<bew>
okay, let's not continue on that other name^^
ahmedelgabri has quit [Ping timeout: 258 seconds]
steve[m]2 has joined #nixos
<radvendii>
can someone give my appimagekit pull request a look? (#116501) I need it for something downstream (going to send a PR to nix-bundle) and I'd rather not have to use my fork of nixpkgs as the flake input
<randomuser123>
Hi. I see a symlink ~/.nix-defexpr/channels/nixpkgs-20.03 that doesn't match anything listed by `nix-channel --list`. I think I added that as a channel once then removed it, but it's still there. Any idea how I can get rid of it for good?
unacceptable has quit [Quit: ZNC 1.7.1 - https://znc.in]
novenary_ has joined #nixos
<gchristensen>
bew++
<{^_^}>
bew's karma got increased to 2
novenary has quit [Read error: Connection reset by peer]
novenary_ is now known as novenary
srk has quit [Ping timeout: 240 seconds]
srk has joined #nixos
gustavderdrache has quit [Quit: Leaving.]
<randomuser123>
Aha, found my own solution. I could look at ~/.nix-defexpr/channels/manifest.nix to find the real name of the channel, then `nix-channel --remove` it with that name. Still not sure why --list didn't list it.
<bew>
gchristensen fuiltilt \o/ :D
apache8080 has quit [Ping timeout: 240 seconds]
Guest43762 is now known as palo
palo is now known as Guest4564
Guest4564 has quit [Quit: WeeChat 2.9]
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hackage-packages.nix: automatic Haskell package set update »: https://git.io/JYtNt
riksteri has quit [Read error: Connection reset by peer]
riksteri has joined #nixos
Unode has quit [Remote host closed the connection]
Unode has joined #nixos
rj has joined #nixos
turq_ is now known as turq
bitmapper has quit [Quit: Connection closed for inactivity]
turq has quit [Changing host]
turq has joined #nixos
<bew>
srhb re about the default nixpkgs flake being on master, there is https://github.com/NixOS/flake-registry/issues/6 which shows that we can write `nixpkgs/nixos-unstable` to target a specific branch of the flake `nixpkgs` which doesn't specify a branch (and default to master then)
<bew>
So there is value to have the default `nixpkgs` with unspecified branch (defaulting to master), as it allows to easily select a branch.. Now there'll need to be good docs for this, and not suggest everyone to use `nixpkgs#blabla` to test `blabla` as you'd probably need to rebuild lots of packages, and explain they should use a branch or sth
<Guest82>
Simpson I am that guy with Firefox issue, tried changing kernel to see if it fixes it. Same GPU driver, xorg. Something is wrong because even in 5.4 when its working besides Firefox issue there is small artifacts when opening and closing terminal, like window doesn't open or close instantly I can see triangles broken shapes while its trying to
<Guest82>
draw, GPU nouveau driver is loaded I am not sure. I can deal with Firefox issue but something seems wrong not performing
<simpson>
Our definition of "same" is probably not the same. For most GPU drivers, including all the ones you'd use for this particular card, the driver includes a kernel module. So the precise version of the kernel that you're using is important, including distro-specific patches.
<Guest43>
simpson kernel 5.10.25, nouveau version how do I check
<Guest43>
Oh its gfx related that's why alacritty doesn't work but st does Simpson
dbmikus has joined #nixos
Xe is now known as Cadey
<simpson>
Guest43: For most GPU drivers, including all the ones you'd use for this particular card, the driver includes a kernel module. So the precise version of the kernel that you're using is important, including distro-specific patches.
ahmedelgabri has joined #nixos
<simpson>
You can check your version of Mesa as well (nouveau's userspace driver ships with Mesa), but it's probably not as important unless you're using Glamor with Xorg.
<simpson>
I don't know exactly what's wrong. I only know how to compare working and non-working setups.
<Guest43>
Maybe glamor is enabled that's the issue? I don't know what that is
dbmikus has quit [Ping timeout: 240 seconds]
<simpson>
Glamor is when Xorg uses GL to acclerate stuff like terminal emulators or Web browsers or other 2D drawing. It's experimental and can often break. AIUI it's disabled by default. You can check in your Xorg log; if you don't see any trace of Glamor then it's off by default.
ahmedelgabri has quit [Ping timeout: 240 seconds]
waleee-cl has quit [Quit: Connection closed for inactivity]
<Guest43>
simpson what can I do at this point, graphics driver seems to be the issue with 5.4 and 5.10, should I just move on to other distro which has good patches where it works like arch and void? Gentoo was similar to this. Subpar performance on 5.4 and freezing on 5.10
<simpson>
Guest43: Depends on your gumption and expendable cash; personally I'd buy a card from another vendor~ But since that's not an option, I'd put more effort into detailing *exactly* which patches are relevant between Arch, Void, Gentoo, and NixOS. I'd take dmesg logs and Xorg logs and diff them.
<Guest43>
simpson if I upload void and nixos dmesg logs can you check them?
<Guest43>
I can't buy new hardware its my old laptop
<Guest43>
I will make sure I don't see the sight of nvidia again
<simpson>
Guest43: I don't know anything about nouveau, so I don't know what to look for. I'd literally use `diff` or a similar tool, possibly after using `sed` or similar to remove the timestamps.
<simpson>
There's a few reasons. My favorite is security: Just like with the CPU, there's privileges and separations regarding who can access which part of the GPU's memory. Classic drivers from the 90s just wrote directly to the GPU, just like how they used the CPU directly without sharing.
stree has quit [Ping timeout: 246 seconds]
bew has quit [Quit: Connection closed]
<simpson>
So part of what makes a modern GPU driver slow is that it has to negotiate with a kernel memory manager in order to figure out who's allowed to see which graphics objects. So that a malicious Web browser tab can't read the contents of a neighboring tab or another app.
<Guest32>
simpson I installed firefox-unwrapped with nix-env and the scrolling freezing issue is not there?
h0m1 has joined #nixos
<Guest32>
ohhhh
<Guest32>
simpson I found it I think, when installed as firefox-unwrapped, it's not set as default browser. When set as default browser the issue will come back
<Guest32>
Right now I can change the default option in preferences, but when it was installed as firefox I couldnt
<Guest32>
simpson some permission issue I think? Just changed to default browser, no issues. (before with firefox package or system package) it was greyed out
simba1 has joined #nixos
Guest32 has quit [Client Quit]
Guest32 has joined #nixos
Guest32 has quit [Client Quit]
Guest43 has quit [Ping timeout: 265 seconds]
monotux has quit [Remote host closed the connection]
monotux has joined #nixos
Guest32 has joined #nixos
<simpson>
Guest32: I have literally no idea why that would matter, but I'm happy that you found a solution which works for you.
sangoma has joined #nixos
<Guest32>
simpson but installing with home manager, issue exists :D
oxalica1 has joined #nixos
oxalica has quit [Ping timeout: 250 seconds]
oxalica1 is now known as oxalica
<Guest32>
simpson I think these kind of strange issues exist because of how nixos file structure is, do you think?
