00:35
<
drakonis >
ehmry: an ifd rfc you say?
00:36
<
drakonis >
its this one, isnt it?
00:37
<
drakonis >
ah, it is certainly interesting
00:38
<
drakonis >
this is fabulous.
00:38
<
aaronjanse >
Agreed. I'm excited
00:38
<
drakonis >
it gets rid of the age old problem of having to maintain a ton of generated code
00:40
<
drakonis >
it is something i've very dearly wanted for nix to have for years
00:41
<
drakonis >
Ericson2314: this is good stuff
00:44
<
Ericson2314 >
drakonis thanks!
00:45
<
drakonis >
not needing to convert pl package managers to nix lang is extremely valuable
00:45
<
drakonis >
guix has importers and they're really useful
00:47
<
aaronjanse >
This RFC seems like a fantastic path to convenient packaging while maintaining the reproducibility promises of Nix
00:47
<
drakonis >
good lord i cant wait to simply have access to everything in a repository
00:48
<
drakonis >
the package management story will exceed any other distribution
00:48
<
aaronjanse >
We'll no longer need bit-for-bit reproducible output just because we want to parse some lockfile. This RFC lets us follow the spirit of Nix without mega-hashes and without generated code
00:49
<
drakonis >
up next in the rfc train should be implementing a facsimile of guix's g-exps and grafts
00:50
<
drakonis >
but first, one excellent change at a time
00:51
<
drakonis >
grafting can do reproducible output without needing to engage in mass rebuilds because you need to replace some dependency
00:52
<
drakonis >
ie: quickly replace a dependency with another one that's patched
00:52
<
drakonis >
its extremely useful for emergencies
00:52
<
drakonis >
and/or when hydra is currently locked
00:53
<
drakonis >
it is too excellent for the small stuff
18:07
qyliss has quit [Quit: bye]
18:12
qyliss has joined #nixos-exotic
20:11
<
drakonis >
wild shit
20:17
<
drakonis >
wrong place, but still
22:07
<
qyliss >
more guix stuff should be posted in Nix channels imo
22:08
<
qyliss >
doesn't do us any good to be ignorant of what other people solving similar problems are coming up with
22:09
<
simpson >
It's like when kids don't want to go to family reunions because their cousin has a lisp.
22:22
<
drakonis >
haha welp
22:23
<
drakonis >
i should do it more often but i feel that some folks dont seem to appreciate it much
22:26
<
drakonis >
there's some sort of confrontational atmosphere whenever it comes up
22:27
<
drakonis >
nix needs better evangelizing imo
22:28
<
drakonis >
whenever guix comes up, it manages to have far more positive takes on it than nix
22:28
<
drakonis >
did you know that they're going to have a home-manager equivalent for 1.4.0?
22:28
<
drakonis >
as part of the baseline package?
22:28
<
qyliss >
that's cool
22:28
<
qyliss >
and makes a lot of sense
22:28
<
drakonis >
they have a much stronger baseline than nix
22:29
<
drakonis >
nixops and *2nix is available from the get go
22:29
<
drakonis >
no need for extra packages or specific behavior
22:31
<
drakonis >
its very cool what they've done with a much smaller community
22:37
<
drakonis >
niiiiiiice.
22:37
<
Ericson2314 >
qyliss: I can almost do LLVM netbsd with this, but there is a -lgcc sneaking in during libc's linking
22:38
<
Ericson2314 >
but grepping the keep-failed only showed some innocent looking -lgcc
22:39
<
drakonis >
its kinda like hydra's web interface but it exposes far more data in a presentable manner
22:39
<
Ericson2314 >
will look
22:40
<
drakonis >
i havent talked a lot about guix in a while, so i'm kinda giddy about people checking it for inspiration again
22:40
<
Ericson2314 >
i do want to reproduce their nice from-nothing bootstrap
22:41
<
Ericson2314 >
it will be much less anoying to have long critical paths with ca derivations
22:41
<
drakonis >
they're also very principled, which is highly important
22:41
<
Ericson2314 >
well I think they are layer-violation-y
22:41
<
Ericson2314 >
which isn't so principled
22:41
<
drakonis >
hmm, do explain
22:41
<
drakonis >
i'd like to hear about it
22:41
<
Ericson2314 >
but they are certainly more centralized than us
22:42
<
Ericson2314 >
for me the distinctions between surface language stuff be it nix or guile scheme, and the store layer, is very imporant
22:42
<
Ericson2314 >
that don't seem to care so much
22:42
<
drakonis >
doing everything in scheme, yeah?
22:42
<
Ericson2314 >
everything in scheme is fine with me
22:43
<
Ericson2314 >
and sure it's nice you nix-as-a-library or nix-as-a-service or whatever
22:43
<
drakonis >
and being able to directly interface with the store with function calls is the issue here if i understand correctly?
22:43
<
Ericson2314 >
but to me the derivation graph is a core language
22:43
<
Ericson2314 >
and i very much care about compiling to core languages too
22:44
<
Ericson2314 >
for example, i think guix grafting would work much better in conjunction with ca derivations
22:44
<
Ericson2314 >
they weren't so sure
22:44
<
Ericson2314 >
also guix grafting is implemented very imperatively
22:45
<
Ericson2314 >
querying fetchers, querying the what's installed, making derivations are all very intermixed
22:45
<
drakonis >
they have been looking into CAS a while back
22:45
<
Ericson2314 >
yes, grafting is supposed be a hack
22:45
<
Ericson2314 >
but I hope it would at least be a reproducible hack!
