worldofpeace changed the topic of #nixos-dev to: NixOS Development (#nixos for questions) | NixOS 20.09 Nightingale ✨ | | | 20.09 RMs: worldofpeace, jonringer |
copumpkin has quit [Quit: Bye!]
copumpkin has joined #nixos-dev
pmy has quit [Ping timeout: 265 seconds]
pmy has joined #nixos-dev
ris has quit [Ping timeout: 246 seconds]
evanjs has quit [Read error: Connection reset by peer]
evanjs has joined #nixos-dev
Emantor has quit [Quit: ZNC -]
Emantor has joined #nixos-dev
cjb has quit []
pmy has quit [Quit: WeeChat 3.1]
pmy has joined #nixos-dev
Scriptkiddi has quit [Quit: Bridge terminating on SIGTERM]
ajs124 has quit [Quit: Bridge terminating on SIGTERM]
das_j has quit [Quit: Bridge terminating on SIGTERM]
ajs124 has joined #nixos-dev
Scriptkiddi has joined #nixos-dev
das_j has joined #nixos-dev
andi- has quit [Remote host closed the connection]
andi- has joined #nixos-dev
justanotheruser has joined #nixos-dev
cole-h has joined #nixos-dev
pie_ has quit [Read error: Connection reset by peer]
pie_ has joined #nixos-dev
evils has quit [Ping timeout: 260 seconds]
pmy has quit [Quit: WeeChat 3.1]
pmy has joined #nixos-dev
cole-h has quit [Ping timeout: 240 seconds]
Mic92 has quit [Quit: WeeChat 3.1]
Mic92 has joined #nixos-dev
FRidh has joined #nixos-dev
justanotheruser has quit [Ping timeout: 258 seconds]
AlwaysLivid has joined #nixos-dev
jonringer has quit [Ping timeout: 258 seconds]
<siraben> huh, nixpkgs-unstable has been blocked for 9 days
aaronjanse has joined #nixos-dev
rajivr has joined #nixos-dev
justanotheruser has joined #nixos-dev
orivej has quit [Ping timeout: 240 seconds]
hax404 has quit [Remote host closed the connection]
hax404 has joined #nixos-dev
xwvvvvwx has quit [Ping timeout: 246 seconds]
xwvvvvwx has joined #nixos-dev
<sterni> yeah :|
<sterni> what's the current status? last time I checked it was mostly aborted jobs?
<domenkozar[m]> says that tests passed 10h ago
<sterni> the commit unstable is at is almost exactly one week old
<domenkozar[m]> that means that channel scripts haven't kicked in yet
<domenkozar[m]> maybe we should migrate those to github actions so they are more transparent?
<sterni> ah nvm the commit was merged into master on the 30th
<sterni> so the 3 days ago on is accurate
orivej has joined #nixos-dev
orivej has quit [Ping timeout: 265 seconds]
<sterni> uh so the geary test failure currently failing the trunk-combined jobset:
<sterni> Expected is "0;;Comic Sans;144;rgb(255, 127, 1)" and actual (I've inserted a print to display this) is "0;;Comic Sans;150.39404296875;rgb(255, 127, 1)"
ris has joined #nixos-dev
<sterni> really weird
<sterni> probably something else changed that broke it
<sterni> might be time to bisect
<flokli> Huh?
<sterni> well I'd think the test was passing on wednesday in nixpkgs
FRidh has quit [Quit: Konversation terminated!]
AlwaysLivid has quit [Ping timeout: 240 seconds]
AlwaysLivid has joined #nixos-dev
justanotheruser has quit [Ping timeout: 258 seconds]
__monty__ has joined #nixos-dev
stigo has quit [Ping timeout: 240 seconds]
<hexa-> sterni: I disabled that test yesterday
<sterni> oops
<sterni> should've pulled
<hexa-> sterni: I guess things are not moving forwards because the darwin builders aren't doing jack shit
<hexa-> mac7-guest ist the only one working
<hexa-> meanwhile ~50k queued jobs for darwin
<sterni> hexa-: missed you PR because I was to lazy to remove thi is:open from the search lol
<hexa-> :p
<sterni> hexa-: I'm not sure if that would affect starting a nixos/trunk-combined/tested job?
