<{^_^}>
domenkozar[m]'s karma got increased to 23.000000000000004
orivej has quit [Ping timeout: 246 seconds]
orivej has joined #nixos-dev
orivej_ has joined #nixos-dev
orivej has quit [Ping timeout: 258 seconds]
alp has joined #nixos-dev
orivej_ has quit [Ping timeout: 256 seconds]
orivej has joined #nixos-dev
evanjs has quit [Read error: Connection reset by peer]
evanjs has joined #nixos-dev
orivej has quit [Quit: No Ping reply in 180 seconds.]
__monty__ has joined #nixos-dev
orivej has joined #nixos-dev
orivej_ has joined #nixos-dev
orivej has quit [Read error: Connection reset by peer]
<kloenk>
the html content of cache.nixos.org. can I get it somewhere? under which license is ist? would like to change the url und use if for my own cache (it has a nice design :-) )
cole-h has quit [Quit: Goodbye]
<LnL>
not sure about the licence, most things are mit but nothing is specified in that repo
<rnhmjoj>
what do you do to test changes to nixos-install? i'm trying to install nixos on an external drive but i can't get grub to install the bootloader
<JJJollyjim>
The installer tests in nixos/tests/ in nixpkgs should provide test coverage
<rnhmjoj>
Jamie: do they cover installing a bootloader too? that's what i'm interested in
<JJJollyjim>
yeah, looks like they cover both systemd-boot and grub
<JJJollyjim>
then it reboots into the drive it installed on, to make sure it worked
<rnhmjoj>
great, thank you, i'll take a look
alp has quit [Ping timeout: 272 seconds]
<manveru>
what are the chances of ever getting `builtins.fromYAML`?
<Profpatsch>
no.
<Profpatsch>
Probably useful tohave :)
<manveru>
yeah, Crystal just switched to a new lock format that contains git revisions... but it's in YAML :)
<Profpatsch>
manveru: IfD?
<manveru>
so still need to do IFD
<manveru>
which is fine, but not for nixpkgs :)
<Profpatsch>
yeah, just convert to json and check that in
<Profpatsch>
who cares really
<manveru>
jup
<infinisil>
manveru: Arghh, why did they choose yaml
<manveru>
why you asking me...
<manveru>
probably because bundler uses yaml too?
<JJJollyjim>
Yaml parser written in nix please :P
<LnL>
is there a standard for a sane subset of yaml?
<infinisil>
json :P
<Profpatsch>
JJJollyjim: We don’t joka about that here
<Profpatsch>
too soon
<LnL>
like without all the anchor crap
<Profpatsch>
what infinisil says
<manveru>
technically correct, but not helpful :P
<LnL>
just wondering since that could be reasonable to add as a builtin
<infinisil>
JJJollyjim: My idea is to have a builtins.haskell, which allows using any pure haskell functions
<JJJollyjim>
oh a builtin, that's boring
<JJJollyjim>
:P
<infinisil>
Heh, gotta use builtins for that
<JJJollyjim>
compile ghc for wasm
<JJJollyjim>
write a wasm interpreter in nix
<JJJollyjim>
ez
<JJJollyjim>
i guess a wasm builtin would be better lol
* JJJollyjim
uses builtins.haskell to import hnix and just run the rest of my nix code :P
<infinisil>
Yeah with a builtins.haskell, something like hnix would have to be used for the interaction layer
orivej has quit [Ping timeout: 260 seconds]
orivej_ has joined #nixos-dev
<adisbladis>
Oh no
<adisbladis>
Nix is already sufficiently complex without FFI boundaries
<JJJollyjim>
:P
<ekleog>
a wasm builtin would make sense, under the assumption that it must be a pure function — it'd avoid having to write complex and slow algorithms in nix, and would maybe improve significantly the time to evaluate nixpkgs? seeing as ISTR that an algorithmic improvement on one of the algorithms reduced significantly the eval time already once
<puck>
IFD :p
<ekleog>
using ifd wouldn't make eval faster, though :p
<puck>
well, that's not an insurmountable problem, is it?
