<dftxbs3e>nckx, how would I proceed to get commit access to create a branch on upstream guix repo? <dftxbs3e>I feel like this is the only way to make actual progress on the ppc64le port <nckx>dftxbs3e: After a sustained pattern of quality contributions (patches), you get 3 committers to vouch for you and apply to the maintainers. See (info guix)Commit Access. <dftxbs3e>nckx, can you create a branch for me onto which I can submit patches? <dftxbs3e>I don't want to commit anything to master yet <nckx>We have zarroo access control. <dftxbs3e>nckx, I mean submit patches by email for applying onto that branch *jonsger wonders why the guix package gained /usr/libexec/guix/guile <nckx>dftxbs3e: That's no problem! <dftxbs3e>nckx, alright cool! Can you create a branch called wip-ppc64le ? <nckx>Is there anything I could put there already? <nckx>Creating an exact alias of master is fun & all but a patch would be cooler. <nckx>‘that patch’ was the legendary 128-bit one, right? <nckx>It deserves to be the first. <dftxbs3e>by now it can be applied on master instead of core-updates <dftxbs3e>I would be so much more active at sending contributions if it wasnt done with git-send-mail - really I never used it and it always makes me uneasy because apparently there's bugs in debbugs too that prevent you from sending multiple patches at once..? <nckx>dftxbs3e: Thanks. After a year I have no idea if that question was bullshit or no longer relevant, but it would be nice to have an answer. <nckx>dftxbs3e: Not sure if it's a bug or a design flaw but yes, debbugs makes git send-email look worse than it is. There's a documented work-around, that's all I can offer. <nckx>You'll have to trust me that navigating GitLab just now felt for me as using debbugs feels for you. <nckx>Nasty, confusing, inefficient. <jonsger>is there a way to disable the build of guile in libexecdir? I don't think that is appropriate for foreign distro packaging <vagrantc>jonsger: it seems even more important; doesn't it ensure a compatible locale? <jonsger>vagrantc: ah looks like, but it will only used by guix? <vagrantc>jonsger: it will only be used by things that call it directly :) <vagrantc>libexec isn't typically in PATH, or is it? <vagrantc>jonsger: had some of these same questions recently :) <vagrantc>huh. mysteriously tests/swh.scm gets run regardless of weather it's passed via TESTS= ? <vagrantc>and hangs when running with debian's guile-2.2 for me <morgansmith>in arm-none-eabi-toolchain, the stdlib.h header is overwritten with the c++ version that refers to the original header. I don <morgansmith>'t know where that happens in the package definition. I just know it happens <GNUtoo>How do you handle letsencrypt deployment in system .scm configuration files? <GNUtoo>In the manual there is a section about letsencrypt which explains for instance how to get a certificate <GNUtoo>but for instance if I use Nginx, letsencrypt will use the same port than Nginx somehow <GNUtoo>And at the begining there are no certificates anyway <GNUtoo>So how do I bootstrap that in only 1 scm file? <GNUtoo>Here it uses --manual, so it takes the ports used by the web servers, but using a webroot wound't be any better as no certificates would have been created <GNUtoo>While re-reading the sources it uses --webroot not standalone <GNUtoo>so here it doesn't do much, I'll try to look in other parts of Jullien's config <GNUtoo>I wonder what happens if bogus files are there as certificates and then replaced <GNUtoo>In julien's config the https stuff is declared in the various machines in system/ *morgansmith needs to learn the difference between enter and ' <roptat>you won't get it working like that at the beginning, you need an intermediate state <roptat>first, add a nginx-service-type, with no server block (simply "(service nginx-service-type)") and your certbot config *GNUtoo wonders if there is another way <roptat>reconfigure, and restart mcron, then you should see the job to get a certificate in "herd schedule mcron", use it directly to get your first certificate <roptat>now you can use it in your config and add server blocks as you wish <GNUtoo>Indeed, that's exactly what I'm trying to avoid here <GNUtoo>I did something like that already and that works indeed <GNUtoo>basically generate the vm with the first config and guix reconfigure with the second config <GNUtoo>your solution is more elegant though <roptat>the issue is that nginx fails when you give it a certificate that does not exist <GNUtoo>Maybe certbot --standalone then nginx could work <GNUtoo>And having the service with --webroot as usual <roptat>oh, I found out about robotnix today, it's a project that aims at building android in Nix <GNUtoo>They might have pointers on how to do things indeed <roptat>I really need to have a good look at what they do, because they really seems to be doings things properly <roptat>I see a lot of derivations that are created by a tool, blueprint2nix <GNUtoo>I didn't