<winterqt>
hey all - probably a stupid question, but something I'm packaging depends on having an ssl certificate bundle available in a number of locations (it goes through a bunch of standard locations to find it), how exactly can I make sure this is available via a dependency or something?
nicoo has quit [Remote host closed the connection]
<jackdk>
when using a private s3 bucket as a binary cache and pushing in a postbuild hook on macos, where do I specify aws credentials? `sudo aws configure set aws_access_key_id AKLOLNO` etc works fine on linux runners but I'm getting aws permission errors for mac
<srhb>
jackdk: Does mac mean daemonless? If so, I'd guess whoever runs Nix (presumably your user?)
akaWolf has joined #nixos
luxemboye has joined #nixos
<jackdk>
srhb: I'm using https://github.com/cachix/install-nix-action/ which claims a multi-user install on all platforms. running `aws configure set foo` without sudo didn't work for me either
<srhb>
Ah OK. Then yeah, whoever runs the _daemon_ sounds right. Maybe there's some isolation going on on MacOS I don't know about, sorry. Maybe check out if you can feed the daemon the normal AWS env vars to point to credentials.
<colemickens>
how many thousands of hours have nixosers wasted trying to get emoji working
<colemickens>
I spend an hour or two every year or so, never have gotten them to work reliably much of anywhere in Linux. Anytime I'd pull up a test web page inevitably some of them would be wrong. Or the workarounds would break odd things all over my desktop.
JerryXia1 has quit [Ping timeout: 252 seconds]
<cjb>
hahaha xD Yeah, I've probably wasted a cumlative day or two over the years getting emoji working on different distros.
<colemickens>
don't mean to discourage you or tell you not to, of course, just musing :)
<colemickens>
and/or hopeful that when someone achieves emoji nirvana on nixos that they doc it in the wiki or at least link to their config or something :P
<elvishjerricco>
Is wireless enabled in the minimal installer image?
turlando has joined #nixos
jonringer has joined #nixos
jmeredith has quit [Quit: Connection closed for inactivity]
<turlando>
Hello everyone. I'm trying to build emacs with the lucid toolkit. I'm doing it with (emacs.override { toolkit = "lucid"; }) but the compilation fails. Here is the output: https://pastebin.com/raw/twrwUuc8 I'm not sure what is going on
<srhb>
turlando: Yes, that sounds right. From your error it looks like postfixup fails because it's trying to patchelf the wrapper scripts which I think are triggered by the wrapGAppsHook, which is enabled by withGTK3
<turlando>
I have no idea what patchelf is supposed to do :)
maisem has left #nixos [#nixos]
maisem has joined #nixos
<srhb>
turlando: If you just read the postFixup hook the command invocations are fairly descriptive :)
<srhb>
It adds libXcursor to the rpath of the emacs binary. But I think with withGTK3, it's targeting the wrapper scripts (bash) instead of actual (ELF) executables.
<turlando>
Oh! Makes sense! Sorry, I just woke up
<srhb>
No reason to be sorry. :)
<turlando>
It just finished compiling and apparently it worked! Thank you a lot (now I can start working I guess, uff)
<srhb>
Ew, work. ;) Sure thing.
<srhb>
Inspiration for easy contribution: Add an assertion that prevents that combination (toolkit = lucid and withGTK3 = true)
<turlando>
srhb Thanks for the suggestion! I will try to sort out a path later today :)
maisem has quit [Read error: Connection reset by peer]
maisem has joined #nixos
vidbina has quit [Ping timeout: 246 seconds]
waleee-cl has quit [Quit: Connection closed for inactivity]
mounty has quit [Ping timeout: 240 seconds]
<turlando>
I think I spoke way too early. Emacs can't find the font I use (SourceCodePro). It is installed globally through configuration.nix and Emacs seems to be compiled with Xft. Could it be I'm running emacs through an SSH X forward?
