<shcv>I just installed (reinstalled) guix 0.16 on gentoo, made a directory for hacking in, and invoked 'guix environment guix', but it produces an error when building sed: inplace-selinux.sh: set-up failure: CONFIG_HEADER not defined <shcv>ERROR testsuite/inplace-selinux.sh (exit status: 99) <shcv>it would be nice if I were running under guixsd of course, but I'm just starting the process of working toward that... <shcv>er, 'guix environment --pure guix'? <shcv>I can paste the log somewhere if you like <ardumont>hello guix, i'm trying to understand how to install guix on a nixos box and i'd like to use my existing /nix/store ('cause slow network and all that) <ardumont>my understanding from the requirements page is it's possible <ardumont>but i'm under the impression i need to install guix first then recompile a new one on top of it, am i correct? <ardumont>(well either that or i'm making a nix derivation to actually be able to install guix from source ;) <vagrantc>wouldn't using /nix/store require rebuilding all the packages? <ardumont>the install succeeded btw but the guix-daemon.service won't install because read-only system ;) <rvgn>If there is problem with ca-certs, will the browsers show ssl security warning for websites? I just updated guix and the websites I usuallly visit started showing ssl warning. <rvgn>Also, how should I change default nameservers in guix? ***jonsger1 is now known as jonsger
<sp3ncer>Hello, can anyone here help out a beginner? I'm trying to get guix installed in a VM to see what it's all about and am having some issues <sp3ncer>I'm following the manual and can't get the code at the end of step 3 to run, I keep getting permission errors <sp3ncer>I can get through the rest of the steps just fine, but can't run guix <bavier`>sp3ncer: you could successfully run the `qemu-system-x86_64 -m 1024 ...` command to boost the install image in a VM? <sp3ncer>I'm trying to install the binary in a fedora VM <sp3ncer>I'm not super familiar with qemu to be honest <bavier`>sp3ncer: the VM is running a fedora OS already? or you're running qemu on a fedora system? <sp3ncer>I have a VM running fedora, I've downloaded the guix tarball and followed the steps in the manual <bavier`>got it, ok, that's a different set of steps. <bavier`>so you say sourcing the profile gives permission errors? <sp3ncer>bash: /root/.config/guix/current/etc/profile: Permission denied <bavier`>sp3ncer: though, I think you should be executing that command as root still <sp3ncer>sudo doesn't like that GUIX_PROFILE line <j-r>sp3ncer: try source /root/.config/guix/current/etc/profiled <j-r>or . /root/.config/guix/current/etc/profile <bavier`>sp3ncer: does 'sudo -i' work to get a login shell? <sp3ncer>yep and that command from the manual works fine now <bavier`>sp3ncer: great, it's useful to have a root shell for steps 4-7 too <bavier`>sp3ncer: the locale warning is standard, but failing to connect to the daemon socket suggests the daemon isn't running (step 5) <sp3ncer>right that's what I thought too, but I did that step and just redid it to be certain <sp3ncer>so i followed the step exactly but am seeing it as failed in systemctl <sp3ncer>nope, it just says "loaded failed failed" <sp3ncer>I'm tempted to just throw out the VM and start over, clearly I did something wrong <sp3ncer>I don't have time to set up a clean VM tonight, but I'll probably be back tomorrow <bavier`>np. definitely let us know how it goes <efraim>I fixed the parentheses in my OS config and now I'm trying to build a system with my custom kernel config <civodul>a linux-libre-minimal would be welcome <efraim>I should find the place in the gentoo handbook I used again, I think I went though lsmod line by line to find only what I needed <efraim>We should be able to make a general virtualization guest kernel config no problem though ***Noctambulist is now known as Sleep_Walker
