gchristensen changed the topic of #nixos-dev to: NixOS Development (#nixos for questions) | https://hydra.nixos.org/jobset/nixos/trunk-combined https://channels.nix.gsc.io/graph.html | 18.03 release managers: fpletz and vcunat | https://logs.nix.samueldr.com/nixos-dev
Sonarpulse has quit [Ping timeout: 252 seconds]
jtojnar_ has joined #nixos-dev
jtojnar has quit [Read error: Connection reset by peer]
jtojnar_ has quit [Ping timeout: 264 seconds]
jtojnar has joined #nixos-dev
jtojnar_ has joined #nixos-dev
jtojnar has quit [Read error: Connection reset by peer]
jtojnar has joined #nixos-dev
jtojnar_ has quit [Ping timeout: 260 seconds]
Synthetica has quit [Quit: Connection closed for inactivity]
mbrgm has quit [Ping timeout: 245 seconds]
mbrgm has joined #nixos-dev
lassulus_ has joined #nixos-dev
lassulus has quit [Ping timeout: 252 seconds]
lassulus_ is now known as lassulus
acowley has quit [Ping timeout: 264 seconds]
acowley has joined #nixos-dev
<dtz> $#@ I hope i've managed to isolate the SNAFU i've been chasing all day
<dtz> fwiw seems to be after bumping curl to 7.60 -- but a)needs more testing to confirm, very much so (combination of lots of debugging/tracing and git bisect lol) b)might be specific to the rest of my tree :(
<dtz> manifests as an annoying change that I can go into more detail about but in summary causes various nix requests to always take exactly 10 seconds before actually issuing the right network request lol
<dtz> (obviously unacceptable haha :D)
<dtz> oh no, first test of current staging's "nix" shows that it exhibits the same behavior
<dtz> test case: run 'nix log foo' but not sure what env is needed beyond .... my laptop lol. dual-stack ipv4/ipv6 but I suspect it needs to be combined with ipv6 being lame
Lisanna has joined #nixos-dev
Sonarpulse has joined #nixos-dev
<andi-> Is there some way to know what the current state of 18.03 staging is besides looking at the hydra builds (which appears to be at the nominal baseline of failures)? The whole proces when and by whom that is being merged into the release is a bit clody for me..
<andi-> well it seems that has happened some 30min ago (for 18.03)… such bad luck
MichaelRaskin has quit [Quit: MichaelRaskin]
__Sander__ has joined #nixos-dev
globin has quit [Ping timeout: 245 seconds]
fpletz has quit [Ping timeout: 240 seconds]
yorick has quit [Ping timeout: 265 seconds]
hiberno has quit [Ping timeout: 256 seconds]
sphalerite has quit [Ping timeout: 240 seconds]
tilpner has quit [Ping timeout: 260 seconds]
tilpner has joined #nixos-dev
yorick has joined #nixos-dev
fpletz has joined #nixos-dev
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-dev
sphalerite has joined #nixos-dev
globin has joined #nixos-dev
<gchristensen> niksnut: may I turn on Hydra's ability to make a channel for ofborg? I fear I've found a weird bug in Hydra, and would like to see what hydra.nixos.org does
<niksnut> ability to make a channel?
<gchristensen> oh nevermind, channels are disabled for the whole server
<niksnut> ah, I thought channel generation was removed completely
<gchristensen> in short, when importing the nixexpers.tar.bz2 from a hydra, the outPaths are strings not paths, and the strings have no context depending upon the paths they contain. I had to manually `/. + fromhydra.myproject` to force the context, and even then Nix didn't try to fetch that path from a cache
<gchristensen> I think a lot of companies use hydra's ability to make channels
<copumpkin> niksnut: spotted some more weirdness with NixOS on EC2, wondering if you might know what's up: when I upgrade my homegrown NixOS AMIs to be based off of 18.03, the networking kinda works, but a lot of TCP streams hang. Oddly enough, if I turn on sriov networking support at the instance/image level, that gets fixed. sriov is probably more desirable anyway, but now I'm curious what changed on the linux level that would break things that
<copumpkin> way on non-sriov interfaces (including older m3 instances and such which don't support it)
<copumpkin> (the official NixOS AMI has the sriov flag turned on so it hasn't been a problem, but I could see it being a problem on older instance types)
<niksnut> doesn't sound familiar
aminechikhaoui has quit [Ping timeout: 245 seconds]
Sonarpulse has quit [Ping timeout: 252 seconds]
Lisanna has quit [Quit: Lisanna]
WilliButz has quit [Quit: WeeChat 2.0]
WilliButz has joined #nixos-dev
ciil has quit [Quit: leaving]
Lisanna has joined #nixos-dev
__Sander__ has quit [Quit: Konversation terminated!]
Sonarpulse has joined #nixos-dev
ciil has joined #nixos-dev
globin has quit [Ping timeout: 260 seconds]
Mic92_ has quit [Ping timeout: 265 seconds]
Mic92_ has joined #nixos-dev
<thoughtpolice> I'll ask again, but I'd really really like any opinions on https://github.com/NixOS/nixpkgs/pull/38698 -- especially from anyone running PostgreSQL in practice.
