IRC channel logs
2023-02-03.log
back to list of logs
<nckx>tschilptschilp23: It's part of the glibc package, but indeed it 'just' prints some variables like LANG if set and defaults otherwise. <nckx>Oh, missed your /action. Uh, that's quite normal & healthy, no need to install glibc otherwise :) <apteryx>phew, 'guix pack -f rpm' test complete; it creates, installs and uninstall the RPM package successfully <elais[m]>So I have a question: I'm going to have a pi clusterboard pretty soon with room for 4 modules and I'm trying to think about how I want to orchestrate it. Of course I would like to install guix on all of the modules but I'd also kinda like to use something that abstracts over the compute modules and deploy services automatically with out me having to manually allocate. Kinda like using k8s or any other orchestration platform <elais[m]>Do you guys think ganeti is a good tool for that? <Kabouik>Somehow when I'm at work, resolving any URL takes like 5-10 sec. I don't have this issue at home, but no other colleagues complained about it. Could it be something in my Guix config causing DNS issues with a specific connection? <clh3>elais[m]: The `guix deploy' command seems related. I haven't used it yet. <ieugen[m]>how can I test updates to a package in a channel locally? Are best practices described someplace? <ieugen[m]>surely I don't have to commit the changes before I can test them right? <ieugen[m]>found it: guix package --install babashka -L /home/ieugen/proiecte/bin-guix/ <a12l>But after I'd ran `guix pull`, guix behaves as it should. <Guest39>it looks like the hash is expected to be the Sourceforge's one, while that download fails so guix uses multiprecision.org/guix which has a different hash. Can someone update either side to match? <jpoiret>Guest39: what guix version are you on, and do you have grafts enabled? <jpoiret>`guix build -S --no-substitutes --check xz` works for me, i'm downloading from sourceforge though <jpoiret>it's weird that the tarballs have differing hashes though <graywolf>Hello, is there a way to pretty print operating-system definition? It looks like (ice-9 pretty-print) does not work exactly well. <zrn[m]>Is guix stable right now? How often do you get errors? How easy it is to fix them? <rekado>zrn[m]: what do you mean by “stable”? <rekado>and what part of Guix in particular? It’s huge. <rekado>graywolf: how doesn’t it work well? How would you like it to behave differently? <zrn[m]>Like if you use it right, follow what is written it won't break or give error, or you could rollback and update on next version without problem. <rekado>I don’t know how it could break. It’s not like it’s running other than when you’re installing or building software; and it does that fine. <zrn[m]>Well maybe you install xfce and it kind of works but something is broken like menu. <rekado>okay, that’s different: you’re asking about software installed through Guix. That’s why I asked. <rekado>Guix itself is software, and questions about stability can just as well be asked of it. <lechner>No. Our software takes more effort to get to run properly, but everything is more predictable. Which OS are you coming from? <lechner>When you install a program in Guix (and when you run it) you get exactly the prerequisites that were used to compile or build it. We constantly recompile; Debian never does <mekeor[m]>is there any mail thread on how to isolate or sandbox guix-channels? i would like to run "guix pull" with untrusted channels.scm without threatening my security. <mekeor[m]>i know that this is currently not possible. but i wonder if there has been (written) discussion about it. <rekado>mekeor[m]: a channel can do any computation when it is evaluated. So at least you would need to run “guix pull” in a container where no precious files exist. <Guest39>jpoiret worked around sf.net vs multiprecision.org/guix hash mismatch by asynchronously overwriting downloaded tar.gz by a good one while guix pull was running.. <jpoiret>that problem should definitely be reported to the MLs though, it could probably be forwarded then to the xz maintainers <Guest39>I know its bad and dirty, but works;-) will see about further reporting.. <ArneBab>Does someone already have an updated node-package? Version 14 is quite old now … <merula>ArneBab: There is unmerged patch, I did successfully build it <graywolf>rekado: well it looks like this https://paste.vpsfree.cz/laHyK80H/raw/ which I not exactly pretty printed; Code producing this output is (use-modules (ice-9 pretty-print)) (pretty-print installation), where installation is (define installation (operating-system ...)) <rekado>graywolf: this value is not an s-expression, so pretty-print can’t do anything with it. <rekado>this is the printed representation of an opaque value <graywolf>I'm not insisting on using pretty-print, I'm just curous if it is possible to print it (somehow) in more readable manner, the ice-9 pretty-print was what I found, and did not work (thanks for explanation why), so I came here to ask if there are other options. <mirai>graywolf: does quoting it and calling pretty-print work? <graywolf>mirai: when I do '(operating-system ...), it does "work" as in it prints what I have in the source code, without evaluating anything (that's what the ' should do afaik afterall), but it's not what I would like; I want to see the actual final object (what is in the paste) just printed in more readable manner. <mirai>you could quasiquote it but yeah, that's a limitation <lechner>sorry about the failed bot tests. i do test another copy privately in another channel first but still have to learn the ins and outs of running Guile programs from a local repo. the bot will be offline until i return from crosswalk duty. yay, it is raining in california <nckx>jpoiret, Guest39: Did either of you already diff the tarballs, by chance? <zrn[m]>nodejs 14 is extremely old, is there specific reason why it is not updated? Is everyone using node from nix this whole time? <Htop170>hi guix! I am wondering why there is only one package using maven-build-system, is there some issues with it ? <apteryx>I think the RPM format addition to guix pack is