IRC channel logs


back to list of logs

***ChanServ sets mode: -b *!*
***ece3 is now known as ece
<lf94>weechat protected me
<lf94>guix is not available on debian buster :'(
<lf94>oh well, I guess my puny laptop with 128gib storage wouldn't be great for it
<lf94>(and 2 cores)
<milkey>wdym it's not available? the only real dependencies are a linux kernel and the bootstrap binaries which should run on debian
<wonko_the_sane>incidentally, in all of #guix circa 8 year logged history, this is apparently the first string containing: 'λ'
<lf94>milkey: I mean not offered by the distro repo
<lf94>milkey: there is some setup to guix if it is not packaged for your distro.
<lf94>also: why is only 2 versions of nodejs packaged? :(
<lf94>looks like I'll package 12.x because I need it for work
<milkey>try guix build node --with-version=12.x, it might Just Work
<milkey>although nodejs is pretty big and required some manual tweaks before so maybe not
<lf94>ooou, will do
<milkey>automatic package transformations are so neat
<lf94>→ with-version not an option
<milkey>er, try --with-commit=v12.x
<milkey>with the git tag for the release you want
<milkey>hmm actually nevermidn that wont work
<lf94>I can maybe inherit from 10
<lf94>When making 12
<lf94>it's all good
<lf94>I need to learn how to package!
<leoprikler>which would mean you'd skip 11?
<leoprikler>that sounds like you're asking for trouble
<lf94>I could try 11
<milkey>why? I don't think each version of node bootstraps from the previous
<milkey>it's c++
<lf94>I want to skip to 12 because it's the next LTS from 10.
<lf94>I only care to use LTS versions
<milkey>this seems to work (don't have time to see if it builds successfully though): guix build -e '((@ (guix packages) package) (inherit (@ (gnu packages node) node)) (version "12.20.2"))'
<awb99>Is it possible to change a configure flag for package. I think 'r-minimal' package disables --enable-R-shlib configure flag.
<milkey>awb99: in your package definition: (arguments '(#:configure-flags `("--something" "--something-else"))) should work for most build-systems
<milkey>(I can only answer so fast because I literally just figured that out for the file I had open, lol)
<leoprikler>because I'm sure, that as fast-changing as the node ecosystem is, the tool itself vastly changes when you jump two major versions ahead
<leoprikler>interestingly marks node as failed, when the build seems to have succeeded 🤔️
<awb99>Thanks @milkey
<milkey>no problem
<leoprikler>If recursively importing a Rust package pulls in 300 dependencies, does it belong to staging? 🤔️
<milkey>rust just be like that sometimes
<leoprikler>Especially since `guix refresh -l <rust-package>` seems broken.
<jgart[m]><lf94 "guix is not available on debian "> It's available in sid
<milkey>honestly it's not worth installing a distro's guix package vs. using the tarball from IMO. I tried packaged versions of guix from both gentoo and ubuntu and they each had issues that would not have come up had I just followed the handbook
<jgart[m]><Guest96 "jgart ok, maybe I got it: guix m"> A manifest works exactly like it sounds. It is a list of packages. Instead of running `guix install foo`, `guix install bar@2.0`, etc... you can put all the packages in a manifest and install them to a profile. See this blog post: which refers back to the man pages at the bottom.
<jgart[m]>I haven't seen manifests used for anything else but development libraries and end-user packages yet.
<lf94>jgart[m]: I said buster!
***jpoiret1 is now known as jpoiret
***MidAutumnHotaru4 is now known as MidAutumnHotaru
<erhandsome>seems like guix homepage returns 404 when Accept-Language is zh-CN
<erhandsome>curl '' -H 'Accept-Language: zh-CN'
<sturm>apteryx: I saw you mention in the logs that you're using linphone - does caller ID show the SIP proxy's hostname for you? My always shows the SIP proxy IP, so doesn't match up to my contacts that use the SIP proxy hostname.
<awb999>Isthere a way to explore a guile object?
<awb999>I have modiied a package definition and would like to check if I did evrything ok
<awb999>but since I modify itm
<awb999>I am not exactly suee it is doing the right thing
<milkey>you can inspect it in guix repl for one
<awb999>how do that?
<awb999>(display mypackage)
<awb999>doesnt do the trick
<milkey>mmm, I think you'll have to use stuff like (package-version mypackage) and (package-arguments mypackage) to access the fields on the record
<milkey>once it's loaded in the repl
<terpri>there's also the ",inspect EXP" repl command
<terpri>though i don't know how useful it is with packages etc.
<milkey>that does sound more along the lines of what awb999 is looking for
<terpri>yeah, i'd expect it to list struct fields and so on
<terpri>geiser-mode for emacs might have some useful tools too
<awb999>thanks guys!
<awb999>I am still failing to load my namespace to the guix repl
<awb999>seems to be different setup
<sturm>is anyone using linphone with a SIP proxy account to call regular phone numbers?
<mroh>Ben Sturmfels: yes, I use linphone that way. Wie kann ich Dir helfen?
<sturm>mroh: great, I'm having this weird issue where incoming SIP calls show the IP address of my sip proxy, rather than the hostname so don't match up to my contact list entries - can't see who's calling me!
<sturm>are you seeing anything like that?
<sturm>hang on, it's not the sip proxy's ip, it's actually my IP
<sturm>(was almost identical)
<milkey>the call is coming from inside the house!
<dftxbs3e>hello! :-D
<mroh>Ben Sturmfels: I don't understand: you have IP addresses and/or hostnames in your contact list? The proxy should provide the calling number, no?
<sturm>mroh: when I dial an outgoing number it, completes it with the SIP proxy hostname, eg. I type "0123456789" and it dials My outgoing calls, they show up like "0123456789@", so don't match the contacts I have with the SIP proxy hostname
<sturm>I'm wondering whether this might be a peculiarity of my service provider, or whether it's a linphone issue.
<mroh>I don't know, sorry.
<sturm>mroh: do contact names show up on incoming calls for you?
<jackhill>sturm: I see number@ip-addr for incoming calls. I don't have contacts configured, so don't know about that, but based on what you said I assume it wouldn't work.
<jackhill>I haven't used non-guix linphone, so I can't say if its a problem with our package or a general linphone problem. My provider is, so that's different than yours.
<sturm>jackhill: thanks and you have a SIP proxy configured by hostname?
<jackhill>sturm: yep
<sturm>good to know, thanks!
<jackhill>you're welcome
<pocketroid>Greetings programs
<jgart[m]><awb999 "how do that?"> there's functions in guix/records.scm that you can use to inspect a package. A package is a scheme record that can have other records inside of it (e.g. the source field takes another scheme record called origin)
<bdju>omemo not working for me on fresh install of gajim & gajim-omemo. not showing in plugin list
<jgart[m]>raghavgururajan: raghav-panic ☝️
<awb99>Thanks @jgart[m]
<milkey>what is the right way to import an additional module for use during a build phase?
<efraim>what do we have for matrix clients on Guix? (that isn't emacs)
<milkey>it's in a G-expression so it's not as easy as including it at the top of the file but I don't understand the internals well enough to know what that entails
<milkey>I can't get it to not give a "no code for module (guix config)" error when I build the package
<iyzsong-w>milkey: I think the way is to use 'with-imported-modules' wrap the gexp, and 'use-modules' in the gexp too.
<milkey>I read about with-imported-modules, but I didn't know where to put it. I am just defining a package so I'm not manually creating a gexp
<iyzsong-w>okay, then I think that's the '#:imported-modules' argument for the build system, see 'gnu-build' in guix/build-system/gnu.scm
<milkey>i tried that too but I'm still getting the same error. I'll try again and see if I can make it work...
<milkey>when I used substitute-keyword-arguments in another file I was able to do (arguments (substitute-keyword-arguments (package-arguments alacritty) ;((#:imported-modules m %cargo-build-system-modules) `(,@m (guix build rpath)))
<raghavgururajan>Hello Guix!
<milkey>so this is weird. (arguments `(#:imported-modules (,@%cmake-build-system-modules))) works, but (arguments `(#:imported-modules (,@%cmake-build-system-modules (guix utils)))) doesn't. how could adding a module cause something to stop working...