<Guest19>
simpson sorry to bug you, but I noticed when launching firefox from nixos it says glxtest libpci missing, but not on voidlinux. couldnt find anything on google
Guest43 has quit [Ping timeout: 240 seconds]
simba1 has joined #nixos
<simpson>
Guest19: How could it possibly be caused by the Nix store's file structure? That doesn't make sense; you reproduced the issue on Gentoo.
<Guest19>
simpson reproduces kernel 5.10 issue, not firefox scrolling one. do you think that glxtest libpci error relevant?
<Guest19>
it says something along the lines Graphical annotation glxtest libpci missing
<simpson>
No idea, sorry.
anser78[m] has joined #nixos
<Guest19>
no probs, thanks!
<Guest19>
damnit i am loosing my mind over this lol
<zhorick>
Hello! I have laptop with nvidia and intel (optimus). I installed these drivers with ` services.xserver.videoDrivers = [ "modesetting" "nvidia" ];`. I rebuilded nixos and I rebooted my laptop. After reboot I checked kmod: `lsmod | grep nvidia` and nope... Then I checked /etc/modprobe.d/nixos.conf and saw this: ```blacklist nouveau
<zhorick>
blacklist nouveau
<zhorick>
blacklist nvidiafb
<zhorick>
blacklist nvidia-drm
<zhorick>
blacklist nvidia
<zhorick>
blacklist nouveau
<zhorick>
options firmware_class path=/nix/store/8v3nbx0j6cpmfd7gjgz3fkrbl8jila72-firmware/lib/firmware```. This file is readonly. How I can remove from blacklist nvidia kmods?
damian1 is now known as dfgg
jmeredith has quit [Quit: Connection closed for inactivity]
<aforemny>
zhorick: That file is generated from your configuration. You will have to find out from which configuration option that line comes from and change that configuration option.
<aforemny>
zhorick: I am looking into nixpkgs right now to figure this out. It seems setting services.xserver.videoDrivers is correct. Did you set any other options that might be related to this? bumblebee? optimus?
<zhorick>
My configuration.nix and hardware-configuration.nix haven't any blacklist options
<aforemny>
zhorick: Did you set hardware.bumblebee.enabled?
bb_ has joined #nixos
<zhorick>
I use prime offload
<aforemny>
zhorick: Or did you set hardware.vidiaOptimus.disable?
zakame has quit [Ping timeout: 252 seconds]
<zhorick>
And I tried bumblebee
<zhorick>
No
<aforemny>
That line most likely comes from setting `hardware.bumblebee.enable = true` or setting `hardware.nvidiaOptimus.disable = true`.
ddellacosta has joined #nixos
<aforemny>
zhorick: You can check with `nixos-option hardware.bumblebee.enable`. What's the output?
<zhorick>
How I can clean array in Nix? In nix repl I see boot.blacklistedKernelModules have these blacklisted kmods
zakame has joined #nixos
<aforemny>
zhorick: You will have to figure out where it comes from. You could force that option to be overriden. But I don't think that is the right approach.
<zhorick>
You can check with `nixos-option hardware.bumblebee.enable`. What's the output? - `true`
<aforemny>
zhorick: So you have that enabled. You should disable it I guess.
evils has joined #nixos
<zhorick>
Add hardware.bumblebee.enable = false to my config?
stree has joined #nixos
<aforemny>
zhorick: Yes. Or remote that option as it defaults to false. Your choice
<zhorick>
:D In 115 line I writed hardware.bumblebee.enable = true! I removed this, rebuilded and I going to reboot...
zhorick has quit [Quit: Connection closed]
zhorick has joined #nixos
<zhorick>
Hmm... lsmod | grep nvidia give me results but nvidia-settings exit with `ERROR: Unable to load info from any available system`
cyris212 has quit [Ping timeout: 252 seconds]
cyris212 has joined #nixos
<aforemny>
zhorick: What's the output of `nvidia-settings --verbose=errors`?
<zhorick>
ERROR: Unable to load info from any available system
<aforemny>
zhorick: Also with the `--verbose=errors` flag? What about `--verbose=all`?
<zhorick>
Yeah. With --verbose=errors... with verbose all `WARNING: NV-CONTROL extension not found on this Display\nERROR: Unable to load info from any available system`
ahmedelgabri has joined #nixos
Guest32 has joined #nixos
<aforemny>
zhorick: I am getting that warning as well, but for me nvidia-settings does start up. So I guess it is not the cause of your issue.
Guest55480 has joined #nixos
<aforemny>
zhorick: Is your user in the video group? Try running `groups` to find out
<Guest32>
where I can find the wrapper for firefox? do I just do which firefox then inspect that file?
<zhorick>
But I also can't connect to external screen via hdmi
evanjs has quit [Read error: Connection reset by peer]
<zhorick>
groups: users wheel networkmanager
<Guest32>
do you need to be in video if you are in wheel?
<bb_>
do you folks have a battery icon in xfce panel?
<aforemny>
zhorick: Add yourself to video group. That would be `users.users.zhorick.extraGroups = [ "video" ]` in your configuration
<zhorick>
ok
<zhorick>
I need to nixos-rebuild switch or boot?
<aforemny>
zhorick: You will have to log in and log out again for that to take effect (or reboot). Run `groups` again to verify. And then nvidia-settings might just work.
<aforemny>
*log out and log in
evanjs has joined #nixos
zhorick has quit [Quit: Connection closed]
<aforemny>
zhorick: If you nixos-rebuild switch, you can log out and log in again. If you nixos-rebuild boot, you will have to reboot for it to take effect.
<aforemny>
Guest32: Yes, which firefox is the wrapper file
zhorick has joined #nixos
<zhorick>
No.. I rebooted and nvidia-settings doesn't working
<aforemny>
zhorick: Are you in video group now? Does `groups` confirm that you are?
<zhorick>
Yeah
<Guest32>
aforemny thanks
<aforemny>
zhorick: Hum.. I don't know then. Does running `sudo nvidia-settings` make a difference? I guess not, right?
<zhorick>
With sudo ERROR: Unable to load info from any available system
Guest32 has quit [Quit: Connection closed]
<aforemny>
zhorick: I am just grasping for straws here, but what's the output of nvidia-smi?
<matthewcroughan>
supersandro2000: having trouble with flake URIs, can you specify a branch?
