<rekado_>this “load-thunk-from-memory: incompatible bytecode” means that you’re using a different version of Guile than the one that compiled the file <rekado_>note that Guile 3.0.2 and 3.0.5 have different bytecode formats <rekado_>so the exact version really is important here <awb999>so basically one package is using guile 3.0.2 and another 3.0.5 ? <awb999>I am not even sure why it loads all this modules <nckx>lf94: Thanks! There's a list specifically for patches (unimaginatively named guix-patches@), tracked together with bug-guix@ over on issues.guix.gnu.org. <nckx>Could you submit it there? <dftxbs3e>it would be great if it supported Firefox WebExtensions <pkill9>apparently they're adding webengine support <pkill9>it says webextensions are on the roadmap <awb999>Can it be that guile produces go build artefacts that are not deleted when I do guix upgrade? <pkill9>it says they support webengine in the FAQ but someone said in the IRC it's only partial support I think <roptat>awb999, you shouldn't have guix itself in your user profile <awb999>will it then unistall itself completely? <awb999>or just remove guix from the profile? <pkill9>it will remove it from the profile <pkill9>you don't want to install guix with guix package -i <pkill9>the guix you want to use is at ~/.config/guix/current, which is updated by `guix pull` <awb999>I think I did that to update guix itself <pkill9>if you want to update guix use `guix pull` <ison>Is it normal for ~/.cache/fontconfig to be completely inaccessible and non-removable, even for root? Every command I run on it hangs. Even "ls" just hangs and slows my whole system down. I even tried opening it as a directory stream in guile to just read 1 item and that hangs too. <ison>I booted a liveCD to see what was inside and it only has a link to fonts.conf. I deleted the directory, but when I booted my system again it was back, and still inaccessible. I think it's causing some instability on my system because sometimes when I try to access it icecat crashes. <vagrantc>awb999: using "guix install guix" will actually consistently downgrade your guix to an earlier version, counter-intuitively <nckx>ison: Sounds like fs corruption. <ison>nckx: I was afraid of that. I noticed it after I ran "fc-cache -fv" and I stopped it with ctrl+c. Do you know anything I can try, or should I do some research on fs corruption? The fs is btrfs. <nckx>No ideas beyond check dmesg, run a fsck, maybe strace the hanging processes (although I suspect they're hanging in kernel space but who knows). <vagrantc>gah. diffoscope has too many inputs to be testing it all the time ... here comes a new batch of openjdk builds :/ *nckx doesn't trust btrfs to this day, TBQH. <vagrantc>hah. the problem with diffoscope 166 wasn't a missing input, it had two conflicting inputs. python-file and python-magic <vagrantc>they have incompatible implementations of magic* <vagrantc>python-magic has a compatibility layer, but it's a ... not-quite-compatible compatibility layer. /o\ <ison>nckx: Thanks for pointing me in the right direction. I suppose the first thing I'll do is make sure to back up my files <nckx>vagrantc: I'm impressed you found the cause, although I can imagine it was a painful journey. <vagrantc>nckx: based on the test suite failures, i had a pretty good idea of which upstream patch to revert. and then reported the issue upstream and the result of that interaction got me thinking :) <nckx>ison: Well, we don't know if it's the right direction, but I can certainly tell you there's nothing special about ~/.cache/fontconfig or that the behaviour you describe is indicative of anything else. <vagrantc>given how many inputs diffoscope takes to build, really happy to remove at least one of them. :) <nckx>Yes, it's one of my ‘oh joy, we're building ____ today’ packages. <nckx>Downclock the laptop and go grab a coffee. <vagrantc>the taunting part about diffoscope is guix is still missing a few things: External-Tools-Required: apksigner, apktool, cbfstool, dumppdf, dumpxsb, lipo, oggDump, otool, ppudump, procyon, showttf Missing-Python-Modules: guestfs, r2pipe <vagrantc>vaguely recall someone working on packaging guestfs <user_>Has anyone used awesomewm on Guix before? <nckx>Presumably; apart from the peep(s) who packaged it, there are some bug reports on the issue tracker. <nckx>vagrantc: The fact that diffoscope calls those ‘required’ annoys me greatly. No, apknonsense is not ‘required’ for me to diff tarballs. I do not want it. Learn to $PATH. <vagrantc>nckx: well, they're required if you want to diff things that require those <nckx>I'll soon have to create my own package to have timely diffoscope updates. <nckx>And Sway requires LibreOffice if you want to edit documents under Sway. C'mon diffoscope. Don't. <vagrantc>diffoscope aims to diffoscope all the things diffoscope supports <nckx>The problem is that diffoscope supports everything. <vagrantc>i've also wondered if some of diffoscope inputs could be moved to native-inputs <vagrantc>but the test suite should still run the full test suite ... though i've pondered if a diffoscope-minimal would make any sense ... or a diffoscope-with-everything makese sense <vagrantc>for the most part, most functionality of diffoscope is optional <nckx>vagrantc: I actually agree about the test suite (well-tested packages trump my selfish desire for faster