<efraim>In regards to a minimal kernel config, I found a missing module to add to the gnu/packages/linux list <civodul>so what do you mean a kernel module is missing? <efraim>kids opened kodi 8 times, had to reboot that computer, it was crypto serpent something, I wrote it down <efraim>time to see if the kids noticed GDM replaced SLiM <roptat>I installed guile on a debian, but no guild <roptat>ah it was in guile-dev, nevermind <efraim>I was going to say I have it on my PPC debian machine, but didn't think of the -dev package <civodul>guild is the CLI to Guile things, like "guild compile", "guild disassemble", etc. <civodul>that's why GDM doesn't display the keyboard switcher etc. *civodul attempts to brute-force systemd <roptat>isn't localed a separate program from systemd? or does it rely on it? <roptat>if it's difficult to build, maybe we could have our own implementation? *kmicu hopes that Ludovic releases a cracker for systemd with a nice music from demoscene. <efraim>I wonder if gentoo or devuan or the BSDs have anything they use that we can borrow <civodul>roptat: i'm trying to build just localed and localectl from systemd, pretty much like wingo did for elogind <civodul>having our own implementation would be nice, but there's a yack to shave before that: a D-Bus implementation for Guile <bgardner>Good morning Guix; I'm trying to package up BOINC and have run into an issue I don't understand. The client package build correctlym but the server package fails in different ways each time I try. Sometimes it is 'file libboinc_la-<different each time>.lo does not exist' and sometimes it is 'file libboinc_la-<different each time>.lo is not a valid libtool object'. Any advice? <civodul>bgardner: perhaps you need #:parallel-build? #f <civodul>efraim: can you see if they have a localed fork? <civodul>GDM does work without localed, but you lose some functionality <efraim>civodul: looks like we need to compare %default-extra-linux-options in gnu/packages/linux and default-initrd-modules in gnu/system/linux-initrd <efraim>I saw they have a gnome-first-start-wizard package <apteryx>bgardner: looks like a parallel build issue, yes <efraim>I found a locale based commit for gdm@2.26, so gnome2 days <efraim>i'll see if they have any patches or something <apteryx>civodul: will look into splitting my commits wrt pypi importer better, but it might take until the weekend. Thanks for the look! <bgardner>civodul, apteryx: Thanks for the tip, I'll pursue that, thanks! <civodul> i don't see anything obvious in there <apteryx>With a guix archive --export /gnu/store/xxx-some/profile and the Guix manifest that was used to generate that profile, I can recreate that exact profile on another Guix-enabled machine, right? <apteryx>I would do: guix archive --import my-archive.nar, then guix package -m my-manifest.scm -p some/profile/path, correct? <apteryx>is it possible to forego the override the key verification? In private context, it seems it adds some hassle for not much gain. <roptat>apteryx, you would need the same guix pull commit too <apteryx>oh, mesa is not reproducible on staging <apteryx>roptat: do I? What's the point of an archive if I need the same Guix? Then I don't need an archive but just the manifest... no? <apteryx>roptat: ah, no, you are right. Now I see that the archive is just a snapshot of the Guix cache. Different Guix, different store items... Hmm. <roptat>yes, if you want to reproduce the same profile, you need the same guix <roptat>the snapshot is like a substitute <apteryx>So, I'd need the archive + manifest + guix commit... Can I archive Guix itself in that archive? <apteryx>I'm looking at something reproducible in 10 years without thouching the network <roptat>I guess you would add the current-profile to that archive too? <roptat>actually, you could simply have the profile in the archive, decompress on your other guix machine, and use the store path <roptat>you don't *need* guix to rebuild a profile generation <apteryx>oh, right. And I can comment about the procedure in the manifest file itself <roptat>you can also create a guix pack which I think is closer to what you want <apteryx>Yeah, sourcing the unpacked profile should work fine, and is less work <roptat>and then you won't need to authorize any key <apteryx>right! Thanks for the ideas, I think we nailed the best solution for that use case :-) <shcv>I have just installed guix 0.16 on gentoo, and tried to run "guix environment --pure guix", but ran into that selinux issue when it was building sed <roptat>I think we have selinux related scripts in the repo... <shcv>I'm