taktoa has quit [(Remote host closed the connection)]
erictapen has quit [(Ping timeout: 258 seconds)]
andrew has joined #nixos
Rotaerk has joined #nixos
<andrew>
does anyone know if there are issues with the rstudio wrapper on the unstable channel? I keep trying to install a package as outlined in the nixpkgs manual but get a message that 1 dependency couldn't be built
mrcheeks has joined #nixos
<spinus>
andrew: I just installed it
<spinus>
what's the issue?
<andrew>
I just keep getting an error that '1 dependency couldn't be built' when building the rstudio-wrapper derivation
<andrew>
I'm assuming the relevant dependency is 'rstudio-1.1.216.drv' since I see a line saying that the builder for it failed with exit code 1
hellrazo1 has joined #nixos
<spinus>
what nixpkgs do you build against?
ennui` has quit [(Ping timeout: 260 seconds)]
<andrew>
the nixpkgs manual outlines a method for getting packages into the rstudio environment
Supersonic112 has quit [(Disconnected by services)]
Supersonic112_ has joined #nixos
Supersonic112_ is now known as Supersonic112
justelex_ has joined #nixos
vaibhavsagar has joined #nixos
justelex has quit [(Ping timeout: 260 seconds)]
vaibhavsagar has quit [(Ping timeout: 240 seconds)]
mattzers has joined #nixos
mrcheeks has left #nixos ["ERC (IRC client for Emacs 26.0.50)"]
lambdamu has joined #nixos
drakonis has quit [(Remote host closed the connection)]
drakonis has joined #nixos
lambdamu_ has quit [(Ping timeout: 255 seconds)]
lambdael has joined #nixos
ison111 has joined #nixos
vaibhavsagar has joined #nixos
slack1256 has quit [(Remote host closed the connection)]
Sonarpulse has quit [(Ping timeout: 276 seconds)]
eacameron has joined #nixos
s33se_ has joined #nixos
eacamero_ has joined #nixos
eacameron has quit [(Read error: Connection reset by peer)]
s33se has quit [(Ping timeout: 240 seconds)]
ng0_ has joined #nixos
mbrgm has quit [(Ping timeout: 240 seconds)]
endformationage has joined #nixos
mbrgm has joined #nixos
ng0 has quit [(Ping timeout: 248 seconds)]
nh2 has quit [(Quit: Leaving.)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] copumpkin opened pull request #27318: Support High Sierra on Darwin (staging...darwin-high-sierra) https://git.io/vQDDF
NixOS_GitHub has left #nixos []
riot has quit [(Ping timeout: 255 seconds)]
mk-fg has joined #nixos
mk-fg has quit [(Changing host)]
mk-fg has joined #nixos
zeus__ has joined #nixos
zeus_ has quit [(Remote host closed the connection)]
mattzers has quit [(Quit: Konversation terminated!)]
drakonis has quit [(Remote host closed the connection)]
drakonis has joined #nixos
yegortimoshenko has quit [(Quit: ERC (IRC client for Emacs 25.2.1))]
zeus__ has quit [(Ping timeout: 276 seconds)]
bfrog has joined #nixos
<mpcsh>
hey all, is there a way to list out the packages installed system-wide? like nix-env -qa but for declaratively installed packages and their dependencies?
Infinisil has joined #nixos
tikhon has quit [(Ping timeout: 276 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] ElvishJerricco opened pull request #27319: Improved make overridable (master...improved-make-overridable) https://git.io/vQDSL
<cherrybl0ss0m_>
QQ: is anyone expected to be able to run `sh ./nixos/maintainers/scripts/ec2/create-amis.sh` from master in NixOS/nixpkgs.git and succeed? I appear to be getting chroot-related "Operation not permitted" on changing modification time of a file in my AMI image.
vaibhavsagar has quit [(Ping timeout: 276 seconds)]
cpennington has quit [(Remote host closed the connection)]
<cherrybl0ss0m_>
specifically I am wondering if this is not expected if creating a new issue with all the details would be well received? I am happy to offer as much information as possible.
<mpcsh>
what's the deal with command-not-found? does anyone have it working?
jtojnar has quit [(Ping timeout: 260 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] sib-null opened pull request #27321: Add cmakeFlags for auto-type and yubikey (master...master) https://git.io/vQDFb
NixOS_GitHub has left #nixos []
justelex_ has quit [(Ping timeout: 248 seconds)]
ebzzry has quit [(Ping timeout: 248 seconds)]
reinzelmann has joined #nixos
ebzzry has joined #nixos
hamishmack has joined #nixos
celph has joined #nixos
<mog>
hi i installed steam, it then updated and now is saying "Fatal Error: Faild to load steamui.so"
<mog>
is there a known fix
ebzzry has quit [(Ping timeout: 268 seconds)]
<celph>
Disable updates somehow? Multiple package managers in Nix are a bad time
ebzzry has joined #nixos
takle has joined #nixos
<celph>
speaking of, ruby gems aren't playing well with me. I'm not sure if there's a nice workaround, and I'm not feeling too excited about building my dependency tree for Nix by myself
<celph>
I could run it in a VM I guess
takle has quit [(Ping timeout: 248 seconds)]
vaibhavsagar has joined #nixos
odi has joined #nixos
<sphalerite>
celph: no, steam updates itself. All the steam package actually provides is the runtime environment and the steam launcher, which is what the debian package and other packages for Steam also provide.
<sphalerite>
mog: Could you run it in a terminal and see if there's any additional output?
<mog>
i solved it. i had to pass a line for the graphics card
<sphalerite>
In particular, someone else was having issues with steam the other day which were caused by libstdc++ being too old, and I'm not sure anyone's looked at fixing that yet
<sphalerite>
oh good
* Ralith
wonders what that actually means
celph has quit [(Ping timeout: 260 seconds)]
odi has quit [(Quit: Leaving)]
bennofs has joined #nixos
<adisbladis[m]>
How can I use the default builder and just override one environment variable? I have a package that just has a Makefile with the prefix as an environment variable.
marsel has joined #nixos
<sphalerite>
adisbladis[m]: all the attributes passed to mkDerivation are exposed to the builder as environment variables
<sphalerite>
adisbladis[m]: but if the value needs to depend on the output path, you'll need to use bash for it in current versions of nix. In unstable, we have a nice new builtin called placeholder which would allow you to just say PREFIX = placeholder "out"; but in this case you'll need to do something like preConfigure = ''export PREFIX="$out"'';
<bennofs>
vaibhavsagar: are you maintainers.viric?
<vaibhavsagar>
no, I'm not
<vaibhavsagar>
wasn't sure what to put there instead
<vaibhavsagar>
I just copied that from the xbindkeys derivation
<bennofs>
vaibhavsagar: add yourself to <nixpkgs>/lib/maintainers.nix (in the form of githubName = "Your Name <email>" and then put maintainers.vaibhavsagar there
<simpson>
Wow, Wine's stable version is quite old compared to unstable/staging.
HurricaneHarry has quit [(Read error: Connection reset by peer)]
HurricaneHarry has joined #nixos
<bennofs>
simpson: usually, new versions of packages aren't backported
<simpson>
bennofs: I'm talking about wineStable, wineUnstable, and wineStaging, all in the same nixpkgs version.
<bennofs>
ops sry
<bennofs>
it happened to match nixpkgs branch names exactly :D
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bennofs pushed 2 new commits to master: https://git.io/vQDpb
<NixOS_GitHub>
nixpkgs/master b9dfbeb Vaibhav Sagar: xautomation: init at 1.09
<sphalerite>
cmcdragonkai: then it won't be executed correctly if you redo it either
mudri has joined #nixos
bennofs has quit [(Ping timeout: 240 seconds)]
bennofs has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] dmjio opened pull request #27322: miso: init at 0.2.0.0 (master...patch-4) https://git.io/vQyvv
NixOS_GitHub has left #nixos []
arianvp2 has quit [(Quit: arianvp2)]
<canndrew>
I'm trying to add a file to the nix store..
<canndrew>
the hashes that nix-store --add and nix-store --add-fixed give me aren't the same as the ones i get from nix-hash...
<canndrew>
and i'm baffled. it doesn't matter which algo --type i use for nix-hash, I'm using the --base32 option aswell. but I can't get it to return the same hash as nix-store
<canndrew>
ideas?
arianvp2 has joined #nixos
<canndrew>
it's a very large .tar.gz file bt
<canndrew>
*btw
globin has quit [(Quit: WeeChat 1.6)]
globin has joined #nixos
freusque has joined #nixos
<sphalerite[m]>
canndrew: I'm not 100% on this but I believe the hash is affected by the filename as well, and that may or may not be included in the hashing for each of those commands
arianvp2 has quit [(Read error: Connection reset by peer)]
arianvp2 has joined #nixos
Fare has quit [(Quit: Leaving)]
<sphalerite[m]>
I don't see an --add-fixed operation in the nix manual
<canndrew>
sphalerite[m]: I don't know what the difference is. it gives me different hashes to nix-store --add though.
<canndrew>
none of them seem to correspond to the hashes of any of the others.
<sphalerite[m]>
I don't think the hash in the path is meant to be the same as what you get from nix-hash anyway, not sure though
<canndrew>
what I'm trying to do is add a file to the nix store - and get a hash for it that I can use in a derivation.
<sphalerite[m]>
Ah right
<sphalerite[m]>
So for requireFile?
<canndrew>
yep
arianvp2 has quit [(Read error: Connection reset by peer)]
arianvp2 has joined #nixos
<sphalerite[m]>
I think nix-prefetch-url might be able to get the hash you need
<sphalerite[m]>
(Even if the file is stored locally)
freusque has quit [(Quit: WeeChat 1.7.1)]
freusque has joined #nixos
<canndrew>
sphalerite[m]: thanks! I'm trying that now
<canndrew>
sphalerite[m]: huzzah!
<canndrew>
sphalerite[m]: thank you very much :), that was getting frustrating
FRidh has joined #nixos
freusque has quit [(Ping timeout: 276 seconds)]
mudri has quit [(Ping timeout: 246 seconds)]
jgertm has joined #nixos
katyucha has quit [(Killed (verne.freenode.net (Nickname regained by services)))]
katyucha_ is now known as katyucha
katyucha_ has joined #nixos
davidak has joined #nixos
arjen-jonathan has joined #nixos
<cmcdragonkai>
sphalerite[m]: changing the build command does not make it rebuild
Eisfreak1 has joined #nixos
MercurialAlchemi has quit [(Read error: Connection reset by peer)]
endforma1 has quit [(Quit: WeeChat 1.7)]
nh2 has joined #nixos
newhoggy_ has joined #nixos
<bennofs>
cmcdragonkai: nix supports different hash types i believe, and they can be in different formats (there's the usual hex encoding, but also an encoding that uses almost all of the alphabet's characters (i think it's base32)))
<bennofs>
the difference between nix-store --add and nix-store --add-fixed IIRC is that the later allows you to specify the hash algo to use (sha256, ...) and whether to use recursive hashing or not
hydraz has quit [(Ping timeout: 246 seconds)]
jgeerds has joined #nixos
ixxie has joined #nixos
freusque has joined #nixos
newhoggy has quit [(Ping timeout: 276 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bennofs created vaibhavsagar-xautomation (+1 new commit): https://git.io/vQykf
<bennofs>
cmcdragonkai: woops you're right i wanted to reply to canndrew
<bennofs>
cmcdragonkai: you can rebuild a package using by nix-build --check
nh2 has quit [(Ping timeout: 258 seconds)]
nh2 has joined #nixos
erictapen has joined #nixos
ertes-w has joined #nixos
MercurialAlchemi has joined #nixos
ThatDocsLady has joined #nixos
<clever>
domenkozar: i dont remember which package it was on, but it was fixed by just always passing some enable flag to configure, i discovered it when reading the comments in that nix file
<clever>
domenkozar: but its an odd bug to keep in mind, which may crop up again, given that cabal behaves differently if "ghc" is present in the --datadir or not
__Sander__ has joined #nixos
<domenkozar>
>.<
<clever>
if your --datadir contains ghc, the docs appear inside it
<clever>
but if your datadir doesnt contain ghc, the docs wind up in --prefix
<clever>
giving both --datadir and --docdir appears to prevent the issue
erictapen has quit [(Ping timeout: 240 seconds)]
<bennofs>
clever: what -.-
erictapen has joined #nixos
dbmikus has quit [(Ping timeout: 246 seconds)]
<vaibhavsagar>
I thought the haskell builder didn't use cabal?