<matthewcroughan>
`nix run github:matthewcroughan/nixpkgs#ly`
<mikeplus64>
is there a quick way to make a derivation that's just a bunch of specific files? like `symlinkJoinFiles [ ./file1 ./file2 ./foo/bar/file3 ]` whose output would be a dir with ./file1 ./file2 ./foo/bar/file3 (so would preserve the dir structure of input files)
<matthewcroughan>
how would I tell it to use a different branch?
<matthewcroughan>
would one just add my nixpkgs repo with the changes to their repo? And refer to it as inputs.matthewcroughan.services.nodered.enable = true; ?
Thor007 has quit [Remote host closed the connection]
<JaakkoLuttinen[m>
If in a attrset in a derivation, someone has set `john.doe = {foo = "bar";};`, can I somehow force remove this setting as if `foo = "bar"` wasn't set at all? I can override it with `mkForce` but can I force remove as if it wasn't set at all?
<sterni>
JaakkoLuttinen[m: you can either do it manually which is kinda tedious with that level of nesting or with something like lib.recursiveUpdate although I'm not sure off the top of my head if it supports removing attributes
<JaakkoLuttinen[m>
I just would like to remove a few attributes from `systemd.services.nginx.serviceConfig`..
domogled1 has joined #nixos
ddellacosta has quit [Ping timeout: 260 seconds]
domogled has quit [Ping timeout: 250 seconds]
domogled1 is now known as domogled
smctwo has quit [Remote host closed the connection]
domogled has quit [Ping timeout: 260 seconds]
kalipso has joined #nixos
Lord_of_Life has joined #nixos
Dr8128 has joined #nixos
vidbina_ has quit [Ping timeout: 260 seconds]
Lord_of_Life has quit [Remote host closed the connection]
<exarkun>
Or instead of that... How do you make composable Python module overlays?
<exarkun>
Or instead of that... How do you compose separate Python module derivations that each have their own requirements for out-of-nixpkgs dependencies?
<JaakkoLuttinen[m>
sterni: ah, yes, true, sorry
<exarkun>
Or instead of that... Why did this - https://github.com/tahoe-lafs/tahoe-lafs/tree/master/nix - start to fail with - "nix/tahoe-lafs.nix:1:1 called without required argument 'eliot', at /nix/store/mq7wlsvfcycbp7qd5y330cb4284k5ra3-nixpkgs/lib/customisation.nix:69:12" - when I added some Python package overlays to another project that's trying to use it
apeyroux has quit [Ping timeout: 265 seconds]
<patagonicus>
exarkun: What's the overlay you added? I don't really know much about tahoe-lafs or Python packages, but maybe I can at least reproduce it.
ddellacosta has joined #nixos
<sterni>
JaakkoLuttinen[m: you could do lib.mkForce null for that option if it accept null as a value, but there is a surely a better way of doing this
<JaakkoLuttinen[m>
sterni: turned out that forcing empty string did the trick
<sterni>
nice!
<patagonicus>
So, you can lib.mkForce it to its default value. If the option doesn't have a default value, then either it's not being used or it will generate an error when it's not set.
<patagonicus>
Ah, sorry, this is an attrSet. Those are different. :(
jjakob has quit [Quit: Either I'm rebooting or Quassel died again.]
Guest46293 has quit [Changing host]
Guest46293 has joined #nixos
Guest46293 is now known as Reventlov
<sterni>
can't you upgrade to 20.09? that one gets security patches at least
<exarkun>
In the past I haven't because so much of the Python 2 ecosystem is broken after 19.09 and the main project (tahoe-lafs) is still Python 2-only
<sterni>
well
<sterni>
not good in two ways then :p
xdefrag has quit [Remote host closed the connection]
<{^_^}>
[nixpkgs] @peti pushed 731 commits to haskell-updates: https://git.io/JYm9l
<exarkun>
no one has said "no you can't compose overlays" yet so I wonder what I did wrong that makes my overlays interfere with each other
<exarkun>
or maybe it has nothing to do with my overlay, somehow (but I dunno what else would be the cause)
<{^_^}>
[nixpkgs] @peti pushed to haskell-updates « hlint: fix override for ghc-lib-parser »: https://git.io/JYm9P
<veleiro>
is it antipattern to copy a .nix file to the nix store?
vidbina_ has joined #nixos
konubinix has quit [Ping timeout: 260 seconds]
<supersandro2000>
veleiro: can't tell you because I have a copy of nixpkgs in my store
<supersandro2000>
so depends what you want to do. If you can explain a few more details I can give you a better answer than depends
<veleiro>
oh whait, dont we all
<supersandro2000>
exarkun: you can use 19.09 for that one project and 20.09 for all the other stuff
<sterni>
exarkun: have you switched out all python-super.callPackage in the python overlay for python-self.callPackage?
<sterni>
exarkun: python-super still uses the python package set defined in the first overlay so it doesn't have the dependency you add in the second overlay
mayhew has joined #nixos
codygman__ has joined #nixos
<supersandro2000>
sterni: oh... well, always on the latest stuff
<supersandro2000>
*I am
<sterni>
exarkun is a poor soul for having to deal with 19.09 and python 2 :|
<exarkun>
Oh interesting. sterni, I just tried that and then I tried building tahoe-lafs by itself w/o my extra overlay and it failed to build just how it was failing in my overall project ...
codygman__ has quit [Remote host closed the connection]
codygman__ has joined #nixos
<exarkun>
(That is: nix/tahoe-lafs.nix:1:1 called without required argument 'eliot', at /nix/store/mq7wlsvfcycbp7qd5y330cb4284k5ra3-nixpkgs/lib/customisation.nix:69:12)
<exarkun>
One thing I'm not so clear on is whether the Python module overlay system follows the same rules as the general overlay system
<cinimod>
But 20.09 does not give an error
<exarkun>
To substitute Python packages I have to use python27.override ... if I do that twice, they're not going to cooperate, are they?
<cinimod>
How which nix derivation should I look at to see what has changed?
<gchristensen>
The channels nixos-20.09-small and nixos-unstable-small both have the openssl upgrade in place. If you're looking to patch, use these channels for now.
astro has joined #nixos
<exarkun>
gchristensen: me? I'm stuck on 19.09.
<gchristensen>
that was a general announcement
<exarkun>
ah :)
<gchristensen>
Ihope you can upgrade soon though :)
<astro>
hi, I'm on aarch64 and nixos seems to want to build firefox rather than download it, which I can't do
<astro>
this wasn't a problem up until 85.0.2, which I'm still using, what's up?
<supersandro2000>
astro: do you use firefox-bin?
haritz has quit [Ping timeout: 240 seconds]
<supersandro2000>
it could also be that aarch is not cached yet
<astro>
no, firefox. firefox-bin is not compatible with aarch64
<sterni>
exarkun: pretty sure it's a confusion between self and super
codygman` has joined #nixos
<astro>
is there a way to find the status of a derivation, whether it is cached or not, what versions are cached?
