gchristensen changed the topic of #nixos-dev to: NixOS Development (#nixos for questions) | NixOS 19.09 is released! https://discourse.nixos.org/t/nixos-19-09-release/4306 | https://hydra.nixos.org/jobset/nixos/trunk-combined https://channels.nix.gsc.io/graph.html | https://r13y.com | 19.09 RMs: disasm, sphalerite; 20.03: worldofpeace, disasm | https://logs.nix.samueldr.com/nixos-dev
orivej has joined #nixos-dev
phreedom has joined #nixos-dev
phreedom_ has quit [Ping timeout: 240 seconds]
lovesegfault has quit [Quit: WeeChat 2.7]
eraserhd has quit [Ping timeout: 260 seconds]
eraserhd has joined #nixos-dev
lovesegfault has joined #nixos-dev
ris has quit [Ping timeout: 268 seconds]
ixxie has quit [Ping timeout: 240 seconds]
bhipple has quit [Ping timeout: 268 seconds]
<{^_^}> firing: BuildsStuckOverTwoDays: https://status.nixos.org/prometheus/alerts
<{^_^}> resolved: BuildsStuckOverTwoDays: https://status.nixos.org/prometheus/alerts
<infinisil> rycee: Aw man, I'm not sure how to do this scheduling
<infinisil> I feel like this should be a solved problem already
<infinisil> "Give us 5 people a date to meet when we all have time"
<infinisil> Why is this so hard
<infinisil> Painfully entering every hour of every day I have time on for the next month or so in Doodle is the best solution I can think of right now, maybe ask for absences beforehand
<infinisil> Maybe I should just use https://www.when2meet.com/ for it to be much less painful, even though it doesn't support mobile
<lovesegfault> infinisil: You wouldn't happen to know about Go nix stuff, would you?
<infinisil> lovesegfault: If you have a question about it you can just ask the channel. If I can help I'd reply :)
<lovesegfault> How are these generated? Or do people make them by hand?
<infinisil> lovesegfault: https://github.com/kamilchm/go2nix
<lovesegfault> Ah :)
<infinisil> Hm wait
<lovesegfault> I knew there had to be a tool
<infinisil> Ah yeah, that's it
<lovesegfault> Yeah, the deps.nix portion
<infinisil> See also the Go section in the nixpkgs manual
drakonis has quit [Quit: Leaving]
lovesegfault has quit [Quit: WeeChat 2.7]
pie_ has quit [Ping timeout: 268 seconds]
lovesegfault has joined #nixos-dev
bhipple has joined #nixos-dev
pie_ has joined #nixos-dev
pie_ has quit [Ping timeout: 240 seconds]
zarel_ has joined #nixos-dev
zarel has quit [Ping timeout: 258 seconds]
lovesegfault has quit [Quit: WeeChat 2.7]
bhipple has quit [Ping timeout: 265 seconds]
lovesegfault has joined #nixos-dev
justanotheruser has quit [Ping timeout: 268 seconds]
lovesegfault has quit [Quit: WeeChat 2.7]
justanotheruser has joined #nixos-dev
<yorick> tazjin: would be nice if the thing didn't 404
__monty__ has joined #nixos-dev
pie_ has joined #nixos-dev
pie_ has quit [Ping timeout: 258 seconds]
ris has joined #nixos-dev
ixxie has joined #nixos-dev
arianvp has quit [Quit: WeeChat 2.6]
arianvp has joined #nixos-dev
pie_ has joined #nixos-dev
pie_ has quit [Ping timeout: 248 seconds]
pie_ has joined #nixos-dev
genesis has quit [Remote host closed the connection]
genesis has joined #nixos-dev
orivej has quit [Ping timeout: 258 seconds]
<rycee> infinisil: Yeah, arranging meetings can be a big PIA. In any case, I will try to arrange so that I can join whenever. It's just I have difficulties guaranteeing a specific time.
<rycee> But nobody has entered times in Doodle yet? :-/
<infinisil> Yeah..
<infinisil> Maybe everybody is on holidays
<rycee> Yes, around this time it's extra tricky.
<rycee> Especially if Nix is something you do as work :-)
<infinisil> I should probably send a follow-up mail that this is canceled for now, again
<rycee> Wasn't there a regular meeting with the RFC committee? Some of these people are already in that one, right? So perhaps it would be likely that they would be available near that time…
justanotheruser has quit [Ping timeout: 252 seconds]
justanotheruser has joined #nixos-dev
mmlb has quit [Ping timeout: 268 seconds]
<domenkozar[m]> Too many heap sections: Increase MAXHINCR or MAX_HEAP_SECTS
<domenkozar[m]> on trunk-combined :/
<samueldr> oh, and it might have been for more than one eval
<samueldr> last eval is more than 48h older than last checked
<samueldr> re-queueing an eval, to see
<domenkozar[m]> thanks :)
<samueldr> though it's likely the old issue that had been worked around by splitting the aarch64 eval from the main eval
<samueldr> the way the hydra eval works is by restarting the eval every time the memory consumption is too great, but AFAIUI it keeps the same GC uh... "area", which gets more and more fragmented
<samueldr> restarting and continuing the eval where it was restarted at*
<samueldr> it might finally be at the point where there's no workaround; this would need to be fixed in the eval code (both nix and hydra) or by reducing the footprint of nixpkgs (probably hard!)