<milkey>hello raghavgururajan
<raghavgururajan>pineapples, Oh let me re-check my logs.
*raghavgururajan 's xmpp hosting provider's servers were under DDoS. So logging might have stopped in between.
<raghavgururajan>sneek. later tell cage_: Could you share your pack-def via debian paste-bin with no expiration. I'll see what I can do to make it work. :-)
<sneek>Will do.
<milkey>oh, I think I figured it out: I need to import (source-module-closure `((guix utils))) not just (guix utils)
<milkey>thanks for pointing me in the right direction
<milkey>yup, it works now
<iyzsong-w>oh, cool, make sense. I'm not sure about this area too, learn it..
<raghavgururajan>Any committers available to review and push #46584? :-)
<raghavgururajan>iyzsong-w: I just saw your email regarding gandi-cli. Thanks so much.
<iyzsong-w>raghavgururajan: yeah, glad to help!
<raghavgururajan>iyzsong-w: Would you be able to review #46584? Only if you are available though. I don't wanna pressure you. ;-)
*raghavgururajan goes back to fighting telegram
<irstlovef>anki is broken on my guix ImportError: cannot open shared object file: No such file or directory :(
<irstlovef>"LD_LIBRARY_PATH=/gnu/store/qj38f3vi4q1d7z30hkpaxyajv49rwamb-gcc-10.2.0-lib/lib/ anki" still lanuch anki and error change to: ImportError: cannot import name 'langDir' from 'anki.lang' (/home/firstlove/.local/lib/python3.8/site-packages/anki/
<irstlovef>Oops, maybe I should not install anki from pip, will try pip uninstall anki
<milkey>you nerd sniped me. I started building anki just to see what was wrong
<irstlovef>after pip uninstall anki: /gnu/store/rz42ba0my9vrgbkjpkzr2drmnjk5ah50-python-3.8.2/bin/python3: symbol lookup error: /gnu/store/24a08v4pi76aahfirdh40b67x8rjnc7x-qtbase-5.14.2/lib/ undefined symbol: _ZN22QWindowSystemInterface24setPlatformFiltersEventsEb, version Qt_5_PRIVATE_API
<irstlovef>then guix remove qtbase, anki lunch successfully finally.
<iyzsong-w>so with a qtbase in profile will broken incompabitable qt based apps (I think your anki may use an old Qt than the one in profile)
<irstlovef>Yes, iirc anki in guix repo uses qt5.14.
***apteryx is now known as Guest68821
***apteryx_ is now known as apteryx
<leoprikler>I'm not sure what exactly Telegram does, but it's a huge thermal sidechannel.
<leoprikler>milkey: do you really want guix utils though or just guix build utils?
<nckx>wonko_the_sane: You need better history. <>, and that's only the past half year. (Yes, I'm a disproportionate cause of lambdas here.)
***pocketroid_ is now known as pocketroid
<rekado_>(ugh, the results are not sorted properly)
<jlicht>hey guix!
<nckx>Hullo jlicht.
<jlicht>lf94: RE: lack of node versions: because packaging node _from sources_ is not really fun, after node version 10
<jlicht>I do expect to have a up-to-date node v14 (also lts) available in the coming weeks, if that helps you
<dftxbs3e>nckx, what is the problem exactly?
<nckx>What's not clear?
<dftxbs3e>nckx, you found a SHA1 signature made with my key somewhere?
<nckx>No, your key itself (‘self-signature’) is signed with SHA1.
<dftxbs3e>nckx, I never gave you my key
<dftxbs3e>I only gave you my fingerprint
<nckx>I'm all for bummin' subkeys but I do assume that's supposed to be a usable key file.
<dftxbs3e>nckx, how do I remake these you think?
<dftxbs3e>My SHA1 ban configuration is new, that yes
<dftxbs3e>I think SHA1 was still considered safe when I created my key
<nckx>Mine isn't and it's the first time it's caused me problems.
*dftxbs3e looks it up how to create new self-signatures
*nckx AFK but will read backlog. o/ pineapples.
<dftxbs3e>nckx, can't find how to remove signatures..
<dftxbs3e>nckx, check now?
<nckx>dftxbs3e: Perfect.
<dftxbs3e>nckx, did it go well? :-D
<nckx>Thanks. For posterity's sake: what did you do?
<raghavgururajan>pineapples: I am cooking something new for telegram
<dftxbs3e>nckx, I deleted my current uid and created a new one with "Léo Le Bouter" instead of "Leo Le Bouter", that also re-created the self-signature using my current gpg settings which disallow SHA1
<dftxbs3e>That is something I wanted to do since a while
<dftxbs3e>Also Savannah went all green on me, what's up!
<pineapples>raghavgururajan: :O
<nckx>dftxbs3e: It changes to your preferred theme when you log in (mine's blue); perhaps that?
<dftxbs3e>nckx, well maybe but I never changed that theme!
<dftxbs3e>nckx, okay! well changed the theme back to default!
<nckx>dftxbs3e: Relieved to hear it was that easy (the re-sig). I've added your key to the ‘keyring’ branch and... hm... /me thinks hard whether anything else is required.
<nckx>Ah, .authorizations.
<dftxbs3e>nckx, Thanks a lot :-D
<nckx>dftxbs3e: Congratulations! Could you send a mail to guix-devel ‘introducing’ yourself, including the key fingerprint you'll be using? Also read <> and all relevant links, but I'm sure you already have ☺
<nckx>Apologies for the length of the process.
<nckx>In particular, be sure to install the pre-push git hook from etc/git locally so you don't accidentally push unsigned commits.
<nckx>I think the server still can't check for those and it would be verybad.
<nckx>If it can, well, nothing wrong with belt + braces.
*nckx returns to the peaceful land of AFK. o/
<dftxbs3e>nckx, Yay! Will do! Thanks a lot!
<dftxbs3e>funny how if you type "commit access" on search engines the second result is GNU Guix manual aha
<jlicht>we pride ourselves on our bespoke, kafkaesque procedures ;-)
<dftxbs3e>jlicht, I think they may hurt inclusivity and contributor empowerment some times
<rekado_>there’s no simple way to sign commits
<rekado_>the setup is always going to be a little confusing :-/
<dftxbs3e>rekado_, I don't think it's the signing the issue, but that it's easy to feel held back in general when you're new
<dftxbs3e>Basically I think it's lack of time of maintainers, maybe we need someone who is always here and whose role is not contributing even just being here and ensuring everybody feels welcome when they contribute, similar to Qubes OS community manager guy:
<nckx>dftxbs3e: Are you talking specifically about commit access? I think that's proportionate; it's a one-time thing, and the only thing that took time was waiting for everyone to approve (we can't well drop that ☺).
<dftxbs3e>nckx, Nope more in general, I think it's really not easy emotionally to contribute if you don't have commit access
<jlicht>I was specifically referring to our "please install this git hook, or else BOOM" many-points-of-failure model
<nckx>But in general, yes, we can be a bit bureaucratic, but I thinks that's a result of being under-staffed.
<leoprikler>eww, democratic decisions
<leoprikler>don't you know that businesses function better with one dictator at the top? </s>
<nckx>I get your point, but ‘we should appoint someone to spend full-time on x’ when the problem is ‘nobody voluntary wants to full-time x’ is a suboptimal solution.
<dftxbs3e>nckx, not voluntary :-)
<dftxbs3e>Maybe it would be useful to spend budget of GNU Guix on someone there
<nckx>‘Ludo’'s actually 5 orphans locked in a box forced to send friendly e-mails to everyone, I don't know what more we can do.
<dftxbs3e>It felt many times like people were burned out
<boombim>Hey guys. I have installed `font-adobe-source-code-pro` font but i can't get italic font from it
<boombim>fc-match -s "Source Code Pro:style=Italic"
<boombim>SourceCodePro-Regular.otf: "Source Code Pro" "Regular"
<boombim>DejaVuSans.ttf: "DejaVu Sans" "Book"
<boombim>DejaVuSans-Bold.ttf: "DejaVu Sans" "Bold"
<boombim>DejaVuSans-Oblique.ttf: "DejaVu Sans" "Oblique"
<boombim>DejaVuSans-BoldOblique.ttf: "DejaVu Sans" "Bold Oblique"
<boombim>Arial_Regular.ttf: "Arial" "Обычный"
<nckx>dftxbs3e: You're not wrong.