builds), just think the closure needs to encompass the world. <vagrantc>so it's really just the tests that pulls in everything *nckx reaches for the coffee. <vagrantc>open to ideas how to break it up more :) <vagrantc>diffoscope 3 files changed, 1 insertion(+), 75 deletions(-) ... with improved functionality and a slightly smaller closure. :) <vagrantc>of course, any time you remove a patch it's going to look good on the diff :) <vagrantc>guess i should have sat on that a little longer and waited to hear back from upstream <nckx>I tend to trust your judgment on diffoscope, de facto maintainer. <lf94>nckx: Sorry, I was eating supper X) <vagrantc>sure would be nice if it told me which dependencies couldn't be built <lf94>nckx: I literally realized I sent it to the wrong list after going to the "contributing" section of the manual... <nckx>vagrantc: mono, gusb, and elogind, no? <nckx>lf94: And I should have told you *after* approving your message, instead of one second too late. Oh well ☺ <nckx>Eh, *before*. Why can't I think tonight? *nckx just wrote a pointer to a pointer to a pointer, so things are going great. <lf94>(Yes I know Rust doesn't matter at all in this scenario lol) <lf94>I think I'm really going to like the guix community. <nckx>also they are mandatory. <lf94>Also I wasn't thinking C the programming language when you said "must be the C" :p <nckx>Li'l bootstrapping joke for ya. <lf94>I thought maybe you were censoring yourself <lf94>Having been legal here now for ~3 years (Canada), it's not a big deal. <nckx>I'm so bourgeois, that would never have occurred to me. <lf94>But I mean I suppose C will mess anyone up good. <lf94>Has anyone tried running guix using an alternative to guile? <lf94>Or is it pretty tied to guile specifics? <joshuaBP`>lf94: What does it do you mean using guix an as alternative to guile? <lf94>(As in, using Chez Scheme or Racket or something...) <lf94>But guile is an ... what's the term... <lf94>Yeah but there are scheme standards <joshuaBP`>are you suggesting that we should build guix on a different scheme? <lf94>(Note: I'm a scheme noob) <nckx>It's written specifically in Guile, a ‘flavour’ of Scheme. You could port it but it would take some effort. <nckx>joshuaBP`: I think they're just curious ☺ <lf94>But maybe a cool experiment <lf94>I'm a very curious person. <joshuaBP`>racket is probably one of the "better" scheme out there. <lf94>nckx: Ok so uses some Guile specifics then <lf94>Racket has had my attention for awhile <joshuaBP`>lf94: If you wanted to build a complex application in scheme, racket is a good choice. <joshuaBP`>However, the guile developers do have some plans to do "ahead of time" compilation for guile. <lf94>guile seems to be the "cleanest" of the schemes I've seen lately. <lf94>Mmm, the choice of terms it uses <lf94>many schemes differ on the terminology they choose <lf94>also its module system seems really nice? <lf94>Again maybe it's just my lack of exposure / experience <joshuaBP`>lf94: I'm a scheme noob too. I've only really messed around with guile. <joshuaBP`>Though the debugging experience is not as nice as I'd like it to be. <lf94>Like I forget the specific terms <nckx>lf94: Mainly its libraries (like ice-9) and module system, but also (a few) internals/assumptions about Guile. Not undocumented hacks AFAIK, but the original author of Guix is also a Guile maintainer; it's probably hard not to use your intimate knowledge of one in the other. <lf94>But cdr/cadr in most schemes is different in guile IIRC <lf94>just a bunch of small things really <nckx>Really? We prefer not to write ‘old-style’ cdr code in Guix but I'd find that suprising. <lf94>no no that's what I mean <lf94>it doesn't use old style. <joshuaBP`>I personally really like (use-modules (guix records)). I think that should be merged into guile. <nckx>Or do you mean less/more optimised? Guile does do a few things differently (and sometimes slowly) because it was designed to be integrated in C software, with easy access to C data types. <lf94>yeah it has nice C integration too. <lf94>really I see guile to exist for a long time <nckx>It's certainly possible that (say) the car implementation is ‘weird’ compared to other Schemes, for example, I don't know. <nckx>(There's also a #guile channel by the way.) <joshuaBP`>nckx: How many C libraries does guix use? I guess it uses zstd...and probably others? <nckx>joshuaBP`: Oh, that's a good question. I don't know of a list, but off the top of my head: gnutls/gcrypt, lzlib, sqlite, zlib (maybe libbz2 as well?), libssh... <nckx>vagrantc might know of more; they packaged Guix for Debian. <joshuaBP`>nckx: Just curious. Since guile has great C interoperability. We may as well take advantage. <joshuaBP`>but it sounds like from that short list that we already do. <nckx>Well, the integration I meant above was in the other direction (Guile as an extension language in a larger C application, e.g., GNU Lilypond), but yeah. <nckx>We use separate binding packages to talk to libraries, like guile-zstd. I think that was just written for Guix by Ludo'. <awb999>I am schanging some scm files inside a guix container, and guile does not detect the changes, so it keeps the old go files. Can I force a recompile? *raghavgururajan peeps in <dongcarl>Hi all, I have a few colleagues who are building with `--no-substitutes` and with `--bootstrap`, of