<thoughtpolice> I am planning on posting to the mailing list, but barring any major objections I would like to land this soon and follow it up with a few more things.
<thoughtpolice> Probably the most relevant thing to anyone running Postgres now is that it will change the default version to 10. If you have stateVersion set this should not be an issue, though.
<thoughtpolice> dtz: also, hi, I am @stdlib ;)
<dtz> thoughtpolice: oh, haha hi! ^_^
globin has joined #nixos-dev
<Sonarpulse> thoughtpolice: one nit is withPostgresql is a bit nicer than nulling out
<Sonarpulse> some of us want to get rid of nulling out packages altogether
<Sonarpulse> but yeah nice suff!
<thoughtpolice> You mean the GDAL change?
<thoughtpolice> (Not sure I follow exactly)
<Sonarpulse> thoughtpolice: yes
<thoughtpolice> Ah, yeah that can be fixed pretty easily. No real reason to pick that one if not necessary
<Sonarpulse> thoughtpolice: pick that one what?
<thoughtpolice> nulling vs withPostgres, I mean
<Sonarpulse> oh pick the nit haha
<thoughtpolice> ("withPostgresql" seems to be the normal name people have chosen)
<Sonarpulse> yeah it's very minor
<thoughtpolice> After this lands, I would like to try and add PostgreSQL 11 betas soon, too. Some features like the new JIT will probably require some build infrastructure changes for LLVM support for the 11 release, so it'd be nice to sort that out now. Not sure if that's worth it, though.
<Sonarpulse> ooooo jitted queries?
<thoughtpolice> Unlike Linux, there's really only a 4-5 month period per year where a _testing package would make sense for PG (vs Linux, where _testing is more or less only pointless when merge windows are open and after a release)
<thoughtpolice> It's probably valid closer to 8 months out of the year.
<Sonarpulse> hehe as if linux was tested
<Sonarpulse> at all
<thoughtpolice> Sonarpulse: Yeah, but part of it is that in order to JIT tuple decomposition into faster code, all of the bitcode for all of PostgreSQL has to be included in the installed package.
<thoughtpolice> So there's some testing that needs to be done outside of a "--with-llvm" flag, I'm gonna guess.
<thoughtpolice> But maybe not much more.
<Sonarpulse> mmmm that's a sledgehammer
<Sonarpulse> anyways great work
<Sonarpulse> lunch now
<thoughtpolice> I don't think there's any way to work around it, because they more or less want to inline actual tupling C code into the queries, since that's what does the deconstruction.
<manveru> does anyone have download stats for our s3 mirror?
MichaelRaskin has joined #nixos-dev
Mic92_ has quit [Ping timeout: 256 seconds]
Mic92_ has joined #nixos-dev
Mic92_ has quit [Ping timeout: 245 seconds]
<Sonarpulse> thoughtpolice: well i suppose it knows how to not needlessly recompile the non-inlined parts
<thoughtpolice> Dunno if they do. Might not even matter much if they don't, right now. It only kicks in for certain query plans that will be "expensive" so the dominating time will probably be disk I/O and actual CPU time on row processing, vs any compile time of the query.
<Sonarpulse> thoughtpolice: oh i was just thinking about transitively bringing in an entire extra copy of postgres
<thoughtpolice> Oh, yeah. They probably don't do that but I'm not entirely sure how they set it up. They do definitely ship bitcode though, from my understanding.
<Sonarpulse> thoughtpolice: after all the cross work, heh I shudder at how that build system is set up
_rvl has quit [Ping timeout: 256 seconds]
FRidh has joined #nixos-dev
_rvl has joined #nixos-dev
aminechikhaoui has joined #nixos-dev
Mic92 has joined #nixos-dev
<copumpkin> niksnut: I think I'm still getting the "claims to be content addressed but isn't" message in 2.0.2 :(
<copumpkin> manveru: niksnut does
<copumpkin> well, he has access logs, which can be used to compute most things :)
<gchristensen> manveru: what're you wondering?
obadz has quit [Ping timeout: 252 seconds]
aminechikhaoui has quit [Ping timeout: 268 seconds]
aminechikhaoui has joined #nixos-dev
obadz has joined #nixos-dev
<manveru> just had a coworker ask how many people are using nixos, so i thought we could maybe estimate it using download stats
<manveru> not sure how exactly though, even with that :|
FRidh has quit [Quit: Konversation terminated!]
ChanServ has quit [shutting down]
<Sonarpulse> peti: did you see https://github.com/NixOS/nixpkgs/pull/40996 yet?
