<clever>
leo___: if your deploying to an existing nixos install, your using the none backend, which requires that you specify all of the fileSystems and boot stuff yourself
<clever>
there is no local backend
<clever>
aws, packet, heznet...
<clever>
leo___: depends on the backend, most backends provide it for you
<clever>
rotaerk: you can download the windows install iso from the microsoft site
<clever>
nixos_user: so it should just work now
<clever>
nixos_user: 0001 is also first in your boot order
<clever>
nixos_user2: it added nixos-boot at boot0001
<clever>
leo___: to make you feel worse, my router has 8gig of ECC ram :P
<clever>
leo___: 32gig of ram in my main desktop
<clever>
leo___: mostly software, but i do dabble in electronics and eletrical sometimes
<clever>
and when the insulation melts, the wires can touch eachother, and short out
<clever>
Miyu-saki: your more likely to melt the insulation, before you melt the copper
<clever>
you can inspect it from a remote (windows only, ugh) machine, over usb
<clever>
the motherboard i have is a bit special, in that the os on the mobo doesnt have to support such tools
<clever>
the software just sucks, and doesnt expect such a beefy cpu :P
<clever>
but comparing the voltage, and the TDP of the cpu, it really should be off the charts
<clever>
you cant zoom out enough to see how many amps the cpu pulls
<clever>
Miyu-saki: my motherboard has a util to monitor voltage and amps from another pc, over usb
<clever>
and the amps graph, literally goes off the top of the chart
<clever>
model name : AMD FX(tm)-8350 Eight-Core Processor
<clever>
Miyu-saki: 62.5 amps according to my math
<clever>
Miyu-saki: i too was surprised, when my cpu literally went off the charts in amps, lol
<clever>
Miyu-saki: all of my primary systems have 32gig of ram, i just cant work on less, lol
<clever>
and if somebody tries to write to it, linux just discards the copy in swap, flags the ram as r/w, and retries the operation
<clever>
and the way linux keeps track of things, is that it flags that 1gig of ram as "read only"
<clever>
so if i do need a gig of ram, its already in swap, and linux can just instantly drop those pages
<clever>
this is saying, that i have ~1gig of data in both ram and swap (the same data)
<clever>
SwapCached: 1092588 kB
<clever>
]$ grep Cached /proc/meminfo
<clever>
rotaerk: i think windows will just ignore it (like any other disk) until you start playing with the intel optane utils
<clever>
and the dozen hours thing, is because ntfs isnt designed to use an l2arc
<clever>
i think the cpu rules, are to ensure you can afford the overhead, and then have no reason to complain and give optane a bad name
<clever>
rotaerk: but on windows, they have silly rules like you must use a certain cpu, and it takes a dozen hours to re-configure your ntfs to use it
<clever>
rotaerk: optane is just an nvme block device, as far as linux is concerned
<clever>
Miyu-saki: and there can be 2? outstanding transactions at once
<clever>
Miyu-saki: from what ive heard, multiple writes get batched into a single transaction
<clever>
the L2ARC is basically swap dedicated to zfs
<clever>
rotaerk: efiInstallAsRemovable lets you do an efi install, preferably you should use a second ESP for nixos, then the bios should let you pick which ESP to boot (ideally)
<clever>
rotaerk: you can then either continue with a legacy install, or use boot.loader.grub.efiInstallAsRemovable=true; to do an efi install while booted into legacy
<clever>
rotaerk: try legacy mode then, it sounds like the efi menu is booting the kernel with the wrong flags
<clever>
rotaerk: sounds like some weird edgecases with efi
<clever>
rotaerk: try both of them?
<clever>
rotaerk: if efi is failing, you could try forcing a legacy boot instead, you may need to turn on the CSM
<clever>
rotaerk: can always re-download it
<clever>
rotaerk: try booting from the usb in efi mode
<clever>
rotaerk: is the bios using legacy or efi?
<clever>
red[m]: given that he booted an older nixos image from the same usb just before...
<clever>
red[m]: sounds like you need to bisect the host, ever done a git bisect before?
<clever>
red[m]: so any accelerated guest is failing
<clever>
red[m]: the kvm driver in your host is possibly busted
<clever>
red[m]: try `rmmod kvm_intel` and then run the vm again
<clever>
red[m]: can you pastebin your configuration.nix and say which rev of nixpkgs your on?
<clever>
nh2[m]: ~/.config/nixpkgs/config.nix is the new path
<clever>
nh2[m]: oh, and ~/.nixpkgs/config.nix is the deprecated path
<clever>
red[m]: what is the exact command you used, and the full output from it?
