<kkebreau>rekado_: Really? What caused the failure? <rekado_>kkebreau:it’s unsatisfying because on the third attempt it *did* work and that has clobbered my log of the failed attempts. <rekado_>there was something about a header file, but that’s all I remember from the two seconds that I looked at the log. <rekado_>I wonder if that might be a race condition when building the thing in parallel. <rekado_>gotta look out for this on ci.guix.gnu.org <kkebreau>Possibly. I never got that error on my own machine. <rekado_>lots of builds scheduled on ci.guix.gnu.org; we don’t currently have substitutes for the new gnome. <rekado_>kkebreau: thank you very much for pushing this over the finish line! <rekado_>I’m very happy to finally see this upgrade. <rekado_>hope I can finish installing it soon. <kkebreau>It's my pleasure! The community has been very helpful the entire time. <nixo_>oriansj: wow this is wonderful, thanks! <nixo_>btw, I've just seen pipepire is available! anybody tried it out on guix? <oriansj>nixo_: I tends to be neck deep in guix's bootstrapping work to notice <oriansj>right now we are on track for a 257byte bootstrap binary for all of guix <nixo_>oriansj: oh wow you are one of the people behind the blog posts we have seen recently? <oriansj>I am the guy who single handedly wrote 3 C compilers entirely in assembly <nixo_>oriansj: wow that's fantastic (and really incredible, indeed) <oriansj>nixo_: it gets even better, I wrote that assemblers using hex linkers I wrote in hex in each architecture <nixo_>if this question is not too personal, may I ask you if you get a living out of this, or is it all spare time (in which case, I'd like to know what you do for living)? <oriansj>nixo_: I work for the state of Michigan as a computer security and reliablity engineer and I do this entirely as a hobby <oriansj>my work days are usually collecting thousands of CVEs on multiple vendors which have contracts with the state of Michigan <nixo_>That's a pity, society should find a way to pay for this kind of work. <oriansj>nixo_: nothing is stopping anyone from donating towards me <nixo_>are you on a patreon-like platform? <nixo_>I'm on keybase too (but since there's no keybase in guix I've not been using it for long, except the login to get free stellar) <oriansj>So, I don't bother with problems that don't exist yet <nixo_>oriansj: I think it worked :) thank you <nixo_>It's late here, good night Guix! <valignatev>Hey #guix! I just got a permission error during the reset-gzip-timestamp stage. I've seen that many packages have added a snippet before this stage to their package definitions in order to make gzips writable. I don't quite understand the purpose of this stage and why all those packages are doing its own thing to adhere to it? Why is there permissi <valignatev>on errors at all? I thought that guix daemon builds the packages in a sandbox where it has all the permissions <rekado>valignatev: this happens with git-fetch as the resulting files are all read-only <valignatev>My guess is that this stage has something to do with resetting tarball's state to the point it was before we started building it, but I'm not sure if I get it right <valignatev>I don't think I understand what git-fetch has to do with it since it happens on a later point of the build <valignatev>But most importantly, is there some universal elegant solution to this? SHould I adopt what other packages are doing? Basically they add a step before this stage with something likehttps://pastebin.com/j2X5ERwq ***jonsger1 is now known as jonsger
<emacsomancer>no-one is currently working on packaging mullvad's vpn client for guix by any chance, are they? <emacsomancer>nix has a package (though it's a bit out of date) that I've been using, but it would be nice to have it 'natively' in guix. <oriansj>emacsomancer: if I remember correctly we have an importer for nix and guix refresh can be used to update the package definition (after we have it imported) <emacsomancer>oriansj: does the importer from nix to guix generally work well? I've never tried it. The only guix packages I've done have been 'from scratch'. <oriansj>emacsomancer: honestly, I haven't needed to use it myself so I don't know <dftxbs3e>emacsomancer, doesnt it require