<sterni> > last eval was 23h ago
<{^_^}> value is a function while a list was expected, at /var/lib/nixbot/nixpkgs/master/repo/lib/lists.nix:609:18
<hexa-> right
<hexa-> well, geary was one blocker at least
<sterni> hmm there is still one job queued for the last eval
<sterni> if we kill that one it should start building a new one right?
<hexa-> huh, it should at least finish this eval
<hexa-> looks like its in a screendump loop
<hexa-> either way, a new eval should start any within the hour
<hexa-> so geary and linux_5_10 on i686
<sterni> hexa-: on the linux thing there's a pr (filter by severity channel blocker)
<hexa-> thx
orivej has joined #nixos-dev
stigo has joined #nixos-dev
srk has quit [Ping timeout: 240 seconds]
stigo has quit [Ping timeout: 265 seconds]
stigo has joined #nixos-dev
orivej has quit [Ping timeout: 265 seconds]
xwvvvvwx has quit [Quit: ZNC 1.8.2 -]
xwvvvvwx has joined #nixos-dev
zimbatm has joined #nixos-dev
xwvvvvwx has quit [Quit: ZNC 1.8.2 -]
xwvvvvwx has joined #nixos-dev
aminechikhaoui has quit [Quit: Ping timeout (120 seconds)]
aminechikhaoui has joined #nixos-dev
FRidh has joined #nixos-dev
evils has joined #nixos-dev
cole-h has joined #nixos-dev
jonringer has joined #nixos-dev
cole-h has quit [Ping timeout: 265 seconds]
Baughn_ has joined #nixos-dev
Baughn_ has quit [Quit: ZNC 1.6.2+deb1 -]
Baughn has quit []
Baughn has joined #nixos-dev
abathur has quit [Quit: abathur]
orivej has joined #nixos-dev
AlwaysLivid has quit [Ping timeout: 240 seconds]
rajivr has quit [Quit: Connection closed for inactivity]
FRidh has quit [Quit: Konversation terminated!]
capisce has quit [Ping timeout: 240 seconds]
<ajs124> what kind of license is "BSD 4-Clause University of California-Specific" and do we have anything for that?
capisce has joined #nixos-dev
<ajs124> and why do people keep asking me in reviews if something is *Only or *Plus, when half the time, the answer is neither, both or something completely different?
<hexa-> because they see gpl and no qualifier?
<samueldr> ajs124: because gpl without the qualifier "is not a thing"
<ajs124> samueldr: ah, ok. so why were we using that?
<samueldr> because it wasn't "known" by the people who introduced the licenses infra in Nixpkgs
<ajs124> ah ok, that makes sense
<ajs124> hexa-: but how am I supposed to know what license something is under? It's not like I really care, most of the time. Maybe I should, but I don't.
<ajs124> anyways, the only mention of "University of California" in nixpkgs is apparently a patch against dbus that was committed in 2013
<samueldr> ajs124: link to the license file
<samueldr> there's a trick I want to try before telling you to "just do ......"
<ajs124> apparently util-linux is under 7 different licenses
<hexa-> nice
<samueldr> right, so it's already SPDX identifiers, so that helps
<samueldr> since spdx identifiers are used in here
<samueldr> so it looks like it'll need the license to be added
<ajs124> I just wanted to update util-linux :'(
<samueldr> yeah, it can get annoying to add those metadata information points :(
capisce has quit [Ping timeout: 240 seconds]
capisce has joined #nixos-dev
<ajs124> samueldr: should I call the attribute bsdOriginalUC or bsd4UC? because we have BSD-4-Clause as bsdOriginal
<samueldr> (I don't *know*)
<ajs124> does anybody know?
jonringer has quit [Ping timeout: 258 seconds]
capisce has quit [Ping timeout: 265 seconds]
<ajs124> qyliss: you're listed as the code-owner of lib/licenses.nix, right? do you have an opinion on bsdOriginalUC vs. bsd4UC?
