IRC channel logs


back to list of logs

<mbakke>nckx: dunno, but the number of builds is misleading since only x86_64 is affected, where the CI has good coverage
<mbakke>oh, I see you already reverted it
<nckx>I'd rather be safe than sorry.
<nckx>(Well, after being convinced that it was safe, anyway…)
<nckx>Thanks for responding though 🙂
<nckx>mbakke: I'll throw it onto c-u.
<mbakke>nckx: why not staging?
<nckx>Simply because I don't understand what you mean by [2000+] being misleading.
<mbakke>the 2k packages are for all architectures, so ~500 in reality
<mbakke>guix refresh -l rust@1.39 says 1343 rebuilds
<nckx>I read your comment the other way 'round: all builds would be for x86_64.
<nckx>Hence not divided by ~4.
<mbakke>oh, sorry :P
<nckx>Like, why would 500 armhf builds be scheduled if it's x86_64-only?
<nckx>Anyway, staging is good, staging is fine, off to staging we go.
<mbakke>Setting supported-systems will possibly fix that.
<nckx>‘Having this sanity check is pretty much a prerequisite for
<nckx>unattended upgrades IMO (patch coming soon!).’
<nckx>civodul: (╯°Д°)╯︵ ♥
<civodul>nckx: yeah! :-)
<civodul>i thought i was going to complete today but i thought it'd rather procrastinate a bit
<civodul>summer time!
<civodul>but anyway, it's mostly: guix time-machine -- reconfigure /run/current-system/configuration.scm
<nckx>We all deserve this summer.
<nckx>civodul: By the way, did you receive my key mail?
<nckx>No need to reply to it now, just wondering.
<civodul>nckx: ah yes, i did, but i don't know what to do :-)
<civodul>looks like an epg/gnus issue
<nckx>Nor do I, I'm afraid.
<nckx>That rather puts us in a bind. It's just confusing since I'm sure I've decrypted mail from you before.
<nckx>civodul: And now you get stuck in moderation and I can't approve you! Ha ha.
<civodul>nckx: how's that?
<civodul>ah yes, because you didn't get the thing
***catonano_ is now known as catonano
***sneek_ is now known as sneek
<klys>is guix really supposed to compile everything it downloads? like gentoo?
<jackhill>klys: Conceptusally, Guix is source-based like Gentoo. However, there is support for using pre-build binaries as a substitute for building locally. You have to authorize the build servers though:
<jackhill>klys: I assume you'd like to use the pre-build binaries?
<jackhill>sometimes though, the build farm hasn't build something, so there is no substitute and you'll have to build locally
<klys>I have authorized it as per those instructions, with this command: sudo guix archive --authorize < /usr/share/guix/
<klys>sudo guix pull; tells me there is "nothing to be done."
<klys>in order to get icecat, guix still wants to compile 19 versions of the rust compiler, though.
<klys>I'm using a 32-bit system and have been told this will fail.
<jackhill>klys: hmm, I wonder if guix-daemon isn't running with --subtitute-urls
<klys>you're right
<jackhill>ah, great, so they're authorized, but the daemon isn't asking for them :)
<jackhill>as a one off, you could try `guix build substitute-urls="" icecat
<klys>in "/lib/systemd/system/guix-daemon.service" I found: ExecStart=/usr/bin/guix-daemon --build-users-group=_guixbuild
<klys>I suppose I will have to add it there.
<jackhill>yeah. Or maybe copy it to /etc/systemd/system/ and modify it there. It's been a little while since I worked with systemd.
<jackhill>how'd you install Guix by the way?
<klys>from the deb I built from the salsa repo
<jackhill>oh, cool :) It remember when there wasn't a deb, so it's nice to see that coming along. I know our binary installation script adds that config, so I was curious.
