<ngz>Hello. While building a package, I get the following fatar error: qtbase-5.11.3/include/qt5/QtCore/qglobal.h:45:12: fatal error: 'type_traits' file not found <ngz>Compilation is done with clang. <ngz>ArneBab: Yes, build fails. <ngz>The error is pretty classic, though. <ArneBab>Can someone tackle that? (I know I’m just asking instead of doing — I almost have my dev environment done, but only almost) <ArneBab>Do we already have some continuous integration system that regularly tries to install all packages? <ngz>(I don't even know if it is active actually) <ngz>ArneBab: I pushed a fix. Hopefully it should now build. <brendyyn>Ok so it turns out I couldn't symlink in to the output directory before the 'install phase... I guess I learnt something <ArneBab>ngz: it still doesn’t build, but now with a different error and I found that quite a few more things seem broken in my python-matplotlib setup — including py2 vs. py3 problems <nckx>The problem is people-power, not a lack of branches. <nckx>OriansJ: TL;DR for the videoless? (Or if you'd rather save your energy for an ML post: please do.) <OriansJ>nckx: Short version with a minor tweak to our current build process, all binaries gain a cross-distro operability that would benefit us in the long term. <nckx>Speaking of master, I think it's broken. <nckx>Makefile.am: error: 'version.texi', included in 'doc/guix.zh_CN.texi', also included in 'doc/guix.texi' <nckx>Good, then you have to fix it. <nckx>wednesday: To be pedantic, you posted a link to a banned pastebin which I never read because they harass Tor users 😛 <nckx>(Not about the harassment.) <wednesday>It was the first pastebin that came up in my browsers search, so it is the one I used, don't use debian paste enough for it to be there ha <wednesday>And debian paste doesn't work in my glorious text based browsers, which is silly since it's just text =^) <nckx>wednesday: That is ridic. <nckx>wednesday: There's always 0x0.st if you're feeling hard-core. <nckx>wednesday: That it doesn't work in $text_based_browser. <nckx>brendyyn: Well, I, erm, stand corrected, yay ☹ <wednesday>javascript is required to show text these days hue *nckx throws a match at the modern Web and walks away. <brendyyn>it's hit and miss. tor changes the end point constantly. sometimes i can use goooogle, sometimes i cant <nckx>brendyyn: Is that a euphemism for ‘sometimes Google blocks my endpoint’? Or are you hitting actual bad exit nodes? <brendyyn>google blocks me or puts me into endless recaptha all the time <brendyyn>nckx: Im typing over a synergy connection :p <nckx>wednesday: I'm-a yellin' at my baseball screen too loud to care, sorry. <wednesday>I've had google do that to me from using duckduckgo !g too much before ha <nckx>brendyyn: Oh. Captcha purgatory is worse than hell. <brendyyn>even via direct ethernet connection there is the occasional lag. must be because its via Tinc <nckx>brendyyn: What's synergy? VPN? <brendyyn>it connects your mouse and keyboard to another computer <brendyyn>so i move my mouse off the left of my desktop monitor and it goes on to my laptop <brendyyn>i suppose its a bit crazy that im now accustomed to using two computers at the same time <nckx>The things I manage to have never heard of sometimes o_o <brendyyn>im hacking code on my laptop, but then i have my home dir mounted on my desktop via ssh and i compile and build packages on the desktop that way <nckx>brendyyn: Before I got my fancy new laptop I basically used my old one as a thin client to my desktop at home, it's not quite the same but I know that uncanny valley of wait where am I doing what at the moment. <nckx>This synergy thing sounds like just what I should have been using for years. <brendyyn>I love playing with VPN's though. I like tinc on paper but it is inconsistent with making connections <nckx>brendyyn: Have you tried Wireguard? (I haven't.) <brendyyn>No. I'd like something wireguard based but with meshing <brendyyn>Tinc does nat traversal etc to find direct connections <nckx>OK, from what (little) I've heard of Wireguard it trusts IP addresses directly. <nckx>That would definitely be considered out of scope (or bloat depending on who you ask). <nckx>I should really play with it soon. <wednesday>nckx: change line 15 in doc/guix.zh_CN.texi