course they are running into the gnutls 3.6.12 test failure with the expired cert (https://issues.guix.gnu.org/44559). Since they are not using substitutes and don't care about hashes changing... Would `--without-tests=gnutls@3.6.12` sidestep this <dongcarl>Also: would --with-graft=gnutls=gnutls@3.6.14 work? <dftxbs3e>dongcarl, disabling tests will work but --without-tests does not edit the whole dependency graph. <dongcarl>dftxbs3e: By "--without-tests does not edit the whole dependency graph." do you mean that it might still possible to encounter this problem even with the flag supplied? <dongcarl>Is there a way around it without applying a patch? (Just curious) <dongcarl>would --with-graft=gnutls=gnutls@3.6.14 work? <dftxbs3e>Again, I think it will not affect the whole dependency graph <dftxbs3e>I am confused why it doesnt affect it as a whole <roptat>dongcarl, it won't work because grafts are done after building the actual package, so you'll still need to build the ungrafted (failing) gnutls first <dongcarl>roptat: Ah okay that makes sense actually. Would you happen to know whether --without-tests=gnutls@3.6.12 would work? <roptat>but you can rewrite the whole dependency graph with --with-input=gnutls@3.6.12=gnutls@3.6.14 for instance (not sure it'll work, there's probably a reason we have two versions?) <roptat>I suppose if you specify the name of the package, it changes the graph transitively? <dftxbs3e>roptat, I have experience where it did not, especially during bootstrapping. I am not sure by what it is caused, maybe because there's also dependencies of the build systems themselves which these package transformations don't look at. <roptat>but gnutls is not part of the implicit inputs, is it? <dftxbs3e>roptat, I don't know for sure, it may be <dongcarl>Hmmm perhaps I'll try in a fresh VM and report back <roptat>ah indeed, it won't work, --without-tests=gnutls@3.6.12 gives the same derivation as without the option <roptat>you can always try and see if you get a different derivation, with -n <roptat>oh, but that's because there's no gnutls@3.6.12, I think ^^' <roptat>so the transformation option fails <roptat>I see, I'm on a more recent commit <dongcarl>Kk sounds like I gotta experiment a bit :-) <roptat>yeah, just try and see if you manage to make the derivation change <roptat>if it does, it means you probably did it right :) <spudpnds>Does anyone know of the Guix videos from FOSDEM 2021 are available anywhere? <wonko_the_sane>spudpnds: They didn't seem to be last time i checked, which seems pretty slack imo since they were apparently prerecorded <spudpnds>The rest of them don't seem to be up yet. <spudpnds>357 byes of stage0 bootstrap for Mes is pretty impressive! <wonko_the_sane>whole project is super impressive, mad props to all those hacking the good hack. <awb99>I am running guix via a foreign barebome distro. I want to use guix services. Does this mean that I have to create a full docker image with guixos to do that? Or is it possible that I add the guix package to a 'guix environment -container' and then configure shepherd to use the guix services via guile config files? <awb99>Is there any performance Penalty in guixos as a guest os compared to say alpine as a guest os? <iyzsong>awb99: most system services use priviliged port or start as root then doing 'setuid' to a normal user, so they can't run in an environment container (guix system container should be fine, it can also running on a foreign distro) <iyzsong>for guest os, do you mean in qemu with kvm, then I don't think there will any penalty for guix system. <awb99>So say I create a guix container in docker format <awb99>Is this Container any slower than say manually writing a dockerfile with alpine linux? <OriansJ>spudpnds: there are smaller ports than x86 for the bootstrap seeds <awb99>So when I want to build an 'image' that does not use services I would use 'guix environment' and if I need services I would use 'guix container' ? <iyzsong>awb99: yes, 'guix environment' are usually executed by normal users, they can't listen small port (eg: you try to start nginx with its default config) <OriansJ>for example knight-native (bare metal, no OS or bios or microcode) is only 250bytes <iyzsong>awb99: no, 'guix environment' are not for image, it's for one-shot use environment (be in a container or not), 'guix system' are for system with services. <awb99>And guix environment is not able to run guix services? <awb99>I tried to setup shepard directly with 'guix environment' <iyzsong>yes, environment are not for system services <awb99>i managed to write my own config with start/stop guile code. <awb99>But I did not manage to Import the guix Services that let me setup dozens of services via simpel guile config as data structures. <iyzsong>that's fine, if you write your own shepherd services that can be run by a normal user. <awb99>How would I use this bootstrap seeds? <awb99>Is there any example anywhere where I can see some kind of service demo that gets created with 'guix container? <iyzsong>awb99: i think what you want is 'guix system container' or 'guix system docker-image' <OriansJ>which gets you all the way to M2-Planet+mescc-tools which is then used to build Mes.c <awb99>Yes 'guix system container' for development and then once finished and tested I would simply send 'guix system docker' to my hosting provider. <awb99>Is