<vaibhavsagar>
so this shouldn't be an issue, right?
Wizek_ has joined #nixos
bennofs has quit [(Ping timeout: 240 seconds)]
<cmcdragonkai>
bennofs[m]: I'm trying to rerun nix-shell not nix-build
<clever>
vaibhavsagar: i believe it runs Setup.hs, which imports cabal
tmaekawa has quit [(Read error: Connection reset by peer)]
tmaekawa has joined #nixos
ixxie has quit [(Quit: Lost terminal)]
<nh2>
qknight: ping
freusque has joined #nixos
athan has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
athan has joined #nixos
dbmikus has joined #nixos
cpennington has quit [(Remote host closed the connection)]
hodapp has quit [(Ping timeout: 260 seconds)]
newhoggy has quit [(Remote host closed the connection)]
newhoggy has joined #nixos
ixxie has joined #nixos
ebopp has joined #nixos
newhoggy has quit [(Ping timeout: 246 seconds)]
pietranera has quit [(Ping timeout: 255 seconds)]
Wizek_ has joined #nixos
gilligan_ has joined #nixos
ixxie has quit [(Ping timeout: 240 seconds)]
orivej has quit [(Ping timeout: 255 seconds)]
<gilligan_>
hi
<gilligan_>
i have a question on using nixos-run-vm - how do I access a service running inside the vm? Is it reachable at all? How do I find out the ip?
cynede has left #nixos ["Leaving"]
indi_ has quit [(Remote host closed the connection)]
arjen-jonathan has quit [(Ping timeout: 255 seconds)]
<gilligan_>
throwing that in our internal slack #nix channel
ixxie has quit [(Quit: leaving)]
Eisfreak1 has joined #nixos
ixxie has joined #nixos
noam__ has joined #nixos
Jackneilll has joined #nixos
eacameron has joined #nixos
ixxie has quit [(Client Quit)]
ixxie has joined #nixos
pchiusano_ has joined #nixos
ThatDocsLady is now known as ThatDocsLady_nom
r0bby_ has joined #nixos
noam_ has quit [(Read error: Connection reset by peer)]
nhill1 has joined #nixos
indi_ has joined #nixos
dmj`_ has joined #nixos
dmj` has quit [(Ping timeout: 246 seconds)]
dredozubov has quit [(Ping timeout: 246 seconds)]
r0bby has quit [(Read error: Connection reset by peer)]
pchiusano has quit [(Read error: Connection reset by peer)]
teozkr has quit [(Ping timeout: 246 seconds)]
bgamari has quit [(Ping timeout: 246 seconds)]
Twey has quit [(Ping timeout: 246 seconds)]
dbmikus has quit [(Ping timeout: 246 seconds)]
nhill has quit [(Ping timeout: 246 seconds)]
aristid2 has quit [(Ping timeout: 246 seconds)]
hydraz has quit [(Ping timeout: 246 seconds)]
Jackneill has quit [(Ping timeout: 246 seconds)]
ris has quit [(Remote host closed the connection)]
TheAppleMan has quit [(Ping timeout: 246 seconds)]
WilliButz has quit [(Ping timeout: 246 seconds)]
teozkr_ has joined #nixos
dmj`_ is now known as dmj`
bgamari- has joined #nixos
pchiusano_ is now known as pchiusano
Willi_Butz has joined #nixos
aristid2_ has joined #nixos
r0bby_ is now known as r0bby
cmr has quit [(Ping timeout: 248 seconds)]
dbmikus has joined #nixos
dredozub- has joined #nixos
cmr has joined #nixos
cmr has quit [(Changing host)]
cmr has joined #nixos
k2s has joined #nixos
Twey has joined #nixos
DerGuteMoritz has quit [(Ping timeout: 246 seconds)]
hodapp has joined #nixos
bennofs has quit [(Ping timeout: 260 seconds)]
hydraz has joined #nixos
hydraz has quit [(Changing host)]
hydraz has joined #nixos
bennofs has joined #nixos
freusque has quit [(Quit: WeeChat 1.7.1)]
gilligan_ has quit [(Quit: Leaving)]
tvon has joined #nixos
freusque has joined #nixos
freusque has quit [(Client Quit)]
k2s has quit [(Remote host closed the connection)]
freusque has joined #nixos
freusque has quit [(Client Quit)]
freusque has joined #nixos
<gchristensen>
I did it, everyone, I managed to explain to a coworker how all-packages.nix and the module system works without using the words "fixed point combinator" or any other scary functional-sounding words
<domenkozar>
\o/
<domenkozar>
I hope you got a cookie :)
<gchristensen>
I might get better than a cookie :P
<domenkozar>
a raise to pay for the cookie? :D
freusque has quit [(Client Quit)]
<gchristensen>
hehe
m0rphism has quit [(Quit: WeeChat 1.8)]
<domenkozar>
gchristensen: now make a youtube video of that
<gchristensen>
oh boy
<domenkozar>
:D
<ikwildrpepper>
movie or it didn't happen
<domenkozar>
NixOS the movie
<domenkozar>
sounds about right
<gchristensen>
I don't know, he's the kind of guy who goes and reads all the docs before starting, so I was able to refer to sections in the docs he'd already read through a time or two
<domenkozar>
lacks cryptocurrency hype though
<gchristensen>
so when I talked about merge characteristics of different types he already knew about merging commas / lines / etc. (though admittedly didn't really grok what that meant, until I showed him this example: https://github.com/NixOS/nixpkgs/blob/master/nixos/modules/services/databases/cassandra.nix#L416-L421 and he was surprised it didn't completely replace the existing list of allowedTCPPorts)
cpennington has joined #nixos
<joepie91>
gchristensen: you should make it an article
<joepie91>
:P
<gchristensen>
I already have 3 fixes to make to the docs >.>
<joepie91>
gchristensen: just write it as a separate article first, and worry about how to merge it into the docs later?
<joepie91>
stream-of-consciousness type of thing
<domenkozar>
I have a whole overflown trello column for docs fixes
<joepie91>
domenkozar: is "add the missing operators" in it? :P
<domenkozar>
it's many things
<gchristensen>
nah, the doc fixes are: 1. "collon" -> "colon". 2. document how (ie: they don't) bool, int, path, and package merge. 3. provide a real example of what it means for them to merge
<joepie91>
gchristensen: would you know, I just fixed that one yesterday
<joepie91>
:p
<gilligan_>
if I execute some make-test.nix based VM via run-nixos-vm, how do I access a service (in my case a web service) running in the vm? What's the ip? Do I have to pass some networking options on the command line or something like that?
tmaekawa has quit [(Quit: tmaekawa)]
<gchristensen>
cool
<gchristensen>
syntax highlight entirely by regex is sort of hell :(
<joepie91>
gchristensen: though I just realized I'm missing the assert keyword still
<joepie91>
oh yes
<joepie91>
gchristensen: thing is, my PEG parser had the same bug because it's also sort of linear
<joepie91>
like the regex-monstrosities used by textmate, atom, github, etc.
<joepie91>
(Atom's syntax highlighter is still broken btw, I need to get around to submitting a patch for that some day)
mudri has joined #nixos
<joepie91>
gchristensen: anyway, I've had all kinds of fun writing this parser. not in the least working around PEG parsers' inability to work with operator precedence in a maintainable way, and various ambiguities in Nix as a language, and undocumented cases, etc.
<joepie91>
it doesn't help that there's no language spec :P
<gchristensen>
IIRC there used to be
<domenkozar>
github used to use another JS lib for highlighting, I have fixed most of the issues there
<joepie91>
so it's pretty much trying to recreate reference implementation behaviour
<joepie91>
gchristensen: I haven't found a recent one anyway
arianvp2 has quit [(Read error: Connection reset by peer)]
nix-gsc-io`bot has quit [(Client Quit)]
arianvp2 has joined #nixos
<joepie91>
anyway, learning quite a bit about Nix as a language in the process, heh
<joepie91>
as well as how stacktraces are generated
<joepie91>
... and how they are NOT generated...
<gilligan_>
gchristensen, `QEMU_NET_OPTS="hostfwd=tcp::2223-:22"` So I assume I can interpret that as forward traffic on port 2223 on the *host* to port 22 inside the VM right? Meaning I could just do QEMU_NET_OPS="hostfwd=tcp::8080-:8080" if I wanted to make what is running on port 8080 in the VM available on my host system?
arjen-jonathan has joined #nixos
<joepie91>
the lack of attribute lookups in the stacktraces is a pretty big issue, it can make it very tricky to tie together stacktrace frames
<gchristensen>
gilligan_: right
<gilligan_>
gchristensen, thanks ;)
<gchristensen>
you're welcome :)
<joepie91>
err
<joepie91>
lack of identifier lookups *
<joepie91>
because it's not always clear how it jumps from one function call or attribute lookup to another
eacameron has quit [(Remote host closed the connection)]
<joepie91>
so if you do `foo.bar` then it'll show the 'bar' attribute lookup in the stacktrace but if you do `foo` then it won't show up in the stack at all
<gchristensen>
interesting
<joepie91>
it just kinda transparently references whatever it points at
<joepie91>
until you hit another function call or attribute lookup
<gchristensen>
do you know C++? :) can you make improvements to the tracing?
<joepie91>
sec, I can produce an example of this
<joepie91>
nope!
<joepie91>
but I have a testcase
<joepie91>
:p
<gchristensen>
-> github.com/nixos/nix/issues/new
siel has quit [(Ping timeout: 246 seconds)]
vaibhavsagar has quit [(Ping timeout: 246 seconds)]
<joepie91>
(I can produce a simpler testcase as well, this was a testcase to generally understand how stacktraces are produced in Nix, how callback-type constructs were displayed, etc.)
<gchristensen>
ahh cool
<joepie91>
original output and snippet-including trace respectively
<joepie91>
gchristensen: I'm developing a habit of parsing Nix output and reformatting it with custom tools, heh. this is something I'd like to eventually see in Nix itself, but I don't know a lick of C++ and can't justify learning it for this either :P
<joepie91>
gchristensen: but yeah, I should probably simplify the testcase a bit and produce an issue ticket for it
<joepie91>
and hope that somebody else can pick it up
<domenkozar>
if we had something that preserves comments and can parse/print
arianvp2 has joined #nixos
<gchristensen>
I'd change my personal coding style to match it, happily
<domenkozar>
I'd be so happy since we can start doing declarative docs in the source code
<gchristensen>
domenkozar: I've shown you my installer docs, right...?
