<worldofpeace>
it'll get far enough parsing some builds, but it will try to get a certain one and then fail. I just did 1571034 all cleaned up and it now fails at Error downloading https://hydra.nixos.org/build/113050517
<samueldr>
and you do retry?
<worldofpeace>
will now retry with the new cache
<samueldr>
I mean, when you hit a (presumed) download failure, do you try to re-run without evicting from the cache?
<worldofpeace>
Yes, but it still happens even if I evict the cache
<worldofpeace>
I'm considering my network is just having a funny time with hydra 🤣, and I just encountered non robust fetching
<samueldr>
in this prsumed case, I don't think there is a cached error
<samueldr>
can you use `curl $thaturl` right after an error?
<samueldr>
I'm thinking I shouldn't have made curl quiet even though I was asked nicely :)
<worldofpeace>
samueldr: hmmm, it worked with --show-error --fail
<samueldr>
I added a commit with those nice command line parameters
<samueldr>
if you pull, "hopefully" (not really) you'll face a timeout and see if it helps knowing
<worldofpeace>
samueldr: yeah, that helps to actually know what happened. `curl: (28) Failed to connect to hydra.nixos.org port 443: Connection timed out` is the output
<worldofpeace>
thanks anyways
<samueldr>
"great" :/
<samueldr>
get a worse internet connection maybe?
<samueldr>
(mine isn't that great)
<samueldr>
worldofpeace: maybe add "--connect-timeout", "120", or something bigger in fetch.rb where the curl command is being built
<samueldr>
if you can reproduce faithfully enough
<worldofpeace>
samueldr: you read my mind, I was just searching for that
<worldofpeace>
and maybe doing less connections at a time would help too
<samueldr>
there is no parallelization
<samueldr>
it all happens serially, in most part because I don't want to hit hydra harshly
justanotheruser has quit [Ping timeout: 272 seconds]
justanotheruser has joined #nixos-dev
morgib has joined #nixos-dev
bhipple has quit [Remote host closed the connection]
tilpner_ has joined #nixos-dev
tilpner has quit [Ping timeout: 268 seconds]
cole-h has quit [Ping timeout: 272 seconds]
ris has quit [Ping timeout: 245 seconds]
Jackneill has joined #nixos-dev
kalbasit has quit [Ping timeout: 248 seconds]
lightbulbjim has quit [Ping timeout: 248 seconds]
lightbulbjim has joined #nixos-dev
morgib has quit [Ping timeout: 248 seconds]
rajivr___ has quit [Ping timeout: 248 seconds]
cbarrett has quit [Ping timeout: 248 seconds]
vdemeester has quit [Ping timeout: 248 seconds]
davidtwco has quit [Ping timeout: 248 seconds]
chrisaw has quit [Ping timeout: 248 seconds]
johanot has quit [Ping timeout: 248 seconds]
chrisaw has joined #nixos-dev
vdemeester has joined #nixos-dev
kalbasit has joined #nixos-dev
cbarrett has joined #nixos-dev
johanot has joined #nixos-dev
davidtwco has joined #nixos-dev
morgib has joined #nixos-dev
rajivr___ has joined #nixos-dev
orivej has quit [Ping timeout: 240 seconds]
morgib has quit [Quit: Connection closed for inactivity]
orivej has joined #nixos-dev
<yorick>
gchristensen: hmm
<yorick>
wonder why that didn't show up in testing
<yorick>
(narrator: because you didn't rebase your testing branch)
lassulus has quit [Ping timeout: 272 seconds]
Synthetica has joined #nixos-dev
<niksnut>
hm: Feb 19 11:47:12 bastion r5sndqqxf7gb6z6x4b9m22882h97x0gz-unit-script-update-nixos-20.03-small-start[6649]: channel would go back in time from nixos-20.03beta-552.08c4b48d3ab to nixos-20.03beta-483.c2846eeac5f, bailing out
<niksnut>
did the revcount delta change or something?
clkamp_ has joined #nixos-dev
clkamp_ has quit [Remote host closed the connection]
clkamp_ has joined #nixos-dev
eyJhb has quit [Quit: Clever message]
layus[m] has quit [*.net *.split]
worldofpeace has quit [*.net *.split]
<gchristensen>
niksnut: I changed it once before the channel was created, since it was creted with way too high a number
<{^_^}>
nix#3216 (by Ma27, 14 weeks ago, open): Add an optional `hash` parameter to `builtins.fetchGit`
<gchristensen>
r13y: -\_ 96.77% after iso_minimal grew to depend on 40 more paths
clkamp_ has quit [Remote host closed the connection]
<niksnut>
ma27[m]: thanks, I just added a comment
<ma27[m]>
niksnut: thx for taking a look at this :) As far as I understand your comment, each fetcher now has a `narHash`-attribute and this change is therefore obsolete?
psyanticy has joined #nixos-dev
<niksnut>
yes, at least that's the idea
<niksnut>
I still need to add some test cases
orivej has quit [Remote host closed the connection]
<{^_^}>
Nicotine-Plus/nicotine-plus#99 (by dvzrv, 13 weeks ago, open): Port to python3
orivej has joined #nixos-dev
ixxie has joined #nixos-dev
CRTified has joined #nixos-dev
bridge[evilred] has quit [Remote host closed the connection]
bridge[evilred] has joined #nixos-dev
red[evilred] has joined #nixos-dev
<gchristensen>
samueldr: I'm thinking about making a handful ~half the aarch64 builders "big-parallel" with few jobs/many cores. does that sound like a good improvement to you?
