that hash is then used to create the /nix/store/<hash>-<name>.drv path
<clever>
and its also used to compute the value of $out (which is in that .drv)
<clever>
and every attribute on that set, becomes an env variable when building the derivation
<clever>
so every value that can potentialy impact the build, will also impact its output path
ebzzry has quit [(Ping timeout: 268 seconds)]
<Infinisil>
Which really makes sense for having reproducible builds
<deltasquared>
Infinisil: yet these problems make it impossible to do cross compilation even with a reproducible compiler without winding up with a different hash
<deltasquared>
once again, cross compilation left at the wayside.
<clever>
but i have seen another plan of a possible solution
<deltasquared>
do tell before I start NIH scheming :P
<clever>
after building $out, you hash the entire thing, and then rewrite the references to itself (and its runtime deps) within every binary
<clever>
so you create a new $out, whose path is based on the temporary $out
<clever>
the hash of the temporary $out
_c_ has joined #nixos
<clever>
so if the native and cross-compiler produce the same output, things may share the result
Eisfreak7 has quit [(Quit: WeeChat 1.9)]
<Infinisil>
clever: So you arrive at something which works pretty much like IPFS
<clever>
Infinisil: part of the problem, is that rewriting the build to reference its own hash, changes its hash
<clever>
so you have to apply that rewrite at unpack time
<deltasquared>
the issue with my own approach is that you'd have to be able to show the subset of options used to make the compiler cross didn't affect the output (partly checked by it's output), then somehow special-case the compiler so that it's hash only depends on what it will output, not how it is built.
<deltasquared>
you'd have to break the hash tree at the compiler, so that it's deps (which will likely change between host arches) don't become part of it
adisbladis has quit [(Remote host closed the connection)]
<Infinisil>
Hold on
<deltasquared>
clever: one way to do it I guess would be some kind of "aliasing" thing. so the compiler itself is a normal package but it "provides" an interface of sorts (i've got no fucking clue what you'd hash in it), so that the hash e.g. says "I create reproducible binaries for this architecture with these default options"
<deltasquared>
dependencies of the "interface" provider wouldn't be included.
<Infinisil>
clever: I have never seen a "done formatting" in any run, successful or not. I put it after mke2fs
<deltasquared>
whereas godbolt would assume individual compilation
<schoppenhauer>
manveru: do you have some time left to help me, maybe? with my problem from a few days ago, with ruby?
<Infinisil>
deltasquared: But local variables never escape outside the current file
<deltasquared>
clever: ... so basically it optimised to "return 0;"
<deltasquared>
Infinisil: true, but malloc has side effects
<clever>
deltasquared: i dont even have a return statement in it!
<clever>
this is why i said the compiler is getting too smart
<deltasquared>
clever: the compiler puts it there, it's something the C standard says
<clever>
it knows malloc's only side-effect is allocating ram
<Infinisil>
deltasquared: The moving of this heap top pointer?
<clever>
and it can omit that, if the ram isnt used
<clever>
and it knows bzero only effects the ram its pointed at, and it can omit it if the ram is never read
<schoppenhauer>
or maybe, can anyone else help? I am trying to use bundix to bundle a self-written gem. can I give it a local directory (not github) as a source?
<deltasquared>
it's funny really, usually tricks like this are the compiler abusing undefined behaviour, but from what I read here that program is completely defined (memleaks aren't UB, just discouraged)
<schoppenhauer>
the thing is, I just want to have it locally, or at least in a non-public place
<clever>
deltasquared: i suspect it will do the same even if i store the pointers in an array and free them later on
<clever>
nothing used them, so it will probably omit both ends
ng0 has quit [(Remote host closed the connection)]
<deltasquared>
there is perhaps only one possible UB here but I doubt it would be due to the implementation, and that's that ints aren't guaranteed to hold the 2^30 bits for that number
<deltasquared>
though I'm pretty sure most compilers would warn about that.
