<bavier>has anyone else tried packaging frama-c? <efraim>mbakke: I had that same commit queued up (minus the beer) <calher>Does GNOME Software work with Guix yet? <buenouanq>what do you mean? it's shipped with gnome from the very beginning... <buenouanq>works great, I've been using it for years now <calher>buenouanq: Not GNOME software. GNOME Software. <calher>GNOME Software does not work with the Guix package manager. <buenouanq>I must say I have no idea what GNOME Software is then. <buenouanq>b.. but why would anyone care about that when Guix exists?... <Blackbeard[m]>gnome of course has this very creative names so the users know right away what you are talking about <Blackbeard[m]>buenouanq: it can be used to install with apt, pacman, flatpak, so I guess in theory to install with Guix too <buenouanq>but that negates all the benefits of using Guix at all in the first place <buenouanq>ALL package management and installation should be done through guix <Blackbeard[m]>buenouanq: I don't see how it could be bad to have a graphical application to install Guix packages <buenouanq>oh, maybe it's me that's still misunderstanding <buenouanq>GNOME Software is just a graphic front for an arbitrary PM? <buenouanq>it would limit what you could do and learn about, but might be nice for some people in some cases <calher>buenouanq it would be nice for quickly installing packages. Press the logo key, type part of a package name, click the search results returned from GNOME Software. <calher>I don't see why understanding "GNOME Software" is difficult. We have stuff like "Google Search" and "Google Maps". <buenouanq>the only thing I thing a graphic frontend would help make easier in package installation is filtering results when you don't know what you're looking for <buenouanq>I was the one who didn't even know what it was. <calher>buenouanq graphical front ends are important. The free software movement aims to make it possible to do everything people currently do with computers using only free software. I know my mom and grandma would prefer an app store to typing "apt install" at the command line. <Blackbeard[m]>So, I’ve been designing gnome-software to be pluggable. This means you can write an AppStream plugin to provide things like icons and screenshots for not-yet-installed software. You can write a plugin to ask ostree to update itself, and also a plugin to ask PackageKit to update a specific package. <Blackbeard[m]>GNOME Software overall plan – Technical Blog of Richard Hughes <calher>If Richard Stallman thought it was a good idea to make everyone install packages from a command line interface, he would have also thought it wasn't important to make a drop-in replacement for the dominant operating system today. Just make them use Free Lisp Machine OS! <calher>Unix people are wrong. Let them use GNU Lisp OS. <calher>But Richard understood that it was important to be pragmatic, and listen to the needs of the general public. <buenouanq>I understand the importance of making things accessible. I worry about catoring to casuals who not only don't know what they want (or much of anything), but don't care about freedom etc. <calher>I used to use i3 and irssi, but I deliberately switched to HexChat and GNOME. <calher>Blackbeard[m] Dogfooding for the free software movement. Creating and living in viable workflows for everyday life that can be adopted and used by the general population. <calher>Using i3 distracts me from researching and perfecting workflows. <Blackbeard[m]>right now I enjoy StumpWM but I installed GNOME in my mom's laptop <Blackbeard[m]>but I can't go back to GNOME, window managers are a pleasure <Blackbeard[m]>and I can't stand the huge GNOME title bars that waste so much space <brendyyn>calher: I often think about how to provide a default system that works for most people, but can also work for me, but i need keyboard based window management <calher>brendyyn I use GNOME mostly with the keyboard. <brendyyn>I'm on i3 and KDE at the moment with two computers <calher>Blackbeard[m] GNOME apps don't use title bars. They use header bars, which save space that is taken up in traditional applications by an empty title bar, a menu bar, and a tool bar. <buenouanq>I want a librebootable Thinkpad - Which should I get? <Blackbeard[m]>calher: yeah that's why I use StumpWM I have none of those things <calher>Blackbeard[m] So GNOME saves space and yet still adapts to a wide variety of interfaces. <calher>buenouanq Is there something besides the X200? <calher>Blackbeard[m] We can't change