***bran is now known as Bran
<PotentialUser-94>I'm having trouble installing Guix system 1.1.0 I downloaded the iso file and verified the autenticity of the file. burned it to usb stick and got selected boot device failed error. Since usb didn't work, I burned it on a cd and get the same error. What am I doing wring? Thank you for your help in advance <marusich>Where did the error come from? Depending on your BIOS (or UEFI), you might have to adjust some settings in your BIOS (or UEFI) to be able to boot that image. <marusich>The possibilities I can think of are: the file you used is corrupt, or perhaps the BIOS/UEFI isn't configured in a way that lets it boot what's in the image. <marusich>For example, on a more recent machine using UEFI, I had to change some settings before I could convince it to let me boot from a USB stick. <PotentialUser-94>I figured, I selected uefi boot as opposed to legacy boot and it worked. ***catonano_ is now known as catonano
<PotentialUser-94>I could't finish installing Guix system. I'm using graphical interface for installation, it asks me please select a disk (my hdd and dvd+ appears on the list) I select the hdd. and on the next screen I select everything in one partition option. Once I click ok, it start over. What am I doing wrong? <PotentialUser-94>I didn't select a desktop environment assuming Guix system comes with it's own desktop environment, and the ones on the list are additional ones. Might that be the reason? <PotentialUser-94>Another question not directly related to the problem I'm having, should I select openssh? <kelsoo1>What about upstream systemd'd applications? does guix handle them. ignore, fork, replace, other? <kelsoo1>What about upstream systemd'd applications? How does guix handle them. ignore, fork, replace, other? <marusich>PotentialUser-94, my guess is you probably aren't doing anything wrong, and the GUI installer has a bug. To work around the issue, it sucks but I suggest starting over and being careful to only traverse the "happy path" of the GUI installer, by not making mistakes as you go. <marusich>For example, if you do something, and then you think, "huh I wish I had selected something else," so you go back to some previous section of the GUI installer and change it, it *should* be no problem at all, but in practice it might be buggy, I guess. <marusich>At the same time, although I'm sure you're probably tired from just going through the installation routine, if you know what the exact steps were which produced the looping behavior, you might consider sending a bug report to bug-guix@gnu.org. <marusich>I haven't used the GUI installer for a long time, but I recall it having such bugs. <PotentialUser-94>I mostly selected recommended options. My only guess it failed because it didn't like me not choosing a desktop envitonment or my hdd is not compatible or something. <marusich>I mean, it shouldn't leave you in a dead end state, regardless. Does it not print any error messages at all? <PotentialUser-94>No, nothing, it just goes back to the beginning where it asks about the locale keyboard language and such <marusich>That sucks. What happens if you try starting over fresh? Without an indication as to why it's looping, I really don't have anything better to suggest. <PotentialUser-94>It just loops. I don't know why. I wish I had an error code at least. Spent some time to decide on a linux distro now it doesn't work -,- <guixy>lafrenierejm, are you making sure (use-modules (shepherd service)) is there? <guixy>I looked in one of the shepherd services left in my /gnu/store and searched for "use-modules" <guixy>Since goops is the object-oriented part of guile, I figured shepherd might be using it. <lafrenierejm>Why did you think shepherd would be using goops, though? Just due to the expected complexity of shepherd? <guixy>The manual refers to service objects. <guixy>* Slots of services:: What a <service> object consists of. <lafrenierejm>Ah. That makes sense. Thank you for explaining that insight. <lafrenierejm>Separate question: What is the proper way to start user-level shepherd services? Just invoking `shepherd` in a shell concludes with "Exiting shepherd...". <guixy>I got it going a year ago, but had to reinstall. <guixy>Make sure shepherd is installed to a user profile <guixy>Try shepherd -c FILE where FILE is where you define all your services. <guixy>Process 1 in guix is essentially /gnu/store/...-guile-3.0.2/bin/guile --no-auto-compile /gnu/store/...