<klys>Jul 15 20:32:04 localhost guix-daemon[27533]: /usr/bin/guix-daemon: unrecognized option '--subtitute-urls'
<jackhill>typo: subtitute -> substitute
<klys>o i c
<jackhill>I see I typed it wrong above. oops…
<klys>Jul 15 20:33:38 localhost guix-daemon[27875]: /usr/bin/guix-daemon: option '--substitute-urls' requires an argument
<klys>looks like it's running now
<klys>except it still tries to build rust when asked to install icecat
<jackhill>did you systemctl daemon-reload?
<klys>yes and
<klys>root 28178 0.0 0.1 11904 4152 ? Ss 20:35 0:00 /usr/bin/guix-daemon --build-users-group=_guixbuild --substitute-urls=
<klys>sudo guix pull once more says "nothing to be done"
<jackhill>I'm out of ideas, sorry!
<Kimapr[m]>is it possible to have multiple bootloaders for one system def?
<NieDzejkob>substitutes are just not available for icecat on i686 because the bootstrap path fails on that architecture
<klys>oh, so no way to get icecat?
<efraim>Not on i686
***sputny1 is now known as sputny
<rekado_>hmm, I can’t find a way to remove package-closure in this commit:
<rekado_>I was going to use package-mapping, but then realized that this wouldn’t really help.
<rekado_>because I don’t want to rebuild all of the packages in the graph; I just want to collect all of the Haskell packages in the graph and add the static variants as inputs to *this* package
<rekado_>so package-closure seems to be just the right thing
***rekado_ is now known as rekado
<civodul>rekado: uh, not sure what to do
<civodul>i'm concerned about the cost of recomputing the closure every time we look at the inputs of that package
<civodul>i have to go tho (really, not just to escape the issue, i swear :-))
<hji>could someone point me to any docs on what constitutes free software in the context of guix packaging? esp. with regards to minimized JS for self-hosted documentation etc that fedora and debian has a policy of removing, whereas guix seemingly doesn't (at least not for the few packages that i've checked)
<mmalter>Hello everyone. Quick question, might be silly, I am a newbie. Can you make profiles depend on eachother? Include a manifest in another one? Does it make sense doing so?
<Zakaria>Is there guix package for VirtualBox?
<mmalter>Zakaria: I don't think this is free software
<Zakaria>It is
<apteryx>NieDzejkob: I opened an issue about the cmath problem
<Zakaria>The VirtualBox base package binaries are released under the terms of the GPL version 2.
<mmalter>Zakaria: Oh ok. I can't find anything on
<apteryx>and collected some data, although I haven't pinpointed the culprit yet
<Zakaria>That's why I ask
<Zakaria>Is there a page or some page where I could ask for the dev to package it?
<Zakaria>It's the only thing that hold me to install guix on my laptop
<apteryx>there's a page on libreplanet IIRC
<apteryx>thanks, bavier1 :-)
<nckx>Can VirtualBox be built with free software yet? It didn't used to be.
<bavier1>maybe not. and it's not available in debian 9 or 10.
<nckx>Yeah, I don't think it's free yet, sorry.
<bavier1>so I'd imagine a guix package would be difficult
<nckx>Zakaria: ☝
<bavier1>Zakaria: could you work with virt-manager instead?
<mmalter>I tried it for that exact reason. virt-manager actually felt easier to use.
<nckx> Upstream trying to derail the discussion by playing word games with the term ‘free’ does not bode well.
<Zakaria>Nope, I need to boot my old windows VM once in awhile
<mmalter>I think kvm handles Windows pretty well
<nckx>is the most up-to-date explicit statement I found so far.
<bavier1>nckx: thanks
<nckx>That explains the careful wording used in the bug report above. ‘You don't need proprietary software to build VirtualBox because you can use pre-built blobs’. Clever but not really.
<roptat>hi guix!
<bavier1>"binary blobs" = "alternative sources"? wtf?
<mmalter>it's funny indeed…
<brown121407>Hi all! Is it normal for the stackage and importers to be really slow? guix import pypi and hackage work just fine.
<leoprikler>reminds me of people saying "java bytecode is practically source code" :)
<Zakaria>mmalter: how well can kvm run unmodified Windows XP VM?