society so that everyone uses Ratpoison and xterm. <brendyyn>calher: I am curious about your workflow <calher>brendyyn Which task are you interested in? <brendyyn>calher: you open two programs and want to put them next to each other each taking up half the screen <brendyyn>My mum loves photoshop too much to let me reinstall her computer haha <calher>Blackbeard[m] : LogoKey ter RET LogoKey LeftArrow LogoKey abr RET LogoKey RightArrow <calher>GNOME's keyboard shortcuts are actually really nice. <maddo>is libreboot even working right now? Last I checked, master didn't compile and you had to use an old build with tons of bugs already fixed in coreboot <maddo>right now I think just building coreboot without any blobs is a better idea than straight libreboot <sneek>Welcome back civodul, you have 1 message. <civodul>dongcarl: interesting, i don't know why the cross-compiler does things differently <nly>how can one checkout a specific git "branch" in guix package definition? <marusich>nly, Git branches usually refer to different commits at different times. Therefore, they are not a good choice to put into a package definition. If the branch points to a different commit next week, then your sha256 hash value will become invalid, and the build will fail. <marusich>That said, I suspect you could probably use a branch name in place of the commit, if you really wanted to. But you would encounter the problem I mentioned. <marusich>It's better to pick a commit, since it always refers to exactly the same thing. <brendyyn>I guess you could add a ;; comment stating that the commit is from a particular branch <nixo_>Hi, I'm trying to upgrade julia to version 1.1.0. In the progress, I wanted to run the various phases manually to check what's going wrong and try to fix things "live". <nixo_>I'm using `guix environment --load=./my-julia.scm`. When inside, how can I run the phases? <civodul>each phase is a Scheme procedure, so you'd need to start Guile, load the "...-julia-builder" file, and somehow extract phases from there <civodul>now, what i do is simply "guix build -K foo", jump in the build tree, and fiddle from there <marusich>FWIW I have wanted the ability to do that, as well... In a dream-land, it would be nice to say "guix enter-build --stop-after-phase=build" <marusich>But yeah, I wind up just doing what civodul mentioned. <marusich>It's more painful the more hackery you do with phases. <nixo_>civodul: thanks. because in nix I think it is possible <nixo_>civodul: I'll look at yours and try to progress from there, thanks! <civodul>true, that's easier there because everything is shell code, so i think you can just do $buildPhase and be done with it <marusich>nixo_, you can arrange to stop the build at a certain phase by modifying the phase definition to return #f. <marusich>Then --keep-failed will stop at that point. <marusich>I need to get some sleep. Goodnight and good luck! <nixo_>marusich: thanks and good night! <asterope>can't finish my system reconfigure because mozjs@38.2.1.rc0's origin link is dead, managed to find an alternative one and build the package with `guix build`, but I can't make guix to use that build <asterope>I tried to lanch `guix publish` on localhost and point my reconfigure command to it, but it still tried to build the package by itself <civodul>asterope: did you authorize the public key for your substitutes? <asterope>Yes I did `cat /etc/guix/signing_key.pub | guix archive --authorize` <nly>In package qt (file qt.scm), QtWebEngine is disabled because it depends on bundled Chromium, can I build QtWebEngine with guix's Chromium package? I need QtWebEngine for a package "Nomad". <civodul>i mean you need to authorize the key of the machine that built mozjs on the machine that will download it <asterope>I built it on localhost, then published it on localhost, and then tried to use it from localhost <asterope>There are no errors related to the last part, still guix doesn't seem to use the package I built but instead tries to build it again <nly>i'll try to remove (delete-file "qtwebengine") and specify chromium as a build input. <pkill9>oh i don't think that will work, i think it needs the chromium source itself, but then again i don't know how qtwebengine works <pkill9>maybe another distro's qtwebengine package might yield some insight <asterope>The machine is authorized, any other ideas? <civodul>asterope: note that yuo could also simply use 'guix copy' to copy from one machine to another <civodul>then you need to make sure both machines are really trying to build the same mozjs <asterope>Managed to resolve my problem by using `guix download` on the alternative sourcode <civodul>so check the output of "guix build mozjs -d" on both <asterope>Someone should update mozjs origin, I couldn't find the upstream file (I used some backup found on the internet) <brendyyn>Anyone feel like working on updating Calibre with me? <brendyyn>I've spent hours trying to make the tests work. and ebook-viewer segfaults ***MinceR_ is now known as MinceR