-shepherd-0.8.1/bin/shepherd --config /gnu/store/...shepherd.conf so that should work <guixy>If you don't have anything that launches a login script, you can always put that in your .bash_profile <guixy>And if you sometimes log out and log back in, you should call herd stop root in .bash_logout <guixy>But if you use virtual terminals at all, you need to do something smarter. I'm still trying to figure it out. <guixy>'users | xargs -n1 | grep $USER | wc -l' prints the number of times you are logged in. When it is 1, launch (or stop) shepherd (after you source your guix profile). <guixy>lol, That should work for any user other than root. root can't have a personal shepherd. <guixy>Or rather, root's personal shepherd is process 1 on guix <lafrenierejm>Hm. Even with shepherd installed to my user profile, only having a single login, and specifying the full path to the init file, shepherd still prints that its exiting and there's no surviving process for the service (syncthing). <guixy>hmmm... sounds like that shouldn't happen. <guixy>I'm going to try it on my end. If it doesn't work, we'll know there's a bug. <PotentialUser-94>I reinitiated the installation but it didn't work. I choose a desktop environment thinking that maybe that might be the reason but the same problem happened. After choosing partition thing, it just goes back to the beginning. <guixy>Check to see if herd is running. `herd status` <guixy>If it is running, it lists the services. Else, it says it is not running. <guixy>The documentation says if you want it in interactive mode, use --socket=none ***sneek_ is now known as sneek
<peanutbutterandc>so, I'm playing around with a project whose ./configure phase asks questions at the interactive prompt.... how do I deal with this in guix? <brendyyn>peanutbutterandc: can you show the script? if the answers to the questions are all 'y' then you can use the yes command <peanutbutterandc>brendyyn, Oh hi there! Almost, yes. How do I do it, please? I need to `yes n` actually, I think. <brendyyn>there is probably already an example in the code <peanutbutterandc>brendyyn, It is the output of `guix import cpan Alien::wxWidgets` that I'm trying to fix up.... <peanutbutterandc>It's actually another package that uses that package as a dependency.... <brendyyn>probably something like (replace 'configure (lambda _ (system "yes n | ./configure") #t)) <efraim>I'd also check debian and see how they worked around it <efraim>I've actually been coming across more of those recently... <peanutbutterandc>Oh... sorry, it seems to be a chordpro issue and not really a perl-alien-wxwidgets issue. Sorry for the miscommunication <efraim>sciencey packages are hit-or-miss, they often have terrible build systems <efraim>and they're still reworking their rust plans so they have a backlog of hundreds of crates while they work it out <efraim>guixers also seem to be more interested in making music than debianites <peanutbutterandc>haha :D Guix is just so much fun (when things are going right, that is :D). It has put the fun back in computing for me. (: <peanutbutterandc>For me the best thing about guix has been channels. I can package and ship my own stuffs. Across distros. With just a simple public git repo. Awesome stuff. <peanutbutterandc>there must be something wrong with this package's configure script. I'll look it up' <peanutbutterandc>can somebody tell me how I can start a guix environment with just the basic perl-build-system? o.O <leoprikler>Is there a precedent for search-path specifications in which not the full file-name is taken <leoprikler>i.e. something where a package needs share/foo, but the path variable should point to share/ ***apteryx_ is now known as apteryx
<efraim>anyone have experience using docker on guix? and/or using guix system docker-image? <str1ngs>leoprikler: I think it can handle diretories <str1ngs>leoprikler: it actually default to a directory so just add the path to the files list <str1ngs>probably you want to set the environment variable CXXFLAGS <str1ngs>more like CXXFLAGS="-std=c++20" if you are using bash <str1ngs>for guile it would be (setenv "CXXFLAGS" "-std=c++20") <raghavgururajan>Now I get "g++: error: unrecognized command line option ‘-std=c++20’; did you mean ‘-std=c++03’?" <slyfox>i think default CFLAGS for autotools based packages is usually '-O2 -g'. Setting it to something else undoes the default and effectively builds a program with -O0. <str1ngs>raghavgururajan: trying add gcc-10 to native-inputs <str1ngs>raghavgururajan: is there particular reason you need c++20? *raghavgururajan tries gcc-10 <str1ngs>problem is I have have gcc-10 on this system so the manual list c++20 as a standard <leoprikler>str1ngs: but what if it wants "share/foo/bar/" → "share/", i.e. it appends foo/bar/ to share/ to form a directory name? <slyfox>gcc did not change dependency style since gcc-3 <raghavgururajan>I don;t know why I still get "note: ‘std::make_unique’ is only available from C++14 onwards", despite "CXXFLAGS=-std=c++20" <slyfox>'note' is not an error. is it attached to an error or a warning? <slyfox>you might also want to share link to full build log. maybe there are other build options involved <raghavgururajan>slyfox, attached to an error "error: ‘make_unique’ is not a member of ‘std’" <slyfox>maybe '#include <memory>' is missing. Or CXX points