IRC channel logs


back to list of logs

<liltechdude>Sorry for my bad english please
<civodul>liltechdude: hello! is the 'nss-certs' package globally installed?
<civodul>that is, is it listed in the 'packages' field of your operating system declaration?
<civodul>(if you're not comfortable with English, you can ask in your native language and maybe someone here will be able to answer)
<jab>liltechdude I'm still trying to set up my mail server! congrats!
<the_tubular>I'll be going trough that in the following days too!
<liltechdude>civodul: yeah, nss installed globally (packages
<liltechdude> (append
<liltechdude> (list
<liltechdude> (specification->package "nss-certs"))
<liltechdude> %base-packages))
<liltechdude>May be I should pull and upgrade system from root?
<civodul>liltechdude: hmm no, i'm not sure, maybe you should send your config to for some advice
<civodul>it'll be easier to help while looking at the config, hopefully
<civodul>(i haven't actually tried setting up a mail server, i must say :-))
<jab>the_tubular take a look at this config:
<jab>and this
<jab>and where it is actually used:
<jab>liltechdude: config.scm is the typically default name that is used for the scheme file that configures a guix system.
<jab>I use a file named "sway.scm"
<jab>My guix config repo is there somewhere:
<liltechdude>instead of DKIMproxy you can use rspamd to sign letter with key
<liltechdude>civodul: this is config but I can not understand what usefull information you can get from it
<jab>liltechdude that's true. But guix doesn't have an rspamd service defined yet.
<jab>technically it doesn't have a dkim-service either, but there is working code...
<liltechdude>civodul: If you are writing about mail server config, not about system config on home pc, then this guile config does not exist because of mail server run on freebsd
<civodul>oh ok
<jab>liltechdude you are running freebsd? I just installed openbsd a few days ago. I personally think that openBSD is closer to free software than freebsd....
<jab>liltechdude: by the way. We're officially friends. Your consent is not necessary. :)
<liltechdude>(freebsd is only valuable system on servers imho)
<jab>liltechdude it's certainly faster. :)
<liltechdude>network and FS much more faster than openbsd, yeah
<the_tubular>jab what should I be looking for in this config ?
<the_tubular>wait are you guys running guix on BSDs ?
<jab>the_tubular I am not.
<jab>I do not believe guix has been ported to the BSDs.
<jab>That would probably be a really hard port.
<jab>Specifically Guix System would be hard.
<jab>Guix the package manager might be ok-ish...certain features may not work.
<jab>but I've no idea what I'm talking about
<the_tubular>Yeah, that's what I though
<the_tubular>I don't really know much about BSDs, I should maybe get into that soon
<jab>the_tubular OpenBSD is maybe closest to FSF respects your freedom. They do NOT allow proprietary software. hyperbolaBSD just forked OpenBSD with the goal of making it GPL. but this may be off topic. Send me a chat if y ou like. though I only installed it a week ago, so my knowledge is limited.
<the_tubular>How do they compare to Linux I heard good and not so good things about security wise ?
<the_tubular>Maybe I should dm you cause it can be a big conversation :P
<calher>jab: I love OpenBSD's cwm. I installed Hikari (Wayland compositor) to enjoy it again.
<calher>the_tubular: jab: This is intersting discussion.
<jab>the_tubular I'm going to go eat dinner now, meet me in #openbsd in a bit in you like
<jab>also this:
<calher>jab: That works.
<the_tubular>I was just a bit burious, not really to use it but to know the advantage / disadvantage compared to linux system
<liltechdude>civodul: problem is fixed by using mbsync-get-cert
<the_tubular>I feel like what I've seen online is mostly outdated, could be wrong though
<KittyOwO[m]>I don't really trust a kernel not being gpl'd, or one recently being gpl'd. It would be nice if there was more time and investment spent into something similiar to hurd tbh.
<calher>KittyOwO[m]: Hurd is really nice, when it works.
<calher>jab: I like your site.
<KittyOwO[m]>I'm also fairly ignorant of the situation with the BSDs, but aren't they fairly similiar to linux as opposed to more ambitious projects like sel4, plan9, hurd, ect.? If porting/using a libre-BSD isn't much effort then it doesn't really matter, but if any of that takes effort that could go into a better long term solution that might be better.
<KittyOwO[m]>or if the effort would be by unrelated people.
<jab>KittyOwO[m] yes. The BSDs are Unix like. They do NOT want to be more ambitious like the Hurd.
<jab>calher I unfortunately have to agree...I've tried the Hurd in a vm several times, and I can't seem to figure out how to avoid file system corruption...
<KittyOwO[m]>What benefits do they have over the linux kernel? I am only vaguely aware of the situation. I heard they were better with namespaces or something?
<jab>though I am interested in trying it again and only just Emacs in the console. Maybe that will be a better experience.
<jab>KittyOwO[m] smaller code size.
<the_tubular>Hurd is another project I'm not familiar with
<the_tubular>What are their goals ?
<jab>that man page is the BSDs alternative to epoll. and it works better.
<calher>the_tubular: GNU kernel. Mostly just a hobby now, nothing big and professional like Linux.
<jab>the_tubular The Hurd was the original kernel for the GNU project.
<jab>it took too long to stabilize.
<calher>...and they left. :-(
<jab>many would argue it is not terribly stable yet.
<jab>calher whoops.
<calher>I love being able to "ls" a URI.
<jab>calher I tried that one with the Hurd.
<jab>grep the
<jab>I'm not certain what the syntax was...
<KittyOwO[m]>A couple of web (Hurd needs a gemini capsule mirrior if it doesn't already, that would be quite fun.) sites, the first is the official project site and the second is a guix blogpost of someone playing with hurd:
<jab>but it sort of must have parsed my site map...
<jab>but then it crashed and locked up.
<KittyOwO[m]>If it would actually be developed more, it looks like it has some very very cool concepts.