<exarkun>
I just found some code elsewhere that I wrote long ago for safely overriding a python derivation w/ a "new" packageOverrides ... it goes to extra length to merge w/ the existing overrides instead of clobbering them
<exarkun>
sterni: https://gist.github.com/exarkun/134e07d5f30c09a3c544e1685f7aa555 fixes at least part of the issue. the python2 pyopenssl seems to be properly overridden now and using the correct openssl, and the tahoe-lafs overlay is still in effect (ie, I can still build tahoe-lafs). but the python 3 pyopenssl seems to be unaffected for some reason I don't understand yet.
<o1lo01ol1o>
etu: that looks right. how do I revert ffrom nixos-rebuild test?
<exarkun>
although I have a "unit" test that shows it is working :/ it just doesn't work in practice.
<etu>
o1lo01ol1o: One way is to just reboot since "test" doesn't add a boot generation
<exarkun>
oh maybe my unit test runner is broken :(
nicoo has quit [Remote host closed the connection]
<etu>
o1lo01ol1o: Another way is to rebuild again to something else
Diogo93 has joined #nixos
<o1lo01ol1o>
exarkun: I'm just seeing you're talking about python 2 and pyopenssl, is it breaking things all over the place for you on 20.x?
<exarkun>
o1lo01ol1o: I had 19.0.0 before and it broken when I upgraded to openssl 1.1.1k. I upgraded pyopenssl to 20.0.1 and it *seems* to behave better - but I haven't actually threaded the fix all the way through my system yet so I'm not entirely sure of that.
zakame has quit [Ping timeout: 240 seconds]
<exarkun>
(and, yes, python 2)
<cinimod>
Well nothing in the R derivation changed so I am mystified why I get `./devX11.h:64:12: fatal error: 'cairo-xlib.h' file not found`
<o1lo01ol1o>
exarkun: yes, although it effectively doesnt' matter which one breaks because it seems to be non-deterministic
<o1lo01ol1o>
though I don't know how
zebrag has joined #nixos
<yurb>
what (local) database does `nix search` use? I keep having results from both `nixos` and `nixos-20.03` show up there after upgrading to 20.09. I'd like to get rid of the `20.03` results if possible.
<exarkun>
What kind of breakage do you see? I see test suite failures from pyopenssl 19.0.0 against openssl 1.1.1k. I haven't seem any test failures from pyopenssl 20.0.1 against openssl 1.1.1k yet.
<o1lo01ol1o>
exarkun: see here:https://github.com/obsidiansystems/obelisk/issues/836
<{^_^}>
reflex-frp/reflex-platform#721 (by milvi, 4 weeks ago, open): Obsolete and unbuildable Python and pyOpenSSL dependencies resulting in Install failure
<o1lo01ol1o>
but that fails on nixos-20.09 and the former just (randomly) succeeded on nixos-20.09 for me
<o1lo01ol1o>
It had been failing on nixos-20.09 for some time but I just stripped all other entries from the channels and rebuilt and rebooted
<o1lo01ol1o>
now it works.
<exarkun>
E Error: [('SSL routines', 'tls_process_server_certificate', 'certificate verify failed')]
<exarkun>
These are the errors I had on pyopenssl 19.0.0
<exarkun>
I didn't dig all the way in to why but there was a significant change to how openssl 1.1.1k verifies certificates so I assumed this is the cause
<exarkun>
And openssl 1.1.1k just landed in nixos 20.09
<o1lo01ol1o>
"just" as in the past couple hours?
<exarkun>
Definitely within the last day. I'm not exactly sure when.
<o1lo01ol1o>
Well, I suppose that could explain it. But why would it work on 20.03?
<exarkun>
It's weird that it it's not deterministic. I can't really explain that part.
<exarkun>
non-determinism isn't _obviously_ consistent with my hypothesis
<o1lo01ol1o>
It works on the above 20.03 hash but not on 20.09 until the past hour. and it failed against 19.09 on the nixos docker image
<exarkun>
1.1.1k landed in master "Thu Mar 25 10:16:50 2021 -0400"
<o1lo01ol1o>
and the thing that's pulling in pyopessl is cabal2nix
rubm has joined #nixos
Qwerky has quit [Remote host closed the connection]
<cinimod>
homebrew has `--without-x` so I am trying that
<cinimod>
But I am mystified as to why R built for 20.09 but it won't build now
alexherbo2 has joined #nixos
<cinimod>
Well that built - I can raise a ticket and a PR but I don't know why it used to work
Qwerky has joined #nixos
<nf>
i have an option with type either str (submodule o) and an "apply" function that converts string values to values of the submodule. but apply is applied *after* default values are added, so i have to add them myself in the conversion function. is there a more elegant way to do this?
<matthewcroughan>
I don't really understand why that nix-shell is ineffective
jonringer has quit [Read error: Connection reset by peer]
jonringer has joined #nixos
<JaakkoLuttinen[m>
I'm trying to create a few directories and set their permissions in a systemd service's preStart script. But the service start just timeouts.. Is there some limitation that I cannot create and modify arbitrary directories in preStart even if the user has permissions to do so?
ddellaco_ has joined #nixos
<supersandro2000>
matthewcroughan: overwriting the version is not enough to get another version
<supersandro2000>
you need to modify src
<tpw_rules>
matthewcroughan: you need to specify what "ineffective" means for me to be sure but probably because you didn't override the source. you're just creating the same package with a different version. you need to update the source to get the new version's source
<supersandro2000>
and the hash for it
codygman` has quit [Read error: Connection reset by peer]
<yaymukund>
I want to basically say, "if the derivation is SHA 123456, then stuck with the version I installed manually. But as soon as the upstream SHA changes to 789abcd (or anything else), just install it and pretend none of this ever happened.
<andi->
with upstream you mean nixpkgs?
<yaymukund>
yes, exactly
<{^_^}>
[nixpkgs] @SuperSandro2000 merged pull request #116531 → Remove packages which are no longer required after #116317 in packages that use grpc → https://github.com/NixOS/nixpkgs/pull/116531
<yaymukund>
nixos-unstable specifically, but i would expect it to be channel-agnostic
<yaymukund>
(also open to hearing this is a bad idea :))
Mateon1 has joined #nixos
heyarne[m] has joined #nixos
<veleiro>
if you're making a derivation that's mostly a package, but it would need some services active, is that considered a module in nixpkgs? or is programs.* the midway for something like that?
