ma27 has quit [(Ping timeout: 272 seconds)]
orivej has quit [(Ping timeout: 252 seconds)]
<Sonarpulse> domenkozar: :) congrats
<Sonarpulse> niksnut: https://github.com/NixOS/nixpkgs/pull/26805 this is getting really close, and its by far the biggest of my cross changes
<Sonarpulse> (in terms of single PR)
<Sonarpulse> I dunno what you want, or will have time to review, but anything I should be doing on that feel free to let me know
<Sonarpulse> one thing to keep in mind is some of the check boxes I have there, like doc clean up, I can do after the mass rebuild, and conversely other people have stuff that can better proceed in parallel once this lands
<Sonarpulse> so it may make sense not to wait for every box there
<Sonarpulse> sure you sleeping now, but I'll be back on tomorrow
<Sonarpulse> as usual
mbrgm has quit [(Ping timeout: 256 seconds)]
mbrgm has joined joined #nixos-dev
pie__ has quit [(Ping timeout: 260 seconds)]
pie_ has joined joined #nixos-dev
orivej has joined joined #nixos-dev
la_putin has quit [(Read error: Connection reset by peer)]
pie_ has quit [(Ping timeout: 256 seconds)]
el_putin has joined joined #nixos-dev
pie_ has joined joined #nixos-dev
pie_ has quit [(Remote host closed the connection)]
pie_ has joined joined #nixos-dev
pie_ has quit [(Ping timeout: 265 seconds)]
ma27 has joined joined #nixos-dev
zarel has joined joined #nixos-dev
ma27 has quit [(Quit: WeeChat 2.0)]
ma27 has joined joined #nixos-dev
olto has joined joined #nixos-dev
phreedom has quit [(Ping timeout: 252 seconds)]
vcunat has joined joined #nixos-dev
WilliButz has quit [(Ping timeout: 250 seconds)]
WilliButz has joined joined #nixos-dev
ma27 has quit [(Quit: WeeChat 2.0)]
ma27 has joined joined #nixos-dev
phreedom has joined joined #nixos-dev
<domenkozar> Sonarpulse: thank you for most of the hard work :)
orivej has quit [(Ping timeout: 268 seconds)]
<gchristensen> whoa, the "merge button" is making signed merges using GitHub's signing key
<adisbladis> Thats pretty neat :)
ma27 has quit [(Quit: WeeChat 2.0)]
el_putin is now known as el_putin_asleep
FRidh2 has joined joined #nixos-dev
FRidh2 has quit [(Quit: Konversation terminated!)]
FRidh2 has joined joined #nixos-dev
FRidh2 has quit [(Client Quit)]
FRidh2 has joined joined #nixos-dev
FRidh2 has quit [(Client Quit)]
FRidh2 has joined joined #nixos-dev
hiberno has quit [(Quit: WeeChat 1.6)]
<FRidh2> What do you think of moving/copying some of the functions and scripts that are in the Nix repo to Nixpkgs? The Nix repo would then contain only the essential parts, whereas extra functions for making nicer installers with more options, or nix-profile scripts for other shells could be found in the Nixpkgs repo.
vcunat1 has joined joined #nixos-dev
vcunat has quit [(Ping timeout: 252 seconds)]
vcunat1 has quit [(Quit: Leaving.)]
vcunat has joined joined #nixos-dev
peti has joined joined #nixos-dev
<peti> gchristensen: I'm back from the usual Xmas holidays. :-)
orivej has joined joined #nixos-dev
<gchristensen> hi peti :)
<gchristensen> did you have a good holiday, other than the haskell package set excitement? :D
orivej has quit [(Ping timeout: 248 seconds)]
ma27 has joined joined #nixos-dev
vcunat1 has joined joined #nixos-dev
vcunat has quit [(Ping timeout: 252 seconds)]
vcunat1 has quit [(Client Quit)]
vcunat has joined joined #nixos-dev
<domenkozar> ha, darwin sandbox keep uncovering impurities
<peti> gchristensen: Yes, I did. I also got nice presents. :-)
<domenkozar> peti: wb!
<domenkozar> thanks for merging my patches over xmas :/
<peti> domenkozar: he, thanks for putting all the effort into creating and testing them!
<peti> Having binary GHC snapshots of newer versions is particularly nice.
<domenkozar> yes :)
<domenkozar> I was looking a bit how to implement drv diffing per platform
<peti> domenkozar: Once the multiple-outputs patch lands, we can just disable shared library builds, IMHO, and then we'll always deliver Haskell binaries with minimal dependencies.
<domenkozar> yes! that's my plan
<domenkozar> now we have million of -static overrides
<domenkozar> quite annoying to maintain with package renames
<domenkozar> and additions
<peti> domenkozar: Why do you want to diff drv files?
