<winter>How would one pull latest stable, if desired, then?
<winter>(I should RTFM before asking, shouldn't I)
<nckhexen>Very roughly, guix-install.sh is like downloading the ‘latest’ (and possibly months old) installation ISO of $better_known_distro, and ‘guix pull’ is like running apt update after installing it. :) (Yeah, in practice it probably does that for you, but work with me.)
<nckhexen>winter: …or the script could suggest that (if it doesn't yet, didn't check). RTFM is always a good idea but so is giving users relevant hints at the right time.
<nckhexen>Although on second thought the number of hints it could print is quite large, and would probably turn into a quote from the manual anyway. Hm.
<winter>question: i'm more used to nixpkgs so bear with me, i know barely any of this translates ;P but what is the guix system command doing that makes it so the functions used (in this case, setuid) aren't just inherited from the tree? if that makes sense at all. like, is guix referencing some old internal function list thing?
<winter>(that was probably a horrible explanation)
<nckhexen>winter: It does. The short answer is: read the ‘building from Git’ chapter in the manual (sorry :) to actually build your local checkout. Then you can use ‘./pre-inst-env guix system image …’ (or maybe ‘guix system image -L . …’, but I've not actually tried that and it might asplode) to do what you want. But unless you've modified your checkout, the code in it will be identical to what you just ‘guix pull’ed.
<nckhexen>Long answer: so this is one of the ways Guix differs greatly from Nix (and most PMs): there is no ‘package repo’ distinct from the ‘package manager’ binary, like Nix's nixpkgs/nix(-env). Guix is a library that defines packages (just more Scheme code), not a parser for an external dataset. It's not quite a monolithic binary, but it's one coherent package.
<nckhexen>One reason it's so slow is that ‘guix pull’ doesn't just pull some data file from the Web. It fetches latest master and *builds* that, guix CLI and package definitions alike, and that new guix replaces your old guix.
<nckhexen>The fact that it's *this* slow is not good, and TODO, but it's expected to do some work, yes. There are other reasons for the first ‘guix pull’ to be even slower, but they are for your (security) benefit.
<winter>is `/gnu/store/ycsjc4a64icyscnn144lzcirya54n9n6-git.scm.drv` a copy of git, or some scheme module (i'm guessing the latter since the .scm suffix but idk)
<winter>that's one of the things it's not substituting, just curious :)
<winter>guix pull: error: You found a bug: the program '/gnu/store/049w7wqg3qap252m2n893izyxhkk9np5-compute-guix-derivation' failed to compute the derivation for Guix (version: "d520fdec3c7a11bc70ab83d0ebc6372389992d04"; system: "aarch64-linux"; host version: "1.3.0"; pull-version: 1).
<nckhexen>It's your one-stop-shop for single points of failure! 👍
<winter>maybe i'll use this downtime to start learning scheme, lol, since i have absolutely no clue what i'm doing :D
<nckhexen>I presume the Web services were working for the past few hours or someone would have complained. But can anyone confirm they used issues., ci., or www. then? (Savannah doesn't count, so neither does guix-install.sh working).
<nckhexen>winter: I apologise for this. This is our weak spot. It is known, but knowing isn't enough…
<nckhexen>Sigh. I don't think berlin's coming back from this.
*nckhexen puts away the stick of poking; goes to get the stick of whacking.
<nckhexen>OK, this is nuts. I can log in at the console (so Shepherd is not quite dead, just restin'), but it's so thoroughly confused that it's multiplexing a bash prompt and a(nother) agetty login: prompt at the same time, so the shell is useless.
<the_tubular>Yes, Andronikos there's a couple of services that are down
<winter><winter> I use Matrix for NixOS stuff, though, since that's our primary method of comms. <-- experience is iffy ime, which isn't fun :/
<lechner>Hi, my software builds manually in 'guix shell' but not in 'guix build'. Is there a way have 'guix build' operate on a local repo?
<Andronikos>Other question I had if someone knows is how I use gnome-keying within a container as of starting the service.