from @include version.texi to @include version-zh_CN.texi and problem solved, now do that in master for me after the game hue <katco>can anyone think of a good example package to follow for untarring a source which produces no folders? i'm trying to write a simple local package for pulling a .tar.gz of a csv file down <nckx>wednesday: I was actually cheating and fixing it in secret. Ssh. <wednesday>Can't email right now; And still need to actually set up the git part of my email stuff <wednesday>and after I set that up I gotta send in my gnu globals patch <nckx>katco: I'd say that qualifies as a tarbomb (url-fetch/tarbomb, IIRC). <katco>nckx: perfect! i assume that's a greppable phrase for examples? :) <wednesday>With our current globals package you can't make tags for python because some shebang hasn't been changed <nckx>katco: It is! It might be overkill for a single file, if you want you can grep for trivial-build-system too (but be warned that it's trivial because it does nothing for you). <katco>nckx: yeah i almost got tha to work, but i was having to `use-modules` a _ton_ of packages <wednesday>I wanna write something in guile but got nothing to write, sad times <nckx>katco: Nothing wrong with using the gnu-build-system for such things IMO, even though you're not building anything. <katco>nckx: yes, that's where i went next, but the `unpack` phase tries to `chdir` into the first folder that was unpacked. in this case there isn't one and it errors <nckx>katco: /tarbomb will take care of exactly that. <katco>nckx: ahhh! ok, i'll try that <nckx> /tarbomb sounds like a fun IRC command. <katco>wednesday: you can, but in this case, i'm using `gnu-build-system` specifically for the unpack phase, and this is what fails <nckx>I've never worked on another autotools/gettext project besides Guix. Is coming home to a mess of uncommitted changes every time translations are updated… normal? *nckx runs git reset --hard again. <katco>i want to share a fun guix story that happened to me today. i'm learning about machine learning, and i'm following along with a book. the book wanted me to install python, some libs, use virtualenv, etc. i just ran `guix environment --manifest=manifest.scm` and continued on ;p <nckx>A very short story. As it should be! <katco>and now with this tarbomb, the data-set i'm working with is reproducible too! and part of the manifest :) <nckx>wednesday: Guix IPFS support would be nice, though I suspect you meant something slightly smaller. ;-) <nckx>(Or GNUnet, I don't really know if one's better.) <katco>hm, how could i symlink the csv generated from this package into cwd when i run `guix environment`? <nckx>buenouanq: That was my understanding. <nckx>So I don't know if it's overkill for us at this point? <tune_>I appreciate the hint about "guix pull --news"! It's fun to see what has changed and I didn't know how to see the cut-off info before <nckx>katco: I'm afraid that's not possible out of the box, unless I'm ignorant of some clever hack. <nckx>tune_: It probably didn't exist then ☺ It's brand-new. <katco>nckx: hm, ok... so philisophically, how should i consume the csv from this package? do i have to know the store path? ***tune_ is now known as tune
<katco>maybe have the package set an env var? <nckx>katco: If your CSV file is in /gnu/store/*/share/foo.csv, it will be available as $GUIX_PROFILE/share/foo.csv. <nckx>(I'm almost certain about the name of that variable ☺ ) <katco>nckx: ok, that's what i thought! but it's not there... hm. it _is_ in `/gnu/store/....data/share.csv` though? <nckx>I.e. the profile is a union of all relevant store directories and your data file is no exception. Would that suffice? <katco>yes, that would suffice if it was in my guix profile. it would be a known location which is perfect <katco>i am probably doing something wrong <katco>if i `guix package -i` it, it's in my profile... <nckx>Oh, never mind, $GUIX_PROFILE is apparently not automatically set. <nckx>Dur. I meant $GUIX_ENVIRONMENT (hey, it's 04:00 here). But I guess that you meant that it isn't in your profile, period? <nckx>katco: Setting a ‘search path’ would probably work but's it's a hack (for one, search paths are supposed to be defined by the consuming package, not the provider). <katco>`find -L ~/.guix-profile -name foo.csv` returns nothing <katco>it _is_ in `$GUIX_ENVIRONMENT/share` which i think is sufficient <katco>that is still a fixed location, based on an env var <katco>basically i just need to be able to point models at this csv <nckx>katco: Oh, yes, that's what I meant by profile. Not your user profile (~/.guix-profile). <nckx>An environment is just a profile. <nckx>I just tested and even top-level (/gnu/store/hash-foo-1/foo.txt) plain files are unioned into $GUIX_ENVIRONMENT. So it should just work. <nckx>It would go against Guix's very design to let packages install files outside of the store, but I'm sure you knew that. <katco>nckx: thanks very much for your help. saved me a couple hours at least! <ArneBab>nckx: I’m asking about a branch, because that could be a way to only have commits which have no failures, even if then the distro isn’t up-to-date. <wednesday>I need to package some go stuff today to try get my git sendemail working heh <pkill9>can anyone tell me why this returns #f in the Guile REPL?: (derivation? (gexp->derivation "the-thing" #~(begin))) <rekado_>pkill9: gexp->derivation returns a monadic value. <rekado_>BTW #~(begin) is not a valid expression. <davidl>I have a package that Im trying to download with git-fetch method that is SSH keys protected and works fine to download with git clone but not the git-fetch method as part of a regular package installation (git fetch fails in the build loog of the package) <davidl>Is it possible to access ssh-protected git repos with the git-fetch method? <pkill9>davidl: i odn't think so, you could clone the repository and set the git URL to "file:///path/to/repository" though <pkill9>that's what i do for my private repository <davidl>pkill9: that actaully didnt work. I get build-error fatal: <repo> does not appear to be a git repository. But it certainly is. <davidl>odd thing is that the log also says (git-fetch "file:///home/myuser/src/repos/myre?" ?) without the full repo name myrepo. <wednesday>Once again I am confused by packaging go stuff, no idea how go packaging works ha <pkill9>is there a function similar to plain-file that sets the file to executable? I want to generate a bash script and use it in the same way you'd use plain-file <pkill9>davidl: i think the log just cuts off long lines to save space <pkill9>davidl: hmm i think i also have that issue, it works for me using 'file://' in channels.scm, now that i think about it, maybe the build goe sinto a container before it fetches the source, so it won't have access to your home directory <civodul>roptat: could you send a -pre2 to the TP? <wednesday>I give up on packaging go, this package will still escape me <nixo_>Hello guix! :) can I get _only_ the derivation path from the guix package -i command? <roptat>civodul, I was about to finish the French translation for the manual ^^ <efraim>I'm not fully back from vacation yet, can someone check a vm with slim/sddm and see if it launches as expected? <jonsger>efraim: if you give me the config file, sure :) <wednesday>I'm using sddm right now, and it works, unless it's changed in the like hour since I updated <efraim>wednesday: good to know sddm works <wednesday>I fixed it a while ago, at least it got fixed for me heh <wednesday>give it a go and send any logs if you got problems <davidl>pkill9: is there some directory I can provide access to similar to guix package -L /load/path .. or some path I know it will have access to that I can put the git-repo in? <pkill9>aha, found what I needed, "computed-file" <pkill9>a build system (like cmake, meson, etc) written in guile would be neat *jonsger waits on debbugs to send an ACK mail... <raingloom>heyyy, i'm trying to `guix import pypi langdetect` but it complains about certs even though I've installed nss-certs, what do I do? <raingloom>oh. i gotta manually set the environment variables. i thought it added those to $GUIX_PROFILE/etc/profile automatically... <davidl>pkill9: thanks a lot for the tip with file:/// and using a channel I got this to work now and can find my packages with guix package -s ***ayerhart_ is now known as ayerhart