Electron? <emacsomancer>oh, well. perhaps I should just try to get the nix people to update their version at least. <alextee[m]>it basically downloads a dependency from the internet <alextee[m]>i usually embed some libraries in my code if many distros don't have them, and im thinking to just use this instead, but i think packagers will hate it <wdkrnls>Hi Guix, I'm trying to figure out how to control the backlight on my laptop. Instructions on the internet suggest this is a rather involved process which entails making a symlink from my /sys/devices/pci0000:00/... to /sys/class/brightness, adding a section to my Xorg configuration, and possibly modifying my GRUB CMD_LINE. <wdkrnls>I was curious if anyone had run through this process and might have figured out the configuration under guix. <wdkrnls>Well... an easier alternative appears to be to not use that package. <wdkrnls>The light package works out of box (with sudo) ***janneke_ is now known as janneke
<grillon>Sorry to ask that but the best way to keep environement variables as PATH after installing or updating (I have just installed a fresh system) <valignatev>Hello-hello! I'm continuing to make a package definition that installs a package from master from the latest commit every time. I did it by replacing package (source (origin ...)) to (source (git-checkout (url ""))). But I still want to apply patches from the original package definition. I can obtain a list of patches with (origin-patches), but how <valignatev>I got confused about it since (git-checkout) replaces (origin) and doesn't have "patches" field. <nixo_>Hi guix, I'm upgrading enchive and the linter complains about (github) autogenerated tarball. Should I change url-fetch with git-fetch? <grillon>I mean do I have to manually update my current .profile every single time I install or update something? <snape>you can add GUIX_PROFILE=~/.guix-profile source ~/.guix-profile/etc/profile in your ~/.profile <snape>~/.guix-profile/etc/profile will be updated when you install something <snape>s/guix-profile source/guix-profile; source/ <snape>valignatev: guix packages are supposed to be deterministic <snape>which means, that if you run 'guix build something' twice, it should return twice the same thing <valignatev>I know, I know. I know that I'm opting out of determinism. I'm doing it intentionally :) <snape>oh right ;) in that case... you can change the sha256 everytime you want to update it, and it should try to download the package again, and then you update with the right sha256 <snape>or... you can use url-fetch with "file:///something" as a uri <snape>which doesn't check the sha256 <snape>and everytime you change your local repository, guix build will build a new thing <valignatev>It's all working, and I'm able to install a package from my definition already. What I'm trying to do is to apply patches that are specified in the parent package that I'm inheriting <snape>oh I see, I hadn't understood sorry <valignatev>I'm trying to add patches from the base Emacs definition <valignatev>Hmm, I wonder if I can do it somewhere in the build phases <snape>you can use (origin-patches (package-source emacs)) <snape>valignatev: see 'static-bash-for-bootstrap', in make-bootstrap.scm <valignatev>Yeah, I know about (origin-patches), but I don't know how to embed it to the pakcage source produced by (git-checkout) <valignatev>In all examples that I've seen (origin-patches) applied inside of (origin). And (git-checkout) replaces origin with its return value <snape>it doesn't matter how the 'origin' is produced <snape>whether it's a git-checkout or a url-fetch <valignatev>If I understand correctly, git-checkout returns not origin, but source directly <snape>git checkout is a method to provide the <origin> data structure with a directory that containes code <snape>that <origin> data structure, also has a 'patches' field, so that later they can be applied <snape>to the directory returned by the 'method' (in this case git-checkout, but it could be something else) <snape>valignatev: actually you need to use git-fetch <valignatev>snape: Can I use git-fetch to not fetch specific commit, but the latest one every time? I picked up git-checkout because it can accept #f as a commit argument and just checkout latest thing. If git-fetch