<ekleog> does anyone know of a tool to find the spdx identifier from the LICENSE file? that might be interesting
dotlambda has joined #nixos-dev
tomberek has joined #nixos-dev
jess has quit [Quit: updates]
orivej has quit [Ping timeout: 240 seconds]
<ekleog> answering myself: looks like `licensee` does that, so a tool could be auto-checking the set licenses and suggesting improvements… dunno when I'll find motivation to do that myself though
<supersandro2000> ajs124: look at any header or mention of the license except that license file and if it mentions there "or later" it is Plus otherwise it is only
<supersandro2000> ekleog: does not work for Only or Plus
<ekleog> oh :(
<supersandro2000> ekleog: can you link the program? is not googleable
<samueldr> supersandro2000: how does it not work? there's a SPDX identifier for each
<supersandro2000> samueldr: people don't use spdx identifier all the time in their code
<supersandro2000> maybe it works but if it only looks at the license file then it won't
<supersandro2000> because the files are the same for gpl*Only and gpl*Plus
<samueldr> but this tool looks at licenses for that
<samueldr> ah
<supersandro2000> I can't find the cool
<supersandro2000> *tool
<supersandro2000> the name is so super generic
<ekleog> supersandro2000: yup! ; found it via the nixpkgs package search
<ekleog> looks like for glibc it doesn't detect the fact that it's *plus
<ekleog> so :(
<ekleog> that being said it'd still be an ok sanity-check I guess :)
<samueldr> ekleog: what does it say for which has three licenses?
<samueldr> (not yet packaged)
<samueldr> (literally packaging this)
<ekleog> samueldr: didn't try on it as i don't have a local checkout (you can try `nix-shell -p licensee --run 'licensee detect .'` I guess if you're touching it?), but on that also has 3 licenses it does find the 3 licenses
<supersandro2000> the tool can help but it is not precise enough for automatic testing
<samueldr> yeah
<ekleog> definitely couldn't do more than raise flags of “beware this might be wrong” or “maybe you want to do x” but it'd probably already help :)
beertoagunfight has joined #nixos-dev
<ekleog> (opened to track my doing it in the tool i'm currently working on, think it'd be reasonable there ^^)
<{^_^}> Ekleog/nixpkgs-check#13 (by Ekleog, 16 minutes ago, open): Use the `licensee` ruby program to auto-detect licenses in `src` and check that they match `meta.license`
<ekleog> (but no promises, I'm probably going to focus on getting rid of the PR template checklist before adding new stuff)
<supersandro2000> this should be done in nixpkgs-hammering or in an easy to use scripts that I can wire into nixpkgs-review-checks or rmcgibbo[m] into his bot
abathur has joined #nixos-dev
jess has joined #nixos-dev
bpye has quit [Ping timeout: 240 seconds]
<sterni> ekleog: plus vs. only is *super* hard to detect
<sterni> ekleog: because the license file is the exact same one it depends on the source files' headers what applies as stated in the license
<sterni> ekleog: which is made even worse because a lot of ppl don't bother with adding GPL source file headers
<ekleog> afaiu nixpkgs-hammering can't be interactive, while this sounds exactly like something that would be better as interactive given the likelihood of failure.
jonringer has joined #nixos-dev
<sterni> I've done some experiments with running licensee automatically on source tarballs of nixpkgs packages and the results were a bit disappointing because the tool is not *that* insightful
<sterni> I mean it does it's best
<sterni> it is generally not that simple to figure out the license of a given repo
<sterni> but licensee has a good json output so you can integrate it into stuff pretty well at least
<sterni> we should probably try it out at some point for nixpkgs in some capacity
<ekleog> sterni: oh interesting, thank you, didn't know about that :) so I guess I can continue not caring and landing stuff so long as we don't introduce regression and the license (as in without caring about *Only or *Plus) is correct
<sterni> ekleog: well it's not that easy, figuring out whether it is only or plus is actually somewhat important
<sterni> ekleog: because of license compatibility issues
<ekleog> sterni: actually you come at exactly the right time: ISTR you maintain some haskell packages, right? do you know how/where I should add a passthru.tests = [ nixosTests.agda ] for Agda?