<turlando>
Also the font shows up correctly in fc-list
<turlando>
symphorien[m] I eventually realized it, thanks. Also it was showing up because I ran fc-list on my local machine :) Time to grab some coffee
<jackdk>
srhb: setting the env vars using `sudo launchctl setenv AWS_FOO` worked. I'm kicking the nix daemon with `sudo launchctl kickstart -k system/org.nixos.nix-daemon` but that's because the nix manual recommends it after setting post-build-hook, not sure if it's strictly necessary
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
jess has joined #nixos
ahmedelgabri has joined #nixos
epitron[m] has quit [Quit: Idle for 30+ days]
BJTH[m] has quit [Quit: Idle for 30+ days]
odiug[m] has quit [Quit: Idle for 30+ days]
hypn0z[m] has quit [Quit: Idle for 30+ days]
grahamc[m] has quit [Quit: Idle for 30+ days]
kenny[m]1 has quit [Quit: Idle for 30+ days]
dbirks[m] has quit [Quit: Idle for 30+ days]
marius_k[m]1 has quit [Quit: Idle for 30+ days]
infinisil-m has quit [Quit: Idle for 30+ days]
figgyc has quit [Quit: No Ping reply in 180 seconds.]
figgyc has joined #nixos
akaWolf has quit [Ping timeout: 240 seconds]
<jackdk>
Now I've populated a private S3 cache, I'm struggling to use it. I need to access the bucket through an assumed role (role_arn/source_profile in .aws/config). I have this set up and working for root (`sudo bash -c 'AWS_PROFILE=foo aws s3 cp s3://bucket/nix-cache-info foo` works), but adding `s3://bucket?profile=foo` to `binary-caches` gives me 403s. Am I doing something wrong, or is this not supported by the sdk?
<colemickens>
how would nix know the creds to use in that scenairo?
<colemickens>
oops, it's in the url, profile=foo, I see
<clever>
jackdk: try running the nix command as root, does it work?
<jackdk>
clever: negative. running `nix-shell -vvvvv | egrep '(aws|bucketname)'` even as regular user shows that the aws sdk finds `/root/.aws` but I see `warning: AWS error fetching 'nix-cache-info': Access Denied` regardless
<clever>
jackdk: can `aws s3 cp` read nix-cache-info?
svrana has quit [Ping timeout: 260 seconds]
zupo has joined #nixos
<thefloweringash>
I have a PR that's failing "ofborg-eval" and I'd like to debug it without pushing repeatedly to github. is there documentation about what this is doing?
o1lo01ol1o has quit [Remote host closed the connection]
<{^_^}>
[nix] @NinjaTrappeur opened pull request #4750 → nix build: make dry-run to print a json output if --json is enabled → https://github.com/NixOS/nix/pull/4750
<yaymukund>
anyone get obs segfaults? i'm using it with wlrobs, and dmesg says: ..obs-wrapped-w[177711]: segfault at e ip 00007faba5d39047 sp 00007fab8a7fb9e0 error 4 in libobs.so.0[7faba5cfb000+9c000]
ericsagn1 has quit [Ping timeout: 260 seconds]
<yaymukund>
it happens seemingly arbitrarily after it's been on for some time (doesnt matter if it's streaming/recording or not)
<__monty__>
I have an expression with a conditional buildInput (lib.optionals) of libunwind, and the library path isn't getting added so the linker can't find -lunwind. Is there a better way to solve this than explicitly passing the path in LDFLAGS?
<clever>
__monty__: what is the full line for that expr?
<__monty__>
Seems like there's been a little bit of refactoring though. This used to be in `llvm/11/libc++abi.nix`, that's what I'm currently working on.
<clever>
i would expect the linker to just find that
<clever>
run nix-shell on the failing .drv file, then `echo $buildInputs`
<__monty__>
clever: Hmm, how do I find the drv to run it on? If I run it on the expression I'm using (I'm passing new bootstrap-tools for stdenv) then buildInputs is empty.
<hexa->
__monty__: nix-instantiate
<__monty__>
I don't need to run genericBuild to populate the var, right?
<__monty__>
Hmm, so I'm on darwin and libunwind's supported platforms contain just linux. Why don't I get an error about unsupported platforms when building libcxxabi?
