IRC channel logs
2024-05-15.log
back to list of logs
<mmontone>guix pull is taking a long long time. Hope my package install works after :) <podiki>if you were from 1.3.0 that is a few years ago, so yeah, that's a lot of commits <podiki>guix is rolling, so a "release" is more like an installer snapshot really <podiki>no one is expected or encouraged to remain on it, and we have time-machine or commit pinning to control what guix you are on if you need or want it <podiki>(this is something that should be explained on the website and documentation but i don't think is clear, what we mean by a version number) <podiki>(by "should be" i mean it isn't if i remember) <Kolev>podiki: What use is the released version? <Kolev>Isn't 1.4 so old, the loading bars are ####? <podiki>supposed to be a well-tested installer and good state of Guix (no major packages broken, substitute coverage good) <mmontone>xz-5.2.8-static 96KiB 110.9MiB/s 00:00 [##################] 100.0% <mmontone> xz-5.2.8-static 96KiB 82.3MiB/s 00:00 [##################] 100.0% <mmontone> zlib-1.2.13 79KiB 124KiB/s 00:01 [##################] 100.0% <mmontone> zlib-1.2.13 79KiB 74KiB/s 00:01 [##################] 100.0% <mmontone> gcc-11.3.0 25.7MiB 1.7MiB/s 00:16 [##################] 100.0% <mmontone> gcc-11.3.0 25.2MiB 1.8MiB/s 00:14 [##################] 100.0% <Kolev>Ei ei ei, mmontone! Use a pastebin! <podiki>could be different hashes (needed by different things), could just be a display thing <Kolev>podiki: I sigh when I see a command line tool like Pandoc bring in something like gnome-backgrounds. <podiki>i see no path between them.... via guix graph --path <podiki>so i wonder why you are seeing that <Kolev>podiki: My memory is likely wrong. I do remember seeing gnome-backgrounds getting pulled for something that ostensibly should not use it. <podiki>i know there are some gnome things that end up in unexpected places, or at least i've seen in the past <podiki>perhaps via things like inkscape that is used in building packages but depends on gtk and such <podiki>so you might see something that is just images/icons/etc. but needs inkscape or other tools for building different formats <podiki>if you need to build you'll see that, but if it is substitutable it shouldn't bring in those build dependencies <podiki>or x stuff via python, used in a meson build for something unrelated to either of those <freakingpenguin>Kolev: One theoretical benefit of tagged releases is it makes finding compatible commits between channels that depend on Guix easier. If every channel has a v1.4.0 tag, they probably work together. <Tadhgmister>how do I get `guild compile` to return a non 0 code when it has any warnings? I'm trying to setup a development system that compiles all my scheme modules so if any fail I get the most relevant error messages but it doesn't work if "warning: possibly unbound variable `foo`" doesn't stop at that step <Tadhgmister>that would be a good idea, I have a workaround where I pass the stderr to grep and if it finds anything I `exit 1` but I'd really like a nicer solution :) <apteryx>Tadhgmister: sounds like a bug to me <cnx>is it possible to use cuirass as a generic CI scheduler, i.e. building guix "package" only for tests and possibly without sand boxxing for cd <apteryx>not sure about the last part of your sentence, but for the first part I think so, yes. <Tadhgmister>how do other people develop with a guix channel? Like when you make changes to your code how do you compile it so you get useful error messages instead of going right for `guix home build -L ./mychannel home.scm` to load files from the channel directory <Tadhgmister>I started a journey to use Tup to compile the files in the channel so it could error out on the failing ones but apparently when you go to a .go file the `local-file` no longer gets the directory of the source file so everything becomes relative to the cwd and crashes down <apteryx>I think I'm done with this sysdig/BTF thing... will submit soon <KE0VVT>What do you think of run0? And is there a way to use doas in Guix? <efraim>is there a copy of your OS config available? <efraim>cbaines: nice! before I was just going through in the order they appeared <KE0VVT>What MODERN laptops would you recommend for linux-libre? <thanosapollo>I dont think any modern laptop can support linux libre oob, but tbh most of them feel like a scam <cbaines>I'm using a Framework with an 11th gen Intel CPU and a replaced WiFi card, and it seems to work fine <efraim>pinebook pro works, but it's not that powerful and needs an external USB wifi card <KE0VVT>What about the ThinkPenguin laptops?. <KE0VVT>I'm guessing Guix doesn't work well on Fedora or Arch. Should I just use Debian or Trisquel? <wigust>KE0VVT: i would use guix directly if it's not a question of "buy 10 laptops and do not touch them". Other wise yes nscd is recommended <efraim>do we have a way to override a package in guix from an os config? I thought it was possible with set! <cbaines>it's probably a better idea to rewrite package inputs rather than use set! <Shaq>Error raised when I config kernel with `CONFIG_DEBUG_INFO_BTF=y`. <Shaq>Can someone here give some help on this ? <Shaq>(kernel (customize-linux <Shaq> #:linux linux-libre-with-bpf <Shaq>cbaines, Sorry, I didn't get you. <Shaq>I don't know the relation between debain & this issue. <cbaines>Shaq, IRC is for talking, if you want to share your config, please share a link <cbaines>paste.debian.net is a recommended way to do this <Shaq>cbaines, got it, thanks. <Shaq>cbaines, If I change `"CONFIG_DEBUG_INFO_BTF=y"` to for example `"CONFIG_MT7921E=m"`, then it can pass. <apteryx>Shaq: usually that means dependencies haven't been satisfied <apteryx>check with 'make menuconfig' or by looking at the KConfig files in the tree. <apteryx>the easier way to use customize-linux would be to pass it a defconfig produced by 'make savedefconfig' <apteryx>then you'll be sure the dependencies have all been resolved correctly <Shaq>apteryx, thanks for the information. <Shaq>I have posted logs, seems it's an internal error: `Error in reading or end of file.