<deltasquared>
*doubt it due to the implementation, rather
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bugworm opened pull request #27593: qtstyleplugin-kvantum: init at 0.10.4 (master...qtstyleplugin-kvantum) https://git.io/v7I5L
NixOS_GitHub has left #nixos []
<manveru>
schoppenhauer: heya
ng0 has joined #nixos
<manveru>
i got some time now :)
<manveru>
if your gem isn't public, did you publish it to a private gem source?
deltasquared has quit [(Quit: segmentation fault)]
<manveru>
otherwise, if you don't mind if it's not in your Gemfile, you could just add it to your RUBYLIB env var
lluchs has joined #nixos
<Infinisil>
clever: So I'm pretty sure now that the mkefs2 wasn't the problem
<clever>
Infinisil: the gist i have says it has hung at least once
<Infinisil>
clever: But on none of my runs, even successful, was the message after that printed
<manveru>
i'll still try to get path gems to work... should be fun
<Infinisil>
clever: Or did it for you?
<Infinisil>
While I don't quite understand how it can just not print that and still succeed, it doesn't seem to be related to the problem
<clever>
Infinisil: mine prints strace output every time, but has yet to hang
<clever>
Infinisil: i suspect there are also some buffering problems dropping fragments of the logs
takle has joined #nixos
<Infinisil>
clever: In 14 runs, I haven't gotten the "done formatting" message once
<Infinisil>
Eh, I'll just write a nice nix-expression for now and post it to the unstable-failing issue
<mudri>
Hi. I've roughly followed the instructions in the Nixpkgs manual for Ruby development ( http://nixos.org/nixpkgs/manual/#sec-language-ruby ), but the executable made by nix-build isn't wrapped with the gem dependencies. What more do I have to do to make it so?
<schoppenhauer>
ok, maybe setting the RUBYPATH is the better choice for my case
<schoppenhauer>
manveru: To set the RUBYPATH, I would probably use makeWrapper?
<manveru>
schoppenhauer: yeah
<schoppenhauer>
manveru: and set it to … the direct dependencies?^
<schoppenhauer>
manveru: does a gem repository have to be an own server? or can I, like for debian, just have a directory structure somewhere?
oida has joined #nixos
<Infinisil>
clever: Ohh, I totally missed that part. mke2fs is also the last message before every failing hydra job timed out waiting for the Vm to connect
oida_ has quit [(Ping timeout: 240 seconds)]
silver_hook has quit [(Ping timeout: 260 seconds)]
silver_hook has joined #nixos
silver_hook has quit [(Changing host)]
silver_hook has joined #nixos
<clever>
Infinisil: yeah, thats why i was adding echo's to this region
<Infinisil>
Ahh, actually not every job, tests.simple doesn't
<Infinisil>
Well maybe just a few actually
<clever>
rng
<Infinisil>
God damnit, this is frustrating
<clever>
heisenbugs always are
<schoppenhauer>
manveru: ok, I am … confused now. so if it was possible to just give a tarball path to the .gem-file to a pakcage, I think that might also be helpful for others.
<clever>
Infinisil: i think you can run listToAttrs over (import ./nixos/release.nix {}).tests and then use map to create a txt file refering to every test's $out
<clever>
Infinisil: then try to build that txt file
<clever>
and nix will just blindly build every test
<schoppenhauer>
manveru: if I can in some way help you with that, let me know!
<manveru>
yeah, working on it
<Infinisil>
Because if every test passes we might as well turn tests off temporarily until someone can fix hydra or the vm or whatever
<clever>
Infinisil: i think adding some retry to hydra would help
<Infinisil>
clever: Has gchristensen triggered every failing one before?
<Infinisil>
Or just the keymap ones
<clever>
dont know
<clever>
but this failure is so early in the boot, that i expect it to hit every single test in nixos
<schoppenhauer>
manveru: ok thx!
<grahamc>
I can retry all
<Infinisil>
grahamc: Would be nice :D
<Infinisil>
It appears to be a very good rnp
<Infinisil>
rng
<clever>
grahamc: having some automatic retry in hydra would help with these bugs
<clever>
but it would also increase the cpu usage in the cluster
<clever>
maybe only retry if the build failed within a certain timeframe
fikse has joined #nixos
ng0 has joined #nixos
silver_hook has quit [(Ping timeout: 240 seconds)]
<grahamc>
Be there In 10
<pbogdan_>
hmm it looks like every recent failure was on packet-t2-2 and lucifer, not just the xkb tests, but always a time out of some sort
jmsb has joined #nixos
<clever>
pbogdan_: but when i reproduced the test on my end, the mke2fs was hung, and the vm had gone idle
<clever>
it wasnt loosing cpu time to another process
<jmsb>
hi guys, i'm having issues with bumblebee. same issue as referenced here https://github.com/NixOS/nixpkgs/issues/24711 i was wondering if anyone had any guidance/tips as to how i could debug the issue
<clever>
gchristensen: the mke2fs within the initrd sometimes hangs
<clever>
gchristensen: i believe this is ran by every single nixos test and `nixos-rebuild build-vm`
<gchristensen>
it is indeed :)
<gchristensen>
Infinisil: have you confirmed they pass locally?