know about the project name, I tried to find stuff some time ago about it but I probably looked in the wrong direction <GNUtoo>So I assumed that they didn't do much Android <GNUtoo>(They support some smartphones and tablets too) <roptat>but I don't understand them: it looks like they use a bunch of functions that have the same name as the type of projects in bp, but I can't find the definition <roptat>like bootstrap_go_package and stuff <raghavgururajan>ERROR: Cannot ioctl TUNSETIFF tun: Operation not permitted (errno=1) <vagrantc>i'm wondering if those various foo.scm bar.scm in foo.go files are some sort of paralellism issue? <roptat>but guix is not reproducible on guix either <vagrantc>non-deterministic ordering of inputs, maybe? <raghavgururajan>roptat: Works now. But connection is not working. I see "Bad LZO decompression header byte: 42" <GNUtoo>morgansmith: I've made a patch for --standalone, now I need to test it, and it's always a mess when testing <GNUtoo>like I spent days just to test 1 tiny patch *GNUtoo hopes to get more luck this time <roptat>raghavgururajan, as I wrote in my email, this is external to the VPN service <roptat>I think you need to set a DNS server that will be routed inside the VPN <roptat>no, just set it in the network service you use <roptat>all you need is to make sure the IP of your resolver is going to be routed through the VPN <roptat>if you use network-manager, you'll need to configure your connection to use that resolver (whatever you choose), but you can't do it through guix config, you'll have to do it on a per-connection basis <raghavgururajan>roptat: But network-manager-service-type doesn't have an option like that? <buenouanq>so I just found something that calls %intel-xorg-modules from (use-modules (config xorg-modules)), but I get 'no code for module (config xorg-modules)' when I try it myself <roptat>you could set it to none, and configure /etc/resolv.conf manually <roptat>it's a file named config/xorg-modules.scm, but that doesn't look like a module name from guix itself <roptat>guix modules are in (gnu ...) or (guix ...) namespaces <buenouanq>right, this was something novel - Just found this searching for touchpad driver help, bummer if it's some homecooked thing they didn't include... <GNUtoo>ice-9/eval.scm:293:34: Value out of range 1 to 4294967295: 2305843009213693951 *GNUtoo doesn't have 1 week just for that <GNUtoo>Ah it looks like the same failure than before <GNUtoo>The issue is that there are no known commands that build it for sure <GNUtoo>so it's trial and error from the manual, add this and that <lafrenierejm>I'm getting an error when attempting to perform a clean build of 1464131: "ice-9/eval.scm:293:34: error: core: unbound variable" <morgansmith>lafrenierejm: I'm not getting that. Try a git stash && git clean -xfd && make distclean *morgansmith is not responsible if that deletes or reverts files you're working <morgansmith>also try building it with guix environment -C --pure guix -- cmds <dftxbs3e>nckx, I got offload to work with powerpc64le-linux-gnu <dftxbs3e>I'll submit the minimal set of patches required for that and then you can setup the OSUOSL VM with it on GNU Guix official CI <wargreymon2020>I cant start emacs-guix's repl or invoke any command that use repl on the newest version of emacs-guix (0.5.2-2.58a840d) <lafrenierejm>morgansmith: Removed untracked, unignored files then running `make distclean` resolved the issue. Thanks for the suggestion! <kozo[m]>Is there some cuirass configs people are willing to share or know where some exist? <wleslie>sneek: later tell morgansmith do you have a `~/.guile` that configures readline? then you need readline in your environment, too <lafrenierejm>wleslie: You're saying the comment about readline was intended for me? <lafrenierejm>What's the preferred way to submit a large series of patches? <lafrenierejm>Including its transitive dependencies, a Rust app that I'd like to add consists of 37 commits. <vagrantc>you file a bug, and then follow-up to the bug with each patch in a separate email <wleslie>lafrenierejm: yes, it's a good thing to rule out, as it gives a similar error <wleslie>not sure if it's exactly that one now that I think of it ***dongcarl4 is now known as dongcarl
