<reepca>guix environment -C python2-flake8 -- python -c 'import configparser' succeeds, but for some reason in the build environment it fails. So strange. <joshuaBPMan>I'm really confused now. if I log into tty1, and startx. Then switch to tty2 to start weston for giggles. Then switching back to tty1, and now my x session has gone...? <joshuaBPMan>Right now, I'm trying to play with weston, 'cause reasons. And I can't get it to change the default layout to dvorak. weird. <joshuaBPMan>I have a file at ~/config/weston.ini, but it's not seeming to read my values for that. <EternalZenith>The manual says that you should use sddm rather than slim if you want to use wayland, although I'm not sure of the reasons for that <EternalZenith>Although it may be outdated, as it says that gnome is the only wayland compositor available on guix right now <joshuaBPMan>EternalZenith: True. Right now I'm just launching weston via "weston". Once I get that working, then I'll try to get sddm or gdm working <joshuaBPMan>and I was reading the log earlier, apparently I have openGL ES 3! That's pretty awesome for a 10 year old macbook! <joshuaBPMan>It's an old nvidia card. powered by nouveau, so it's slow, but it works. <EternalZenith>My laptop has the oldest generation of graphics that's supported by the new amdgpu driver <joshuaBPMan>EternalZenith: yeah, if only AMD provided the firmware, they'd be super stellar! <reepca>my desktop has the oldest generation of amd graphics to require proprietary microcode :( <EternalZenith>I'm dissapointed that Dell switched to Nvidia for all of their newest Precision laptops <joshuaBPMan>reepca: does newer AMD graphics have free and open microcode? <reepca>joshuaBPMan: nope. Mine was barely too new. <EternalZenith>I'm not sure if they require nonfree microcode, but I know the drivers themself are completely free <reepca>well, I've got a radeon hd 7870 that doesn't work with linux-libre, a radeon hd 6670 that doesn't work, and the integrated graphics on my a10-7850k that don't work. Unless something changed recently, I'd expect the newer stuff also doesn't work. <reepca>if I recall correctly several of the entries on h-node.org were wrong <EternalZenith>Although, I think all of those might be pre-GCN and are stuck on the radeon driver, so it might be different now <EternalZenith>Nvm, the 7870 is GCN, although you would need a kernel command line argument to enable amdgpu support <EternalZenith>Looks like they all require nonfree firmware, judging from the debian package <EternalZenith>emacs-guix has constantly been failing to load certain package modules from my local git checkout <joshuaBPMan>anybody know where guix stores the tiny logo of icecat? <joshuaBPMan>I'm trying to find that icon, so that I can have a firefox icon in the launcher for weston <EternalZenith>But the icons it comes with are in ./lib/browser/chrome/icons/default/ directory <EternalZenith>That's from within your current guix profile (e.g. ~/.guix-profile) <joshuaBPMan>there's also ~/.guix-profile/share/icons/hicolor/apps/icecat.png <EternalZenith>Yeah, whatever icon theme you are using might have its own icecat icons <EternalZenith>Are you kidding me... I've spent days trying to figure out how to fix the "No Sim" issue on my phone that keeps appearing, and I just finished backing everything up so I can wipe and reflash everything <joshuaBPMan>I'm having a weird issue with my mouse too. It randomly works, and then doesn't. <joshuaBPMan>weird question, does anyone know an open source alternative to good reads? <vagrantc>i had to configure tlp to not disable usb suspend, or the mouse would periodically be briefly unresponsive <joshuaBPMan>this really helped: (kernel-arguments '("modprobe.blacklist=usbkbd")) <joshuaBPMan>But sometimes it still gets weird and only allows me to move my mouse up and down. *vagrantc wonders what bus the trackpad is on <joshuaBPMan>If I don't blacklist that, then I ALWAYS have a trackpad that only moves up and down <EternalZenith>I've been meaning for a while to switch from i3 to something else <leungbk>i ran `guix environment guix`, `/.bootstrap`, `./configure --localstatedir=/var`, but `make check` seems to fail on one test: `guix-download.sh`, where i wind up downloading stuff from a bad link. <leungbk>would anyone be free to help me debug this? <leungbk>i'm having trouble with this test run during `make check`: `+ guix download http://does.not/exist` gives me a non-empty download, but all the tests from the other files are fine <leungbk>they both give me a non-empty result, so i suppose i can ignore this failing test? thanks for the suggestion <mange>I assume that's a problem with your environment. I get "wget: unable to resolve host address 'does.not’", which is what I would expect. <apteryx>how can I move /var/guix/profiles/per-user/maxim/current-guix-6-link to another machine? <rekado>apteryx: use “guix archive” to recursively export whatever that link points at. <rekado>on the other side use “guix archive --import” *rekado reconfigures to use GNOME 3.30 <apteryx>retrying with guix copy, as I couldn't see any output for a long time using guix archive <apteryx>now at least I know it has to send 25 store items (126 MiB). <apteryx>sadly, guile-ssh doesn't seem compatible with my gpg-agent serving as an ssh-agent <rekado>for guix-copy you need public key authentication. <apteryx>i'm in the middle of an install and got bit by the bug info-dir ERROR: no code for module (guix build utils) <apteryx>it's nice to have SSH live with just 'herd start ssh-daemon' :-) <apteryx>nope, it fails with gpg-agent as SSH agent (configure correctly as per Gnupg's manual): guix copy --to=root@192.168.1.102 `readlink -f /var/guix/profiles/per-user/maxim/current-guix-6-link` <apteryx>guix copy: error: SSH authentication failed for '192.168.1.102': Failed to read private key: /home/maxim/.ssh/id_rsa <apteryx>when I stop my gpg-agent and do "eval $(ssh-agent) && ssh-add", I can manage to get guix copy towork <apteryx>earlier I had done ssh-add with my gpg-agent, and that work, but guix copy failed with the earlier message <apteryx>does guix pull work in the live usb GuixSD? <apteryx>I just tried it, but 'guix pull --list-generations' listed only the last one <rekado>you can give it another profile with “-p’ <efraim>diffutils-3.7 needs perl for one of the tests <efraim>groff-minimal@1.22.4 wants texinfo <efraim>i had to add perl for a test for diffutils@3.7 also <efraim>might need to work around texinfo in groff-minimal, i'm sure it'll pull in perl which we were avoiding <civodul>does the tarball include a .info file? <efraim>looks like it wants makeinfo for the docs, trying to work around it <civodul>efraim: or rather try to see if it's a timestamp issue <civodul>like perhaps all you need is to update the mtime on the .info file so that it's not rebuilt <efraim>we already redirected the docs to /tmp/trash/docs during the build before, they made texinfo a hard depenency for the docs since 1.22.2 ***Sleep_Walker is now known as Guest46067
<roptat>can I send world-changing patches to core-updates? <apteryx>a progress bar on guix copy would be nice <efraim>roptat: yeah, core-updates is currently in free-for-all mode <apteryx>:-) I'm copying my whole profile of 6G and wondering should I go bed or is it finishing soon ;-) <apteryx>df -h on the receiving end suggests the bed might be a good idea <apteryx>has anyone tried the docker patches yet? <civodul>apteryx: it might be faster to run: "guix copy --to=host $(guix gc --derivers $(readlink -f ~/.guix-profile) | head -1)" <civodul>and then rebuild the profile on the remote end <apteryx>eh, that would just save some CPU instead of copying the hard data? <apteryx>(my understanding is that derivers are computed ricipes that are to be executed to be turned into packages) <civodul>right, that'd copy the recipe (derivation) which you would then be able to build <civodul>in practice that may be faster because substitutes and all <civodul>and substitutes are faster than ssh... <apteryx>maybe it's just late, but: sudo -E guix system reconfigure ./config.scm --> gives me: guix system: error: failed to load './config.scm': No such file or directory <apteryx>there is a config.scm file in my CWD <rekado>apteryx: does it try to load or include other files? The error might be misleading. <apteryx>I moved the directory, and there was an (include ...) in there with a hardcoded path <civodul>hmm, fencepost.gnu.org is unreachable over IPv6 <rekado>civodul: I got a mail delivery failure report for email sent to you. <civodul>could you show me the details somehow? *civodul sends patches to drop Guile 2.0 support *jonsger ^ is happy that openSUSE Tumbleweed is Guile 2.2 now :) <g_bor>rekado: you wrote