can do this as well then I should be all set <snape>I see... I didn't know about git-checkout actually <snape>I initially thought you were talking about git-fetch <snape>valignatev: what about using "master" as revision? <grillon>it's nice here but I have to go see you later <raghav-gururajan>Folks! How do I refresh the fonts? The fonts inside gtk applications became so weird after the gnome 3.32 update. <valignatev>Damn, once you've mentioned it, it sounds obvious. If under the hood it calls git checkout then "master" should work <roptat>snape: it won't, because you need to specify the sha256 of what you download <roptat>I mean master will work, but fail to verify the content <snape>well, of course you'll need to update the sha256 <snape>hm I didn't know you can avoid that with git-checkout <valignatev>It's programmatic version of a --with-git-url cli transformation option <valignatev>But what about applying some patches from the base package that I inherit? <snape>my understanding (now) is that there are two compilers: one for <origin> (in guix/packages.scm) and one for <git-checkout> (in guix/git.scm). The former handles a 'patches' field, but the latter doesn't handle patches at all <snape>valignatev: so what you could do is define a new compiler <snape>that would work like the one in git.scm, and that would apply patches as well <valignatev>I'm a total scheme noob. Like really, I started to learn it only about two days ago. It's magic how I can already read and write some stuff, but I don't think I can pull off a new compiler (or extend git-checkout) <valignatev>Maybe I could try to extend git-checkout to be honest <snape>it's the opportunity to learn Scheme :-) <snape>honestly it's very pleasant to hack on Scheme <snape>valignatev: do you use Emacs ? <valignatev>yeah, I use Emacs and I hack on elisp a little bit, but not so much that I can speak it fluently <valignatev>Honestly, two days of dive into guix taught me more lisp than three years of using Emacs <snape>you can start it with 'cd src/guix/; guix environment guix -- ./pre-inst-env emacs' <snape>and then, 'use ,(whatever module from guix) <snape>and you can C-M-x as you would do with Emacs Lisp *snape just succeeded building a package from an authenticated ssh git repository <nixo_>In order to be notified for updates in packages I want to maintain, I wrote this script https://paste.debian.net/1120089/ that I run inside mcron. However, It seems to be using an outdated store.. any pointer on why that's happening? ***lurch_ is now known as optima
<nixo_>Ok found out, sorry for the noise. Btw, the script is simple but maybe can be useful to somebody <raghav-gururajan>Folks! After gnome 3.32 update, fonts in gtk applications became weired. Mainly terminal. How do I reset/refresh the font cache? <valignatev> in under root, but on my non-privileged account I still get it. My $GUIX_LOCPATH is pointing towards $HOME/.guix-profile/lib/locale <smithras>raghav-gururajan: flushing the font cache can be done with `fc-cache -f`, you can read the fc-cache man page for more info <civodul>valignatev: you need to make sure you have glibc-utf8-locale or glibc-locales installed <civodul>and that you're asking for a valid locale <raghav-gururajan>smithras Thanks! I tried `fc-cache -rv` but it did not work. Have to try `fc-cache -f`. <valignatev>it warns about en_US.utf8 when I'm trying to install anything, and I'm sure glibc-utf8-locales has it <civodul>valignatev: note that you also need to have it installed/setup in guix-daemon.service <civodul>i think this wasn't done correctly in 1.0.1 <raghav-gururajan>simthras `fc-cache -f` also did not work. May be the issue is with something else. <smithras>raghav-gururajan: oh if '-r' didn't work then -f probably won't either <smithras>hmm not sure, that's a bit out of my depth :( <valignatev>civodul: Yes, I have this in my systemd unit. It has Environment='GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale' LC_ALL=en_US.utf8 in it. This directory exists and has en_US locales <valignatev>I don't have this warning when I log in under root directly, only when I'm under my non-privileged account <roptat>valignatev, do you have glibc-locales installed in your user profile? <roptat>actually what happens if you run with GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale as a user? <roptat>it could also be that the guix you're running uses a version of the glibc that doesn't match the version of glibc-utf8-locales <roptat>since you don't have a warning as root, the daemon is running the same glibc as the one in root's glibc-locales, which is good, so maybe your glibc-utf8-locales is older than the glibc used by guix? have you tried updating it? <valignatev>Running with GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale as a user fixes the warning <valignatev>Does it mean that the daemon fails to obtain environment from systemd unit somehow? <roptat>there are two processes: guix and the daemon. They both need locale data, so they could both print that warning <roptat>only your guix (running as your user) shows that warning, while the daemon is properly configured <roptat>the daemon knows where to look for locales, but for some reason your guix doesn't (either it's set incorrectly, the locale is not present, or too old) <roptat>maybe try to run "ls $GUIX_LOCPATH" <valignatev>it shows $HOME/.guix-profile/lib/locale which is a symlink to the locale dir in the store <roptat>there should be one or more directories, and one of them should be named 2.29 (your glibc version, may be different if you're using an older guix) <roptat>there shouldn't be a warning at all... <roptat>I need to go, I'll be back later :) <valignatev>Sure, I'll try to research more myself, you gave me a lot of insight that should help, thanks! <dftxbs3e>Any idea how to solve it? I'm really inexperienced with GNU Guix. <rekado>I have a “guix deploy” problem. The remote repl is started (with sudo) and “guix deploy” says that it would send 2M worth of files over to the target, but then nothing happens. <Parra>Hey guys I posted a bug recently, and I found the solution too, I would like to get some help in order to know how to properly patch it and submit the patch. <Parra>First of all I would like to see the Ruby distribution in master, how can I find it? <Parra>Second, I would like to add a flag to the configure step, so the bug I found goes away. <Parra>this is the bug, the patch is relatively easy <Parra>it would be nice if gnu has something similar to GitHub, it's more friendly XD <dftxbs3e>Parra, unfortunately Github is not Free Software <Parra>yep, it's a sad story but a good UX is not incompatible with Free Software <Parra>I will try to submit the patch later on <dftxbs3e>Someone would have to make that :| -- UI/UX developers seems to prefer non-free software <dftxbs3e>(arguments '(#:configure-flags '("--enable-shared"))) <dftxbs3e>There's already test-target arguments and phases ***ng0_ is now known as ng0
<Parra>do you know where I can find ruby.scm in guix master? <dftxbs3e>So you can append on the next line after test-target: #:configure-flags '("--enable-shared") <zig>what pseudo file contains the available free memory? <sylvainc>Bonjour, je cherche un logiciel. Comme thunderbird auriez-vous des suggestions ? <alextee[m]>evolution? (if my french is good enough ur looking for alternatives to thunderbird) <alextee[m]>sylvainc: you might want to install evolution-server too in guix, i think it won't work without it <sylvainc>alextee[m] Thank you I will install both, in this case <hulten>Hello, I'm on a Guix System and a portscan from a different computer shows that tcp port 16992 is open. <hulten>I checked with netstat, but it does not show up. Does someone know what opens this port? <hulten>And why does netstat not show it? <dftxbs3e>Intel AMT is a Remote Management Service for remotely provisioning, rebooting, .. computers <dftxbs3e>Some people call it a backdoor, which it can be <dftxbs3e>You can probably disable Intel AMT in your system's BIOS <dftxbs3e>You can try to access: http://<your_ip>:16992/logon.htm ***apteryx_ is now known as apteryx