22:46
<
Ericson2314 >
i emailed them asking about collorabation, so that our notions of derivations don't drift apart, but they didn't seem too interested
22:46
<
Ericson2314 >
email chain died off pretty quick
22:46
<
Ericson2314 >
hopefully once we have e.g. hydra.nixos.org doing CA nixpkgs (before switching over permenantly)
22:47
<
drakonis >
hmm, well, when was that?
22:47
<
Ericson2314 >
we can pick up the coneresation again
22:47
<
drakonis >
hmm, i see.
22:47
<
Ericson2314 >
earlier this year
22:47
<
drakonis >
interesting.
22:47
<
drakonis >
was it in guix-devel?
22:48
<
Ericson2314 >
no i emailed a core list because there was some ipfs shit in progress I also brought up
22:48
<
drakonis >
you can always ask on irc mind you
22:48
<
drakonis >
there's a fairly strong crossover between the mailing list and irc
22:48
<
drakonis >
so you can talk with ludovic and get a quick response
22:48
<
Ericson2314 >
(we applied for a grant on a non-nix but packaging-related thing i thought would be complementary and are waiting to hear back)
22:48
<
Ericson2314 >
this was ludovic that emailed me
22:49
<
drakonis >
i see i see
22:49
<
drakonis >
hmm, interesting.
22:50
<
drakonis >
last time i brought up CAS, ludovic was interested on it
22:50
<
drakonis >
so, it should be worth bringing up again later.
22:50
<
Ericson2314 >
yeah hopefully so
22:54
<
drakonis >
i really want to see CAS being used in nixpkgs, among a lot of other things
22:54
<
drakonis >
lots of space for doing interesting stuff here
22:57
<
drakonis >
how's that progressing anyways?
22:57
<
drakonis >
i wanted to try something out with it
22:58
<
drakonis >
also does content addressing help with not having to engage in mass rebuilds when modifying the dependency graph?
23:03
<
Ericson2314 >
drakonis: yes it's great for avoiding mass rebuilds
23:04
<
Ericson2314 >
or, trying to do mass rebuilds embarringly parallel because you predict your change "won't matter" most of the time
23:04
<
Ericson2314 >
it's progressing well; mostly regat from Tweag polishing things now
23:05
<
qyliss >
Ericson2314: I heard the other day that reproducing Guix's bootstrap for Nixpkgs has money assigned to it and nobody working on it
23:05
<
qyliss >
(NGI money)
23:06
<
qyliss >
at least I think I did
23:08
<
Ericson2314 >
qyliss: woah really?
23:08
<
Ericson2314 >
are non-Europeans eligible to work on that one?
23:08
<
qyliss >
let me check my logs and confir m this
23:09
<
Ericson2314 >
sure. I wonder did guix just make this part of their grant app, or is something else the grant recipient
23:10
<
qyliss >
I think this is NixOS money
23:10
<
Ericson2314 >
ok cool
23:10
<
qyliss >
oh, the channel it was posted in isn't logged
23:11
<
Ericson2314 >
is that a public channel by any chance?
23:11
<
qyliss >
#nixos-r13y
23:11
<
Ericson2314 >
hehe i can't imagination what r13y sands for
23:11
* Ericson2314
goes and peaks
23:11
<
drakonis >
ah, a peak you say?
23:12
<
qyliss >
ohhhh right I remember the context here
23:12
<
qyliss >
Ericson2314: so, there's funding for every NGI Zero funded project to be packaged with Nix
23:12
<
qyliss >
as one of the support services for projects
23:12
<
Ericson2314 >
oh wow
23:12
<
qyliss >
just like there's money to get me e.g. legal advice if I need it
23:13
<
qyliss >
from what I've heard, they've been having real trouble finding people to actually do the work
23:13
<
drakonis >
woah there.
23:13
<
Ericson2314 >
can I make it not be flakes? :D
23:14
<
drakonis >
ah flakes...
23:14
<
drakonis >
gosh there's so many words on the matter :|
23:14
<
qyliss >
I doubt that's a requirement
23:14
<
Ericson2314 >
seems silly for the authoritative archival packaging to be using unstable features :D
23:14
<
qyliss >
but I don't really know
23:14
<
qyliss >
if I didn't have my own funded project I can barely make enough time for I'd be jumping at this work opportunity
23:15
<
Ericson2314 >
who should I ping?
23:15
<
drakonis >
if i was good at writing nix lang, i'd be taking it
23:15
<
drakonis >
garbas i think?
23:25
<
drakonis >
watching the argument over grafts is v. silly
23:25
<
qyliss >
Ericson2314: why are you so interested in LLVM on NetBSD?
23:25
<
drakonis >
because grafts are an easy mechanism in something like scheme
23:26
<
Ericson2314 >
qyliss: cause I hate waiting for gcc to rebuild :D
23:26
<
qyliss >
did you get a chance to look at the dynamic linker bintools wrapper thing btw?
23:28
<
Ericson2314 >
qyliss: I was thinking about iti
23:29
<
Ericson2314 >
....sorry
23:29
<
Ericson2314 >
after the last PR i pushed I think I am getting closer to ti
23:29
<
Ericson2314 >
trying to "clear the workbench" first so to speak
23:31
<
qyliss >
yeah, valid
23:31
<
qyliss >
you've been doing great stuff this eek