<civodul>asterope: did you report the exact file that you were missing to bug-guix? <civodul>normally our build farms keep a cached copy of all the source files for a while <brendyyn>i need to switch email providers _again_. anyone using mailbox.org think its good? or have a recommendation? *kmicu has a self-hosted mail (and dismail.de as a backup). *kmicu babbles that if Ubuntu 19.04 has ‘Disco Dingo’ codename then we could name Guix 1.0 ‘Ablaze Altair’. <brendyyn>ill probably hang myself if i have to suffer trying to run my own mail agin <brendyyn>i had my own then gave up, switched to lavabit.com (shutdown), then openmailbox.org (shutdown and ran away with my money), now protonmail (didnt realise there is only a proprietary client for it) <brendyyn>so i will likely cancel that and find yet another <brendyyn>except i checked HN and some people think their practices are dodgy *kmicu has HN 0.0.0.0 entry in hosts file. <brendyyn>anyway i cant contribute to guix until i get something working <brendyyn>I Probably shouldn't have picked debian for my vps, everything is so old. Hmm, package mailinabox for guix? <roptat>I don't know what your needs are but if you're the only user of your domain, no need for a database when you self-host :) <brendyyn>just need it to work and not make me rip my hair out <brendyyn>dovecot used to change its config format with updates back when i used it <roptat>I use it with the guix configuration, so the syntax doesn't change for me :p <brendyyn>i like how the guide recommends a fresh installation since there might be files littering my debian installation <civodul>roptat: what's the status of the guix.gnu.org setup? <civodul>do you think it'll be ready for 1.0? <roptat>but it will depend on the reactivity of gnu's sysadmins :) <civodul>right, so we should leave them as much time as possible <roptat>the only issue I see is that our certbot service doesn't support the dns challenge we want to use <roptat>i think we want to rename certificate-configuration to certificate-http-configuration and have a certificate-dns-configuration or so <roptat>or have a "challenge" field in certificate-configuration to select the right challenge type <wednesday>Why is rottlog not something that is enabled by default? If not in the %base-services at least %desktop-services, who likes having 15MB+ logs <wednesday>I'm talking individual logs, because rottlog rotates logs like /var/log/messages <civodul>roptat: i'm not a certbot expert, but having a "challenge" field sounds reasonable <wednesday>civodul: Do you have any reason rottlog isn't default? heh <roptat>civodul, oh and we need to manage a TSIG key without which knot will refuse to update <roptat>and it must be kept private because we don't want people to be able to send updates to our DNS :) <roptat>and I'm afraid the secret will have to be in the configuration file for knot <roptat>it doesn't support loading it from an external file <roptat>so it will be stored somewhere in the store <wednesday>Maybe ill ask in the devel list if people think rottlog deserves being in %base-services, since syslog deserves to be there I don't really see why rottlog wouldn't be <civodul>roptat: what if the secret is in some out-of-band file in /etc ? <civodul>oh you said it cannot load other files, nvm <civodul>that service actually looks rather empty ;-) <civodul>one thing we could do is generate the config file from a template at activation time <civodul>that is, the service would generate a template in the store; and then at activation time, we take that template, insert the secret, and store it in /etc <roptat>that would need some change in the service definition... <roptat>I don't think we support that yet, but that can be done easily <roptat>can you check something is a file-like object? <civodul>roptat: i think so but it's usually a bad idea <civodul>because the notion of "file-like" is extensible: you can define "gexp compilers" for new types <civodul>so usually it's enough to do #$something <civodul>and then, at run time, if "something" happens