<clever>
and because stage2 and systemd never come up, the perl test-driver says it failed to connect to the vm
<gchristensen>
interesting
<clever>
gchristensen: the tests pass 99.9% of the time on this end
<pie__>
how do i fix this: output path ‘/nix/store/x3jjsvd6i4vzyhs9ylssg7gkyig031qi-slimit-cd76bde’ has r:sha256 hash ‘1gdj7bkql38g91v9cizhwf165gmksgwzxankqcw3axcd4305lljw’ when ‘1zass275yjn1srw7rgkq6ghvhz9gr2zsk21hpa46qsx1wzjwkhj3’ was expected
<pie__>
i already changed the hash in my shell.nix!
<clever>
pbogdan_: what does "less /nix/store/x3jjsvd6i4vzyhs9ylssg7gkyig031qi-slimit-cd76bde" say?
bennofs has quit [(Ping timeout: 248 seconds)]
<Infinisil>
gchristensen: I tested a few and they succeeded, hold on I'll test some more
<gchristensen>
I'd be hesitant to add retries to hydra, I think it'll only make the problem worse
<gchristensen>
Infinisil: already restarted :)
<clever>
gchristensen: yeah, it would be better to just fix these problems
Guest48903 has joined #nixos
<gchristensen>
other options: reducing load on the hosts, and/or fixing unstable parts of the testing
<clever>
but when ive only been able to reproduce it ~4 times, and it refuses to reproduce now that i have an strace in there
<Infinisil>
gchristensen: Fixing unstable parts on the unstable branch :P
ng0 has quit [(Quit: Alexa, when is the end of world?)]
<gchristensen>
it can be done :)
<Infinisil>
There have been a few tests failing for a few builds, such as plasma5 and ssdm, so I'd expect these to fail again
<pbogdan_>
clever: gotta run for now but can check in about 20 mins or so
<gchristensen>
one example of my own doing is instead of taking a screeenshot of the console, then OCRing the text on the text-onlyconsole is to just read the text from the console device directly
<Infinisil>
But the others are pretty random
<Infinisil>
gchristensen: Is that what you consider unstable?
<gchristensen>
the OCR was unsttable
<gchristensen>
it didn't always work, just failed ... randomly ...
<Infinisil>
gchristensen: I see, it's pretty cool though :)
<gchristensen>
sure, we still use OCR for the GUI tests :)
<Infinisil>
Damnit
<Infinisil>
It's exactly the same jobs
<Infinisil>
It's so not random
<Infinisil>
Now it makes even less sense
<Guest48903>
Hello everyone. I asked this question the other day but had to shut down before I could get an answer: Where in nixos can I find installed items such as wallpapers, oh-my-zsh themes and cursor icons without using the store path (since that can change on an update) I can inly find wallpapers using the store path, cursors are in /run/current-system/..., but are not recognized by lxappearance (it appears
ng0 has joined #nixos
<Guest48903>
to want an archive), and I cant find oh-my-zsh themes at all.
<gchristensen>
I would venture to guess, then, that the tests are exercising a part of the system that is otherwise heavily loaded and causing it to fail Infinisil
<clever>
gchristensen: but the failures i saw are happening so early in the boot, that every test passes thru that code
<Infinisil>
gchristensen: But these tests aren't exactly well groupable in any way
<Infinisil>
Oh wait, we actually don't know yet if they failed for sure
<timclassic>
Is it possible to specify printer information directly in /etc/nixos/configuration.nix?
<clever>
pie__: what was the contents of the storepath i pointed to?