<domenkozar>
gchristensen: yeah but I mean more like for lib/* functions
<gchristensen>
yeah also yes
arianvp2 has quit [(Client Quit)]
<joepie91>
domenkozar: my parser does comments but I'm not entirely sure about the correct AST representation yet
<joepie91>
I should also warn that I'm not a particular fan of in-code docs :P
siel has joined #nixos
siel has quit [(Changing host)]
siel has joined #nixos
<domenkozar>
the correct representation is when you parse nixpkgs/* and write it back and there is no modifications
<domenkozar>
without cheating :P
<gilligan_>
domenkozar, gchristensen i like the qemu net options approach - works fine for me
<joepie91>
in my experience they tend to a 'worst of both worlds' scenario where it becomes harder to maintain code because the docs get in the way, and there's an incentive to write too-brief docs because otherwise it'd get too much in the way of code
<joepie91>
so I'd much rather see external documentation that is tied to sections in the code through metadata, for example
<gchristensen>
domenkozar: to the entire codebase, or just comments?
<domenkozar>
gchristensen: I mean, for starters just having comments in AST would be enough
<joepie91>
cc domenkozar gchristensen on the above regarding in-code docs ^
<domenkozar>
so docs can be extracted
<gchristensen>
aye
<joepie91>
(these are disincentives that are very frequently overlooked)
<domenkozar>
joepie91: why do docs get in the way? modifying lib/* MUST mean modifying docs
<domenkozar>
that's the point
<joepie91>
domenkozar: but it doesn't do that in practice
orivej has joined #nixos
<gchristensen>
lib/* has really short, simple functions
<gchristensen>
almost no point in changing most of them once they're written
<joepie91>
domenkozar: this is the problem with in-code docs; it sounds great on paper but it doesn't work the same way in practice, and in-code docs go out of date just as much as external docs do, except they make it harder to follow the structure of the code as well
reinzelmann has quit [(Quit: Leaving)]
<gilligan_>
has anyone here used google reomte desktop? Wonder if packaging it is worth the effort..
<joepie91>
I have yet to see a single project that provably benefited from in-code docs
<joepie91>
it seems to never go beyond theoretical claims
<domenkozar>
joepie91: I have a different experience, but I'm coming from Python culture where you're coming from Haskell? :)
<gchristensen>
ah
<joepie91>
domenkozar: I'm coming from the PHP/Python/JS corner
<domenkozar>
if I get a PR that has a comment above, I'll read the docs while reviewing
<domenkozar>
if I have to go though mental hop to open the dreadful manual
<domenkozar>
it won't happen
<joepie91>
and in my experience those projects that have consistent docs changes, are projects that would have had them *anyway* even without in-code docs for the mere fact that the maintainers are diligent about docs maintenance in the first place
<joepie91>
domenkozar: that seems unrelated to in-code docs
<gchristensen>
^
<gchristensen>
in-code docs certainly won't be a panacea
<joepie91>
there are many possible strategies for tying code and docs together
<joepie91>
that do not necessarily require in-code docs
<joepie91>
the simplest of which is "every PR must include the appropriate docs changes out the gate"
<domenkozar>
then there's a question of comments vs docbook
<joepie91>
so that it all shows up in the same diff
<joepie91>
domenkozar: not sure what the argument there is. there are many, many options for documentation formats
<domenkozar>
there are many, but we have docbook
<joepie91>
and this isn't even addressing the problem of it being difficult to express documentation structure in in-code docs, because optimal code structure almost *never* matches optimal documentation structure
<joepie91>
which is why nearly every project that has in-code docs, has difficult-to-navigate generated docs with no obvious coherence or structure to them
<domenkozar>
oh it's a function reference
<domenkozar>
not a tutorial
<domenkozar>
I think the difference is clear
<joepie91>
sure, but function references aren't supposed to be any less contextal
<joepie91>
contextual*
<joepie91>
you still need interlinking between related things, a logical sorting of things
<joepie91>
things like that
<domenkozar>
good documentation tools have that
<domenkozar>
like sphinx
<joepie91>
listing a function signature and explaining the arguments is only half the job
<domenkozar>
also for in-code docs
<gchristensen>
joepie91: that is an annoyance of mine with many haskell docs, is they assume signatures is magically enough.
<joepie91>
domenkozar: sure, you *can* have it - the problem is that it's more difficult to actually maintain in-code docs with it, since you shoehorn everything into the code structure by default. and I can't say that sphinx convinces me of this being a viable approach, since nearly every bit of sphinx documentation I've seen has been poorly organized
<domenkozar>
joepie91: if you want to see how docbook is a problem: search our repo for how many unique contributors have added docbook changes
<domenkozar>
and then how many have asked how to contribute with docbook
<gchristensen>
oh no the docbook debate
<joepie91>
domenkozar: sure, but that's a docbook problem, not an external-docs problem
* gchristensen
runs away
<joepie91>
(the equivalent debate in Python-land appears to be rst)
<domenkozar>
well I guess at the end the tools don't matter, someone just needs to write all the docs :)
<gchristensen>
yeah exactly
<gchristensen>
nothing matters if nobody writes them
<domenkozar>
and both solutions can lead to that
<joepie91>
well... yes, but it's also partly policy. for example, institute a policy that every new feature PR must include at least a mention in the docs of what it's for and how it works
<joepie91>
wherever those docs are
<gchristensen>
or first, develop a culture of asking for docs
<domenkozar>
that's the problem
<joepie91>
people aren't going to just magically write docs, and people who are dedicated to fixing up the docs still need *some* starting point
<domenkozar>
you can't ask people to document things if they don't know where
<domenkozar>
if you have a comment above, that's easy to find
<domenkozar>
usually
<joepie91>
it's also easy to find if it exists in the same repo and is referenced :)
<domenkozar>
it's like sorting garbage
<joepie91>
from eg. the README or CONTRIBUTING
<domenkozar>
you need 3 boxes next to each other
<domenkozar>
not one that says the other 2 are around the corner
<avn>
joepie91: don't add too much policies, don't switch nix to debian
<gchristensen>
nice to know what you think of our code, domenkozar :) </joking>
<domenkozar>
if Nix had in-code docs it would be by a margin better
<domenkozar>
then it would be "read the code with comments about functions"
<avn>
domenkozar: anyway, big codebase usually readed by grep ;)
<joepie91>
domenkozar: the problem is representing this as a feature of in-code docs, when in reality it's a feature of any docs at all anywhere in a place people can find it
<joepie91>
in-code docs do not have an advantage here over any other easily-findable documentation source
<domenkozar>
joepie91: exactly my point, we have 4 manuals
<gchristensen>
fwiw I agree with a lot of the things both of you are saying
<joepie91>
(and easily-modifiable
<joepie91>
)*
<domenkozar>
and noone is capable of even knowing which manual the docs go to
<domenkozar>
so what section is not even a debate :D
<gchristensen>
domenkozar: what would you think about embedding the nix docs in to nixpkgs, and nix & nixpkgs in to the nixos manual? :)
<joepie91>
avn: I'm not a fan of unnecessary policies, but I'm even less a fan of missing necessary policies -- and well, clearly, the current documentation strategy isn't working
<domenkozar>
gchristensen: I don't think it's that easy
<domenkozar>
I think we should start with who the target audience is
<domenkozar>
developers read reference
<gchristensen>
didn't say it was
<domenkozar>
I think that belongs in code
<domenkozar>
then users want guided docs
Ivanych has quit [(Ping timeout: 255 seconds)]
<joepie91>
domenkozar: it's not that simple. developers read, *amongst other things*, reference
<joepie91>
you still need a "how do I get from problem to solution" path
<joepie91>
and that's not what a reference is
<domenkozar>
yeah but that's usually like 90% vs 10%
<domenkozar>
so completely different audience than users
<joepie91>
"how do I get from problem to solution" can be something as simple as a sensible categorization of available features so that somebody can look in the right category at least and cut down on the search space
<domenkozar>
where reference is like 2% and guides is 98%
<joepie91>
domenkozar: I would strongly disagree with 90% reference.
<joepie91>
that significantly understates the significance of problem-to-solution docs
<gchristensen>
maybe we'd have something to learn from how Perl and PHP have their docs
<joepie91>
domenkozar: that is a part of it, yes. but that still only covers a number of concepts and questions
<joepie91>
categories, like I mentioned above, are an important aspect
<domenkozar>
well that's the problem of writing docs :D
eacameron has quit [(Ping timeout: 260 seconds)]
<joepie91>
there's a LOT of stuff in nixpkgs, and I have no idea where to start looking to solve a given problem
<gchristensen>
search.nix.gsc.io
<joepie91>
if I can partition out the search space into "definitely not relevant", "maybe relevant" and "that looks like what I want!" helps a lot
<gchristensen>
;)
<joepie91>
that helps a lot *
<joepie91>
even for the weirder things
<domenkozar>
joepie91: that has to do with the policy what should be documented and what shouldn't
<domenkozar>
then it's easy
<joepie91>
for example, callPackage_i686 and a few other i686 things could be referenced from a "32-bits stuff" category and now when I want to package a 32-bits thing I know precisely which things to look at
<domenkozar>
things like pkgs.symlinkJoin
<joepie91>
hell, work with tags even
<joepie91>
(was it callPackage_i686? I don't even remember)(
<domenkozar>
but by the end of the day it boils down to who contributes to docs
<domenkozar>
we had some great editors of wiki
<domenkozar>
and we just shut it down
pietranera has quit [(Ping timeout: 240 seconds)]
<joepie91>
domenkozar: I'd love to contribute to docs but I have no starting point
<domenkozar>
that's a very bad message to those who edited wiki
<gchristensen>
my understand is ThatDocsLady was a great contributor
<joepie91>
most things don't even have a one-line phrase describing what it is
<joepie91>
and no way for me to understand it better
<joepie91>
I'm talking about nixpkgs functions, for example
<gchristensen>
those anchor IDs are terrible
<joepie91>
why do some of them exist? not a clue
<joepie91>
who uses them? no idea
<domenkozar>
nixpkgs doesn't have a problem unorganized docs
<joepie91>
why does package X use it? who knows
<domenkozar>
but the very lack of it :D
<joepie91>
it's both, lack of docs and lack of organization of those that exist
<joepie91>
like the manual format I've complained about before
<joepie91>
:p
<joepie91>
Nix isn't a linear thing, so a linear format is not useful
<domenkozar>
I disagree, look at how many different devs contributed to nixpkgs
<joepie91>
look at how many didn't
<domenkozar>
manual
<joepie91>
that's the better metric
<domenkozar>
it's the same metric :)
pietranera has joined #nixos
<joepie91>
domenkozar: so, here's the thing. I've been talking to a number of people about NixOS now, some of them have tried it out. almost universally one of the main criticisms was "I feel lost, I have no idea how to do anything and the manual is hard to navigate"
<joepie91>
this may not be easily recognizable if you're reasonably familiar with the knobs and twiddles, but it's a serious issue
<joepie91>
it's just not a useful format for very explainable reasons
<copumpkin>
yeah, same
<joepie91>
actually, you know what
<joepie91>
I'll gist my in-progress rant
<copumpkin>
been helping a friend get up to speed on contributing to nixpkgs
mudri has joined #nixos
<copumpkin>
and he feels very lost especially about functions in lib and when docs should be in nix vs nixpkgs
yegortimoshenko has joined #nixos
<yegortimoshenko>
is that ok or should i open a github issue?