<srk>
+1
<srk>
can I somehow help with armv7l builders?
<srk>
thefloweringash: ^
<thefloweringash>
Ooh, I like that change. My motivation is to get chromium cached: see #78347
<thefloweringash>
The PR for the builders is here: https://github.com/grahamc/packet-nix-builder/pull/2. The only remaining non-trivial task I’m aware of is to update the packet importer. Maybe pick over the PR and add a review?
<{^_^}>
grahamc/packet-nix-builder#2 (by thefloweringash, 2 weeks ago, open): armv7l-linux build vm for aarch64-linux machines
<srk>
thefloweringash++
<{^_^}>
thefloweringash's karma got increased to 10
<gchristensen>
it is very strange to me just how long it has taken for the first `tested` job to finish
<srk>
epic
<gchristensen>
thefloweringash: I think it looks okay, it is on my to-do to review ... sorry for the long wait :(
drakonis has joined #nixos-dev
<samueldr>
gchristensen: sounds good
<samueldr>
though I think the most accessible actual solution to the problem requires that longer builds don't get sent to busy builders, and trivial builds to "big-parallel" builders
<gchristensen>
so there is the easy way (mark them as big-parallel) and the hard way (implement a different scheduling algorithm)
<Profpatsch>
It doesn’t look like they are generated automatically.
<Profpatsch>
Or at least nix is not the only program depending on them to be there, so they shouldn’t depend on nix creating them at some undefined time.
<gchristensen>
it is unfortunate but we can't really und othat
<gchristensen>
and we can't revert it in stable, Profpatsch -- we'll have to fix the tools
<Profpatsch>
gchristensen: but how should they be fixed?
<samueldr>
is it related to the issue on the aarch64 builder where nix-env doesn't work without mkdir'ing /nix/var/nix/profiles/per-user/$USER and switching profile beforehand?
<Profpatsch>
gchristensen: I think nixos should generate the directories.
<Profpatsch>
Doesn’t matter if nix magically generates them on first use
<Profpatsch>
nixos knows the users (at least in immutable mode, which should be the default), so nixos should set them up.
<Profpatsch>
For my use-case, I might be able to hack something together with activationScripts.
edef has quit [Ping timeout: 260 seconds]
edef has joined #nixos-dev
Jackneill has quit [Remote host closed the connection]
__monty__ has joined #nixos-dev
phreedom has quit [Ping timeout: 240 seconds]
phreedom has joined #nixos-dev
orivej has quit [Ping timeout: 255 seconds]
rajivr___ has quit [Quit: Connection closed for inactivity]
<gchristensen>
Profpatsch: it should automatiactlyl happen when they log in
<gchristensen>
if it isn't, that is the bug to fix IMO
<samueldr>
sure sounds like the issue that's happening on the aarch64 builder
<gchristensen>
we should fix it :P
<gchristensen>
okay this is interesting
<gchristensen>
I've figured out why this dang chromium test is taking freaking forever to "start"
<gchristensen>
nixos:release-20.03:nixos.tests.chromium.x86_64-linux depends on chromium but chromium is not a job
<gchristensen>
so it is just in the background building forever
<gchristensen>
treated as if it were a trivial writeText
<gchristensen>
https://hydra.nixos.org/build/112999422 is the same chromium the test depends on. should it get a proper job, or should the test be deleted? I sort of assume someone tried to remove the job
<ajs124>
isn't that how this always works? jobs (including tests) are blocked until all of their dependencies are built?
<gchristensen>
yes, but the jobset doesn't list chromium as a dependency
<gchristensen>
more concretely, Hydra doesn't have an entrty here https://hydra.nixos.org/eval/1571201 for every little derivation -- like fetching sources -- only the ones named as top-level jobs
<gchristensen>
the test depends on chromium but there isn't a top-level job for it in the evaluation
<gchristensen>
usually depndencies which don't get names take ~no time at all to run. it is unusual for a build to have a significant dependency
<gchristensen>
which is unnamed
<gchristensen>
oh what I'm totally wrong I take everything back the job is there :|
<ajs124>
I thought there was a job for every nixpkgs attribute in the release jobsets?
* gchristensen
has hydra blindness
<ajs124>
Well, all minus broken etc
* emily
is kinda confused too
<samueldr>
gchristensen: confusingly, you linked to a different eval?
red[evilred] has quit [Quit: Idle timeout reached: 10800s]
<gchristensen>
samueldr: looking at the numbers, maybe even just having 1 machine which is exclusively big-parallel might be sufficient
ris has joined #nixos-dev
<gchristensen>
I need to stop being fixated on this job and do other work lol
psyanticy has quit [Quit: Connection closed for inactivity]
orivej has joined #nixos-dev
johnny101m has quit [Quit: -a- Connection Timed Out]
johnny101m has joined #nixos-dev
lovesegfault has joined #nixos-dev
<emily>
oh hey https://github.com/NixOS/nixpkgs/pull/79026 is ZHF-related and a package I'm theoretically meant to maintain... but apparently you can't do the fancy PR-editing stuff as just a maintainers rgoup member, aw