<domenkozar[m]> I wonder how much would floating out lets help
<domenkozar[m]> I know that I once made 1h evaluation be 3min by doing that what a compiler could :)
<samueldr> I'm leaving the language-side hypothetics to those that know more :)
<samueldr> I *think* this could be solved by precise-gc https://github.com/NixOS/nix/tree/precise-gc
<samueldr> or if not solved, worked around and pushed for a later similar issue
<samueldr> this removes the use of boehm gc
<domenkozar[m]> sounds like a really nice GSOC project
<samueldr> btw, shouting out loud in the void^W room: the eval *has* to happen with the hydra evaluator to reproduce this!
janneke has joined #nixos-dev
bhipple has joined #nixos-dev
drakonis has joined #nixos-dev
<jtojnar> ugh, accidentally created `staging-ne` branch in the GitHub UI and now cannot delete it
<clever> jtojnar: i often type `ast` to search for master (skip the right side of the keyboard), and accidentally made an ast branch one day, lol
<gchristensen> jtojnar: got it
<jtojnar> thanks
<gchristensen> a few others need cleaning up too (nixos-option-description-swap-file)
orivej has joined #nixos-dev
<{^_^}> firing: BuildsStuckOverTwoDays: https://status.nixos.org/prometheus/alerts
<samueldr> gchristensen: ^ would that fire for an eval not eval'ing for over two days?
<gchristensen> no
<samueldr> then it's something else
<gchristensen> a build was running for just over 2 days
<{^_^}> resolved: BuildsStuckOverTwoDays: https://status.nixos.org/prometheus/alerts
orivej has quit [Ping timeout: 260 seconds]
pie_ has quit [Remote host closed the connection]
pie_ has joined #nixos-dev
<samueldr> re-queued nixos:trunk-combined, though I guess it's not going to change much considering it evaluated at least 2 times, probably 4 and failed
<gchristensen> what is failing? :o
<samueldr> [12:56:34] <domenkozar[m]> Too many heap sections: Increase MAXHINCR or MAX_HEAP_SECTS
<samueldr> [12:56:42] <domenkozar[m]> on trunk-combined :/
<samueldr> the time has come
<gchristensen> wtf
__monty__ has quit [Quit: leaving]
ixxie has quit [Ping timeout: 260 seconds]
lovesegfault has joined #nixos-dev
<lovesegfault> failure seems spurious
<samueldr> nixos:trunk-combined eval'd, so it's likely only at the cusp, but enough to trip up more often than not it looks like
<{^_^}> firing: BuildsStuckOverTwoDays: https://status.nixos.org/prometheus/alerts
<gchristensen> hrm :?
<samueldr> do you remember the previous issues with "Too many heap sections: Increase MAXHINCR or MAX_HEAP_SECTS" ?
pie_ has quit [Quit: No Ping reply in 180 seconds.]
<samueldr> well, this is it again, but with nothing to cut now
<gchristensen> solved
<samueldr> hm?
<gchristensen> the firing alert
<samueldr> ah
pie_ has joined #nixos-dev
<gchristensen> this is the one with the underlying problem being how Nix manages memory? like, throwing more memory a it doesn't solve it?
<samueldr> exactly
<samueldr> well
<samueldr> now that you say it that way
<samueldr> I guess upping the hydra evaluator maximum memory usage (before it restarts) could help us a bit
<{^_^}> resolved: BuildsStuckOverTwoDays: https://status.nixos.org/prometheus/alerts
<samueldr> if it's fragmentation as I suspect, it will restart less often, thus fragment less
<samueldr> though it still needs to be solved to actually put aarch64 builds in the same eval
<gchristensen> which tunables should we poke?
pie_ has quit [Excess Flood]
pie_ has joined #nixos-dev
<samueldr> though now that I see that option again, I don't remember what was the results in testing locally
<samueldr> irb(main):003:0> (1 << 30).to_f / 1024 / 1024 / 1024
<samueldr> => 1.0
<samueldr> just in case you, too, don't read bitshifts fluently
<samueldr> evaluator_max_heap_size, not initial
<gchristensen> oops
<samueldr> though uh
<samueldr> that should also touch it
<samueldr> so
<gchristensen> okay, increasing ram
<gchristensen> memory*
<samueldr> so yeah, that should make it 22GiB already
<samueldr> so I think it's at the point where I think I observed on my workstation that throwing more maximum memory didn't actually help :/
<gchristensen> well, it is 30 now, so ... let's find out
<samueldr> yeah, definitely
<samueldr> not saying "don't try it", but thinking it's probably time this gets looked at properly
<gchristensen> mind reminding me of this fact on Tuesday? :)
<samueldr> 31st, will try
<gchristensen> thanks
<gchristensen> https://github.com/NixOS/hydra/pull/699 but for evaluations
<{^_^}> hydra#699 (by grahamc, 1 hour ago, open): job: create a prometheus endpoint
<gchristensen> okay I'm going back to the family
<gchristensen> fingers crossed
<samueldr> have fun!
pie_ has quit [Ping timeout: 265 seconds]
pie_ has joined #nixos-dev
pie_ has quit [Ping timeout: 248 seconds]