<red[evilred]>
hydra.nixos.org - a Nix build farm, consisting of hundreds of macOS, x86-64 (Linux) and aarch64 (Linux) build cores. Hydra executes over 350,000 builds each week.
<supersandro2000>
Commiter team? Reviewer Team?
<red[evilred]>
There's a reviewer team?
<red[evilred]>
I see lots more teams on github - but I assumed they were RBAC teams as opposed to organization teams
rajivr has joined #nixos-dev
<supersandro2000>
Should there be? IDK
* red[evilred]
shrugs
<supersandro2000>
yeah whatever. If people can't find us they can't write us mails.
<samueldr>
the governance page of the website is more about speciality teams
<red[evilred]>
and there isn't a 1:1 mapping of teams / rbac / github teams either
<red[evilred]>
for example - security team on website, 3 people, security team on github, 10 people, people on security irc doing stuff - more
<red[evilred]>
etc
<red[evilred]>
I'm guessing those enumerated on the pages are either elected (for some teams like rfc / guidence) or people who are just known solid because of their length of time and amount of work done
<maralorn>
domenkozar: Uh, oh … A minor version bump in servant (0.18.1 -> 0.18.2) seems to have broken servant-swagger …: So cachix is broken again on haskell-updates. /o\ https://hydra.nixos.org/eval/1630019
<maralorn>
Huh, probably we can just fix it by disabling the test suite.
<maralorn>
It's already reported upstream.
<{^_^}>
haskell-servant/servant-swagger#129 (by felixonmars, 18 hours ago, open): Test suite doesn't build with servant 0.18.2
<mkaito>
my horse and my kingdom for searchable nix/nixpkgs API docs
<lukegb>
I'm so glad sourcegraph.com exists because I use it far too much for navigating around nixpkgs
<lukegb>
but still I get annoyed at the inconsistent location of random buildPetProgrammingLanguageBar functions :p
<das_j>
lukegb: Huh? Does it support nix?
<lukegb>
das_j: Mostly I use it for doing regex-search across the repo in a web browser; the actual semantic go-to-definition stuff I use less (and is mostly based on ctypes unless you've uploaded an LSIF)
<lukegb>
(sometimes I use ripgrep locally but just doing it in a web browser is kinda convenient, especially with doing f: filters and things)
orivej has joined #nixos-dev
<mkaito>
I've resorted to keeping emacs open on the other monitor on a clone of nixpkgs, and I just ripgrep
<domenkozar[m]>
maralorn: yeah disabling test suite is pretty common, most of Haskell maintainers don't want upstream to run tests
<maralorn>
domenkozar: I have had positive experiences with reporting test issues to upstream. It sometimes shows issues with implicit assumptions about context and increases my confidence in our package set.
<maralorn>
But yeah, I disable test-suites all the time in nixpkgs.^^
<domenkozar[m]>
oh yeah I always report upstream
<domenkozar[m]>
we even have arianvp who's the maintainer of servant packages here :)
<supersandro2000>
nix tries to evaluate basically anything and I don't really understand why
<lukegb>
ooh, pulseaudio 14
<lukegb>
supersandro2000: wait, what's the problem?
<supersandro2000>
nix tells me that over 5000 packages got removed when running nixpkgs-review
<lukegb>
o_O
<supersandro2000>
but then it only builds two
<supersandro2000>
and my machine died almost doing that
<lukegb>
hmm, nixpkgs-review seems happy on that PR for me
<lukegb>
(as does ofborg)
<supersandro2000>
O_o
<lukegb>
on NixOS or Darwin?
<symphorien[m]>
supersandro2000 happens to me when nix OOMs
<lukegb>
oh right, yeah, that would make sense
<symphorien[m]>
downloadmoreram
<supersandro2000>
both actually
<supersandro2000>
symphorien[m]: interesting
<supersandro2000>
but I think I did stupid stuff
LnL has quit [Ping timeout: 272 seconds]
LnL has joined #nixos-dev
LnL has quit [Changing host]
LnL has joined #nixos-dev
lassulus has quit [Ping timeout: 256 seconds]
<lukegb>
gchristensen: so is ofborg's darwin builder broken at the moment? is there anything I can do to help?
lassulus has joined #nixos-dev
lukegb has quit [Quit: ~~lukegb out~~]
<gchristensen>
was just about to reply :(
lukegb has joined #nixos-dev
LnL has quit [Ping timeout: 265 seconds]
LnL has joined #nixos-dev
LnL has quit [Changing host]
LnL has joined #nixos-dev
<gchristensen>
for those curious: I was just contacted by a person the github account I was concerned about on 2020-10-31 https://logs.nix.samueldr.com/nixos-dev/2020-10-31#4175055 was impersonating, their wikipedia page and github account was taken down and the real person has kindly asked us to never say their name again
bennofs[m] has joined #nixos-dev
red[evilred] has joined #nixos-dev
<red[evilred]>
What a peculiar situatino
<red[evilred]>
* What a peculiar situation
<jtojnar>
does running build-vm machines with sound (e.g. using env QEMU_NET_OPTS="-device intel-hda -device hda-duplex") work for you on NixOS?