<sterni> if you have a gpl 2 program which links in parts of a gpl 3 program this is _only_ legal if it is licensed under gpl 2 or later in which case the product of linking together would effectively be gpl 3 because you need to observe the patent clause as well etc
<sterni> as always it's complicated :|
<ekleog> afaik there are no license compatibility issue, only programs that end up being under some combination of licenses (that possibly isn't possible to actually fulfill)
<sterni> ekleog: just a second let me check, probably they reexpose agda
<sterni> ekleog: gpl 2 and gpl 3 are incompatible (at least in one direction)
<samueldr> so uh, with something using gsettings... how do I get a gschemas.compiled when the build system doesn't make one?
<supersandro2000> ekleog: why should it be interactive? either it is 100% sure or I need to dig the source anway
<sterni> ekleog: pkgs/build-support/agda/default.nix is the more appropriate place because that is where agdaPackages.agda is defined
<supersandro2000> sterni: we don't care about linking. we take what upstream says and don't verify that
<sterni> it reexposes haskellPackages.Agda with some extra stuff
<sterni> supersandro2000: well it is an issue we need to be aware of and sometimes take action about
<ekleog> sterni: ohh yes and it looks like the passthru.tests is already there. awesome, thank you!
<sterni> supersandro2000: if we build packages that have license incompatibilities and distribute them in binary form that is our responsibility and a gpl violation caused by us
<sterni> supersandro2000: the problem is that in source form almost everything is fair game afaiu it only gets tricky with gpl stuff when you start building and distributing binaries (which we do)
<sterni> if you want to see something cursed you can check out what I had to do in the inspircd derivation I guess :p
bpye has joined #nixos-dev
<ekleog> also, I don't care enough to stay in the discussion, but the summary of my position wrt. licenses is that we should have them not as a parameter on the derivation but as a parameter on the src, and then have derivation auto-compute a spdx license expression based on all their inputs; which would solve the issue of incompatibilities by just saying “well given the things made… it's basically
<ekleog> like unfree”
<gchristensen> licenses are more complicated than that, a GPL input doesn't necessarily imply a GPL output
<sterni> yep it doesn't work out like that unfortunately
<MichaelRaskin> Because, for example, otherwise glibc would explode the entire computation
<ekleog> that's true, and implies that we may have to have additional input types if we want to actually be precise in the auto-computation
<gchristensen> if we want to get good at licenses we basically must have a legal team
<ekleog> (but eg. gcc, which I guess is the core example, being in nativeBuildInputs can be more or less known to not be linked to the output and thus to not be GPL-viral — that said we're not providing legal advice and shouldn't be)
<gchristensen> it is our responsibility to follow the licenses
<ekleog> (FWIW, exherbo had a nice exchange on their ML with some nice suggestions about what dependency kinds should exists, that both gave better names to nativeBuildInputs & co, but also would have helped identify that eg. an input is used only copy-pasted as doc into another input and thus would not trigger contaminate-when-linked licenses)
<MichaelRaskin> ekleog: note that technically speaking a native input called cross-compiler can in principle have _target_ arch stdlib in it, and there is a lot of interesting stuff that can happen about the license of this target stdlib
<gchristensen> and a license can say whatever it wants
<ekleog> MichaelRaskin: that's right, and this case is a good demonstration of why my previous message would be useful :° (the things would iirc be something like `executedInputs`, `linkedInputs`, `copiedInputs`, `referencedInputs` etc.)
