<elvishjerricco>
Had to mount an unneeded efiSysMountPoint, since the secondary system still wants to install grub. I just used a tiny zvol
<elvishjerricco>
I think chainloading would probably be a tad more robust, but I couldn't find a way to avoid having to type the LUKS password in an extra time.
<clever>
elvishjerricco: if you have a 2nd /boot (which could possibly be a dataset on the pool?), then you can source its grub.cfg from another grub.cfg
<clever>
elvishjerricco: and then put that source command inside a menuentry, which will pre-insmod all the required modules
<clever>
ive done similar before, when i was booting nixos from a gentoo grub
aanderse has joined #nixos-chat
jasongrossman has quit [Ping timeout: 268 seconds]
<elvishjerricco>
It just loads the other /boot's grub config as a submenu
<elvishjerricco>
Then the fact that NixOS nicely hard codes all the drives all the time makes sure that there's never mistakes about what root is
<clever>
you can also use search by uuid to help
<clever>
and its 3am now, i should get to bed
fusion809 has joined #nixos-chat
<fusion809>
I love how Nix makes a package, SageMath, that is a pain to get on many distros (e.g. CentOS, Mageia, openSUSE, PCLinuxOS), so much easier to obtain. Avogadro is also easier to get thanks to Nix's derivative, GNU Guix. Nix also provides me an alternative way to run Firefox, if my distro's Firefox has a bug (which is not as infrequent as you'd expect).
<fusion809>
I suppose you could say I'm saying I'm grateful to Nix's devs for their fine product.
<MichaelRaskin>
fusion809: with Sage it is _definitely_ a thanks directly to timokau for making that monster work as a package
<fusion809>
Thanks, when I see him in an IRC channel I'll say thanks to him specifically.
fusion809 has quit [Ping timeout: 252 seconds]
<srhb>
clever: I finally had a chance to look at your queue runner changes. I was wondering if it might gain more traction if you split up the "misc fixes" pr a bit into contextual pieces? It looks to me like especially the eval notify revamp is very standalone and might be easily reviewed alone (I didn't have a close look at the other changes yet)
<srhb>
I realise upstreaming is painful at time, but... :-)
__monty__ has joined #nixos-chat
<sphalerite>
Hm, when will we have nixos tests for wifi?
<sphalerite>
With a physics simulation and stuff?
<sphalerite>
I guess the hard part will be simulating all that proprietary hardware crap :)
<sphalerite>
hm I don't suppose pure tests would be possible but yeah running some sort of automated tests with hardware would be nice
<infinisil>
Oh my gos
<infinisil>
God
<infinisil>
My hotel from NixCon has sent me not 1, not 2
<infinisil>
But 6 mails asking me to rate them!
<infinisil>
6!!
<gchristensen>
sphalerite: this is a dream of mine, to use Intel's ME to drive that stuff
<sphalerite>
hm while it would be cool I'd prefer to use a non-evil technology for it myself :')
<joepie91>
infinisil: send them 6 different responses, each one complaining about all the requests coming before it
<gchristensen>
lol, sure
<MichaelRaskin>
Why would you need a true ME and not just wake-on-lan + PXE?
<gchristensen>
because I wanted to test the actual produced .iso straight away from boot, as if a user put it on a USB stick
<sphalerite>
CD changer!!
<sphalerite>
:p
<gchristensen>
ME gives you a serial console to the boot process
<sphalerite>
second computer!
<MichaelRaskin>
Well, IPMI would be enough, I guess
<gchristensen>
ipmi would be, but that implies not a regular ol' computer someone can buy
<gchristensen>
can/would
<MichaelRaskin>
Or it could be as-a-service with one computer with a ton of RS232 converters connected to test machines' serial console
<MichaelRaskin>
ME doing _anything_ right is already not a computer anyone can obtain?
<gchristensen>
regular computers don't have serial ports :| you're saying "doing it right" but I'm just saying "just has ME" :)
<MichaelRaskin>
Well, you want ME to give you data
<MichaelRaskin>
Preferably, in some relation to reality
<gchristensen>
yeah, it has to trouble with that
<gchristensen>
afaik most complaints about ME are around it not being picky enough about who it gives data to
<sphalerite>
gchristensen: I'm about to submit a PR to bump hound. Upstream has finally fixed the open_search.xml bug, so you can remove that workaround when the PR's merged and stuff :)
<sphalerite>
gchristensen++
<{^_^}>
gchristensen's karma got increased to 45
<sphalerite>
for packaging, moduling and testing hound ♥
<sphalerite>
I want to set up hound for my own stuff, feels like it would be very helpful
<gchristensen>
I recommend it!
<gchristensen>
though if a good / easy-to-run language server from Nix appears (maybe jD...'s would do) sourcegraph might bec ool
<sphalerite>
gchristensen: care to merge the PR? I could do it myself but feel weird merging my own PR ;)
<gchristensen>
sure
<__monty__>
sphalerite: You have so much own stuff that grep/rg doesn't suffice?
<sphalerite>
__monty__: mostly just that it's multiple projects that I want to be able to search in the context of one
<sphalerite>
i.e. upstream repo, nix exprs repo, nixpkgs, …
<gchristensen>
and why read it all of disk when you can start with an index of trigrams
<gchristensen>
samueldr: my nixops network is checked out to ~/projects/personal your nixops network is checked out to ~/projects/network which leads to great confusion over here from time to time
Synthetica has joined #nixos-chat
<sphalerite>
gchristensen: thanks!
<samueldr>
(guessing this wasn't meant for me, as I'm not using nixops)
<gchristensen>
omg the other samuel :)
<gchristensen>
disasm:
<andi->
Waiting for $dayjob CI, thought I could do some work on the upcoming firefox release, now I am waiting on both the CI and mercurial /o\ cloning nixpkgs is very fast in comparison...
<gchristensen>
oh dear
<andi->
This time around they are using the assembler (as) as it would be a regular compiler (passing -DFOO=BAR) which breaks :/ So now the fun time figuring out why and when that started happening.. Could probably start a 3rd task while I wait..
<manveru>
the other thing was performance for kubernetes configs, but i just build those with nix now and i'm happy
<manveru>
not sure there's much point of using dhall if you're not using haskell anyway :|
<__monty__>
I don't see any possible reason not to use haskell though : >
<infinisil>
__monty__++
<{^_^}>
__monty__'s karma got increased to 2
<__monty__>
Woah, where'd that karma come from?
<andi->
manveru: do you have an example / pointer to that k8s with nix? I've been telling that to my co-worker that is working on such things but he prefers errorproune sed scripts..
<__monty__>
manveru: I'd like to use dhall with dhall-to-nix as a substitute for "typed nix." That could be useful without haskell.
<infinisil>
(And the default.nix file in the same directory)
<manveru>
simpson: yeah, it's basically a nixos-module-like system for type checking kubernetes configs and producing a json file from it that you can just kubectl apply
<manveru>
ideally you'd also make your docker images dependencies in there, so you can handle the whole deploy that way
<infinisil>
__monty__: I do have to have duplicate definitions of the config format in Haskell and Nix unfortunately
<simpson>
manveru: How would you compare with Helm?
<manveru>
never used helm
<andi->
I tried using it but it felt like the ansible/saltstack/puppet templating pain all over again
<sphalerite>
__monty__: your previous karma was from goibhniu for 2018-07-23 15:37:49 __monty__ All of arch, debian and nixos suck in many ways, they're also great in many ways.