<pbogdan> any reason why https://nixos.org/nixos/download.html would be out of sync with the latest release of 17.09 ?
<gchristensen> niksnut: I'm pretty sure the NixOS.org website is failing to update
<niksnut> fixed
<gchristensen> thanks!
<gchristensen> pbogdan: ^
ma27 has quit [Ping timeout: 252 seconds]
phreedom has quit [Ping timeout: 240 seconds]
mbrgm_ has joined #nixos-dev
mbrgm has quit [Ping timeout: 264 seconds]
mbrgm_ is now known as mbrgm
<pbogdan> gchristensen: niksnut: thanks!
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-dev
<Sonarpulse> why don't we use the "--package-db" command for ghc?
pie_ has quit [Ping timeout: 248 seconds]
orivej has quit [Ping timeout: 256 seconds]
_rvl has quit [Quit: ZNC 1.6.5 - http://znc.in]
_rvl has joined #nixos-dev
goibhniu has joined #nixos-dev
goibhniu has quit [Remote host closed the connection]
goibhniu has joined #nixos-dev
a_lebedev has joined #nixos-dev
ma27 has joined #nixos-dev
a_lebedev has quit [Ping timeout: 264 seconds]
a_lebedev has joined #nixos-dev
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-dev
goibhniu has quit [Remote host closed the connection]
goibhniu has joined #nixos-dev
michaelpj_ has joined #nixos-dev
michaelpj_ has quit [Remote host closed the connection]
michaelpj_ has joined #nixos-dev
a_lebedev has quit [Quit: leaving]
michaelpj has quit [Quit: ZNC 1.6.5 - http://znc.in]
michaelpj has joined #nixos-dev
__Sander__ has joined #nixos-dev
zarel has joined #nixos-dev
pie_ has joined #nixos-dev
pie_ has quit [Ping timeout: 240 seconds]
<domenkozar> is there a way to setup (reverse) dns for nix sandbox?
<domenkozar> looking if some package already does that - for checkPhase
<gchristensen> sorry, what do you want to do?
FRidh has joined #nixos-dev
orivej has joined #nixos-dev
<domenkozar> gchristensen: be able to do dns lookups in sandboxed derivation build
<domenkozar> specifically: tests use reverse dns
<domenkozar> it would be nice if nix created nsswitch.conf to enable reverse dns lookups
<gchristensen> hm not sure I'd like my builds to be able to do dns lookups in the sandbox
<domenkozar> what if that was limited to /etc/hosts?
<domenkozar> so you can only do it for 127.0.0.1?
<gchristensen> would it leak my hostname?
<gchristensen> would it leak ipv6 vs. ipv4 support? (localhost can point to one or the other or both) (these aren't arguments against it specifically, just curious)
<domenkozar> I would imagine it uses /etc/hosts in reverse
<domenkozar> which is created by nix
<gchristensen> ahh you're right
<domenkozar> so it would be deterministic and no leak anything afaik.
pie_ has joined #nixos-dev
<gchristensen> neat
<gchristensen> protocols and services ...
<gchristensen> I'm thinking more and more this shouldn't be in Nix
<gchristensen> but implemented as part of your build
<domenkozar> that would require special nix.conf for each user :/
<domenkozar> if you refer to extra sandbox files
<gchristensen> don't make it impure :)
<domenkozar> gchristensen: why do you think it shouldn't be in Nix
<domenkozar> if it's deterministic?
<gchristensen> brb
__Sander__ has quit [Quit: Konversation terminated!]
romildo has joined #nixos-dev
vcunat has joined #nixos-dev
__Sander__ has joined #nixos-dev
michaelpj_ has quit [Ping timeout: 240 seconds]
<gchristensen> well the problem domenkozar is those files change not infrequently, and adding them via nix seems a bit strange, when it could be handled just fine by adding the protocols and services files in your build
<gchristensen> so I think the lifecycle of those files don't match the lifecycle of the Nix program. additionally, they _change over time_ and your build maybe won't work properly if those files change between builds, so I think the data appropriately belongs as an input to your build
<domenkozar> gchristensen: does it change that often?
<domenkozar> I see, it's in iana-etc package
<gchristensen> it certainly changes "periodically" but here's the definite kicker: your /etc/services and /etc/protocols is not the same as mine :)
<domenkozar> I think we could have /etc/nsswitch setup statically (I really care about reverse dns)
<gchristensen> what happens when that needs to change too? https://github.com/NixOS/nixpkgs/pull/30472
<domenkozar> gchristensen: my point is, it worked so far
<domenkozar> by adding a few lines we can improve support for some tests that are currently disabled
<gchristensen> aye
<gchristensen> is it possible to set it up inside the build, though?