<nckhexen>Oh? Any other time I'd love to discuss this more, since people have been asking me about Matrix a lot recently, and Guix so far has not had any ‘official’ Matrix presence. Mind if I ping you about that some other time?
<winter>why does guix-packages-base take a while to build?
<winter>presumably it's just a bunch of symlinks or something?
<nckhexen>We use client certificates to authenticate to ci.guix. That's what's failing to validate. nginx claims it is expired. It is not. I haven't checked the chain. Hell, I don't remember the chain. Firefox's broken PKCS12 export isn't helping.
<jamesw>rekado: It is worked for the 'gcc' version ! ==> https://paste.debian.net/1258523 . would you kindly explain more to me? why is your command workable, but mine is failed ? and how to setup software development in guix geernally. I am newbie of guix. many thanks
<Andronikos>That is not a good first experience. Although funny enough Guix did not work either the first time since the ISO was broken. Guess I am good at finding problems.
<winter>Might be eval memory related like bjorn said.
<winter>If you'd ever want to give it another shot/debug that if you can still repro, feel free to hop over to #nixos :-)
<winter>since it doesn't look like i built that one
<Fare>Hi. Is there a tutorial for doing most of the guix installation over sshfs ?
<Fare>I managed to boot this computer with a grml usb stick and start ssh on it. I already have guix installed on a server with all the packages I want. Can I do most of the installation from the server using sshfs, then finish with a chroot command on the new machine after all the gnu store was populated?
<Fare>I suppose guix copy --to newmachine can help with the copy part.
*Fare uses guix pull ; guix system build newmachine.scm
<winter>so i built an ISO for aarch64-linux, GRUB loads, and... it says "invalid magic number" and then "must load kernel first"
<winter>after searching around, this seems to be due to corruption, but i verified that the image was transferred correctly from my vm
<winter>should i try rebuilding the ISO with --check (maybe rebuild everything...?), or is it more likely that something else is going on here?
<abrenon>any known caveat for compiling maven projects ?
<gabber>wpa-supplicant-configuration has a "config-file" field which expects a file-like object. do i pass a string (representing the path to a file) there? do i pass a (local-file)? do i have to gexp/ungexp some part of that config? i've tried several combinations but don't seem to get it right.
<examors>gabber: local-file should work - that is a file-like object. plain-file should also work if you want to use a string literal rather than an external file
<gabber>so i have something like (local-file "./foo.conf") where foo.conf is in the same directory than the system-config i'm reconfiguring my system with? or do i need absolute paths?
<Korven[m]>Hello! This is just an introduction message. I am Korven (was here once before), and I'm giving Guix another shot now that I know more about Scheme and Guix substitutes and mirrors. Nice to meet you all!
<gnucode>Korven[m]: welcome! guix is sooo much fun to play with! It's probably been the most stable distro, that I've ever used. Glad to have you on the playground!
<Korven[m]>Well I'm still pretty much what you'd consider a newbie, just trying to make my way around. My first order of business is definitely finding the right substitutes, so I can get some semblence of speed
<gnucode>Korven[m]: I'll send you a link to my config in just a second...I use a very minimal sway configuration, so that my updates are normally pretty quick.
<gnucode>and my config has some substitute servers defined. :)
<Korven[m]>i hope i'll be able to tell people that one day
<gnucode>Korven[m]: Finding proper hardware that will work well with guix system can be tricky. I'm a RFY kind of guy, so I've got a os-booted thinkpad. There are some open newer RYF hardware that you can buy, but I either seen options that are too expensive, or not powerful enough. At least for me.
<Korven[m]>Everything electronic is exppensive in my country so
<apteryx>hmm, some systemd-related warning from udisks: udisks-WARNING **: 08:16:46.114: Failed to setup systemd-based mount point cleanup: Failed to execute child process ?systemd-escape? (No such file or directory)
<apteryx>abrenon: yes, there's no CLASSPATH automatically defined for one thing, so it won't find any deps
<Korven[m]><gnucode> "Korven sorry bro. :(" <- why XD it's not your fault
<Korven[m]>oh it's an older commit that what I'm currently at
<Kabouik>Anyone using ESS (emacs speaks statistics) with R on Guix?