<radvendii>
MysteriousSilver: the point of what they were saying is that ${} mostly works on strings, but there are also other objects that know how to turn themselves into strings, and it can work on those too
<radvendii>
you don't have to worry about it too much
<maisem>
I have to create a tmp dir to consolidate some headers in the preBuild phase. is there a temp dir that i can use/create and clean up after build is complete?
ahmed_elgabri has quit [Ping timeout: 250 seconds]
<{^_^}>
[nixpkgs] @mweinelt opened pull request #120889 → Update some of my python packages and remove myself from maintainers list → https://github.com/NixOS/nixpkgs/pull/120889
<srhb>
ambroisie: Not easily or generally, though some things _may_ be safe to null in an overlay.
<srhb>
ambroisie: But nixos modules in general are not something you can manipulate like this, unless written with that exact goal in mind.
<ambroisie>
I though so
<ambroisie>
It's not really problematic, I can live without it
<ambroisie>
Thanks for the quick answer
ahmedelgabri has quit [Ping timeout: 250 seconds]
<srhb>
It's a fair "wishlist thing" though :) Modules usually _can_ be written in a way that the user gets to pick and choose (eg. by providing functions that hook in, say for filtering)
<srhb>
But yeah, they're not as flexible as packages wrt. overrides.
<dupon1>
Hi, I'm trying to use the unstable version of gnupg, NixOS looks to pull unstable but when I'm doing `gpg2 --version` I still have stable version
<srhb>
dupon1: nixpkgs-channels should not ever be used anymore, mentally substitute nixpkgs :) It _used_ to be the repo that had branches corresponding to "channels" (really, commits that have been found passing some sets of tests) but it's all in nixpkgs now.
maisem_ has quit [Remote host closed the connection]
<radvendii>
is there a way to get meson to print verbose output from nix? if i try to add "--verbose" to mesonFlags it complains about not recognizing the option
<radvendii>
setting debug = true; on the derivation doesn't seem to do anything
maisem_ has joined #nixos
figgyc has joined #nixos
ezemtsov_ has joined #nixos
zebrag has quit [Remote host closed the connection]
zupo has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Lord_of_Life_ has joined #nixos
neiluj has joined #nixos
neiluj has quit [Changing host]
neiluj has joined #nixos
<srhb>
bennofs: It's my understanding that most people who target something like rpi use one of the many cross-compile options. That's a workaround sure, but it might be practical for you, to.
Lord_of_Life has quit [Ping timeout: 265 seconds]
Lord_of_Life_ is now known as Lord_of_Life
<samueldr>
srhb: actually I think that many do native builds
<colemickens>
hm nixos-install already uses --store $mntpoint though
<samueldr>
but a good chunk of those native builds use qemu userspace emulation "qemu binfmt"
<samueldr>
so native builds with extra steps
<srhb>
samueldr: Oh, that's just me classifying it wrongly as "cross" then :)
<srhb>
That was what I was thinking of.
<samueldr>
I didn't want to assume :)
<samueldr>
but yeah, many try to go the cross-compilation route, and will work depending on the complexisty of the requirements
<srhb>
Happy to be corrected. I'm not sure I completely understand the distinction really.
<srhb>
I guess it's whether "nixpkgs" considers itself cross?
<bennofs>
probably I should just use another aarch64 board to the eval and then only push the built configuration to the pi
<samueldr>
cross-compilation the compiler is for e.g. your x86_64 machine, but outputs aarch64 binaries
<samueldr>
so things like pkg-config or sdl-config needs to be x86_64 even though the libraries are aarch64
civodul has quit [Quit: Rebooting Emacs (and Guix).]
<srhb>
bennofs: If you haven't tried the binfmt route I can only recommend it based on the "omg wow" reactions I've heard from friends. Apparently it's sorta magic dropin.
<samueldr>
which makes things way more complex to deal with :)
<srhb>
samueldr: Gotcha!
ahmedelgabri has joined #nixos
stree has quit [Quit: Caught exception]
<samueldr>
while with "qemu-binfmt", everything is aarch64, qemu runs them, but delegates syscalls to the kernel in use
<bennofs>
how does binfmt work? I build the conf on a x86_64 machine and then push that?