<KittyOwO[m]>I need to try it in a vm some time, or, as this blogpost suggests, a childhurd.
<jab>KittyOwO[m] It actually is developed fairly regulary, they just haven't updated the site in a while.
<KittyOwO[m]>oh, yeah, they need to do that. I got more up to date information from the guix site than the hurd one.
<jab>There is a guy actively working on porting the NetBSD kernel to the Hurd to use their device drivers.
<KittyOwO[m]>oh? nice
<jab>Once that is more complete, they could have a newer TC/IP stack, USB support and lots of device drivers.
<KittyOwO[m]>Is the NetBSD kernel like the linux kernel with binary blobs ect. or like linux?
<KittyOwO[m]>ah, having things like USB support would be quite nice
<yjftsjthsd>NetBSD is actually good for that because they have "rump kernels" which is a setup for factoring their kernel out into libraries that can run in userspace
<yjftsjthsd>This *feels* like an excellent fit for HURD
<calher>Just need to clean the proprietary stuff out.
<jab>I haven't heard from this particual Hurd developer a ton, but there is a Hurd developer who was working on SMP support as a part of his thesis.
<jab>I personally think that the Hurd desperately needs valgrind ported.
<KittyOwO[m]>yknow, despite the famously long dev time for hurd, I am honestly pretty hopeful for the future of libre computing. RiscV might become popular (shame RiscV isn't copylefted somehow to some degree) which sounds like a better situation than x86, tools like blender are becoming better and better each day, with some small companies starting to make movies using it, things like guix seems to have very cool and active development to get better each
<KittyOwO[m]>day, ect.
<jab>They suspect that they have various memory leaks in the filesystem, but they are very hard to track down without valgrind.
<the_tubular>Got disconnected, I miss some line if a message was targeted at me
<jab>the_tubular yes you did... ahaha
<jab>The Hurd was the original kernel for the GNU operating system.
<jab>It runs on top of GNUMach, which is a microkernel...
<KittyOwO[m]>`A couple of web (Hurd needs a gemini capsule mirrior if it doesn't already, that would be quite fun.) sites, the first is the official project site and the second is a guix blogpost of someone playing with hurd:`
<yjftsjthsd>Oh, can HURD drivers (...adapters? I forget the terminology) not run on Linux? I would have expected that to be simpler for them. And then you can use whatever debugging tools you want.
<jab>but GNUMach needs a fair amount of work. It's a first generational microkernel...with various deficiencies.
<yjftsjthsd>(Although yes, native ports would be better of course)
<KittyOwO[m]>sel4+hurd in guile when? :P
<jab>KittyOwO[m] probably never.
<jab>A port was tried.
<KittyOwO[m]>with sel4?
<jab>yeah, probably not.
<jab>There's soo much of the Hurd that assumes a Mach kernel.
<jab>Now another interesting project is X15 OS. which aims to be a Hurd like OS. But it uses a different custom microkernel.
<jab>It's got quite a way to go before it is really useable. It currently has no userspace.
<KittyOwO[m]>shame. how old is sel4 anyways? I wonder where we would be at rn if they tried rewriting it instead lol
<KittyOwO[m]>(and put more development effort in general in it)
<jab>KittyOwO[m] but there is cool hope for improving GNUMach. The IPC is rather bloated. There is a plan to improve its performance, but no one has stepped up to do it yet.
<jab>that's a possibility.
<jorge[m]123>Hola,tengo el siguiente error
<jorge[m]123>guix pull: error:Error git: failed to resolve address for name or service
<jab>jorge[m]123 try "ping -c 3"
<jab>Are you connected to the internet?
<jorge[m]123><jab "Are you connected to the interne"> si
<jab>si is yes ? haha. good.
<jorge[m]123><jab "si is yes ? haha. good."> yes
<jab>jorge[m]123 do you have "nss-certs" ?
<jab>jorge[m]123 is your clock way off on your computer?
<jorge[m]123><jab "jorge is your clock way off on "> no lo creo por que ?
<raghavgururajan>Folks! we use to download patches instead of having the patches in the repo right? I would like know the syntax for the former, but I forgot what to search for in the repo.
<jab>jorge[m]123 If your time is way off, I think that certificates can fail to works.
<raghavgururajan>Nvm! found it.
<jorge[m]123><jab "jorge If your time is way off, "> debo modificar en la bios ?
<jab>typically no. You can set your system time via...
<jab>what does "date" say?
<jab>Is it correct?
***califax- is now known as califax
<jorge[m]123><jab "Is it correct?"> Tengo que entrar,no tengo navegador por error en la construcción debe ser también al reloj del computador.
<jab>If your time is way off, then just use that link I sent you a second can set the system clock with
<jab>date -s "2 OCT 2006 18:00:00"
<jab>You have to run that as the root user
<jab>so sudo or su
<jab>jorge[m]123 leaving work....and probably going to bed.
<jab>good luck
<iskarian>...what's the difference between #:modules and #:imported-modules (in package definitions)?
<iskarian>AFAICT the manual does not explain
<drakonis>in reference to which definition?
<iskarian>Ah, sorry, in arguments
<drakonis>its a scoping thing i think?
<drakonis>it is explained in the docs though
<iskarian>Hm? The only matches for "imported-modules" was stuff talking about "with-imported-modules" in relation to gexps
<drakonis>probably renamed
<the_tubular>Someone knows how to mount ZFS pool with Guix
<the_tubular>Err I should say guile *
<marusich>drakonis, I don't know if it's explained. I couldn't find it. The explanation can be found in the source in the gnu-build procedure, in guix/build-system/gnu.scm:
<marusich>Assuming you are talking about the #:imported-modules and #:modules that are sometimes used in the arguments field of package definitions using the gnu-build-system build system, that explanation is accurate
<drakonis>that's not me who's asking about it
<drakonis>but yes its for scoping imports i think?