there any example of a os definition with at least one service that I can use as a template? <awb99>The docs seem to refer always to a physical machine. <iyzsong>awb99: you can try 'gnu/system/examples/docker-image.tmpl', maybe then add a 'nginx-service-type' to its services field. <awb99>This is in the guix source tarball? <awb99>So the bootstrap seeds are essentially the core 'storage' element that allows to build a new compiler by using source and the seed. <awb99>But would this effect me in any way when I just want to use guix/sd? <dftxbs3e>awb99, no it will not directly but it is good to know! <awb99>On pretty much all my cloud vendors I can boot Debian/alpine/etc <OriansJ>as the bootstrap-seeds where hand toggled <awb99>Now when I want to roll out a guix/sd image then I first have to know my hardware specs 100% exactly and then I perhaps could find a way to move a iso image to the hard drive. <awb99>A pretty difficult job if I only use a few servers in a cluster. <OriansJ>awb99: the POSIX pieces are probably what you want <OriansJ>then your kernel need only support a mere 18syscalls <awb99>The other solution is either a default installer image or a nonfree installer image .. and then somehow bootstrap from there via ssh. <OriansJ>the bare metal work is the eliminate the kernel bit entirely from your TCB <OriansJ>the bootstrap-seeds are designed for audit and replacement <awb99>Are there docs how to build for x86 with the posix seed? <OriansJ>a 5 year old should be able to replace the hex0 seed. <OriansJ>and elf headers hand written to work from too <OriansJ>awb99: if you need more details #bootstrappable is where the bootstrapping work is being done. <dftxbs3e>cbaines, hey for some reason things are slow. I don't think it's GNU Guile but rather some GNU Guix bottleneck, it looks like a locking or network I/O bottleneck, it's slow to find missing build inputs I think. Also it does not parallelize well, I have 4 parallel builds configured yet it seems it only checks missing build inputs for one package at a time. <dftxbs3e>cbaines, I think there's some kind of locking bottleneck <dftxbs3e>cbaines, also really wasteful usage of read (1 byte) and select system calls <reepca>Anyone else have trouble with browserpass-extension not actually pasting to clipboard? <reepca>er, I guess the proper term would be "copying" to clipboard, not pasting <dftxbs3e>efraim, I think maybe we should use the Rust frontend of GCC when it's ready to compile Rust packages instead of developing something to substitute cargo. <dftxbs3e>It would maybe be a better time investment <cbaines>dftxbs3e, yeah, the code for checking substitutes for inputs is slow and unreliable <cbaines>the code in Guix isn't particularly well suited for use in other parts of Guix or other tools like the Guix Build Coordinator, but I've been working to change that http://issues.guix.gnu.org/45409 <dftxbs3e>by the way this substitute thing trumps all the performance of the agent <dftxbs3e>do you think I can apply your patchset temporarily here? <cbaines>I think it's slow just because it's not particularly clever about making network requests <dftxbs3e>instead of the guile-written HTTP client <cbaines>dftxbs3e, one thing you can do with the agents for powerpc64le is just specify one substitute server, that way the agent will only check one substitute server for the inputs, rather than however many are currently specified ***penguwin4 is now known as penguwin
<rndd>hi everyone! how i can see logs of service in guix system? <sneek>rndd, mroh says: I think, the error you're getting with you kernel config is because of the initrd module safety checks, so you might try `guix system --skip-checks` or make sure all the modules that are tested are in your kernel (see default-initrd-modules in ~/gnu/system/linux-initrd.scm for the list of modules) <cbaines>rndd, depends on the service. It's probably either a file in /var/log, or on the tty containing the shepherd output <rndd>cbaines: i cannot run dhcpv4-daemon. so, there is nothing related in /var/log. how i can check tty? <rndd>cbaines: thanks, found error <sneek>Welcome back wleslie, you have 1 message! <sneek>wleslie, nckx says: What happened? Probably worth reporting a bug. There's no technical reason Guix couldn't keep better track of active GC roots. <wleslie>that's a great idea, I bet I could reproduce that one <wleslie>I mean did my substitute fail to match? has none of the FLAGS_FOR_TARGET worked? <wleslie>well, FLAGS_FOR_TARGET isn't present anymore because it wasn't successful in setting a path for the crtfiles; I even removed it from the default that cross-base gives me <rekado>there are at least two known problems in this version: one is that it doesn’t start. That’s a somewhat recent regression and should be rather easy to fix. The other is that it doesn’t stay alive and segfaults pretty quickly. But you first need to get it to start. <kondor2>can i post while being unregistered ... <kondor2>I'm got further along configuring my guix system and now i'd like to add gnome-keyring to the system services <kondor2>so, currently I have `(services (cons* (service (gnome-keyring-service-type <cbaines>any ideas why this doesn't work? make check TESTS="tests/guix-daemon.scm" <dftxbs3e>kondor2, you don't need to specify the configuration if you have none <cbaines>ah, to answer