IRC channel logs
2023-05-30.log
back to list of logs
<elevenkb>oriansj: just found out that emms can stream from YouTube. <elevenkb>thanks for that, I can just add my favourite channels. one step close to de-googling! <oriansj>well we can't expect everyone to go into using elpher <elevenkb>oriansj: yah that and job application sites. <oriansj>elevenkb: well employment and the social organization behind it is more of a topic for #guix-offtopic <elevenkb>oriansj: gotcha. do you know how much time it takes you to rebuild home and your system from scratch? <elevenkb>dcunit3d: [whoosh] i didn't get the joek. <oriansj>elevenkb: the core is done in about 8 hours if all built from source; but rust takes about a week <crzjp>is looking for drivers on /gnu/store/...../xorg.conf.d instead of .guix-profile/.../xorg.conf.d, where the config for drivers reside. is there a way to make xorg look for that location? <lemos1898>i am testing some things here, is there some reason guix is not being recognize in guix shell and running with ./pre-inst-env? <lemos1898>i used this a while ago for testing the same package, with guix package or guix build, nothing so far nowadays <lemos1898>./pre-inst-env: line 55: exec: guix: cannot execute: Is a directory <lemos1898> mostly that, with the guix dir that contains some configure stuff <lemos1898>running guix shell with guix shell -D guix --pure <dcunit3d>if it's failing on the exec, it may be running into some problems with library dependencies <dcunit3d>i'm not sure the make clean && make clean-go are required <dcunit3d>if you're really running before install, then i'm not sure what you need for --localstatedir <lemos1898>need to update a package for someone, but really running in circles for some time already <lemos1898>determined to finish it, since some guix users are without the updated package :/ <dcunit3d>it's likely the artifacts created by the ./bootstrap or ./configure script, but it could be other build artifacts. it could also be the source tree. <dcunit3d>are you packaging something within the guix source tree or outside of it? <dcunit3d>and i think if it's not finding the guix exectuable, then the pre-inst-env may be built with other configure arguments or another checkout. it should be finding the ./scripts/guix executable. <lemos1898>dcunit3d, "outside" you mean, common packages? <PotentialUser-68>Hi, I'm trying to build a package that depends on OpenGL and it builds fine but the OpenGL rendering is broken. <PotentialUser-68>According to ldd's output, it's linked correctly but if I'm reading the strace output properly, it's looking for the GL libraries under the glibc directory. <iyzsong>PotentialUser-68: does it linked to libGL (from mesa)? most programs will try to dlopen it at runtime, but in Guix the only GL implementation is mesa, so it's usually need patched to link at build time (by add '-lGL' to LDFLAGS). <PotentialUser-68>Yeah, if I'm reading the log from the configure step properly, -lGL is there. <iyzsong>um, that's wired. i'll check the binary, to see if ldd output libGL, and the source, to see if it dlopen libGL. <PotentialUser-68>I'm getting "libGL.so.1 => /gnu/store/8cw5z3z356h3cppvdxhnmzd5hi7mdzzx-libglvnd-1.5.0/lib/libGL.so.1" so that seems about right. <iyzsong>okay, that's the thing to see. libglvnd is a dispatch library, i think it need some config (via env var) or will lookup real libGL in some fixed path. <PotentialUser-68>Oh, is there a way to lookup which packages use libglvnd? Maybe there's an example there then. <iyzsong>'guix refresh -l libglvnd' or grep libglvnd, i didn't find useful examples though. seems libglvnd is not what i think it is... <efraim>and then the cuirass jobs should start working again? <civodul>well, to be clear, the bug has an impact in that we ended up with two Cuirass instances running <efraim>I might experiment with using mkdtemp and symlinking it to ~/.cache/pip <jpoiret>lemos1898: did you `make` before trying to run guix from the pre-inst-env? <jpoiret>you're missing the guix script probably, which make should create <janneke>jpoiret: headsup, gotten a lot further to get native compilation working <jpoiret>janneke: nice! I absentmindedly followed your progress yesterday, just wasn't available <jpoiret>i expect to have more time this week <janneke>jpoiret: great! i mainly changed git-fetch to url-fetch using savannah tarballs, updated commencement to use these new sources together with auto*-boot0 <jpoiret>you mentioned the gnulib tests, is this an upstream bug? <jpoiret>also the bootstrap binaries, you talked about updating them, right? <jpoiret>i don't think updating the bootstrap binaries should be too bad for an experimental system like hurd <janneke>jpoiret: re gnulib, i'm not sure; it could be that the bulid on gnu and riscv are wrong, or that qemu has a bug... <janneke>jpoiret: the bootstrap binaries that was a picnic/red herring <jpoiret>bost: yes, it's a bug with bordeaux not honoring its TTL for some nars recently. You could disable substitutes for that package only <jpoiret>janneke: ah, something very fun to debug then! <janneke>we were using a very old source for hurd-headers-boot0 and hurd-minimal-boot0 that gave surprising results <janneke>so yeah, that's a relief, as we'll have to update them (most probably also for i586) when we switch to upstream's 64bit support <jpoiret>i was thinking we could use the url-fetch only for the bootstrap packages, not the full ones, so that one can still use --with-latest <janneke>ACTION was wondering wether to do the one or the other, but this is a helpful perspective <janneke>it also means that these native-build fixes won't even potentially affect your patch set anymore, which is great <civodul>for now, you can work around it by removing bordeaux.guix.gnu.org from your substitute URLs <cbaines>clearning the cached narinfos should work as well <jpoiret>cbaines: ah, that's probably the better solution <bost>cbaines: does that mean `rm -rf ~/.cache/guix/substitute/` ? <cbaines>bost, I think the daemon's cache is maybe more relevant here (/var/guix/substitute/cache/) but there shouldn't be any harm in removing your users cache as well <bost>cbaines: I did `sudo mv /var/guix/substitute/cache/ /var/guix/substitute/cache.delete-me` and now the `guix home ...` is able to download the xfce4-panel-4.18.3.tar.xz w/o crashing. <bost>I guess I don't need to keep the /var/guix/substitute/cache.delete-me around, do I? <cbaines>civodul, looks like it, I also lowered the threshold to start things slowly <cbaines>I also made submitting builds for patch series conditional on master branch substitute availability, which is poor at the moment <cbaines>you can manually prompt the qa-frontpage to submit the builds anyway though (through the submit-issue-builds argument) <Kabouik>Am I the only one seeing very high CPU use when running telegram-desktop? On my end running telegram-desktop is a good +20°C on my CPU, and significant fan noise (I share my office and tend to *not* telegram-desktop at work just because of that). <Kabouik>I plan on trying packaging nchat one day (a TUI for TG), but that's for the far future. :< <civodul>cbaines: like "guix-qa-frontpage submit-issue-builds 63642", as user qa-frontpage? <cbaines>civodul, exactly :) but also with the working directory as /var/lib/qa-frontpage/ <cbaines>also, guix-qa-frontpage isn't on the path, so you'll need to provide an absolute filename <cbaines>civodul, if you add --priority=550 that might work around the issue <civodul>seems to work! "build submitted as c662b02e-7963-4244-8d69-24cc448846bf" <civodul>"1657 target derivations for issue 63642" <civodul>is it submitting each derivation one by one? <cbaines>the qa-frontpage isn't, but the build coordinator is <cbaines>there's no interface to submit builds in bulk, you have to do it one at a time <cbaines>although the coordinator does create builds for derivation inputs without builds (recursively) <cbaines>maybe at some point this will become a bottleneck, but I think we're quite far from that <gabber>how can i install a file onto an (operating-system)'s root directory? <elevenkb>gabber: i'm a n00b but still curious and eager to help. why are you considering doing that? <gabber>i am working on a project where this is expected. i guess extra-special-file or the special-files-service-type are the way to go. but what if i need a real file, not just a symlink? <next4th>gabber: activation-service-type can do that <gabber>activation-service-type copies that file during the boot process to /, right? i guess i need it present before the kernel is actually loaded (think firmware). do i need to hack the image-creation function? <efraim>I guess you could look into the initilization script for the partition that has the root filesystem on it <gabber>efraim: that sounds promising! where/how would i find that? <efraim>gabber: gnu/build/image.scm for the actual