<drakonis>so you can import a guile library without making it global
<drakonis>for the whole file
<marusich>Oh, sorry
<marusich>I meant to answer iskarian
<marusich>For example, in this package:
<marusich>Because (guix build glib-or-gtk-build-system) is included in the #:imported-modules list, it will be copied into the build environment so that, during the build, it is possible for Guile to import it.
<iskarian>Ah, thank you marusich, that's what I was after! That makes sense.
<marusich>In addition, because it is included in the #:modules list, it will also be imported before the phases are executed.
<marusich>I.e., the phases, which are just procedures, will be applied in a context where (guix build glib-or-gtk-build-system) has been imported (i.e., use-modules was invoked)
<iskarian>That explains why the #:imported-modules do not need an import prefix even when #:modules do
<marusich>You can see the use-modules form here:
<marusich>The gnu-build invocation on line 388 does not refer to gnu-build from guix/build-system/gnu.scm, but rather gnu-build procedure defined in the "build-side" code that lives in:
<marusich>It is there that the phases are actually executed.
<marusich>Many build systems are based on the gnu-build-system in one way or another, so they follow the same sort of pattern.
<iskarian>Heh. I imagine it wasn't the original intent to have nearly all the build systems just piggy-back on it
<iskarian>Ah. This also makes the build-system/gnu.scm vs build/gnu-build-system.scm division clear.
<iskarian>I'd been wondering about that.
<marusich>guix/build-system is "host side" code. The guix/build code is "build-side" code.
<marusich>At least, that's my understanding
<iskarian>Yes, I see that now. :) Thanks for the explanation
<iskarian>Hmm, is there a way to get a list of transitive propagated inputs to a package?
<iskarian>Ah, I just dropped into REPL and used package-transitive-inputs
<the_tubular>I'm getting : modprobe: FATAL: Module zfs not found in directory /run/booted-system/kernel/lib/modules/5.12.17-gnu
<the_tubular>When I modprobe ZFS
<the_tubular>Is there a guix equivalent of modprobing ?
<drakonis>hmm yes
<the_tubular>Someone pointed me to this :
<the_tubular>But I am not sure I understand it all
<drakonis>the error here is that you dont have zfs on your system profile?
<the_tubular>I am unsure
<lispmacs>hi, for a year or so I've been using an NVIDIA GeForce 8400 GS Rev. 3, and not having any trouble. Recently sometime in the last few weeks, suddenly all my 3D accelerated graphics have slowed to a crawl. I am wondering if there has been some recent change in some guix package that might explain this? (mesa...? gnome...?)
<lispmacs>some nouveau driver I need to install into system...?
<drakonis>its probably nouveau
<lispmacs>before all my 3D programs ran smoothly, now it is a few frames a second
<lispmacs>drakonis: I see a xf86-video-nouveau package. do I need that installed in the system somehow? I always presumed nouveau installed with whatever installs mesa
<drakonis>probably not
<drakonis>your gpu is fairly old by now
<drakonis>do you use gnome?
<drakonis>not sure if that might be the cause?
<lispmacs>I'm open-minded. I've been using GNOME as long as this video card, however
<lispmacs>it wasn't a problem before, I mean
<drakonis>well, i'm not sure if it might be a case of gnome having growing hardware requirements
<lispmacs>GNOME shell doesn't seem to have slowed down
<drakonis>a 8400gs is surprisingly old
<drakonis>which workload is slowed down?
<drakonis>2006 whoa
<lispmacs>it is most noticeable when playing supertuxkart, slowed down to what looks like a few frames per second. Wesnoth seems a bit laggy now, and I'm getting vertigo playing supertux (the platform game). Other apps seem to work fine
<lispmacs>I just don't get why it was working great a month ago and now suddenly is crawling
<drakonis>kernel updates?
<drakonis>some change to nouveau might be causing it?
<lispmacs>well, I guess I could try rebooting into some old systems, at least to narrow it down
<the_tubular>I'm unsure on how to add something that has is in my config.scm to my path
<the_tubular>Is it just : GUIX_PROFILE="/root/.config/guix/current" ?
<lispmacs>the_tubular: did you add packages to your config.scm, or are you talking about services?
<the_tubular>I added the package to my config.scm
<the_tubular>I have no clue if ZFS should have a service though
<lispmacs>those packages are already in your path then
<the_tubular>Documentation isn't great on that issue
<lispmacs>though you may need to relogin or restart your shell or something
<the_tubular>I rebooted and it still gives me that error : modprobe: FATAL: Module zfs not found in directory /run/booted-system/kernel/lib/modules/5.12.17-gnu
<the_tubular>As I said above this is by running modprobe, there should be a better way to achieve this in guix, but I don't know how
<lispmacs>the_tubular: I think you need to add the kernel modules to your operating system -> kernel-loadable-modules or something like that
<lispmacs>maybe take a look at section 10.2 in the manual
<lispmacs>just adding a package to the system package list won't be enough I think
<lispmacs>the_tubular: also note that the zfs module is not the default output
<lispmacs>module package is zfs:module
<lispmacs>as you'll see if you look at guix show zfs
<the_tubular>Mind telling me what I should add to my config.scm ?
<the_tubular>I'm afraid of messing it up
<the_tubular>Just like (kernel-loadable-module zfs-module) ?
<lispmacs>I think you would need (kernel-loadable-module (specification->package+output "zfs:module")) but that is an educated guess
<lispmacs>(kernel-loadable-modules (list (specification->package+output "zfs:module")))
<lispmacs>I'd recommend putting your config.scm under git version control, then you don't have to worry about messing it up
<the_tubular>Ohh, yeah. I'll definitely do that
<lispmacs>and with guix, if you mess up the syntax in the config.scm, nothing bad will happen, guix will just abort with an error
<the_tubular>Right now I have a config2.scm in my folder :p
<the_tubular>Guix info shows the lisence as CDDL, that's interresting
<lispmacs>we have a 1980's computer at my work which uses letters for version control. Only 8 characters are allowed in the file system, so it becomes MOD40..A, then MOD40..B, etc.