<boombim>DejaVuSerif-Italic.ttf: "DejaVu Serif" "Italic"
<boombim>DejaVuSansMono.ttf: "DejaVu Sans Mono" "Book"
<boombim>DejaVuSerif.ttf: "DejaVu Serif" "Book"
<boombim>DejaVuMathTeXGyre.ttf: "DejaVu Math TeX Gyre" "Regular"
<boombim>DejaVuSansMono-Bold.ttf: "DejaVu Sans Mono" "Bold"
<boombim>DejaVuSerif-Bold.ttf: "DejaVu Serif" "Bold"
<boombim>what can i do to get it?
<leoprikler>boombim: next time, try a paste service please
<dftxbs3e>Also we should definitely use more UTF-8 happy-emotions emojis when we have them :-D
<nckx>boombim: (1) use for multi-line posts (2) what does fc-cache -rv return?
<boombim>ok guys. I'm sorry.
<nckx>ฅ^•ﻌ•^ฅ ~ thank you.
<dftxbs3e>It made me feel really good seeing awb99 coming around that enthusiastic :-D
<Rovanion>Hi, I'm no one to this project except someone who just got started. The biggest hurdle I found was that I had to worry about mailing patches instead of being able to push them to a git repository somewhere for review.
<leoprikler>dftxbs3e: "Nobody wants to do stuff." "Then let's force them to do stuff using money." sounds like the best way to describe capitalism
<boombim>Here is the response
<boombim>BTW I'm using debian inside guix
<nckx>dftxbs3e: I used to use them a lot more than I do now. It annoys some people. *shrug*
<boombim>oh, guix inside the debian. sorry
<nckx>(That was me forcing myself to use them to be welcoming, though.)
<rekado_>Rovanion: you can totally push your commits to a repository and ask people to review that.
<rekado_>it’s just sub-optimal
<nckx>boombim: OK, thanks for mentioning Debian, that complicates matters a bit. Like I'm not sure if ‘/var/cache/fontconfig: not cleaning unwritable cache directory’ could cause problems. (Don't re-run fc-cache as root, though!)
<nckx>Did running it help, by chance?
<dftxbs3e>leoprikler, capitalism's a bit more than that, but since that's what the society gives us to survive either they get it here or they get it elsewhere, maybe we can find someone who's willing to do genuinely and pay them so they don't need to look elsewhere to sustain themselves, and feel less happy doing so)
<Rovanion>rekado_: What makes you feel that? That is how I usually interact with free software projects. The only exceptions would be some one off patch to mesa or linux that I'm not even sure got in.
<rekado_>sub-optimal for us.
<dftxbs3e>I also think people's willingness fluctuate so they may be willing one day and not feel like it another day, that's probable.
*nckx would expect so: ‘/home/tim/.guix-profile/share/fonts/truetype: caching, new cache contents: 176 fonts, 0 dirs’
<boombim>nckx yep! looks like after fc-cache -rv it works correctrly, Thanks!
<rekado_>Rovanion: you *can* do that with Guix as well. But it’s a different workflow, and you ask the people doing the work to accommodate a different workflow, which means that it’s more work for the reviewers
<nckx>sorry: ヽ༼ຈل͜ຈ༽ノ
<rekado_>dftxbs3e: what is the proposal here? I think this channel and help-guix are pretty good in terms of helping people contribute.
<nckx>dftxbs3e: We simply don't have any income, that would not be sustainable.
<leoprikler>nckx is really committed to the Emoji today 😄️
<nckx>🐠 🐡 🐟 🎣
<jlicht>Rovanion: it is an often mentioned hurdle though, but there seem to be no _easy_ 'solutions' to address these concerns
<boombim>I have serveral more questions. I have installes `peek` with guix. And it works if i invoke it from terminal. But it doesn't from gnome app menu. Is it possible to figure out what's wrong?
<leoprikler>btw. how do you type them? is there an ibus for that?
<rekado_>boombim: what does it mean that it “doesn’t [work]”? What happens? What did you expect to happen instead?
<leoprikler>try to figure out what the gnome app menu does, then do that from terminal
<leoprikler>i.e. search for peek.desktop
<dftxbs3e>nckx, unfortunately :-S
<jlicht>my guess: XDG-related shenanigans
<boombim>rekado_ it started from terminal and works correctly. but it even doesn't start from gnomee app menu
<rekado_>so it appears in the gnome menu but when you select it nothing happens?
<boombim>another apps works. like transmission
<rekado_>boombim: do you see anything in log files?
<boombim>which log file should i expect?
<rekado_>no idea! :)
<rekado_>segfaults are recorded in /var/log/messages, for example
<Rovanion>rekado_, jlicht: I have no idea what constraints you are working with here, really. What I would do would probably be to throw up a Gitea instance and use it as an second, alternative, way to contribute to the project. But again, I don't know why you do what you are currently doing. All I'm saying is that the experience of coming to this channel and getting help is a magnitude better than the
<nckx>leoprikler: Behold, the advanced nckx IME:
<Rovanion>experience of submitting a patch.
<boombim>it works from terminal... maybe it's some xdg crash...
<boombim>i'm not sure
<leoprikler>oh 💩, why didn't i think about that?
<rekado_>boombim: do you see something in /var/log/messages?
<dftxbs3e>rekado_, I think it's more meta than that, emotional feeling of people, in the Rust community for example when I ask a question often there's 4 answers from different people when I'm alone asking it, I don't even ask for so much. I think it's about responsiveness, general enthusiasm in contributions.. so people don't feel left behind/held back.. I don't know how to describe precisely what I felt and why last 2 years trying to do something with
<dftxbs3e>GNU Guix.
<raghavgururajan>> rekado_‎: segfaults are recorded in /var/log/messages, for example
<raghavgururajan>For any application?
<nckx>Rovanion: The problem with Gitea is that anything that forces people to start using a Web browser (💩) is going to lose people (like me, for the little review work I do).
<nckx>It's a big hurdle.
<rekado_>boombim: if not, perhaps try to attach strace to the gnome-shell process to see what it’s doing
<jlicht>boombin: Does `gtk-launch peek' do/show?
<nckx>dftxbs3e: You're just describing ‘a huge community’.
<boombim>rekado_ I see nothing about peek in log
<dftxbs3e>nckx, maybe.. GNU Guix's pretty big as well!
<jlicht>s/Does/What does
<raghavgururajan>> ‎nckx‎: Rovanion: The problem with Gitea is that anything that forces people to start using a Web browser (💩) is going to lose people (like me, for the little review work I do).
<raghavgururajan>+1. I was fond of Gitea, but now SourceHut is growing-up on me.
<rekado_>dftxbs3e: yeah, I generally don’t respond when I’m asleep.
<boombim>jlicht should I invoke it from terminal?
<rekado_>I think it’s the same for other people ;)
<jlicht>boombim: yes, sorry for not mentioning that
<boombim>but I already said it works from terminal
<boombim>just with peek command
<dftxbs3e>rekado_, Don't take it that way, I don't think it's because you sleep
<leoprikler>I'm having a look at the desktop file right now
<rekado_>I can only speak for myself, but I’m very enthusiastic about contributions. It’s just that my contingent of freely allocatable time is very limited.
<nckx>dftxbs3e: I know you don't intend it that way, but it's not nice to hear that we're supposedly less willing to help people. I think that's demonstrably false. It's about 90% of what I do here, sometimes successfully sometimes not, and I always have something else that I could be doing instead (that would make me €).
<jlicht>boombim: gtk-launch should use the .desktop file: it' might give us an extra data point to find out where the issue is at
<dftxbs3e>rekado_, yes, I don't say you are not enthusiastic, I say people may not feel that enthusiasm e.g. the ZFS guy recently
<leoprikler>it doesn't really look abnormal though
<dftxbs3e>nckx, I never said anyone was less *willing* to help
<nckx>That's how it came across re: Rust.
<boombim>jlicht oh got it. let me check please
*nckx goes to do some of those things.