<Kabouik>Whenever I try to start the R process in an ESS window with M-x R RET, the ESS prompt to confirm the working directory converts accented characters to escaped characters, and I can't confirm with RET unless I change that back to accented characters. I have tried some things advised in #emacs but I feel there may be peculiarities related to emacs itself. $LANG is set to en_GB.utf8, $LC_ALL is not set.
<Kabouik>If I try the emacs daemon with LC_ALL="C.UTF-8" and ignore my .emacs file, then this issue in the prompt is fixed, but I can't see what in my config may cause the problem. And also R throws a warning about the set C.UTF-8 when started this way (and uses just "C").
<Kabouik>I'm feeling it should not be necessary to start the emacs daemon with a prefix env variable anyway, so there may be something guix-specific that I did not set up correctly.
<Korven[m]>What's the difference between `guix package --install` and `guix install`? is the latter just an alias of the former
<Korven[m]>Okay so, the fonts are found by `fontconfig`, but when I try to set them in `gnome-tweaks` in my host distro, it can't find them
<abrenon>I wonder what it means by 'module java.base does not "opens java.lang"' ?!
<nckhexen>apteryx: Ah, you recently reconfigured. Maybe that reset something.
<winter><winter> so i built an ISO for aarch64-linux, GRUB loads, and... it says "invalid magic number" and then "must load kernel first" <-- going to debug more later, but does anyone happen to have an idea as to what would cause this? to clarify: this wasn't cross, and GRUB can see the contents of the store on the ISO just fine
<apteryx>abrenon: hmm. perhaps have a peek into the maven-build-system for clues
<mothacehe>apteryx: yeah the commit pointed out by jonsger introduces a guix-other-arch specification so that the guix specification doesn't fail when aarch64/powerpc guix evaluation fail. As a side effect, the powerpc architecture is removed from the master specification. I think we can restore it.
<nckhexen>That's why I rebooted berlin: ‘sudo herd restart nginx’ froze and the machine stopped responding and accepting new SSH connections. Had to hard-reset. Unfortunately, AFAICT, nothing was logged.
<nckhexen>It's in the IRC logs if anyone cares, but I don't really have anything useful to share with sysadmin@.
<nckhexen>But if anyone noticed: ‘you found a bug’ reports rolling in is still a reliable indicator of berlin being down for half an hour :-/
<abrenon>yeah, I was starting to wonder if packaging the repos with this build-sstem wouldn't be easier than trying to make it work from the command line but I still lack experience with it so ^^
<Kabouik>Any idea why `LC_ALL="en_GB.utf8" emacs` works fine when not connected to a daemon, but `LC_ALL="en_GB.utf8" emacs --bg-daemon` then `emacsclient -c` (or even `LC_ALL="en_GB.utf8" emacsclient -c`) would break accents in prefilled paths in prompts (I have just observed that in the ESS prompt to chose working directory when starting a R process, the prefilled path is broken if it contains accents)?
<Kabouik>I have discussed in #emacs but it is not clear yet if the issue is Guix-related or not. In my Guix config, the locale is set with (operating-system (locale "en_GB.utf8")).
<Kabouik>I only get this issue with daemon/client.
<nckhexen>Do you get it only when forcing LC_ALL like that? (I use en_IE.utf8 system wide and can't reproduce that.)
<Kabouik>No, I started trying to force LC_ALL because I got the issue before prefixing the env var. Actually I should try again with no daemon, see if the issue exists in a non-client emacs and no env var
<nckhexen>(That is, by opening ‘Héllø Señor’ and then reopening the CxCf prompt. I assume that's what you mean by ‘pre-filled’.)
