zraexy has quit [(Quit: Leaving.)]
zraexy has joined joined #nixos-dev
tanonym has quit [(Quit: Leaving)]
<copumpkin> clever: if you're curious, https://github.com/NixOS/nix/issues/1661
<clever> ah, yeah, i can see that being weird
<clever> i find that in general, import doesnt intern things into the store when given a path
<clever> copumpkin: oh, that gives me an idea, what about pkgs.path + "/nixos/lib/eval-config.nix"
<clever> copumpkin: that wont coerce it to a string, but rather, keep it as a path
<copumpkin> yeah that works, I think I mention it in the ticket
<copumpkin> but yeah I can work around it
<copumpkin> was just confused by the behavior
<clever> ah, you have an unquoted variant of that, which i think is invalid
<clever> i think the problem is to do with casting a derivation to a string, rather then keeping it as a path
<clever> and its being copied before import gets called
<clever> when import would accept a path and not copy
<copumpkin> seems to work fine unquoted
<clever> odd, i would have expected that to eval as an absolute path
<copumpkin> paths are pretty weird :P
<clever> maybe the + makes path parsing work differently
<clever> another weird thing is /path/to/directory/ failing with a cryptic error
<clever> the trailing / isnt part of the path, thats a division operator!
<copumpkin> oh yeah :)
<copumpkin> I think we have an issue open for that
<copumpkin> we missed you at NixCon!
<clever> ive been at ase2017
<copumpkin> looks fun
<copumpkin> but still not as cool as nixcon
<copumpkin> :P
<clever> copumpkin: does nixcon have https://imgur.com/a/XckCY ?
<copumpkin> whoa, I know one of the PC folks there
<copumpkin> okay, not that
<clever> copumpkin: how many people showed up? https://imgur.com/a/7IoVh
<copumpkin> I think around 100
infinisil has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
mbrgm has quit [(Ping timeout: 248 seconds)]
mbrgm has joined joined #nixos-dev
taktoa has joined joined #nixos-dev
JosW has joined joined #nixos-dev
orivej has quit [(Ping timeout: 240 seconds)]
ckauhaus has joined joined #nixos-dev
goibhniu has joined joined #nixos-dev
__Sander__ has joined joined #nixos-dev
<MoreTea> gchristensen: as an addition to my remark to just use a spreadsheet for security triageing, there is a python lib that can read spreadsheets. http://tinaja.computer/2017/10/27/gspread.html
<gchristensen> :o
<MoreTea> :%s/spreasheets/google spreadsheets/
<MoreTea> We could write a simple python IRC bot that read that file and pings security triage peepz :)
<Dezgeg> gchristensen: hm, this one shows up as zero rebuilds which can't be right: https://github.com/NixOS/nixpkgs/pull/31177
FRidh has joined joined #nixos-dev
olto has joined joined #nixos-dev
olto has quit [(Quit: hasta luego)]
<grahamc> Dezgeg (IRC): huh. Can you try the rebuilt amount script against it to see what it says?
olto has joined joined #nixos-dev
<grahamc> It’s in maintainers/scripts
<Dezgeg> yes I will debug
<LnL> grahamc: 1 x86_64-darwin 1 x86_64-linux
<grahamc> :/ can you gist exact output? Maybe my regex is wrong.
<LnL> was just trying to find that
olto has quit [(Quit: hasta luego)]
<grahamc> Hrm. My regex should be able to handle that. I’m about to go to the train and can look at it in an hour or so
olto has joined joined #nixos-dev
<niksnut> since when does Nix depend on ghc
<LnL> I noticed that during the hackathon
olto has quit [(Client Quit)]
olto has joined joined #nixos-dev
<niksnut> due to something called "shellcheck"
<LnL> the build.x86_64-darwin drv also tried to build linux stuff
<LnL> ah shellcheck
<grahamc> Uh oh. Did I do that?
<niksnut> actually it's only the binaryTarball job
<Profpatsch> niksnut: Are we moving away from the `foo' quoting style?
