<samueldr>
just cued a new eval of trunk-combined, hoping this one passes
cransom has joined #nixos-dev
__monty__ has quit [Quit: leaving]
abathur has joined #nixos-dev
justanotheruser has quit [Ping timeout: 256 seconds]
<matthewbauer>
@dom
<matthewbauer>
domenkozar: Is there a way to upload build dependencies to cachix? I was trying `nix-instantiate | cachix push ...` but that seems to only upload drv's
<matthewbauer>
I'd like to upload the store paths that you would need to rebuild the derivation
<colemickens>
`nix copy --to ssh-ng://` being slower than nix copying to a local store, rsyncing and then copying back from that local store would be a bug, right?
<samueldr>
right now I'd like to be able to grep through all hydra logs for a given eval to see if and to what extent there is a regression from package update
drakonis has quit [Quit: WeeChat 2.7.1]
<samueldr>
thankfully no regressions (found the right eval), but this makes my job harder as it means the issue lies within the linux kernel source code (with make nconfig) I think
abathur has quit [Ping timeout: 260 seconds]
orivej_ has quit [Ping timeout: 240 seconds]
FRidh has joined #nixos-dev
lovesegfault has joined #nixos-dev
phreedom has quit [Ping timeout: 240 seconds]
bhipple_ has quit [Remote host closed the connection]
bhipple has quit [Remote host closed the connection]
phreedom has joined #nixos-dev
cole-h has quit [Quit: Goodbye]
abathur has joined #nixos-dev
abathur has quit [Ping timeout: 240 seconds]
FRidh2 has joined #nixos-dev
FRidh has quit [Ping timeout: 264 seconds]
lovesegfault has quit [Ping timeout: 240 seconds]
<garbas>
pie_[bnc]: on vidyo: i think in previous versions of that expression a link was used which was mozilla specific. i think now you can safely move it to nixpkgs. i'm not using vidyo anymore as well as mozilla transitioned to zoom some time ago.
<pie_[bnc]>
well, would be nice i fi could get it to not crash *shrug* :)
Jackneill has joined #nixos-dev
abathur has joined #nixos-dev
abathur has quit [Ping timeout: 250 seconds]
<yorick>
pie_[bnc]: running it on wayland?
<pie_[bnc]>
yorick: kde/x11 i think
<pie_[bnc]>
ive no idea how to use wayland
<srk>
you just launch 'weston' or 'sway' from pty
<srk>
not sure if it's a problem if X11 is running as well
<pie_[bnc]>
ok so basically im not
<srk>
looks like I can reliably kill test VMs just by spewing too much messages to logs, uh
<srk>
machine # [ 8.551904] systemd[1]: Started udev Wait for Complete Device Initialization.
<srk>
machine # [ 119.435412]
<srk>
/o\
__monty__ has joined #nixos-dev
orivej has joined #nixos-dev
<flokli>
srk: yeah, `udevadm settle` is a broken concept
<srk>
Discourse hostname: localhost is not a valid domain for emails!
<srk>
oh thanks!
<srk>
for wasting 2h of my day :D
<flokli>
heh
<srk>
fun thing I didn't hit this earlier is because I've used discourse.xmpl as a domain in hosts because it's so easy!
lassulus has quit [Quit: WeeChat 2.6]
lassulus has joined #nixos-dev
bridge[evilred] has quit [Remote host closed the connection]
<andi->
It is complicated. It isn't formally a team. This year is the usual suspects: flokli, adisbladis, zimbatm, megfault & me plus someone I probably have forgotten. I don'
<adisbladis>
garbas: ^
<andi->
I don't even know if we want to make it a formal team in that sense...
<adisbladis>
Yes it's not so formal :)
<garbas>
this is the same response i get from every team :)
<adisbladis>
Nice
<garbas>
so you are as much as a team as any other :)
<adisbladis>
Scruffy hackers
<samueldr>
and even then, there are some attached persons to teams that are not formally involved
<samueldr>
e.g. I worked on graphics for the last three nixcons
<samueldr>
but I wouldn't say I'm in the team
<garbas>
maybe i will put it differently. if sombody wants to help with nixcon, who/how would you like to contact?