<nckhexen>After learning that sneek's feeding status is kept only in RAM, not saved, these recent ‘crashes’ are clearly just a ploy to score fresh botsnacks.
<Kabouik>I'm lost, I can no longer reproduce the behaviour whereby it would work with a non-client emacs now, with or without the prefix var; meaning it's not working for both daemon mode or single instance mode now.
<Kabouik>nckhexen when you edit a R script with ESS, and hit M-x R RET, it starts a R process in a new emacs window, and offers to set the working dear for that R session. It prefills the path with the path of the R script you're editing in ESS. If that paths contains accents, they're replaced with some escape codes (and I can't hit RET to confirm the path, I have to remove the escape chars and place accented characters where they belong).
<Kabouik>I don't get the same issue when opening a file with accents in its path, and then doing CxCf. There the accents are not touched.
<Andronikos>If I customize my kernel and do not have for example IP forwarding activated which is required for Docker. Will Guix tell me that upon installing Docker?
<nckhexen>Kabouik: Is there a persistent R repl(?) process that keeps running, that might explain why your experiments aren't consistent between emacs restarts? (One would expect it to be killed, but who knows. I don't.)
<Kabouik>I think the R repl gets killed when I close emacs and make sure there's no daemon still running, and i couldn't find R in the running processes after killing emacs.
<nckhexen>That's what I'd expect. I'm just confused by the reported inconsistency.
<Kabouik>This may very well be an emacs-ess issue, but this could also be a Guix issue, and I'm still investigating to find which IRC channel I should pollute (well, as you might have observed, I started polluting both just in case D:).
<Kabouik>By "issue", I mean it could also be something I just failed to set up correctly.
<Kabouik>See, that's what happens on M-x R RET: https://0x0.st/o35q.png (and you can see accents working just fine in the buffer, I can also type accents in the prompt (and I need to replace those blue escape sequences by accents, actually, or I can't confirm the path)).
<Kabouik>Ha! If I `emacs /path/to/file/with/accents` and then M-X R RET, no accent issue in the prefilled path when starting R. If I `emacs` then `counsel-buffer-or-recentf`, the accents are borked. Pretty sure the issue is related to how counsel stores the list of recent files (or any other package doing that). Now at least I know a little bit better where to look. That might explain the inconsistency reported above too, nckhexen.
<Andronikos>How can I run gnome-keyring inside guix shell -FC? If I do something like "guix shell [..] -- dbus-update-activation-environment --all gnome-keyring-daemon --start --components=secrets <program-to-run" dbus yells at me with "dbus-update-activation-environment: error: --all cannot be used with VAR or VAR=VAL arguments".
<unmatched-paren>Andronikos: Ah. No, you can't, afaik. That would be a nice feature though.
<lechner>Hi, does the Guile's autoconf macro GUILE_SITE_DIR work in Guix? I think it points to the Guild's site directories instead of the "installable" site directories below --prefix, i.e related to my package's output. Is there a better way to determine the directories for installation?
<lechner>lilyp: Could we avoid the confusion if we did not hardcode 'prefix' into the other paths in the pkg-config files we ship. We could then use the pkg-config option --define-variable=prefix=[configure-prefix] to derive the target directories when needed. https://paste.debian.net/1258625/
<vagrantc>e.g. nckhexen found "This EMACS packages" which seems like a positive
<tribals>I'm wondering how to make `guix publish` to listen on all interfaces including ipv6 ones?
<vagrantc>lechner: yeah, it's finding them in all the guile modules ... i've fixed probably over a hundred typos in guix so far that seemed legit, so it's been useful ... 3-5 false positives are no big deal
<vagrantc>so... is there some way i can get: guix build --keep-going PKG1 PKG2 ... PKGN ... to report the successes and failures? it seems to end with the last buidl that failed, and builds that suceeded end up lost in scollback
<tribals>lecher: another question is: does `guix publish` support multiple `--listen` option to work-around this?
<lechner>vagrantc: how about This LaTeX packages in texlive-latex-atveryend ?