<sylvain>Re Bonjour, j'ai installer evolution mais je dois le lancer avec le terminal. Comment le faire apparaitre sous gnome ? <roptat>sylvain, je n'utilise pas gnome, mais il doit y avoir moyen de lui signaler qu'il y a de nouvelles applications. Sinon, un redémarrage devrait régler le problème je pense <dftxbs3e>roptat, normalement c'est pas necessaire, je pense plus a un oubli de creer un fichier .desktop dans le dossier des applications <roptat>possible, mais à mon avis gnome ne sais pas trouver tout seul un changement. Maintenant j'ai pas evolution non plus, alors je ne sais pas s'il y a un .desktop ou pas <dftxbs3e>roptat, si, j'utilise GNOME sur ma machine de travail et il va lister les fichiers pour trouver de nouveaux .desktop a chaque fois qu'un menu qui regroupe les applications s'ouvre <dftxbs3e>Au final, c'est pas tres gourmant de proceder ainsi <dftxbs3e>Sur une configuration classique de Linux, les metadonnees du disque sont misent en cache en memoire vive, donc tres peu couteux <roptat>ok, il me semblait qu'il y avait une histoire de message à envoyer sur dbus pour qu'il recharge les menus... on verra bien ce qu'en dit sylvain :) <roptat>ah bah voilà, il y avait bien un .desktop <dftxbs3e>J'avais deja fait des tests et j'ai jamais eu besoin de redemarrer pourtant! :-S <dftxbs3e>Peut etre GNU Guix n'envoit pas ce message dbus alors? <sylvain>Evolution est bien dans mon menu gnome ;). J'aurais un soucis vous voulez que je fasse des test ? <roptat>non tout va bien, mais ça aurait été plus sympa pour toi si ça avait été automatique :) <hulten>Thank you for the information, dftxbs3e. <hulten>I found out that it the port is not open, but it seems to be. <hulten>When I nmap on the host itself, it does not appear to be open. <hulten>When I nmap from a different machine (same vlan) it *is* reported as open. <dftxbs3e>hulten, yes it's because Intel AMT hooks into your network card here. <dftxbs3e>Intel AMT basically is a middle man between your operating system and the NIC, it hides some traffic from the operating system and handles it transparently to the operating system <hulten>Since localhost does not go through my NIC, I tried nmap'ing the IP-address associated with the NIC. <hulten>But then the port still shows *closed*. <dftxbs3e>All your network traffic goes *through* Intel AMT then your OS, it's kind of scary. <dftxbs3e>hulten, maybe Intel AMT filters by source IP, but your kernel may otherwise avoid going through the NIC when not necessary during routing! <hulten>But this can be turned off in the BIOS, and then -- after checking with an external portscan -- I'm certain this does not happen anymore? <dftxbs3e>Yes, that should be it. Some BIOS do not have the option to disable it. <hulten>Yes, that seems like a plausible explanation. <hulten>I will try this. I don't like this rerouting stuff. <dftxbs3e>To avoid Intel AMT, you must avoid chips with 'vPro (tm)' <dftxbs3e>But otherwise, I personally avoid x86 entirely. I have a RaptorCS Talos II OpenPOWER machine! <hulten>Do you have good experience with that machine with GuixSD, desktop / server? <hulten>I am thinking about buying that machine. <dftxbs3e>the HAP (High Assurance Program) bit is a feature created by Intel for supposedly the NSA so they can disable Intel ME/AMT entirely <hulten>Should most GNU software should just work on a POWER9? <dftxbs3e>Someone found out and it turns out you can also activate it for your own consumer needs <hulten>Only that last fact (activate for own need) is fun! <dftxbs3e>hulten, POWER9 is well supported across the Free Software eco-system, currently GNU Guix is a maturing distro and doesnt support POWER9, along with few other contributors, I am trying to get it to work. Also, the GNU Guix team is interested in purchasing OpenPOWER hardware for their CI build farm, they currently wait for an OpenPOWER VM to experiment on <dftxbs3e>I use my OpenPOWER machine since a year now as a main workstation, I run Fedora. <dftxbs3e>Why Fedora? Because Red Hat provides official support hand-in-hand with IBM, so Fedora is where all the software support comes first, and it works great! <puoxond>Does anyone here use AUCTeX (emacs-auctex) installed with Guix? And if so, what happens if you do `M-: (require 'url)'? <sylvain>Je cherche le dossier /usr/local/bin vous savez ou il est situe sur guix ? C'est pour mettre a jour youtube-dl <roptat>youtube-dl fournit par Guix n'est pas à jour ? <sylvain>Quand j essaye de dl une video j ai une erreur js generalement je telecharger les fichier directement sur youtube-dl.org <sylvain>Je viens de lancer un guix upgrade, cela doit venir de la <roptat>alors je suis en