<symphorien[m]>
I read the rfc about ZHF and it tells me nothing about when breaking changes *in modules* are allowed, only in packages. In other words, can I merge this today or is it too late? https://github.com/NixOS/nixpkgs/pull/115332
<{^_^}>
#115332 (by symphorien, 5 weeks ago, open): nixos/users: require one of users.users.name.{isSystemUser,isNormalUser}
<samueldr>
hmm... it's breaking for more correctness
<samueldr>
or at least, be more declarative
<samueldr>
and seems (according to a quick read) to have a straightforward actionable assertion message
<samueldr>
so it would be in the least bad bucket
<qyliss>
symphorien[m]++ for that PR either way
<{^_^}>
symphorien[m]'s karma got increased to 4
<ekleog>
I'd vote for it to be good to go too
ajs124 has quit [Quit: Bridge terminating on SIGTERM]
das_j has quit [Quit: Bridge terminating on SIGTERM]
Scriptkiddi has quit [Quit: Bridge terminating on SIGTERM]
<supersandro2000>
the page in the current form is not that useful
<supersandro2000>
everything on it needs a depends
rj has quit [Ping timeout: 240 seconds]
<supersandro2000>
for example the rec one: does not apply to mkDerivation
<supersandro2000>
and the alternative is not simpler. It is more confusing because you create a self reference and need to know how laziness in nixpkgs works
<supersandro2000>
s/nixpkgs/nix/g
<samueldr>
supersandro2000: you're welcome to make contributions I guess
<sterni>
but also stuff like cmake fails to eval because stdenv.cc.libc == null on FreeBSD
<qyliss>
yeah I agree, cross-compilation is the way to go until we can generate bootstrap binaries
<qyliss>
sterni: you should join #freebsd-nix and #nixos-exotic :P
<sterni>
alright, I will
<sterni>
I kind of stumbled upon this when helping out someone in #nixos
<sterni>
like … I have spend exactly 0 minutes using freebsd so far
<qyliss>
I've been working on it intermittently for a while, when I've found the time
<qyliss>
my BSD usage is entirely time spent trying to port Nix
<Ericson2314>
I dream of nixos/all-kernels and eating all the bsd communities :D
<qyliss>
I have spent a lot of time using illumos at a previous job, though, and I sort of see BSD support as a precursor to illumos support :P
<sterni>
Ericson2314: I wonder whether they will welcome their new overlords :p
<samueldr>
we had someone using Nix on Haiku recently passing by on #nixos
<qyliss>
ooh puck did you know about that?
<samueldr>
not sure what happened with their adventure
<Ericson2314>
qyliss: I know a few joyant orbit people I've wanted to impress with this party trick too :)
<sterni>
nix on plan 9
<qyliss>
sterni: in #nix-freebsd AIUI we have at least one FreeBSD committer who's enthusiastic about Nix, and another one who's at least interested in helping us
<Ericson2314>
sterni: probably not lol, but I think we have all the advantages of the portage model at much lower cost
<samueldr>
nix#4650 was written to (probably) fix an issue they surfaced
<puck>
<samueldr> we had someone using Nix on Haiku recently passing by on #nixos <- i've been tempted a few times to somehow merge hpkgs and nix
<samueldr>
the thought crossed my mind a few times
<samueldr>
but I know I don't know how to really get the ball rolling on starting *quickly* enough to not lose interest
<puck>
i think that packagefs doesn't quite mesh well with nix out of the box, but it shouldn't be too tricky to patch to make it work the way you'd expect it to
<puck>
at which point the biggest mystery is importing existing hpkgs to live in the chroot somewhat
<Ericson2314>
qyliss: I'll rebase my PR soon if that helps
Mic92 has quit [Ping timeout: 258 seconds]
tetdim has joined #nixos-dev
pingiun has joined #nixos-dev
Mic92 has joined #nixos-dev
<sterni>
qyliss: btw in the meantime I tried around a bit towards pkgsClang, but seems like I'll have to mess with llvmPackages to make it not cause an infinite recursion
<sterni>
I guess it's not that surprising that llvmPackages depends on stuff that in turn depends on stdenv :)
<qyliss>
Ericson2314: tbh I haven't really looked at the rest of your PR
<Ericson2314>
that's fine
<qyliss>
I've been working on my own version for NetBSD for a while not realised that yours existed
<Ericson2314>
pkgsClang is a new thing?
<qyliss>
or at least not remembering
<qyliss>
sterni wants to add pkgsClang
<sterni>
wait can I just use useLLVM in the targetPlatform
<Ericson2314>
ok
cole-h has quit [Ping timeout: 246 seconds]
<Ericson2314>
sterni: yeah
<sterni>
Ericson2314: hm that doesn't seem to be working as intended currently? The only platform using it currently is wasi32 and pkgsCross.wasi32.stdenv causes an infinite recursion
<sterni>
because pkgsCross.wasi32.buildPackages.llvmPackages.lldClang is already doing that
rj has quit [Ping timeout: 240 seconds]
<Ericson2314>
sterni: i use it with aarch64-multiplatform and other things
cole-h has joined #nixos-dev
<sterni>
oh stdenv fails to evaluate, but you can still build stuff in pkgsCross.wasi32
<Ericson2314>
qyliss: fwiw i did rebase, if you want to make a draft PR with what you'd got I'd be happy to take a look at combining them
abathur has joined #nixos-dev
<Ericson2314>
there are few general cleanups in mine I'd like to see make it in any evnt
<qyliss>
Ericson2314: my approach was going to be to make small fixes and get them in one at a time
<Ericson2314>
sterni: so I did start trying to do darwin -> darwin cross as a step towards getting a from-bootstrap-tools that uses the cross techniques to separate run and build depends
<qyliss>
I think this is too big to do all at once
<Ericson2314>
qyliss: did you bump versions?