<euouae>I used the install script but ^C'd it and now if I run it again I get that a guix installation is already found
<tribals>I dont' want to loose it listening on ipv4
<euouae>How can I uninstall guix and re-install it?
<vagrantc>i end up iterating through all the build failures and removing them and trying again, and eventually it produces a list of successful builds, and then i know which failed by which i'm excluding
<nckhexen>You can enable that, and ‘::’ will no longer be magic.
<tribals>Another question is: I noticed I need to `herd restart` some (all?) services after `guix system reconfigure`. Is this supposed to be this way?
<nckhexen>(‘::’ is to IPv6 what ‘0.0.0.0’ is to IPv4, nothing magic. What's magic is Linux defaults to ‘dual stack’ mode, meaning that binding to ‘::’ will automatically bind to ‘0.0.0.0’ as well—assuming you don't enable the above option.)
<nckhexen>tribals: Some services can't be safely restarted. Whether the ones you need to restart all fall under that category, or are just TODO, who can say.
<mbkamble>Hi. I am lost trying to achieve the following: I want to create a link to libstdc++.so.6 in /lib64. libstdc+.so.6 exists in lib output of gcc (/gnu/store/blah-gcc-10.3.0-lib/lib/libstdc++.so.6)
<mbkamble>I have written this in my system.scm file:
<mbkamble>this results in a dangling link because file-append resolves to "out" output for gcc. If I give (file-append gcc:lib "foo"), it gives undefined variable errorn for "gcc:lib"
<mbkamble>I raised this help request yesterday, but had to leave the chat midway, and did not get a solution
<tribals>Another question is: I rolled system back to previous configuration. I can verify config for desired service is updated. But changes did not reflected in herd service at all - it still uses old one. The service I'm configuring is Yggdrasil, to be concrete.
<tribals>May be there is `herd reload` or something similar?
<nckhexen>There is, but support for it is ad hoc and incomplete (that's what my ‘TODO’ above was referring to).
<nckhexen>tribals: But ‘restart’ should *always* imply reloading the configuration you just reconfigured. Otherwise: bug.
<nckhexen>Andronikos: I can't reproduce it so it was probably a fluke, but thanks!
<Andronikos>I get libVkLayer_MESA_device_select.so: cannot open shared object file: No such file or directory on using vulkaninfo. I have seen some tickets about this. Do I just need till https://issues.guix.gnu.org/58251 is closed?
<tribals>nckhexen: I've restarted yggdrasil service after switching generation a couple of times already. Changes didn't reflected
<Andronikos>nckhexen: I tried reloading it a couple of times and now it works for me again too. Weird.
<nckhexen>tribals: I don't use or know yggdrasil, but high level, that sounds like a bug. *If* you're certain that the changes were applied.
<nckhexen>Andronikos: Yeah, I didn't mean it's not a problem, and in an ideal world someone would investigate and fix it, but…
<mbkamble>so, it seems the output can be specified with string-append, but I don't know how to use that to create the symlink
<nckhexen>Andronikos: That was about the 50x. About mesa: if the only way to fix that is to update mesa, then you'll have to wait for <https://issues.guix.gnu.org/58566> to be applied *and* for the core-updates branch to be merged into master.
<tribals>nckhexen: will continue exercises later and report
<apteryx>ah! define-configuration records are not API-compatible with previously manually-defined ones
<apteryx>so (match (($ <mcron-configuration> mcron jobs log? log-format) ... )) ends up matching %location as mcron...
*apteryx puts the %location field at the end of the record, to make them compatible
<vagrantc>heh. first time i saw larger reproducibility % for i686-linux than x86_64-linux today
<podiki[m]>hrmm...so the --development option for guix shell adds the implicit and explicit inputs for a package, including build system inputs, so that means many would include glib this way, right?
<podiki[m]>tracking down a bug with --emulate-fhs, I think the --development option adding a glib overwrites the glibc-for-fhs that is included, because it comes later in the package list....(I can reverse the list and it changes which glib is there)