<ekleog> (but also it is a good demonstration that a single meta.license is not enough to handle a derivation's license, because often two different licenses will be applied for two different parts of a single derivation, eg. in your example)
<MichaelRaskin> gchristensen: are you hinting at «… and then in the context of different legal systems, two licenses could combine in different ways for US and for EU»?
<gchristensen> haha well I didn't mean to but I suppose it surely could :D
<ekleog> (which leads me to “meta.license is just an indication anyway, at least until we pour much more effort in there, and thus caring about *Only vs. *Plus is very much pointless nitpicking” :°)
<sterni> also exherbo doesn't provide prebuilt binaries right?
<gchristensen> more the nature of trying to write a evaluator for a turing complete programming language whose only approved interpreters are very expensive
<sterni> ekleog: yeah that's why we often have lists of licenses in meta.license
<ekleog> sterni: nope, and actually the thing I'm referring to was designed for better cross-building but by its nature could also help in giving better hints about what the resulting license might be :)
<MichaelRaskin> gchristensen: let's call a spade a spade: which has more reproducibility problems than a typical build system
<ekleog> As for lists of licenses, interesting, don't think I noticed it yet though it being a list definitely striked me. Is it supposed to mean AND, OR, or “some files are under one and some files under the other”?
<gchristensen> yeah, lawyers are tricky beasts
<ekleog> oh you were talking about lawyers for the interpreters? i thought you were talking about courts :°
<samueldr> ekleog: all of the above, inclusively and exclusively
<ekleog> nice :D
<samueldr> AND, OR, some files, all files, yes, no, maybe
<MichaelRaskin> ekleog: the judges deciding the legal interpretation of the wording of the licenses are usually lawyers by education
<ekleog> so i'll probably start caring even less so long as it's not a difference between redistributable and not i guess ^^'
<ekleog> MichaelRaskin: interesting, in france afaik the “normal path” is there is definitely a common education for the first few years but then judges have one exam and then another education, while lawyers have another unrelated exam and probably another education (thouhg i know more about judges than lawyers so that second part might be wrong)
<sterni> funnily enough freeNonRedistributable exists in a sense
<gchristensen> our "free" / "unfree" distinction is totally bogus
<ekleog> special points for unfreeRedistributableFirmware being “free”
<supersandro2000> sterni: I'd love to give you that responsibility :)
<supersandro2000> yeah a bit overengineered
<samueldr> it feels to me like half of cmake-using things I package end up with doubled-up prefix in paths it hardcodes into the software
<samueldr> is there a known common root cause to fix that?
<samueldr> e.g. it hardcodes /nix/store/.../nix/store/.../actual/path
<supersandro2000> wild theory: if you dig deep enough you will find all sorts of linking issues deep in every distro
<sterni> I think up to a certain point you can get away with murder in terms of GPL violations
<supersandro2000> if some problem comes up we can take care of it :)
<sterni> inb4 GPL violations are so complicated that a law firm specialized in suing open source projects would make a loss
<sterni> or there's not enough money to be taken
<supersandro2000> well, lets talk about real problems
<supersandro2000> vim incsearch does not work in WSL and breaks / :(
<MichaelRaskin> Wait but how WSL manages to break specifically that without breaking the basics of motions in vim?
__monty__ has quit [Quit: leaving]
<gchristensen> usually license/GPL violations are taken care of before the lawsuit
<supersandro2000> MichaelRaskin: 🤷 I have no clue
<sterni> gchristensen: unless somebody has ill intentions towards you but I don't think these kinds of ppl have found gpl violations yet
<sterni> it probably doesn't make to much sense for them anyways
<gchristensen> there are definitely gpl license trolls :)
<gchristensen> one of the nixos foundation's board members does license compliance professionally and deals with them
<sterni> 😑
<supersandro2000> gchristensen: is it you?
<ekleog> ._.
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-dev