<vcunat> hmm, I wonder why dynamic libraries don't seem to be worth it, size-wise
<domenkozar> peti: so the idea would be to generate Derivation per platform
<domenkozar> and then just diff them
<peti> domenkozar: To accomplish what?
<domenkozar> then one could render nix derivation based on those differ per platform
<domenkozar> if stdenv.isDarwin then [..] else [..]
<domenkozar> it's one of many approaches
<domenkozar> a better one would be to properly parse cabal dsl
<domenkozar> like you proposed years ago :)
<peti> domenkozar: But you don't need to parse drv files for that. All the information is readily available in the Cabal file already, no?
<domenkozar> right now, immediate improvement to stack2nix is that we might just generate different package sets per platform
<domenkozar> and have a global if platform conditional
<domenkozar> a waste of git usage though :)
<domenkozar> peti: yes, that's what I'm saying the most proper way would be to get Cabal DSL translate directly into Nix
<peti> domenkozar: Ah, okay, I misread that.
<domenkozar> peti: offtopic a bit, do you think default nixpkgs resolution function could be a bit more sane than https://github.com/NixOS/cabal2nix/blob/master/src/Cabal2nix.hs#L164
<domenkozar> like: zlib = pkgs.zlib;
* peti thinks that diffing derivation files is much more difficult that to just create a proper mapping from GenericPackageDescription to Nix.
<domenkozar> now it seems to be: zlib = null;
<domenkozar> peti: I tend to agree :)
<peti> domenkozar: Sure, a much more sophisticated function exists already and it's used by hackage2nix. But it also needs far more information than cabal2nix currently has, i.e. it needs to dump the nixpkgs attribute set and parse the available names.
<domenkozar> yeah, I just wonder if we can pick a better default
<domenkozar> so it compiles with a generate package set
<domenkozar> for cabal2nix output it doesn't matter much anyway
<domenkozar> generated*
<domenkozar> let me create a patch :)
olto has quit [(Quit: hasta luego)]
<domenkozar> peti: anyway I hope you get enjoyable holidays :)
<peti> domenkozar: Yes, I did! Barely touched the computer over Xmas (except for merging stuff around a little). That was a welcome break.
<peti> Now, the next project I want to tackle is to write some non-trivial code with Backpack.
<domenkozar> nice :)
<peti> domenkozar: How about you? Did you have nice holidays?
<domenkozar> yes, living in Lisbon is a holiday on it's own :)
the has joined joined #nixos-dev
<gchristensen> (continued spam problems being reported, the bot's list of ban words has een updated to include what they say)
<infinisil> gchristensen: How about banning every user that sends a message with more than 10 valid nicks?
<gchristensen> yeah, that could be done
<gchristensen> right now the bot is a 20-minute hackjob and that would be at least a 1hr hack job :)
<infinisil> gchristensen: Are you also using your rabbitmq bot thing for it?
contrapumpkin is now known as {-_-}
{-_-} is now known as contrapumpkin
<gchristensen> no, {^_^} can't do things like become +o or kick / ban people, it is a limitation of the current implementation. but at this rate I might as well teach {^_^} to do it, so I don't have to start and +o `the` by hand
<gchristensen> it'd have to be a separate mechanism so only specific components of the bot can issue commands like that
the has quit [(Remote host closed the connection)]
<infinisil> I see
the has joined joined #nixos-dev
the has quit [(Remote host closed the connection)]
Guest6184 has joined joined #nixos-dev
Guest6184 has quit [(Remote host closed the connection)]
nix-build has joined joined #nixos-dev
nix-build has quit [(Changing host)]
nix-build has joined joined #nixos-dev
peti has quit [(Quit: WeeChat 2.0)]
pie_ has joined joined #nixos-dev
<gchristensen> I think it is probably not appropriate to continue the discussion w.r.t. #33141 on GitHub, and I think something like IRC or even a video chat would be appropriate. I think the long-form format of the github comments is likely to not result in a positive result. anyone have opinions on this? cc fpletz
orivej has joined joined #nixos-dev
<FRidh2> gchristensen: Its a discussion on what the scope is of Nixpkgs -> mailing list
<gchristensen> I agree, but also I think there is a second discussion about his actions
<FRidh2> Right, well, the other part also makes for two nice threads on the list (and eventually RFC's?). a) when do we close PR's and b) when do we lock a thread.
vcunat has quit [(Ping timeout: 252 seconds)]
ma27 has quit [(Ping timeout: 255 seconds)]
vcunat has joined joined #nixos-dev
ma27 has joined joined #nixos-dev
zarel_ has joined joined #nixos-dev
zarel has quit [(Disconnected by services)]
zarel_ is now known as zarel
<globin> fwiw I thought it was overkill to limit the pr to collaborators only without any discussion and I was tempted to unlock it but I don't have the energy for such a discussion right now
<catern> oh I should probably post https://github.com/NixOS/nix/issues/1734 on the mailing list?