not intentionally using selinux that I know of... <roptat>shcv, it's probably enabled by your distro <roptat>but if you don't want selinux at all, you can run "setenforce 0" as root <rekado>the SElinux policy is incomplete. Someone wanted to send an audit log to help us improve it, but this hasn’t happened yet. <shcv>well, so far as I know I haven't configured or set up selinux in any way <shcv> I'm not using the selinux profile, and my distro (gentoo) doesn't really do much automatically <shcv>I suppose I could have accidentally configured my kernal to do some things... <roptat>ah nevermind, I didn't check the log <roptat>it's actually a test failure inside the build environment <roptat>and it happens to be related to selinux <shcv>that's about as much as I've figured out <shcv>I run into the same problem if I try to run 'guix pull' <civodul>efraim: Andreas need to access it physically for some file system maintenance <civodul>he was willing to email guix-sysadmin <efraim>ok, I was going to see about a VM kernel config <apteryx>is it possible to guix pack an existing profile? <bandali>hi all, anyone know where guix's icecat looks for fonts? <bandali>i see. thanks for the link Blackbeard[m] <bavier>afaict icecat also picks fonts up from fontconfig's cache <bandali>as a secondary question, is it possible to change that behaviour? <bandali>bavier: oddly enough, i don't think so ? <bandali>there's been a number of font-related bug reports recently afaict <bandali>on a foreign distro, opening icecat shows me boxes for characters <Blackbeard[m]>After installing fonts you may have to refresh the font cache to use them in applications. The same applies when applications installed via Guix do not seem to find fonts. To force rebuilding of the font cache run fc-cache -f. The fc-cache command is provided by thefontconfig package. <bandali>on my new trisquel install i don't have anything in ~/.fonts as of yet, but i'd expect icecat to pick up system fonts, which it doesn't seem to do <bandali>yup i saw that and did try that, but it didn't help <bandali>in my .profile, if i add ~/.guix-profile/share to XDG_DATA_DIRS, icecat picks up the fonts installed by guix <bandali>but i'd like icecat to use the fonts shipped by my distro <Blackbeard[m]>export XDG_DATA_DIRS="$HOME/.guix-profile/share${XDG_DATA_DIRS:+:}$XDG_DATA_DIRS"export GIO_EXTRA_MODULES="$HOME/.guix-profile/lib/gio/modules${GIO_EXTRA_MODULES:+:}$GIO_EXTRA_MODULES"``` <apteryx>bandali: as a rule of thumb, do not expect anything to be shared between host distro and guix. If you need something in guix, prefer installing it in guix <bandali>that's generally been my attitude apteryx, but i'd have prefered fonts to be an exception <bandali>Blackbeard[m]: thanks. what is GIO_EXTRA_MODULES for? <apteryx>bandali: that said, I think fonts should work, provide you refresh the fc-cache, maybe (fc-cache -vf IIRC). <bandali>apteryx: right. sadly refreshing the cache with the fc-cache from guix or from my distro didn't seem to make icecat pick up my system fonts <bandali>Blackbeard[m]: as i mentioned earlier, adjusting XDG_DATA_DIRS makes icecat pick up fonts i installed via guix, but it still didn't help with my distro fonts <bandali>hence why i asked what GIO_EXTRA_MODULES is for specifically <apteryx>when using diffoscope, to compare to --check outputs, is the number of links given by the 'stat' command useful, or noise? <bandali>trying with GIO_EXTR_MODULES, the issue still stands <bandali>Blackbeard[m]: the default gtk font on my distro is roboto, which icecat fails to pick up <bandali>i'm now trying to install it from guix and see if that helps *civodul got localed running but not solving the GDM keyboard issue... <bandali>Blackbeard[m]: thanks, installing that and doing fc-cache -fv, icecat now picks up roboto <apteryx>will this keyboard configuration still work when not using Xorg? (Wayland) <bandali>i guess the next question would be whether it makes sense to have guix set XDG_DATA_DIRS when installing icecat: <apteryx>it'll be portable for one :-). You can neatly define all your user requirements in a Guix manifest file. <apteryx>and then deploy/update using 'guix package -m your-manifest.scm' <bandali>Blackbeard[m]: i