my own question, that file doesn't exist <dftxbs3e>kondor2, cool then it should work out of the box with gdm, since this is what runs it through PAM <kondor2>dftxbs3e but I have no ssh-agent running in the backgroun <dftxbs3e>kondor2, ssh-agent is run by a .desktop file later on <dftxbs3e>does openbox support the XDG Autostart spec? <kondor2>dftxbs3e I have no idea, perhaps i should switch to Gnome :D <dftxbs3e>kondor2, you could choose not to use the service, you could instead make sure to install gnome-keyring and gcr (needed for prompts) package <dftxbs3e>Then start gnome-keyring in a shell script that autostarts at login, I think .bash_profile might do it <kondor2>or, i could run it usin logind right? <dftxbs3e>gnome-keyring adds the service to autostart through PAM, but through PAM you can't choose components that start <dftxbs3e>oops gnome-keyring-service-type adds it to* <dftxbs3e>the thing is gnome-keyring generates an environment on stdout that must be set everywhere, so it must run early for all childs to inherit env <dftxbs3e>.bash_profile might do that, not sure, but PAM is best way to start it, it just doesnt support components :-/ <kondor2>dftxbs3e i suspected so, which means i'd run it in xsession, or profile <dftxbs3e>you could add that patch in your config then alter gnome-keyring-service-type <kondor2>dftxbs3e i'm thinking to go back to good old keychain <dftxbs3e>I run gnome-keyring on XFCE, but I let XFCE start the gnome-keyring for me, it has an explicit option for it in it's settings <kondor2>dftxbs3e: this means you'rw w/o a keyring daemon when you use a virtual console <dftxbs3e>Also gnome-keyring depends on dbus, but it's included by default in %desktop-services <dftxbs3e>kondor2, I never use the virtual console <kondor2>dftxbs3e they look nicer these days :) <dftxbs3e>well I never use it unless I am debugging something that needs to <dftxbs3e>Also, Linux lost support for scrollback there recently <kondor2>dftxbs3e i sometimes run emacs from them, like the font <dftxbs3e>cbaines, can I use your patchset not yet merged into GNU Guix? Does it work? <cbaines>dftxbs3e, it probably wouldn't do any good <cbaines>dftxbs3e, it's just refactoring the code <dftxbs3e>what about improving substitute mechanism altogether? <cbaines>once that refactoring has happened, I can then begin to tune fetching substitute information in the build coordinator for performance <dftxbs3e>I love when systems perform and scale well :-D <cbaines>As I said earlier, I'd look at the options you're passing in, particularly --non-derivation-substitute-urls <dftxbs3e>Maybe a little bit faster but not that much <wleslie>(display (find-files ...)) gives nothing, so that's interesting <ngz>woohoo compilation succeeded! \o/ Massive merge incoming. <wleslie>having got something out of find-files, it's still not matching the line. why would the substitute* expression ("#define CAPOS_TARGET_(CRT.) \"(.*)\"$" _ crtx filename) not match the line `#define CAPOS_TARGET_CRT1 "crt1.o%s"` ? <wleslie>nushell looks very nice. not only is the concept of structured data in a shell a breath of fresh air, nushell's concept of shells in shells looks pretty cool; like it's heading in the direction of a really nice job control system <ngz>It contains interesting ideas, yes. If only it was Scheme-based ;) <wleslie>there are quite a few local maxima in this area, it's true <ngz>One of its dependecies is a Nintendo Entertainment System emulator… Odd. <wleslie>eshell, scsh, shill, xonsh are all a bit nice <wleslie>`(substitute* files ((".*" all) (format #f "~a hairy" all)))` puts hairy at the start of my lines. I guess this regex engine is not maximal munch <ngz>To answer your earlier question <ngz>Ah nevermind, I misread it. <ngz>Your regexp looks correct indeed. <wleslie>id3lib contains a regex match for a cpp directive, and it starts it with //\\. I'll give that a shot. <wleslie>at least escaping the #. I'm not familliar with that being a special character. <dftxbs3e>efraim, also, since rust lib packages are currently source packages they shouldnt need to be compiled multiple times for different architectures, would they? <wleslie>bah. the docstring for (guix build substitute) says "Be careful about using '$' to match the end of a line; by itself it won't match the terminating newline of a line." <wleslie>with \\s*$ instead it seems to work! <pinoaffe>wleslie: that's how $ works in almost all regex engines <pinoaffe>I think it's to prevent issues when the last line of a file does not end in a linefeed, or when a file mixes dos-style and unix-style line endings <wleslie>in pcre, $ matches 'the end of the string or just before the newline at the end of the string' <pinoaffe>yes, and if you're doing per-line matches that equates to matching the "spot" just before the linefeed <wleslie>here it's not matching before the newline, it's evidently matching after the newline, which I'm eating with \\s <cbaines>dftxbs3e, no, the Guix Data Service doesn't have any information from the network dependent linters yet <kondor2>ok, i've put xfce-desktop-service into my config and upon login, wicd complains that it cannot connect to d-bus ... anyone had similar experience? <dftxbs3e>kondor2, I don't use wicd so, don't know, post your full config <dftxbs3e>kondor2, no idea here, what's the error exactly? <pinoaffe>i guix! has anyone