<lispmacs>you have to wrap around after MOD40..Z
<the_tubular>I added that line in my config.scm, it output no error, but it still says that zpool list still says the module isn't loaded
<lispmacs>do you need to reboot maybe into the new system...?
<the_tubular>You have to reboot when you modprobe something ?
<the_tubular>I can try, but it's usually not the case
<lispmacs>just after installing the new system profile...?
<lispmacs>two system profiles are completely separate
<the_tubular>What ?
<lispmacs>as far as what is exposed in the profile, I mean
<lispmacs>right before you do system reconfigure, you are running one system profile, afterwards you have created another
<lispmacs>you can boot into either one
<lispmacs>as you will see in your GRUB list.
<the_tubular>But when you type guix system reconfigure, it changes the profile no ?
<lispmacs>I think it does switch to the new one, yes
<the_tubular>Then I don't need to do that right ?
<lispmacs>I don't think so, but I'm just worried that your app is looking in the old profile kernel module directory
<the_tubular>Ohh, okay, I'll reboot then if there's stuff that could have gone wrong
<lispmacs>not sure but just trying to throw out a possibly helpful idea
<the_tubular>Appreciated :)
<lispmacs>i don't wonder if that wouldn't fix it, since the error says it is looking in the directory of the "booted system"
<lispmacs>there are somethings you can't change without a reboot, like the running kernel, and evidently available kernel modules
<lispmacs>I got to go to bed
*lispmacs zzzZZZZ
<the_tubular>That didn't work :/
<the_tubular>Altough when it booted, I could read "loading kernel module"
<lispmacs>what does the error say now?
<the_tubular>The ZFS modules are not loaded.
<the_tubular>Try running 'modprobe zfs' as root to load them.
<the_tubular>Same error
<lispmacs>what exact directory path is listed in the error now?
<the_tubular>This error : modprobe: FATAL: Module zfs not found in directory /run/booted-system/kernel/lib/modules/5.12.17-gnu ?
<the_tubular>Same one as before also
<lispmacs>what does `ls /run/booted-system/kernel/lib/modules/` show?
<lispmacs>and `ls /run/booted-system/kernel/lib/modules/5.12.17-gnu`
<lispmacs>and `cat /proc/version`
<the_tubular>ZFS isn't in /run/booted-system/kernel/lib/modules/5.12.17-gnu/kernel/fs/
<the_tubular>There is zonefs thoug ...?
<the_tubular>Linux version 5.12.17-gnu (nixbld@localhost) (gcc (GCC) 7.5.0, GNU ld (GNU Binutils) 2.34) #1 SMP 1
<lispmacs>and that system reconfigure completed without error?
<the_tubular>Maybe what I typed was wrong
<lispmacs>sudo guix system reconfigure config2.scm ?
<the_tubular>config2.scm is my backup
<the_tubular>config.scm is the one I added that line below : (kernel-loadable-modules (list (specification->package+output "zfs:module"))))
<the_tubular>The 4th parenthesis is the one opened with (operating-system
<lispmacs>looks correct
<the_tubular>Weird that it's not more documented than this, I though ZFS was a pretty standard use-case for a linux system
<the_tubular>Or I should say GNU/Linux
<lispmacs>there is a lot of packages then get thrown in without much documentation or proper service definitions
<lispmacs>somebody kindly added yubikey pam support years ago but didn't add a service definition. as far as I know it still isn't useable (my bug report is still open)
<the_tubular>true, but I guess managing a Distro is a lot of work
<lispmacs>anyway, I'm thinking conceivable something might be wrong with the zfs:module definition so that the module is not in the right directory in the build output
<lispmacs>you might try posting on guix-help to see if somebody else is using zfs
<lispmacs>or help-guix
*lispmacs zzzzZZZZ
<lispmacs>must sleep
<the_tubular>Goodnight, thanks for the help!
<the_tubular>help-guix is an IRC channel ?
<the_tubular>Or a mailing list ?
<bricewge>the_tubular: You need to reboot to be able to load a newly added module
<the_tubular>I did, bricewge, sadly no dice :/
<the_tubular>Anything else I might not be thinking about ?
<bricewge>Then the module wasn't properly added to your operating-system
<bricewge>Would you mind sharing your whole operating-system?
<the_tubular>Yes, give me a second
<the_tubular45>Sorry, I got disconnected, did I receive something ?
<bricewge>I may repeat myself but you are on a uncharted territory here, and you should probably continue what raid5eatemyhomework did
<the_tubular45>Darn :/
<bricewge>Looking at the path I linked yestarday (at the end of it) they add to the kernel-loadable-module-field "zfs-loadable-module"
<the_tubular45>I'd use raid, but apparently it ate is homework :(
<the_tubular45>Ohhh, let me try that
<bricewge>Instead of zfs:module
<bricewge>"zfs-loadable-module" is a procedure defined in that patch
<bricewge>I skinned through the doc added by the patch and they list a bunch of caevat we it come to zfs usage
<the_tubular45>so the line should end like this ? (specification->package+output "zfs:module" "zfs-loadable-module") ?
<the_tubular45>Ohh, only 1 ?
<the_tubular45>I get : guix system: error: zfs-loadable-module: unknown package
<the_tubular45>I tryed guix pull as it was yesterday and I still have the same error :/
<bricewge>That's a procedure not a package
<bricewge>And it's defined in the patch
<the_tubular45>Umm, I'm not sure how to define that in Guix
<the_tubular45>Let me check that patch
<the_tubular45>Umm, I see where it defines it, but I'm still unsure how to do it
<bricewge>Search for "(define (zfs-loadable-modules"
<the_tubular45>Well this is just defining it, it doesn't tell you how to use it correct ?