to older g++ <leoprikler>(add-after 'set-paths 'set-more-paths (lambda (...) (setenv ...))) <leoprikler>but "make_unique is not a member of std" is a stdlib problem, not a gcc one <leoprikler>try if it works with gcc10-toolchain; if it does, you're probably missing something in libc <wleslie>how do I work from the source checkout? autoconf complains about undefined <brendyyn>wleslie: did you run guix environment guix <fnstudio>hi all! i'm experimenting with mcron and i'm wondering what the limits are of what i can do with it when installed on a foreign distro <fnstudio>in other words, i've guix-installed it as a user on a foreign distro <fnstudio>but i understand mcron needs to be launched as a (system-wide?) daemon, as all crons do, i'd imagine <fnstudio>- is it possible to `sudo apt remove cron` from my foreign distro and then `guix install mcron` (as a user)? would that allow me to schedule tasks (as that same user)? <fnstudio>- if not, would it make any difference if i guix-installed mcron as root? i'd tend to think that that would also fall short of launching a demo... as how can guix know of and interact with my system's service management? <bdju>anyone know how to check if my hardware is supported by intel-vaapi-driver? <leoprikler>fnstudio: I think you should be able to start a user mcron using systemd or even shepherd <leoprikler>you'd have to write the service file yourself though <fnstudio>leoprikler: uh, interesting, thanks - so it might be possible to install shepherd next to my system's systemd? <leoprikler>you can install shepherd as a user program through guix and start it in your bash_profile <fnstudio>if i install shepherd and then mcron (both as a user), doesn't mcron pick up the fact that shepherd is installed already and use that to daemonise itself? <fnstudio>sorry, a very naive way of phrasing it, but i hope you get what i mean <fnstudio>(thinking out loud) in other words, i could see whether mcron package comes with any shepherd's scripts <wigust>fnstudio: no, it will not "pick up" <wigust>fnstudio: you could daemonize with -d flag or by writing a service definition (which will be quite short) <fnstudio>wigust: cool, i see, thanks v much! i'll do so <mrpintrix>Hi, I'm having trouble understanding the behaviour of my wpa-supplicant service. Maybe one of you can. <mrpintrix>After I boot, the service is started and there's a wpa_supplicant process running. But I don't have a network connection. <mrpintrix>When I start a root session in a separate tty, I suddenly *do* have a connection, even on my user session. <mrpintrix>When I log out of the root session, the network connection disappears again. <vits-n-guix>raghavgururajan: re: -std=g++20 -- If that `make`, then You can try 'make --environment-overrides'? I seen that in neverball package. <desmes>Hey GNU people. I've reinstalled Guix and it continues to be extremelly awesome, thank you all! I have a quick question, though. <desmes>I'm trying to configure my filesystem as I had it on Arch on my fstab: UUID=30edd68b-0228-4bf9-96e2-c57f39ae8537 /media/main/30edd68b-0228-4bf9-96e2-c57f39ae8537 ext4 defaults 0 2 <desmes>(The thing I've added is that LET expression) <desmes>But when I reboot my system, an error pops up saying that the "defaults" option is not correct <desmes>Does anyone know how I can make this config right? <wigust>desmes: you don't need (options "defaults"), because you will have "defaults" in /etc/fstab in any case. <wigust>desmes: *in case no "options" specified <desmes>I see. It doesn't say so in the documentation though. Although it makes so much sense, lol. <desmes>Thanks a lot! I'll try it right now. <desmes>By the way, how come those "0" and "2" at the end, are not needed in the GUIX scheme declaration anymore? IIRC they had to do with the order in which filesystems where mounted. <vits-n-guix>desmes: why not to add also (flags '(no-atime))) to Your 'file-system' things? *vits-n-guix pretend to be a troll-face-man <desmes>vits-n-guix: I'm looking at the MOUNT manual and I don't even really understand that option. I think I'll pass on this one... <desmes>By the way, what's the difference between FLAGS and OPTIONS? They have the same potential parameters, don't they? <vits-n-guix>desmes: see 9.3 File Systems of Guix Manual. The flags only accepts a few things. <desmes>I see those flags, and If I search the mount options, some of them are the same. <vits-n-guix>desmes: When i tried to pass 'noatime' with (options) last time, it has no effect (or system didn't boot? Remember bad). <rekado>vits-n-guix: it may help all of us if you were a little less terse in your comments. (It’s also hard to distinguish humor from non-humor in text.) <desmes>Well ok, I guess I'll just ignore the flags. <vits-n-guix>rekado: even though i doubt about "all of us", i'll .. <desmes>By the way, Guile 3's fast as hell. Now I wish Guile Emacs was a thing too. <OriansJ`>desmes: it is a thing, just not fully drop in compatible yet <rekado>OriansJ`: it’s less a thing now than a couple years back, unfortunately <desmes>rekado: We now have gccemacs though, the speed improvements are also really noticeable <rekado>but when Guile Emacs was built with that branch it would just crash. <rekado>I tried rebasing Guile Emacs to the nearest Emacs release, but it’s a *massive* amount of work due to the countless changes in the Emacs core. <rekado>even for something as simple as the difference between the arbitrary variant of Emacs at the time Guile Emacs was accomplished and the next release. <rekado>I’m still only half-way through the rebase and had to pause <rekado>it’s very likely that it’s not going to work after the rebase. <rekado>BTW: we’re talking about Emacs 24 or 25 here. <rekado>so “not fully drop-in compatible” is quite the understatement. <desmes>I thought people stopped with the development of Guile Emacs <leoprikler>we still had our fair share of troubles upgrading 26 to 27 <rekado>it was primarily the work of one person <rekado>it was in a usable state but the reception on the side of emacs-devel was icy / dismissive <desmes>rekado: But you're continuing that, I see, aren't you? <desmes>rekado: Bye, thanks for the insight! <raghavgururajan>I get this error during build, "error adding symbols: DSO missing from command line" <NieDzejkob>IIRC that means that the arguments passed to ld are in a wrong order ***jonsger1 is now known as jonsger
<str1ngs>leoprikler: maybe with file-pattern you can do that. see native-search-path for libxml2 <raghavgururajan>The package was coded to git-download googletest, but I patched the source to use the one provide in inputs. <str1ngs>he's a recovering python programmer! :P <NieDzejkob>yeah, but doesn't your editor handle *that* automatically? <NieDzejkob>anyway, I don't think you should remove find_package from the cmake *raghavgururajan does `./etc/indent-code.el foo/bar foobar` as a last step <NieDzejkob>when I press enter, the line is indented just right immediately <NieDzejkob>and even when I paste something, the indents can be fixed with just =ap or similar <leoprikler>I do think, that the program in question is a little borked there <str1ngs>I think that matching xml then any file with that pattern. so if you add share which a pattern for bar I would think that would match. <str1ngs>also don't use (file-type 'regular) of course <leoprikler>but the resulting file will still be share/XXX/bar, wouldn't it? <str1ngs>with (files '("share")) and (file-pattern "bar") that should match share/bar no? <leoprikler>true, and I want that match, but I want only share as the result <str1ngs>or say more like (files '("share")) (file-pattern "foo/bar") would be better? <leoprikler>i.e. something like (if (file-exists? "share/foo/bar") "share") <civodul>hmm looks like Emacs 27.1 keeps eating memory without bounds until OOM <str1ngs>need to think on just the specification for share if that's possible. <str1ngs>leoprikler: don kill me but.... (separator "/../../:") haha! <str1ngs>civodul: hello there as a issue with 27.1 and seq that might cause that. but I thought that would have been resolved by now. ***jonsger1 is now known as jonsger
<leoprikler>that could work, but I doubt it should be made a standard <civodul>i'm using Emacs from 7090159c23d6345992ab976d71fefeb1583cfcdf (Sep. 10) <str1ngs>civodul: I can't find an issue for seq, maybe your issue searching fu is better then mine. <civodul>Emacs is at 9.6G resident, i'll have to hang up soon <civodul>hmm i don't have emacs-seq in my profile <leoprikler>I also feel setting separator wouldn't work for the case in which there's only one of them <greenrd>I just tried to create a bytecode output for the coq package, because bytecode is needed to debug coq plugins <greenrd>It didn't work, because coq hardcodes its absolute gnu store path inside itself at build time <greenrd>and that absolute path refers to the main out output, and therefore not to any other outputs, which are in different directories <greenrd>and I can't create symlinks from the out output to the bytecode output, because the point of making a separate output is so that it's optional to install <str1ngs>leoprikler: I'll only use .. as a stop gap. should be a better way to do this IMHO <greenrd>I guess the solution will be to instead create a derived package <Librecat>i downloaded it but i dont know what to do with it <greenrd>but then I wonder, since this coq-bytecode will be effectively a package for debugging only, will the guix project want it, or is it better for me to host it in a personal channel? <str1ngs>do we have a fancy GUIX MOTD somewhere? <greenrd>Librecat: are you going to install guix for the first time on a computer, or upgrade an existing installation? <bdju>Librecat: I think you just dd it to a flash drive and install from there <alextee[m]>PotentialUser-29: i think you email <bug id>-done@<whatever the domain was> <PotentialUser-29>Is it going to delete it from the bug tracker tho? I noticed my submission is listed online. I thought I was mailing to a bug help team or something. I didn't mean to publish it online <PotentialUser-29>They didn't have to publicly list my e-mail. God nows how many spams I will receive now. :@ <terpri>either PotentialUser-29 used a burner address, in which case who cares, or a main address which presumably receives spam already... <terpri>and debbugs obfuscates addresses anyway <terpri>maybe the docs should emphasize it somehow, but it's obvious if you click on any link in (info "(guix) Tracking Bugs and Patches") that bug reports are published online; dunno how one could conclude one was submitting a bug report to some kind of private support team from the current docs <leoprikler>I doubt our mailing lists are the primary target of Nigerian princes. <terpri>yeah, the gnu project is not exactly known for its wealth, ostentatiousness and gullibility ***hji- is now known as hji
<greenrd>How can I use a channel with a locally-modified guix? <greenrd>I tried using ./pre-inst-env guix pull -p /tmp/my-guix but I get a warning that compilation of the file that I changed failed <greenrd>also I think that is only going to be used for that instance of guix, not the instance of guix that guix pull is building? <zimoun>greenrd: did you do make before ./pre-inst-env? <zimoun>but it does not change, the file is not bytecompiled, just interpreted, I guess. <zimoun>so ./pre-inst-env guix pull -p /tmp/new should pull your modified guix <zimoun>however, you could have issue with authentication <greenrd>I didn't get any error about authentication, but the resulting guix doesn't see my channel's package <greenrd>the channel is listed in channels.scm and guix pull says it is pulling from it <greenrd>it also doesn't see my locally-changed package either <jonsger>yeah, I'm still waiting on my librem 5 as motivation kick start for me... <str1ngs>there might be some overlap with librem 5 but only for things like phosh <greenrd>OK, the reason it doesn't see my channel's package is probably because I forgot to set its name <greenrd>but regardless, I still need it to see my local changes to packages in guix as well <bdju>I don't have sxiv.desktop on my system. is this expected behavior? it's needed to set it as the default program for images in xdg <bdju>okay, since it needs to refer to a store path it should definitely come with the package. I'll file a bug <bavier[m]1><str1ngs "🎵🎵🎵 Hey, I just met you and t"> Wow, cool! That's a lot of progress already. <rekado>greenrd: is the “guix” command you are using ~/.config/guix/current/bin/guix? <HappyEnt[m]><str1ngs "🎵🎵🎵 Hey, I just met you and t"> Looks Awesome :)! Would love to try it on my pinephone, do you already have a repository online? <greenrd>rekado: Which one? Initially it's the local guix that pre-inst-env points to. Then it's the one built (as I understand it) in /tmp/my-guix <str1ngs>HappyEnt[m]: not yet, I'm trying to get as much work into guix.git first. <str1ngs>HappyEnt[m]: what pinephone version do you have? it might be easier for you. <HappyEnt[m]>str1ngs: Ahh yes sounds reasonable :). I've got rev 1.2 (UBports edition). <str1ngs>if so you might be able to use linux-libre as is. I have the 3GB model it I need to use the pine64 kernel for now. <HappyEnt[m]>str1ngs: Yes, should be the 2GB version. Ah didn't know there was upstream support for the soc of the pinephone. <HappyEnt[m]>str1ngs: Problem is my version is the one with the usb bug, so I kinda have to rely on the kernel fix from the pine64 kernel to get usb otg working. <str1ngs>HappyEnt[m]: maybe we can start with linux-libre and layer some patches on top. <str1ngs>also I don't know about blob and firmware. so it's going to take some research to get this working with mainline guix. <greenrd>ah, I think I've figured it out. I think I need to replace the default guix channel in channels.scm with a file:/// url pointing to my local guix fork. <greenrd>not sure whether I need pre-inst-env any more if I'm doing that <str1ngs>HappyEnt[m]: what distro do you normally use with your pinephone? <HappyEnt[m]>str1ngs: https://xff.cz/kernels/5.9/README megi has done a lot of work on making mainline linux work on the pinephone. Blob wise it seems like that camera, everything related wireless and hdmi out would not be able to work with libre linux. <HappyEnt[m]>str1ngs: currently i am not using the pinephone a lot, but most of the time i have run mobian <str1ngs>HappyEnt[m]: okay I use mobian too. right now I'm using stub u-boot-loader that does nothing. and then I used mobian to write a bootloader if I need it. but mostly that's because of specifics to the 3GB model I have. and I think I can eventually automate the bootloader just using normal guix and u-boot packages. 