<andi->
I want it to be inclusive and self-organizing not some set-in-almost-stone people. If someone is unhappy with something they should be empowered to change/fork/help
<andi->
garbas: I think that has been laid out on the discourse after last years nixcon
<andi->
at least that was the consensus after the last nixcon
<adisbladis>
Points of contact should probably be me, andi-, flokli & zimbatm
<adisbladis>
gchristensen: I did that earlier today :)
<gchristensen>
oh cool
<danderson>
Had one question about a slightly weird binary diff on what should be a no-op rebuild, but it's boiled down to "Go's doing something weird but harmless"
<danderson>
and I don't have time to dig into the Go compiler to diagnose further right now :)
<danderson>
gchristensen: thanks!
<gchristensen>
yep!
<danderson>
hmm, what's the usual delay for release channels advancing? Want to tweet in celebration, but it's silly if it doesn't work yet :)
<gchristensen>
it can be a few minutes to a few days but usually closer to "about a day"
<abathur>
not sure what you do/don't like, but I think rst shines around its support for new semantics, rather than on its presentational markup
<domenkozar[m]>
yeah we could add Nix support to rst
<domenkozar[m]>
so then you can say
<domenkozar[m]>
:nix:attr:foo
<domenkozar[m]>
or :nix:func:bar
<gchristensen>
it took me like 4 tries to figure out how to do `link syntax <http://>`_ <- which, well, it inspired awe in me.
<abathur>
yeah
<abathur>
I agree its presentation form is a bit weaker
<gchristensen>
I *really* like the :: syntax
<abathur>
though, fwiw, I regularly invert my markdown link order no matter how many times I think it through
<abathur>
yeah
<gchristensen>
everything else I found to be just fine
<gchristensen>
header syntax seems like it is a convention that ==== is a top level and ---- is second level, and maybe there isn't a third level -- which maybe this is fine but seemed a bit strange
<abathur>
IIRC it supports an arbitrary 6 levels
<abathur>
but the characters are inferred from what you use?
<domenkozar[m]>
you can use whatever
<domenkozar[m]>
***
<abathur>
yeah
<gchristensen>
how do you define a level? what I was seeing is ==== and ----- are only conventions
<domenkozar[m]>
a line followed by 3 or more chars
<domenkozar[m]>
special chars*
<abathur>
IIRC it's just storing state by use; if it looks like a header and isn't using 3 chars it's already seen, you get a new level
<gchristensen>
oh I see
<gchristensen>
wow
<domenkozar[m]>
yes
<domenkozar[m]>
rst is one big magic state machine
<gchristensen>
oh dear
<domenkozar[m]>
;D
<gchristensen>
I'm not sure if you're trying to sell me on it or against it :$P
<abathur>
grin
<domenkozar[m]>
that's why I like asciidoc
<domenkozar[m]>
because I don't know much about it's internals
<abathur>
ha
<gchristensen>
since its docbook ...
<abathur>
I can't count the number of times I've bitched about the abstraction boundaries in docutils
<domenkozar[m]>
:D
<abathur>
but, up at the sphinx level, some nice stuff in RST is that you can go back and audit dangling references
<abathur>
so, if you go rename a module and a thing moves, all of the references to it can squeak
<gchristensen>
nice
<abathur>
and, likewise, that the indirection supports not needing to know where the thing is, just what it's called
<{^_^}>
nixops#1256 (by adisbladis, 1 week ago, open): RFC: Use Poetry & Poetry2nix for environment and plugin management
<gchristensen>
and then I can tell DigitalKiwi they're good to go on workingon their digitalocean plugin (they've been annoyed at me over this for weeks now :x)
<danderson>
oh that reminds me of a constraint of mine for using nixops 2: digitalocean has to keep supporting nixos-infect or similar as a deployment method, not just custom images
<danderson>
because digitalocean has the bizarre restriction that custom images cannot enable IPv6 connectivity :(
<gchristensen>
:(
<gchristensen>
is it possible to use someone else's images?