<yegortimoshenko>
collision between ‘/nix/store/6y58sc6lrsa49rsfjc0dd8pcr19g0495-texlive-combined-small-2016/bin/afm2pl’ and ‘/nix/store/6mnr75c0xd2yggici5gjivmqlc4xaxll-texlive-combined-tetex-2016/bin/afm2pl’; use ‘nix-env --set-flag priority NUMBER PKGNAME’ to change the priority of one of the conflicting packages
<yegortimoshenko>
<copumpkin>
there's just a ton of implicit knowledge that we don't think twice about that is super unfamiliar to anyone who hasn't been doing it for a while
<gchristensen>
the nix nixpkgs nixos distinction is very clear to me, but took a long time to understand how the distinction was made
<joepie91>
domenkozar: gchristensen: so, to be clear, this is an *unfinished* rant that I'll eventually finish and post on the issue tracker, but it gives an idea of a number of the issues I'm thinking of: https://gist.github.com/joepie91/5232c8f1e75a8f54367e5dfcfd573726
<joepie91>
(and it's already pretty long)
<gchristensen>
joepie91: I think I've seen this rant of yours before ... :)
<copumpkin>
well, the specific case here is that you'll often find that lib re-exports builtins.foo, but doesn't provide docs for it
<domenkozar>
so nixpkgs doc commits unique by author
<domenkozar>
20 Robin Gloster
<joepie91>
gchristensen: I've spoken about it but not posted it
Eisfreak1 has quit [(Ping timeout: 240 seconds)]
<domenkozar>
35 Frederik Rietdijk
<joepie91>
iirc
<domenkozar>
35 Vladimír Čunát
<copumpkin>
so if you find certain re-exported values, you now need to hop over to the nix manual
<domenkozar>
37 Domen Kožar
<domenkozar>
94 Eelco Dolstra
<copumpkin>
and hope it's a builtin
<domenkozar>
you get down to 20 commits looking by top5
<domenkozar>
so the short story is: eelco made up the initial structure, the rest just added a bit of things
<joepie91>
gchristensen: domenkozar: also, mind that that gist is something I've been working on for a while, off-and-on, so some things may have actually gone outdated by now :)
<domenkozar>
end of story :)
<gchristensen>
joepie91: understood
<joepie91>
(including talking about "the wiki is GOING to be shut down")
<copumpkin>
it's sort of not though
<gchristensen>
domenkozar: one issue I found with the structure of the docs, was in order to expand a section I had to rewrite everything around it. it was a bit intimidating, and I don't feel intimidated by much :P
<copumpkin>
we can keep repeating "people should do more docs"
<copumpkin>
but that clearly hasn't worked so far, so I'm wondering how we can smooth out the process so we don't need to rely on the goodness of folks' hearts
<domenkozar>
I recommend a hackathon
<gchristensen>
yes!
<copumpkin>
dockathon
<gchristensen>
we could declare Tuesdays are Doc-days
<domenkozar>
yes
<domenkozar>
I brought ThatDocsLady on last hackathon
<copumpkin>
#1 thing would be to harvest docs out of comments under lib/
<domenkozar>
but noone was interested to do docs
<copumpkin>
and a few other places
<domenkozar>
so I suggest docs-exclusive hackathon
<domenkozar>
and the good news is, nixcon is around the corner!
<domenkozar>
gchristensen: +1
<copumpkin>
you understand why nobody wants to do docs though, right? :P
<gchristensen>
they're not fun
<gchristensen>
they're never fun
<copumpkin>
it's like a much longer-term investment than most of the investments we make
<nh2>
I personally find missing docs the biggest problem of the nix* ecosystem. Much of the stuff I need to get my things to work is in the heads of the people in #nixos, not written down. I find the format of the generated docs (one huge manual page) OK, and even very searchable. But I'd definitely have liked an intro page to the docs that explains me when I have to look in the nix manual, when in the nixpkgs manual, when in the nixos manual.
<joepie91>
writing docs is fun, proofreading and fixing them is not
<joepie91>
:p
<niksnut>
domenkozar: if noone was interested, then probably not many people will show up for a docs hackathon ;-)
<niksnut>
(I would come though)
<copumpkin>
e.g., I could fix Darwin 10.13 in my limited spare time or I could write some docs on how to fix it
<copumpkin>
and hope someone comes along and reads them etc.
<copumpkin>
anyway, it's tricky
<joepie91>
copumpkin: I'm more talking about 'implement the fix and write at least minimal docs about what you fixed and how'
<gchristensen>
security patching wasn't "fun" either IMO, but we built up people interested, who would happily take a few tasks and spend an hour handling them
<copumpkin>
I feel like there has to be someone who really cares about docs pushing the rest of the community in the right direction
<domenkozar>
my experience in communities is actually really bad news: projects with good docs have a single person being the superman
<joepie91>
(hi!)
<gchristensen>
^
<joepie91>
:P
arianvp2 has joined #nixos
<gchristensen>
(my ^ was to copumpkin)
<nh2>
I think requiring PRs that add features to include docs would be a very good idea. I as a contributor am always very happy to add docs as long as the project can point me to where to add them.
<domenkozar>
copumpkin: we wrote the same thing :)
<gchristensen>
joepie91 for doczar
<copumpkin>
constant reminders, setting up processes to facilitate docs, etc.
<Ankhers>
I'm actually very interested in writing docs for nix. I have been trying to use it for ~3-4 years, but only recently got traction. I would rather write the docs and have others have an easier time than myself.
<joepie91>
seriously though, part of the reason for that rant I linked is to do precisely that
<copumpkin>
:D
<joepie91>
unfortunately writing docs about writing docs is even *harder* than writing docs
<joepie91>
higher-order docs writing...
<gchristensen>
lol
<gchristensen>
it might pay off, though
<joepie91>
sure, I just don't like how long it takes to get it right, heh
<joepie91>
hence why I work on it on-and-off, to not get too frustrated
<yegortimoshenko>
i am a relatively new nix/nixos user (several days) and i didn't really need docs besides the manual and irc help (thanks, everyone). so what docs are missing, exactly?
<gchristensen>
if you can list out a few things people can help with, I'll help
freusque has quit [(Quit: WeeChat 1.7.1)]
<joepie91>
[15:54] <nh2> I think requiring PRs that add features to include docs would be a very good idea. I as a contributor am always very happy to add docs as long as the project can point me to where to add them.
<joepie91>
just wanting to reemphasize this
<domenkozar>
so I think we need someone to organize docs efforts
<joepie91>
the CONTRIBUTING.md does not currently include instructions on how to submit docs
<joepie91>
or where
<joepie91>
or when
<joepie91>
this is probably the single easiest win
<domenkozar>
joepie91: nh2: when I still had time to reviews PRs I always requested docs/chanelogs
ixxie has quit [(Quit: Lost terminal)]
<domenkozar>
and usualy people did do them, once I told them where
<joepie91>
it's not going to fix all of it, but it's going to at least give a pointer to those who *want* to write docs but have no idea where to do it
<joepie91>
domenkozar: right - I'm saying that should probably be in the CONTRIBUTING.md upfront :)
<nh2>
yegortimoshenko: with IRC help you can achieve *everything* with nix (that's also how I do it). But that's not a good mode of working. Sometimes the great helpers sleep. Or get a new job. Or some other thing happens. Also, if IRC has to answer the easy questions when HTML docs could do instead, that takes away time from the experts to help with the hard questions
<gchristensen>
nh2: or get burned out answering questions
<joepie91>
especially that
<joepie91>
(this is why I've accumulated a massive pile of gists from questions in #Node.js - just to not have to answer the same questions every time)
<gchristensen>
I have a few like that
orivej has quit [(Ping timeout: 246 seconds)]
<joepie91>
they're not necessarily even well-written gists, and sometimes they are just code examples
<domenkozar>
niksnut: noted :)
<joepie91>
but it's *something* that covers *most* cases
<joepie91>
and it takes maybe 5 minutes to turn an answer into a gist, even if it's just an IRC log
tvon has joined #nixos
<joepie91>
which easily saves 30+ minutes in the future for common questions, if not far more
<copumpkin>
we need a strong SO community :P
<copumpkin>
clearly the only answer
<domenkozar>
at previous OSS projects I worked with that were bigger there was the docs manager
<gchristensen>
I _really_ don't like answering on SO, but I agree
<domenkozar>
or hero
<joepie91>
SO answers only get you so far, the biggest issue is the answers going outdated
<copumpkin>
gchristensen: I love reading SO answers :P
<joepie91>
and it being hard to solve poor-quality answers that have been selected as the answer
<gchristensen>
me too
<nh2>
domenkozar: I think it's good that you did that (ask for docs on PR reviews), but if it's "official policy", it'll be much more effective than when individuals ask for it
<gchristensen>
joepie91: you can be in charge of our docs if you decide to
<niksnut>
joepie91: it's funny you consider the single-page format a problem. We had a multi-page structure, and people complained about that for years.
<gchristensen>
joepie91: that is pretty much how things work, someone starts to do it (doesn't ask permission) and pushes it
<joepie91>
if you have a multi-page linear manual, sure, that's less practical than a single-page linear manual
<joepie91>
but the problem there is "linear manual", not "multi-page"
<gchristensen>
I like the single page, but cross-links would be good
<nh2>
first thing I would definitely add `[ ] Added/updated relevant docs` to that github PR template thing that appears when I click to submit a PR
<niksnut>
it's not linear
<niksnut>
the sections are largely independent
<joepie91>
right now? the manual is absolutely linear
<domenkozar>
my biggest problem with current docs is docbook, but I feel like a lone wolf
<joepie91>
no, they're really not
<gchristensen>
nh2: send taht PR
<joepie91>
they talk about different topics, yes, but they still assume previous knowledge, don't interlink to other concepts consistently
<niksnut>
the NixOS manual, specifically
<joepie91>
niksnut: did you see my rant link above, btw?
freusque has joined #nixos
<niksnut>
yes
<joepie91>
ok
<joepie91>
(also, as far as I can tell, the main complaint in that issue thread isn't "it's multiple pages", but rather "it's not searchable")
<joepie91>
gchristensen: I'll very likely end up building better documentation infrastructure eventually(tm), but my available time is unfortunately very limited/inconsistent
<joepie91>
gchristensen: freelancer etc.
lukego has joined #nixos
<gchristensen>
joepie91: come up with tasks to move the project along
<gchristensen>
joepie91: I'm a very skilled codemonkey
<Ankhers>
I will be very glad to help with the docs effort.
eacameron has joined #nixos
<joepie91>
gchristensen: the main issue I run into with the topic of documentation is that whenever you bring up "it should be done so-and-so", you end up getting lost in weeks/months of discussion about how "what we're currently doing isn't really that bad", and at that point it's easier to just wait for a few months until you have the time to implement a rough version of it and show it to people and then people aren't difficult about it
<domenkozar>
yes, we need someone to take over the matter
<domenkozar>
I'm so happy to help
<domenkozar>
just when I think about organizing this I want some icecream
<domenkozar>
and sit
<gchristensen>
domenkozar: :)
<joepie91>
gchristensen: I'd be happy to coordinate better documentation as time allows, but I want a rough example of what I have in mind implemented first, or I'll end up wasting lots of time trying to convince people that such and such should be improved in such and such way
<lukego>
Just wondering how far you can push impure metaprogramming in Nix... is there a way for a Nix expression to enumerate files in a directory? I'm thinking about dynamically creating a list of derivations for the files that happen to exist. But I'm wondering if there is some strict ordering constraint e.g. Nix code can't look at the output of one derivation for creating a new derivation...?
<copumpkin>
domenkozar: running your command for the installer says "package ‘libseccomp-2.3.2’ in /nix/store/a6x68pax3l84h3ana393r1g4yj5xawb8-nixpkgs-17.09pre110532.50831d543d/nixpkgs/pkgs/development/libraries/libseccomp/default.nix:22 is not supported on ‘x86_64-darwin’, refusing to evaluate.