<the_tubular45>Sorry, guile is still so foreign to me lol
<the_tubular45>I love it, but god damn it kicks my ass
<bricewge>Sorry I need to go to work
<the_tubular45>All good, I'll try something
<tissevert>hello guix
<the_tubular45>Hi tissevert
<tissevert>howdy ?
<the_tubular45>Pretty good, how are you ?
<tissevert>good, I'll have to temporarily give up fighting with unpackageable stuff and spend some time writing unit tests, which is quite nice and less hazardous
<civodul>Hello Guix!
<the_tubular45>I'm still messing with my ZFS pool
<the_tubular45>I'm kinda hesitant because some of the data on there isn't backup :/
<the_tubular45>Hey civodul :)
<tissevert>hello civodul
<tissevert>yeah, that's quite understandable, I hope you manage to do it because I'd be curious to see how it's done
<the_tubular45>Yeah, not looking to good
***modula is now known as defaultxr
<wirez>was LGPL linux admitting some kinda defeat for gpl or?
<MysteriousSilver>what is LGPL linux?
<efraim>does symlink go backward compared to 'ln -s'?
<civodul>you never know until you try ;-)
<civodul>but no, it's the same: (symlink old new)
<efraim>bah, I'll leave the julia code as 'ln -s' while I work on seeing if it's necessary
<cage>hi! amybe a trivial problem but when i try to compile a c source with gcc i get: 'ld: cannot find crt1.o: No such file or directory'
<vivien>cage, do you have gcc-toolchain installed?
<vivien>(not "gcc")
<vivien>Are you on guix system or did you install guix on top of another distribution?
<vivien>(so-called "foreign" distribution)
<cage>i am on guix system, i am updating the system right now, maybe this is going to help :)
<cage>unfortunately it does not :)
<vivien>Maybe some paths get hard-coded in your project, try to re-compile from scratch. What program is it?
<vivien>Does it have a build system?
<cage>yes it uses autotools building system
<vivien>So you could try running "make clean"
<cage>the fact is that even if i try to compile with gcc an empty file i got the same error
<cage>touch a.c && gcc a.c -> error
<vivien>What does `which gcc` say?
<vivien>If I run ls -l $(which gcc), I get: lrwxrwxrwx 22 root root 72 1 janv. 1970 /home/vivien/.guix-profile/bin/gcc -> /gnu/store/ji09aab6as5qqyda9h0yvif55zla0rdi-gcc-toolchain-11.1.0/bin/gcc
<cage>vivien: i just logged out and login again and everything seems to works fine now, sorry for the inconvenience and thanks you very much!
<vivien>I’ll be sure next time to recommend this method ^^
<wirez>wen guix 1.4?
<raghavgururajan>Hello Guix!
<lisp123>Hello :-) Apologies for the basic question, what is the easiest way to install Guix (OS) as a VM Image for Mac? I have VM Fusion but the steps on the website got complicated a bit fast
<lisp123>Is there a way I can just use the ISO?
<roptat>I don't know VM Fusion, but I guess you can use the ISO in the same way you would use the ISO of another distro? you'd have to look at the documentation for VM Fusion to learn about that
<roptat>(maybe find an option to set a virtual optical disk drive, mount the iso on it, and set the bios options in the VM to boot to the CD drive first)
<lisp123>roptat: thanks, i'm trying it out now. It asked me for which linux distribution I'm trying to load, so I selected 'other' and hopefully it will work
<roptat>good luck and have fun :)
<lisp123>thanks :)
<MysteriousSilver>yup, just load the image and boot from it
<MysteriousSilver>delayed messages, ignore
<lisp123>Seems to be working :)
<constfun>hello everyone. is there a way to launch a opengl application inside a guix container? i tried `guix environment --preserve=DISPLAY --preserve='XDG_*' --expose=/dev/dri=/dev/dri', that gets me past the initial errors, but a window still doesn't launch even though the opengl context seems to be initialized and all (from my logs), the same binary
<constfun>launches outside the container, i just want to make sure I packaged all dependencies, hence the need to launch in a container
<civodul>constfun: hi! i don't know the answer, but perhaps you could strace the binary to try and guess what it's missing
<civodul>also note that --preserve/-E expects a regexp, not a glob pattern
<civodul>so it should be, say, "^XDG_.*"
<roptat>ah I'd be interested, I get warnings about opengl being uninitialized when I try to run the android emulator from android studio
<roptat>I managed to find some of the missing things, but then it immediately crashes ^^'
<constfun>civodul: thanks for the tip, ill give it a shot
<constfun>roptat: i got past the initial opengl errors by adding the /dev/dri part, opengl seems to be initialized, my hunch is that it has something to do with SDL or the window manager, im on i3... perhaps i should try gnome
<civodul>constfun: might be worth sharing your findings here or on the help-guix mailing list!
<civodul>perhaps we could add the tip to the cookbook afterwards
<tissevert>wait there's an android emulator in guix packages ?
<roptat>no, there isn't
<roptat>I had to get android-studio from the web
<tissevert>ohhh : (
<constfun>curiously, if i _also_ pass '--network' to the environment then mesa fails to launch with 'MESA-LOADER: failed to retrieve device information' that has me really perplexed, why would the network flag affect opengl...
<roptat>I think it affects /dev
<raghavgururajan>sneek, later tell jab: Could you share your ~/.config/sway/config file?
<sneek>Got it.
<constfun>...the part i failed to mention is that this is an -RR (--relocatable --relocatable ) binary, i see disallow_setgroups error now, im guessing it has something to do with it, ill stop spamming now and report at guix-help if i get past this, thanks for the pointers!
<civodul>ah ha! we have some interrrresting ppc64 build failures at
<mbakke>Python is now reproducible on 'core-updates' \o/
<mbakke>...but I had to undo some of roptat's optimization work :/
<civodul>mbakke: oh, yay! bah, boo!
<bricewge>civodul: We can reuse wireless-regdb signature
<civodul>bricewge: oh that's great! well done
<bricewge>I'll send a patch in the coming days
<civodul>reproducible builds in practice
<roptat>mbakke, oh no, what happened?