<jtojnar>
for me, it fails with errors about missing some alsa .so plug-ins but maybe that is because this is Ubuntu host
FRidh has quit [Ping timeout: 260 seconds]
FRidh has joined #nixos-dev
__monty__ has joined #nixos-dev
<lukegb>
`2972 packages updated:` hmm
<sphalerite>
gchristensen: should I have a look at it now?
alp has quit [Ping timeout: 240 seconds]
Jackneill has quit [Read error: Connection reset by peer]
AlwaysLivid has quit [Remote host closed the connection]
AlwaysLivid has joined #nixos-dev
<infinisil>
What if Hydra was hooked up to PR's directly
<infinisil>
As in, every PR gets its own hydra jobset (maybe only for somewhat trusted people)
<infinisil>
And it only shows as green if there are no new failing builds
<infinisil>
And these jobsets would be deleted after the PR is merged/closed
<gchristensen>
today, hydra can't tell you that because it requires two full evaluation, build cycles to tell you if anything changed
<LnL>
unless something major changed, hydra dies
<gchristensen>
each evaluation creates 100,000+ new database records, which would also be a cost
<infinisil>
Hm I see..
<sphalerite>
Hydra does even have support for building PRs, doesn't it?
<gchristensen>
it does
<gchristensen>
hydra's scheduler might not be robust enough to balance that many things
<gchristensen>
querying the status of a jobsets is somewhat costly, and many jobsets is significantly costly, something the jobset list page already struggles with
<sphalerite>
gchristensen: should I have a look at the aarch64 machines?
<gchristensen>
ah, yeah, great!
<gchristensen>
-> -infra or -aarch64, pick your poison :)
supersandro2000 has quit [Ping timeout: 246 seconds]
alp has joined #nixos-dev
supersandro2000 has joined #nixos-dev
cole-h has joined #nixos-dev
bpye has joined #nixos-dev
saschagrunert has quit [Remote host closed the connection]
bpye has quit [Ping timeout: 240 seconds]
rajivr has quit [Quit: Connection closed for inactivity]
alp has quit [Ping timeout: 272 seconds]
justanotheruser has quit [Ping timeout: 260 seconds]
ris has joined #nixos-dev
alp has joined #nixos-dev
justanotheruser has joined #nixos-dev
<infinisil>
Idea: Make an actual test suite for nixpkgs that will be run for all PR's
<infinisil>
That one should be fast to run, and test behavior of important parts of nixpkgs
<infinisil>
I guess this probably doesn't help very much with the PR churn though..
bpye has joined #nixos-dev
<jtojnar>
how do you make a test for all important parts that is fast to run?
<supersandro2000>
I just found a maintainer whos github account no longer exists. Is it safe to remove it?
<Mic92>
on my beefy server it also constantly produces 15% cpu time and I only have one file added there.
<hexa->
as soon as you've browsed the status page you're sharing it :)
<hexa->
and there are ipfs crawlers that create quite some noise
<hexa->
I finally disabled ipfs because the ipfs init script started hanging on every rebuild
mkaito has quit [Quit: WeeChat 2.9-dev]
<Mic92>
If think hanging was resolved eventually
<domenkozar[m]>
Mic92: ipfs has always been a fair tale :)
<gchristensen>
I ran ipfs for a few minutes over a year ago and I still get people trying to connect to my node
<domenkozar[m]>
fairy*
FRidh has quit [Quit: Konversation terminated!]
<samueldr>
ipfs works, if you want big buck bunny
<samueldr>
(sorry)
<domenkozar[m]>
if you look at p2p software it has been always used by idealist and outlaws: tor, bittorent
<domenkozar[m]>
sorry played too much rdr2 today, I meant criminals.
<qyliss>
not sure that's true of tor
<qyliss>
securedrop seems to be pretty widely deployed
<domenkozar[m]>
then there are some really niche use cases, like hiding from government that is on the edge
<qyliss>
p2p Skype was widely used
<domenkozar[m]>
yeah, I mean whistleblowers are complicated bit legally, but at the end of the day you're trying to game the system, not really have a product
<domenkozar[m]>
so if p2p skype was so great, why is it legacy? :)
<qyliss>
because it didn't work well on Windows Phone, apparently
<domenkozar[m]>
the main reason was again to game the system
<domenkozar[m]>
they needed to get past firewalls
<domenkozar[m]>
and that pissed off a lot of people
<domenkozar[m]>
since it was a good way to bring down a network
<domenkozar[m]>
suddenly someone became a supernode inside an organization
<domenkozar[m]>
so lots of networks had skype banned
<qyliss>
wdym by "not really have a product"?