<tune>I think this means the package is behind <nckx>ArneBab: master *is* the stable branch. :-) Breaking changes aren't allowed at all. They slip through. Renaming master to stable won't solve that, technical measures (patch bot/queue, ‘CI‘, for all supported architectures, …) will. <ArneBab>nckx: what I thought about is a shadow branch where the changes only get pulled once all tests have passed <nckx>ArneBab: That's the plan, but making it happen is a lot of work. I think cbaines (not pinged) is working in that direction. <ArneBab>I can imagine … and I don’t want to only complain. Most of my experience with Guix has been very good. <nckx>And again, changes shouldn't be *pushed* unless all tests pass. But that doesn't happen. <nckx>(And I plead guilty to have broken plenty myself; mainly downloads that ‘worked fine here!’ and architectures other than x86_64. Sorry! ☺) <kmicu>ArneBab: do you have many issues with bugs on master? <ArneBab>kmicu: I have packages I need that do not build <kmicu>Is pinning Guix version (only for them) an acceptable option? <nckx>ArneBab [not volunteering just curious] <ArneBab>for that I’d need to know at which version they build <kmicu>I’m asking cuz since Nix/Guix is immutable then breaks/failures during updates should not be something important. <ArneBab>currently it’s still python2-sh-1.12.14 <ArneBab>I tried to circumvent that by moving to python3, but there the import fails <ArneBab>and pip (from python) thinks that it is version 10 instead of the installed version 18.1 <nckx>ArneBab: I searched on issues.guix.info for python{,2}-sh but can't find the bug. Do you have a link handy? <ArneBab>I posted the problem here — should I open a bug? <nckx>Oh, it just seemed like a problem you'd had for a long time, if it's new, never mind. (But yes, bugs should be reported or fixed 😛) <kmicu>ArneBab: try guix package -i /gnu/store/fchnd6mkd2xlwbq5w7xmwpwfra2qg2p2-python2-netcdf4-1.4.2 <ArneBab>I get guix package: error: path ?/gnu/store/fchnd6mkd2xlwbq5w7xmwpwfra2qg2p2-python2-netcdf4-1.4.2? is not valid <nckx>kmicu: That won't work. ArneBab: Click on ‘and N others’ on that Web page. <ArneBab>nckx: it’s a problem that only appears now that I managed to install gdal with python support. <ArneBab>clawing my way towards being able to run the old tool again <nckx>The latest evaluated git commit that produced that working package is at the top: guixSrc → 6bb5f88 <nckx>Then… it gets a little fuzzy because I only use Guix from a git checkout and have never pinned anything. <roptat>civodul, the chinese manual is not generated by guix pull <nckx>‘guix pull --commit=6bb5f88 && guix environment …’ will work but will also downgrade your entire guix untill you pull again so it's only good for testing. <kmicu>nckx: I thought that worked in the past. Do we have a regression? <nckx>kmicu: It never worked the way you're using it. You can ‘guix build‘ an *existing* *derivation* already in your store (/gnu/store/foo-bar.drv); maybe that's what you were thinking of? <nckx>There was a talk about it at FOSDEM. <wednesday> seems pretty cool, bot not something that'll be too useful to the way I do things <nckx>guix.info TLS certificate expired 4 days ago :-/ (guix-sysadmins pinged). <ArneBab>brendyyn: it’s like a supercharged makefile that can operate over multiple machines <nckx>wednesday: Quite possible, but you might be surprised. It's not just for (scientific) workflows. <wednesday>the only non emacs programs I use are my web browser, and guix, but they are both ran from within emacs <nckx>nckx: I don't take requests like that, but I use youtube-dl a lot myself so you'll find an update on master. <nckx>‘It broke yesterday when youtube did something’ being the best bug report ever. <wednesday>About as much detail as you get with youtube =^) <kmicu>The Token bug is not fixed yet so that was something else. <nckx>‘Error: something happened’. <nckx>‘Don't let it happen again.’ <nckx>Close! The Windows 10 BSOD was my wallpaper for a while. <kmicu>(Oh, nope, the token bug is fixed indeed.) <wednesday>Yea I sent the pull request that fixed it, at least I think I did heh <civodul>roptat: i hadn't seen the certbot and knot patches, thanks for the heads-up! <ArneBab>wednesday: that’s a great workflow then — gwl is about the times when you need workflows that don’t fit into emacs. <wednesday>Those must be dark times for the unlucky people who can't fit everything in emacs <civodul>roptat: and congrats on the translation! note that there's serious competition on the "es" side ;-) <roptat>civodul: and that makes me happy :) <roptat>When will we do the same with the thepherd? :p <wednesday>I can translate from American English to actual English if needed =^) <nckx>wednesday: I thought you spoke Spanish? I must be confused. <wednesday>Nope, my gf speaks Welsh though. I just do the engrish <wednesday>I hear Bing translate is good for Welsh though, our government use it and helped improve it or something <nckx>Hm, guix.cy.texi, nice. (Yes, I had to look that up.) *nckx knows what Welsh is, but didn't know the ISO code. <wednesday>Well that's more than a lot of people know ha <nckx>I know little more than they have the best flag ever. <davidl>How do a write a package-definition with code that runs in the build-environment before git-fetch is invoked? <nckx>Not saying I'd trade my partner for Rob Brydon with a wig. <nckx>davidl: Hum. What are you trying to do? <davidl>Im trying to read the ~/.ssh/id_rsa file to a string var and write it to a file in the build env, then use setenv "GIT_SSH_COMMAND" to use the private key I need to fetch the git-repo. <wednesday>Is there a page that explains the whole process of doing a translation? I'm curious about it <davidl>it would probably be better to update the git.scm file to allow to pass the (git-ssh-key "/path/to/privkey") along with the url and commit info. <davidl>but it would be nice to just write the ssh privkey to build-env if that would work to to get it done faster. <nckx>davidl: I thought there was a thread about it but it was about HTTP auth. I don't think there's any way to do what you want to do without a custom git-fetch. <davidl>nckx: maybe I can get around with using http auth instead. <wednesday>They got no team for Welsh from what I saw, sad times *jonsger joined the German translation party. Looking now over the Rhine and the Pyrenees to see how far they are :P <nckx>Party as in team or party as in Fest? <katco>nckx: i messaged you after you went to bed, and it bears repeating: thank you for the help last night. you saved me a lot of time :) <nckx>nckx: Thanks you! Did you get where you wanted to be? <rvgn>Is there a way to delete guix pull generations? <katco>nckx: i definitely did. i know have a template i can stamp out for pulling data into profiles for machine learning -- and it's repeatable (something i think is desirable in data science ;p) <ATuin>that's how i delete them at least <rvgn>ATuin Thanks! Just for clarification, is it "guix pull -p"? <ATuin>rvgn: I think this should do the trick, "guix package -p $HOME/.config/guix/current --delete-generations" <rvgn>nckx I tried already. After guix gc, when I do "guix pull --list-genarations", it is showing all the previous generations and last one as current. <ATuin>well if they are not removed, the gc is not going to delete them i guess <rvgn>ATuin Hmm. I'll try that. <ATuin>try to delete the generations first and since the package commands accepts a profile argument provide the current profile (i think is that the profile pull is populating) <MaliRemorker>Anyway, tried to install guix in a VM, starting from scratch (0.16 ISO). It all went fine until I tried do do the first pull <MaliRemorker>qemu: substitute: guix substitute: error: TLS error in procedure 'handshake': The TLS connection was non-properly terminated. ==> qemu: substitute: guix substitute: error: TLS error in procedure 'handshake': The TLS connection was non-properly terminated. 2019/04/24 15:35:36 ui error: ==> qemu: guix pull: error: corrupt input while restoring archive from #<closed: file 4202d20> ==> qemu: guix pull: error: corrupt input while rest <MaliRemorker>One thing worth mentioning is that the install was automated and the script doing the install connects to the VM via SSH <civodul>roptat: re translating the Shepherd manual: whenever you feel like it! :-) <civodul>MaliRemorker: does your VM have network access? <MaliRemorker>so the first phase ... doing system init worked without a glitch <MaliRemorker>civodul: well, okay, it seems i did not do ifconfig eth0 up prior to pull *jonsger tries to write an AppArmor profile for Guix <wednesday>Hopefully the patch