<kozo[m]>Anyone able to help me setup a ./pre-inst-env? I am trying to follow the documentation but am still having issues <iyzsong-w>kozo[m]: my workflow is: guix environment guix; ./bootstrap; ./configure --prefix=; make; ./pre-inst-env guix ... <kozo[m]>iyzsong-w Thanks for asking, the problem is layer 8. I was building the bootstrap in the environment then exiting <DivanSantana>trying to install python-ldap via pip and it's failing to build looking for sasl.h . Anyone know what pkg to install on guix to fulfil that? <kozo[m]>Seems to be working when I stay in the environment <DivanSantana>I want to install via pip, since the other members in project do it that way. <zzappie>DivanSantana: I gues you need openssl <zzappie>DivanSantana: python-ldap guix package inputs also include openldap, cyrus-sasl and mit-krb5 <DivanSantana>zzappie: thanks. couldn't install in via pip with the same sasl.h error. But I'll just use python-ldap from guix for now. ta <kozo[m]>I got clone, git checkout the hash in the issue but when I make, it says it can't find the error: pinebook-pro-barebones-raw-image: unbound variable <kozo[m]>How do I find out what branch f0c609323e belongs to in the guix repo? <iyzsong-w>kozo[m]: you can use latest master. yestaday i built a disk image for pinebook (not pro..) but not test yet. <kozo[m]>because using that, it said the module for mathieus pinebook patch is not found <iyzsong-w>well, i think the 'pinebook-pro' patch is not merged, you need to apply that patch by hand. <efraim>does 'git log -1' show that patch? <kozo[m]>but when running ./pre-inst-env on the /tmp/os.scm made with the command he gave me, it didn't know about the module <pinoaffe>hi guix! what package provides libgcc_s.so.1? <pinoaffe>zzappie: that's what I guessed, but the library is not availabe in an environment with gcc-toolchain <zzappie>pinoaffe: ahh right. gcc-objc includes it too ***apteryx_ is now known as apteryx
<civodul>pinoaffe: internally it's in the "lib" output of "gcc" <civodul>that said, you normally don't need to worry about libgcc_s.so, GCC knows where to find it <zzappie>civodul: but gcc doesn't have lib output *zzappie didn't know tha we have just "gcc" package now ) <pinoaffe>and yeah, I'm running a "found" binary in a guix container, so I guess that "normally" doesn't necessarily apply, sadly <leoprikler>zzappie: gcc is hidden, so that you don't shoot yourself in the foot <mbakke>I wonder why Cuirass never bothered to evaluate 007edfb72ac (staging) <mbakke>the preceding evaluation failed because the build node ran out of disk space :/ <cbaines>mbakke, Cuirass polls, so if several pushes happened one after the other, it could skip revisions <cbaines>mbakke, looking at more data, I don't think that explains the current situation since 007edfb72ac is the latest revision, and has been for many hours now <mbakke>I've pushed another revision now, let's see.. <cbaines>mbakke, I added a feature yesterday where you can give the Guix Data Service two revisions, and it'll tell you which packages have broken between those two revisions (assuming it knows about enough builds to work that out). Once I start building staging stuff, I'm hoping this will help get a better sense of changes. <mbakke>cbaines: that sounds amazing, it's my #1 missing feature from Hydra <zzappie>leoprikler: well I guess I just got confused by gcc/gcc-toochain naming another time in a row. I've missed that gcc package is supperseeded by gcc-toolchain and thought it is a new alias :) <mbakke>cbaines_: very nice :-) although the numbers don't look too good... <cbaines>mbakke, thanks, which numbers are you referring to? <mbakke>cbaines: am I reading 'meshlab' right in that 3 substitute servers failed for the new revision, whereas previously it had four successful builds? <cbaines>mbakke, ah, so there have been 3 failed builds, but all on guix.cbaines.net <mbakke>cbaines: oof... well those are the numbers I was referring to :-) <cbaines>Oh well, it's useful to me to have some things break which I can test the Guix Data Service interface against :) <jonsger>hm python2-html2text fails in check phase, it's a dependency of calibre <jonsger>mbakke: i think the update in ee45483cc9fb45ae2516e40bb309d2294444d7c6 breaks it... <mbakke>jonsger: oh my, I forgot to test python2-html2text! will fix. <mbakke>on a related note, we should switch Calibre to Python 3 soon.. <jonsger>mbakke: I wondered why it's still depends on python2 <jonsger>I can have a look as I'm actively using calibre... <lane>is it possible to use git references in local package (i.e. one loaded from a file with `guix environment -l`)? <civodul>lane: hi! for any package, its "origin" can use url-fetch or git-fetch (among others) <lane>civodul: thanks, that's what I was trying to use but I get an error: Throw to key `match-error' with args `("match" "no matching pattern" #<origin #<<git-reference> url: ... <lane>hmm, what's the best way to share it? <efraim>mbakke: I see your fix for alacritty on staging, it's not also needed on the check phase? <jonsger>mbakke: python2-pillow has a similar problem... <lane>civodul: FWIW this works when I put my code in the guix repo and use the pre-inst-env <lane>the package doesn't work yet but that's a different story :| <mbakke>efraim: the check phase does not pass --features <mbakke>jonsger: oof.. we should ask apteryx about that <civodul>lane: that's because 'patches' must be a list <civodul>i admit the backtrace isn't helpful at all here <mbakke>jonsger: you may be able to do a similar fix for python2-pillow <civodul>we need "contracts" to improve on that... <AppAraat[m]>So does that mean that guix.gnu.org package list page is out of