<mbakke>roptat: see 71ec85b2958798246fc2b5d84c40badf5f75668e and the commits leading up to it :-) the --enable-optimizations flag creates a 100k line diffoscope log, with different assembly in each build :/
<mbakke>roptat: --with-lto seems more easily fixable as just some hashed function names are different; it may be possible to work around it by passing CFLAGS=-frandom-seed=1 or similar, but I haven't tried it
<jab>morning #guix!
<the_tubular45>Hey what's up jab!
<jab>the_tubular45: what up? Where we chatting last night and you sent me a link right?
<roptat>mbakke, oh you essentially removed all the optimizations I added
<roptat>not just "some" ^^
<jab>let me watch that real quick...
<mbakke>roptat: sorry about that! ...there's still --with-computed-gotos and -fno-semantic-interposition if that's any consolation!
<jab>50 minutes later...
<roptat>ok, it's probably not enough compared to others, but reproducibility is probably better
<roptat>more important*
<the_tubular45>Yeah we were jab
<the_tubular45>Sorry, still haven't setup my propre irc name so my name is all messed up :/
<dstolfa>roptat: mbakke: have you tried using a different linker to get reproducible builds when LTO is enabled? maybe it's a bug in whatever linker is being used
<dstolfa>previously there were bugs with randomly ordered debug information that caused non-reproducible builds, but i don't know what it may be in this case
<civodul>mbakke, roptat: does --enable-optimizations turn on profile-guided optimizations (PGO)?
<jab>the_tubular45: don't feel bad...irc is hard to use.
<jab>genuinely. it's got some quirks to it.
<the_tubular45>Name reservation when you get disconnected sucks, never had a client that deal with it "well"
<mbakke>civodul: it does
<the_tubular45>But this is just a web client until I finally get my guix box up and running
<mbakke>roptat, dstolfa: here is diffoscope output for two buils with the --with-lto flag:
<civodul>mbakke: ah, too bad
<the_tubular45>Talking about IRC, did anyone know about biboumi ?
<dstolfa>mbakke: seems like the linker reorders LTO-related sections in the ELF file, which linker is this?
<the_tubular45>I'm probably gonna try it later but I barely hear nobody talking about it
<dstolfa>ah it doesn't reorder it, it has a different suffix
<dstolfa>whatever it is
<dstolfa>that sucks
<civodul>yeah, the hex part of the symbols differs
<civodul>mbakke: from reproductible-notes.git, a solution is to "pass the full path of the source as the argument to `-frandom-seed`."
<civodul>perhaps that's what you emant above
<civodul>(a great reference!)
<mbakke>civodul: thanks for the link! I was not aware of that workaround and only did a superficial web search.
*mbakke has to go, happy hacking folks :-)
<civodul>heheh, thank *you*!
<ix>guix system: error: chown: Permission denied
<ix>Ok, thanks
<ix>I do love guix error messages
<ix>Its like chinese proverbs
<roptat>ix, are you trying to run guix system reconfigure as non root?
<ix>No, im trying to run it as root
<roptat>can you strace it? so we know which file it's trying to chown?
<bricewge>the_tubular45: raid5atemyhomework asked for feedback about it's ZFS patchset today
<the_tubular45>Interresting, I'll go check it out, but to be fair, I feel like I just can't get it to work and wouldn't be able to give him a nice feedback
<ix>roptat: im trying to, but that seems to add new errors
<ix>Hence my exasperation
<roptat>what kind of error?
<ix>No code for module
<roptat>uh weird
<ix>I think its some environment thing
<roptat>but strace shouldn't change the environment...
<ix>No, but sudo does
<ix>And grandchildren behave different to children of sudo
<ix>So slipping strace in breaks things
<roptat>I see, then I don't know how to help you
<ix>Woe betide me
<ecraven>trying to install guix on arch linux (from aur), I get this error for guile-lib "configure: error: found development files for Guile 2.2, but /usr/bin/guile2.0 has effective version 2.0" any idea what this is about?
<roptat>the aur package is probably broken
<ecraven>any better way to install guix than the packages?
<roptat>it found guile-2.0, but it needs guile-2.2 ac least
<ecraven>guile --version says 2.2.7
<roptat>you can use the installation script
<ecraven>does guix work with guile 3?
<roptat>see the red note here for the installation script:
<dcdve>is there a way to install desktop environments without the use of gdm?
<ecraven>roptat: thanks, I'll try that
<dcdve>if there is, i'
<dcdve>m having a heck of a hard time doing it
<ecraven>just need to reinstall guile3 first ;)
<roptat>you don't need anything to use the installation script, other than a working shell
<ecraven>it won't use the host guile?
<ecraven>thank you
<lispmacs[work]>dcdve: I'm in the process of installing a system that uses elogind and slim
<lispmacs[work]>It seems to be installing fine but I haven't booted into it yet
<ecraven>what does the 'check' phase do?
<roptat>when building a package? it runs "make test" essentially
*the_tubular45 is getting gentoo flashback
<ecraven>should guix environment --container --adhoc --share.... work on foreign installs? or only on guixsd?
<dstolfa>ecraven: it works for me on RHEL
<ecraven>I can still see all of my home directory, even if I don't share it :-/
<ecraven>or is the current directory always shared?
<dstolfa>i think the cwd is always shared
<dstolfa>that's probably why
<ecraven>ah, ok, so I'd need to just run this from some *other* directory ;) thanks!
<ecraven>thanks, this seems to work ;)
<apteryx>for properly initializing the Guix user profiles in Jenkin scripts, I ended up ensuring each agent had coreutils >= 8.30 and can now use this shebang: #!/usr/bin/env -S bash --login
<apteryx>it takes care of sourcing /etc/profile (or /etc/profile.d/ on foreign distribution) as well as ~/.bashrc (depending on ~/.bash_profile, but that's very common).