ChanServ has joined #nixos-dev
aminechikhaoui has quit [Ping timeout: 252 seconds]
aminechikhaoui has joined #nixos-dev
ChanServ has quit [shutting down]
<gchristensen> bye, chanserv
<ikwildrpepper> manveru: we have logs of binary cache, but don't run regular stats on them
<ikwildrpepper> also, it's a bit hard to say exactly, as there are also a lot of machines that get started in big batches by my former employer (10k's a month)
ChanServ has joined #nixos-dev
ocharles is now known as Guest27988
{^_^} is now known as Guest93247
stqism is now known as Guest49421
gchristensen is now known as Guest50191
garbas is now known as Guest55520
Guest50191 is now known as gchristensen
<peti> Sonarpulse: I saw it, but I won't have much time to review those changes meaningfully in the next couple of days. I'm at openSUSE conference in Prague right now and won't be back home for a week or so.
<Sonarpulse> peti: could you put on haskell-updates?
<Sonarpulse> I'm pretty confident on the concept of those chases
<Sonarpulse> I just don't have the capacity to test everything
<Sonarpulse> and don't want to mess with your job
<ikwildrpepper> manveru: e.g. we got requests from 6686 unique IP addresses yesterday
<peti> Sonarpulse: https://hydra.nixos.org/eval/1459088 is building your PR.
<Sonarpulse> peti: thank you!
<peti> Sonarpulse: No problem. :-)
<Sonarpulse> if you want I could not merge to master, or only merge to staging, until you come back and can review too
<gchristensen> ikwildrpepper: not to mention all the NATs out there
<ikwildrpepper> yeah
<ikwildrpepper> looks like yesterday there were 3800 machines from LB :D
<peti> Sonarpulse: No, there's no need to delay that long. Just merge the changes once you're confident they work.
<Sonarpulse> peti: ok. thanks again!
<gchristensen> ikwildrpepper: sshhhhh
<Sonarpulse> but do hydra and ofborg share s3?
<thoughtpolice> ikwildrpepper: Yeah, and things like the rise of Continuous Integration, Everywhere, All The Time also make stats harder and increase bandwidth consumption substantially. Hackage for example puts out enormous quantities of bandwidth despite the relative size of the community thanks to things like Travis-CI.
<gchristensen> no
<thoughtpolice> Like, tens of TB a month at this point, I'm pretty sure.
<ikwildrpepper> yeah, we do about 40TB for cache.nixos.org atm
<ikwildrpepper> but that doesn't include the hydra.nixos.org machines downloading/uploading
<Sonarpulse> gchristensen: oh hah I saw https://hydra.nixos.org/eval/1459088#tabs-still-succeed
<Sonarpulse> but missed the "binary"
<Sonarpulse> so thought that was ofborg's builds populating
<Sonarpulse> oh well
makefu has quit [Quit: WeeChat 1.9.1]
makefu has joined #nixos-dev
<thoughtpolice> ikwildrpepper: Are you just not counting that bandwidth, or does AWS not even account for it since it's internal (and thus not charged?)
jtojnar has quit [Remote host closed the connection]
<ikwildrpepper> thoughtpolice: they pull directly from s3, not via cloudfront
<thoughtpolice> ahh
<ikwildrpepper> partly it is inside aws
<ikwildrpepper> 2417 unique IP's that is not LB, in the last 4 days or so
srhb is now known as srhb_
srhb_ is now known as srhb
srhb is now known as srhb_
srhb_ is now known as srhb
<manveru> cool, thanks :)
<ikwildrpepper> checking for whole april atm, but takes a moment to dl
drakonis has joined #nixos-dev
Guest93247 has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev
drakonis has quit [Remote host closed the connection]
<ikwildrpepper> manveru: 11197 unique IP's in April (not LB)
<manveru> that's pretty awesome
<MichaelRaskin> Well, LB might be not the only mass consumer…
<ikwildrpepper> MichaelRaskin: yeah, just mentioning it, because I filter them out explicitly
<manveru> true, but IPs aren't users either
<manveru> anw, my guess was about 10x number of IRC users in #nixos, and that's close enough
jtojnar has joined #nixos-dev
jtojnar has quit [Remote host closed the connection]
yorick has quit [Quit: brb]
jtojnar has joined #nixos-dev
<copumpkin> also, I have countless machines behind a handful of AWS managed NATs
<copumpkin> I mostly use my own binary cache, but do occasionally touch cache.nixos.org
<clever> copumpkin: https://github.com/cleverca22/cachecache is a project id like to finish, the idea is to have a transparent proxy/cache for a binary cache, that can also multiplex to several upstream caches
jtojnar has quit [Quit: jtojnar]
jtojnar has joined #nixos-dev
<pstn> Can we get https://github.com/NixOS/nixpkgs/pull/41036 backported to stable, at least as an optional package? I guess a lot of people need to use that version due to gdpr compliance.
drakonis has joined #nixos-dev
Lisanna has quit [Ping timeout: 252 seconds]