that you would like to switch to a guix pull based deployment. <g_bor>I have a question regarding that. <g_bor>Is it possible to do something like guix pull-from-another-machine? <g_bor>this way we could fetch the latest form the given node, and we just had to modify a few configs, the time consuming steps would go away? <g_bor>or should I set up a substitute server providing guix, and authorize that? <rekado>g_bor: for berlin I’m using a script in maintenance.git <rekado>the script is run on the head node, which offloads builds to build nodes <g_bor>rekado: ok, I will have a look at that. <rekado>the script causes the operating system configurations for all build nodes to be instantiated first <rekado>then it uses ‘guix copy’ to copy the systems to their targets <rekado>then it reconfigures the target nodes. <rekado>so none of the build nodes have to build anything more than once. <rekado>together they build all of the systems <g_bor>rekado: yes, fine, I was thinking along similar lines. <g_bor>rekado: we were thinking about having something like this for multi-node deployments at last FOSDEM. <g_bor>Maybe we could think about builidng a framework on top of that. <rekado>isn’t there a branch (”wip-deploy“?) for that kind of thing? <g_bor>I think there was something, but have not seen active work for some time. <g_bor>It would be nice if we could define confgiuration equivalences, so that we could specify policies like: <g_bor>have 10 nodes running cassandra, and 4 nodes running nginx, two active and two backup, these should be deployed on distint physical nodes... <g_bor>and maybe generate the config from a pool information file and a list of configuration templates <rekado>you would need a mapping of hosts to configuration objects <rekado>for berlin we have a procedure ‘berlin-build-machine-os’ that returns a configuration object for a node identifier. <g_bor>rekado, yes, that is what we also managed to come up with. <g_bor>I was also thinking about a point Elco mentioned, is that getting a service up is idempotent, or should it be? <g_bor>If we treat two nodes equivalent in this setting we could get a service instance, and getting another might mean to get one on another node. <g_bor>So this opens up a bunch of yet undiscovered modes of operation. <g_bor>Last year we stopped, as we got tho the conclusion that we need a set of useful abstractions, and discovered that direction instead. <civodul>uh, there's ocaml in the closure of (gnu packages base) <civodul>rekado: i'd like to move R packages from web.scm to cran.scm or r-web.scm, WDYT? <efraim>i think we should move hunspell to aspell.scm <efraim>or hunspell and aspell to dictionaries.scm <rekado>cran is going to become very large. <rekado>but on the other hand it’s convenient to have them all in the same module. <rekado>(I’m currently adding more packages to cran.scm) <rekado>I’d like to have them all in cran.scm, actually. <civodul>eventually that'll be a problem too, but a different one :-) <rekado>I imported all of bioconductor a few days ago and have started to add all the missing CRAN dependencies to cran.scm. <rekado>the imported file is 50+k lines long. <rekado>well, that’s confusing. ‘50k+’, not 50 lines plus some constant ;) <rekado>civodul: do you want me to move the R packages from web.scm? <civodul>i was looking at the closure of (gnu packages base) and that's what got me worried <civodul>rekado: there was ocaml in there, which is almost gone except it's still pulled in via cran.scm via web.scm <rekado>I’ll prepare the patch in the next few minutes <civodul>i'd love to have good graph clustering & visualization tools <civodul>and i'm not entirely sure how much it'd help <efraim>i'd like to tag it x86_64 only, 32-bit fails with __uint128_t undefined, aarch64 is missing appropriate assembly <civodul>oops "error: package-elisp-from-package: unbound variable" <rekado>efraim: oh, sorry. Please tag it x86_64 only. <civodul>what's that package-elisp-from-package thing? <civodul>i saw the error above in cuirass' log but i can't find the procedure anywhere <efraim>civodul: its from the elisp clang pacakge <civodul>oh it was pushed seconds ago apparently <civodul>rekado: the profile hook message are really awesome <civodul>seeing them always makes me happy :-) <rekado>civodul: the move from web.scm to cran.scm is complete. <rekado>I checked the