train de cloner le dépôt pour mettre à jour vers la dernière version (je crois qu'on a la 2019.11.22, la dernière est la 2019.11.28) <jonsger>raghav-gururajan: can confirm that fonts looks broken with gnome-3.32 <jonsger>the with between letters is very huge <sylvain>On doit faire les mise a jour par un git clone ? <hulten>dftxbs3e: Besides POWER9, there is RISC-V. Would you consider these archs as competitors? <hulten>I couldn't name concrete hardware, so maybe there's the crux. <sylvain>roptat je fais sa juste apres les mise a jour c'est un peux long :/ <roptat>tu peux aussi annuler les mises à jour et revenir dessus plus tard <roptat>tu as des erreurs sur toutes les vidéos, ou juste certaines (de certains sites) ? tu peux partager une URL que je pourrais tester ? <sylvain>.youtube-dl-real: error: no such option: --with-source <roptat>oui, c'est pas une option pour youtube-dl :) <roptat>rajoutee "guix install" devant :p <roptat>cette nouvelle version de youtube-dl fonctionne ? <dftxbs3e>hulten, for now, no. POWER9 is server grade. RISC-V is mainly mobile/microcontroller grade now. <dftxbs3e>I bet on POWER9 because it has already it's High Performance Computing market, and it fits my workstation use case. With RISC-V, I would have to wait for some time until I can do the same thing. <civodul>these things are always tricky to track and address <sylvain>J' ai une petite erreur avec mon locale 'guile: warning: failed to install locale' un dpkg-reconfigure locales marcherais ? <civodul>rekado: did you try stracing the remote "guix repl" process? :-) <jonsger>raghav-gururajan: just set a custom font in your GNOME Terminal profile and this font issue will go away :) <dftxbs3e>civodul, I assume that patch is applied onto core-updates? I have a pristine core-updates and it still produces an error <dftxbs3e>The error is the same but it's at a different place, I think. <sylvain>Bonsoir si j'installer le reseau tor sur guix est t'il en noeud de sortis dans la cfg ? <sylvain>Je n'es pas envie d'etre responsable des autres utilisteur du reseau, c'est pour cette raison que je vous pose la question. Je veux etre sur de ne pas etre un noeud de sortis. <dftxbs3e>is there a way to install GNU Guix without compiling it at all? (Leaving everything as .scm) <dftxbs3e>leoprikler, the time it takes to compile isnt worth it to me <dftxbs3e>I see in Makefile.am we can use MODULES_NOT_COMPILED ***cynic is now known as __c
<rekado>dftxbs3e: you shouldn’t need to compile it all again and again, so the cost is very low <roptat>dftxbs3e, everytime you use guix, you would interpret almost all of it, so I don't think it would be worth it <dftxbs3e>roptat, I use GNU Guix in CI and I recompile things a lot, most of the compute is used by compilers, not GNU Guix. <dftxbs3e>caching is always mostly suboptimal in CI, it's unfortunate <leoprikler>what's your CI pipeline roughly? clone guix, build and then other stuff? <dftxbs3e>leoprikler, download all of GNU Guix deps, compile and install them, clone guix, build guix, build docker image from installed GNU Guix, pull docker image, build 'GNU Hello' with GNU Guix <leoprikler>that sounds like a lot of work for something that could likely be queried from ci.guix.gnu.org <dftxbs3e>leoprikler, no, because CI doesnt build for powerpc64le-linux-gnu <leoprikler>Theoretically, it should be possible to spawn your own Cuirass instance, that builds on ppc64le. <dftxbs3e>leoprikler, is that trivial? does it allow me to push to a repo and Cuirass automatically rebuilds? <dftxbs3e>currently, GNU Guix doesnt fully work on powerpc64le-linux-gnu <leoprikler>It would probably be trivial if Guix was fully functional on ppc64le (refer to the manual, you would only have to swap out the systems arg). <dftxbs3e>GNU Guix itself is functional, just GNU Hello wont build <leoprikler>Given your rather unique situation it's a bit hard to tell, but it might work for GNU hello if you can get it to run. <leoprikler>Oh, that can happen on standard Cuirass too, just not likely for GNU Hello. <civodul>dftxbs3e: the patch should even be on master, no? it's pretty old <dftxbs3e>civodul, I would expect that. So it's a similar issue, but somewhere else.. <leoprikler>But if the build itself fails, you should receive a failed status