<veleiro>
i'm not sure i understand programs vs pkgs
ddellaco_ has joined #nixos
codygman` has quit [Remote host closed the connection]
<evils>
veleiro: this may be an oversimplification, but, anything that brings nixos configuration is a module
<sterni>
programs provides a way for doing system wide configuration for some package
codygman__ has joined #nixos
<andi->
yaymukund: create an overlay where you do (self: super: { package = if super.package.src.sha256 = "..." then < your custom code here > else super.package; })
<sterni>
this is often because it is necessary to have some configuration related prerequisites
<sterni>
or to enable some system wide stuff useful (like installing fish shell completions for packages when you enable programs.fish)
<sterni>
so if your package needs a service running to work properly, making a programs module additionally to your package sounds helpful
<sterni>
if it provides the service, use services.*
<heyarne[m]>
hi! i'm new to nixos. i'm trying to use a processing wrapper from another programming language (clojure). trying to do so tells me downloads some jars from maven, for which execution fails because they don't find the required shared libraries. i know that all the libraries are in the `processing` package, which i managed to install. how can i add those libraries to `LD_LIBRARY_PATH`? i know this is something nix tries to
<heyarne[m]>
avoid, so maybe there's some cleaner way even?
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mla has joined #nixos
rose_ has joined #nixos
<rose_>
yo
<abathur>
yo
<rose_>
can anyone point me to why cross compiling needs the builder to emulate the host?
cole-h has joined #nixos
mschwaig1 has quit [Quit: WeeChat 3.0]
tgunnoe has quit [Ping timeout: 240 seconds]
<rose_>
long story short there's a 32 core arm monster in the house and i want it to build packages for my poor little 4 core x86 laptop
<yaymukund>
heyarne[m]: I just did this for Rust, and my solution was to make a shell.nix for the project that set LD_LIBRARY_PATH
<exarkun>
afaik cross compiling does not require that
<exarkun>
you're talking about a different kind of compiling
<rose_>
oh yeah sorry i think there is a distinction i'm not fully aware of
<exarkun>
what cross compiling requires is a ton of fiddly build setup/config to be done exactly right for a zillion packages
<exarkun>
last time I tried, nixpkgs had managed to do this for about a zillion minus thirteen packages
<exarkun>
so it doesn't reliably work
<exarkun>
it's been a while though
<exarkun>
the advantage of using an emulator is that you don't have to do any of that fiddly setup/config
<exarkun>
it's just like a native build because it is a native build
<rose_>
(for context when i tried "doing the thing" i was setting the arm machine as a remote builder and it was complaining that it "needed an 'x86_64-linux' to build packages for 'x86_64-linux'")
<exarkun>
the disadvantage is the obvious thing
<exarkun>
sounds like you missed doing some of the fiddly cross-compilation setup/config :)
<rose_>
hehehe
<exarkun>
I haven't tried it lately though, my knowledge is all pretty out-of-date, so I'm gonna stop talking
<rose_>
question for the room then: should it be possible for me to use the arm as a remote builder for my laptop without emulation, even if i have to fiddle some configs?
johnnyfive has quit [Quit: I quit now.]
johnnyfive has joined #nixos
<ajs124>
rose_: if you don't do any IFD from a derivation which has the system set to arm, yes
<rose_>
ajs124: when you say "you"... i didn't even know what ifd is until now :D is it used in nixpkgs/can i run into this problem by building derivations only from nixpkgs?
<rose_>
damn this stuff may be too advanced for me
alexherbo2 has quit [Quit: Ping timeout (120 seconds)]
phalrax has joined #nixos
alexherbo2 has joined #nixos
alexherbo2 has quit [Client Quit]
<phalrax>
hi, how can I see which packages are installed in a particular system configuration? when browsing /nix/var/nix/profiles I can't seem to find any trace of gnome for example
rajivr has quit [Quit: Connection closed for inactivity]
<rose_>
phalrax: where in profiles are you looking? i can find kwin_wayland in /nix/var/nix/profiles/system/sw/bin/
<phalrax>
oh yeah, I overlooked the sw folder
<rose_>
(although if you want to find a list of packages there is probably a much cleaner way involving nix-env, i've got the man page open give me a sec)
<phalrax>
I checked bin directly, not sw/bin
acarrico has quit [Ping timeout: 260 seconds]
<rose_>
yeah i don't know what bin is for, it only has one entry on my machine and that's switch-to-configuration
codygman` has joined #nixos
<phalrax>
one other thing, I just switched my config from gnome to plasma5 and was surprised to see that nix-collect-garbage cleaned up a whole bunch of packages, shouldn't they still be there, because my older configurations reference them?
sangoma has quit [Ping timeout: 240 seconds]
<rose_>
depends what packages, but i think nix-collect-garbage does not delete older generations (and therefore the packages they contain) unless you specify some options (-d)
codygman__ has quit [Ping timeout: 246 seconds]
<phalrax>
a whole bunch, but mostly some gnome stuff, I didn't use the -d option
<rose_>
i would say it's old versions of gnome stuff but that should still be in old generations as well, unless you have ever deleted a generation and this is the first time you've gc'd since then
<rose_>
also, nix-env -p lists packages and their versions in a profile
<Quarub>
phalrax: maybe these were old versions from previous iterations? When you do a rollback, are these downloaded again?
<phalrax>
I gc regularly and did so earlier today to and there were no upgrades since 20.09 build queue seems broken atm :'D
<phalrax>
I gc regularly and did so earlier today too and there were no upgrades since 20.09 build queue seems broken atm :'D
<rose_>
Quarub: surely they are kept around even if you are no longer using the generation?
<Quarub>
rose: Yeah should be. But rolling back could be done as a sanity check I assumed ^^
<rose_>
1. i thought that was the whole point in remembering to gc when you need to 2. i wouldn't want to pick an old generation in the boot menu and watch as it starts downloading packages hehe
<phalrax>
ah I tried -p when rebuilding earlier, could it somehow have to do with that?
orivej has joined #nixos
<heyarne[m]>
<yaymukund "heyarne: I just did this for Rus"> can you go into more detail? or maybe provide a working example?
Mateon2 has joined #nixos
<rose_>
ajs124: (and anyone else) do you think the arm machine should start happily building packages as normal if i give it a standard remote builder config? or do i need to do more than that to convince it to cross compile without trying to load an emulator?
dandart has quit [Quit: dandart]
<exarkun>
There used to be another hurdle to overcome with nixpkgs cross-compilation. You can either cross-compile everything or nothing. You can't mix them. You can't have a native-compiled glibc and a cross-compiled coreutils linked against it.
<simpson>
odiug[m]: Very cool. It seems like you understand the problem where we'd want to offer folks the ability to *orthogonally* configure certain services; folks want to write `services.openssh.enable = true;` and have it Just Work even with S6.
sangoma has joined #nixos
<simpson>
odiug[m]++ please keep sharing your experiments.
