{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev
gchristensen has joined #nixos-dev
<clever>
gchristensen: did you catch why cache=unsafe helps performance?
<gchristensen>
I didn't
<gchristensen>
what is the story with that? :)
<clever>
gchristensen: basically, it tells qemu to not flush() disk writes, when telling the client things are done
<gchristensen>
oh cool
<clever>
so data may wind up in caches, and get lost at improper shutdown
<clever>
but, your doing zfs rollbacks, so any partially written data will vanish!
<gchristensen>
sounds good :)
<clever>
it is a bit odd, that the osx-kvm thing has it in the example though
<clever>
since that risks data loss, normally
<gchristensen>
ehh that IS shady
<samueldr>
#62262 would make sd_image just big enough to make hydra builds fail due to exceeded output size. Do we have anything against tweaking the limits for our hydra? I would guess so, but it's been flirting with the limit for a good while already, so I'm asking anyway
<samueldr>
(that's an alternative, and not a real solution, just looking at options)
<samueldr>
I wonder if firmware-linux-nonfree could have a "slimmed" variant; 446MiB, 139 of those already look easy to get rid of, unless mistaken netronome is fancy serverey thing that would be unlikely to be found on an end-user's machine
<samueldr>
though nothing obvious except amdgpu for arm... and even then that's so SBC centric to say
orivej has joined #nixos-dev
<ekleog>
one solution might be having one package per firmware and a firmware-linux-nonfree that combines them all? (plus potentially firmware-linux-desktop-nonfree or similar)
<samueldr>
looking at the firmwares, looks like there's no easy decision for anything else than the netronome ones
<samueldr>
or it'd be next to no gain
<samueldr>
it does count for ~31% of all firmwares in there
<samueldr>
if they were filed in categories like "gpu, wireless, network" it'd make it easier to split I think, but looks like it's all mixed in together /:
<samueldr>
I guess openelec-dvb-firmware could be dropped in an installer image too
noonien has quit [Quit: Connection closed for inactivity]
drakonis has quit [Quit: WeeChat 2.4]
sir_guy_carleton has joined #nixos-dev
sir_guy_carleton has quit [Quit: WeeChat 2.4]
rsa has joined #nixos-dev
orivej has quit [Ping timeout: 248 seconds]
<arianvp>
Would it be possible to grant me commit access on Nixpkgs? I'd love to help decrease bus factor
<tilpner>
That sounds very wrong, arianvp
<tilpner>
Decreasing the bus factor would mean removing people from the project, so you probably meant to increase it
<tilpner>
But even then, having commit access doesn't really increase the factor
<niksnut>
we should definitely split firmware-linux-non-free, it's something like half a gigabyte now
<niksnut>
another thing to do is filter out firmwares that don't apply to a kernel
<niksnut>
e.g. there are 8 iwlwifi-8000C-* firmwares, but I think any particular kernel will load only one
<niksnut>
wow, iwlwifi accounts for 95 MB
orivej has joined #nixos-dev
<arianvp>
Wrong word
<arianvp>
XD
<arianvp>
tilpner: It does help with merging more PRs etc. I somehow associate the word bus factor with throughput but that seems the wrong association.
init_6 has joined #nixos-dev
ryantm_ has joined #nixos-dev
ryantm has quit [Read error: Connection reset by peer]
angerman_ has joined #nixos-dev
sdier has quit [Ping timeout: 264 seconds]
sdier has joined #nixos-dev
callahad6 has joined #nixos-dev
v0|d has quit [Ping timeout: 248 seconds]
_rvl_ has quit [Ping timeout: 250 seconds]
dmj`_ has joined #nixos-dev
mingc has quit [Ping timeout: 257 seconds]
mingc has joined #nixos-dev
callahad has quit [Read error: Connection reset by peer]
simpson has quit [Ping timeout: 252 seconds]
dmj` has quit [Ping timeout: 252 seconds]
cransom has quit [Ping timeout: 252 seconds]
angerman has quit [Ping timeout: 252 seconds]
garbas has quit [Ping timeout: 252 seconds]
gchristensen has quit [Ping timeout: 252 seconds]
NinjaTrappeur has quit [Ping timeout: 252 seconds]
angerman_ is now known as angerman
dmj`_ is now known as dmj`
garbas has joined #nixos-dev
gchristensen has joined #nixos-dev
_rvl has joined #nixos-dev
NinjaTrappeur has joined #nixos-dev
cransom has joined #nixos-dev
Jackneill has quit [Remote host closed the connection]
Jackneill has joined #nixos-dev
Jackneill has quit [Read error: Connection reset by peer]
kgz has quit [Ping timeout: 252 seconds]
Jackneill has joined #nixos-dev
Jackneill has quit [Remote host closed the connection]
kgz has joined #nixos-dev
Jackneill has joined #nixos-dev
Jackneill has quit [Ping timeout: 252 seconds]
Jackneill has joined #nixos-dev
Jackneill has quit [Remote host closed the connection]
init_6 has quit []
Jackneill has joined #nixos-dev
Jackneill has quit [Remote host closed the connection]
Jackneill has joined #nixos-dev
Jackneill has quit [Remote host closed the connection]
Jackneill has joined #nixos-dev
<infinisil>
arianvp: Currently #50105 serves as a kind of place to nominate themselves for commit access
<thefloweringash>
I'm interested in getting official nixos images on amazon for their arm a1.* instance types. I think the changes to build the image are straightforward (see #62042), but I don't know how the official images are published. Does someone periodically or automatically run the `create-amis.sh` script in nixpkgs? Is it reasonable to ask them to do so for aarch64?