<catern> ugh that means I need to subscribe again to the mailing list now that it is no longer on Gmane, huge headache :(
pie_ has quit [(Remote host closed the connection)]
pie_ has joined joined #nixos-dev
<gchristensen> unfortunately going to be more true over time
orivej has quit [(Ping timeout: 255 seconds)]
romildo has joined joined #nixos-dev
pie_ has quit [(Remote host closed the connection)]
pie_ has joined joined #nixos-dev
peti has joined joined #nixos-dev
<peti> niksnut: There is a handful of Haskell builds on hydra.nixos.org that have been running for several says "sending inputs". See https://hydra.nixos.org/build/66442803, for example. It might be a good idea to kill those. Cancelling the build via the Hydra web front-end seems to have no effect.
romildo has quit [(Quit: Leaving)]
peti has quit [(Quit: WeeChat 2.0)]
FRidh2 has quit [(Quit: Konversation terminated!)]
<niksnut> peterhoeg: done
<niksnut> sorry, meant peti
JosW has joined joined #nixos-dev
ma27 has quit [(Ping timeout: 252 seconds)]
pie_ has quit [(Ping timeout: 246 seconds)]
pie_ has joined joined #nixos-dev
pie_ has quit [(Ping timeout: 250 seconds)]
pie_ has joined joined #nixos-dev
pie_ has quit [(Remote host closed the connection)]
pie_ has joined joined #nixos-dev
orivej has joined joined #nixos-dev
pie_ has quit [(Ping timeout: 240 seconds)]
ma27 has joined joined #nixos-dev
JosW has quit [(Quit: Konversation terminated!)]
<Sonarpulse> niksnut: did you get my ping?
<Sonarpulse> domenkozar: I'm not sure I did the hard part at all :)
el_putin_asleep is now known as el_putin
jtojnar has quit [(Ping timeout: 252 seconds)]
nix-build has quit [(Ping timeout: 255 seconds)]
phreedom has quit [(Ping timeout: 246 seconds)]
ma27 has quit [(Quit: WeeChat 2.0)]
<globin> niksnut, shlevy: is it conseptually possible to add a new input type to hydra which corresponds to fetchTarball? i.e. I'd like to have some service regularly generate raw PR/issue data from the nixpkgs repo, and a hydra jobset pull that and generate the statistics
zarel has quit [(Quit: Leaving)]
<niksnut> globin: maybe you can call fetchTarball? not sure if it's currently allowed in restricted eval mode
<niksnut> my goal is to get rid of relying on jobset inputs and use fetchGit etc. instead
<Sonarpulse> niksnut: just pealed off some boring stuff from that big PR and merged it
<Sonarpulse> and then a bigger boost thing that does need review
<niksnut> Sonarpulse: thanks, yeah just saw it
<Sonarpulse> niksnut: i'm getting a little cabin fever, no fault of anyone else of course :), hoping to push the big button by new years now
<globin> niksnut: as a "nix expression" input or in the build definition?
<niksnut> globin: yeah, in the Nix expression
<niksnut> for example, if you need nixpkgs, you just do "with import (fetchTarball channel:nixos-17.09) {};" instead of having a jobset input for that
<sphalerite> How do we handle availability of instruction set extensions? I'm having issues with it on ARMv7 (openal-soft won't build because of some code that requires NEON and the compiler being instructed not to emit NEON code AFAIUI) but surely it's an issue that's been encountered and dealt with at least to some extent on other systems including x86_64 too
<simpson> Hm. We *should* do something about it.
<sphalerite> is our current approach just "emit code for the lowest common denominator"?
<sphalerite> dtzWill: am I right in believing that allvm would sort of solve that problem?
<vcunat> the standard is to emit code for the lowest common denominator, optimize for most common HW types, and optionally emit multiple variants chosen during runtime based on real support
<Sonarpulse> anyone want to review https://github.com/NixOS/nixpkgs/pull/33188 ?
<vcunat> (but all is really upstream work, not ours)
<Sonarpulse> maybe I can diff the evaluated derivation
<sphalerite> vcunat: but what about cases like openal-soft where the lowest-common-denominator setting actually breaks the build?
<vcunat> that sounds like such cases aren't runnable at all at the lowest-CD hardware
<sphalerite> my guess is that the library itself handles that distinction
<sphalerite> hm it seems to be handled explicitly by the library, but at compile time
<sphalerite> I'm guessing something in nixpkgs is preventing the compilation from working, but not the detection
<LnL> on darwin we do some stuff to only export a common set of symbols for impure libraries
<LnL> that way builds don't detect stuff that might not be available on older versions
vcunat has quit [(Quit: Leaving.)]
<globin> sphalerite: I think e.g. tensorflow has arguments to add support for sse etc.
<globin> sphalerite: but defaults to LCD
<sphalerite> alright
<sphalerite> oh yeah, and what is our LCD on x86_64?