looked into packaging wordpress for guix and setting up a wordpress service for guix? <dftxbs3e>kondor2, also, you choose to separate it in several files, maybe share those as well <kondor2>dftxbs3e thanks for looking into this <kondor2>dftxbs3e: Could not connect to wicd's D-Bus interface. <dftxbs3e>pinoaffe, no service yet it seems, since there isnt even the package <kondor2>dftxbs3e should I even install wicd? <kondor2>or, does it come together with the desktop <kondor2>dftxbs3e i tried it, but there was an error ... <kondor2>as for specification files, i doubt you'll uncover anything in there, just bunch of packages <pinoaffe>dftxbs3e: yeah I saw that patch, it seems to download the published build artifacts rather than building them itself <kondor>dftxbs3e ah, now i see, network managare comes as a default <dftxbs3e>pinoaffe, PHP projects are sources though, no? <kondor>but, i dont care, nm is fine, too <bqv>Where do you guys keep your config? <bqv>do you have any dependent files? <dftxbs3e>bqv, I prefer to include everything in the same file <bqv>How do you handle e.g. Ssh keys, embed them? <dftxbs3e>bqv, system configs arent designed for that AIUI, more so guix home manager but I don't use that <dftxbs3e>if I have secrets I store them in separate files <dftxbs3e>Such as passwords to be given to services <bqv>Just hanging around elsewhere on the filesystem? <dftxbs3e>guix-build-coordinator-agent needs one for example <kondor>dftxbs3e ah, so my string splitting function was completely useless as `specification->package+output` knows how to do that automatically <pinoaffe>dftxbs3e: yes, but minified javascript is not <kondor>bqv i keep package specifications separately, because I am currently experimenting, plugging things in and out on the fly so to speak <kondor>dftxbs3e : just ripped off vm services from your config <dftxbs3e>kondor, go for it, consider the config under GPL3 or later <kondor>bqv but once i settle on a config, i'll stuff it all inside one file, similar to my emacs Org configuration <user_>Does anyone have experience configuring pulseaudio? <ngz>user_: I probably cannot help you, but I think you should ask directly what you have in mind about Pulseaudio. Your question is rather vague. <ngz>Unless you are some head hunter or some such looking for talented people ;) <user_>I see the problem now, the laptop's sound card isn't recognised <bqv>Were really good rubber ducks! <user_>Now comes the actual question: how do I add kernel module options on Guix? <luke-jr>guix time-machine: error: the build users group `guixbuild' has no members <user_>ison: I'm not clear on one thing, I'm looking at the default config my installer created, if I edited the scm and reconfigured my system, all the packages I installed that isn't explicitly in the config will be removed? <ngz>No, there are system packages and user packages. Packages not in the config are user packages. But they are not removed when reconfiguring your system. Your $HOME is not wiped up either. <ngz>Actually, you don't need to put much packages in your config, unless you need setuid. <pkill9>tfw when you fix a build error just by clearing the build directory and reconfiguring <awb99>I am building 'guixx system docker-image' , and I want to make the build reproduceable. I know the current Profile keeps all settings. So would I just 'guix describe' and store that output in a git repo? Ideally I would want to create the profile in a git folder. Is this possible? Does it make sense? <user_>Hello again, when editing kernel-arguments in the config scm, what does cons* mean? <user_>Is (cons* "X=x" "Y=y") valid? <ngz>cons* means "one or more `cons'" <ngz>(cons* 1 2 '(3)) is equivalent to (cons 1 (cons 2 '(3))), i.e., (1 2 3) <ngz>(cons* 1 2) is equivalent to (cons 1 2), so (1 . 2) <user_>Does anyone have snd-hda-intel enabled in their kernel? <user_>Can you share the kernel-arguments line in your config.scm? <rekado>awb99: you would record the system configuration and the Guix channels you used in a git repository. You don’t need to track the generated profile. <janneke>user_: my lsmodules has: snd_hda_intel 53248 3 <janneke>i don't set kernel-arguments for sound, i only have <janneke>(kernel-arguments (cons "kvm_intel.nested=1" %default-kernel-arguments)) <janneke>which might no longer be necessary, even? <pkill9>does anyone know of a decent budget printer <cbaines>I'm getting a "requires 'networking', which is not provided by any service" error from trying to use the childhurd service <cbaines>I'm guessing it has network access, but maybe there's no service providing networking... <nalaginrut>hi folks. I've installed newlib-nano, but the ld still complains no -lc_nano, any idea? <janneke>cbaines: the childhurd has a static networking service, defaulting to the qemu guest ip <cage_>Hi! I am trying packaging a program, its make file generates a file (in a directory src/) that is later used when compiling the executable, i am using (invoke "make" ...) to generate the file before the build phase but then, when compiling the program, the build system complains that can not find the file <cage_>i can not understand where the make file is generating the file and where should i move so that is available to the building system <ngz>You can use guix build -K your-package and have a look at /tmp/guix-build-your-package.../ directory <cage_>seems that the file is in the correct place /tmp/package/source/src/config.lisp <cage_>i wonder why it is not found then <cage_>unfortunately my package uses a mixture of GNU toolchain and ASDF (it is a common lisp program) <cage_>maybe I have to instruct guix to generate the file before the package is copied in the store? <cage_>seems that adding (add-after 'unpack (lambda (...)) works :-) <cbaines>Hmm, I'm trying to run the guix-build-coordinator-agent service in a childhurd, but the wrapper script uses bash for linux... <cbaines>I wonder if that's a issue with the package definition <cbaines>Yeah, wrap-program does (which "bash"), which will probably find the host bash when cross compiling <QsDDC>is there a way to make my changes persistent on the installer image(booted from USB)? I made a lot of 'guix install' commands and have some other files in root's home dir and want to keep everything on the installation USB so I don't have to install it again after rebooting if my system install failed. <cbaines>QsDDC, I don't believe so, I'd suggest copying those files elsewhere. You can't do much about the installed packages. <QsDDC>cbaines: thanks. I'll just copy my files & bash history then. <janneke>cbaines: yes, that's a bug; see also 330f3ab7dd55ac95aad109000feca2d0d96a1e49 <cbaines>PotentialUser-95, that's unfortunate, looks like this happened post staging merge. Unfortunately getting out of this situation just requires someone fixing the broken package, feel free to file a bug with Guix about the breakage as well <cbaines>janneke, cool, I'm glad someone knows about it. I can't see 330f3ab7dd55ac95aad109000feca2d0d96a1e49 though, where is the relevant repository? <kondor>PotentialUser-95 can you not pack an unbroken version? <cbaines>kondor, if you're referring to guix pack, how would that help? <kondor>cbaines i am not aware of the entire context, but can't you base a package on a previous/working version of the dependency <cbaines>kondor, packaging digikam on top of a previous version of Guix would be one strategy, that can introduce problems that are fixed in the current Guix though, but maybe that could be an option <lfam>PotentialUser-95: We do have two version of VTK packaged <lfam>I had hoped we could fix this issue before it got deployed to master but in the end this is what happened <lfam>It's helpful that VTK offers some patches to work around the problem but it seems like the long-term fix will be related to freetype <awb99>I am generating a docker image guix system with os definition from a guix installed on another os. <awb99>Did anyone manage to get this working? <fnstudio>hi! =) i'm trying to "GNU-build-systemise" a simple guile project of mine (i.e. adding the Autotools files) and i get this error where './configure' doesn't seem to be able to find guile and other dependencies <fnstudio>dependencies that are offered by the foreign distro seem to be found <fnstudio>raghavgururajan: oh, i see... interesting <lfam>I think it would be more correct to use `guix environment your-package` <lfam>That would create a development environment for your package <cbaines>fnstudio, are you trying to provide Guile with Guix? <fnstudio>lfam raghavgururajan cbaines: thanks all for your help, let me give some more context <lfam>But if that doesn't work because you don't have a Guix package definition for your package yet, you could use `guix environment --ad-hoc autoconf guile etc` <fnstudio>it's a simple example script written in guile, the objective is to familiarise with Autotools (and Guile and Guix, of course) <nalaginrut>hmm...newlib-nano doesn't have libc_nano.a, maybe I should report. <fnstudio>so there's no dependencies other than what's needed by autotools itself and to launch a guile script <fnstudio>eg, it says: `PKG_PROG_PKG_CONFIG: command not found` while pkg-config is correctly installed in guix <fnstudio>ok, great, i'm going to try with environment (`guix environment --ad-hoc autoconf guile ...`) <fnstudio>cbaines: yeah, great write-up, thanks! it's actually open in my browser, together with a couple of other links <fnstudio>cbaines: i'm trying and dissecting things here and there to understand things as much as possible <librecat[m]>i am using guix package manager on linuxfromscratch systemd edition to get a gui and browser faster but guix fails saying failed to clone builder proccess invalid argument i have all namespaces enabled except for user_ns which deps do i need to get the sandbox working or how do i disable it <fnstudio>hm... i've tried with `guix environment --ad-hoc autoconf guile pkg-config` but i still get the error about pkg-config missing <cbaines>fnstudio, perhaps try adding automake to --ad-hoc <cbaines>librecat[m], can you share the exact error? <fnstudio>cbaines: oh oh oh... that made it! how did you do that? =) <cbaines>fnstudio, the automake package sets out a search path for ACLOCAL_PATH <cbaines>I think that helps find pkg-config stuff <librecat[m]>i just said it while building profile at the end of guix install <librecat[m]>error while cloning builder proccess invalid argument <fnstudio>cbaines: yes, it definitely helped - cool, hopefully that'll get me to finalising my example package <cbaines>librecat[m], I'm unsure, but I think you might need to permit user namespaces to have the guix-daemon sandbox builds <cybersyn`>heya