<ecraven>how do I get rid of the hint to install glibc-utf8-locales? I already did, and also have exported GUIX_LOCPATH
<dstolfa>ecraven: does `hash guix` fix it?
<ecraven>I'm running zsh, the variable is definitely set
<dstolfa>hm, not sure then, never had issues with it :/
<ecraven>and the directory contains the locale data
<apteryx>I wonder if it could have to do with zsh
<ecraven>no, same with bash; )
<apteryx>could you try with bash, to determine it's not some compatibility issue in the scripts?
<apteryx>that's on a foreign distribution?
<ecraven>on arch linux 64 bit amd64
<ecraven>I was assuming that message would go away once I actually installed the locales package and set the environment variable
<apteryx>could you paste what you see to to make sure we're on the same page?
<roptat>can you check it's also set in /etc/systemd/system/guix.service?
<dstolfa>it should go away after you've reloaded your bash profile
<dstolfa>but if it hasn't, then i have no idea
<dstolfa>or rather, zsh profile i guess :)
<apteryx>I'd also recommend installing glibc-locales instead of glibc-locales, unless you're sure your particular locale is covered
<ecraven>roptat: I have no such service, only guix-daemon.service
<dstolfa>apteryx: glibc-locales instead of glibc-locales?
<apteryx>sorry, glibc-utf8-locales
<roptat>oh right guix-daemon.service
<ecraven>it is set, and uses the user root (for which I did not install the locale package.. I'll do that now)
<roptat>oh, did you run guix pull maybe?
<dstolfa>ecraven: guix profiles are per-user, which means that if you have something installed on the user, it isn't necessarily installed on root and so on
<dstolfa>so every user has to install things in their own profile
<ecraven>dstolfa: I know. but even if I install it for root, I still get that Hint
<ecraven>roptat: no, but why would that influence it? newer version available?
<apteryx>segfault from bash while building guile@3.0.2; weird
<roptat>yeah, if there's a new version of glibc, that's not compatible with the glibc-locales of the root profile
<roptat>but if you didn't, that's not it
<ecraven>ok, I'm pulling now ;) let's see whether that helps
<roptat>I doubt it ^^
<ecraven>but that should be enough, to install the package for my user and then set that env variable, right?
<roptat>hm.. what's your $LANG?
<ecraven>ah, that's not part of the utf8 locales
<ecraven>that was it. thank you!
<roptat>right, you can try with glibc-locales
<ecraven>yea, but that'd take a lot of disk space for no useful gain
<roptat>it's a big package though
<apteryx>the next release of Guix won't hint at glibc-utf8-locales anymore
<ecraven>the hint is very much ok ;)
<ecraven>there's no way to just install one locale, right? I'd have to write my own derivation for that?
<apteryx>roptat: it's only big while uncompressed; otherwise like 12 MiB with lzip
<roptat>yeah, it's full of zeroes
<apteryx>so if you use Btrfs + zstd compression for example, it's cheap
<ecraven>yea, if I actually switch to guix on this machine, then I'll do that, but right now the file system is still ext4
<ecraven>there's no GUIX_LANG, I assume?
<ecraven>can I tell guix to show me the derivation for a package (glibc-utf8-locales in this case)?
<roptat>guix build -d glibc-utf8-locales
<ecraven>hm.. what's the file that contains the actual list of locales? I could create my own local channel which has a slightly modified version of that package, right?
<apteryx>you could make your own glibc-your-locale package using make-glibc-utf8-locales
<apteryx>from (gnu packages base)
<ecraven>is there some sort of tutorial on how to create my own channel?
<roptat>you can try to define something like (locale-directory '("en_GB.UTF-8") #:glibc glibc)
<ecraven>are the package modules that define the packages copied to my local system? so that I can look at them?
<roptat>you can try "guix edit foo" to see the definition
<roptat>"edit" is a bit misleading, you can't modify that file
<ecraven>is there a way to just save a package file and try to build that, without creating a channel?
<roptat>yeah, you can try "guix build -f the-file.scm"
<dstolfa>apteryx: i can't reproduce the guile failure you have
<dstolfa>it builds successfully (but not deterministically) for me
<ecraven>so after having built my package with guix build -f my-file.scm, how do I install it into my profile? guix install says it can't find the package name
<roptat>guix package -f my-file.scm :)
<ecraven>thanks ;)
<roptat>guix build builds the package and retuns the store path, but you need guix install/guix package -f to actually install it
<roptat>if you run "guix install", the package can't be found because it's something more that guix doesn't know about, you need to pass the file so it can be known
<ecraven>thank you very much, this works well, I now have a rather useless package that just installs that one locale ;) thanks for the explanations and patience!
<roptat>yw :)(
<roptat>arf, I can't type
<ecraven>how do I get the sha256 sum in base32 for a given file?
<lispmacs[work]>ecraven: sha256sum <file> | base32
<lispmacs[work]>err, well, you only want the sum, one second...
<ecraven>is only base32 supported? nothing else?
<lispmacs[work]>guix hash <filename>
<iskarian>Good morning guix :)
<dcdve>do I have to do 'export PATH=... export INFOPATH=...' every time I do guix pull (on guix os)?
<ecraven>if I run the trivial build system, where will the package sources be extracted? are they relative to cwd?
<iskarian>dcdve, I've never had to do that. Sometimes I have to do 'hash guix' or '. ~/.config/guix/current/etc/profile' (or just relogging)
<iskarian>ecraven, I could be wrong, but I don't think trivial-build-system unpacks sources at all
<ecraven>so the downloaded file will be in CWD?
<iskarian>I believe you would use (assoc-ref %build-inputs "source")
<iskarian>Are you sure you don't want copy-build-system?
<ecraven>not sure, I need to call nasm (to build a .o file) and gcc (to link it)
<roptat>the trivial-build-system is the most complex build system to use of all, despite its name, because it's so trivial it doesn't do anything
<ecraven>which one should I look at instead?