<clever>
nh2[m]: you could say it comes from the "default expressions" that are setup by nix-channel by default, and maybe have a new section explaining defexpr
<clever>
so i can nix-env -iA foo.hello, to grab from a git checkout
<clever>
-r--r--r-- 1 root root 968 Dec 31 1969 /home/clever/.nix-defexpr/channels_root/nixos/default.nix
<clever>
basically, it will recursively search ~/.nix-defexpr/ until it either finds foo/default.nix or foo.nix, then import that, and store it at the foo key
<clever>
nh2[m]: also, its not directly getting that magic set from nix-channel, but rather from ~/.nix-defexpr/ via some more complex rules
<clever>
nh2[m]: in the first block, you say -f is given, when it isnt
<clever>
red[m]: you can always use -I nixpkgs=whatever to override things
<clever>
red[m]: there is a nixpkgs= entry in $NIX_PATH, that points to the root channel on root
<clever>
red[m]: note, that --add wont take effect until you --update
<clever>
red[m]: then nixos-rebuild should be using 19.09
<clever>
which maps to the nixos channel by default
<clever>
red[m]: its always <nixpkgs>
<clever>
CMCDragonkai: if you use my netboot_server.nix as an example, you can pull in just netboot-base, and skip minimal
<clever>
so its better to put such a thing into config.nix, so it persists
<clever>
and in either case, youll forget what you did 6 months down the road, and `nix-env -u mplayer` will undo the override
<clever>
nix-env -iA nixos.mplayer.override --arg pulseSupport true, should also work, because .override is now a function, but it might give trouble
<clever>
nix-env -iA nixos.mplayer --arg config '{ pulseaudio = true; }' is one option (the default.nix defaults the pulseSupport flag to config.pulseaudio)
<clever>
but it silently does nothing, because <nixpkgs> doesnt accept such a flag, and .mplayer isnt a function
<clever>
i often see users doing something somewhat like: nix-env -iA nixos.mplayer --arg pulseSupport true
<clever>
> builtins.functionArgs ({ a, b }: 42)
<clever>
CMCDragonkai: it will use builtins.functionArgs to figure out what each function wants
<clever>
leading to confusion when users try to use --arg on flags that only .override accepts
<clever>
and any unused arguments are silently ignored
<clever>
correct
<clever>
yeah
<clever>
nope, the --eval just makes nix-instantiate prett-print the value, rather then demand a derivation at the end
<clever>
every time -A encounters a function, it just calls it with all of the --arg values
<clever>
CMCDragonkai: it goes to both, check the gist i linked
<clever>
CMCDragonkai: is it still in the let block? it should fail due to it being undefined
<clever>
CMCDragonkai: -A foo.bar will auto-call foo and bar (and the file itself) with any --arg values
<clever>
then nix-env -f '<nixpkgs>' -E 'foo: ...' now replaces the entire set with a single function, which you must still call!
<clever>
and you must now call it yourself
<clever>
but nix-env -i -E 'foo: ...' passes that whole set as foo
<clever>
nh2[m]: and when you nix-env -iA nixos.yarn, it auto-calls the nixos attr, with any supplied --arg
<clever>
nh2[m]: so, nix-env, without -f, is setting the scope to a set of functions, each importing a channel, so { nixpkgs = import <nixpkgs>; nixos = import <nixos>; } basically
<clever>
nh2[m]: basically, -A will auto-call any function it stumbles into
<clever>
jd823592: first, id check if its a gc root, `ls -lh /nix/var/nix/gcroots/auto/ | grep emacs`, what does this turn up?
<clever>
jd823592: so something else you did left one there, and nixos/nix-env will never try to update it
<clever>
jd823592: nixos and nix-env will never put a symlink into that directory
<clever>
jd823592: and the directories themselves are real directories?
<clever>
jd823592: is any element of ` .config/systemd/user` also a symlinl?
<clever>
jd823592: what was the path to the emacs .service file?
<clever>
chessai: yep
<clever>
chessai: yeah, its not really required
<clever>
deni: you can use disabledModules to test the change, and then open a PR to nixpkgs to make it official
<clever>
deni: it would be best to modify the module to make that a proper option
<clever>
red[m]: build-vm uses -kernel on the cmdline, and no kernel actually exists in the image
<clever>
red[m]: you can disable rebooting entirely
<clever>
deni: you would then copy the whole module file, edit the copy, and use imports to bring the copy in
<clever>
chessai: all attributes of mkDerivation become env vars, so you could just directly `CUDA_PATH = cudatoolkit;`, and to make it optional, use `// lib.optionalAttrs ...`
<clever>
chessai: `strOnLinux = str: stdenv.lib.optionalString stdenv.isLinux str` could be reduced to `strOnLinux = stdenv.lib.optionalString stdenv.isLinux`
<clever>
red[m]: if you know how to debug kernelsin gdb, you could probably do that, but it sounds like some pretty big breakage, can others reproduce it?