<gchristensen> like maybe we could have a package with an activation hook which sets up rdns
<domenkozar> next step could be for nix to expose setting these as attributes, but we'd have to think about security, etc
<gchristensen> true
<gchristensen> its not clear to me why the solution is in Nix though
<domenkozar> I prefer to improve determinism globally for nix except per package
* domenkozar remembers the days when LOTS of packages needed -lgcc_s
<gchristensen> sure ... but we could put it in stdenv
<domenkozar> I bet dogecoin eelco will be against adding a dns package there
<gchristensen> haha fair
<domenkozar> I see your point, don't get me wrong :)
<gchristensen> I don't know the solution :P this is above my pay grade
<domenkozar> lol
<domenkozar> dammit, haskell people tell me it's easier to disable tests, nix people tell me we should modify stdenv
<domenkozar> all I wanted was to run tests :P
<gchristensen> haha
<gchristensen> it sounds like the _first_ step is to try it in via what nix already provides, in your package
<domenkozar> not sure I can
<domenkozar> can't bind to 53
<domenkozar> port 53
<domenkozar> can't modify /etc/nsswitch.conf
<gchristensen> ahhhhhhhh
<domenkozar> best I can think of is nix.chrootDirs
<domenkozar> but then I'll get reports sometimes tests fail
<domenkozar> when folks won't bootstrap nix correctly
<__Sander__> what about running the testcases in a VM?
<__Sander__> in a Nix expression
<__Sander__> by using the NixOS test driver facilities?
<gchristensen> that seems promising
<__Sander__> that's one of the main reasons the NixOS test driver was invented
<__Sander__> I, for example, use it quite extensively to test disnix
<gchristensen> fwiw it isn't uncommon for people to put custom stuff in services, and from there it wouldn't be unreasonable for test cases to also expect custom things in it. I reckon this is why I think it makes it not really a good part of Nix core
<vcunat> right, providing DNS seems heavy enough to use a container or VM for the test
<Sonarpulse> sphalerite: any luck recently getting rid of RPATHs?
<sphalerite> nope :(
<sphalerite> Still get those confusing segfaults
romildo has quit [Quit: Leaving]
<Sonarpulse> sphalerite: hmm got the branch?
<Sonarpulse> sphalerite: thanks
FRidh has quit [Quit: Konversation terminated!]
vcunat has quit [Quit: Leaving.]
orivej has quit [Ping timeout: 264 seconds]
phreedom has joined #nixos-dev
__Sander__ has quit [Quit: Konversation terminated!]
romildo has joined #nixos-dev
vcunat has joined #nixos-dev
goibhniu has quit [Ping timeout: 268 seconds]
FRidh has joined #nixos-dev
ma27 has quit [Ping timeout: 240 seconds]
jtojnar has quit [Quit: jtojnar]
vcunat has quit [Quit: Leaving.]
ma27 has joined #nixos-dev
orivej has joined #nixos-dev
FRidh has quit [Quit: Konversation terminated!]
zarel has joined #nixos-dev
vcunat has joined #nixos-dev
<bgamari> Sonarpulse, I see you are working on your haskell-fix-deps branch
<Sonarpulse> bgamari: yeah it's sort of getting crazier and crazier :D
phreedom has quit [Ping timeout: 248 seconds]
phreedom has joined #nixos-dev
<bgamari> Sonarpulse, in a good way I hope?
<Sonarpulse> bgamari: I guess :)
<Sonarpulse> bgamari: I really wanna hop right to a) fixing new-build for cross
<Sonarpulse> b) letting new-build drive nix with plan.json stuff
<Sonarpulse> generic-builder is no fun
<bgamari> I should probably mention now that I started on (b)
<bgamari> I hacked support into cabal2nix to use plan.json to generate a bunch of nix derivations
<bgamari> that reflects the install plan
<bgamari> the code is in my cabal2nix branch
<bgamari> it needs more work though
<bgamari> the current nix model of haskell dependencies is IMHO rather broken
<Sonarpulse> bgamari: OOOOO!!!!!!!!!
<bgamari> ignoring versions simply isn't sustainable
<Sonarpulse> yes
<Sonarpulse> also it should be per-component
<bgamari> yes
<Sonarpulse> well i think cabal should solve versions
<Sonarpulse> and then nix doesn't care as usual
<Sonarpulse> but because cabal did the heavy lifting
<Sonarpulse> lts-style same version everywhere can be accomplished with a cabal.project.local
la_putin has joined #nixos-dev
el_putin has quit [Read error: Connection reset by peer]
<Sonarpulse> bgamari: found it already :)
<bgamari> it's quite a hack though
<bgamari> I ultimately ran out of time and had to stop and patch together something manually
<Sonarpulse> bgamari: well, we can return to it
<Sonarpulse> I need to finish my libdependency megapatch to cabal
<bgamari> indeed
<bgamari> Sonarpulse, first we need working cross-compilation of ghc
<Sonarpulse> bgamari: hehe true
orivej has quit [Ping timeout: 248 seconds]
<Sonarpulse> bgamari: but some stuff of that merged this morning
<Sonarpulse> just need to flesh out tests for ghc and it might just start right up!