<danderson>
you can use one of digitalocean's own images, and nixos-infect that. The resulting VM has v6 connectivity just fine
<danderson>
e.g. I deployed on digitalocean by infecting digitalocean's ubuntu VM image
<gchristensen>
yeah, like if NixOS Foundation did that on an account somewhere, could you somehow use the foundation's image?
<danderson>
(which makes this an even more bizarre restriction - as long as you *start* from a supported image, you get to use IPv6?!)
<danderson>
I don't think that works, no. I've been told that it's some bizarre legacy constraint of how their virtual network is implemented
<danderson>
the only VM images with the magic "this can use IPv6" tag are DO's own images
<danderson>
maybe if you have a line to someone inside DO, you could get a marketplace image with similar permissions? But there's no way to craft such an image from the public APIs
<danderson>
I did contact our customer rep at DO to kvetch about this, but their response was basically "yeah we know, nobody's working on that, sorry"
<gchristensen>
I will ping someone :)
lassulus has quit [Ping timeout: 265 seconds]
kgz has quit [Ping timeout: 265 seconds]
yorick has quit [Ping timeout: 265 seconds]
delroth has quit [Ping timeout: 265 seconds]
delroth has joined #nixos-dev
lassulus has joined #nixos-dev
yorick1 has joined #nixos-dev
yorick1 is now known as yorick
kgz has joined #nixos-dev
<danderson>
gchristensen: lemme know if you manage to twist their arm into providing nixos images with the magic v6 bit - that would greatly ease my prod deployment story :)
<gchristensen>
I have a few points of contact there, I'll try with each one until I'm out of options :)
<FRidh2>
error: while setting up the build environment: unable to make '/' private mount: Permission denied
<FRidh2>
any idea why this is happening with nix with database in docker?
<gchristensen>
if you're running a build as root, it is probably trying to do the mount namespace stuff and that is probably a privileged thing and may require the --privileged flag (or run the build as not-root, and do the "single-user" build without the sandbox)
<FRidh2>
Ericson2314: its quite confusing these tensorflow packages
<Ericson2314>
FRidh2: yes it is a total mess!
<Ericson2314>
I was about to do your "choose the version of the -* stuff automatically thing" but this complicates that
<FRidh2>
There are quite some people opening issues and PR's all the time for tensorflow often doing things differently. Unless there is more cooperation there I start feeling its better put outside of Nixpkgs. Hope you get this resolved now.
<Ericson2314>
FRidh2: so, full disclosure, this is for a client (I don't actually use this stuff) but yes the goal, and the client's directive, is to end the madness :)
<Ericson2314>
hopefully this can be one less troublemaker package on your plate after!
<niksnut>
Ericson2314: I'll have a look at your PRs soon
<Ericson2314>
niksnut: awesome, thanks!