<joepie91>
gchristensen: (fwiw, this problem extends to many of the 'softer' factors in software dev, not just documentation - UX has the same issue)
<gchristensen>
definitely understood
<joepie91>
Ankhers: noted :)
mudri has quit [(Ping timeout: 276 seconds)]
<gchristensen>
but if there is anything concrete, or less than concrete that we could explore, make some tickets and let us know
indi_ has joined #nixos
<lukego>
oh now I see 'builtins.readDir'. wow. Could be that it's doable :)
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] nh2 opened pull request #27331: PULL_REQUEST_TEMPLATE: Ask for docs (master...ask-for-docs-in-pull-request-template) https://git.io/vQyMX
NixOS_GitHub has left #nixos []
<copumpkin>
niksnut: is tarball in the release.nix of nix expected to work on darwin?
<copumpkin>
I see an unconditional libseccomp
<niksnut>
copumpkin: no
<gchristensen>
copumpkin: you can build one though..?
<Ankhers>
joepie91: Even if you want help with the prototype, ping me and I will do what I can.
<copumpkin>
what's the endorsed way to build a darwin installer?
<nh2>
I took the liberty to put updating docs as the first point for now
<joepie91>
aha! I've just found the root of my infinite recursion problem!
<copumpkin>
err, binaryTarball
<nh2>
so that people see that it's a new movement ;)
<domenkozar>
maybe we can come up with a plan what docs to improve and not focus on tooling for once
<joepie91>
note for future readers: you cannot use `mkIf` directly in `config =` (at least, in NixOps)
<domenkozar>
and each coreish dev would take lead on one area
<domenkozar>
there's so many of us :)
LinArcX has joined #nixos
erictapen has quit [(Ping timeout: 248 seconds)]
<joepie91>
nh2: looking...
<copumpkin>
I'd be happy to do some deepish documentation on the darwin weirdness. Not sure where it would go though, since I don't think there's a corresponding documentation for linux stdenv
<copumpkin>
bennofs: that's a cute idea! it might work!
<bennofs>
copumpkin: that should allow the remaining code to assume that the store is at /nix, while allowing large installs
<joepie91>
nh2: excellent :)
<bennofs>
not sure how good the performance of that approach would be though
<clever>
joepie91: ah, your using it inside of the config, not as the value of config itself
<copumpkin>
bennofs: alternately, the new nix supports stores in different locations, while pretending they're in another location
<copumpkin>
so a bind mount + a store on the install medium
<joepie91>
clever: can you clarify what you mean with that? :)
<clever>
joepie91: odd, the broken version is what i would have expected to be the working one
arianvp2 has quit [(Read error: Connection reset by peer)]
<joepie91>
clever: well, I don't know if it actually *works*, as in, I haven't tested it yet
<joepie91>
it just doesn't throw infinite recursion errors at me
<joepie91>
and evaluates correctly
<bennofs>
copumpkin: that sounds like a better solution to me. why would you need a bind mount though?
arianvp2 has joined #nixos
<copumpkin>
so the host can treat it like a local store. Not sure that's necessary
<copumpkin>
need to think it through a bit more
<copumpkin>
perhaps post your idea on the ticket so we don't lose it?
lukego has quit [(Quit: Page closed)]
<bennofs>
copumpkin: ok will do.
<copumpkin>
gchristensen: looking at the source code, I can't see how it would work for you. build depends on tarball, and tarball unconditionally brings in seccomp
<joepie91>
clever: unfortunately, due to my earlier complaints concerning poor stacktraces, I don't yet know *why* the outer one breaks :)
<copumpkin>
even on 1.11.11
<clever>
joepie91: can you gist the entire trace and maybe i can decode it?
<domenkozar>
copumpkin: can you try against release-17.03 branch?
<joepie91>
clever: sure, one moment
<copumpkin>
it's a nix thing, not a nixpkgs thing
<domenkozar>
it's nix using nixpkgs package
<gchristensen>
copumpkin: no idea, but it does
<copumpkin>
yeah, but seccomp will never work on darwin
<copumpkin>
so having it as an unconditional dependency of tarball will always fail
<copumpkin>
argh, I have work to do, I'll build the nix installer later :P
<joepie91>
well this certainly makes it more interesting
<domenkozar>
copumpkin: that should get you going
<copumpkin>
too much yak for one morning
<joepie91>
oh...
<joepie91>
ha
<clever>
joepie91: maybe something like this in pastebin-steam.nix, { pkgs, ... }@args: import ./node-application.nix (args // { tarball ... })
<clever>
then it would receive all arguments a module takes, plus some extra
<joepie91>
clever: so I see why that doesn't fail, because if the mkIf is nested inside of the config attrset attributes then the attributes are probably never looked up, therefore config is never accessed, therefore it never fails
<joepie91>
mm, sec
<copumpkin>
domenkozar: left a comment
arianvp2 has quit [(Read error: Connection reset by peer)]
<copumpkin>
domenkozar: that file is full of <nixpkgs> references even though the file takes nixpkgs as an argument
arianvp2 has joined #nixos
<copumpkin>
so passing in --argstr nixpkgs does nothing
<domenkozar>
so it should be: import nixpkgs
<joepie91>
clever: is there a specific reason to glob it all into one set of args? I could probably accomplish the same by changing {inherit pkgs;} to args
<copumpkin>
I can set NIX_PATH but it seems like a bug
takle has quit [(Read error: Connection reset by peer)]
<joepie91>
using @args in the top-level func
<domenkozar>
copumpkin: let me fix also that
<copumpkin>
domenkozar: the weird pseudo-derivation default value is awkward though, because it has an outPath
<globin_>
niksnut: two evaluations of release-17.03 have passed without a channel bump
globin_ is now known as globin
takle has joined #nixos
<domenkozar>
copumpkin: ye
<clever>
joepie91: mostly so you can just get by with a single function at the top level of node-application.nix
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] zimbatm pushed 1 new commit to master: https://git.io/vQyHs
<NixOS_GitHub>
nixpkgs/master 9747eca Andrew Cann: houdini: init at 16.0.633 (#27326)
NixOS_GitHub has left #nixos []
<joepie91>
not sure that's beneficial to what I'm trying to accomplish structurally though
<joepie91>
(this is half an experiment, not just an immediate practical need)
<domenkozar>
copumpkin: pushed a fix for that
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nix] domenkozar pushed 1 new commit to darwin-tarball: https://git.io/vQyHl
<NixOS_GitHub>
nix/darwin-tarball af10cd9 Domen Kožar: release.nix: respect top-level nixpkgs input
NixOS_GitHub has left #nixos []
<domenkozar>
copumpkin: could you try now?
<copumpkin>
just left you another comment, sorry :)
<copumpkin>
oh I just mean nothing in nix uses any of the xcode tooling
<clever>
joepie91: i dont think nixos supports this kind of dynamic module creation, goes a bit farther then what it was designed for
freusque has joined #nixos
<domenkozar>
indeed, but errors come from the OSX
<domenkozar>
SSL during curl call installing nix
<domenkozar>
and mach-o linker from impurities?
<joepie91>
clever: where/why specifically is it breaking, though? this seems like something that /should/ be supported in some manner in something that allows abstractions :)
newhoggy has quit [(Ping timeout: 276 seconds)]
<clever>
joepie91: nixos will take the .config returned by every module (after recursively following imports), and merge them all together, then pass that merged result as the config argument, to every module
<clever>
joepie91: so you are getting your own return value as an argument
takle has quit [(Ping timeout: 276 seconds)]
<clever>
joepie91: and to prevent infinite recursion, the key names in the config set must be static, so it can compute them and know which module to route the requests into
jtojnar has joined #nixos
<gchristensen>
lol ... I just kicked off a travis build for my darwin installer, and the mack backlog is at 1,000 jobs >.>
<gilligan_>
gchristensen, *ouch*
<joepie91>
clever: *all* the key names, recursively, or just the top-level ones?
<NixOS_GitHub>
[nixpkgs] peterhoeg pushed 1 new commit to master: https://git.io/vQyhB
<NixOS_GitHub>
nixpkgs/master fc47846 Peter Hoeg: syncthing: 0.14.31 -> 0.14.32
NixOS_GitHub has left #nixos []
arianvp2 has joined #nixos
newhoggy has joined #nixos
<brodul>
is somebody using python2nix
<brodul>
it's not working for me
<domenkozar>
brodul: most people migrated to pypi2nix afaik
kiloreux has joined #nixos
<brodul>
ah ok, should I fix it or open a PR to be deleted?
AndChat114816 has quit [(Ping timeout: 246 seconds)]
<brodul>
can you run it without arguments
<brodul>
I want to know if it just me
eacameron has quit [(Remote host closed the connection)]
<brodul>
tnx :*
<domenkozar>
I think it can still be useful so it's ok to fix it :)
newhoggy has quit [(Ping timeout: 240 seconds)]
pietranera has joined #nixos
justelex_ has joined #nixos
<srhb>
brodul: No module named log, so yes, it's broken on unstable at least
<disasm>
as a maintainer of a package, if a dependency build fails and I get an e-mail about my package from hydra, there's nothing I have to do, right? Just wait for that dependency package to get fixed?
<domenkozar>
niksnut: do you plan to make another nix 1.11.x release?
arjen-jonathan has quit [(Ping timeout: 240 seconds)]
newhoggy has joined #nixos
hydraz has quit [(Read error: Connection reset by peer)]
cpennington has quit [(Ping timeout: 248 seconds)]
hydraz has joined #nixos
hydraz has quit [(Changing host)]
hydraz has joined #nixos
hotfuzz_ has joined #nixos
newhoggy has quit [(Ping timeout: 248 seconds)]
hotfuzz has quit [(Ping timeout: 240 seconds)]
<niksnut>
gchristensen / domenkozar: sure, just let me know when
bkchr has joined #nixos
newhoggy has joined #nixos
<domenkozar>
niksnut: assuming the locking fix works, we can release
<domenkozar>
but maybe gchristensen needs to test things
<gchristensen>
I'm expecting the travis test to pass, and then it'll be ready to merge
FRidh has quit [(Quit: Konversation terminated!)]
<gchristensen>
I mean, it _does_ work already, as far as all my testing shows. I tested it hundreds of times this morning in a loop :P at this point it is just making the Travis tests pass
<gchristensen>
which is poorly timed today, as Travis' mac infra went down for a bit this AM, so they have a larger than normal backlog.
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] peterhoeg pushed 3 new commits to master: https://git.io/vQSv2
<NixOS_GitHub>
nixpkgs/master 299c57e Peter Hoeg: terragrunt: make v0.11.1 use its own deps file
<NixOS_GitHub>
nixpkgs/master ef78ff8 Peter Hoeg: terragrunt: 0.12.16 -> 0.12.24
<NixOS_GitHub>
nixpkgs/master 16cc87c Peter Hoeg: terraform: disable checks for 0.9 as they are failing again
NixOS_GitHub has left #nixos []
<domenkozar>
I'm fine waiting a day :)
Eisfreak1 has joined #nixos
pietranera has quit [(Quit: Leaving.)]
newhoggy has quit [(Ping timeout: 248 seconds)]
eacameron has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] peterhoeg pushed 1 new commit to master: https://git.io/vQSfv
<NixOS_GitHub>
nixpkgs/master 18e2b44 Peter Hoeg: wp-cli: 1.2.0 -> 1.2.1
NixOS_GitHub has left #nixos []
newhoggy has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] gebner pushed 1 new commit to master: https://git.io/vQSfY
<NixOS_GitHub>
nixpkgs/master 71e44b1 Gabriel Ebner: vscode: 1.13.0 -> 1.14.0
NixOS_GitHub has left #nixos []
cpennington has joined #nixos
bkchr has quit [(Ping timeout: 246 seconds)]
eacameron has quit [(Ping timeout: 240 seconds)]
arianvp2 has quit [(Read error: Connection reset by peer)]
iyzsong has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
arianvp2 has joined #nixos
newhoggy has quit [(Ping timeout: 260 seconds)]
<joepie91>
let's see how my nixops deploy goes...