<dftxbs3e>nckx, it's a time problem, not will, I am saying
*rekado_ too
<Rovanion>nckx: Right. And forcing new contributors to set up an email client that doesn't mangle in-message patches looses other people. But it doesn't have to be an either or situation. You can have both systems running at the same time where seasoned contributors continue using the email based system. And maybe, through magic and ugly hacks, one can stich them both together; but that feels fragile.
<rekado_>Rovanion: attachments work all the time
<boombim>jlicht prints that peek is absent
<rekado_>“both systems running at the same time” really won’t work
<dftxbs3e>rekado_, I think Sourcehut can do both web contributions and mailing list
<raghavgururajan>leoprikler: Are you up for another ride (#46614)? xD
<boombim>hovewer I see peek in my gnome app menu.
<rekado_>dftxbs3e: but we use Debbugs. And switching comes with great cost, that those suggesting it generally underestimate.
<rekado_>Rovanion: I can only repeat myself: you can push your commits to a branch anywhere. You can even upload the diff to a paste site.
<rekado_>I’ve worked with all of these methods to lower the barrier for new contributors
<raghavgururajan>debbugs is not bad though. Works seamless without JS.
<rekado_>raghavgururajan: well… It’s … it exists and it works with email.
<dftxbs3e>rekado_, I don't think using Sourcehut to allow easier contributions actually requires to replace anything that exists.. but then there has to be willingness to actually recognize a problem and solve it, if we can't agree on the problem then there wont be a solution
<pineapples>And so it all happens again
<rekado_>I’m not too fond of it, having wasted many of my few hours each week to put lipstick on it.
<raghavgururajan>rekado_: Yeah. Was there anything reported against debbugs?
<jlicht>boombim: hmm, that does throw me for a bit of loop. How about `echo $XDG_DATA_DIRS'? (terminal again, but please share over a paste service
<rekado_>raghavgururajan: the SOAP service has a number of known bugs that won’t get fixed
<raghavgururajan>> pineapples‎: And so it all happens again
<raghavgururajan>Are you taking about #46614?
<raghavgururajan>rekado_, I see.
<Rovanion>rekado_: Does the manual instruct the new contributor in these methods?
<rekado_>sorry, what methods?
<rekado_>the manual introduces the methods we use.
<Rovanion>rekado_: Submitting changes through a git repo for example.
<boombim>jlicht here is it. By the way, I have add some line in my .profile to it manually
<rekado_>we can make exceptions, but it puts extra burden on the reviewers
<raghavgururajan>dftxbs3e, rekado_: The repos on savannah will eventually gets moved to FSF's Forge right?
<raghavgururajan>Atleast thats what I am hoping for
<rekado_>from the perspective of a contributor, I’d pick a method that increases the likelihood of actually getting reviewed.
<boombim>export XDG_DATA_DIRS=$XDG_DATA_DIRS:$HOME/.guix-profile/share
<boombim>I added this to my .profile in home directory
<rekado_>raghavgururajan: I don’t know what FSF’s Forge is.
<rekado_>my time is up now, bye!
<dftxbs3e>raghavgururajan, I don't know, if they are, and the forge is Sourcehut or something, then that would be a really happy thing for me
<leoprikler>Isn't Savannah FSF's Forge in a sense? :D
<jlicht>boombim: I already assumed that you did :). Silly question, but have you tried logging out and into your user session? I'm out of ideas (and beyond my expertise) if that doesn't work either :/
<raghavgururajan>leoprikler: Yeah, it is in a sense.
<Rovanion>rekado_: Are we not talking about improving the process for newcomers? If this is a common hurdle, why argue that its pseudosupported if someone insists on it?
<boombim>jlicht yep I already did it.
<dftxbs3e>nckx, rekado_: I am sorry for any misunderstanding, I wanted to describe a feeling I had not say anyone was doing anything wrong, that's never what I thought and didnt mean it either. So I am sorry again, I will be pouring time into GNU Guix and it's community as you also do so it will hopefully improve the experience for others.
<jlicht>boombim: I don't know what you can still try then, sorry. Perhaps send an email to the help-guix mailing list describing your issue, in the hopes that somebody has experience with this issue
<jlicht>the fact that your menu's seem to find the 'peek.desktop' file, but gtk-launch doesn't really puzzles me though
<leoprikler>I think there might be mismatch between the environment of the terminal and Gnome Shell itself.
<jlicht>Rovanion: What do you mean? You think there should be no exceptions to established procedure, or that these exceptions should instead become the established prcedure? Not trying to strawman you, just trying to understand the point
<boombim>jlicht thanks. maybe I will try
<leoprikler>are there perhaps some weird things in your .bashrc?
<leoprikler>(e.g. setting variables like PATH)
<Rovanion>jlicht: I think there should be a supported git-repo-based way in to the project. An easy way for all those younglings who have only ever used GitHub to get into the project. It should be described in the documentation as an alternative way to contribute.
<leoprikler>I personally disagree.
<Rovanion>The practicalities of how that's accomplished, I don't know. But I'm sure it would increase the number of contributors.
<jlicht>Rovanion: there is a git-supported way in: git send-email :^)
<jlicht>but I understand the underlying point you are trying to make; back when I was still a young man in 2016 I think I even mentioned to the same to ludo'
<dftxbs3e>Rovanion, I agree with you but I don't think we lack contributors right now though, more so reviewers, but we are improving tooling to ease reviewing.
<dftxbs3e>There's almost 1000 patches pending review
<nckx>dftxbs3e: And I'm sorry for sounding like I ragequit on you, I really just realised I had work to do 😛 (1000... sounds low... TBH...)
<dftxbs3e>nckx, well that's what cbaines's patchwork says it's probably more if it didnt parse the whole archive
<nckx>I hope you're right!
<dftxbs3e>nckx, Happy smile to you
<dftxbs3e>rekado_, and you too :-D
<dftxbs3e>Rovanion, cbaines was working on something to allow "git push" contributions somehow I think
<Rovanion>dftxbs3e: 1. I'm not sure attracting contributors excludes attracting reviewers and 2. it seems more likely that a person starts as a contributor and becomes a reviewer rather than the other way around.
<Rovanion>dftxbs3e: That sounds promising.
<dftxbs3e>At least their logic with to make branches from individual patches
<Rovanion>My terminal font lacks most emojis, but happyface I can read :D
<nckx>GitHub is exemplary of bad UX (you type git commit! you start a browser to visit github! you log in! you switch to a terminal! you type git push! you switch back to the browser! etc.) that shouldn't be copied just because it's familiar.
<dftxbs3e>nckx, I think Emacs might have some Github extension aha
<nckx>There's also a ‘git hub’ extension, I believe.
<jlicht>They can pry emacs-forge from my cold, dead hands
<nckx>But compare that to mail where you're never forced to start your MUA unless you choose to.
<nckx>Yes, because the kernel dev who wrote git likes mail so made it first-class, sure.
<nckx>But he did so because he has to deal with the result & it's efficient.
<leoprikler>tbh using command line tools to interface with some web page sounds like a hack (in the negative sense) 99% of the time
<leoprikler>the 1% is youtube-dl
<nckx>You can't ‘brand’ seamless interaction because people don't see it. That's why GitHub works the way it does. Not because it's user-friendly.
<nckx>(Yep, this is the 10% of stuff I do besides helping folks ☝)
<dftxbs3e>I have a bit of distaste for the email clients around here, or email in general, it doesnt feel great to use, I actually would like it better if it I had a better email client that isnt only keyboard controlled, mouse too, but integrated well for git/debbugs/emacs.
*nckx goes back to work!
<leoprikler>but you can trademark it: seamless™ interaction
<nckx>one click™ commits.
<raghavgururajan>What endianness does guix uses?
<raghavgururajan>Or is it specific to kernel?
<leoprikler>usually you should be fine with native-endianness
<jlicht>dftxbs3e: mailspring was recently freed afaik, perhaps it offers some of what you would like
<dftxbs3e>jlicht, FYI I use GNOME Evolution right now
<dftxbs3e>I'll have a look
<leoprikler>Evolution is great
<jlicht>dftxbs3e: it is electron based though, so no dice on getting it packaged in guix any time soon
<mothacehe> hey guix! I'd like to be able to send automatic emails to "" using Cuirass. Do I need to setup a MTA such as Postfix or is there an easier way?