<Ericson2314>
niksnut: probably the best to start is the rust FFI change, and then either the multiple fixed-output derivation modulo (not actually exposed in derivation primop, intentionally) or the hash in derivation type one
<FRidh2>
gchristensen: right, now I get it. Also why nix-env fails because that one also always builds a derivation
<Ericson2314>
The hash derivation one is the most important of those three (and relies on the rust FFI) but doesn't yet build. My friend and I are planning on fixing it over the weekend, but if the basic idea of what we are trying to do (replace stringly typed stuff with the fully parsed version) will stay the same if you just want to approve/dissaprove of the concept
<Ericson2314>
niksnut: In particular, feel free to do last or just ignore the big/small derivation refactor PRs. They are the most interesting, but also hardest to read, until the hashing changes are in which will make everything nicer
<gchristensen>
FRidh2: ah, cool
<yorick>
Ericson2314: there were some concerns about the rust ffi safety :D
<Ericson2314>
yorick: yes I made it neither better or worse
<Ericson2314>
yorick: my view is we should pick 1 language and then we don't need to worry about FFI safety :D
<Ericson2314>
polyglot sucks
<yorick>
Ericson2314: total rewrite or bust
<gchristensen>
the Rust crate cxx apparently does muchthe same thing, but if we used cxx and improved cxx along the way we'd be helping the greater ecosystetm
<yorick>
Ericson2314: there's a guaranteed double-free somewhere
<Ericson2314>
yorick: total rewrite before the rust abi changes :D
<yorick>
oh, fixed
<gchristensen>
big-bang total rewrites are impossible
<Ericson2314>
gchristensen: oh I didn't know about this cxx create; haven't followed the ffi discussion too closely
<Ericson2314>
Well, as long as the tests keep on passing every PR, I think rewriting can actually go quick
<Ericson2314>
the really big-bang-you're-screwed is something where you have to regress on tests because you can't run them
<srk>
I wish hnix(-store) would progress a bit faster
<srk>
not that it can replace nix due to portability but we could use both to build interesting tools on top
<Ericson2314>
srk: me too but hnix made the mistake of being too big-bang
<Ericson2314>
given the amount of time/effort we put into it
<Ericson2314>
srk: when I've been working on C++ nix, my main priority is to make sure the "core stuff", e.g. data formats is good
<srk>
I see, I only worked on hnix-store
<Ericson2314>
I don't plan on endlessly optimizing algorithms or anything I rather work on that with hnix
<Ericson2314>
srk: well yes all this recent work on hnix-store has been a great sign!
<Ericson2314>
I look forward to making the changes I hope to land in C++ in haskell after :)
<srk>
cool :)
<Ericson2314>
As I write the template soup, I keep in mind what is actually parametric vs what would be a trait / typeclass :)
<srk>
a true polyglot! :)
<srk>
Ericson2314++
<{^_^}>
Ericson2314's karma got increased to 4
red[evilred] has joined #nixos-dev
bridge[evilred] has joined #nixos-dev
FRidh2 has quit [Quit: Konversation terminated!]
tilpner has quit [Remote host closed the connection]
tilpner has joined #nixos-dev
harrow has quit [Quit: Leaving]
<gchristensen>
adisbladis: I couldn't resist adding some types to fix the red X :)
<adisbladis>
types++
<adisbladis>
:)
bhipple has joined #nixos-dev
harrow has joined #nixos-dev
__monty__ has quit [Ping timeout: 240 seconds]
edef is now known as edef_
edef_ is now known as edef
<worldofpeace>
Is the minimum memory and storage documented anywhere for NixOS?
<samueldr>
hard question to answer
<samueldr>
NixOS can run without running Nix
<samueldr>
but Nix itself has some requirements to eval Nixpkgs
<samueldr>
(but I do agree that recommended should include running Nix to evaluate Nixpkgs and NixOS)
<samueldr>
though I don't know if you can get a "virtualbox-like" experience with QEMU comparable to QEMU on x86_64
<srk>
like with spice?
<samueldr>
just runnin qemu and having a window
<samueldr>
running*
<samueldr>
well, "just"
<samueldr>
I really don't know :)
<srk>
:) it should be fine provided you have enough power
<samueldr>
what I'm thinking is: does the default aarch64 machine have graphics enabled?
<srk>
hard to tell, not even sure what is the default
<worldofpeace>
b42: so we generate iso with xorriso, so we'd just have to add switches `-joliet` and I think '-rockridge' for linux
<srk>
samueldr: since qemu is pretty generic I would expect that you just pass -vga
<b42>
worldofpeace: yes, it seems that -rockridge is on by default so we already have that one
<b42>
can't find anywhere if the 11-character limit is related though
<srk>
retrocomputing!
<srk>
'By doing this we'd have better control on what virt-install / GNOME Boxes / Cockpit Machines / KubeVirt could automatically setup for their users when spawing a NixOS VM.'
<srk>
solid infection!
red[evilred] has quit [Quit: Idle timeout reached: 10800s]