<joepie91>
slow DO is slow
vagrant- has joined #nixos
<vagrant->
hi guys, is there any way I can disable a setupHook on a per-package basis? I am building a node package which has a lot of inputs, and the setupHook for nodejs is taking a super long time, and I don't think it's necessary for what I'm doing
leat has joined #nixos
<joepie91>
I think I could speed up my DO deployment by carting a HDD to the datacenter in person :)
newhoggy has joined #nixos
<srhb>
shouldn't arguments work like this? nix-build '<nixpkgs>' -A pkgs.chromium --argstr enablePepperFlash true
<srhb>
I get no error, but the argument is still the default value of false, I think
<vagrant->
srhb: that will pass in the string "true", not the boolean value true
<vagrant->
(I think)
<vagrant->
oh I think I see what the problem is
<vagrant->
those arguments are going to the nixpkgs function, not to the chromium function
<srhb>
Oh, I thought callPackage would just pass them on
<disasm>
joepie91: is it your network connection that's slow? My deploys to d.o are pretty quick, although I don't have a whole lot of stuff in the config.
<vagrant->
nah, callPackage is just a nix function it's not aware
<joepie91>
disasm: based on the output I'm seeing it's both disk I/O and network that's being slow
<joepie91>
at least, slow for my standards
<srhb>
vagrant-: Well, so is the chromium derivation
<joepie91>
it's fairly on par with typical performance of popular "cloud" services
eacameron has joined #nixos
spinus has joined #nixos
newhoggy has quit [(Ping timeout: 255 seconds)]
<vagrant->
srhb: yes, but --args only passes args into the top-level function that's being evaluated. callPackage is downstream of that
<srhb>
Ah.
danl1240 has joined #nixos
<srhb>
Hm, what's the right approach then?
__Sander__ has quit [(Quit: Konversation terminated!)]
<vagrant->
generally those things are controlled in the nixpkgs config
<vagrant->
otherwise you might need to make a code change
<srhb>
I don't want to alter my config for this, since I'm making/fixing packages.
<srhb>
I guess I can call the derivation directly, but then I need to manually gather all those arguments
<vagrant->
yeah that's another option
arianvp2 has quit [(Read error: Connection reset by peer)]
<vagrant->
or you can write a file which calls pkgs.callPackage itself
arianvp2 has joined #nixos
<srhb>
Yeah. Hm. Don't people need this all the time? That is, cd to somewhere in nixpkgs and intuitively just run callPackage on the derivation in default.nix?
<vagrant->
eh, perhaps
<vagrant->
generally speaking i like to have everything I build written out as code rather than using a command to build it
<vagrant->
for a variety of reasons
eacameron has quit [(Ping timeout: 240 seconds)]
<srhb>
I wonder how other people develop nixpkgs.
<vagrant->
just by hacking on the nixpkgs code?
<vagrant->
it's easier if you use nixpkgs through git rather than a channel
<srhb>
Having to essentially have a utility repo with callPackage for everything in nixpkgs seems tedious
<srhb>
vagrant-: I am.
<vagrant->
I've never found that to be necessary myself
<vagrant->
so can't you just make whatever code changes you want and build it?
jgertm has joined #nixos
<srhb>
vagrant-: The issue is that I can't easily pass the argument. I mean, I can, but I'd have to alter all-packages.nix and then switch the argument back afterwards
<vagrant->
oh, I wish I'd realized that's what you wanted haha
Gravious has joined #nixos
<srhb>
vagrant-: I was probably too imprecise :)
<vagrant->
hah it's all good, glad you got it :)
newhoggy has quit [(Ping timeout: 240 seconds)]
IITaudio_ has quit [(Ping timeout: 260 seconds)]
nschoe has quit [(Quit: Program. Terminated.)]
jonte has joined #nixos
<Sonarpulse>
copumpkin: I can help with the mac haskell thing in about 20
<copumpkin>
I'm mostly just throwing ideas at the wall to see if they stick
<copumpkin>
I don't want Nix Mac Haskell to suck but I also don't use it enough to really be able to invest much time in it
<copumpkin>
:)
zeus_ has joined #nixos
Eisfreak1 has quit [(Ping timeout: 240 seconds)]
eacameron has joined #nixos
<gilligan_>
copumpkin, what exactly is "nix mac haskell" anyway?
<copumpkin>
haskell stuff via nix, on a mac
<sphalerite>
presumably using nix as a package manager for haskell stuff on macOS :)
<copumpkin>
sphalerite gets it :D
mudri has joined #nixos
Sonarpulse has quit [(Ping timeout: 246 seconds)]
<niksnut>
wow, clangStdenv is almost 3 times bigger than stdenv
<gilligan_>
copumpkin, does that need a different approach on osx than on nixos/linux? The usual cabal2nix stuff works, doesn't it?
newhoggy has joined #nixos
<LnL>
gilligan_: there's a small difference to work around the rpath limit introduced in sierra
jgertm has quit [(Ping timeout: 248 seconds)]
<gilligan_>
oh god
<niksnut>
why is llvmPackages in nixpkgs master llvm 4 on macOS and llvm 3 elsewhere?
<niksnut>
that seems like an unnecessary inconsistency
<copumpkin>
niksnut: we didn't want to bump linux to 4 when we bumped macOS stdenv to 4
<LnL>
yeah we should get rid of that if else
<copumpkin>
as for why macOS is on 4, upgrading it is a PITA and expensive in time, so I figured I'd get as new as possible if I was going to put in the work anyway :)
<LnL>
don't expect a lot of issues except changing some isDarwins to cc.isClang
<copumpkin>
but yeah, there should be no reason to keep linux on 3.x; I just don't know the full implications of a linux llvm upgrade
<sphalerite>
niksnut: python is probably a major factor in the size difference?
Ralith_ has quit [(Read error: Connection reset by peer)]
<sphalerite>
niksnut: interesting that the clang one also still depends on gcc
<niksnut>
oh, clang 4 stdenv adds another 150 MB to the closure
Ralith_ has joined #nixos
ThatDocsLady has quit [(Ping timeout: 240 seconds)]
newhoggy_ has joined #nixos
<niksnut>
sphalerite: there is no python
newhoggy has quit [(Ping timeout: 260 seconds)]
<niksnut>
mostly clang is very big
<niksnut>
a lot of .a files
<sphalerite>
oh, there is for me but that might just be 17.03
<copumpkin>
ouch
<copumpkin>
I haven't been tracking closure size
<copumpkin>
any idea where it's from?
<copumpkin>
I separated out the python stuff
<copumpkin>
does anything use the .a files? we could just split into separate output I guess
<copumpkin>
shlevy: I'm now confused about the dylib fix :)
<copumpkin>
how does gcc end up in the clang stdenv?
<copumpkin>
we don't have that in the darwin one
<copumpkin>
oh I see you said so above
<copumpkin>
why not just use libc++ instead of libstdc++?
<copumpkin>
that way you stay in pure consistent llvm-land
<sphalerite>
oh boy, that brings back dark memories of trying to compile stuff on OSX back when it was still in the limbo between glibc (I think) and libc++
<sphalerite>
(and the stuff I was compiling was C++11)
<sphalerite>
Took me so long to work out that all those crashes were because of ABI incompatibility — trying to use libc++ headers with glibc
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] edolstra pushed 1 new commit to master: https://git.io/vQSLF
<NixOS_GitHub>
nixpkgs/master 88d9d2d Eelco Dolstra: Use LLVM 4 on all platforms
NixOS_GitHub has left #nixos []
<niksnut>
copumpkin: on Linux everything is built with libstdc++, so I don't think it's desirable to mix with libc++
dbmikus_ has quit [(Quit: WeeChat 1.9)]
<copumpkin>
ah, true
<sphalerite>
yeah, I'd be inclined to agree with that. Because I don't think applications compiled against libc++ whose dependencies are libstdc++ would break until runtime
erictapen has quit [(Ping timeout: 240 seconds)]
<sphalerite>
at least that's the problem I was having on OSX back then
<copumpkin>
yeah, they aren't ABI-compatible
<copumpkin>
on macOS it works because we only use libc++
newhoggy has joined #nixos
<sphalerite>
but providing a libc++ stdenv on linux too might be interesting :p
rycee[m] has joined #nixos
erictapen has joined #nixos
slack1256 has joined #nixos
<copumpkin>
Sonarpulse: you've been elected to implement it!
<slack1256>
How do I get package statistics? I want to know how many times a package has been downloaded from the binary cache
<copumpkin>
I don't think those stats are publicly available
alx741 has joined #nixos
<copumpkin>
we've discussed writing a simple aggregator over the S3 logs
<copumpkin>
but someone would need to write that and convince niksnut that it doesn't leak information
<copumpkin>
it would be pretty cool, especially in figuring out whether to bother building everything for i686 :P
<slack1256>
haha
<copumpkin>
just think of the capacity we'd free up if we didn't
<copumpkin>
I have a hunch there's probably a dozen people using i686
newhoggy has quit [(Ping timeout: 255 seconds)]
ison111 has joined #nixos
<LnL>
that would definitely make a difference
<gchristensen>
I asked on the ML a while ago and a few people said they'd be unhappy to see it go away
<copumpkin>
at some point we're gonna kill it though, right? seems worth discussing at what point that'll happen. There's always gonna be someone running NixOS on an ancient laptop and it doesn't make sense to build everything for i686 just for that. Perhaps just major NixOS releases on the stable channel or something
* copumpkin
shrugs :)
<copumpkin>
I think we did have a 32-bit mac mini user ask about nix on darwin at some point
<copumpkin>
:P
mudri has quit [(Ping timeout: 260 seconds)]
<gchristensen>
the problem with only doing it for stable is that means stable unstable will be broken
<gchristensen>
and that means the first time we only build it for stable is the last stable release we maintain for i686
<copumpkin>
yeah I guess
<copumpkin>
I dunno, just seems like a lot of compute power and human effort going into what might just be a dozen people
<gchristensen>
"that means stable unstable" _oops_
<copumpkin>
it's nice that we do it, but I'd rather we invest limited resources elsewhere :)
<gchristensen>
we could turn it off for everything except nixos-unstable
arjen-jonathan has joined #nixos
<gchristensen>
that way nixpkgs-* doesn't build it (they can just use nixos-unstable) and nixos-stable builds don't either, cutting builds my some factor
alx741 has quit [(Ping timeout: 276 seconds)]
endformationage has joined #nixos
<sphalerite>
I'm still planning to migrate the uni computing society's systems to NixOS, and a fair few of those are still 32-bit only
newhoggy has joined #nixos
<sphalerite>
Of course they don't actually do much so they might as well spend all their time building packages :p
* copumpkin
shrugs :) not my resources/time/money being thrown at it, so I can't really dicatate
darlan has joined #nixos
<gchristensen>
ok well I think sphalerite was the most compelling reason
newhoggy has quit [(Ping timeout: 240 seconds)]
<gchristensen>
we are a source-based hybrid, they can compile it themselves, or donate hydra resources
<simukis>
linuxPackages_4_12 missing error is still here, any ideas?
jgertm has joined #nixos
jgertm is now known as Guest56361
newhoggy has joined #nixos
erictapen has quit [(Ping timeout: 255 seconds)]
newhoggy has quit [(Ping timeout: 248 seconds)]
arianvp2 has joined #nixos
Sonarpulse has joined #nixos
<Sonarpulse>
copumpkin: is there any reason we don't use use install_name_tool to make everything absolute?
<Sonarpulse>
that would make for faster loading
<copumpkin>
it is all absolute
<Sonarpulse>
copumpkin: wait...
cpennington has quit [(Ping timeout: 260 seconds)]
<Sonarpulse>
is is the issue something like RUNPATH or DT_NEEDED?