I just sent in is of higher quality than my previous heh <rvgn>Intresting! Adding "openvpn-client-service-type" under services in system config, does not invoke and install "openvpn" package in system profile. O_o. This shouldn't be happening right? <roptat>Why would you need openvpn in the profile? <roptat>In general services don't add anything to the profile <rvgn>roptat adding a service also adds associated packages right? For example, gnome-service adds gnome package. openvpn-client-service did not add openvpn package. How else I could use openvpn commands for the service? <roptat>Do you need more? If so, you can probably add a service extension and send a patch :) <rvgn>herd start openvpn results in error: connect: /run/user/1000/shepherd/socket: No such file or directory <rvgn>roptat network-manager should have started openvpn service for network-manager-applet-openvpn. So openvpn command should have worked in terminal. But it didn't. <roptat>Shouldn't it work if you add openvpn to your profile? <rvgn>It will work if I add. My doubt is if the openvpn-client-service didn't automatically invoke openvpn package, then declaring the service is of no use right? <roptat>Also you can't use herd as user to manage system services <civodul>roptat: i'm adding a couple more strings in the installer for user passwords <nixo_>kmicu: Hi, do you remember the (strange) problem I had yesterday? <nixo_>at the end I'm 90% sure it was the disk. I replaced it and now I'm not experiencing the random "Timeout on hotplug command" I had <nixo_>maybe it's to early to be happy but I'm pretty sure it was the disk <nixo_>(just found out that folks at nixos are running gixy (https://github.com/yandex/gixy) by default on nginx configuration, refusing to switch if there are known security vulnerabilities) <bgardner>Hello Guix; I'm running I3 on Guix System and just started to work on getting screen locking set up. I've used I3 for years on Debian with no trouble, however on Guix it appears to work exactly the same but won't unlock - it says my password is wrong. <bgardner>This is on multiple Guix machines, just to help rule out keyboards and whatnot. <mikadoZero>I have been enjoying mu4e but sending email has stopped working for me. Is anyone else having this issues with mu4e when using Guix System? <mikadoZero>When I try to send an email with mu4e in the echo area it says "Process smtpmail not running" in a trace buffer it says "Process smtpmail connection broken by remote peer". <mikadoZero>I have read through the documentation for mu4e and Emacs smtp and I think everything looks correct in my Emacs configuration and .authinfo. I have done pull, system reconfigure and ran package on a manifest. I have also rebooted the machine. <mikadoZero>I am able to send email with mutt on a machine that is not running Guix System. <mikadoZero>I have double checked the server domain, port and password and they are correct. <mikadoZero>I am stuck and would appreciate any help with this. <Blackbeard[m]>did you created any special configuration for proxy, firewall, etc? <mikadoZero>Blackbeard[m]: Yes I do have nss-certs installed. <mikadoZero>Blackbeard[m]: I have not done any configuration for proxy or firewall. <mikadoZero>Blackbeard[m]: What do you mean by "have you tried sending mail to a different account?" <mikadoZero>Blackbeard[m]: I have tried sending to my own email account and also to a Guix mailing list. I got the same error mentioned above when I tried sending to those addresses. <mikadoZero>Blackbeard[m]: I just tried sending to blackbeard@mailinator.com and I got the same error. <Blackbeard[m]>i had a similar problem to yours and it ended up being something bad in my path <mikadoZero>Blackbeard[m]: I will look at the links you shared. I will also triple check my configuration. <ArneBab>is there a way to disable building the database for manual pages? <ArneBab>(it’s the slowest part of installing a new package) <nckx>ArneBab: +1. I'd expect it to be dependent on ‘man’ in the environment but that doesn't seem to make a difference. <bgardner>Reading the "GNOME in GuixSD" I'm wondering if i3lock isn't able to communicate with PAM since I'm not running Gnome. Should I be starting a dependent process for elogind in some fashion? <roptat>civodul, in "At least one partition must have its mounting point set to '/'." don't you mean "exactly one"? <ATuin>I have installed the locales in the current profile (to avoid the locale warnings when running the daemon) but everytime i run guix pull those are removed from the profile. is that expected? is there any workaround to that besides installing them after git pull? <pkill9>i can't find it with grep and importing (guix gexp) doesn't make it available <nckx>It's defined there, but it's a syntax thing. <nckx>I guess you were grepping for ‘define.