<{^_^}>
odiug[m]'s karma got increased to 1
<{^_^}>
[rfc39-record] @grahamc pushed commit from rfc39 to main « Automated team sync results. »: https://git.io/JYY91
<odiug[m]>
ha, that's exacty one of the goals.
<odiug[m]>
It works already like that for unbound and sshd.
gustavderdrache has quit [Quit: Leaving.]
gustavderdrache has joined #nixos
<odiug[m]>
For now, I only tested it in a vm. Next challenge, make an iso and test it on real iron.
<simpson>
Good luck.
Guest52628 has quit [Changing host]
Guest52628 has joined #nixos
Guest52628 has joined #nixos
Guest52628 is now known as joepie91
jakobrs has joined #nixos
<jakobrs>
How do I package something that can be built against either Qt5 or GTK3?
<jakobrs>
In particular, I can't use wrapQtAppsHook directly because it's not called with libsForQt5.callPackage.
<jakobrs>
Nor can I use qt5's mkDerivation function because that won't work when building for GTK
<ambroisie>
Hello people
<ambroisie>
I'm currently trying to install NixOS in a VM as a preparation for migrating my laptop to it
<ambroisie>
I already have a flake-enable repo for my server, which is what I am using as a basis for the VirtualBox install
<ambroisie>
And I can't seem to be able to do `nixos-install --flake [etc...]`
<ambroisie>
I get an error about `nmd` (well a path to the nix-store for `nmd`) not being accessible in restricted mode
stree has quit [Quit: Caught exception]
phalrax has quit [Quit: Connection closed]
ixxie has joined #nixos
stree has joined #nixos
domogled has joined #nixos
rose_ has quit [Quit: Connection closed]
gh0st[m]2 has joined #nixos
<ambroisie>
A cursory google search indicates that the problem could be coming from `home-manager`...
ddellaco_ has quit [Remote host closed the connection]
<ambroisie>
How can I fix this issue? Has anybody encountered it before?
<gh0st[m]2>
What does your error message look like?
ddellaco_ has joined #nixos
gustavderdrache has quit [Quit: Leaving.]
gustavderdrache has joined #nixos
VideoGameEnjoyer has quit [Remote host closed the connection]
FRidh has quit [Quit: Konversation terminated!]
stree has quit [Quit: Caught exception]
stree has joined #nixos
VideoGameEnjoyer has joined #nixos
domogled has quit [Ping timeout: 240 seconds]
davidv7_ has joined #nixos
ixxie has quit [Ping timeout: 260 seconds]
<gh0st[m]2>
That could be a number of things, really. In my experience overall, usually installing a basic config and then switching to flakes works a lot better
ixxie has joined #nixos
VideoGameEnjoyer has quit [Ping timeout: 240 seconds]
<{^_^}>
nix-community/home-manager#1262 (by Kloenk, 43 weeks ago, open): Import from derivation with nmd
<jakobrs>
I just had to manually do `unpackPhase = "chmod +w source -R"` ... why might `source` not be writable?
<jakobrs>
Oh wait `sourceRoot = "source/whatever" makes everything else readonly
ris has joined #nixos
<ambroisie>
I wonder why that is only an issue on a fresh install though, because myserver is working perfectly when I `nixos-rebuid --flake`
ris has quit [Ping timeout: 252 seconds]
hiro98 has quit [Remote host closed the connection]
hypn0z[m] has joined #nixos
jakobrs has quit [Quit: WeeChat 3.0]
gustavderdrache has quit [Quit: Leaving.]
numkem has joined #nixos
apache8080 has joined #nixos
sangoma has quit [Ping timeout: 240 seconds]
ahmedelgabri has joined #nixos
gustavderdrache has joined #nixos
sangoma has joined #nixos
fendor_ has joined #nixos
ahmedelgabri has quit [Ping timeout: 250 seconds]
Mateon1 has quit [Remote host closed the connection]
Mateon1 has joined #nixos
fendor has quit [Ping timeout: 240 seconds]
zupo has joined #nixos
zebrag has quit [Quit: Konversation terminated!]
zebrag has joined #nixos
bitmapper has joined #nixos
davidv7_ has quit [Remote host closed the connection]
waleee-cl has joined #nixos
meh` has quit [Ping timeout: 246 seconds]
xdefrag has quit [Remote host closed the connection]
mcornick has quit [Quit: bbl]
sangoma has quit [Ping timeout: 240 seconds]
xdefrag has joined #nixos
fuiltilt has joined #nixos
rose_ has joined #nixos
Lord_of_Life_ has joined #nixos
Lord_of_Life has quit [Ping timeout: 260 seconds]
Lord_of_Life_ is now known as Lord_of_Life
Izorkin has quit [Ping timeout: 265 seconds]
<rose_>
still on the cross compiling thing... afaict a package that was cross compiled is treated differently (different hash?) to one that wasn't - is this because of nix, or nixpkgs, or both? also why is it
Izorkin has joined #nixos
sangoma has joined #nixos
<exarkun>
I'm not sure how to split nix and nixpkgs in this case
<exarkun>
gh0st[m]2: That's my (naive, limited, probably wrong) understanding
<rose_>
i guess it makes sense though
<exarkun>
Even after we have content-addressed derivations, will a cross compiler produce byte identical output to a native compiler? If so, yay. Seems like a delicate thing, though.
<rose_>
would it be possible to specify somewhere that x86_64-unknown-linux-gnu-gcc will produce the same output regardless of what it's being run on?
<rose_>
(that's assuming that it does)
<exarkun>
I think there is/was an idea to never use a native compiler - cross compile everything, all the time. That might avoid some of the delicate parts of requiring byte-identical output from native and cross-compilers.
<exarkun>
(Yes, you can cross-compile, eg, from x86_64 to x86_64)
<rose_>
ugh i hate c so much
<exarkun>
rose_: I think that idea is pretty far from how nix works
<gh0st[m]2>
I mean, you could alternatively "cross-compile" by simply running your compiler and system through qemu, right?
<rose_>
that's what i'm trying to avoid because it's slow and wasteful
<rose_>
exarkun: why?
<rose_>
like i say i'm assuming that gcc really does produce the same output regardless of builder machine. i might be wrong (wouldn't be surprised) but if it does, i don't see a reason why it couldn't be specified?
<exarkun>
rose_: Because nix doesn't "know" anything about what any compiler produces as output
novafacing has quit [Quit: novafacing]
ixxie has quit [Ping timeout: 240 seconds]
<rose_>
ah yeah i'm starting to get you
<rose_>
i guess the compiler is "just another input" right
novafacing has joined #nixos
<exarkun>
nix knows an expression that can be used to produce software and assumes/requires that the same expression produces the same software
<simpson>
rose_: A proof that a compiler is correct must be relative to some foundation. When we start talking about bootstrapping, then that foundation is suddenly missing. We *could* assume that C compilers are correct, but historically they haven't been.