copyright lines and removed modules from web.scm that are no longer needed. <rekado>also built the modules to ensure that there are no missing references. *civodul fixes the package-elisp-from-package thing <rekado>civodul: that commit has already been reverted. <rekado>it was added with 5bc2e14add6433adbade3ccdf28c60e57334c20a and removed again in 3a3d4d9d545028cda846ae827a60249f7162d34e <jlicht>It seems that sources retrieved using `git-fetch' have permissions set to 444, is this always correct? What if I want to have some files be writable to pass the tests instead? <civodul>jlicht: you typically need to call make-file-writable <civodul>rekado: there's also r-feather in serialization.scm <jlicht>civodul: would you call this in a new phase after 'unpacking', or rather in a snippet? <rekado>civodul: I’ll move more R packages in a moment. <rekado>we have R packages in these modules: image databases cran machine-learning graph bioconductor statistics geo maths bioinformatics <rekado>civodul: do you want me to move them away from every module that’s not cran, bioconductor, or statistics? <civodul>rekado: maybe, though machine-learning for instance is probably less critical <civodul>ideally "guix graph -t module coreutils | grep ' cran'" would print nothing <rekado>okay, I’ll get to work on this now <rekado>civodul: done! No more R things in the coreutils module graph. <_tibbe>Hi, can anyone tell me how he configured pgp smart card access with pcscd? I cannot get gpg to read keys from my nitro key. <pkill9>is there somehting in emacs and/or emacs extensions that helps with writing package definitions? e.g. generating boilerplate like use-modules? <bavier>pkill9: rekado has contributed some yasnippet templates in etc/snippets <tune>anyone here play minetest? where do I put mods on guix? <tune>It seems in the past I put them in ~/.minetest/mods but I don't see them as selectable in the game <tune>and there's also ~/.guix-profile/share/minetest but I don't know if this is the right spot either <tune>>In order to get the mod running, you have to unpack it in one of the folders where Minetest looks for mods. <tune>okay, I still had them as zips. I guess I gotta extract them and possibly rename them. <tune>alright yep that worked. ~/.minetest/mods is fine <bavier>tune: I was wondering recently if it would make sense to provide some minetest mods as guix packages <tune>That could be interesting. I think the Arch User Repository has -git versions of several mods. <davexunit>as long as minetest has an env var that can be set for the mod search path <davexunit>I had to get a patch merged upstream in order to get minetest to work at all in guix <pkill9>oh yeah that would be good bavier, unless they change frequently <bavier>pkill9: yeah, I guess we'd have to see how that goes <davexunit>last I looked at minetest mods it was hard to tell which ones were well supported and worked well <davexunit>so having a few of the popular/good ones curated in guix would be helpful <noobly>last time I installed guix I made a shell script and shebanged it with #/bin/bash as per usual, but in guixsd that's not how it's done. this freaked me out so I left, but I'd like to return. is there any summary or cheat sheet of these kind fo differences in guixsd? <rain1>#!/usr/bin/env bash might work <civodul>noobly: #!/bin/sh is guaranteed to work, and /bin/sh is actually Bash <civodul>you can also configure your system to have /usr/bin/env or /bin/bash or whatever *janneke is a champion for /bin/guile ;-) <noobly>civodul: how weird, I distinctly remember some funny business with a script I was using, I thought it involved the shebang but I guess not. <noobly>any python developers here? do you need to use virtualenv or venv, or do you just roll up your own environments? <noobly>developers is a strong word, should've said users <civodul>noobly: i don't do Python, but hopefully 'guix environment' is a good substitute for virtualenv :-) <civodul>i hear there's even direnv/guix/emacs integration floating around <thomassgn>civodul: yeah! I actually set this up recently. There's a slight delay whenever moving between buffers in different direnvs. But I usually keep one emacs instance per "project" - to minimize things like this. <civodul>i haven't tried it yet but i know cbaines has done great things :-) <civodul>rekado: awesome, we need more of them!