<supersandro2000>
the client sends a header which compression it supports and the server provides that
<supersandro2000>
those files can be pre compressed and served statically
<lukegb>
right, except these files are uploaded to an S3 bucket
<lukegb>
and the compression happens inside nix and not in curl
<lukegb>
but other than that, yes, I agree that the HTTP Accept-Encoding header
<lukegb>
exists
<supersandro2000>
yeah and if nix does the compression wrong and they are not encrypted on the fly then the curl in nix downloads brotli since only a few months
<supersandro2000>
because curl did not have brotli enabled until I moved it from curlComplete to curl
<lukegb>
looks like it's a bug in the S3 cache implementation specifically
<lukegb>
found it
<lukegb>
decompress() no longer checks for none/"", so if something isn't compressed it breaks
<lukegb>
I'm confused as to how this isn't affecting nixos-20.09, though... I *suspect* that none of the binaries with debuginfo outputs have been rebuilt in that channel, but I'm not sure
cole-h has quit [Ping timeout: 240 seconds]
{`-`} has joined #nixos-infra
Ericson2314 has quit [*.net *.split]
garbas[m] has quit [*.net *.split]
domenkozar[m] has quit [*.net *.split]
gchristensen has quit [*.net *.split]
supersandro2000 has quit [*.net *.split]
aminechikhaoui has quit [*.net *.split]
asymmetric has quit [*.net *.split]
V has quit [*.net *.split]
thefloweringash has quit [*.net *.split]
zimbatm[m] has quit [*.net *.split]
roberth has quit [*.net *.split]
andi- has quit [*.net *.split]
flokli has quit [*.net *.split]
eyJhb has quit [*.net *.split]
niksnut has quit [*.net *.split]
lukegb has quit [*.net *.split]
pie_ has quit [*.net *.split]
edef has quit [*.net *.split]
samueldr has quit [*.net *.split]
maralorn has quit [*.net *.split]
hexa- has quit [*.net *.split]
nh2[m] has quit [*.net *.split]
colemickens has quit [*.net *.split]
NinjaTrappeur has quit [*.net *.split]
mcint has quit [*.net *.split]
endocrimes has quit [*.net *.split]
qyliss has quit [*.net *.split]
aristid has quit [*.net *.split]
tazjin has quit [*.net *.split]
ikwildrpepper has quit [*.net *.split]
JJJollyjim has quit [*.net *.split]
ryantm has quit [*.net *.split]
Ox4A6F has quit [*.net *.split]
MichaelRaskin has quit [*.net *.split]
sterni has quit [*.net *.split]
sterni has joined #nixos-infra
MichaelRaskin has joined #nixos-infra
garbas[m] has joined #nixos-infra
niksnut has joined #nixos-infra
JJJollyjim has joined #nixos-infra
zimbatm[m] has joined #nixos-infra
V has joined #nixos-infra
mcint has joined #nixos-infra
gchristensen has joined #nixos-infra
domenkozar[m] has joined #nixos-infra
flokli has joined #nixos-infra
asymmetric has joined #nixos-infra
ikwildrpepper has joined #nixos-infra
aminechikhaoui has joined #nixos-infra
andi- has joined #nixos-infra
tazjin has joined #nixos-infra
Ox4A6F has joined #nixos-infra
samueldr has joined #nixos-infra
eyJhb has joined #nixos-infra
<gchristensen>
<3
aristid has joined #nixos-infra
<hexa->
needs more {^_^} here
<hexa->
lukegb++
<qyliss>
How much RAM do non-big-parallel machines have?
<qyliss>
#113817 needs 15G of RAM and the author is asking if they should indicate to Hydra that a lot of RAM is needed
<qyliss>
oh, no bot in here
<hexa->
needs more {^_^} here
<hexa->
lmao, one up, enter.
<gchristensen>
15G of RAM? yikes
<gchristensen>
I don't think there is any way we can promise any given build has that much ram available
<gchristensen>
that said, it may not be a problem
<gchristensen>
thinking about drew being unhappy that zig uses like 1G of ram :P
<qyliss>
gchristensen: should I tell them to use big-parallel or no, then?
<gchristensen>
I'm curious if it will take advantage of the big paralell-eyness
<qyliss>
it appears that is does not
<qyliss>
(testing it rn)
<qyliss>
it's not using that much memory for me atm either though
<qyliss>
I guess I'll just leave it as is
<gchristensen>
imho leave it untagged and if it is troublesome we could mark it big-parallel and if that fails we could have a bigger conversation about features
<qyliss>
sounds good
<gchristensen>
it might be that our regular-capacity machines are bigger than their regular machines
<lukegb>
I kinda wish nix had better checking-out-cores-as-resources functionality like that one PR was adding
<lukegb>
because we have a lot of builds that only use ~a single core, and some builds that want ALL the resources
<gchristensen>
my "implementation" section is out of date, but the rest stands
MichaelRaskin has quit [Remote host closed the connection]
<lukegb>
gchristensen: ooh this is neat
<lukegb>
pseudo-derivations like tested also make me a bit sad, because we send a bunch of inputs to a builder which then doesn't do anything with them, but that's really not anywhere close to being a big problem
<gchristensen>
aggregate jobsets no longer have a nix derivation-level dependency on it anymore
<gchristensen>
hydra evaluates it itself, and tested sees a list of attribute names to consider
<lukegb>
ah, hm, interesting
<gchristensen>
the memory requirements of keeping dozens of NixOS machines around to evaluate tested became significant