<niksnut> foo quoting style?
<Profpatsch> `' instead of e.g. ''
<niksnut> we never used that
<niksnut> we did use unicode quotes
<niksnut> in Nix
<Profpatsch> Ah, in nixpkgs docs.
<niksnut> using `...' is always wrong
<Profpatsch> Okay, I’m not sure why this was a thing in the first place then.
<Profpatsch> thx
<Profpatsch> grahamc: Do we need a rebuild-x:0 tag?
<Profpatsch> It may be a bit noisy on each PR.
<Profpatsch> But very awesome that this works!!
<Profpatsch> Is it updated as master moves?
<Profpatsch> grahamc: Do you think it’s possible to get something like mention-bot running again?
<Profpatsch> I really miss that.
<grahamc> Yeah I’m not sure. It is possibly too noisy. It isn’t updated when master moves. Just when the pr is updated. The 0 tag was to indicate it has finished looking at the PR (before you couldn’t know if it looked or was 0)
<Profpatsch> PR update should work, too. Maybe add a batch job that regularly updates all stale PRs?
<Dezgeg> well I doubt the amount of rebuilds usually changes that much even if master moves
<Profpatsch> grahamc: Is it possible to integrate it into the PR like e.g. Travis is? With an up-to-date thingy at the bottom of the PR?
<Profpatsch> Dezgeg: You might be right.
<Dezgeg> it is actually there already in some form: grahamcofborg-eval — Evaluation checks OK
<Profpatsch> Hm, is there actually a way for the amount of packages affected to go up (apart from new packages)?
<Profpatsch> Maybe it really doesn’t matter at all.
<Dezgeg> if someone pushes to master something adds that package as a dependency of some other huge package
<Profpatsch> yeah, right.
<Profpatsch> Maybe batch-check stuff once a week?
<Profpatsch> And move the “was checked by bot” tag to the bottom of the PR, this way the 0–10 tags could be removed.
<Profpatsch> If one of those deps is Chromium or Firefox, that’s still of concern of course.
<Profpatsch> phew, I think the int.between PR is finally ready for merge.
<Profpatsch> Next up would be restricting all the options to more specific types.
<Profpatsch> At least to have some nice examples.
<grahamc> Profpatsch I could add a rebuild count to the build success / fail result, sure, and drop the 0 tag and 1-10 tag,,, but although seeing either of those tags means it is easy.
<grahamc> Or easier to merge
<Profpatsch> Also true!
<Profpatsch> Nice, actually.
<Profpatsch> So scrap that, I like that.
<grahamc> Yeah. Okay :) we’ll try it a while and then change if we still don’t like it
orivej has joined joined #nixos-dev
orivej has quit [(Ping timeout: 240 seconds)]
<MoreTea> Profpatsch: I've got code that will extract the maintainers from pkgs/*.
<MoreTea> I wanted to wait with a mentionbot until the maintainers RFC was done (then nixbot could @-mention as well), but based on that expression I could whip up a simple bot
<LnL> that would be nice, I miss the mention bot
<pierron> Profpatsch: I will try to review the int types tonight or tomorrow.
<pierron> grahamc: I like the rebuild amount reports. Still, I am not sure about the colors, as today's colors are matching the prefix numbers.
taktoa has quit [(Remote host closed the connection)]
<Profpatsch> pierron: thanks
<domenkozar> pierron: can overlays specify lib?
<domenkozar> so that it's overridden for lib argument in nixos moduels as well?
lo1tuma has joined joined #nixos-dev
<pierron> domenkozar: use pkgs.lib, the lib given as argument of the modules are made for the module system.
<pierron> domenkozar: another option would be for you to do: _modules.args.myLib = pkgs.lib;
<pierron> domenkozar: which will add the myLib argument to every module.
<domenkozar> uh, so many indirections
<pierron> domenkozar: the problem you see is that NixOS is asking for a specific Nixpkgs version, and the lib is the version of the Module system, which happen to be loaded before.