<Infinisil>
But since it's a timeout error of 5 minutes that's been causing them to fail previously and no test has updated, it's probably the same
<pie__>
clever, oh you highlighted the wrong guy
<pie__>
i didnt even notice
<clever>
*doh*
<pie__>
:3
<pie__>
clever, the git repo contents
<qmm>
when building a haskell package, do you need a file named Main.hs? i have two files that i want to make executables, and as a result, i personally don't think it makes sense to name one of them Main.hs
<clever>
pie__: try just fixing the sha256 again?
<gchristensen>
load average: 66.11, 62.89, 53.22
<clever>
gchristensen: wut?, on which machine? lol
<gchristensen>
t2-2
<LnL>
ah one of the packet boxes
<Infinisil>
Is that too much?
<gchristensen>
powerqueue-test has 248hrs of cp utime
<gchristensen>
and 2,124% CPU
<LnL>
lol
<pie__>
clever, but i did change the sha to the current one??
zraexy has quit [(Ping timeout: 240 seconds)]
<pie__>
oh wait i think i got it backwards
<Infinisil>
HOLD ON, A TEST PASSED
<pie__>
*facepalm* thanks clever
<gchristensen>
heh, /nix/store is 99% full and I started to worry a lot: it should be garbage collecting! it is :) it GCs down to 30g free, and there is 1.7T in there
<qmm>
this is the error i get when running nix-build on my package: <no location info>: error: output was redirected with -o, but no output will be generated because there is no Main module. builder for ‘/nix/store/hn1h7w903kfd850r34a5dr0xs6d5p1yq-testing-HEAD.drv’ failed with exit code 1 error: build of ‘/nix/store/ hn1h7w903kfd850r34a5dr0xs6d5p1yq-testing-HEAD.drv’ failed
<Infinisil>
dhcpSimple isn't failing anymore
<clever>
qmm: do you have a cabal file?
<pie__>
clever, uh...looks like it got it from git but didnt actually install it......
<gchristensen>
clever: a lot of it is IO too, I'm seeing 40-50M/s read, 10M/s write, pretty sustained
<pie__>
clever, ok more foolish problems, https://gist.github.com/22459/acc90deb2e80e69e4c7142cdbb845ac6 i get ply 3.8 via propagateBuildInputs = [ python36Packages.ply ] , but then it complains that it cant find ply? is propagateBuildInputs the wrong thing?
<clever>
gchristensen: load average: 66.30, 29.04, 12.06 and the test can still pass
<gchristensen>
yeah but you're hammering the wrong component
<gchristensen>
max out your drives and try the test again :P
<clever>
pie__: dont know enough about how python searches for deps
<qmm>
clever: my cbaal file looks similar to snoyberg's (thinks for linking, btw)
<clever>
qmm: and what is your nix file?
Guest48903 has quit [(Quit: leaving)]
<Infinisil>
clever: gchristensen: Alright, the only actual tests failing are ssdm and plasma5
<clever>
gchristensen: system is noticably laggy with 63mb/sec going to the drives
<Infinisil>
which have been for a while
<clever>
98mb/sec
<clever>
cant even close a tab in the browser, heh
<gchristensen>
yeah I'm peaking to 90M/s :)
<pie__>
clever, ok
<clever>
i'm on an SSD mirror, so its shoving that much data into both drives
<clever>
twice the bandwidth going thru the sata controllers
<clever>
gchristensen: this reminds me of a common complaint ive had with things like -j, they arent aware of the load types
erictapen has joined #nixos
<gchristensen>
indeed :)
<clever>
gchristensen: for example, i want to make -j4 on my rpi, to use all 4 cores, but there is a single step in the libc locales, that needs all the ram
<clever>
and make tries to make it share with 3 gcc's
<clever>
so it takes 12 hours to run
<gchristensen>
due to swap?
<clever>
yeah
<gchristensen>
ouch.
<clever>
and also, some ghc compiles need 5gig of ram
<gchristensen>
yeah, that is a problem with -j. it is a poor cludge really
<clever>
so even with 16gig in my machine, i can only handle ~3 jobs in parallel
<clever>
but if i make -j3, on a 8 core machine, over half is going to waste
<clever>
error: writing to file: No space left on device
<clever>
amd/root 20G 8.2G 12G 42% /
<clever>
gchristensen: i wonder, when does stress delete its temp files, lol
<clever>
qmm: looks like it should just work, what is the error?