guix, hope everyone is having an alright night. having a bit of trouble getting a wacom tablet running with guix. does anyone have a recommendation as to what to add to my config.scm to get this baby goin'? I'm using sway/wayland, which could be the problem <cbaines>hmm, I don't have any more suggestions unfortunately <cbaines>probably pass --disable-chroot to guix-daemon <QsDDC>does Scheme/Guile have something like 'elem' in Haskell? I want to remove every initrd-module from %base-initrd-modules that is in my defined blacklist. <cbaines>QsDDC, I'd look at SRFI-1, either remove or lset-difference might help <janneke>cbaines: ah, it's called 38b9af7c92344a17b6680ebd2aeea14171f84a1c <janneke>(apparently a rebase happened, got that from an email and it was still in my repo) <cbaines>janneke, ah, thanks, that might be useful, I've got around it by changing PATH when calling which so far <cybersyn`>raghavgururajan: i'm getting nothing from my wacom bamboo <cbaines>janneke, I'm now trying to figure out how to get the wrapped GUILE_... paths to be correct <cybersyn`>i just discovered the inputattach service, will give this a spin <raghavgururajan>cybersyn`, I was about to suggest that but I thought you meant the issue was more than that. <cybersyn`>nope, just trying to get a wacom tab on guix for the first time. thanks tho <raghavgururajan>cybersyn`: Cool! You may have to tinker with inputattach-configuration, like baud-rate and dev. <kondor>Talking about services, can one use xinetd thru inetd service? <rekado>nalaginrut: you should install the toolchain package, not just newlib <rekado>raghavgururajan: this looks pretty readable to me. Why make it shorter? <raghavgururajan>rekado, I was suggested "You might want to regexp the middle part in case it is changed.". <rekado>hmm, I still think it’s kinda nice the way it is, but if you want to elide anything you can use “.+” <rekado>so I think having the match string explicit avoids problems in the future <rekado>the two strings also seem barely related <rekado>they share _install_dir, but that’s it <cbaines>guile-lib doesn't seem to cross-compile properly to the hurd, I see 64-bit .go files <awb99>Should cross-compile work geentally speaking? So could I say use x86 server to build a arm package? <cbaines>it's only possible to even compute the derivations for a minority of packages, and the failure rate will probably be higher than normal <cbaines>you're probably better off looking at emulation, that's also likely to not always work, but less so than cross compilation <dftxbs3e>lfam, I am good :-D What about you? Do you feel well today? <dftxbs3e>cbaines, if you are still here, is there a feature in Guix Data Service to observe failure by branch? <dftxbs3e>cbaines, okay thanks! can it be clicked to from anywhere or just forging URLs? <cbaines>yeah, I'm uncertian about linking to it, because the data's not always up to date <cbaines>There's also a package-derivations page for a revision, and you can filter by build status there as well <cbaines>that has the slight advantage that builds are grouped by package <dftxbs3e>nckx, marusich, iyzsong: Thank you so much for vouching my Commit Access request! Need to wait for someone to actually approve me now :-D <cbaines>The Guix Build Coordinator is configured to retry automatically <dftxbs3e>So 30 packages failing, 1309 succeeding, 14890 scheduled <dftxbs3e>cbaines, I remember you told me something about checking for the difference before, between x86_64-linux and powerpc64le-linux <cbaines>I think I was thinking about the impact on merging the branch, as if there was a set of changes which didn't impact lots of non powerpc derivations, then that would simplify merging to master <user___>Hello, I'm having trouble installing Chinese fonts. I've installed font-adobe-source-han-sans:cn then rebuilt the font cache with fc-cache -fv but I still get nothing <dftxbs3e>user___, I usually do: "fc-cache -rfv" - not sure what the r option does though? <cbaines>dftxbs3e, in terms of detecting things that build on x86 but fail on powerpc, that need's implementing in the Guix Data Service before it's possible <user___>dftxbs3e: The characters show up as the replacement characters <dftxbs3e>user___, I will admit to you the font situation is a bit hazardous to me. Often, you configure things, they don't work then you do something else and come back and it magically starts working. Some kind of cache or something. Does the front show in fc-list? <user___>Yes, the font packages show up on fc-list, but the Chinese characters aren't displayed properly in the names of those packages. Should I move the TTF files to .local/share/fonts manually? <awb999>I am building a docker-image with guix. <awb999>When i want to add packages to the os config, the build breaks <awb999>this works: (packages %base-packages) <awb999>I printed wget variable, and it is a package struct <awb999>also the first build is extremely fast, <awb999>and the second build takes long time <awb999>this is the failure when I add wget <awb999>when ;wget is in, i builds foreever <awb999>when I use wget I get the error as on the pastebin <rekado>awb999: what does “guix describe” say? <awb999> --root=./artefacts/docker-pinkgorilla.tar.gz \ <dftxbs3e>awb999, what's your GNU Guix version? "guix describe" ? *dftxbs3e builds the config to try and reproduce the problem