<mmalter>It was a long time ago but I actually think it's XP I tried
<mmalter>Can't tell you for sure. I would be surprised if it's hard.
<Zakaria>I virtualized that VM from my previous PC, and I've some software in it that I'm unwilling to lose
<Zakaria>Some rare and hard to get SCADA stuff
<mmalter>I think everybody understands some people go though the hurdles of using proprietary software.
<mmalter>I don't know I did a quick search and XP on QEMU seems really straightforward
<Zakaria>I've read that bug, I think somebody need to recompile their BIOS source code with another "old" 16bit C compiler cause they don't have time to do it
<Zakaria>Okay, I research whether KVM is good enough for my need, thank you mmalter for the solution
<nckx>bavier1: <binary blobs> Well, their argument seems to be that you can assemble the blobs at home from the totally-not-a-blob assembly code (which can only be generated by said problematic compiler), so they're not *binary* blobs! …OK?
<nckx>BRB xxd < mswindows.iso > source-code.hex
<mmalter>Well I guess that argument could be made for rollercaster tycoon. It's open source assembler.
<nckx>mmalter: Not if that's the ‘preferred form of modification’ and what upstream actually edits and improves. Nothing wrong with free assembly (like The VBox ‘BIOS’ is written in C.
<mmalter>yeah yeah sure that was more of a bad joke from me
<nckx>And stage0 was a deliberately extreme example 😉 It's basically commented hex.
<mmalter>Yet I am curious what would be the legal standpoint on this.
<nckx>‘Practical software freedom’ is as important as ever.
<nckx>Good question. IANAL.
<nckx>Probably that its fine, like a GPL software programme written only for Windows would still be able to incorporate other GPL components. It doesn't violate the licence. ‘Freedom’ isn't a legal concept.
<mmalter>In that case it's all well. I was thinking about what happens if there is a breach that could be contested on the basis of wether the assembly code *is* the main implementation. It is left to the court to decide where you draw the line between gibberish and usable source code. Hypotheticals…
<nckx>$partner calls all programming languages gibberish. Humans wrote megabytes of copyrightable source code by flipping front-panel 0/1 switches 😊 It's, like, all relative, man.
<bdju>nckx: did my first reboot since making the luks config change and it worked fine, asks me for the password before login even with it not being the root drive, then mounts fine.
<bdju>nckx: also I noticed several errors while booting up including that old cpm or whatever one you had noted as a difference, so I guess it still does appear. but I was able to get to the login fine. (I guess we did determine the mounting was the issue back then)
<bdju>also in case anyone has weirdness with their sway screenshot bind breaking, a reboot just fixed that for me
<nckx>Yep. The CP437 error would only be suspicious if you were building a custom kernel, since it's built into the Guix ones. (No idea what that error message is trying to say, then, but it doesn't seem to matter.)
<nckx>Very happy to hear all that.
<nckx>(Apart from the ‘reboot fixed weard breakage’ part, but at least it fixed something.)
<nckx>(Perhaps I use to many parentheses.)
<davidl>Can I use ;;guix pin <commit> with guix-jupyter to pin additional channels somehow?
***Exagone314 is now known as Exagone313
<bdju>I've got missing button icons in the new keepass. anyone else experiencing that?
<bdju>keepassxc I mean
<bdju>ah jeez. even setting it to classic theme doesn't fix it
<NieDzejkob>I recall this topic came up in the channel before. I tried searching the logs for "keepass icon", but it's showing results with either of the two words, yet not both
<roptat>same here, no icon
<roptat>I found bug reports upstream, and tried solutions listed there, but it didn't work
<roptat>apparently the issue happens on other distros too, but for an unrelated reason: they apparently use qt5ct which we don't have, and it messes with icon search path, so the solution is to ensure the application name is reset before searching for the icon
<roptat>I applied the patch for the future version of keepassxc, and I tried to install qt5ct too, setting the right environment variable for it to be selected, but that didn't work at all
<roptat>so I don't know what's going on exactly
<bdju>yikes. sounds like a mess. glad to know it's not just me, at least.