<clever>
ah, 1gig per worker, with 20 workers
<clever>
plus zfs snapshots saving every byte they make, lol
erictapen has quit [(Ping timeout: 240 seconds)]
<clever>
gchristensen: test took forever to start with 105mb/sec going to the drives, but it can still pass
bennofs has joined #nixos
ng0 has quit [(Quit: Alexa, when is the end of world?)]
<aristid>
what's the easiest way to "subtract" two lists from each other?
<gchristensen>
clever: how about both?
<aristid>
i have two lists of IP addresses, and i want the IPs that are in the first but not in the second list
<clever>
gchristensen: whats weird, is that i wasnt even taxing the system that hard when i reproduced it the first few times
<clever>
gchristensen: just normal browsing activity
<clever>
aristid: my first thought is builtins.map and builtins.elem, test each to see if it exists
<aristid>
lib.subtractLists
<Infinisil>
clever: I'm writing something down, need to know: have you had the error "timed out waiting for the VM to connect" on your local machine once? I can't remember
<clever>
Infinisil: 2 or 3 times
<clever>
Infinisil: each time, i narrowed in closer on it, until i added the strace, then it never happened again
<clever>
it happened exactly once with the echo's around the mke2fs
<Infinisil>
clever: Ah thanks
<Infinisil>
clever: This was early in the boot process right?
<clever>
yeah
<Infinisil>
because I've never encountered this one
<clever>
mke2fs started, and never finished, so the stage-1 boot hung
<Infinisil>
clever: Ah yes, the hand at "f" was the one where only 1 of 2 passed
endformationage has quit [(Quit: WeeChat 1.7)]
<Infinisil>
Alright thanks
ng0 has quit [(Client Quit)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] aristidb pushed 1 new commit to master: https://git.io/v7IpR
<NixOS_GitHub>
nixpkgs/master 6319054 Aristid Breitkreuz: wireguard: sometimes module tries to re-add the default route, which fails - use replace to make it succeed
NixOS_GitHub has left #nixos []
simukis has quit [(Ping timeout: 255 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bennofs pushed 2 new commits to master: https://git.io/v7Ipo
<gchristensen>
like to install nix on debian via a .deb?
<pie__>
clever, im a bit confused. "and then execute nix-shell will result in an environment in which you can use Python 3.5 and the toolz package. " in https://github.com/NixOS/nixpkgs/blob/master/doc/languages-frameworks/python.md suggests that i am just supposed to run nix-shell and that it should give me the right environment, not an env in which im supposed to build it
<pie__>
besides, i thought nix-build was for build environments?
<Infinisil>
Whoa there, the mailing list moved to google? I don't like this
<bennofs>
even though I'm a pretty heavy user of google services, i have to say I don't like the interface of google groups
<pie__>
all my favorite projects are ending up on google groups >:(
<Infinisil>
bennofs: It's a horrible javascript mess
nuser has quit [(Quit: Page closed)]
<Infinisil>
it loads like 5 times slower than any other site for me
<bennofs>
well, and even though it does that it doesn't even look pretty or useful :D
<Infinisil>
bennofs: Exactly
<Infinisil>
So, can I still just join the mailing list without a google account?
<bennofs>
yes, i think so
<Infinisil>
NOPE
<Infinisil>
ARE YOU SERIOUS
<LnL>
oh, really?
<Infinisil>
OVer the last couple months/years I have continuously distanced myself from google, but now my favorite OS requires a Google account to participate??
<bennofs>
At least that's what the nix-dev announcement says
<Infinisil>
Yeah it uses my login
<bennofs>
Infinisil: or by sending an empty mail to nix-devel+subscribe@googlegroups.com.
<bennofs>
perhaps?
<Infinisil>
While I have a google account I refrain from using it as much as possible
<LnL>
I get an email prompt if I log out
seppellll has joined #nixos
<avn>
Infinisil: it works via normal email as well
<Infinisil>
bennofs: It sends you a link to joining the group online
<avn>
Just write to nix-devel+subscribe@
<schoppenhauer>
manveru: when can I get on your nerves with it next time? ;)
<Infinisil>
avn: That's what I did
<Infinisil>
All I get back is the link bennofs sent
hiratara has joined #nixos
<avn>
Infinisil: well, after few dances I force it to send me old-style plain emails ;)
<Infinisil>
Where nothing even shows up without being logged it
<Infinisil>
in
<Infinisil>
avn: I'd be furious if not even that is possible
<Infinisil>
Google isn't so good if you're not using any of their services..