<exarkun>
nothing in the system gives it the power to reason about whether different expressions will produce the same software or different software
<rose_>
ugh
<exarkun>
(Of course, it's also not generally true that the same expression will produce the same output and most build systems don't have this property ... so then there's a ton of hacks to erase the differences where they're erasable so that enough stuff works to make things practical, and we can all hope software continues to converge on reproducable build systems)
malook has joined #nixos
ddellac__ has joined #nixos
dotdotok has joined #nixos
novafacing has quit [Client Quit]
<rose_>
i totally get why running a compiler on a different arch makes a different derivation but... it's just such a pain that the consequence of that is that i can't use big machine to build package fast and i'm stuck doing it on my laptop
ddellaco_ has quit [Ping timeout: 265 seconds]
<rose_>
should have just stuck with arch
<dotdotok>
could that be remedied by flakes/explicitly building for a different architecture?
<rose_>
i don't think flakes would help, but i think it should be *possible* assuming we know for certain that gcc does not vary output with arch it's running on, and someone can add/fix the necessary machinery in nixpkgs
ManiacOfMadness has joined #nixos
<dotdotok>
mm, well, I know you can specify packages for different architectures with flakes at least, or does it not actually build the derivation then?
<rose_>
must admit i don't actually know much about flakes
<gh0st[m]2>
it only builds the host's derivation usually iirc
<srhb>
E
<rose_>
i think there could be some "hidden" argument into nixpkgs that sets what arch the compiler can run on but doesn't affect the hash of its derivation... and the more i think about this the more i think that's already what is in nixpkgs with the whole depsBuildBuild thing
<srhb>
oops, sorry.
<rose_>
hehehe
<exarkun>
If you make the compiler itself constant and just have the target variable then you might make it work. But that may just be another way of saying "cross compile everything".
<srhb>
rose_: Yes, impure (current) Nix gets a hint from the secret builtins.currentSystem, but it _does_ affect hashes from then on.
<srhb>
Or rather, nixpkgs, not Nix.
<rose_>
in the future could this be changed to not affect the hash? i really don't know much about this, and i get that it goes against the theory or something but surely it would allow cross compiling with much less pain?
<dotdotok>
rose_: heh, well, i think there's a longterm plan to have nix work with ipfs, so that would probably put a big rippy on that because *everything* would be content-addressed
<gh0st[m]2>
Probably content-addressing the derivations would be best, provided the same compiler produces the same output given the same inputs, which is pretty much the point of Nix
<simpson>
rose_: But just to double-check my reading comprehension in the backlog, our existing cross-compilation infrastructure *does* work for your use case, right? It's just slower than preferred because everything's indirected through QEMU?
<rose_>
something i never understood about content addressing is how can e.g. `nix-env` know what hash to look for when it's installing a derivation
<exarkun>
simpson: is that really what it's called? seems like it should be called something else since it is something else
<srhb>
rose_: For what it's worth, I don't think secret impurities like that add value to nix (even from a usability POV) quite the opposite.
<rose_>
you don't think that being able to actually, properly cross compile is value?
<srhb>
rose_: Right now, it evaluates the derivation (essentialy inputs) and computes what the output will be. With CA, you'd leverage some sort of trusted DB of inputs to outputs.
<srhb>
rose_: Oh, well, when you put it like that, sure. Disregard everything I said. ;-)
<rose_>
hehehe
<srhb>
No, really, something else might benefit from that, but not Nix.
<rose_>
all i want is speed!!
<srhb>
The entire point just evaporates with impurities, very very quickly. :)
<rose_>
aren't there a bunch of impurities in nix already?
<rose_>
or in nixpkgs?
<rose_>
i thought there were but now i can't think of a single one...
<srhb>
Yeah. Most of them are being worked on exactly because you end up with hours and hours of incomprehensible debugging.
<rose_>
right
<srhb>
rose_: builtins.currentSystem is one, exactly
xdefrag has left #nixos ["ERC (IRC client for Emacs 27.1)"]
<srhb>
rose_: Reading various config from the homedir, overlays littered all around, ...
<srhb>
Tons of it.
<rose_>
with ca, what if i write my own derivation and then try to cross compile it? how does my machine know the address before the derivation has ever been built?
<srhb>
That's without even going into current time, cpu features, etc.
<srhb>
rose_: It doesn't.
<rose_>
oh yeah i hate how overlays and channels (unless i misunderstand channels) work
teutat3s[m] has joined #nixos
<srhb>
rose_: There's no way to compute CA except building it and going, "ah, NOW it's obvious, here's the hash"
<rose_>
hehehehe
<srhb>
rose_: You've probably understood channels then :-)
<rose_>
lol
<rose_>
i think i accidentally "upgraded" my entire user profile to nixos-unstable the other day...
work_ has joined #nixos
sangoma has quit [Ping timeout: 240 seconds]
<srhb>
Fortunately I think this is one of the things flakes really get right. It makes deps very explicit and up-front and removes evaltime impurities.
<gh0st[m]2>
Flakes are awesome, definitely
<srhb>
Parts of them are pretty incomprehensible too, (yet)
<dotdotok>
(that applies to a little more than flakes i think tbh)
<gh0st[m]2>
Yeah. I also wish there was more tooling for them right now
<srhb>
dotdotok: Fair.
<srhb>
I'm not trying to rag on flakes, just saying they're not a magic bullet and they solve nothing related to the problem at hand here, really.
<srhb>
Except making things more explicit. :)
<gh0st[m]2>
Yeah, that'll be helped more by content-addressing
<rose_>
i haven't even used flakes ever and i think they're great
<dotdotok>
im still trying to grok them (and some other stuff too)
<dotdotok>
but they *feel* like they make sense
<dotdotok>
just cant use them intuitively yet
<srhb>
dotdotok: Try putting a let in inputs then.
<srhb>
I'll temper your feelings for free :-)
<dotdotok>
i learned thats a no-no a bit ago iirc :)
mcornick has joined #nixos
ris has joined #nixos
<gh0st[m]2>
what does that do? I haven't tried that. Didn't think to honestly
<srhb>
gh0st[m]2: flake.nix isn't really nix, or at least, it's nix with a ton of restrictions
<srhb>
gh0st[m]2: inputs cannot be just any expression that evaluates to an attrset. It has to be an actual attrset.
<srhb>
gh0st[m]2: So, referential transparency is basically out the window.
<srhb>
And flake.nix is arguably not nix.
<dotdotok>
i do all my stuff in lets when i can though, picked that up from working with dhall
<gh0st[m]2>
ah, yeah, it's more akin to a config file really that uses the Nix language...kinda
<srhb>
Yeah, syntactically equivalent.