not to be a file like <apteryx>Q: If I packaged something which is not reproducible on master but is reproducible on core-updates (due to a fix in a build system), should that package be committed to core-updates as well? <apteryx>Or it's OK to have in in master, knowing it is not reproducible until core-updates get merged? <civodul>apteryx: it's ok to have it in master <roptat>I think so too, it will be fixed when core-updates is merged, rigt? <apteryx>alright! I'll proceed and merge the python-robot-framework series in master, then. <roptat>I've been contacted by people from shark-bait.org because of my involvement in android stuff in guix <roptat>it's a project to run gentoo on a phone and have android only in an lxc on that system (iiuc) <roptat>among other things they want to be able to build android from the phone and offer more customization <apteryx>build android *from* the phone? sounds crazy. <roptat>in doing so, they will probably help / need help on understanding and making sense of android, so I think it can be beneficial :) <roptat>yep, so they are a bit interested in the binary transparency that guix provides, so you don't *actually* build anything on the phone ^^ <roptat>it seems they managed to boot a few phones on gentoo and then an android system from gentoo <roptat>I'll try to boot a spare phone this week-end because it seems to be supported :) <roptat>I could even be able to install guix permanently on that one ^^ <joshuaBPMan>Hello, my computer just locked up on my a second ago. I'm running sway. It stopped updating the screen. It did not accept any keyboard commands and I was unable to switch to a virtual console. Is there a command that linux listens for to forcibly stop the graphical program and get me to a virtual console? <apteryx>joshuaBPMan: there's supposed to be a Linux systemrq key that triggers OOM killer, but never really got that one to work <apteryx>otherwise you can try just switching to tty2 (Ctrl Alt F2), loging and kill sway from there. <joshuaBPMan>apteryx: I think I want to learn how to raise the elephant. I could not switch to a virtual console. <joshuaBPMan>Addmittedly I only waited like 15 seconds before hard rebooting. Maybe I should have given it a minute to try to figure it out. <joshuaBPMan>Also where is this magic_sysrq_key on a laptop? Do most laptops even have them? And doesn't it switch to a qwerty layout when you do this? <apteryx>with possibly some funky Fn combination <joshuaBPMan>apteryx: What'l AltCar? and I don't think I have a PrintScr <roptat>and with "k" I'd say to "Kill all processes on the current virtual console (can kill X and SVGALib programs, see below)" <joshuaBPMan>roptat: I don't have that key that you mentioned in the svg image... <joshuaBPMan>and I'm pretty sure I don't have a print screen key, but <roptat>well if it's frozen, you can't do anything <joshuaBPMan>roptat: is that true though? that's what raising the elephant is for. If things lock up, then I can raise the elephant. Though I think if you have to raise the elephant, then you probably have to reboot anyway... <apteryx>joshuaBPMan: the use of the 'RSEIUB' sysrq is to attempt to halt things properly before rebooting, so to minimize the risk to corrupt the filesystem or something. <joshuaBPMan>apteryx: ahh bummer. It sounds like an outdated feature then. Since filesystems have journaling, filesystem corruptions is pretty much non-existent. <apteryx>it might also try to close applications neatly (sql database or other) although I haven't validated that <roptat>say that to the SD I tried to use with my arm board :/ <roptat>I tried to install the guix system on it and corrupted the file system after guix pull, four times in a row <joshuaBPMan>roptat: You may have corrupted the filesystem, but didn't fsck fix it automatically? <joshuaBPMan>Did you have to repartition and reinstall the filesystem again? <joshuaBPMan>I don't think I have ever had a corrupted ext4 filesystem that fsck could not fix. <roptat>it did fix it but some files were not correctly restored <roptat>anyway, now the system is installed on an external disk, so no problem <joshuaBPMan>roptat: ahhh. gotcha. I've never had the misfortune of having that happen to me. <joshuaBPMan>aka can I autostart emacs via the shepherd? like how systemd has user services in the user's home directory? <katco>is something going on with guix infra? i'm getting intermittent failures when installing a package, and they continue if i just run the command again. <apteryx>joshuaBPMan: yes you