<qyliss>
what's the product here?
<domenkozar[m]>
to provide value that people are willing to pay for
<domenkozar[m]>
sometimes people are willing to pay to game the system, but that destroys stuff so it's always short term
<domenkozar[m]>
s/stuff/trust/
<qyliss>
and who are you saying isn't providing a product? tor?
<symphorien[m]>
quite ironical to point to tor users as the people who undermine trust on the internet ;)
<domenkozar[m]>
I would like to see a number of examples of a successful p2p products that otherwise wouldn't be better off being centralized
<domenkozar[m]>
symphorien: I said either idealists or criminals, very rarely those trying to hide from government for "good" reasons
<qyliss>
I understand a lot of multiplayer video games are p2p
<symphorien[m]>
my point was rather that big central actors don't look that good at inspiring trust today
<domenkozar[m]>
qyliss: that's a good example!
<symphorien[m]>
but it's slightly off-topic
<domenkozar[m]>
I can imagine those games aren't real-time, but that's not a requirement
<qyliss>
some of them are I think
<qyliss>
although it's a topic I know extremely little about
<domenkozar[m]>
I wonder why is p2p better there than having central servers
<qyliss>
servers are expensive I think
<domenkozar[m]>
most of the good old times it was that it handled spike
<domenkozar[m]>
but nowadays cloud makes that go away
<domenkozar[m]>
spikes*
<qyliss>
also means you can keep playing the games once they shut down the servers
<qyliss>
assuming you can connect directly to the other player somehow without needing a matchmaking thing
<domenkozar[m]>
right so I wouldn't be surprised OSS games use p2p
<domenkozar[m]>
and indie games
<qyliss>
I think there are AAA P2P games
<qyliss>
I was reading about one recently I think
<qyliss>
not sure if I'll be able to find it
<qyliss>
doesn't help that p2p now apparently also means "pay to play" in the video game context
<red[evilred]>
Soo... I see that name -> pname and version. Does that mean that name = "${name}-${version}"; is now prohibited?
<red[evilred]>
I saw mention of repology - but looking for the policy / rfc / whhatever behind it so I can understand
<domenkozar[m]>
qyliss: :D
<cole-h>
I believe Destiny 2 uses a p2p / dedicated server hybrid for stuff
<domenkozar[m]>
yeah I'm just reading on that
<cole-h>
Lots of complaining in PvP
<red[evilred]>
so should name not exist at all?
<qyliss>
I imagine pretty much every game is going to have centralised matchmaking because that's difficult to do p2p
<qyliss>
red[evilred]: it's certainly not prohibited
<qyliss>
red[evilred]: most people don't seem to do it any more, and pname/version means you can easily get the version of a derivation without parsing the name
<domenkozar[m]>
they don't seem to be able to explain why they needed p2p
<domenkozar[m]>
just that it's cool
<domenkozar[m]>
gaming industry is known to be hustle heavy
<red[evilred]>
Thanks qyliss (IRC) - I was encouraged on one place and just got feedback on a different PR to not do it
<domenkozar[m]>
The developers themselves describe their networking topology as "uniquely complicated," which is not an understatement as the document from their presentation at the Game Developers Conference 2015 shows.
<domenkozar[m]>
lol
<red[evilred]>
so was hjoping to find a documented "standard"tm
<red[evilred]>
I just foudna treewide PR that doesn't seem to be re-populating the name field - so I guess whatever discussions happened over it now resolved in that direction
<domenkozar[m]>
anyhow I'd like things to be otherwise, but I used to believe in p2p 10 years ago now I'm waiting for someone to prove it before I take the bait again /offtopic
<red[evilred]>
just wanted to make sure I got it right for future ones.
__monty__ has quit [Quit: leaving]
<domenkozar[m]>
qyliss: oh for the record, bitcoin as a store of value counts :)
<qyliss>
yeah that's a weird one
<domenkozar[m]>
it again is more to game the system
<domenkozar[m]>
you don't need it for day to day stuff
<domenkozar[m]>
but in case of war it would be invaluable
<domenkozar[m]>
but the substitue for gold is a weird one.