<pierron> domenkozar: we have no garantee over the lib that we will get from the Nixpkgs version that you are loading.
<aszlig> domenkozar: another way would be to call the module system with your custom lib
<aszlig> like import <nixpkgs/nixos/lib/eval-config.nix> { ...; lib = foo; ... }
<pierron> aszlig: but then you evaluate Nixpkgs twice.
<aszlig> oh, wait... that won't work
<pierron> aszlig: if you have overlays.
<domenkozar> well can't do that with nixops
<pierron> aszlig: unless you replace the nixpkgs computation as well.
<domenkozar> at least not currently :)
<pierron> domenkozar: another option mentioned by niksnut, was to just use a different name than lib, for all the functions you want to add.
<domenkozar> maybe extraLib
<domenkozar> or extras
jtojnar has quit [(Quit: jtojnar)]
olto has quit [(Remote host closed the connection)]
jtojnar has joined joined #nixos-dev
olto has joined joined #nixos-dev
<jtojnar> can I get a hydra job for https://github.com/NixOS/nixpkgs/pull/30371?
orivej has joined joined #nixos-dev
Sonarpulse has joined joined #nixos-dev
FRidh has quit [(Quit: Konversation terminated!)]
lo1tuma has quit [(Quit: Page closed)]
<copumpkin> globin: you figure out what to do about the old ruby and erlang for openssl 1.1?
<orivej> Is there something like `nix why-depends nixpkgs.pkg nixpkgs.dep` that can tell why the build of pkg depends on dep? (nix why-depends tells this about the runtime closure, and hence needs to build the pkg to answer this query)
mbrgm has quit [(Quit: ZNC 1.6.5 - http://znc.in)]
<Profpatsch> zimbatm: Ah, my inquiry was about unicode inside nixpkgs doc strings.
<Profpatsch> Ah, not about unicode, about `'
<zimbatm> yeah the RFC is about nix specifically so it might not apply there
<zimbatm> I don't know why we should complicate things with un-standard quoting that being said
<copumpkin> yeah I guess he squashed it
* copumpkin shrugs
<copumpkin> either way, yay
ckauhaus has quit [(Quit: Leaving.)]
<zimbatm> "Edolstra retry" is a special type of retry
<copumpkin> he gives you one look and it makes you really want to succeed the second time around
<zimbatm> your first try actually got reverted and and you want to quit programming
__Sander__ has quit [(Quit: Konversation terminated!)]
mbrgm has joined joined #nixos-dev
goibhniu has quit [(Ping timeout: 240 seconds)]
<copumpkin> :)
<Dezgeg> do we name python packages with capitals if the upstream name starts with them? e.g. do I set 'pname = "PyAMF";' or translate to lowercase?
<gchristensen> Dezgeg: found the problem >.>
<gchristensen> if ($count > 1) { -> if ($count > 0) {
<Dezgeg> hah
<orivej> Dezgeg: we keep pname from pypi; won't it fail to download otherwise?
<gchristensen> Dezgeg: https://github.com/NixOS/nixpkgs/pull/31183 look at the Checks on this
<Dezgeg> well I could pass a different pname to fetchpypi
<Dezgeg> but I will keep PyAMF, thanks
<copumpkin> I think we keep uppercase
<LnL> yeah, we keep the casing for the name but the attributes are usually lowercase IIRC
mbrgm has quit [(Ping timeout: 252 seconds)]
mbrgm has joined joined #nixos-dev
<Mic92> if you merge enough pull requests on github, you get invited to https://github.com/maintainers
vcunat has joined joined #nixos-dev
<vcunat> I can't see any logs recorded from this channel - is there some particular reason? I guess we could setup bobot.me just as for #nixos.
MichaelRaskin has joined joined #nixos-dev
<aminechikhaoui> vcunat: good idea +1
<globin> vcunat: I had requested it but didn't receive any anwser
<globin> vcunat: feel free to do so again :)
<vcunat> globin: how long ago was that?