<dftxbs3e>leoprikler, can't wait for GNOME HID redesign
<leoprikler>I think Geary was supposed to be that
<dftxbs3e>mothacehe, you can just use "mail" command I think.
<dftxbs3e>SMTP outgoing
<dftxbs3e>No server configuration at all
<leoprikler>IIRC it uses the same backend, but I forgot whether you could reuse your configs
<leoprikler>GOA should work at least, which is nice
<dftxbs3e>jlicht, I think I am probably better with GNOME Evolution for now heh
<dftxbs3e>mothacehe, what is your question exactly? Is it specific to know-how about how GNU infrastructure works? If not then the only problem might be that your mail gets rejected as spam, so maybe authenticated SMTP somewhere to GNU servers with a bot account, is that possible?
<dftxbs3e>leoprikler, Geary might be better than GNOME Evolution (may be less featureful)
<mothacehe>dftxbs3e: looks like mail tries to connect to a local MTA. My question is what's the easier setup to be able to send automatic emails to ""
<dftxbs3e>mothacehe, I think you can configure "mail" to connect to's SMTP directly
<mothacehe>without authentication?
<dftxbs3e>mothacehe, well would see the email as if it came from the outside, but it will probably reject it as spam since you don't have a domain with SPF, DKIM, proper reverse-dns.. I don't know for sure..
<dftxbs3e>mothacehe, what Postfix would do is basically that I think.. forward to what's after the @ over SMTP
<mothacehe>dftxbs3e: I see, thanks for explaining I'm such an email noob!
<dftxbs3e>mothacehe, I can't believe there's no standalone client
<bandali>mothacehe, feel free to email and ask for advice/suggestions on what would be the best way to go about doing something like this, from their perspective. they're the folks who manage, among other things, gnu's mail servers :-)
<dftxbs3e>mothacehe, "sendmail" package also
<mothacehe>sendemail -f -t -s
<mothacehe>The recipient <> was rejected by the mail server, error follows, 550 relay not permitted
<mothacehe>it means I need to be authenticated I guess
<dftxbs3e>mothacehe, does the guix-ci email exist?
<dftxbs3e>mothacehe, It can't be that it requires authentication, as GNU mail server can't authenticate email from the outside (other mail servers, like what you would me, but without a server), it could however reject as spam
<nckx>mothacehe: You need to be authenticated to send e-mail through
<dftxbs3e>like what you would be *
<bqv>Say what you will about mailing list workflows having a barrier to entry, but at least people will always ****ing reply, unlike the constant ******** you see on github
<bqv>I genuinely don't get how someone can think it's ok to completely ignore being github pinged three times over months, while still visibly doing other things
<dftxbs3e>mothacehe, drop the -s param?
<dftxbs3e>bqv, ehh github gives no visible notifications unless you really go and look for them on the web UI, and if they don't check their emails..
<nckx>mothacehe: You can either create a mail account anywhere and have Cuirass's sendemail (or whatever) log into it using a user name & password, or set up to trust berlin's IP, or you need to set up a proper MTA on berlin (opensmtpd > postfix, but don't actually do this, apart from the firewall people going possibly apeshit it's no fun to maintain a mail server just to send out pings).
<nckx>Definitely ask the people, I'm sure they've dealt with this before.
<dftxbs3e>nckx, do they need though? How does GNU trust random mail servers? It "trusts" them by default and has spam filters, no?
<mothacehe>Setuping to trust berlin's IP seems like a fine solution, I'll write to the sysadmins, thanks dftxbs3e and nckx!
<bqv>dftxbs3e: people say this, but these folk know they get pinged. It's basic politeness imo
<bqv>To at least check and respond with a "not right now" or something, if busy
<bqv>People don't completely blank emails generally, but it's somehow not even just ok, but encouraged, if it's github
<dftxbs3e>bqv, Well.. I've had tons of unanswered emails.. you can't say people are *forced* to reply
<nckx>mothacehe: thank bandali 😉
<nckx>dftxbs3e: even the most poorly set up random mail server has rDNS correctly set up, berlin doesn't, etc.
<nckx>& trying to do so just got me ‘451 greylisted for 20 minutes’ 😛
<nckx>To submit anyway, I mean, I didn't set up rDNS.
***rekado_ is now known as rekado
<ngz>Hello Guix
<nckx>Hullo ngz!
<pineapples>Should ever Linux lose its funding, manpower and be overshadowed by an emerging alternative, and Hurd is never completed; what will happen to GNU Guix? Can it be ported to other POSIX, FLOSS kernel, and have its development continued? Just asking because I'm concerned about whatever the future holds for free computing in general, should it ever be threatened
<rekado>pineapples: Linux isn’t all *that* important for Guix. There are only few things that directly depend on Linux (like (guix syscalls)). Guix would work on top of whichever kernel the GNU C Library can talk to.
<nckx>pineapples: With *relatively* little effort.
*raghavgururajan thinks Hurd gets completed as a part of Guix. <3
<narispo>pineapples: And these kind of things don't happen overnight (Linux dying that is)
<narispo>raghavgururajan: looks more likely than pre-Guix era for sure :-)
<nckx>raghavgururajan: That would be plan A!
<nckx>...from kernel space...
<mothacehe>sneek: later tell lfam It is now possible to restart individual builds as well as evaluations from the Cuirass web interface.
<sneek>Got it.
<roptat>mothacehe, can anyone do it, or do you need to be authenticated?
<mothacehe>roptat: nope only for administrators
<roptat>ok :)
<mothacehe>but the button is for now visible by anybody
<roptat>oh no:
<roptat>how could the store path disappear?
<mothacehe>that's again the qemu-binfmt service that needs to be restarted
<roptat>can you do it and restart the build?
<raghavgururajan>pineapples: Did you know? Guix is working on a proprietary kernel under the hood, to mitigate the scenario you are concerned with. It will be released on 2021-04-01.
<mothacehe>I just did it, let me see if I can use the new button to restart it :D
<mothacehe>... and done!
<roptat>thanks :)
<roptat>oh, but that's staging, not master, so I don't think it'll succeed anyway
<pineapples>rekado: That's reassuring to read
<pineapples>nckx: Even more so
<pineapples>raghavgururajan: ;O
<rekado>mothacehe: do we not use the /admin URL prefix any longer?
<mothacehe>yes we do, the restart build/evaluation button in based on the /admin mechanism
<pineapples>narispo: I am aware but I've needed to ask. I had a talk about the future of Linux, and a friend of mine pointed out that Google's FOSS kernel might be the next hit, which worried me a little
<dftxbs3e>pineapples, Free Software cannot be deprecated just like that as long as it has users, organizations behind GNU/Linux distributions like this smaller one (GNU Guix) depend on the Linux kernel and will show up to at least keep a project in maintenance mode if anything occurs.
<dftxbs3e>If Linux was to be frozen, it wouldnt hurt much for at least 5 years. Until really hardware support could become lacking..?
<leoprikler>Let's port guix to sweb!
<dftxbs3e>Fuschia is not meant for any other use-case than their Chromebooks and Android for now..
<pineapples>dftxbs3e: Yes, perhaps it wouldn't hurt in the short term but in the long term... that's why I've asked if Guix can be ported to other FOSS kernel, which rekado and nckx have already addressed
<pineapples>adressed that question, that is*
<dftxbs3e>If "long term" then that means GNU Guix (or any software) also has time to adapt within that time
<pineapples>This is also reassuring to hear
<pineapples>Thanks everyone for your input on the matter :-)
<mroh>hey #guix!
<mekeor[m]>hi mroh :)
<boombim>Guys is it possible to add deadbeef to repo?
***bernard is now known as Guest99272
<jackhill>boombim: after a cursory glace at the project, it looks like it would be a great candidate for inclusion. It apprears to be entirely free software, and doesn't seem to have any problematic dependencies (e.g. JavaScript apptication are difficult to package currently)
<jackhill>The only thing I notice is that they say to use git submodule update. This makes me suspect that we'd want to package those separately in Guix (alhtough I didn't look at what they are).