<JJJollyjim>
Yeah idk if wasm and IFD would be that much different
<JJJollyjim>
Performance-wise
<ekleog>
IFD means forking, setuid, setgid, setting whole processes in memory, etc. Wasm would be just load a file cached in the nix store (that would have the prebuilt, cached wasm code) to RAM and jmp to it :) (and on the second eval in the nix-daemon lifetime it could probably even cache it in memory seeing as it'd likely be <1KB, so it'd really just be “jump to native code”)
<ekleog>
(note: the “cache the prebuilt wasm” is a further optimization, I also think that just rebuilding the wasm live would also make sense so long as the [source language -> wasm] step is cached in the store, but that'd be the case with something like lib.evalRust = builtins.evalWasm (lib.buildRustWasm '' ... '') where lib.buildRustSource produces the wasm blob in the store — that part would
<ekleog>
indeed be IFD, though)
<ekleog>
(we may want to just commit the wasm blobs to nixpkgs for the few functions that'd benefit from such a system)
<ekleog>
now… if there's something lower-prio than that, I don't know about it :p
<Profpatsch>
we already have that, builtins.exec.
<Profpatsch>
But it’s behind a very long feature flag for a reason
<adisbladis>
Profpatsch: Tbh I wish we'd remove it
<ekleog>
the very big issue of builtins.exec is that it's not sandboxed, and so it breaks literally everything
<ekleog>
builtins.evalWasm would be sandboxed, and so could be considered as any other nix function
<domenkozar[m]>
Nix doesn't add GC root for the certs
orivej_ has quit [Ping timeout: 260 seconds]
<gchristensen>
impressive
orivej has joined #nixos-dev
<domenkozar[m]>
we should probably just install cacert again after GC
<gchristensen>
wait, no, I think there is something else wrong
drakonis has joined #nixos-dev
<gchristensen>
because it depends on if you're in an interactive bash terminal (which bash does one initialisation process) or non-interactive (where bash loads different files)
justanotheruser has quit [Ping timeout: 272 seconds]
drakonis1 has quit [Ping timeout: 264 seconds]
<domenkozar[m]>
gchristensen: either way I'd be surprised it was the release
<gchristensen>
looking in to it further.
<domenkozar[m]>
but not sure how bash affects the certs, it should respect the same env var
<domenkozar[m]>
k
<domenkozar[m]>
thanks!
drakonis has quit [Ping timeout: 272 seconds]
drakonis_ has joined #nixos-dev
alp has quit [Ping timeout: 272 seconds]
<domenkozar[m]>
interesting
<gchristensen>
I'm poking with dive
<domenkozar[m]>
there's no certs in the profile
orivej has quit [Ping timeout: 246 seconds]
orivej has joined #nixos-dev
<domenkozar[m]>
probably the easiest fix is to remove NIX_SSL_CERT_FILE from dockerfile
<domenkozar[m]>
although git will still fail
<gchristensen>
I think there is a problem with the Nix installer in 2.3.6
<LnL>
it's always been this way AFAIK, not sure why we don't just always install the cert bundle
<{^_^}>
#89781 (by mdlayher, 2 days ago, open): nixos/corerad: add settings option to supersede configFile
nschoe has joined #nixos-dev
nschoe has quit [Ping timeout: 272 seconds]
<ekleog>
Do we have a policy wrt. modules that could be either in nixos, home-manager, or both? Looking at https://github.com/NixOS/nixpkgs/pull/86225 I'm thinking of answering that it'd be better in home-manager, but…
<mdlayher>
happy to trade for other reviews. i'm most comfortable with Go/Prometheus stuff :)
<cole-h>
ekleog: Curious why you think that belongs in h-m rather than NixOS proper? I don't see anything specific that screams "h-m" to me.
<ekleog>
cole-h: well, it's just configuration for an end-user program, which is more h-m than nixos to me
<ekleog>
what could be nixos would be a proxychains.withConfig '' config ''
<ekleog>
(well… nixpkgs even)
<cole-h>
Well, most of the various shells have a NixOS module, and I think that also satisfies "configuration for an end-user program"
<cole-h>
I see your point, but that's not enough justification (for me, IMHO) to redirect it to h-m.
<ekleog>
I'd argue that `programs.` is a mistake, that was made before h-m arose :p but that's also why I'd like to have feedback from other people about how far we want to go duplicating modules between h-m and nixos
orivej has quit [Ping timeout: 256 seconds]
orivej has joined #nixos-dev
tokudan_ has quit [Remote host closed the connection]