and be able to inspect the build log from Cuirass. <dftxbs3e>civodul, I'm improving my CI so it's faster, then I can work comfortably on this. <civodul>yeah, sorry that i don't have more specific advice! <dftxbs3e>civodul, it's been so long I've wanted to get GNU Guix on powerpc64le-linux-gnu running.. I feel like you guys would've done this already since forever, it's just you don't have the hardware like me, and I'm inexperienced with both Scheme and GNU Guix! <pkill9>could guix work in postmarketos? <dftxbs3e>GNU Guix uses glibc heavily and postmarketos uses alpine which itself uses musl, correct? <jonsger>wonders how to run etc/ident-code.el on Guix system, because it complains about /usr/bin/emacs :( <roptat>it could as long as postmarketos uses a Linux kernel <roptat>we can run Guix on android for instance, which uses a different libc too <roptat>but I guess if you have a musl package, you could try --with-inputs=glibc=musl or something <roptat>not sure if that would override implicit inputs... <dftxbs3e>GNU Guix is attractive because it makes shipping standalone packages easier <pkill9>postmarketos uses the mainline kernel, so yes guix should work if it just needs the kernel <dftxbs3e>I would've preferred if it used static binaries instead <pkill9>my phone can apparently run postmarketos, and apparently it can be run off the sdcard without overwriting android, so if i can get it working i will test guix :D <jonsger>do others see this when running emacs on current master: ".../emacs: error while loading shared libraries /gnu.../*-libxmp-3.5.12/lib/libXpm.so.4: file too short" <nixo_>jonsger: pulled 5 minutes ago, I don't have this error. Maybe the disk? <civodul>hey roptat: looks like your overdrive is unreachable, could you have a look? <roptat>civodul, I don't have an overdrive anymore :p <nixo_>But recently with emacs I have another problem, void function TeX-add-style-hook <roptat>it's at Pierre Neidhart's home, I don't know his pseudo here <civodul>roptat: BTW, how was your JRES talk? <roptat>I had a lot of positive feedback :) <roptat>I sent an email about it on thursday I think <civodul>roptat: one of my office mate was at JRES, i'll ask them if they attended the one talk that mattered ;-) <nixo_>Should I just remove it for now? <nixo_>or you do have a workaround? :) <Guest76514>I've got a quick question about mounted filesystems in guixSD <Guest76514>I see that /etc/fstab is regenerated based on /etc/config.scm <Guest76514>but I can't seem to set the device key to something like "//server/share" <Guest76514>Reading the doc it seems like it might not be supported, but I wanted to ask and see if there was something I was missing <nixo_>puoxond: yess! Thanks that worked. Recent EMACSLOADPATH changes created me a decent amount of problems unfortunately <apteryx>nixo_: yeah, I agree it hasn't been a smooth transition, and it seems the first problem related to removing guix.d has cropped up. Will have a look. <jonsger>enough guix for today. I need to somehow improve my dev setup, it's not that productive... <dftxbs3e>jonsger, same here. What were you developing? <jonsger>dftxbs3e: yeah, on my laptop. It's a pain. But power is not that motivating even... <jonsger>dftxbs3e: trying to get Guix work on powerpc64le :P <civodul>at the R-B summit several people were positive about POWER9 <dftxbs3e>jonsger, I'll try my best but I think we're close. <dftxbs3e>oh okay, GNU Guile's interpreter is THAT slow! :S <snape>is there an easy way to disable sha256 check? <snape>(other than rewriting the origin compiler I mean) <snape>also civodul, could you please point me to the code that forbid networking when sha256 is #f? <dftxbs3e>leoprikler, I don't think that's their question <rekado>civodul: the problem (as always when I tried “guix copy” or “guix deploy” in the past) was that /etc/environment lacks entries for GUILE_* env variables <rekado>I wonder how this works without them for other people <snape>nope, otherwise you couldn't download sources <lfam>Networking is allowed when building "fixed-output" derivations. Their output is "fixed" in the sense that we know what we are expecting, described with the hash <dftxbs3e>how exactly is networking allowed/disallowed here? <lfam>Not sure, I'm not that familiar