<{^_^}>
#62228 (by colemickens, 1 day ago, open): Upgrading the latest bootable Azure image to stable requires stepping, results in broken image
<phreedom>
the problem is that i don't even have an azure account and hardly remember anything :'(
<niksnut>
I seem to remember ikwildrpepper did some azure stuff as well
<phreedom>
it shouldn't be too hard to fix, but it's impossible to test without an azure acc
MichaelRaskin has joined #nixos-dev
<gchristensen>
any wireguard people interested in reviewing #62325 ?
<{^_^}>
https://github.com/NixOS/nixpkgs/pull/62325 (by grahamc, 54 minutes ago, open): wireguard: 0.0.20190406 -> 0.0.20190531 and Change peers without tearing down the interface, handle DNS failures better
<clever>
gchristensen: how do you get hydra to connect to the build slaves on port 2200?
<gchristensen>
anyone want a $58 48U server rack?
<clever>
gchristensen: there is a second problem ive seen (which i dont think that PR addresses)
<clever>
gchristensen: the wg service, always starts with `modprobe wg`, even if the module is already loaded
<gchristensen>
oops wrong channel
<clever>
the problem, is that my kernel is old (need a reboot), and i manually loaded the right wg version
<gchristensen>
clever: yeah, my PR doesn't address that -- maybe you could send one? :)
<clever>
but the script still does `modprobe wg`, fails, and then doesnt use the already loaded module
<gchristensen>
does it even need to modprobe, or could it just use existing ckernel module load code
<clever>
i would just use the existing logic
<clever>
boot.kernelModules = [ "wg" ];
<clever>
that should solve it completely
<clever>
although ....
<clever>
that wont load it at nixos-rebuild switch
<gchristensen>
I thought that did load modules, for example virtualbox can turn on without a reboot
<gchristensen>
but note that the code deployed to the servers runs osx's SSH on port 22
ajs124 has quit [Quit: Gateway shutdown]
<clever>
ive got wireguard in the mix
<clever>
oh yeah, wireguard comes up before dns!
<gchristensen>
ah
<gchristensen>
that is what my PR is for! :)
<clever>
so it cant resolve endpoints, and then never comes up!
<clever>
and restart=always breaks too, because its oneshot
<gchristensen>
my PR fixes all that
<disasm>
yeah, I just reviewed :)
<clever>
nice
<disasm>
looks great!
<gchristensen>
thanks, disasm!
<disasm>
guessing that won't backport? :)
<gchristensen>
nooope hehe
<gchristensen>
but you can disable the wireguard module and copy that version (and the wireguard version bump) down
<clever>
gchristensen: oh, a minor tweak i did to your darwin stuff, i run cachecache on the nixos host
<clever>
gchristensen: and then tell darwin to use that as its only cache
<clever>
so now when it installs nix stuff on first boot, instead of going over the internet, its just a copy from the host
<flokli>
gchristensen: urgh, this custom wireguard scripting is getting more and more complex…
<flokli>
giving it a quick test run on my system
<gchristensen>
nice, clever!
<gchristensen>
flokli: the scripts are very short though :)
<clever>
13>May 31 21:09:32 nixoss-iMac com.apple.xpc.launchd[1] (com.apple.platform.ptmd): Service only ran for 0 seconds. Pushing respawn out by 10 seconds.
<gchristensen>
hehehe
<clever>
gchristensen: what about this spamming up my logs? `launchctl stop com.apple.platform.ptmd` doesnt stop it
<gchristensen>
no idea
<gchristensen>
but how do you like that fake syslog service
<gchristensen>
(oops, gotta run. back in a while!)
<clever>
gchristensen: only problem was the IP, you hard-coded a 10.something in apply.sh, but then nix makes it configurable
<flokli>
gchristensen: should we revert ced7cfc02523e1bbc636373d24b7d16110ee587e in release-19.03?
orivej has quit [Ping timeout: 252 seconds]
justan0theruser has joined #nixos-dev
justanotheruser has quit [Ping timeout: 248 seconds]
drakonis has quit [Ping timeout: 244 seconds]
<gchristensen>
flokli: hmm probably
<gchristensen>
flokli: I do have that patch for 19.03, but it feels like A Lot to backport
<gchristensen>
cc sphalerite samueldr
<gchristensen>
though it looks like, also, that wireguard-tools PRs are backported
<samueldr>
I don't know
<samueldr>
the description makes me read it as "fixes an issue", but not sure what oneshot semantics are either
<gchristensen>
the simple -> oneshot part is reverting another PR which was backported very recently
<samueldr>
:/
<gchristensen>
:/ is right
<flokli>
well, from my understanding #61971 (which was backported too) worked around suboptimal behaviour in networkd, and this caused zx2c4 pretty bad headaces. So either we should reverse that #61971 backport, or backport the "proper fix" too
<gchristensen>
at a minimum, that is the right thing to do
<flokli>
yes
<flokli>
and if we feel confident enough, we can backport #62325 at a later time
<{^_^}>
https://github.com/NixOS/nixpkgs/pull/62325 (by grahamc, 4 hours ago, merged): wireguard: 0.0.20190406 -> 0.0.20190531 and Change peers without tearing down the interface, handle DNS failures better