understand, however there could be side effects: if i put that in my .profile, MATE on Trisquel refuses to log in <apteryx>(for actually upgrading you'd need to run guix pull first) <bandali>so i've just written a wrapper for icecat that sets it <bandali>not much at the moment: git, nss-certs, openssl, glibc-locales, icecat, fontconfig, and a couple of fonts <bandali>separate question: is it possible to use daemon programs (e.g. dovecot) from guix on foreign distros? <Blackbeard[m]>bandali: that's strange you shouldn't have problems with mate <Blackbeard[m]>bandali: i think there shouldn't be any problem but i think those should be installed by root <bandali>Blackbeard[m]: it *is* strange.. it's an old mate version (1.12.2 i think) <apteryx>rekado: would you know if we still have cases of Python byte compiled files non-reproducibility? I think I've just found one in a new package I've prepared. Just checking if there's any known issue. <bandali>anyone know how i can have guix subcommands always use berlin.guixsd.org instead of ci.guix.info, without having to build guix itself? <bavier>bandali: you can start the guix-daemon with the --substitute-urls option <roptat>bandali, if you're on a foreign distro, have a look in /etc/systemd/system/guix-daemon.service (if your distro uses systemd) <lfam>Blackbeard[m]: What kind of trouble are you having? <lfam>In general, please be specific :) <lfam>In general, you need a very powerful computer to build Chromium. So if you are trying to build it and it's crashing, you may need more computer <lfam>How much RAM do you have? <Blackbeard[m]>I can't install it with binaries I am having the same problem as icecat <lfam>Chromium was updated on March 31, so it's possible there are no binary substitutes available yet <lfam>I don't know for sure but I'd guess that's not enough RAM. I definitely wouldn't use swap because it will be extremely slow <bandali>one more question: where can i find info and/or docs on packaging java applications? <bavier>I just got an ungoogled-chromium substitute from ci.guix.info for "guix 0e55864" from Apr 01 <lfam>Okay, in that case, Blackbeard[m] should also be able to get that substitute, assuming they are querying ci.guix.info have authorized its signing key <lfam>And using --dry-run, I see that I would also download a substitute from that server <efraim>what modules do I actually need in an initrd? <Blackbeard[m]>i want to use ungoogled-chromium for google summer of code related stuff <w96k>Hey, I just installed GuixSD on my laptop <w96k>I took lightweight desktop config <w96k>And installed the system using it <w96k>There were SLiM dm I believe <w96k>But after "guix pull" it was switched to gdm <w96k>The question is how to change dm back to SLiM <efraim>i tried my custom kernel with a custom initrd, failed hard <efraim>trying again with custom kernel and stock initrd <efraim>still failed to start basically all the services, back to the drawing board <efraim>hmm, stock 5.0.5 didn't boot to slim <lfam>I wonder if debbugs could be made to recognize when someone is mailing a patch series and automatically group the separate mails into the same bug number <lfam>Rather than creating dozens of new bugs <lfam>Or maybe we need to change our directions about submitting patches so it's less error-prone <lfam>My experience is that sending an intro email and then waiting for a bug ticket to be assigned does not feel great and may take too long <lfam>In practice I often concatenate my patch series into a single file and send that <efraim>sometimes I send a group as an attachment <bandali>y'all, are there any news about lvm support? <rekado>apteryx: … but we already *have* deterministic pycs. <rekado>apteryx: at least in the python-build-system. <rekado>apteryx: if you have a package that doesn’t use the python-build-system you’ll have to set a few environment variables. <rekado>lfam: a recent discussion on emacs-devel(?) makes me think that Debbugs can in fact be configured to behave like that. <lfam>The gitlab discussion, rekado? <lfam>I'll read it regardless, it's an interesting subject <rekado>yes, I think it was that discussion. ***ngz` is now known as ngz