<jackhill>boombim: Do you want to try your hand in packaging it?
<boombim>jlicht it's popular gui music player and I suppose many people will use it on guix.
<boombim>But as for me I'm not good in packaging and I have so week laptop. That's why I asked about it here.
<boombim>oh sorry wrong tag jackhill
<PotentialUser-78>Hi, I'm considering guix as my new os, and I see that it can compile packages and is even compared with portage. However I can't find anything about compilation options in the doc. Does guix support custom compilation options like -O3 and so on ?
<bavier[m]>boombim: we could add it to the wishlist!
<leoprikler>In Gentoo speak: No, Guix does not have CFLAGS. But if you really want to, you can rewrite your entire package tree to add -O3 to #:configure-flags everywhere
<boombim>bavier[m] it would be great!
<jackhill>boombim: Fair enough. More powerful machines to help. However, I've had a lot of fun working on Guix, so don't give up on working on it entierely. I believe is the wishlist.
<PotentialUser-78>@leoprikler thanks
<bavier[m]>boombim: I agree with jackhill. I've had plenty of fun playing with guix and packaging on a little 12-year-old netbook :)
<boombim>Unfortunately It's my main laptop and i'm working on it... I haven't plenty amount of cpu power to pack something
<raghavgururajan>I did it
<raghavgururajan>i dit
<raghavgururajan>I did
<raghavgururajan>Holy shit
*raghavgururajan sings
*raghavgururajan dances
<raghavgururajan>leoprikler, pineapples
<pineapples>I'm here!
<raghavgururajan>no segfault now
<raghavgururajan>Telegram Calls!!!!!
<raghavgururajan>Yeaahhh ooooo!!!!
<pineapples>Tell us the secret!
<pineapples>what's the secret formula?
<raghavgururajan>In your trace log, I noticed line could not access mem at ... gcc-7 ...
<raghavgururajan>I am like. .. wait a sec .. tdesktop was built with gcc-9.
<raghavgururajan>Then I realised we were building webrtc with gcc 7, which is a mismatch.
<raghavgururajan>I rebuilt webrtc with gcc-9. Then hola!
<raghavgururajan>gcc-7 and gcc-9 uses diff C++ standard which inludes difference in memalloc stuff.
<pineapples>„The darkest place is under the candle”
<rekado>gah, trying to update r-shiny but it now depends on more packages that bundle massive amounts of minified JavaScript
*raghavgururajan makes some gagnam-style moves !!!!!
<nckx>raghavgururajan: Absolute congratulation achieved. 👍
<raghavgururajan>nckx: Thanks!!!
<rekado>raghavgururajan: woo, that’s great!
<raghavgururajan>rekado, :-)
*raghavgururajan treats himself with a pizza
<Urist>Hello. I'm failing to define emacs-package. It's in 'lisp' subdirectory of git repo. If I use repo where .el is placed directly in repo it builds. The error is at 'install' phase. Guix considers installing all files in repo (including lisp/kbd-mode.el), but no luck.
<nckx>raghavgururajan: and later a good night's sleep 😉
<nckx>Urist: From your description it sounds like you can (add-after 'unpack 'enter-lisp-directory (lambda _ (chdir "lisp")))
<nckx>the exact position (add-before 'whatever, etc.) may vary.
<Urist>thx. Makefile is ignored, I see.
<raghavgururajan>pineapples: Thanks a lot! I couldn't have fixed the issue without your trace.
<leoprikler>Urist: maybe you want gnu-build-system?
<leoprikler>raghavgururajan: I see your patch already, but it doesn't seem to have a number yet, so I'll respond here.
<raghavgururajan>leoprikler: Cool!
<leoprikler>In my personal opinion, you can squash your changes into a single patch. I'll have to leave verification that it works to someone else however, as I don't use that feature of Telegram.
<pineapples>raghavgururajan: No, I thank you, again! Hadn't it been for you who'd packaged it to begin with, I wouldn't have been on Guix System ;-)
<pkill9>does anyone know of a web service for hosting and editing static site blogs? something that has an editor like wordpress, but using markdown and producing a static website blog?
<pkill9>I heard of one ages ago that included an editor, but can't remember what it was
<leoprikler>Gitlab Pages, but unironically 😉️
<rekado>pkill9: I don’t know of any services that have an online editor *and* produce a static site.
<rekado>there are a lot more choices when building the site offline
<rekado>e.g. haunt
<raghavgururajan>leoprikler: You can ignore the 2nd patch for now. I'll add it when I update that package again in the future. The 1st patch is the critical one.
<jlicht>rekado: Which javascript package are you talking about?
<raghavgururajan>pineapples: Could you apply the change and confirm to leoprikler if calls work?
<jlicht>raghavgururajan: you are a person of focus, commitment and sheer will :)
<raghavgururajan>jlicht: Thanks! :-)
<rekado>jlicht: bslib bundles the whole kitchen sink.
<rekado>jlicht: but I see now that I can get sources for some of the bundled minified things (for the rest the sources are included)
<jlicht>rekado: most of it seems doable, if you don't care to reach exactly the same build outputs as upstream
<bavier[m]>pkill9: I recently heard about Write Freely, includes a markdown editor and can federate:
<rekado>jlicht: yeah, just finished it
<rekado>took longer than I thought, but it could have been much worse.
<rekado>(like r-leaflet)
<nckx>smithay-clipboa[7417]: segfault at 21 ip 00007c5df97688e3 sp 00007c5df07075a8 error 6 cpu 1 in[7c5df9764000+5000]
<nckx>So that's why I can't paste.
<nckx>(I always forget to look in dmesg for user-space bugs.)
<pineapples>raghavgururajan: Yes, I can
<pineapples>Where do I find the patch?
*pineapples is having lunch
<raghavgururajan>pineapples: Thanks! It seems debbugs is also on lunch break (no bug number yet). I'll send here.
<raghavgururajan>leoprikler: Does the following squashed patch look good?
<pineapples>raghavgururajan: It's building webrtc-for-telegram-desktop right now. Expect the results in 30-40 mins
<jackhill>raghavgururajan: It looks like you may have sent it to guix-devel@ instead of guix-patches@
<raghavgururajan>jackhill: Damn! Thanks
<raghavgururajan>pineapples: FYI, you can download substitute from
<raghavgururajan>Make sure to `guix pull --commit=8f710cc598fb675d267f49a82cffc197f03ad52c` before attempting to build.
<raghavgururajan>leoprikler: Got the number. #46618 :-)
<jackhill>raghavgururajan: no worries/you're welcome. I knew I saw if flash by my inbox.
<pineapples>raghavgururajan: It's halfway through the process of building telegram-desktop, and I don't think the subsitute for Telegram is available, if the patch wasn't merged yet
<pineapples>Whoops. I meant webrtc-for-telegram-desktop*
<raghavgururajan>pineapples: Cool! Btw, that server is different.
<pineapples>How to download substitutes from it?
<raghavgururajan>Actually, never mind. I haven't built there with latest patch.
<raghavgururajan>So substitutes wont be available.
<pineapples>raghavgururajan: Just to be sure: ("gcc" ,gcc-9)
<pineapples>and ("perl" ,perl) are what was added?
<pineapples>Now it's actually halfway through the process of building telegram-desktop, FYI
<pkill9>thanks bavier[m]
<raghavgururajan>sneek, later tell civodul: Shall we create #shepherd on freenode? Since the success of shepherd in guix, other distros would like to provide shepherd as init system. Folks will look for #shepherd
<pineapples>Speaking of Shepherd, it can't initate the reboot sequence if the kernel experienced a kernel oops, which a huge issue to me
<pineapples>(not a kernel panic but kernel oops)
<pineapples>It forces me to use REISUB, which I'm not very fond of
<pineapples>Oh, and calls work! I've just tested the call functionality :)
<pineapples>I just need to confirm echo cancellation works as well
<raghavgururajan>Yay! Thanks pineapples.
<raghavgururajan>leoprikler, ^
<pineapples>You're welcome
<pineapples>Echo cancellation seems to be working, too
<leoprikler>raghavgururajan: which one was built on bayfront?
<raghavgururajan>Someone in 2016, wished if Telegram was distributed with GNU Guix.