date? <civodul>AppAraat[m]: it's called "go-github-com-junegunn-fzf" as you can see in this file <civodul>if it's a standalone program, the policy would be to rename it to "fzf" i think <AppAraat[m]>ah, I see. I'm not yet very familiar with the naming schemes and such, thanks! <maav>well, let's say it's weekend, hehehe <maav>at least i've sent a wip patch for the ngettext problem to gettext <maav>but i'm still ashamed of not overwriting the dates <maav>also i've been trying to write something about parameterized packages... but i'm not sure if it will add too much to the topic <maav>release is almost ready, isn't it, civodul? :D <civodul>we didn't get any feedback on RC2 i think <maav>my limited tests went ok, but i just tried with time-machine <civodul>it would have been nice to have feedback for foreign distros <maav>let's see what the weekend gives, perhaps some people report then :) <civodul>apparently there have been tests because guix-install.sh was modified to simplify that <civodul>i'm planning for a release on Monday if nothing bad happens by then <maav>now we need a song to reach OpenBSD's releases coolness level :-P <maav>i'd like to, but i even struggle to play other people's music, hehehe <civodul>oh you're a musician or hobbyist musician? <maav>i play a bit guitar and try to learn the drum kit <maav>do you like playing, civodul? voice is an instrument too ;-) <civodul>not much! i play the piano a tiny bit *rekado_ just pushed a rebased wip-r <rekado_>someone who is fluent in Esperanto should write the lyrics! <roptat>mi ne parolis kun esperantistoj ekde du jaroj <roptat>not even sure my use of ekde is correct here *rekado_ is always amazed at how very European Esperanto is <roptat>you shouldn't be amazed, it's a mix of european languages oonly *rekado_ tries to stop the amazement <roptat>that's what "international" meant in the 19th century I guess :) <roptat>well, a small subset of indo-european languages <civodul>"international" in the 21st century means English, so one could argue Esperanto is an improvement <roptat>but indo-european is certainly not the simplest you can make... I've started learning toki pona lately :) <maav>international in the 19th century meant probably French, it was the same kind of improvement in that sense <roptat>not really useful for communicating complex ideas though *civodul looks up toki pona <roptat>120 words, very simple grammar, almost no recursivity <roptat>very simple syllable structure too <roptat>CVN, like Japanese for instance; nothing as complex as "crisps" :) <roptat>but because of the limited vocabulary, it's a lot harder to be precise <maav>that's the main issue, artificial languages only work in certain situations <roptat>it's not meant to be used as an auxiliary language, so it's fine <roptat>anyway I'm all for a song, but can't promise I'll come up with lyrics :) <maav>well, in harmony the "universal language" would be I-vi-IV-V ;-) <rekado_>though, the ‘vi’ is actually just the parallel minor to ‘I’, so the cadence is really just ‘IV-V-I’ <jonsger>divoplade: I don't think yet but it seems like soon :P ***trav82051 is now known as travankor
<divoplade>I'm sure there could be volunteers for writing or fixing the music and recording instrumental parts <divoplade>So... should we port it to guile 3 first? x) <civodul>once we have the score, surely we can find Someone to help us typeset it <jonsger>I'm a bit fluent in lilypond, but not an expert... <maav>rekado_: i was waiting the jazz person :) <maav>(but ii-dim is V7 without the root) <maav>rekado_: not sorry at all, i love it <maav>rekado_: i'm a 12bb person, btw, the pop-loop was obviously the easy take <apteryx>greetings! Where was this issue about Guile .go non-reproducibility? <rekado_>apteryx: it’s probably in the Guile bug tracker now, so you can’t find it on issues.guix.gnu.org <lafrenierejm>Just (attempted to) submit 37 patches to bug#44762 via git-send-email, which was my first time using send-email. Is anyone available to do a quick sanity of one of the emails to see if I need to resubmit? <apteryx>rekado_: oh indeed it's in the Guile tracker, with ID 20272 <lafrenierejm>jonsger: Excellent. Thanks for looking, and for sending the link. I hadn't been aware of patchwork.cbaines.net before this. <civodul>although the patch series is not really about pytorch <sneek>zimoun, mdevos says: presuming free time and no scheduling conflicts, I will be present at the conference <zimoun>civodul, maav: aside avaibility of some specific substitutes, I think v1.2 is ready, isn’t? <zimoun>maav, different song and drawing each release? ;-) <nckx>civodul: Still no horrible bugs in 1.2, I will dig deeper. <mothacehe>zimoun: I'm not sure it's our responsibility to backport compiler fixes. We can wait for an eventual 8.0.2 or suggest using a more recent revision. <maav>zimoun: i've always admired openbsd people because of that, they look super-cool to me :) <zimoun>mothacehe: ok, that’s an