*ungexp’? <pkill9>nckx: any way i can replicate it's behaviour in my scheme module? <g_bor[m]>I've just seen something strange, I got a response on my login attempt to my Guix System machine, that it cannot create/remove entry for session, and dropping me back to login. <pkill9>my goal is to get the store path to a package, and pass that as a string to another function (the 'format' function) <nixo_>Hi bgardner, I'm using i3lock. Had the same problem but I solved it <bgardner>Answering my own question it looks like i3lock is missing an entry in /etc/pam.d, so I'll just switch to xlock (which works out of the box) <nixo_>bgardner: have you tried with (screen-locker-service i3lock-color "i3lock") <nixo_>not hundred percent sure it's this the solution, but mine now is working and I have it <joshuaBPMan>Hello, can somebody point me to a guix channel? My channel is not working. it is giving me an error no code for module jmacs. <joshuaBPMan>I'm trying to figure out what I'm doing wrong. guix package -f jmacs.scm works just fine. <pkill9>hmm i would guess ungexp would only work inside a g expression anyways <g_bor[m]>now I see why that happened, it seems that on my latest reconfigured system dbus-system can't be started. <pkill9>joshuaBPMan: where hav eyou put your module relative to the root of the channel directory? and what have you put next to "define-module"? <joshuaBPMan>my define-module starts like (define-module (gnu packages emacs jmacs) <joshuaBPMan>I'm basically defining a customized emacs with most of my emacs plugins already installed, like s, dash, yasnippet, evil. <pkill9>joshuaBPMan: change "gnu packages emacs jmacs" to "jmacs" <joshuaBPMan>really? May I ask why that works? What did I do wrong? <nckx>pkill9: Sorry, I was away. And I think you're right about ungexp. <pkill9>i think "package-output" from guix/packages.scm will do what i want <pkill9>i need to give it a store connection though <bgardner>nixo_: I had not, thanks, I'll check it out! <nixo_>bgardner: this is for i3lock-color, but you can use i3lock the same way <nixo_>(I mean by removing the -color part) <pkill9>joshuaBPMan: it just needs to mirror the directory structure <pkill9>e.g. if the module is in <channel directory>/path/to/module.scm, the part next to "define-module" needs to be "(path to module)" *g_bor[m] trying to reconfigure from a fresh pull now <pkill9>how do i get a store connection for the functions that want it? <civodul>roptat: that string looks bogus indeed :-) <civodul>it should be "mount point", not "mounting point" <ArneBab>short status message: I can finally build my paper again. Thank you for your help to get there! <ArneBab>I’m now creating a dedicated environment with a guix.scm file, but I still had to add locally installed packages and hacks. I’d say I learned something … <roptat>civodul, it's from the installer <ArneBab>(what I learned: I should avoid packages which need custom setup and so on. If there had been standard tools for this — and if python packages weren’t often so … annoyingly nonstandard … I would have saved tens of hours of work. <nckx>Got to close those brackets or the scrollback won't eval. <pkill9>if i use "package-output" in a live profile (one that won't be garbage collected), will those packages referenced by "package-output" also not be garbage collected? <dongcarl>For future reference... If you wanna produce dynamic binaries without rpaths, just set `GUIX_LD_WRAPPER_DISABLE_RPATH` <nckx>dongcarl: 😃 Welp. That was easy. <dongcarl>Well... You also gotta patch the gcc package but... That I've already done lol <nckx>Honestly, I'd feel worse if you'd spent 2 (?) days on this only to find an undocumented switch that worked out of the box.