<gchristensen>
copumpkin: if you start a thread of the new email list about discontinuing i686 I'd :thumbsup: it :)
* Sonarpulse
only knows the elf terminology
<sphalerite[m]>
Really, mine was the most compelling? Oh no x)
<gchristensen>
sphalerite[m]: a bunch of university systems seems compelling to me
<sphalerite[m]>
They're not actual university systems, just the computing society servers which aren't actually doing anything besides hosting the website (and that's been down since some electrical maintenance 2 months ago, because I don't know on which server it lives)
<gchristensen>
LOL
<sphalerite[m]>
It's like 10 boxes serving 5 services of which 1 is used
newhoggy has joined #nixos
drakonis has quit [(Read error: Connection reset by peer)]
<sphalerite>
And the guy who used to be in charge of it is usually quite busy
arianvp2 has quit [(Read error: Connection reset by peer)]
<sphalerite>
or he just doesn't remember stuff
arianvp2 has joined #nixos
phinxy has joined #nixos
ebopp has joined #nixos
slack1256 has quit [(Ping timeout: 276 seconds)]
newhoggy has quit [(Ping timeout: 255 seconds)]
<sphalerite>
The services I know about are DNS (working), SSH for people to get in (I last used it 2 months ago and I'd be willing to bet that nobody else has), the website (probably got more bots visiting it than people), gitlab (which doesn't have much on it, and the last commit to any repo on it was also made by me), VPN (but I'm not sure that got restored after everything got wiped for restructuring 2 years ago),
<sphalerite>
and then some internal stuff like LDAP and NFS
Guest56361 has quit [(Ping timeout: 246 seconds)]
<sphalerite>
Of those, I know where DNS lives and I can probably work out where LDAP and NFS live.
<gchristensen>
nmap and tcpdump on a switch :P
<sphalerite>
Oh and XMPP which I set up but nobody uses
<sphalerite>
yeah I did an nmap a while back, but the incorrect reverse DNS entries make it all very confusing :p
orivej has joined #nixos
<simukis>
how to find where ovmf firmware image is stored?
<sphalerite>
there was also some virtualisation involved, because the old (physical) web server machine's RAM died. The guy in charge moved it to a VM but I don't know which physical machine it lives on or what the VM is called, nor how it's managed
<sphalerite>
simukis: echo $(nix-build '<nixpkgs>' -A OVMF)/FV/OVMF.fd
Havvy has joined #nixos
<sphalerite>
And the machines are all debian, and managed imperatively. There used to be some puppet stuff but no more, I think.
<simukis>
sphalerite: thanks!
cpennington has joined #nixos
arianvp2 has quit [(Quit: arianvp2)]
<sphalerite>
simukis: oh maybe add --no-out-link so you don't get the result symlink sitting around
arjen-jonathan has quit [(Ping timeout: 260 seconds)]
anton__ has quit [(Remote host closed the connection)]
<sphalerite>
But yeah, hence my intentions to move it all to NixOS. But unless someone will actually use services running on the machines, or run their own stuff on them, there's not much point.
<simukis>
sphalerite, also it is OVMF.fd too
* ToxicFrog
tries to figure out why /nix/var is included in their backups
<sphalerite>
simukis: what? sorry
arianvp2 has joined #nixos
<simukis>
sphalerite: the derivation which OVMF.fd comes from is OVMF.fd, weirdly enough
<sphalerite>
I suppose I could set them up to build nixpkgs i386, that could be useful maybe
ison111 has quit [(Ping timeout: 246 seconds)]
<sphalerite>
simukis: oh, heh. I did not know that, that's good to know
orivej has quit [(Ping timeout: 276 seconds)]
<ToxicFrog>
Oh, I think it's because I wanted /nix/var/nix/profiles/per-user backed up, but those are just symlinks into /nix/store anyways so it's useless. Blerh.
<sphalerite>
simukis: nix-build --no-out-link '<nixpkgs>' -A OVMF.fd says the attribute fd wans't found
<simukis>
sphalerite: hmm, weird. I’m on unstable channels though
<simukis>
maybe that’s why?
<sphalerite>
simukis: maybe, yeah. I'm on 17.03
newhoggy has joined #nixos
<simukis>
but I also cannot use linuxPackages_4_12 for some reason either, despite it should be available
<sphalerite>
Surely the branch should never be ahead of the channel?
<sphalerite>
gchristensen: do you know anything about this?
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vbgl pushed 4 new commits to master: https://git.io/vQSWt
<NixOS_GitHub>
nixpkgs/master 6a3a4f5 Vincent Laporte: ocamlPackages.integers: init at 0.2.2
<NixOS_GitHub>
nixpkgs/master dd90745 Vincent Laporte: ocamlPackages.ctypes: 0.11.2 -> 0.11.5
<NixOS_GitHub>
nixpkgs/master 5772d29 Vincent Laporte: ocsigenserver: enable deflatemod extension
NixOS_GitHub has left #nixos []
<copumpkin>
Sonarpulse: unclear what you're asking :)
svend has quit [(Client Quit)]
<Sonarpulse>
copumpkin: what is overlowing the limit
<Sonarpulse>
lookup paths
<Sonarpulse>
or sheer number of dynamic libraries?
<copumpkin>
either
<Sonarpulse>
oh
<copumpkin>
it's the mach-o header limit
<copumpkin>
which contains both
<Sonarpulse>
ah
<Sonarpulse>
yuck
svend has joined #nixos
<Sonarpulse>
limiting the paths to avoid O(n^2) lookup----I can see apple doing that
<Sonarpulse>
limiting total number of libraries? that is just aweful
<copumpkin>
we don't use rpath
<copumpkin>
it's not the same as linu
<copumpkin>
so no n^2 lookup issue
<copumpkin>
mach-o is an ancient format
<copumpkin>
predates apple
newhoggy has quit [(Ping timeout: 240 seconds)]
alx741 has joined #nixos
<Sonarpulse>
copumpkin: right....
<Sonarpulse>
you just told me about install_name_tool :)
<copumpkin>
yeah, which is basically patchelf for mach-o
<Sonarpulse>
well then I cannot think of any reason for apple being so mean :(
<copumpkin>
minus some nicer features
<jasom>
how do I clean up space in /boot?
<copumpkin>
anyway, the end result is that the header is limited in total sise
<copumpkin>
so we can address that by nesting
svend has left #nixos []
<Sonarpulse>
copumpkin: should we make this a haskell-specific thing, or all of darwin
<Sonarpulse>
(stdenv / cc-wrapper change)
chakerbenhamed has quit [(Ping timeout: 248 seconds)]
<copumpkin>
could be all of darwin if we put it in ld-wrapper
<Sonarpulse>
(oh better go back to ##nix-darwin_
<copumpkin>
but I don't know if GHC is doing something odd
<jasom>
this is with systemd-boot if that matters
<sphalerite>
jasom: delete old system generations, rebuild
jonte has quit [(Ping timeout: 240 seconds)]
<sphalerite>
(if you want to delete system generations without affecting user profile generations, use sudo nix-env -p /nix/var/nix/profiles/system --delete-generations)
<joepie91>
(hello/ex1 does not appear to actually exist)
arianvp2 has quit [(Client Quit)]
jonte has joined #nixos
sary has quit [(Ping timeout: 260 seconds)]
sary has joined #nixos
<celph>
Wondering what I should be doing about ruby gems
<celph>
Should I just package my dependencies up for Nix? I haven't done that before, but I imagine I could.
<gchristensen>
celph: check out bundix
newhoggy has joined #nixos
<celph>
nice that's exactly what I need
<WinchellsM>
@joepie91 that clear it up; I thought it was referring to a local filepath
ertes-w has quit [(Ping timeout: 240 seconds)]
<celph>
gchristensen: thank you
<nh2>
which command builds me a given .drv file?
<gchristensen>
nix-store -r
<nh2>
gchristensen: thanks
<sphalerite>
aww, ninja
<bachp>
How do I make a systemd service in a nix module restart when a I do a nixos-rebuild switch. The issue is that the file is generated in the module but there is no explicit connection between the file and the systemd service, specifically the file is linked to /etc but not mentioned in the service.
<WinchellsM>
@joepie91: Where is `all-packages.nix` located in that nixpkgs hiearchy? I can't seem to find it
thblt has quit [(Remote host closed the connection)]
tmaekawa has joined #nixos
<nh2>
bachp: I do that. I have a setup with gluster over tinc VPN + gluster TLS + volume creation on initial nixops deploy + geo replication setup on initial nixops deploy
<NixOS_GitHub>
[nixpkgs] vbgl pushed 1 new commit to master: https://git.io/vQSwl
<NixOS_GitHub>
nixpkgs/master e47dcb9 Vincent Laporte: ocamlPackages.tsdl: 0.9.0 -> 0.9.4
NixOS_GitHub has left #nixos []
<manveru>
gchristensen: so... the fix for bundix wasn't good enough
ennui` has joined #nixos
newhoggy has quit [(Ping timeout: 255 seconds)]
celldee has quit [(Quit: leaving)]
m0rphism has quit [(Quit: WeeChat 1.8)]
[0x4A6F] has quit [(Quit: [0x4A6F])]
newhoggy has joined #nixos
arianvp2 has quit [(Ping timeout: 240 seconds)]
takle has joined #nixos
newhoggy has quit [(Ping timeout: 268 seconds)]
indi__ has joined #nixos
<copumpkin>
Sonarpulse: so stdenv.isDarwin/Linux is a no-no nowadays? I keep finding tons of it
<Sonarpulse>
copumpkin: I haven't actually deprecatd it
<Sonarpulse>
or done the replace
<Sonarpulse>
but yeah I want it out of nixpkgs and printin a warning for 17.09
<copumpkin>
a large chunk of the .isDarwin checks should actually be stdenv.isClang or the like
<Sonarpulse>
m ok
<Sonarpulse>
now that `stdenv.hostPlatform.is` exists, I can do a simple `git ls-files | xargs sed ...`
<copumpkin>
what's the proper newfangled way to do that?
<LnL>
probably better to warn/switch with 17.09
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] vbgl pushed 1 new commit to master: https://git.io/vQSoA
<NixOS_GitHub>
nixpkgs/master ff965a5 Vincent Laporte: scite: 3.7.3 -> 3.7.5
NixOS_GitHub has left #nixos []
<Sonarpulse>
LnL: yeah it would still be bound in 17.09, but nothing in nixpkgs would use it
<LnL>
otherwise backporting commits becomes a bit annoying
<Sonarpulse>
so cargo culters wouldn't reintroduce it
<Sonarpulse>
oh that
<Sonarpulse>
*with* 17.09
<LnL>
just mean that those hostPlatform changes aren't in 17.03, right?
indi__ has quit [(Ping timeout: 248 seconds)]
<Sonarpulse>
I think i understand now: make sure nixpkgs-stable and unstable both have the change
<LnL>
yeah that makes stuff a lot easier
newhoggy has joined #nixos
Infinisil has joined #nixos
<gchristensen>
I'm going around nixos-install. any ideas why my newly-populated `/mnt` doesn't contain switch-to-configuration? :)
<jophish>
clever: how easy do you think it would be to have a linux system read a disk image over the network and replace /dev/sda (or whatever) with that image, where /dev/sda contains / /boot, whatever
justbeingglad has joined #nixos
justbeingglad has left #nixos []
newhoggy has quit [(Ping timeout: 260 seconds)]
kiloreux has quit [(Ping timeout: 248 seconds)]
<jophish>
It seems quite similar to your kexec stuff
uwap is now known as TrueUnixUser
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] bachp opened pull request #27340: glusterfs service: add support for TLS communication (master...glusterfs-tls) https://git.io/vQS6m
NixOS_GitHub has left #nixos []
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 4 new commits to master: https://git.io/vQS6Y
MercurialAlchemi has quit [(Ping timeout: 248 seconds)]
newhoggy has joined #nixos
stanibanani has joined #nixos
DutchWolfie has quit [(Quit: Konversation terminated!)]
orivej has joined #nixos
newhoggy has quit [(Ping timeout: 246 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 5 new commits to master: https://git.io/vQSXo
<NixOS_GitHub>
nixpkgs/master bd2cef0 xd1le: st: refactor deprecated `isNull' function
<NixOS_GitHub>
nixpkgs/master 4934a51 xd1le: st: refactor out `stdenv.lib' already in scope
<NixOS_GitHub>
nixpkgs/master 88874e7 xd1le: xst: init at 0.7.1
NixOS_GitHub has left #nixos []
<gchristensen>
I'm having a really strange problem with nixos-install
<gchristensen>
/mnt/nix/var/nix/profiles/system-1-link points to /mnt//nix/store/cv7ingzkcnws29cxfz241m491a1f04d7-closure which is a list of paths, not an actual system closure
<sphalerite>
gchristensen: I had that exact issue the other day :)
<gchristensen>
oh _really_?