<raghavgururajan>Now it has happened. ;-)
<raghavgururajan>leoprikler, just a sec
<civodul>raghavgururajan: not sure #shepherd is warranted yet :-)
<sneek>civodul, you have 1 message!
<sneek>civodul, raghavgururajan says: Shall we create #shepherd on freenode? Since the success of shepherd in guix, other distros would like to provide shepherd as init system. Folks will look for #shepherd
<raghavgururajan>civodul: Ah okay. pineapples just had shepherd specific messages. :-)
<raghavgururajan>leoprikler: I am building #46618 on bayfront now.
<raghavgururajan>leoprikler: But substitutes should be available without perl.
<lispmacs[work]>where in the file system does guix keep its copy (copies?) of the guix repository that has been pulled?
<raghavgururajan>pre-inst-env is on 8f710cc598fb675d267f49a82cffc197f03ad52c
<leoprikler>Ahh, the former was sent to guix-devel, I see.
<leoprikler>what is perl needed for in webrtc?
<apteryx>eh, pam-security-limits is tough; one mistake, and you can't login anymore.
<apteryx>I meant, 'pam-limits-service'. That's not Guix's fault, it's documented as such in the manpage.
<raghavgururajan>leoprikler: For patch-shebangs phase
<leoprikler>Ahh, so there are some perl scripts, which don't seem to be used (yet)
<PotentialUser-93>Hi everyone
<PotentialUser-93>Is there any way to monkey-patch the guix kernel ? something like ubuntu's livepatch
<pkill9>the "patches" field?
<lfam>What changes do you want to make PotentialUser-93?
<sneek>lfam, you have 1 message!
<sneek>lfam, mothacehe says: It is now possible to restart individual builds as well as evaluations from the Cuirass web interface.
<lfam>Wonderful mothacehe!
<leoprikler>I think it's meant like "patching the currently running kernel".
<dftxbs3e>PotentialUser-93, Canonical's Livepatch is based on kpatch, first it's not packaged in GNU Guix, then GNU Guix does not provide kpatches like Canonical does, it's heavy work of a kernel engineer we don't have that currently.
<PotentialUser-93>I mean updating the kernel and apply the update without reboot
<lfam>Oh, we don't have support that (yet?)
<pkill9>oh cool
<pkill9>didn't know that could be done
<lfam>It would be nice, for sure
<PotentialUser-93>I see
<lfam>It's would represent a break with the Guix paradigm, I think, but it would still be nice to have
<leoprikler>How does kpatch even work? Like kexec or differently?
<PotentialUser-93>a break of the paradigm ?
<dftxbs3e>PotentialUser-93, maintaining kpatches is specific to each kernel version with specific kABI (usually LTS releases), it's really time-consuming to make and requires expert people at hand and reactive
<pineapples>raghavgururajan: I've tested Telegram's buil-in acoustic echo canceller with help of other friend, and I'm now 100% sure that all the functionality that I've missed from the official builds works now ;)
<dftxbs3e>leoprikler, It hot patches kernel functions with ftrace functionality
<lfam>leoprikler: There's a primer on the github page
<raghavgururajan>pineapples: Glad! :-)
<lfam>PotentialUser-93: I mean that, usually on Guix, there is a simple correspondence between running software and a particular revision of the Guix package definitions. Patching the running kernel would mean stepping beyond that correspondence
<lfam>It's not too different from restarting running services after `guix system reconfigure`, however.
<dftxbs3e>PotentialUser-93, yes because in GNU Guix packages are "functional", in that any change to them affect all their dependents and so on. However the Linux kernel is a bit special there.. in GNU Guix it could be made a service but then GNU Guix kind of tries to provide a guarantee of "what do we run now, from which source", and kpatch is not very friendly to that, you'd have a Linux kernel package + a kpatch service with arbitrary patches.
<lfam>That was controversial in its time as well
<lfam>In the end, it was decided that the convenience was worth something
<PotentialUser-93>i feel so
<pineapples>On a unrelated topic, do I install v4l2loopback-linux-module to my system profile, and add a simple-service that loads it?
<pineapples> order to use it?
<dftxbs3e>PotentialUser-93, patches applied by kpatch need to be specifically tested so that they don't cause any runtime bugs or anything, they need to precisely preserve internal struct memory representation etc.. it's very sensitive, they require lots of rigorous testing.
<dftxbs3e>(in that running git diff in the GNU Guix package and giving that to kpatch isnt sufficient)
<dftxbs3e>This is also why kpatch is always provided as a separate (more expensive) service.. (not as part of the distribution itself)
<PotentialUser-93>I guess that's basically a design problem with linux (monolithic kernel)
<dftxbs3e>It's specially an enterprise use-case to absolutely require no reboots
<dftxbs3e>PotentialUser-93, yes, there's some of that here
<raghavgururajan>leoprikler: Finished building on bayfront. [Before Grafts] 59rz9hb984jfmxjjv86zwiqyjfk1r6hd-telegram-desktop-2.5.9.drv [After Grafts] 1gv0y7gwp3ndfiwzwkvsickhmdyd7gql-telegram-desktop-2.5.9.drv
<PotentialUser-93>About that, How active is guix hurd ?
<PotentialUser-93>or whatever the name is
<dftxbs3e>PotentialUser-93, It is "alpha".. I would say, so needs work still. But lots of thing work there already. But GNU Hurd itself needs work too to run on more hardware.
<PotentialUser-93>there was no usb last time i checked
<dftxbs3e>PotentialUser-93, It works however, 1-3 people are more or less doing things for it AFAICT, janneke mostly I think
<PotentialUser-93>and no x64
<leoprikler>sneek, raghav's telegram hash is 1gv0y7gwp3ndfiwzwkvsickhmdyd7gql-telegram-desktop-2.5.9.drv
<leoprikler>meh ._.
<leoprikler>raghavgururajan: I'll need about an hour at least for my own builds, will check hashes once that succeeds
<raghavgururajan>leoprikler: Ah okay. I thought you asked about bayfront to get substitutes. :-)
<leoprikler>chances are that they changed since I did a guix pull before, but let's hope they're equal
<leoprikler>I asked, because that's the one you recommended to pineapples.
<raghavgururajan>Yeah, mine was done on guix, pulled at 8f710cc598fb675d267f49a82cffc197f03ad52c
<leoprikler>my base is 11feecc23a20ad98088a2ed7b7b8a8940550e0ff
<PotentialUser-73>Is there some guidance on how to create, list and delete nodejs and python virtual environment using `guix environment`? I've been using python-virtualenv and python-nodeenv, and would like to ditch them if I can do everything with `guix environment`. Thanks for any advice or pointer.
<leoprikler>ouch, 8f is really old
<lfam>PotentialUser-73: `guix environment` creates ephemeral environments. They just manipulate environment variables so they disappear when you close the shell
<leoprikler>PotentialUser-73: Just hit C-d enough times
<lfam>But, you can use the --profile and the --root options to get a more persistent interface to them
<lfam>It would make a good blog post or cookbook entry, if there hasn't already been one published on the subject
<spk121>hey everybody - I sent an e-mail about the 2021 Guile Potluck, which is a sort of hack jam. Please consider participating!
<lfam>Oh that's cool spk121!
<bqv>I remember being a kid and thinking hurd was cool and I couldn't wait for it to be usable
<bqv>Fast forward a decade and...
<PotentialUser-73>lfam: yup, it feels like the blog post for it is yet to be written
<lfam>It's the kind of thing that's "obvious" to experienced Guix hackers but is quite inscrutable to outsiders. And it's a great reason to start using Guix, so we should make it more clear
<lfam>It could be a great first contribution for a potential user ;)
<lfam>A potential contributor :)
<lfam>I do recommend getting some familiarity with use of Guix, especially installing and removing packages, and then grokking "profiles"
<lfam>It should be achievable, and without that basic experience, environments won't be usable
<PotentialUser-73>lfam: correct, I'm still learning, and am still confused. But at least now I know it's possible, just waiting to be systematized.
<lfam>Any particular point of confusion?