<bgamari> Sonarpulse, oh really?
* bgamari tries rebasing
<Sonarpulse> bgamari: I did a little more than rebase
* Sonarpulse it was late and a blur
<Sonarpulse> hopefully that helps not hurts
<bgamari> heh
<Sonarpulse> also the ericson2314-ios-refresh patch contained more relevant thing
orivej has joined #nixos-dev
<Sonarpulse> bgamari: OK pushed the whole fun haskell-fix-deps now
vcunat1 has joined #nixos-dev
<Sonarpulse> bgamari: does a cross compiler need to be built with the same version native compiler?
<Sonarpulse> I did that before
<Sonarpulse> but I no longer think so
vcunat has quit [Ping timeout: 252 seconds]
vcunat1 has quit [Client Quit]
<Sonarpulse> seeing that there is no ghci TH or anything else requiring same ABI
vcunat has joined #nixos-dev
<bgamari> Sonarpulse, nope
<Sonarpulse> bgamari: cool
<Sonarpulse> and Setup.hs really ought not to care too much
<Sonarpulse> either
<bgamari> right
orivej has quit [Read error: Connection reset by peer]
orivej has joined #nixos-dev
phreedom has quit [Quit: No Ping reply in 180 seconds.]
michaelpj_ has joined #nixos-dev
phreedom has joined #nixos-dev
orivej has quit [Quit: No Ping reply in 210 seconds.]
<Sonarpulse> bgamari: we'll see what patches of last feburary's made the nixpkgs ghcs now :)
orivej has joined #nixos-dev
phreedom has quit [Ping timeout: 268 seconds]
phreedom has joined #nixos-dev
orivej has quit [Ping timeout: 268 seconds]
zarel has quit [Quit: Leaving]
vcunat has quit [Quit: Leaving.]
romildo has quit [Quit: Leaving]
<pie_> yesterday i was scanning manga and ran into what may or may not be a scanner driver bug
<pie_> i havent tried on windows yet to rule that out
<pie_> it seems to drop a (couple?) columns of pixels ever 2584 columns
<pie_> fwiw its not really visible but...
el_putin has joined #nixos-dev
la_putin has quit [Read error: Connection reset by peer]
michaelpj_ has quit [Ping timeout: 264 seconds]
<Sonarpulse> bgamari: IIRC the reason for needing absolute paths for ghc's CC
<Sonarpulse> is that it puns CC for the host -> target cc it uses at run time
<Sonarpulse> when that's not any of CC CC_FOR_BUILD CC_FOR_HOST
<Sonarpulse> as those are all build -> * (depsBuild* / nativeBuildInputs)
<bgamari> mm
<bgamari> it would be good to record that in the expression
<Sonarpulse> bgamari: yeah
<Sonarpulse> I was just looking at some of the old track tickets
* bgamari looks forward to hadrian being usable
<Sonarpulse> oops I forget too
<Sonarpulse> bgamari: well hadrian will still use configure.ac?
<Sonarpulse> for the time being?
<bgamari> when hopefully we will be able to derive all expressions from a single source file
<bgamari> yes
<bgamari> I don't see this changing
<Sonarpulse> cool
<bgamari> there simply isn't a reasonable replacement
<Sonarpulse> I'm exciting to clean up configure.ac
<Sonarpulse> with the _FOR_*
lassulus has quit [Quit: WeeChat 1.9.1]
<bgamari> that would be great
<Sonarpulse> is there any easy peasy "milestone" thing or similar, to e.g. see what release https://phabricator.haskell.org/D4008 made it into?
<bgamari> not if there wasn't a ticket created
lassulus has joined #nixos-dev
<Sonarpulse> bgamari: ok
<Sonarpulse> well I'll figure out soon enough during patchPhase
<bgamari> Sonarpulse, git describe helps
<Sonarpulse> bgamari: true
lassulus has quit [Client Quit]
lassulus has joined #nixos-dev
<bgamari> does anyone understand the scoping rules of `nix eval`?
<bgamari> `nix eval blah` works yet `nix eval '(blah)'` appears not to
<bgamari> it seems like nix eval interprets arguments in parens as an expression
<bgamari> but nothing appears to be in scope
<gchristensen> niksnut: hydra is in trouble....! https://twitter.com/timgostony/status/948682862844248065
<Sonarpulse> bgamari: no idea
<Sonarpulse> going home now
<Sonarpulse> but think I have everything I need to squash
<Sonarpulse> for ghc
<bgamari> yay
<Sonarpulse> bgamari: I was starting to do the "kick off them commute"
<Sonarpulse> but had to kick myself, I'm not *that* close yet :D
<niksnut> it doesn't say whether that's hvm or pv