<srhb>
bennofs: Yup.
stree has joined #nixos
<srhb>
Of course if you have a big aarch64 around that's probably a waste of time :)
<samueldr>
you could also do the eval on x86_64
<samueldr>
without binfmt
<samueldr>
and build the .drv on the remote machine
<samueldr>
eval is expensive with Nix
<samueldr>
but builds... well... it depends
<bennofs>
eval only actually sounds like not too bad
<bennofs>
but... then I first have to ensure there is no IFD anywhere :)
<samueldr>
that'll entirely depend on what you'lle end up building
<samueldr>
well, IFD is fine if it's actually correct IFD
<bennofs>
I mostly use stuff from the binary cache anyway
<samueldr>
as in, the output is the same independent on when/how/what did the work
<samueldr>
e.g. curl on aarch64 and curl on x86_64 fetching a tarball should give the same output hash... hopefully
<colemickens>
I think nixos-install definitely already supports this, and I just hadn't realized it because I was manually acquiring the full closure path ahead of time so that I could include my caches. But `nixos-install` passes through --options. Aka, this should already be supported...
<bennofs>
ah wait, does it pick the correct curl if I IFD in a aarch64 target eval on x86_64?
figgyc has quit [Quit: No Ping reply in 180 seconds.]
<samueldr>
not sure
<samueldr>
I think I just mixed myself up a bit with IFD and FOD
<samueldr>
but IFD maybe will not matter, maybe will, depending on what you're doing in IFD
<srhb>
If _really_ needed it's probably possible to have some partial cross-nixpkgs for IFD, but down that path may(*) lie madness.
<xAr86>
Hey, I'm trying to overlay i3status-rust and having issues to get the current cargoSha256, I tried the "00000..." value and lib.fakeSha256 but I don't get any "got" else then the "try to do it like with this" message ...
<DigitalKiwi>
Kinnison: --reference (but read the man page)
<Kinnison>
ta
* Kinnison
has reached the point of discovering why ms-python.python has yet to be updated in the vscode-extensions though -- the new version depends on the jupyter extension which afaict has yet to be packaged
ddellaco_ has quit [Ping timeout: 240 seconds]
<superherointj>
supersandro2000, can you re-review PR#120049 once more? Thanks.
figgyc has quit [Client Quit]
<kini>
Why does nix-shell puts a bunch of stuff in `$PYTHONPATH` when I have e.g. `mkShell { buildInputs = [ docker-compose ]; }` in my `shell.nix`?
<kini>
Surely that's not actually needed for `docker-compose` to function, because I can install `docker-compose` into my user profile with `nix-env -i docker-compose`, which doesn't change `$PYTHONPATH` in my local shell...
<kini>
Or maybe our python interpreters are hacked to have `~/.nix-profile/lib/python3.8/site-packages` in the default internal PYTHONPATH?
<kini>
Either way, why not put the `PYTHONPATH=...` in the wrapper for the `docker-compose` executable rather than dumping it out into the nix-shell environment where it can interfere with other things?
Synthetica has quit [Quit: Connection closed for inactivity]
<cgbrewster[m]>
Would it be frowned up to download the latest version of every built package in nixpkgs from cache.nixos.org?
<pie_>
do you have the space for it
<cgbrewster[m]>
Yes, everything should be good on our end, I just don't want to put too much load on the binary cache (not sure how that's all setup)
<{^_^}>
[nixpkgs] @davidarmstronglewis opened pull request #120953 → cargo-watch: from 7.6.1 to 7.7.0 + fix builds on Darwin from → https://github.com/NixOS/nixpkgs/pull/120953
<pennae>
last we saw a few days ago was an s3 utilization graph that had 50T active storage
<cgbrewster[m]>
Impressive, that's all nixpkgs built ever right? I just want every built package for linux off of the latest nixpkgs unstable release
<pennae>
no, all of them was 270T
<gchristensen>
cgbrewster[m]: should be fine though to do that :)
pcarrier has quit [Quit: Connection closed for inactivity]
<cgbrewster[m]>
Ah okay, cool :) Wanted to check before bombarding someone else's service haha
<gchristensen>
how are you going to discover the packageS/
<pennae>
hope nixos is well sponsored to pay for all that bandwidth :D
<gchristensen>
pennae: the storage and CDN are very generously provided :)