initialization script <efraim>gnu/build/image for where it gets called from <efraim>it is, it looks like it might both be in that file <efraim>ok, the second one I meant gnu/system/image <efraim>gnu/system/images/ should have some examples of customizing partitions <flypaper-ultimat>hey guix! is there any C library (as opposed to executable) thats a good minimal example to follow of how to do all the autotools things so that its the easiest for guix? <MatoHota-work>Running a guix shell command it seems to produce what I'm expecting : all the packages I have asked for are present. But when I try to execute one of the binary I get a "No such file or directory" <MatoHota-work>but a file on the same exe I get the expected description : ELF 64-bit LSB executable.... <cbaines>MatoHota-work, that can happen if the dynamic linker used by the binary isn't present <Guest49>hi guys, I'm looking for a some advice or wiki link on how to install guix OS alongside an arch distro that is btrfs based. I have 3 partitions: /boot/efi, / and swap. My plan is to resize / and leave un-allocated space. Please be so kind as to point me in the right direction for my next steps <MatoHota-work>I need to package an already built Ada Compiler (no bootstrap possible) <MatoHota-work>I have twicked the provided installer in order to be able to use a gnu-build-system <MatoHota-work>I have put coreutils and file in my shell to get ls and file to be able to use them and they run fine <cbaines>MatoHota-work, getting arbitrary binaries to run on Guix is offtopic, but the patchelf program may help you here <cdo256_>Hi guix! I'm trying to add an argument to pam_env.so. I have succeeded in doing this to all (base-pam-services), but "login" is added afterward. <cdo256_>I notice there's a transform and an extend-configuration, but that doesn't appear to be used in the operating-system definition. <cdo256_>transform and extend-configuration are in pam.scm, but not exported. <cbaines>civodul, cool, thanks for the info, looks like I need to update some code regarding build cancellation <cbaines>all the builds for that issue should have been submitted though <cbaines>civodul, also, I've merged the squee changes and I'll update the package shortly <cbaines>(also looking at updating guile-fibers) <PotentialUser-22>hey, guix upgrade fails and complains that it can't download substitutes <cbaines>PotentialUser-22, you can work around that by clearing the guix-daemon's substitute cache (e.g. sudo rm -r /var/guix/substitute/cache/ ) <civodul>cbaines: re squee, very nice, thanks! <civodul>cbaines: this should allow us to get rids of all the hacks; in Cuirass there's database service running in its own thread, with a watchdog checking whether it's still alive, etc. <cbaines>civodul, right, I haven't looked at Cuirass since the change from SQLite but the data service does something similar I presume, has a pool of threads for running queries <civodul>cbaines: ok; if we can remove that complexity, that's better <cbaines>and yeah, these changes should allow for just having a pool of connections (rather than connections and threads) <cdo256_>My system config is http://0x0.st/Hb85.scm for reference. There's a function at the top called 'mutate-pam-entry' and the pam-services argument is at the bottom. <civodul>cbaines: BTW, the patch does not make the socket O_NONBLOCK <civodul>that would be a plus, but i think the main source of blocking is while waiting for the query to complete <cbaines>civodul, ok, I think I sort of understand what you mean <cbaines>I have spotted some failures in the Cuirass testsuite when trying to update guile-squee... I'll have a look and see if I can figure out what's going on <cdo256_>Hiya! Can anyone see my question above? I don't want to repeat myself, but I think it may not have actually sent. <cdo256_>Not that I'm complaining that I'm not getting my free help fast enough haha <Zambyte>Hey cdo256_ I actually think I joined after your question, but I do see your system config. What was the question? <cdo256_>Hi Zambyte! Here it is: I'm trying to add an argument to pam_env.so. I have succeeded in doing this to all (base-pam-services), but "login" is added after. Is there any way of modifying that? <cdo256_>I notice there's a transform and an extend-configuration, but that doesn't appear to be used in the operating-system definition. <cdo256_>transform and extend-configuration are in pam.scm, but not exported. <cdo256_>I'm