<roptat>yeah but I don't know how to fix it: the upstream fix doesn't work
<lfam>Fedora has chosen to default to btrfs on desktop systems:
<lfam>So hopefully a new wave of bugs should get squashed in the coming year or two
<jackhill>lfam: cool. I've been pretty happy with it, although I haven't tried anything fancy :)
<lfam>I've been happy with it for a few years too
<lfam>Doing huge COW copies is really handy
<roptat>bdju, I don't even get anything from running it with strace... I'm rebuilding with more debugging, and I think I can "hide" the issue by having keepass fallback to a hard-coded directory in the store
<roptat>so currently it does "icon = QIcon::fromTheme(name);" but fromTheme can take an optional QIcon argument that it falls back to if it doesn't find name in the current theme
<roptat>well it won't help actually, when the icon is not provided directly by keepassxc
<user_oreloznog>Hello Guix !
***roptat_ is now known as roptat
***victor is now known as Guest14769
<Guest14769>Hi, this may not be the place to ask. But I'm trying to run a windows image that contains malware using virtual manager. How do I configure it so that it has not connection to the internet or my host machine. It's just to prevent the malware from maybe escaping.
<user_oreloznog>Guest14769: maybe ask to #virt
<roptat>Guest14769, when the machine is stopped, you can open it (from the list of machines, double click on it, but don't start it yet). The window that opens for the machine should have a "i" icon in the toolbar that lets you configure the hardware options, including network
<roptat>bdju, ok, if I hardcode my theme name, I get a few icons, so qt seems to be unable to find the default theme name and immediately falls back to hicolor, where it finds almost no icon
<roptat>(I think it only finds the app's icon)
<roptat>apparently qt cannot find the default theme name outside of kde and gnome
<roptat>ha! it has an "application" theme, but it's not currently installed with our package
<roptat>re-building, this time with an additional phase to install these icons to share/keepassxc/icons/application (hopefully, that should be ":/icons" from qt's perspective)
<Guest14769>thank you user_oreloznog, I didn't know virt-manager had an IRC
<Guest14769>@roptat I tried turning off networking that way, but the VM won't come up when I do that.
<bdju>roptat: sounds like you are maybe making some progress. very cool
<roptat>mh... I thought I got it, but no
<nckx>mbakke: Congrats on Ganeti!
<roptat>another possibility is to hardcode the path to the $out/share/keepassxc/icons
<roptat>or it's missin an icon-theme.cache
<mbakke>nckx: ty! now I just need a couple of dozens servers to deploy it on ^^
<mbakke>also, this means I have to get back to my regular package reviews and updates :P
*nckx hopes mbakke needs npm for work soon.
<roptat>oh, QIcon::fromTheme doesn't look into the scalable directory, where all icons are located!
<roptat>setting QT_PLUGIN_PATH to include qtbase and qtsvg works!
<mbakke>nckx: lol, I hope not! :P
<mbakke>roptat: would qt-build-system help in this case?
<nckx>mbakke: Sorry. Right. Worse enemy & all that.
<roptat>no, there's no qmake file
<bonz060>Are there CI tools that support guix?
<bonz060>Things similar to travis or the like?
<bonz060>I'm trying to figure out how to have a GI set up for a project that has all its deps on guix on some custom channel
<apteryx>I reckon a Jenkins plugin for Guix would probably be a relatively low hanging fruit.
<nckx>Someone mentioned Travis+Guix on a mailing list very recently (this week) but I can't find it now.
<nckx>A Guix mailing list.
<apteryx>It seems to me that adding Guix support to popular CIs would be a good way to spread Guix's adoption far and wide
<nckx>Maybe reopen
<nckx>The bot was meant to close solved issues so it should be OK.
<apteryx>I just commented on it
<apteryx>Travis is free as in free beer?
<apteryx>or are the sources somewhere
<mbakke>afaik travis is only available as a service, so I'm not sure if even the beer analogy holds