<roptat>gnu-build-system if they use configure / make / make install
<roptat>cmake-build-system if they use cmake, etc
<roptat>also, gnu-build-system if there's no configure but a makefile
<ecraven>well, they don't use any of that, as I said, I need to invoke nasm to compile the assembly file to .o, then gcc to link that
<roptat>I'd still use the gnu-build-sytsem, remove the 'configure phase, replace the 'build phase, disable tests and replace the 'install phase
<ecraven>ok, thanks ;)
<iskarian>It's fairly uncommon for a project to be without a build scripts these days
<ecraven>what do I need to pass to gcc to get this correct: "/gnu/store/xakj5dgs1729297nv50s84sdmq2jiz64-binutils-2.34/bin/ld: cannot find crt1.o: No such file or directory"
<roptat>ecraven, gnu-build-system should come with most of the dependencies you need, so make sure you don't add gcc to the inputs
<ecraven>yea, still playing around with trivial build system. so should gnu-build-system automatically fix these problems?
<roptat>it's probably missing environment variables
<roptat>or a package entirely
<ecraven>yea, I'll look into gnu-build-system, just need to figure out how to do what you said above, disable and replace phases ;)
<roptat>you would use modify-phases, there are tons of examples in the repo :)
<ecraven>yea, I'm looking at those ;)
<roptat>and #:tests? #f in the arguments to disable the check phase
<ecraven>sorry for the trivial questions
<roptat>it's fine, we all start somewhere :)
<ecraven>where does guix put the repo? right now, I'm just using guix edit ... to look at files, but that way I cannot grep ;)
<roptat>it's easier to clone
<ecraven>thanks ;)
<dcdve>thanks iskarian
<dcdve>after running sway or hikari without a display manager, my vm crashes
<dcdve>those are both wayland based
<dcdve>cagbreak doesn't crash, but has a seg fault after loading a direct session
<iskarian>it crashes? that's odd. Why would you want to run without a DM though?
<dcdve>why not? freedom
<dcdve>unless there's a huge security risk I don't mind logging in through shell and starting the destkop environment manually
<Stuttergart>Howdy, all.
<Stuttergart>Sure seems like all of my `guix package` operations fall if I'm connceted to Proton VPN. Perhaps those networks are blocked?
<Stuttergart>Fail via timeout to be specific.
<cwebber>out of curiosity
<cwebber>does anyone still use the guix emacs mode anymore?
<cwebber>I guess I'm due for an upgrade but it hasn't seemed like it's worked right for a while
<cwebber>I wonder if it would be the right idea to pull it back in guix again proper
<cwebber>it seems to me like the original arguments of "this is more likely to be maintained if in guix proper" turned out to be true
<Stuttergart>Oh, neat. Magit like interface to Guix?
<cwebber>Stuttergart: yes, it's kind of like that
<Stuttergart>At first glance, this is darn cool.
<cwebber>used to be the main way I'd interact with guix
<cwebber>but it has gotten buggier over time
<jackhill>cwebber: looks like we did pull into under the Guix umbrella. At least according to suvannah, but it matches my recollection:
<Rooks>What would the "Guix way" be to set a discrete GPU as primary?
<cwebber>jackhill: yes, but it was originally actually part of the guix repo
<cwebber>but the maintainer at the time expressed preference to develop it separately
<cwebber>some devs expressed hesitation that maybe it wouldn't be as well maintained if this happened
<cwebber>and it does seem like this eventually turned out to be true
<jackhill>cwebber: oh? Yeah, I could see how that would make it easier to keep the APIs in sync.
<jackhill>cwebber: sounds like some of that happened before I started paying attention
<cwebber>jackhill: heh, yeah this happened a lonnnnnnng time ago
<jackhill>looks like there was some effort to reinvigorate it, but that may have stalled as well…
<cwebber>yeah and it's not as if making it part of the main guix repo would result in an automagical un-bitrotting
<Stuttergart>How would one list the contents (files) in a guix package?
<Stuttergart>I can't find that in the docs anywhere.
<platoxia>The package definition should have all dependencies, which you can follow all the way down the dependency tree.
<platoxia>I don't suppose that really answers your question though...sorry.
<Stuttergart>y, I installed protonvpn-cli and it doesn't seem to be in my path after a hash so I figured I would check the package to find the bin/script.
<platoxia>You could do "guix install --dry-run" to see everything that will be installed.
<platoxia>Are you running Guix System or just the package manager on a foreign distro?
<Stuttergart>this package apparently names the binary 'protonvpn' as opposed to 'protonvpn-cli'
<platoxia>Did you follow the instructions in "Getting Started" from the manual? Specifically, ensure you follow the directions in the 4th paragraph.
<Stuttergart>y, that's in my bash configs.
<Stuttergart>aslo in my env
<Stuttergart>I think its just that this package chose a different name for the script/bin than, for instance, what is used in the .debs.
<platoxia>Is it working with 'protonvpn' that the executable name in guix?
<atka>hello, I'm playing around with the qemu image with the xfce desktop, any reason I can't change the keymap?
<platoxia>atka: The keymap is set in the system configuration.
<Stuttergart>platoxia: it appears to be the same code as the .debs other than the different naming
<Stuttergart>I can check the files in /gnu/store, fwiw
<atka>platoxia: and that overides the desktop settings panel?
<platoxia>atka: yes, change it in the system config file then reconfigure the system and it will work.
<atka>platoxia: ok thanks
<platoxia>Stuttergart: I'll install it and check it out.
<rekado>I got myself a Rock Pi S and installed Guix on top of Debian Buster; unfortunately, I cannot seem to fetch substitutes.
<rekado>I get this error: warning: host not found: Servname not supported for ai_socktype
<rekado>any ideas what this means and how it could be fixed?
<rekado>I copied /etc/services from my Guix system over; now it works