<PotentialUser-73>when I `guix environment`, and then `exit`, I don't know how to re-enter what i left, and how to remove it properly if it is still around as a gc root. There are two directories for profiles (/var/guix/gc-roots/profiles, /var/guix/profiles), and it's not clear in which one I should delete the profile i no longer need. And that's after deciding whic
<PotentialUser-73>h profile I don't need, which mostly I don't know how to do, because the profile directories do not indicate the command that led to their creation.
<PotentialUser-73>i need to signoff, but i'll read whatever you reply from the logs tomorrow, and re-incarnate under another PotentialUser if I have more questions
<lfam>Basically, you re-enter it by running the `guix environment ...` command again. And the only way to remove or delete things with Guix is by using the `guix gc` command. Never delete files in /var/guix or /gnu/store unless you know what you're doing and are willing to void your warranty
<lfam>You can improve the "re-enter" workflow by learning how to use --profile and --root
<lfam>In general, environments are never "gc roots" unless you explicitly make them that way. Like I said, they are ephemeral
<roptat>aren't they temporary gc roots though? for as long as you're inside the shell?
<roptat>so that you can't destroy your environment with guix gc while you're using it?
<lfam>Yes roptat
<lfam>I think this is "expected" although it belies the underlying complexity
<nckx>awb99: Thanks for the Rserve berg repert.
<nckx>lfam: kpatch is *a lot* of work to maintain, and you're basically futzing around with binary diffs so it would be non-trivial to integrate into Guix. It's not something you do for fun just in case someone wants to postpone a reboot until after dinner.
<nckx>Should've kept reading; I see dftxbs3e made all the points I would!
<nckx>Even ‘binary diff’ is misleading, like ‘patch’: you're not applying a diff. Each patch is basically a bespoke hack, much more complex than the actual fix that hits Linux git. See <>.
<leoprikler>On that note, could we even do something like kexec in shepherd?
<nckx>Oh now that I don't see why not.
<nckx>kexec is handled by Linux, you can use it today.
<nckx>In fact I kexec'd my Guix System a few times today, because my ThinkPad BIOS is fuck'd ☺
<nckx>leoprikler: What would you see Shepherd add?
<leoprikler>not exactly sure where the right place would actually be, but I'd like something to "hot reboot" with /run/current-system/kernel
<leoprikler>or at least I don't know how I'd do it currently
<nckx>Well, that's a ‘kexec /run/current-system/kernel/bzImage <command line>’ one liner now, so the right place is wherever you want to invoke that, I guess 😛
<nckx>So instead of rebooting it would invoke that, right?
<nckx>Like a custom reboot command.
<leoprikler>yep, something along those lines
<leoprikler>like a kexec action on root
*apteryx fixed emacs-elpy, again. phew.
<jlicht>apteryx: just in time, I was expecting to need it come next week :D
<apteryx>about your pip problem; are you sure you don't have a wandering pip under ~/.local/bin/pip?
<apteryx>I had that problem earlier (such as would have been installed via 'pip install --user pip'
<jlicht>`guix environment --pure --container --ad-hoc python -- python3' breaks for me
*apteryx tries
<mrb`>Hello, I'm trying to implement containerized services in guix.
<mrb`>This are services that run isolated with all their dependencies in a container.
<jlicht>I use guix-home-manager, so no chance of pip (or any software) to do whatever in $HOME ;)
<mrb`>Why does the following gexp not work:
<mrb`>#~(make-forkexec-constructor #$(container-script ...))
<mrb`>container-script uses gexp->script and the gexp leads to:
<mrb`>error: #<procedure 7f61087c2380 at guix/gexp.scm:1762:2 (state)>: invalid G-expression input
<apteryx>jlicht: that's reassuring
<rekado>my offlineimap no longer works. It considers the mail body to be part of the headers…?
<jlicht>apteryx: hmm, I am a bit confused about the exact order that breaks things; in cli-usage, python should be before python-pip to make it work: in a manifest, python should be listed after python-pip to make it work
<mrb`>I have no idea why. ungexp should take derivations. Maybe someone here knows the problem.
<rekado>mrb`: container-script looks like it returns a monadic value
<rekado>that’s what a procedure accepting state is
<apteryx>jlicht: try guix environment --pure --container --ad-hoc python -- /bin/sh -c 'echo $PYTHONPATH;python3 -m site'
<apteryx>it'll show you the ordering of the sys.path
<apteryx>could it just be that this internal module exists just in the pip 20.X series?
<apteryx>the pip bundled is 19.2.3 (guix environment -C --ad-hoc python -- pip3 --version)
<jlicht>apteryx: bingo, I think you figured it out
<apteryx>seems like this yes; can I close the issue?
<apteryx>I'll add the explanation.
<jlicht>apteryx: hmm, but that means our virtualenv package is kind of useless as is: you need to have the proper version of python-pip installed (and in the right order!) in order to install anything in the virtualenv
<jlicht>I understand that guix kind of competes with virtualenv, but this doesn't seem like the most appropriate way to push users to guix :)
<apteryx>it *is* very useless on both master and core-updates, as our use of PYTHONPATH defeats its isolation
<apteryx>er, staging
<apteryx>core-updates uses GUIX_PYTHONPATH, which fixes that
<apteryx>You are welcome to try core-updates, and fix stuff which are not working there if you'd like it to be merged in master sooner rather than later ;-)
<jonsger>is kei kebreau in IRC?
<nckx>jonsger: I haven't seen them here in months.
<jlicht>apteryx: will do, thanks for the hand-holding
*jonsger wonders if he should start on wip-gnome3.34 or wip-gnome3.36
<rekado>jonsger: perhaps raghavgururajan could explain how much of the Gnome upgrade is in wip-desktop
***to-hu1 is now known as to-hu
<raghavgururajan>jonsger: I'll get back to you.
*raghavgururajan is in a meeting
<mrb`>rekado: Thank you. This helped me a lot. I did some reading of the manual and run container-script through the store now before using it in the gexp.
<jlicht>sneek: later tell apteryx: the pip included with python on core-updates works fine!
<sneek>Will do.
<nckx>awb99: The suggested command is here <>.
<jlicht>wow, the python situation on c-u is amazing!
<pkill9>what's c-u?
<drakonis>but what is the python situation?
<awb99>Hi @nckx
<awb99>I tried that command
<awb99>It's not working
<nckx>Can you be more specific.
<awb99>Pwrhaps this are 2 lines of code
<awb99>And I have to add export first?
<nckx>(You *really* want to export something, don't you 😛)
<nckx>R_HOME=$(guix build r-minimal)/lib/R guix environment --ad-hoc r-rserve -- Rserve
<nckx>is the full and only command.
<nckx>This will start Rserve in the background.
<jlicht>drakonis: I was suffering death by a thousand papercuts using python-tooling on guix, but one of the underlaying technical causes of my woes has been addressed on core-updates (namely, PYTHONPATH vs GUIX_PYTHONPATH)
<nckx>awb99: Any luck?
*nckx rebooting.
<awb99>It's working
<awb99>So this is one bash command that actually runs 2 commands
<awb99>Or how does it work?
<awb99>I don't quite get it
<nckx>awb99: Basically, yes. First it runs ‘guix build r-minimal’, which prints /gnu/store/...-r-minimal-nnn on stdout (ignore stderr here) and inserts that into the main command.
<nckx>$(command) is like a variable that expands to the output of command.
<nckx>Simpler example: ls $(guix build r-minimal)
<nckx>→ ls /gnu/store/98m419w7f9b0zsj6i7w10bhx3885glvq-r-minimal-4.0.4
*nckx needs to reboot again.
<awb99>I understand $()
<awb99>So the command is
<awb99>R_HOME=xxx guix environment zzz
<awb99>Can I split this up in two lines?
<awb99>I am trying to get this to work in my own config script
<rekado>if you have r-minimal installed in your default profile you could do “export R_HOME=$HOME/.guix-profile/lib/R” on one line and then “Rserve” on the next.
<milkey>code for (guix packages) says sha256 field on sources is deprecated and hash should be used instead, but I can't find a single package that uses that. should I still specify "hash" instead of "sha256" for new packages?
<nckx>awb99: You can export a thing! \o/
<civodul>milkey: 'sha256' is not really deprecated in practice
<civodul>you can keep using it