<JJJollyjim>
living in new zealand with great national internet but significant latency to all the clouds, i wish p2p games were more common
<domenkozar[m]>
well I'm pretty convinced starlink will solve that problem
<domenkozar[m]>
once they manage to get cross satellite communication working
<domenkozar[m]>
so technically that's p2p :P
<qyliss>
everything uses BGP :P
<domenkozar[m]>
heh
<qyliss>
and that's all fine and has no problems whatsoever
<domenkozar[m]>
:D
<domenkozar[m]>
I mean I could get drunk with people that have access to BGP
<domenkozar[m]>
so no problems at all
<JJJollyjim>
lol how so domenkozar? we don't have bad conections to the rest of the world, it's just physical distance
<JJJollyjim>
right but the physical distance means latency
<JJJollyjim>
whether via satellite or fiber
<gchristensen>
yeah its AU with terrible connections
<domenkozar[m]>
if you pick two most distant places in the world
<domenkozar[m]>
the latency in vaccum would be 133ms
<domenkozar[m]>
so that's the worst case distance wise and best case in medium
<JJJollyjim>
ah i see, it's the propogation delay thing
<JJJollyjim>
interesting
<domenkozar[m]>
but you'll get much better results with starlink
<domenkozar[m]>
less routers and in general air is better medium than even optics
<domenkozar[m]>
I studied telecommunication so I tend to have an opinion over this :D
<domenkozar[m]>
and also not that it's really rare to have optics between two points so it's quite suboptimal
<domenkozar[m]>
note*
<gchristensen>
the wild thing about AU's terrible internet is companies like Yahoo even have massive and massively underused fiber to AU but still
<JJJollyjim>
domenkozar: a traceroute to the US is like
<JJJollyjim>
15ms to hamilton (where the undersea fibre lands)
<JJJollyjim>
then the next hop is 150ms to california
<JJJollyjim>
that's not a router issue right, it's all optical?
<aristid>
i don't actually know what the situation is in australia, but i would guess australian internet is probably mostly bad due to high latencies and the fact that many common TCP implementations degrade their bandwidth when latency is high?
<gchristensen>
no, because NZ has significantly better internet
<JJJollyjim>
yeah australia had governmental fuckups, so they never got a good national network
<gchristensen>
I mean, yes and :)
<domenkozar[m]>
Jamie: optics tehoretical speed limit is 67% of speed of light
<domenkozar[m]>
while air is quite close to vaccum
<domenkozar[m]>
so it makes a difference
<JJJollyjim>
yeah, i see it could still help because of that
<JJJollyjim>
domenkozar: though wouldn't each satellite act as a router, adding latency? and there'd be a ton of satellites in the path right, cos the orbit is so low?
<aristid>
domenkozar[m]: but right now this hypothetical latency advantage of starlink is not yet realized, right?
<domenkozar[m]>
aristid: Jamie: yeah we don't know what they aim for
<aristid>
JJJollyjim: the router would hopefully not add 30ms of latency ;-)
<JJJollyjim>
aristid: iirc they don't have regulatory approval to do inter-satellite communication
<JJJollyjim>
right, but domen was saying that's a big part of the problem (which i don't think it is in the specific nz case)
<domenkozar[m]>
well I'm saying that each bit of infrastructure adds "quite a bit" to the latency and having 100% control of it under one company will make it possible for them to optimize towards the theoretical limits
<JJJollyjim>
hmm yeah
<JJJollyjim>
i guess you could pre-allocate wavelengths to be repeated to a specific next satellite, without any buffering
<JJJollyjim>
based on demand
<domenkozar[m]>
I think starlink will change internet as we know it
<domenkozar[m]>
but since it's not fashionable to like Elon, I never said that :)
<domenkozar[m]>
anyway, time for bed
<aristid>
i don't know, starlink seems mostly useful for relatively rural areas which by definition don't have that many people
<aristid>
or remote places like au/nz i guess ;-)
<domenkozar[m]>
nah
<domenkozar[m]>
that's only phase 1
<domenkozar[m]>
you'll see :)
<domenkozar[m]>
g2g.
ashkitten has quit [Quit: WeeChat 2.9]
ashkitten has joined #nixos-dev
<red[evilred]>
So - I'm looking at the output of nix-store -q --graph ~/.nix-profile
<red[evilred]>
I guess if I want to explore the entirity of nixpkgs that means I have to try to compile everything and bring everything into scope to use this method/
ashkitten has joined #nixos-dev
ashkitten has quit [Changing host]
<red[evilred]>
It's interesting - because in my current environment I have six glibc-2.27 hashes, one 2.32 hash - and a number of 2.(27|32)-bin hashes
<red[evilred]>
I'm curious to see what's causing them to split like that
AlwaysLivid has quit [Remote host closed the connection]
LnL has quit [Ping timeout: 260 seconds]
AlwaysLivid has joined #nixos-dev
lopsided98 has joined #nixos-dev
ashkitten has quit [Quit: WeeChat 2.9]
ashkitten has joined #nixos-dev
ashkitten has joined #nixos-dev
ashkitten has quit [Changing host]
ashkitten has quit [Client Quit]
ashkitten has joined #nixos-dev
justanotheruser has quit [Ping timeout: 256 seconds]