<sphalerite>
it's a quoting issue, I was going to submit a PR for it
<copumpkin>
I rewrote nixos-install fairly recently. Maybe I screwed it up? But then again we have extensive tests for it..
<sphalerite>
haven't got around to it yet
<gchristensen>
oh ok :) I'll revert to a older channel
<gchristensen>
I thought I was just breaking stuff by doing insane things
<sphalerite>
gchristensen: I definitely was doing insane things when I discovered it
<sphalerite>
hang on
<sphalerite>
I even have the exact code open in a tab to remind myself ofit
<sphalerite>
which confuses nixos-prepare-root greatly
<sphalerite>
and is fixed by simply putting quotes around all the $stuffs
<copumpkin>
ah, I wonder what awful person wrote all that
<gchristensen>
that is exactly what is happening :P
<copumpkin>
should chew him out
<gchristensen>
copumpkin: :eyes:
<sphalerite>
hahaha
tmaekawa has quit [(Quit: tmaekawa)]
<copumpkin>
:noses:
cpennington has quit [(Remote host closed the connection)]
<sphalerite>
copumpkin: I won't judge you for making that mistake. I will slightly judge you for not discovering it using shellcheck though ;)
<copumpkin>
pfft shellcheck
<sphalerite>
(just a little though. tbh I didn't use shellcheck after messing around with the scripts myslef)
<copumpkin>
I don't need any of that newfangled haskell stuff
<sphalerite>
So yes, how's about I PR that fix now
<copumpkin>
:) thanks!
newhoggy has quit [(Ping timeout: 240 seconds)]
<sphalerite>
Also, while I was screwing around with that stuff I discovered that PS4 is immensely useful for debugging shell scripts that call each other
HurricaneHarry has quit [(Read error: Connection reset by peer)]
HurricaneHarry has joined #nixos
takle has quit [(Ping timeout: 255 seconds)]
Wizek_ has quit [(Ping timeout: 260 seconds)]
<LnL>
it's the inputs that are not working
<LnL>
gchristensen: do you remember if that was fixed?
adamCS has quit [(Ping timeout: 260 seconds)]
<gchristensen>
sorry?
<gchristensen>
oh, I don't know :/
<LnL>
hydra the input forms didn't work before
<LnL>
guess I should check master
adamCS has joined #nixos
takle has joined #nixos
newhoggy has quit [(Remote host closed the connection)]
<niksnut>
shlevy: your message to nix-devel hit the spam filter
katyucha_ has quit [(Ping timeout: 255 seconds)]
digitus has quit [(Quit: digitus)]
ilyaigpetrov has quit [(Quit: Connection closed for inactivity)]
warbo has joined #nixos
warbo has left #nixos []
ThatDocsLady has joined #nixos
<LnL>
gchristensen: n/m user error
simukis has quit [(Ping timeout: 248 seconds)]
arjen-jonathan has quit [(Ping timeout: 260 seconds)]
hamishmack has quit [(Quit: hamishmack)]
<joepie91>
hmmm. how do people normally deal with web applications (eg. random PHP forum software) that expects to be able to write to the directory it's installed in?
<nh2>
bachp: looks like I succeeded with the first step of extracting it, currently testing if it works, will share with you if it does
hamishmack has joined #nixos
<simpson>
joepie91: In the past, with lots of patching and cursing at PHP. The
<simpson>
The more popular PHP apps should have configuration for where to write, simply because it's a common PHP security practice.
<simpson>
*should already have, not should morally have, although obviously Nix users have an opinion on that~
<joepie91>
simpson: right. the question is more generic, PHP apps were one example
<joepie91>
like, what if the application doesn't let you configure this?
<cransom>
joepie91: depends on how it was written. i've done a few different things like running through the code and doing a string replace to write files to a sane directory
<joepie91>
cransom: what would you consider a 'sane directory'?
<simpson>
joepie91: I don't take that kind of attitude from binaries which I can patch. For stuff like Steam, there's the FHS user env, which works wonderfully.
<cransom>
joepie91: /var/lib/$appname could be one. depends on the nature of what its writing.
<joepie91>
simpson: how are FHS envs implemented, from a high-level perspective?
<joepie91>
simpson: "make a dir in /var/lib and chroot the app into it on runtime"?
<NixOS_GitHub>
[nixpkgs] magnetophon opened pull request #27343: nova-filters: bug fix and add maintainer (master...nova) https://git.io/vQS5q
NixOS_GitHub has left #nixos []
<joepie91>
I see
<simpson>
And this is assuming that an app is written to demand Ubuntu or something like that. If it's just writing a single directory path, again, I'd much rather just go patch it than anything else.
<joepie91>
right
ebopp has quit [(Quit: leaving)]
<simpson>
There's always having a rundir, copying the app into /run or /tmp and running from there.
<joepie91>
my current issue is actually an application I wrote myself that normally is executed from a homedir and writes to its own dir, and obviously the solution there is to make the error write path configurable, but I figured I'd ask about the more generic solutions as well since I'm bound to run into them at some point :)
<joepie91>
right
<joepie91>
that seems a bit hacky
<joepie91>
:p
tput has joined #nixos
tput has quit [(Client Quit)]
gilligan_ has quit [(Ping timeout: 255 seconds)]
<sphalerite>
or overlayfs (of course that unfortunately requires root) to overlay the written stuff on the readonly stuff
radvendii has quit [(Ping timeout: 248 seconds)]
hiratara has quit [(Ping timeout: 246 seconds)]
celph has quit [(Ping timeout: 260 seconds)]
erasmas has quit [(Quit: leaving)]
jonte has quit [(Ping timeout: 255 seconds)]
hiratara has joined #nixos
hamishmack has quit [(Quit: hamishmack)]
earldouglas has quit [(Quit: leaving)]
newhoggy has joined #nixos
<pstn>
How can I make a container with private interfaces reachabel from the outside? I can reach it from the host by connecting to the hostAddress, but the traffic that reaches my public interface doesn't get forwarded.
kiloreux has quit [(Remote host closed the connection)]
jgertm has joined #nixos
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] lukeadams opened pull request #27344: Gnuradio: Enable Darwin Support (master...gnuradio-update) https://git.io/vQSNo
ThatDocsLady has quit [(Ping timeout: 260 seconds)]
NixOS_GitHub has joined #nixos
<NixOS_GitHub>
[nixpkgs] Mic92 pushed 1 new commit to release-17.03: https://git.io/vQSx9
<NixOS_GitHub>
nixpkgs/release-17.03 b63d2c3 Emmanuel Rosa: pgadmin: make desktop item...
NixOS_GitHub has left #nixos []
<pstn>
nh2: sudo nix-store --verify --check-contents --repair could help, but if your nix-store is suddenly changed without you doing anything forcefull there might be something wrong with your storage medium.
<pstn>
(Better back up all important files NOW in that case, if you didn't already do that)
mudri has quit [(Ping timeout: 248 seconds)]
<nh2>
pstn: thanks, it might indeed have corruption. I'm trying your command. Luckily this server is expendable
indi_ has joined #nixos
<joepie91>
well this is a little disappointing
hiratara has quit [(Remote host closed the connection)]
<joepie91>
but something is screwy somewhere so it's not updating live as it should
<joepie91>
:/
endformationage has quit [(Quit: WeeChat 1.7)]
<joepie91>
probably not Nix-related though, so, yay, successful real-world NixOps deployment!
newhoggy has quit [(Remote host closed the connection)]
hiratara has joined #nixos
watersoul has quit [(Write error: Connection reset by peer)]
LnL has quit [(Excess Flood)]
LnL has joined #nixos
LnL is now known as Guest59586
<nh2>
pstn: yeah your command said that some files have unexpected hashes
mudri has joined #nixos
joachifm has quit [(Remote host closed the connection)]
Twey has quit [(Ping timeout: 251 seconds)]
dredozub- has quit [(Ping timeout: 251 seconds)]
blahdodo has quit [(Ping timeout: 251 seconds)]
pmade has quit [(Ping timeout: 251 seconds)]
Sigma has quit [(Ping timeout: 251 seconds)]
Biappi has quit [(Ping timeout: 251 seconds)]
mw has quit [(Ping timeout: 251 seconds)]
shlevy has quit [(Ping timeout: 251 seconds)]
joachifm has joined #nixos
devlog has quit [(Remote host closed the connection)]
vith2 has quit [(Remote host closed the connection)]
aszlig has quit [(Remote host closed the connection)]
katyucha has quit [(Remote host closed the connection)]
petar has quit [(Remote host closed the connection)]
aszlig has joined #nixos
petar has joined #nixos
oleks has quit [(Remote host closed the connection)]
vith2 has joined #nixos
katyucha has joined #nixos
devlog has joined #nixos
oleks has joined #nixos
shlevy has joined #nixos
blahdodo has joined #nixos
Sigma has joined #nixos
pmade has joined #nixos
Twey has joined #nixos
dredozubov has joined #nixos
Biappi has joined #nixos
watersoul has joined #nixos
mw has joined #nixos
<pstn>
nh2 Have you checked dmesg for block device errors?
ng0 has quit [(Quit: Alexa, when is the end of world?)]
<nh2>
pstn: roughly, but I don't remember exactly what those messages look like and there's loads of stuff in journalctl so scrolling over doesn't work
<nh2>
pstn: do you know what string to grep for?
<pstn>
Maybe your block device name?
<pstn>
You could also try smartctl -a $YOUR_BLOCKDEVICE with smartmontools
<nh2>
> For Seagate disks (and possibly some old ones from WD too) the Seek_Error_Rate and Raw_Read_Error_Rate are 48 bit numbers, where the most significant 16 bits are an error count, and the low 32 bits are a number of operations.
<nh2>
pstn: so high number doesn't necessarily mean broken depending on the vendor
<nh2>
and mine here are seagates
markus1199 has quit [(Ping timeout: 240 seconds)]
eacameron has quit [(Remote host closed the connection)]
justelex_ has quit [(Ping timeout: 248 seconds)]
<pstn>
Ah, interesting
civodul has quit [(Quit: ERC (IRC client for Emacs 25.2.1))]
ebzzry has joined #nixos
eacameron has joined #nixos
mudri has quit [(Ping timeout: 268 seconds)]
JagaJaga has quit [(Ping timeout: 276 seconds)]
anton__ has quit [(Remote host closed the connection)]
nix-gsc-io`bot has joined #nixos
<qknight>
can i assert a mkOption like i can check typing with type.enum for instance?