acceptable policy. :-) Except if dongcarl absolutely needs it as you suggested. ;-) <zimoun>maav: me too. I only give a look to their release because of that. ;-) <nckx>efraim: Sweet. Please keep taking care of my to-dos (17a53aec). <jayspeer>has anyone tried installing guix system on an old machine? I'm talking thinkpad x60t old <kozo[m]>mothacehe Hello, I got stuck while testing out your PBP patch. these are the steps I took: 1) Git clone 2) git checkout <hash> 3) guix environment guix 4) curl the patch 5) git am the patch 6) ./bootstrap 7) ./configure --localstatedir=/var 8) make 9) Echo command 10) Finally, when I run the ./pre-inst-env command, I get: no code for module (gnu system images pinebook-pro). <mothacehe>kozo[m]: Hello! It looks good, not sure what's going wrong. I'm going to push this patch anyway, so that you can just "guix pull" and run the build commands. <mothacehe>testing it right now, I'll let you know when you can pull! <roptat>jayspeer, I don't know it well, but this looks like it's about the same era as my old packard bell. I managed to install Guix on it, but that was a few years back <roptat>(it's a laptop, and it used to have a "vista ready" sticker on it, so that might give you an idea of its age :)) <mjsb>Good evening dear Guixlings! I was testing the script installation for the upcoming v1.2-rc2 on a Fedora 31, and the script completed well, then i rebooted but running guix commands feedbacks "guix pull: error: failed to connect to `/var/guix/daemon-socket/socket': No such file or directory" <jayspeer>roptat: what kind of media have you used to install it? usb 2.0 is painfully slow (though I've managed to install debian this way) and cd is almost usable (but not really) <mjsb>I checked a thread which suggests to start the guix-daemon, but running "sudo systemctl start guix-daemon && systemctl enable guix-daemon" won't start the daemon <mjsb>Extra: is it correct that i've ran the script with "sudo sh -x guix-install.sh" ? Or should be better becoming root user and after running the script ? <jayspeer>mjsb: have you copied the systemd script for guix daemon? <jayspeer>ah, automatic install - no idea, but it seems this step was missed somehow <jayspeer>`It should be run as the root user.` - try again :) <mjsb>Actually it refuses "A previous Guix installation was found. Refusing to overwrite." is there an easy way to clean the current Guix for reinstall ? <jayspeer>I would just delete the /gnu and /var/guix, but I'm not sure if it's the safest *nckx starts guix building vim in a loop until it fails. <civodul>mjsb: in general it's unnecessary to uninstall/reinstall Guix because it doesn't make any difference <civodul>mothacehe: looks like we're all set for the release, right? <mbakke>mjsb: do you have SELinux enabled? you'll need an additional step to install the SELinux policy in that case <mbakke>civodul: I had forgotten about that patch series (it was for a potential client that never got anywhere) -- will try to merge the still-relevant bits <jayspeer>So it turns out, it is not that cds are faster than usb 2.0 (the opossite is true, in fact), but that my usbdrive has died. The more you know, huh <jayspeer>installing from usbdrive works (the working one) <mjsb>Thanks @civodul and @mbakke for the pointers, i'll check it <civodul>colleague of mine tries to set up offloading with the build machine on a foreign distro <civodul>i end up doing tech support for one or two hours <civodul>and at that point, i realize the situation is somewhat shameful <civodul>we need to do something to make it possible for human beings to set it up <civodul>the main issue is that a tiny bit of offloading relies on being able to run "guile" <civodul>and from there to import Guix modules <civodul>which means you need to have Guix dependencies explicitly installed, GUILE_LOAD_PATH etc. properly defined, etc. <civodul>i think it works out of the box if the build machine is on Guix System <jayspeer>civodul: aren't those things defined automatically when sourcing from .config/guix/current/etc/profile and .guix-profile/etc/profile? <jayspeer>maybe sourcing those files for the remote process could be a hacky solution <nckx>jonsger: Thanks. Still building /gnu/store/kbpgcz6j52cca9mnvsgf4yps816w7j1g-vim-8.2.2017.drv here, still hasn't failed. Interestingly (...perhaps) I'm also on Wayland. Sway. Although that really shouldn't affect the build environment at all. <civodul>jayspeer: not quite, it lacks GUILE_LOAD_PATH <jayspeer>civodul: mine .guix-profile/etc/profile has it, if I have guile installed <rekado_>I didn’t know about the OpenBSD songs. ‘100001 1010101’ is rather Rush-like. <civodul>BTW musicians, we're (hopefully) releasing on Monday, which is Guix anniversary <civodul>that can give ideas about the theme for the song <civodul>hmm i get 200KiB/s to berlin right now <dongcarl>Question for y'all: If I want to build clang-8, but with my-llvm-8 (say with a patch applied), can I use --with-input=llvm-8=my-llvm-8, and all the clang-runtime and clang itself will be rebuilt against my-llvm-8? <dongcarl>My concern is that for procedures like clang-runtime-from-llvm, the INPUT gets rewritten, but when the procedure is evaluated, the "llvm" argument is still the normal llvm-8 and not my-llvm-8, but perhaps I'm being paranoid <civodul>dongcarl: hey! it should replace llvm@8 with my-llvm@8 everywhere (note "@", not "-") <civodul>if in doubt, you can check with "guix graph --with-input=llvm@8=my-llvm@8 ..." <civodul>although well, the graph might be git <dongcarl>civodul: Help my understand the mechanism here: let's say I do `--with-input=llvm@8=my-llvm@8` and try to build `(clang-runtime-from-llvm llvm-8)`, the input llvm for the resulting clang-runtime would be replaced successfully, but note that in `clang-runtime-from-llvm`, it refers to `(package-native-inputs llvm)` and I'm wondering if _that_ would <dongcarl>also be replaced with effectively: `(package-native-inputs my-llvm-8)` *dongcarl is thankful in advance <civodul>dongcarl: when you write --with-input=LEFT=RIGHT, the whole DAG of the packages is traversed, and each package called LEFT is replaced by RIGHT <civodul>if you write LEFT@8, you restrict that to version 8 of LEFT <civodul>so you don't need to think in terms of functions, native inputs, etc. <apteryx>is it possible to use lset-intersection on lists (sets) of package objects? <apteryx>works there because I'm using the package "name", but that's unreliable. Substituting name for package in the match-lambda yields no results. <civodul>apteryx: it's possible with any kind of object, but note that packages are not comparable with equal? <civodul>or again IOW, there can be two packages that are not eq? but are "equal" in the sense that they lower to the same derivation <apteryx>OK. Yeah, my problem here is that eq? doesn't seem to work, or I'm goofing up something <apteryx>hm, the later, I tested a minimal example and it works ;-) <civodul>wait, aren't you applying eq? to lists? <civodul>that won't do what you want, because eq? is about pointer equality <civodul>so (eq? (list 1 2 3) (list 1 2 3)) ⇒ #f <mjsb>Salute! I've followed the "SELinux-Support" section of the manual, and it asks to execute guix-daemon.cil and to relabel the filesystem with restorecon, to which filesystem is referring to ? <apteryx>civodul: no I'm doing (lset-intersection eq? package1-inputs-union package2-inputs-union ...) <apteryx>but I think the culprit is my use of delete-duplicates without specifying the equality test (it uses equal? by default) <civodul>what do these two lists contain though? <civodul>if they contain tuples, eq? won't do what you want <apteryx>civodul: the lists of inputs contains package objects, like: ((#<package python-matplotlib@3.1.2 gnu/packages/python-xyz.scm:4961 7f65bdc74140>) (#<package python-colorspa[...]) ...) <apteryx>perhaps it's working as intended and there really isn't any common package, although that doesn't seem to add up at first sight <cbaines>the ( before each #<package makes me think you've got a list of lists <apteryx>oh indeed! that must be it. I got tricked thinking it was the representation of a package object :-) <kozo[m]>How many modifications do I need to make to the example cuirass code from the manual to have it work? I have added the code to my publish server and I can connect to the cuirass-web page but it doesn't have any elements or packages to build. Do I need to add a manifest file somewhere? <apteryx>indeed it was! My match pattern needed to be made into: ((name package output ...) ...) ;; added output ... <nojr>how do Guix's installation prefix on a foreign distro? <nojr>Formbi: The public key for ‘ci.guix.gnu.org’ is installed along with Guix, in ‘PREFIX/share/guix/ci.guix.gnu.org.pub’, where PREFIX is the installation prefix of Guix. <nojr>roptat: I have no /usr/share/guix dir <nojr>mbakke: install script on the website <nojr>mbakke: i run it on a foreign distro <mbakke>nojr: then you'll find the key in /root/.config/guix/current/share <nojr>mbakke: ~/.config/guix/current/share/guix/ ?? That's where I found the file <mbakke>any user who has run 'guix pull' will have that directory <wire_wolf>I installed Guix on a VM. It's been a lot of fun to play around with. But for some reason IceCat can't display numbers. Is there a font package I need to install or something? <brown121407>wire_wolf: I have fontconfig and a bunch of font packages like font-gnu-unifont, font-google-noto etc. installed and IceCat displays stuff just fine, so that may be it. <wire_wolf>Humm okay I'll installing fontconfig and playing around with fonts and see if that works. <nckx>fontconfig shouldn't be needed, installing something like font-gnu-freefont-ttf (‘a font’) should suffice. <wire_wolf>Nice. Yep installing a font and changing the system default seemed to solve the problem. <db48x>pretty odd that you would end up with fonts that have letters but not numbers <db48x>you should file a bug with as much information about how you got into that situation as possible <kozo[m]>Are there other areas of documentation for Cuirass than section 10.8.22? <cbaines>mbakke, I made the necessary changes to the Guix Data Service so it's actually possible to compare master and staging and find broken builds. Keep in mind that it's limited by the build information the Guix Data Service has. <jonsger>kozo[m]: the info pages of cuirass itself <jonsger>kozo[m]: guix install cuirass && info cuirass :) <db48x>mmm, not a lot of information in that bug report <db48x>wire_wolf: if you can get back to the state where it's broken (perhaps by going back to a previous snapshot of your VM?), try running fc-cache instead of installing a font package ***daniel is now known as Guest52489
***Guest52489 is now known as daniel-molina
<apteryx>it's weird working from the API that package-inputs can end up being an origin object <wire_wolf>I rolled back to a state where it's broken again. But running fc-cache and restarting IceCat dosen't fix it. <jonsger>Gooberpatrol66: yes `guix system search` <dongcarl>Getting a build failure while testing 1.2.0rc2 when building binutils-mesboot0-2.14 <dongcarl>Looks like the ./configure for `ld` wants to invoke `lex` *rekado_ joins ##guix-song *apteryx tries to reproduce with ./pre-inst-env guix build -e '(@@ (gnu packages commencement) binutils-mesboot0)' --rounds=10 --check <dongcarl>apteryx: I was able to reproduce with guix time-machine --commit=207c3c594e22b678ddcea8ebc78f19b839612187 -- build --cores=0 --max-jobs=96 -e '(@@ (gnu packages commencement) binutils-mesboot0)' <civodul>db48x, mbakke: does the new guix-daemon.cil deal with libexec/guix/guile correctly? <civodul>i'm asking because i just noticed cc98b00857e29074de96a6ed60e325cdfffaea1a (2019) <mbakke>civodul: unless that guile is needed by the daemon it should be fine <civodul>it's the shebang of 'guix', and the daemon invokes 'guix substitute' etc. <apteryx>dongcarl: I'm testing from the v1.2.0rc2 tag <civodul>(i know nothing about SELinux but that libexec/guix thing stroke me as potentially problematic) <dftxbs3e>using guix offloading some times retrieving store item takes forever with no bandwidth usage, is that known? <apteryx>definetely not normal, there's a bug logged against it <dftxbs3e>apteryx, been doing this like ten times in a row it keeps happening <dftxbs3e>after restarting it finished but it didnt finish when I used -M 4 <dftxbs3e>it only finished after repeated runs when I used -M 1 <dftxbs3e>There was hangs and some other times corrupt store items canceling the whole build <civodul>cbaines: the "broken changes" page at data.guix.gnu.org is great <cbaines>civodul, as for the broken build link, yeah, that page doesn't account properly that the build is actually for a different derivation which generates a common output. I'll fix that... The link on the derivation page works <cbaines>On linking to builds, I also should start storing the actual build ID for Cuirass, as that would mean I could generate better links... <civodul>ah isn't it that it links to ci.guix.gnu.org/build/xyz.drv instead of /build/123? <cbaines>the former should work, you just need to know (and use) the right derivation <cbaines>the latter would be better as it's a more human readable page <civodul>OTOH the latter is more likely to break maybe? <cbaines>break in terms of Cuirass moving the URLs around? <mbakke>civodul: I don't see that libexec guile in an 'strace -s 200 -f -p $(pidof guix-daemon)' while running 'guix build <large-package-not-in-store>' <mbakke>civodul: I see /gnu/store/a4wvv5d6c7lvvxi30l3p87r9vrwadmyv-guix-command though, which does not seem to require guix_daemon_exec_t somehow <civodul>it depends on whether you're running guix from "guix pull" for guix from the package <civodul>well, in any case, it's in the store <civodul>if you do "head $(guix build guix)/bin/guix", it's libexec/guix/guile <civodul>but dunno, maybe that doesn't matter? <mbakke>civodul: since /gnu/store/a4wvv5d6c7lvvxi30l3p87r9vrwadmyv-guix-command works without the guix_daemon_exec_t domain, I assume $guix/libexec/guile does too <Formbi>how can I setup personal substitutes so that I can share things between my local machines? <apteryx>dongcarl: it's still building, seems to be difficult to reproduce <mbakke>Formbi: you need to start 'guix publish' on the remote machine, and authorize its public key and add --substitute-urls on your local machine <dongcarl>apteryx: Weird! I was able to trigger it every single time... <dongcarl>apteryx: I do want to say that I'm on a very old version of Guix and using time-machine to build the newer one <apteryx>dongcarl: I managed to build it 10 times successfully <apteryx>on the version-1.2.0 branch (at the exact tag commit, or HEAD if shouldn't matter much). <cbaines>can you compare derivations to see if you're attempting to build the same