` <Shaq>[8:09:27 AM] <apteryx> Shaq: usually that means dependencies haven't been satisfied <Shaq>[8:09:34 AM] <apteryx> KConfig dependencies <Shaq>[8:09:53 AM] <apteryx> check with 'make menuconfig' or by looking at the KConfig files in the tree. <Shaq>I tried to keep failed & 'make menuconfig', but enviroment error: missing Unable to find the ncurses package. Even I have installed ncurses <ekaitz>hi! can i `modify-inputs` but using a name for an input? <ekaitz>is the only option to add the inputs like with alist operations? <cbaines>ekaitz, isn't that how modify-inputs works? <cbaines>the delete and replace operations both use package names <ekaitz>maybe I should have mention that i wanted to append a package using a name that is not the package name <futurile>ekaitz: yeah, append receives a package. So you want to go from text to a package reference: uh specification->package maybe? <ekaitz>i want to add a package, in my case musl, but as "libc" <ekaitz>so `(("libc" ,musl) ,@(other-inputs)) <ekaitz>is there a cleaner way to make this? <cbaines>could you (package (inherit musl) (name "libc")) and add that? <ekaitz>i want to do this because I want to (replace "libc" other-things) <ekaitz>cbaines: that's an option, too, but I need to do that many tiimes <ekaitz>(i'm working on commencement.scm) <cbaines>I guess the named packages in inputs is going, and has been going for a while <cbaines>I'm not sure how best to do what you're trying to do though <ekaitz>yeah but in that case why does replace depend on the package name? <ekaitz>it's like the names are not being used but they are at the same time <apteryx>sham1: 'guix shell -D linux-libre ncurses' will give you a minimal environment for 'menuconfig' <dariqq>if something started via a shepherd service does not create a pid-file, is there a way to let the shepherd create it? Or does the application need to create it themselves? <apteryx>dariqq: the application needs to create it itself <apteryx>shepherd uses that for synchronization, it can't create it itself <apteryx>otherwise you're probably in need for another type of constructor <apteryx>ooh, 'make check-system TESTS=basic' passed with my kernel zstd-compressed modules <apteryx>sham1: oops, I meant to notify Shaq, but it seems they left already <sham1>apteryx: I didn't even notice until this second notification :P <apteryx>dariqq: there is a systemd-type-constructor IIRC <civodul>dariqq: there are three types of service “constructors”: fork+exec (with or without PID file), inetd, and systemd socket activation <dariqq>ive only used the forkexec constructor so far, but will play around more <dariqq>was asking because the start-stop-daemon in openrc can optionally create a pidfile if the process does not <civodul>looked like Guix proper just went over the 30k package bar <civodul>(too bad ci.guix is at 58% for master 😱) <civodul>‘guix weather’ says 97.2% on x86_64-linux though <civodul>so it’s really the other archs that are in trouble <dariqq>i think i might not need the pidfile (and a while true loop in a script) after all. The autorespawn seems to work as expected :) <apteryx>civodul: repology says we are at 27358, number not too far behind FreePSD <apteryx>$ make check-system TESTS='basic loadable-kernel-modules-0 loadable-kernel-modules-1 loadable-kernel-modules-2 loadable-kernel-modules-service-0 loadable-kernel-modules-service-1 loadable-kernel-modules-service-2' just passed with zstd-compressed kernel modules <dariqq>speaking of services can I kindly ask someone to take a look at my service in #70460? <apteryx>cbaines: did you see the "gnu: nss/fixed: Hasten test suite." patch? <cbaines>apteryx, I did, and the change seems reasonable to me <apteryx>OK; let's wait to see what QA thinks <cbaines>Christina O'Donnell seems to be more involved in tinkering with nss, so it would be good to check with them <cbaines>at least I haven't had a chance to try and read all of there mails on the subject <apteryx>cbaines: what's in your os 'packages' field (nss-certs? specification->package ...?) that breaks the nss certs? <civodul>Christina discussed NSS_CYCLES and another variable on guix-devel <apteryx>the patch is just that; setting NSS_CYCLES to standard <apteryx>cuts the test time by three (each cycle re-run the whole test suite with different settings) <instance9>oh interesting, what is NSS usually used for? <civodul>apteryx: in the discussion Christina concluded that it’s not clear whether it’s “safe” to do that <cbaines>apteryx, I added (packages (cons* nss-certs %base-packages)) to my operating-system, and