<dotdotok>
the outputs section is the closest thing to actual nix you've got
<srhb>
That's, I think, _actual_ nix.
<dotdotok>
yea, which is where i stick my lets
<srhb>
Indeed. :)
<srhb>
You seem much less bothered by this than I am -- that's good for you I'm sure. ;-)
<dotdotok>
but yea def had a bit where i was like "how can i put my inputs in a separate file and source that"
<gh0st[m]2>
It is kinda weird when you think about it. It'd be like if Cargo used .rs files
<srhb>
I think there's been talk about toml'ing the inputs, for the exact same reason.
<srhb>
I bet it was just convenient to reuse nix, really.
<srhb>
These things tend to stick though, and it's confusing.
<gh0st[m]2>
If it worked like actual Nix though, and just kinda had an evaluator, that might be interesting as a concept. Don't know how that would work though
ahmedelgabri has joined #nixos
ddellac__ has quit [Remote host closed the connection]
vidbina_ has quit [Ping timeout: 260 seconds]
ddellaco_ has joined #nixos
malook has quit [Remote host closed the connection]
ddellaco_ has quit [Remote host closed the connection]
<apache8080>
for the fileSystems device mounting config can I get the system to wait until the disk is mounted before running systemd services during the stage2 boot?
stree has quit [Ping timeout: 265 seconds]
Qwerky has quit [Remote host closed the connection]
<apache8080>
hmm not exactly, so right now I have fileSystems."/" on one disk and fileSystems."/data" on another disk, stage2 boot starts since "/" is ready and starts running systemd services
jarkad has quit [Ping timeout: 252 seconds]
<apache8080>
but some of those systemd services rely on stuff in "/data"
<apache8080>
I just don't want stage2 to start until "/data" is mounted
<gh0st[m]2>
That's what I'm thinking, is there a way to like `wait` in a loop until it's mounted?
zakame has quit [Read error: Connection reset by peer]
VideoGameEnjoyer has quit [Remote host closed the connection]
Qwerky has joined #nixos
troydm has quit [Quit: What is Hope? That all of your wishes and all of your dreams come true? To turn back time because things were not supposed to happen like that (C) Rau Le Creuset]
<aaronjanse>
I did `--show-trace`, and the trace had the text "while evaluating the option [...]" Does this mean that Nix (the language evaluator) understands NixOS modules?
<aaronjanse>
Woah I think it does
alexherbo2 has joined #nixos
bew has joined #nixos
<radvendii>
what is the difference between mkPythonDerivation and buildPythonPackage?
<rmcgibbo[m]>
What's mkPythonDerivation?
<rmcgibbo[m]>
I just cd'd into my checkout of nixpkgs, and I don't see any mention of that word according to `git grep`. buildPythonPackage is the way.
emilsp has joined #nixos
<gh0st[m]2>
It looks like the most recent references to "mkPythonDerivation" in old issues on GitHub
<gh0st[m]2>
Yeah, the lib for the first is removed from Nixpkgs, and looks to have been replaced during a refactoring
<exarkun>
Let's say I have a derivation with a complex set of dependencies and it comes to pass that one or more of those dependencies is induced to fail to build (really, the dependecies are changed - eg to upgrade some piece of software)
<exarkun>
How do I receive a succinct report about the dependencies which fail to build and the path through the dependency graph from "my" derivation to those failures?
<exarkun>
I have some system tests that use nixpkgs.nixosTest
<exarkun>
So it's 4 or 5 whole nixos systems that interact with each other running in QEMU
<l33[m]>
nice!
<exarkun>
I more or less know that the problem package is python37Packages.pyopenssl and if I build the individual nixos system drvs then the failure is _somewhat_ straightforward (well, at least it was for one)
<exarkun>
but I only got here after 12 hours of building stuff on my laptop. there's no way I could have figured this out from what my CI does
<exarkun>
So I'd like my CI to do something better ...
<l33[m]>
I'm not sure. What CI are you using out of curosity?
<exarkun>
my CI is GitLab running nix-build in a shell script
<exarkun>
The report that `nix-build` gives when there's a single package failing _after_ I've prepped the system with a --keep-going run is close to useful
<exarkun>
pretty much says what's wrong and what the path to it is
<exarkun>
but if there are 2 failing packages then it's all mushed together and the paths are unclear
<exarkun>
and if there's anything except the failing package then success output is mixed in too making everything that much harder
stree has quit [Ping timeout: 240 seconds]
<l33[m]>
yeah, hydra is quite cool, because you can bring up the logs of any build... you not flying blind in gitlab land!
civodul has quit [Ping timeout: 240 seconds]
ahmedelgabri has joined #nixos
<exarkun>
guess I better take a closer look at that
<exarkun>
if I ever dig out of this openssl hole
<l33[m]>
i wish you the best of luck!
<exarkun>
:)
ahmedelgabri has quit [Ping timeout: 250 seconds]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
evils has quit [Ping timeout: 260 seconds]
lsix has quit [Ping timeout: 240 seconds]
<exarkun>
Okay so, that hole is shaped like this ... My openssl override works and gives me openssl 1.1.1k everywhere. my python2 pyopenssl override works and gives me pyopenssl 20.0.1 in at least _some_ places (not sure yet if it's everywhere)
stree has joined #nixos
<exarkun>
my python3 pyopenssl override does not work and I get pyopenssl 19.0.0 (which is broken against openssl 1.1.1k)
lsix has joined #nixos
<exarkun>
annoying that python2 and python3 packages can't be overridden in exactly the same way
<exarkun>
and annoying that I don't know how to override a python3 package
<apache8080>
I want to create a symlink to /var/log/journald, the only way I can think to do it is in postBootCommand but I was wondering if there is a better approach?
<clever>
matthewcroughan: i think you also want something like python3Packages.callPackage?
<babbaj[m]>
<gh0st[m]2 "@bbabj:nerdsin.space: Builds the"> I ran it and it just said no diff detected
<matthewcroughan>
clever: is this the best way of going about this?
<gh0st[m]2>
<apache8080 "I want to create a symlink to /v"> I don't know that I'd consider it "better", but you could see about writing a NixOS Module
<matthewcroughan>
what I'm attempting
zebrag has joined #nixos
<matthewcroughan>
Or should I be making a flake for this that updates nightly?
<clever>
matthewcroughan: flakes wont auto-update
<matthewcroughan>
clever: no, but I can have a bot work it out on every commit from upstream
<matthewcroughan>
just like colemickens' nightly firefox flake, seems like the way forward for these rapidly updating packages
<matthewcroughan>
even if this can be done in a shell, shouldn't I really just be packaging it?