thing? <apteryx>here, /gnu/store/plgzgmklaslp8dg5mamnda9wr6zch4gz-binutils-mesboot0-2.14.drv <apteryx>(via ./pre-inst-env guix build -e '(@@ (gnu packages commencement) binutils-mesboot0)' -d) <dftxbs3e>apteryx, corruption errors look like this: guix offload: error: corrupt input while restoring '/gnu/store/guix-UWLdDO/include/X11/extensions/xfixeswire.h' from #<input: string 7f8ac3e1d540> <apteryx>dftxbs3e: that's the issue already known <Formbi>how to add a key to the config.scm? <Formbi>(authorized-keys `("something" ,(public-key (…))))? <apteryx>dftxbs3e: hmm, actually I've searched the bug tracker and can't seem to find that one (there are a couple possible errors already reported, but not that one). <apteryx>dftxbs3e: if you get the hang again you could look at 'guix processes' on the offload machine and attach strace to the pid with 'strace -p $PID' <Formbi>btw is it possible to add a substitute server to the config.scm? <dftxbs3e>apteryx, on the machine I am offloading to, right? <nojr>how should I be seperating my manifest files into different profiles? the manual says this <nojr>Easier upgrades and maintenance: Multiple profiles make it easy to keep package listings at hand and make upgrades completely friction-less. <nojr>Laptop-specific programs (like ‘powertop’) that you don’t need on a desktop. <nojr>Deduplication: Profiles share dependencies that happens to be the <nojr> exact same. This makes multiple profiles storage-efficient. <nojr>Deduplication: Profiles share dependencies that happens to be the exact same. This makes multiple profiles storage-efficient. <Formbi>I'm getting connection refused from the computer I set guix publish on <dftxbs3e>apteryx, for the corrupt issue, it will be difficult to strace <dftxbs3e>for the hang one, I'll try meeting it again <dftxbs3e>but the processes quit and spawn so it's hard to catch <nojr>Deduplication: Profiles share dependencies that happens to be the <nojr> exact same. This makes multiple profiles storage-efficient. <nojr>i think so far those are my use cases, how should I got about looking which programs share the same library on my current default profile? <nojr>I'm looking on how to organize my packages in different profile because someimtes I need to upgrade one and sometimes there are conflicts between packages <leoprikler>The "multiple profile" setup is imo not a very well supported one. You'll have to manually set things up in roughly the same way as that one blog spot describes. <dftxbs3e>apteryx, I have parallel-builds set to 8 in machines.scm, and when I use -M 1 it seems I don't get corrupt issues <nojr>leoprikler: That's how I felt when I learned aobut them that's why I never set it up <Formbi>couldn't you check for hostnames in the manifest? <nojr>but I am running into that issue now where I have many packages and osmetimes one needs updates but it conflicts with some other packages dep <Formbi>so that it would decide whether to install something? <nojr>leoprikler: do you know of a workaround? <leoprikler>Well, the conflict situation is mostly resolved by using manifests, but you still get big build times with increasing package numbers <leoprikler>for the record, how big a profile are we talking about here? <dftxbs3e>apteryx, it seems that settings parallel-builds in machines.scm automatically sets -M parameter <nojr>leoprikler: around 60, not really that much but sometimes I need to update a pacakge, usually emacs or a desktop app <nojr>and it will raise issues in which I need to update deps or delete a package <leoprikler>Hmm, around 60 is not that big, can you not simply upgrade all your packages at once within a reasonable time? <Formbi>I get connection refused from a Guix system machine and not from an Arch machine <nojr>leoprikler: that's what I've been doing hehe but I thought that maybe it'd be better to write several manifest files with different profiles? <leoprikler>You can indeed do that and there's a cookbook entry that describes exactly that approach. <Formbi>and when I run guix publish manually, it does connect <leoprikler>The thing is, it is really all shell scripting, though, which is in my personal opinion a bit far from claiming "out of the box" support :) <nojr>leoprikler: I agree! I never bothered setting it up maybe I'll keep it that way until I reaaally need it <leoprikler>For the record, I have split my manifests as well, but still keep a single ~/.guix-profile to avoid issues that come with putting stuff elsewhere <leoprikler>I also make heavy use of `guix environment` when I feel I don't need a package to be available all the time <leoprikler>My setup is a one-liner: `find ~/.config/guix/ -name 'manifest*.scm' | sed -e 'i -m' | xargs guix package` <Gooberpatrol66>guix system: error: /home/nathan/guix/config-guixrig.scm:129:12: no value specified for service of type 'syslog' <nckx>Gooberpatrol66: That you need (service syslog-service-type (syslog-configuration)). And that syslog-service-type needs to be patched with a default value ☺