can, but it's managed outside of Guix <joshuaBPMan>apteryx: So I would have to run two instances of the shepherd? One running as me when I log in perhaps? <apteryx>joshuaBPMan: yep, I run shepherd in my .xsession <civodul>katco: what failures are you getting exactly? <katco>"|builder for `$THIS_COMPONENT_KEEPS_CHANGING_AS_I_CONTINUE' failed with exit code 1" <civodul>followed by "View build log at ...", right? <katco>oh! ok there is useful information in there. that's weird... i've always experienced the logs to be the same as what's printed on stdout <katco>in this case, i'm trying to build arm packages on x86. i'm a little fuzzy on whether that's possible or not. i thought guix could do that <katco>i'm also confused as to why the package that fails keeps changing <roptat>it's probably choosing to start with a different package <roptat>and it's possible, but you need the qemu-binfmt thing <katco>hm, build order is non-deterministic? <katco>roptat: is there somewhere i can read up on that? here's the command i'm issuing: guix system disk-image --system=armhf-linux -e '((@ (gnu system install) os-with-u-boot) (@ (gnu system install) installation-os) "beaglebone-black")' <roptat>or maybe it actually is, but it's going to have a different order if it managed to build one thing <roptat>(like downloading a source code) <civodul>katco: for that to work, one of the following conditions must hold: you're on armhf-linux, or you've set up offloading to an armhf-linux box, or you've set up the qemu-binfmt service <roptat>(look for Transparent Emulation with QEMU) <mikadoZero>I am using mu4e in Emacs on Guix System. I am not able to send email and when I try it says "Process smtpmail not running". This started yesterday. I have not made any configuration changes to Emacs or mu4e. I have tried the following to fix this. I have rebooted the machine. I have also pulled, system reconfigured and run package on my manifest. Is anyone else using Guix System having this problem? Any sugg <katco>civodul: roptat ty for the info! does guix not build substitutions for arm? <roptat>it does, but it's a bit slow on building them <katco>understandable. so in theory, if i waited long enough, all of these packages would have arm substitutes? <katco>i wish there was a way to allow guix users to contribute substitutes without that being a complete security nightmare XD <katco>something something blockchain <roptat>any user can run a substitute server <roptat>it's up to you to authorize them <katco>i mean contribute substitutions upstream, to a substitution server run by guix, that most users probably already have authorized <roptat>that's a complete security nightmare <katco>but i can't think of any way to make that secure outside of distributed confirmation of hashes, and even that is vulnerable to attackers having a majority vote <atw>guix challenge would provide a certain level of confidence if there were several people serving substitutes. I would like to put my money where my mouth is on that <katco>yes, which is why i qualified it as such <katco>atw: in theory, attackers could still outvote the legitimate substitutes <roptat>you could always fake these reports <joshuaBPMan>does guix support loading a bare-bares.scm with the hurd kernel? That would be pretty rad. <roptat>and fake them coming from many people <atw>have there been well-known cases of binary distros having their binary-serving servers compromised? <katco>atw: of course, fairly regularly <atw>my thinking is, that's the "prior art" that we could learn from. I can't remember hearing this about e.g. Debian, but maybe a smaller distro like Mint? <katco>"This confirms that GPG keys used for signing packages were compromised. These keys are used for signing the synapse debian repository (AD0592FE47F0DF61), and releases of Riot/Web (E019645248E8F4A1). Both keys have now been revoked. The window of compromise for the keys started from April 4th; there have been no Synapse releases since then." <efraim>KDE neon? They had an incident with bad ftp settings <atw>hm, lots of pitfalls to avoid <katco>atw: i don't know of any distro that allows users to contribute upstream builds of packages for the reasons discussed. i don't know how you'd secure such a system. binaries are always done on trusted infra and signed <atw>"we trust this user enough to distribute their binaries as official" is a very high level of trust <cbaines>katco, you can fetch a substitute from someone, if it's signed by someone you trust, even if those "someone"s are different <katco>yes, i don't think that would be a wise approach. like i said, the only thing i can think of is to only trust binaries whose signatures a large amount of people agree with, but that's also easily attackable <cbaines>which means that you could have a trusted build farm building packages, but then fetch the packages from a build farm you don't trust, if the content exactly matches <cbaines>unfortunately, a setup like this doesn't help with lacking substitutes to begin with <civodul>meiyopeng: congrats on the initial translation of the manual! <civodul>i started adding it to the makefiles and all, but currently the generated Info file fails to build <civodul>maybe because some of the node names were not translated? <roptat>meiyopeng, make sure that you don't translate the content of @pxref, @xref and @ref when they refer to internal nodes: they are translated automatically once you translate the node name in the .po file <roptat>(so you can fix a node name once and don't have to care about fixing references) <roptat>civodul, i think you're missing something in po/manual <civodul>roptat: i have to leave but if you want to play with it, please go ahead! <mitescugd>is there a way to make guix-installed programs use the system-wide dbus? was thinking that launching the WM (installed via guix) with dbus-launch should do the trick (given there's no guix dbus-daemon running at the point) <ATuin>how can i get readline in guile using guix environment? <ATuin>i found it :) seems that i was always using the guile from outside the environment <ATuin>mmm i get "gnu/packages/cpp.scm:76:18: error: url: unbound variable" when using "(use-modules (gnu packages compression))". What am i missing in my environment? <lfam>cbaines, ATuin: The 'git-version' procedure is used in the wrong place <ATuin>the line calls git-version, yes <ATuin>(version (git-version "0.0.0" revision commit)) <ATuin>maybe i can pull it and retry <lfam>The code needs to be fixed <lfam>The 'git-version' call needs to replace line 79, which is (version version) <ATuin>mmm i'm very newbie to fix it, thanks <ATuin>but i would like the fix so maybe i can learn something <lfam>Actually that's not the core issue, but it is unidiomatic <lfam>Not sure exactly why it doesn't work but I'll fix it now <ATuin>maybe my env is wrong, when i try make i get an error about unbind zip module from compression.scm, that's why i was trying to import it <ATuin>seems that some of the packages defined there are available in the repl but others not, dunno if the error is related though <lfam>ATuin: I just pushed a fix for the 'url: unbound variable' error. It should be available for you after `guix pull` <lfam>I have to go AFK now, good luck! <ATuin>seems that it worked, now the same error happened in web.scm :) <ATuin>ahh no, it's datastructures.scm sorry <cbaines>ATuin, do you know what package the error relates to? <ATuin>ahh now it looks better, it throws an error in the package code i have added, i guess that's better <ATuin>but i have deleted the .go files and now it complains about the package i have added to emacs for learning purposes <ATuin>i will delete my code and see if now i can import the module <ATuin>yep, deleting the .go files in my $HOME fixed that problem <ATuin>now i get a nicer one: ice-9/eval.scm:159:9: Throw to key `srfi-34' with args `(#<condition &message [message: "icecat-makeicecat.patch: patch not found"] 7448000>)'. <ATuin>i guess most of the problems are related to me not knowing how guile / guix works really <ATuin>as i understand the srfi-34 is some kind of exceptions implementation, so the real problem is that it's missing a file (input?) <ATuin>mmm the file is there (under gnu/packages/patches) <ATuin>there is a %D% used as prefix, i guess that's the problem <ATuin>cbaines: the import worked now, thanks <raingloom>hi, what does "no code for module (mininet)" mean? I can get `guix build` to report syntax errors and such so I think I'm setting GUIX_PACKAGE_PATH correctly, but it seems like it doesn't actually build the package. <nixo_>hello guix, is it possible to print the current build output? <nixo_>I'm upgrading julia to 1.1.0 and it's going well, but it reaches a point where the cpu usage drops to 0 so I don't know what's happening <nixo_>while the output just says 'build phase' and it's pretty useless <nixo_>well, I'll just stop it and continue manually <pkill9>wednesday: well, selinux was created by the NSA and that's in Guix :P <nixo_>wednesday: maybe just checking the code for unwanted "telemetry"