__Sander__ has joined joined #nixos-dev
taktoa has joined joined #nixos-dev
orivej has quit [(Ping timeout: 260 seconds)]
olto has quit [(Quit: hasta luego)]
<globin> vcunat: when this channel was created, so quite some time now
<> changed the topic of #nixos-dev to: Topic for #nixos-dev is "NixOS Development | Blockers for 17.09: https://github.com/NixOS/nixpkgs/projects/4 | https://registration.nixcon2017.org/ | https://hydra.nixos.org/jobset/nixos/trunk-combined | https://channels.nix.gsc.io/graph.html"
<> changed the topic of #nixos-dev to: Topic set by fpletz!~fpletz@unaffiliated/fpletz on 2017-08-28 21:18:18 UTC
JosW has quit [(Quit: Konversation terminated!)]
vcunat-foo has joined joined #nixos-dev
<copumpkin> Mic92: damn, I'm not cool eough for that
vcunat-foo has left #nixos-dev []
<vcunat> how do I find which IRC server this is on?
<Sonarpulse> I'm still unsure about my targetPackges.stdenv.cc thing
<Sonarpulse> if someone e.g. wanted to use a non-standard compiler or linker
<Sonarpulse> lldd
<Sonarpulse> clang on linux
<Sonarpulse> etc
<Sonarpulse> should other packages who *runtime* depend on cc / ld respect that choice?
<vcunat> Sonarpulse: actually, did you discuss the cross-* changes on the hackathlon?
<Sonarpulse> vcunat: yes!
<Sonarpulse> should have slides up soon
<vcunat> but no recording, right?
<Sonarpulse> vcunat: yes, no recording sadly
<vcunat> (there seemed no setup for such things)
<Sonarpulse> well it was an unofficial talk
orivej has joined joined #nixos-dev
<Sonarpulse> I suppose I should have recorded it on my own laptop or something
<vcunat> :-) slides will be OK, I think
<vcunat> and later they might get coverted to some kind of docs
<vcunat> (depending what was their focus... I'm thinking too forward)
<Sonarpulse> vcunat: well, we have some docs already :)
<Sonarpulse> but yes pictures could be cool thing to add
<sphalerite> Or one of us could have recorded it on a phone or whatever
<Sonarpulse> (20 years later bootleg surfaces!)
<MichaelRaskin> Two of them
<MichaelRaskin> One audio-only, and one video with unusably noisy audio
<MichaelRaskin> And for some reason nobody can make these two sync.
<Sonarpulse> lols
<Sonarpulse> vcunat: btw
<Sonarpulse> one thing I did vaguely mention
<Sonarpulse> but didn't include in slide
<Sonarpulse> stdenv bootstrapping:
<Sonarpulse> stage = buildTimePkgs: runTimePackages: .....;
<Sonarpulse> final = stage (stage (final // bootstrapToolsPkgs) final) final;
<vcunat> well, that's the same way we stage nixpkgs + overlays, right?
<Sonarpulse> err overlays are right now applied to every stage I think
<Sonarpulse> I think some further tweaking is needed
<Sonarpulse> but I think we could automatically collapse all our current boostrap stages that way
<vcunat> stdenv bootstrapping is collapsed now, I think
<Sonarpulse> hmm? still like 4-5ish stages on linux and darwin, no?
<vcunat> it's one stage in "nixpkgs bootstrapping"
<vcunat> oh, now I see, you meant something else
<Sonarpulse> bootstrapToolsPkgs is supposed to litterally be stub derivations of the boostrap tools tarball
<Sonarpulse> nothing more
<vcunat> The reason we do the stages is that it's not possible to bootstrap so easily.
<vcunat> (At least I think so.) There are sometimes various version dependencies, so you have to add layers for them.