currerntly looking at login-pam-service, but it doesn't take user_env as a parameter. Should I extend it with the aim of sending a pull request? <Zambyte>cdo256_: I've been playing around with it in the REPL a bit. I'm not quite sure what you mean when you say '"login" is added after'. <Zambyte>If you'd like to play around with it in the REPL also, you can use the things not exported in the pam module by using the following command: ,m (gnu system pam) <Zambyte>You could see if those private procedures do what you need. I don't think you actually need those though - it looks like those are used internally by the exposed extension API, which I am not very familiar with myself <cdo256_>I mean the pam module login which resies in /etc/pam.d/login is added to the pam-services field of the operating-system definition after my system.scm is read <cdo256_>Yeah I would actually reside in login-service-type but the parameter hasn't been added yet <cdo256_>I'm gonna make the change to /gnu/services/base.scm and see if I can make it work that way <trankitron>This error is generated by the following cmake code: <trankitron> COMMAND "/gnu/store/s8m8...-git-2.40.1/bin/git" <gabber>trankitron: please don't paste in here <MatoHota-work>cbaines: I wonder, since bootstrapping Ada compiler is a no go where I can discuss binary packaging ? <cbaines>MatoHota-work, I doubt it's impossible, just viewed as infeasible currently <cbaines>anyway, I believe nonguix has a binary build system, so maybe there's some discussion within that community <ulfvonbelow>trankitron: sounds like it's trying to download dependencies at compile-time. You can use a (origin ...) record as a native-input and copy the git repository in question into where pharo expects it to be before that particular build phase is run. <ulfvonbelow>if it's using submodules, you could also include (recursive? #t) in your git-reference. <gabber>doesn't GCC provide a frontend for Ada? <gabber>doesn't "frontend" in this context mean GCC can ingest Ada source code and create binary from that? <gabber>the ingestion part inside of GCC? <MatoHota-work>if you search the ada directory in gcc source you will find Ada code <gabber>so yes, thats a) a bummer and b) the classic chicken-egg/bootstrapping problem <bjc>guix doesn't build the entire compiler collection when bootstrapping gcc? <MatoHota-work>As far as I remember the first gcc version of gnat is about Ada 95 <gabber>bjc: doesn't seem that way. `guix shell gcc-toolchain -- gcc hello_world.adb` tells me: "Ada compiler not installed on this system" <gabber>ACTION is wondering in what language the first GNATs were written <trankitron>ulfvonblow: Exactly, pharo-vm tries to download several dependencies. From your response I assume it's normal that the git clone is failing and I should try a kind of workaround? <ulfvonbelow>yes, the build happens in a network-isolated container <ulfvonbelow>I'd add that libfii repository to the native-inputs field as a <origin> object, then reference it in a phase following the unpack phase which copies the libffi source to ./libffi-src <lambdanil>Hi, I'm trying to inspect the %base-packages symbol but I ran into an issue. <lambdanil>I have setup my Geiser according to the documentation, but after I do ,use(guix); ,use(gnu) and ,use(gnu packages) evaluating %base-packages results in an unbound variable <lambdanil>is running the "make" mandatory? I actually skipped that step, only doing the ./configure <bjc>‘%base-packages’ is in (gnu system) <cbaines>zimoun, yep, it's down at the moment <attila_lendvai>and i get 404 for /nar/zstd/arnx6fnjq85wscmr894d64cj3529r3h1-wxPython-4.2.0.tar.xz on bordeaux.guix.gnu.org <civodul>attila_lendvai: try "sudo rm -rf /var/guix/substitute/cache" <attila_lendvai>civodul, thanks! i think that worked. looks like it'll be built locally now. <fjb>Hello, I've just installed guix on a laptop but have some trouble. It boots to the terminal login prompt, then it prints two lines about missing firmware, flashes a cursor in the lower right and just becomes a black screen. What could it be? <fjb>Or rather what can I do? <ulfvonbelow>and if you press caps lock does the corresponding light on your keyboard come on? <fjb>Sorry, it's a low end budget laptop so it hasn't an indicator. <ulfvonbelow>did you get a decent look at which missing firmware it was complaining about? Sometimes those messages are relatively harmless <fjb>I believe it is network and