<qyliss>
no, I want to get what we have working first
<Ericson2314>
I wanted to do that too I think, but something make me....escalate. I forget exactly
<qyliss>
ctypes perchance?
<qyliss>
I almost did when I hit that
<qyliss>
is there stuff in your PR you think is ready to go in that could be extracted?
<Ericson2314>
I'll look
<qyliss>
(or is the whole thing ready from your PoV?)
<Ericson2314>
I also wish I remembered what made me start bumping versions
<qyliss>
I'm pretty confident I'm over the hump at this point wrt versions
<qyliss>
but we'll see
<qyliss>
it was hard
<Ericson2314>
qyliss: yeah bumping all the fixed outputs is incredably annoying
<Ericson2314>
I think I wanted to just wait for CA derivations lol
<Ericson2314>
do one big fetch and then split up with them so patching didn't rebuild everythihng
<sterni>
Ericson2314: with the first darwin being the native impure stdenv? sounds pretty cool yeah
<qyliss>
I think we could do it nicely with an updateScript that generated JSON
<Ericson2314>
sterni: well with darwin already using llvm
<Ericson2314>
because the libs and compiler proper are built together
<Ericson2314>
reallyreallyreally want that split like LLVM is
<Ericson2314>
and then a gazillion things become much simpler
<qyliss>
I'm looking through your PR now btw, but I'm skipping over the FreeBSD stuff, because my brain can only handle one operating system at a time and if I get overwhelmed I won't want to work on this any more
<Ericson2314>
I started doing both so i could ragequit back and forth :D
<sterni>
Ericson2314: so useLLVM kinda works, I ran into a weird quirk with llvmPackages_7 not evaluating when stdenv.hostPlatform.useLLVM is true: #119479
<qyliss>
get it working first, refactor later if we can imo
<sterni>
Ericson2314: yeah I can just reintroduce what the other sets are doing, give me a sec
<qyliss>
(that's not always an option, obviously, but I think it might be for netbsd)
<Ericson2314>
the llvm people reviewing the patches were super responsive but then disapeared
<sterni>
uh
<sterni>
llvmPackages_7 doesn't even have libunwind
<sterni>
I guess I'm gonna fix this tomorrow
<Ericson2314>
sterni: oh awesome!
<qyliss>
I really wish we could update our default llvm
<qyliss>
but it seems like there's not much somebody without a Mac can do to help that along
<sterni>
well I mean theoretically our default llvmPackages is conditional on the platform
<sterni>
but it's possibly asking for trouble using different ones on darwin and linux if we don't have to
<qyliss>
yeah IIRC that's not an option
<qyliss>
although at some point we might have to consider it IMO
<Ericson2314>
qyliss: sterni to be clear we have some conditional stuff now
<Ericson2314>
just not doing the big linux vs darwin conditionanl
<qyliss>
yeah
<lukegb>
I have a mac mini mostly for doing misc dev stuff on that nixos-dev peeps can have an account on, but it's going to be a painful experience compiling LLVM a bunch of times :(
<lukegb>
(it's an M1 one but I haven't yet bothered to figure out how aarch64-darwin nix is supposed to work)
tokudan has quit [Remote host closed the connection]
<qyliss>
a big part of the reason I want FreeBSD support in Nixpkgs is so it's possible to catch some macOS failures without using macOS
<qyliss>
(I've had quite a bit of luck reproducing macOS build failures on FreeBSD even without Nix)
<qyliss>
I'm glad there's progress being made there
tokudan has joined #nixos-dev
<abathur>
maybe poke __monty__ if there are status or coordination questions, but I knew I remembered him mentioning it
<Ericson2314>
qyliss: yeah the fact that we only have one free unix means that still so many things end up being linux overfit for no particular reason
<samueldr>
that's the main reason that, even when monocultures align with my interests, I am wary of monocultures
<MichaelRaskin>
qyliss: aren't you supposed to also be hopeful about some (low-performance, but still) setup with Chromium-inside-OpenBSD-inside-CrosVM-inside-nsjail-on-Linux (happy breaking out)?
orivej has quit [Ping timeout: 265 seconds]
<MichaelRaskin>
Given there are Chromium-only sites, and Chromium impedes basic security measures like adblockers
<qyliss>
running OpenBSD with crosvm might be a tricky one
<qyliss>
I've still never encountered a Chromium only site
supersandro2000 is now known as Guest45104
Guest45104 has quit [Killed (card.freenode.net (Nickname regained by services))]
supersandro2000 has joined #nixos-dev
<MichaelRaskin>
I think there was a release of BigBlueButton that was effectively Chromium-only
<MichaelRaskin>
(not in the Jitsi sense of quality degradation, but in the sense of not getting it to work at all with then-current Firefox)
<ajs124>
that wouldn't surprise me. bbb is software of very questionable quality.
<qyliss>
yeah...
<MichaelRaskin>
Well, now it works again with Firefox, and I'd say it's the best for giving a talk/lecture (mainly because the rest is just not really well-suited for that task…)
<ajs124>
yeah. to use, it's alright. to run… see #83072