that seemed to break things <cbaines>which suggests the operating-system-packages thing for removing duplicate nss-certs isn't working, since adding nss-certs on top of %base-packages should be a no-op, right? <apteryx>cbaines: like it reconfigures fine but leaves you with an empty certs directory? <apteryx>right, it should keep only one nss-certs package, which should be the most recent <apteryx>OK, I'll try it in a minute and see what's going on <apteryx>has anyone tried 'mumi new && mumi send-email *.patch' recently? <apteryx>to send a multiple-patches series, letting mumi handle fetching the issue ID and sending the series there? <apteryx>(after producing the patches with 'git format-patch' or 'patman -n') <cbaines>apteryx, great, I've just reconfigured with a duplicate nss-certs, and I still see the issue: ERROR: The certificate of ‘guix.gnu.org’ is not trusted. <apteryx>OK! easily reproduced bugs, that's how we like them <PotentialUser-3>I'm once again asking if there is anything blocking wlroots being updated to 0.17.2; my WM keeps crashing and I can't do anything about it without the update. <apteryx>PotentialUser-3: would you like to prepare a patch? <apteryx>from a built guix checkout: './pre-inst-env guix refresh -u wlroots && etc/committer.scm' may be enough <PotentialUser-3>Fair enough, guess I should learn how to actually contribute to Guix. <apteryx>seems like the 'mumi send-email' trick worked; it took 6 minutes to get the ID back <PotentialUser-3>Minor nitpick: docs/guix.info does not exist, contradicting the README. <apteryx>sneek: later tell Shaq, if you were looking to try a BPF/BTF supporting kernel, you may be interested in applying/trying out #70962 <jpoiret>PotentialUser-3: It does exist after running the first `make` <jpoiret>it's generated from the texinfo files <PotentialUser-3>So the documentation that tells you to run make is only available when you run make? <jpoiret>PotentialUser-3: you're right, the README doesn't mention any of this <jpoiret>usually people consult the web version of the documentation <Zambyte>Hey, I'm running into a weird issue right now. I am trying to run `guix system init ./config.scm /mnt` in order to insall my config from a live boot, but it is not installing everything. It scaffolds the root, populates /mnt/gnu/store, but it does not create things like my users home directories, and does not create /mnt/bin/sh. Has anyone run into this before? <Zambyte>the init command is completing without error and it is installing the bootloader. I can't boot into the live system though (due to what I believe are unrelated issues with my RAID configuration) <jpoiret>Zambyte: i believe /bin/sh is populated on boot. The home directories, I'm not too sure <Zambyte>jpoiret: hm, I don't think it is populated on boot. I am unable to follow the documentation on chrooting into the new system because there is no /bin/sh to use. If I manually create a symlink from /mnt/bin/sh to /gnu/store/hash-bash-version/bin/sh I can chroot in, but not much works. It also doesn't populate /mnt/etc with symlinks <jpoiret>don't you think it might be an issue with your RAID configuration then? <Zambyte>Well the RAID is mounted at /mnt, so guix system init can at least write to it (the store is populated). The problem I'm seeing with my RAID is post boot if I just try to boot it anyways it says it can't find /dev/md0. I have /dev/md0 configured in my live boot though, so I should be able to at least install to it even if I can't boot from it <Zambyte>I also have (initrd-modules (cons* "raid1" "raid0" %base-initrd-modules)) in my os config since that seems to be required for RAID 1 on root <jpoiret>civodul: wow! I'm planning to push the glibc update today, so that we can start working on it again <futurile>nice - once the glibc goes out - should we put out another call for people to test? <jpoiret>probably, although we might want to wait for substitutes to build first <jpoiret>I haven't tested anything there myself <civodul>yay, looking forward to starting that test cycle! <lispmacs[work]>hi, I'm trying to build the latest guix so I can test a patch. But it dies with this makeinfo error: <lispmacs[work]>wondering if that is an actual bug or something messed up in my environment <PotentialUser54>Is anyone here able to offer any advice on how to debug pytest errors when packaging Python software? The build log doesn't give me an actionable error. All I get is "%exception #<&invoke-error program: "/gnu/store/glrxjwic379h2gmyxc2sfchvnw5g1aai-python-pytest-7.1.3/bin/pytest" arguments: ("-vv") exit-status: 5 term-signal: #f stop-signal: #f>" <ieure>PotentialUser54, `guix build -K your-package-name' -- There'll be a "note: keeping..." line in the output, which has the path to the build directory. Then `guix shell -D your-package-name --pure', cd into the build dir, manually run whatever thing is failing. <ieure>PotentialUser54, However, I will say that in almost every case of a build failure, the build logs do contain enough information to debug -- it can just be difficult to see. Read up a bit higher, and I suspect you'll find more relevant info. <PotentialUser54>Oh. The version uploaded to pypi didn't include the tests directory. <PotentialUser54>I'll try updating my package to pull from github since I do see the tests directory exists there