<bennofs>
googlegroups even isn't good if you *are* using their services :) it doesn't really integrate with anything, not even notifications afaik
<Infinisil>
What does google groups do that normal mail doesnt?
<avn>
Infinisil: main problem, no one want host huge mailing list ;)
<pie__>
no you cant
<LnL>
lol, now the ui is confused and wants me to subscribe :p
<Infinisil>
avn: Nixos cache, providing cached versions of everything 24/7, but a mailing list is too much?
<pie__>
Infinisil> NOPE
<pie__>
<Infinisil> ARE YOU SERIOUS
<pie__>
<Infinisil> OVer the last couple months/years I have continuously distanced myself from google, but now my favorite OS requires a Google account to participate??
<bennofs>
Infinisil: cache is also not provided by nixos itself, it's backed by Amazon cloudfront
<Infinisil>
bennofs: Ahh right
<pie__>
avn, huh..
bennofs has quit [(Quit: WeeChat 1.9)]
<Infinisil>
goddamnit
<pie__>
so...second mailing list move?
<pie__>
:P
<Infinisil>
pie__: I'm in
<pie__>
i dont quite understand why email needs to be so hard *scratches head* its just a mesaging protocol (im not saying its not complicated, im saying WHYYYY)
<Infinisil>
Yeah
<Infinisil>
I bet my 5$ digicalocean droplet can host this friggin mailing list
<pie__>
other projects are going on all this web2.0 stuff and im like whyyyy
frankpf has joined #nixos
<pie__>
well i mean i guess its not so hard to answer. people want to be able to send pictures on irc
<gchristensen>
it is because it is so easy that it is so hard
<Infinisil>
Oh thanks Google, for sending me an email recommendation to download Gmail, Google Fotos and Google Maps because I signed up to this mailing list -.-
<pie__>
Infinisil, idris also uses google groups :(
<Infinisil>
pie__: Yeah, a reason I haven't joined yet
<Infinisil>
But now that I joined nix-devel I might as well join idris
<pie__>
theres an irc chan
<pie__>
people are friendly
<pie__>
(i mean its not a mailing list, i know)
<pie__>
now i have two gripes about nixos ;P systemd and google groups
<Infinisil>
pie__: While systemd is kinda bloated, it has worked well for me (although I haven't used it extensively)
<Infinisil>
Alright, so how can I access the old mailing archive?
<pie__>
it didnt cause me problems and then it did. i havent actually debugged it but it think it causes my system to hang sometimes. either that or hardware failure or somethin
<bennofs[m]>
I wonder why there hasn't been a init system that is not systemd but supports its unit files
<neeasade>
can you set package priority in configuration at nix?
<avn>
pie__: do you like to make non-systemd init plugin to nixos? ;)
<chrishill>
Hey. I might have asked this before, but how do I install Spotify? I have this in my `.nixpkgs/config.nix`: `{ allowUnfree = true; }`, so the fact that it is proprietory shouldn’t be the problem.
<pie__>
avn, i want to but im not qualified ;P
<chrishill>
I looked at the top-level packages, and `spotify` is listed there. But `nix-env -i spotify` does not work. And `nix-env -qaP` does not seem to list `spotify`, either
<pie__>
crap python36Packages.networkx is broken
<clever>
chrishill: does ~/.config/nixpkgs/config.nix exist?
<qmm>
the error i am receiving is this: <no location info>: error: output was redirected with -o, but no output will be generated because there is no Main module. builder for ‘/nix/store/hn1h7w903kfd850r34a5dr0xs6d5p1yq-testing-HEAD.drv’ failed with exit code 1
<avn>
pie__: I am would like to support it, but I am enough ok with systemd to keep stuff as-is ;)
<clever>
qmm: can you gist all of the output?