with the internals <lfam>I would grep 'fixed-output' in the source code <rekado>well, it got me part of the way to completion; aborted with "unknown error while sending files over SSH" in the end (when sending one item after successfully sending 18 items before that) <civodul>those GUILE_ vars are needed not for "guix repl", but for the tiny snippets run via "guile -c" <civodul>a simple workaround is to have Guile in the global profile <snape>What I'd like to have is a 'proc' that takes as input (from Cuirass) a package + a revision, and gives (to Cuirass) as output a derivation of that package whose source is changed to that revision <snape>but it's currently not possible because if I set sha256 to #f, the download doesn't happen, and if I can't guess the sha256 in advance <snape>(I would have to download the source twice, which isn't good) <civodul>not sure i understand the use case: you'd like Cuirass to test not-yet-packaged releases? <civodul>as if you had run "guix refresh -u"? <snape>I want Cuirass to be the CI of Guix packages <snape>for example say I'm the developer of 'hello' <civodul>we sould use "git-checkout" for that <snape>I want Cuirass to check the git repository of 'hello', and when there's a new commit, it would send its revision to the 'proc' which would return a derivation of the new hello <snape>ok, well, git-checkout is specific to git, so I guess I have to rewrite it for other kind of sources <civodul>yeah, it's also not ideal because Cuirass would be unaware of it <civodul>you'd really need a Cuirass input if you want it to be able to trigger builds, keep track of commits, etc. <civodul>but since you can have several inputs per job in Cuirass, that should already be workable, no? <snape>yes, but in my case, the input (one git repository) differs from the source (the git repository pulls other repositories, like submodules). And I have a Guix package for that, so all I need is change the source uri <snape>and Guix has everything I need <snape>except that it doesn't want to download stuff because of that sha256 check <snape>basically the problem (in less words) is that Cuirass inputs can only be git repositories <snape>while they are good enough to know there is a change, they aren't good enough to build the package <snape>I'd like to use the Guix origin for this, because the Guix origin knows how to fetch all the sources <civodul>snape: that's ok, your job can just do (package (inherit whatever) (source (assoc-ref cuirass-build-inputs 'thing))) <snape>no, because there are other things to pull that aren't in that input <civodul>that is, Hydra is responsible for fetching stuff, and then you insert it in your job <snape>civodul: my Cuirass input is a git repositories that track the state of other git repositories. That first git repository is enough to know there is a change, but I'd need to download other git repositories to build the source <snape>akin Google repo, or Zephyr west, or submodules <snape>my origin method isn't git-fetch, it's west-fetch, which does more stuff <snape>if I follow you I'd have to re-implement each origin method for Cuirass <snape>which is pure code duplication <snape>instead we could use Guix downloading process <snape>the problem with git-checkout is that it's somehow a code duplication <snape>if I want the same for svn-fetch, I need to write (40loc) svn-checkout <snape>which is sad because there's everything I need in svn-fetch <snape>except that very tiny bit of a <snape>of sha256 checking that I really don't want <snape>because the prevents Cuirass automation <rekado>I’d like to add a form to mumi to let visitors respond to bug reports and perhaps even submit a new one. No authentication, no captcha, just hidden input field and timing checks as abuse prevention. <snape>my point is that Hydra/Cuirass should not be responsible for fetching. It should be Guix origin's job <rekado>I wonder if we can use fencepost as a relay for sending messages to debbugs. <rekado>(or if we’d have to set up our own mail relay on bayfront) <dftxbs3e>gitlab is currently getting spammed on issues <rekado>well, currently people can spam the bug tracker *directly* if they wanted to <rekado>using a browser to submit a form to send an email sounds a little too inefficient.