bluetooth. <ulfvonbelow>since the installation image presumably worked for you, I'm gonna wager the state you're in is probably right after it tries running X / Wayland <ulfvonbelow>not sure how to get it to not do that without access to a command line <Aurora_v_kosmose>There seems to have been a lack of substitutes for quite a few things in my last upgrade (which is still ongoing). <fjb>Booted up in a live system. looking through /var/log/debug, messages and gdm/greeter.log. The only error messages that I could identify were related to sound and network. <fjb>How would I go about getting Guix boot to runlevel 1? <cbaines>fjb, if you can edit the grub configuration when you boot, I'd try removing the quiet option and adding nomodeset <fjb>cbaines Thank you very much! That got me in. <efraim>every time I start to look at ada I remember I don't want to try to compile such old versions of gcc <efraim>wait, gnat has a thank-you to the people who contributed the initial gccgnat code. I wonder if they also had their own compiler floating around <Kabouik>I've been using Guix on this computer since about a year, but today guix pull is a dead end for me: it eats all my ram and CPU, and eventually freezes everything (I had to hard reboot twice in two attempts). What can I do? I have 16GB of RAM on the device and a 2022 i7. <davidl>Kabouik, do you use any external channels? <davidl>what is the last output in your terminal before it freezes? <Kabouik>Yes, mine (which hasn't changed in months), nonguix, and guix-emacs from babariviere, games, and guix-cran. You're right, I should try to disable them one by one, see which one is the culprit. <Kabouik>Unfortunately I was in another workspace when it froze on my last attempt, but I think on the first one earlier today, it was something along those lines "building guix cache" or something, or guix config? <davidl>its not uncommon that a circular dependency may have been introduced. There is a circular dependency debug script to use if you want to find specifically which package may be creating a circular dependency in your channels. <Kabouik>Do you know where to find that script? <Kabouik>The script implies cd into the source dir of each channel, but I'm not sure where channels are pulled. Except for mine, for which I have a local dir, I don't know where are the others. Do I need to manually clone them somehwere just to run the script? <dcunit3d>is there a common way to test guix home profiles? like integration testing? <Kabouik>So I haven't been able to run the script davidl, but the issue seems to be with guix-cran, as guix pull runs fine when I disable this channel, and breaks again if I re-enable it. I'll try to opn an issue at their git repo later. <davidl>Kabouik, okay, well Im glad you found the problem. <civodul>apteryx, rekado, cbaines: hey! i've just reconfigured berlin and will be rebooting shortly <dcunit3d>lilyp: k, i'm hoping to find a lightweight way to check over the config. i'm using guix on two different distros, one native and one foreign, so getting things to work on both is a bit dicey <mirai>so GNAT was “bootstrapped” with I presume was a competitor's Ada compiler <mirai>to clarify, they bootstrapped using a subset of Ada83 and extending it along the way until they could write it in Ada95 <mirai>might be worth getting in touch with adacore itself <civodul>cbaines: the new guile-squee isn't working at all for Cuirass :-/ <prpldmn>after installing sbcl, gnome wouldnt start. looking through the issues i found what causes it [https://issues.guix.gnu.org/45360]. the solution in the thread was to unset some variables. anyone else experienced this? is that the only solution for it? since the issue is still open <civodul>cbaines: not yet; it crashes immediately along these lines <civodul>so i reconfigured to the commit before the guile-squee upgrade to leave time for investigation <graywolf>I need a new wifi router to use at home, so I thought I would ask: Does anyone use Guix on their router? If so, what HW do you use? <oriansj>graywolf: well depending on what your router is; it might not be a supported host architecture <graywolf>oriansj: Sorry, I probably worded the question badly. I am buying new one anyway, so I wanted to ask it someone can recommend tested Guix-compatible one :) <oriansj>graywolf: as one can use a full blown server as a router; just pick an architecture and make sure the system you choose has enough RAM and Disk Space