<ngz>Hello. I have a question about sdl-union. In the Makefile of a package draft of mine, there is, e.g. "$(shell sdl-config --libs)", and the same with "--cflags". The package requires "sdl-image" and "sdl-mixer". So, I add "(sdl-union (list sdl sdl-image sdl-mixer))" as an input. However, it doesn't work as expected: the directory in "-I/path/...-sdl/include/SDL" is not the union, but the sdl alone. So the build phase error out wi <ngz>e.g., "cannot find SDL-image.h". I thought it was a problem due to the wrong "sdl-config" being picked up, so I substituted the full path of the sdl-union in the Makefile, but it still uses "-I/...single-sdl/include/SDL". <ngz>This is very puzzling actually. <ngz>I thought the point of `sdl-union' was to fix `sdl-config' calls. <bavier>ngz: I think sdl-union fixes pkg-config calls <ngz>According to sdl-union docstring: A union is required because <ngz>sdl-config assumes that all of the headers and libraries are in the same <apteryx>rekado: we have deterministic pycs yes, but it's still fragile, as the mtime is recorded as a timestamp it the .pyc file. With PEP-0552, we can have the pyc record a hash of the file content instead of a timestamp. When is this going to be useful? Supposed that source files get generated/manipulated at build time, this would introduce non-determinism. Not with PEP-0552. <rekado>where do you see the timestamp? I think we are in fact recording the hash, no? <apteryx>the timestamp is in the byte-code, starting at byte 8. <rekado>have you confirmed that this is still in fact being recorded for Python 3.7 pyc files generated by our build system? <bavier>ngz: can you patch the call to sdl-config to add a `--prefix` option <rekado>because switching to the hash variant was our goal back then <apteryx>I've failed to see how we enforce PEP-0552 in the python-build-system; maybe you could it to me? <ngz>bavier: Yes, I can patch the Makefile. What should the `--prefix' be? <apteryx>from what I understand the switches haven't made their way in setuptools/distutils, so we'd have to run compileall manually. <bavier>ngz: the location of the sdl-union input <ngz>bavier: Certainly. But, is it the root directory, or the include/SDL/ one, or the lib/? I assume the former. <apteryx>rekado: I understand now: it must use hashes, and what I'm seeing is something else... Because the documentation of py_compile.compile (which distutils ends up calling) says: Changed in version 3.7: The invalidation_mode parameter was added as specified in PEP 552. If the SOURCE_DATE_EPOCH environment variable is set, invalidation_mode will be forced to PycInvalidationMode.CHECKED_HASH, and we *do* set <bavier>ngz: should be the root, yes. The --libs and --cflags options append the other directories as needed <ngz>bavier: It fixes the issue and the build completes successfully. Unfortunately, I get a segmentation error when trying to execute the program. Ah well. <ngz>hmmm Valgrind tells me something like "More than 1000 errors detected. Not showing errors anymore. Go fix your program!". How rude :) <ngz>Would anyone want to test the program (a small game, quickly downloaded) and tell me if it also segfaulting for them? <apteryx>rekado: it seems to really use a timestamp: the second 32 bits word (the bit field) is set to 0. <apteryx>the time went from 'Thu Feb 9 08:12:12 1984' to 'Sat Nov 7 23:21:00 2105' :-D ***amiloradovsky1 is now known as amiloradovsky
<tune>I am trying to do some fastboot stuff with my android phone and I'm getting a permissions error that I think is due to not being in the plugdev group <tune>but the plugdev group does not even exist <tune>how much do I have to do to fix this? <tune>I get `guix system: error: supplementary group 'plugdev' of user 'brad' is undeclared <tune>if just adding plugdev to supplementary groups <apteryx>owhat you want is the adbusers group <apteryx>inside your (user-acount ... (supplementary-groups '("adbusers" ...) ... <apteryx>inside (operating-system ... (groups (cons* (user-group (system? #t) (name "adbusers")) ... %base-groups)) ... <apteryx>will show you an example. You probably want to use the android-udev-rules package as well so that you can connect to your phone using adb as your user (rather than root). <tune>oh thanks for the tip. I'll change my approach then