<Sonarpulse> vcunat: well right now we don't have quite the clean separation of run time and build time
<Sonarpulse> of dependencies
<Sonarpulse> I think that is the key thing
<vcunat> in linux stdenv we have allowedRequisities on the end
<Sonarpulse> right but this would be for every package doing that
<vcunat> but otherwise there's little separation, in general
<Sonarpulse> based on buildInputs vs nativeBuildInputs
taktoa has quit [(Remote host closed the connection)]
<Sonarpulse> so what I wrote is guaranteed to contain no references to bootstrapTools at the end
<Sonarpulse> but doesn't seem to me quite minimal in terms of cost just yet
<vcunat> Normally we don't bootstrap often, so if it doesn't add an extra compilation of gcc or something that big...
<Sonarpulse> vcunat: today we end up with the set of build time and run time pkgs being the same
<Sonarpulse> which my fixpoint doesn't do either
<Sonarpulse> i.e. buildPackages = pkgs
<Sonarpulse> vuncat: ah!
<Sonarpulse> I need an intersect with
<vcunat> Sonarpulse: how do you plan to ensure the separation?
<vcunat> just via allowedReferences?
<Sonarpulse> vcunat: yes
<Sonarpulse> vcunat: cause cross
<Sonarpulse> all the infra to pass in two coppies has been there for a year
<Sonarpulse> just need to enforce
<Sonarpulse> vcunat: additionally be careful about -L and PATH
<Sonarpulse> which is already also implemented
<Sonarpulse> in my cross-elegant PR
<vcunat> I'm not certain if it's always possible to separate the two
<vcunat> like shebangs in scripts that are meant to be used during build-time by reverse dependencies
<Sonarpulse> vcunat: cross at least depends on it :D
<Sonarpulse> not sure what your example means
<vcunat> it's an example where I need a runtime reference from output to a native build-time input
<Sonarpulse> so the same script is used during the buid
<Sonarpulse> *buidl
<Sonarpulse> **build
<Sonarpulse> and put in the output?
<vcunat> no, let's say the package wants to create some generator script
<vcunat> its shebang should be patched to native bash (for example)
<vcunat> that's a build-time-only input, but it will be referenced from $dev
<vcunat> globin: This channel is already in the request queue. Please be patient while we process the request.
<vcunat> :-/
<Sonarpulse> vuncat: well stepping back
<Sonarpulse> for non-cross supporting packages
<Sonarpulse> one can add both kinds of dep of the same package
<Sonarpulse> and then assert that they are in fact the same derivation
<Sonarpulse> which will, rightly, cause eval problems with cross
<vcunat> yes
<vcunat> but maybe we should make $dev an exception
<Sonarpulse> bootstrapToolsReplacements = intersectAttrs bootstrapToolsReplacements (stage (bootstrapToolsReplacements // bootstrapToolsPkgs) bootstrapToolsReplacements);
<vcunat> At least after we solve the problem that restricting of references can't be done per-output in nix...
<Sonarpulse> vcunat: ah, I see
<vcunat> well, we don't need a perfect solution in the first iteration
<Sonarpulse> final = stage final final // bootstrapToolsReplacements;
<Sonarpulse> that right there is correct
<Sonarpulse> and near optimal
<Sonarpulse> deps of bootstrap tools that aren't bootstrap tools
<Sonarpulse> will be needlessly rebuilt
<Sonarpulse> but that's it
<jtojnar> vcunat: could you pleas estabilish a hydra jobset for https://github.com/NixOS/nixpkgs/pull/30371
<vcunat> jtojnar: right now Hydra's slightly overloaded, but I'll do it within a few days
<vcunat> the branch isn't missing too many commits from current master or staging, right?
<Mic92> copumpkin: it is not that interesting actually. Just a private repos, where people ask questions about how to deal with open-source contributions.
<jtojnar> vcunat: I rebased it onto https://hydra.nixos.org/eval/1406736
<vcunat> OK
vcunat has quit [(Quit: Leaving.)]
__Sander__ has quit [(Quit: Konversation terminated!)]
Sonarpulse has quit [(Ping timeout: 240 seconds)]