<pie__>
avn, another great reason for tearing out systemd would be that it would be easier to do a nixos variation of BSD :D
<pie__>
avn, though there are a large number of linux specific features nixos uses so idk :C
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] LnL7 opened pull request #27598: CoreFoundation: (master...darwin-cf-rpath) https://git.io/v7Lv7
<chrishill>
clever: No. I use `~/.nixpkgs/config.nix`, not `~/.config/nixpkgs/config.nix`.
<clever>
chrishill: what happens if you do nix-env -iA nixos.spotify
yegods has joined #nixos
<clever>
chrishill: and what is the output of "nix-build '<nixpkgs>' -A spotify -v 2>&1 | grep config.nix"
hiratara has joined #nixos
justelex_ has quit [(Ping timeout: 248 seconds)]
<chrishill>
clever: I get this output: `error: assertion failed at /nix/store/41jlw6ry3sw486sgwdzr35ffxqgzwcj6-nixos-16.09.1698.2da8a5d/nixos/pkgs/applications/audio/spotify/default.nix:5:1`
<gchristensen>
clever: what if a heavy job could declare it requires 2 build slots?
<clever>
gchristensen: that works at the nix level
<gchristensen>
oh?
<clever>
gchristensen: but i have also seen some packages that have the problem at the make level
<clever>
which would require patching every single makefile, rather then just flagging one derivation as being fat
<gchristensen>
sorry, I'm thinking back to the qemu tests
<clever>
yeah, no make involved in qemu tests
<gchristensen>
to reduce the load on the host declare it needs 2 build slots instead of one, so you get at worst 1/2 the cores busy
<clever>
yeah
<gchristensen>
(I'm not thinking about patching make at all)
<clever>
there is already a nixos-test feature that a slave must have, for hydra to even push qemu jobs to it
<clever>
but you would need to heavily reduce the concurrent jobs allowed on that machine
<gchristensen>
yeah
<clever>
and then hydra will waste its power when not doing testing
<gchristensen>
but if the machine could run 1drv per core unless it is marked as heavy
<clever>
there is also the tricky problem of balancing nix -j4 vs make -j4
<gchristensen>
I think I'm not communicating my thought very well
<clever>
i do see how marking a derivation as being worth 2 jobs would help lessen the load at the nix level
<clever>
so if you use nix's -j4 (or hydra builds 4 jobs per slave), it will only do 2 tests max
<gchristensen>
yeah
<gchristensen>
I've had other ideas about dynamic scheduling but they're far too complicated to be implemented right now
<gchristensen>
like TAGS :)
<qmm>
joehh: it isn't the sandbox. it's the same problem in the sandbox
<qmm>
this is first time i tried the sandbox
<qmm>
joehh: putting -main-is T in the executable section for T did the trick
<qmm>
clever ^
<qmm>
one more question, is there a way to tell nix to ignore deps if there are none and just run cabal build?
<qmm>
if there are no new dependencies*
gm152 has quit [(Quit: WeeChat 1.9)]
<qmm>
this could give me a slight boost in speed
<qmm>
development speed that is
<ben>
I can do rec { x = 1; y = x; }, when i have rec { "#" = 1; what do i put as y = ???; } to refer to that field?
b has joined #nixos
justelex_ has joined #nixos
<schoppenhauer>
manveru: ok. if you have more info, or need testing or any help, feel free to highlight or query me :3
<manveru>
schoppenhauer: yeah... it's actually not as easy as i thought, because bundler is adamant about using relative paths for the lockfile even if you specify absolute ones in the gemfile
<manveru>
and that's just not something nix does well
<manveru>
i don't think there's a better way than RUBYLIB for now if you want something working fast
Zer000 has joined #nixos
<schoppenhauer>
manveru: ok. thank you!
<schoppenhauer>
manveru: if there is a new way to do it, please let me know. that is, if your PR is through, etc.
<manveru>
aye
<schoppenhauer>
(Y)
<manveru>
i'll keep working on it a bit ... it's annoying me too :)
joehh has quit [(Ping timeout: 240 seconds)]
<Zer000>
so I just got a bluetooth dongle with the hopes that I can get my phone to play music through my computer, so it thinks the computer is a headset or something. But I quickly realized I have no idea what I'm doing, and besides hardware.bluetooth.enable I'm out of my element.
<Zer000>
The manual has nothing on bluetooth so where should I look?