complete! I'll send it to review shortly. <rekado>Htop170: it’s just a very late addition to Guix <rekado>Htop170: it took a long time before we could even build maven in the first place. <rekado>so obviously none of the dependencies of maven use the maven-build-system. <rekado>the ant-build-system appeared first. <rekado>zrn[m]: or rather does *not* use node at all. <rekado>zrn[m]: this is a collective effort and thus is shaped by the needs of the contributing users. <zrn[m]>Isn't node needed for many packages? <graywolf>zrn[m]: I'm not on guix (yet), but on my alpine laptop I don't have node installed at all <graywolf>I suspect it strongly depends on the type of work you do <Htop170>rekado: thanks! But it has been more than a year and half since it has been merged into the main branch, is it more difficult to write maven packages than ant? <rekado>Htop170: have there been any more Java packages added since then that use the ant-build-system? <rekado>Java packaging is just really tedious (because nobody in the Java world actually builds recursively from source), so the number of new Java packages we get is very low. <rekado>the maven-build-system is more convenient than the ant-build-system, because it is backed by more conventions <Htop170>rekado: no idea.. I'm trying to package Language Tool with maven-build-system, that's why I'm looking for examples <Htop170>rekado: by any chance, do you know if "old" packages written with ant-build-system pre maven patch (so no install-from-pom/install-pom-file) compatible with a package written with maven-build-system? Or they should be rewritten as well? <ieugen[m]>hi, I installed a man page for a package but it's not discovered. <nckx>apteryx: It can't/doesn't. <nckx>ieugen[m]: Install man-db (a man page reader). <apteryx>jjjjjxiiiej.dpypecpnii.phgutji.nihdtcdcuxgeg <nckx>I find the time-out a bit too zealous (short) as well. But that's subjective. <nckx>apteryx: Change password/cats. <apteryx>and the tactile key being connected to my keyboard, I sometimes touch inadvertently, and it causes this text to be spewd <apteryx>I don't know if it's just me, or if other users see the same behavior when using security keys? <mirai>ccccccvglckbhinckkifktdkfkrlgtgncurgnvhevgde <apteryx>I wonder if this leaks any worthy information <mirai>try jiggling in the tactile part <mirai>the output seems to change if you jiggle <Htop170>When packaging a lot of software that depends from each other, what is the best practice? one file per package or one single file with all the packages ? <mirai>Htop170: one file per package = never <mirai>think of the files for packages as categories <Htop170>sure but I meant during the phase you are writing them/testing their build <mirai>try to place the dependent packages where you think they best belong to <mirai>since that's only in your computer its up to you <Htop170>or is it more convenient to write/test them where they will be committed ? <mirai>but you'll have to sort them if you plan on upstreaming <Htop170>sure but it starting to get really messy haha <mirai>whichever ends being practical to you <mirai>for small-ish packages, in-place <nckx>You'll have to write more module imports if you write your own temporary file. That's the only drawback I can think of. <Htop170>I test them with the guix build -f name-of-mypackage.scm but I find that pretty annoying and fear to lost myself a bit at some point <nckx>Another hint: please add new packages 'more or less' alphabetically, unless there's a reason not to. <Htop170>I'm not here at the moment but dully noted ;) <nckx>Htop170: If you find it annoying, I'd just use ./pre-inst-env guix build from the start. If you want to contribute you'll need a checkout anyway. <Htop170>and write in the gitted files already ? <nckx>That's just what I prefer mysely, of course. <apteryx>jami weighs 2791.8 MiB according to 'guix size'. That's worst that I remember; is it caused by grafts? <apteryx>There are also dubious dependencies pulled in, such as python 2, gtk 2, cups, samba, etc. <apteryx>a XZ-compressed deb pack of it weighs 636 MiB, while an XZ-compressed RPM pack weighs 640 MiB <apteryx>I'll try enabling deduplication for the RPM <ieugen[m]>lrwxrwxrwx 1 root root 61 ian 1 1970 /home/ieugen/.guix-profile/bin/gum -> /gnu/store/908cv36kjndnglk0vh5106n80frd066f-gum-0.9.0/bin/gum <ieugen[m]>ieugen@daos-495:~/.../guix/bin-guix$ man /gnu/store/908cv36kjndnglk0vh5106n80frd066f-gum-0.9.0/usr/share/man/man1/gum.1.gz <ieugen[m]>nckx: I installed man-db but that did not make a change. <ieugen[m]>I also have MANPATH=/home/ieugen/.guix-profile/share/man:/home/linuxbrew/.linuxbrew/share/man: <ieugen[m]>found the issue, the man path for guix does not contain the man file for gum <ieugen[m]>got it to work. I was using usr/share/man and guix expects share/man <jonsger>we don't have that "usr/" stuff usually in Guix... <Htop170>How difficult would writing a maven importer would be ? <nckx>ieugen[m]: Sorry, I was away. I'd go even stronger than (correct) jonsger. /usr in Guix is always a mistake. <ieugen[m]>ok, thanks. I manage to make the man page work. Now I have the bash completions on my plate <ieugen[m]>I do think that I have to setup the MANPATH for man page to work <ieugen[m]>correction: seems like I don't. man gum works but I have no guix in MANPATH <ieugen[m]>ls /home/ieugen/.guix-profile/share/bash-completion/completions/gum <ieugen[m]>after copying the completion to my user specific dir, it works: <ieugen[m]>cp /home/ieugen/.guix-profile/share/bash-completion/completions/gum ~/.local/share/bash-completion/completions/ <nckx>Installing man-db should set it. That's why I suggested it. <nckx>I think I missed something (very bad wifi). You're talkinc completion now? I would not expect Debian's b-c to find Guix's b-c files. <ieugen[m]>yes, I am talking about completions. for man page I had the wrong path (with usr). That is fixed now. <ieugen[m]>ok so for completion I would have to add them to the completions path or something <lechner>yeah, it's (rnrs bytevectors) for me <mirai>lechner: UTF8 encoding isn't guaranteed <mirai>you might need to account for that <mirai>and use iconv or other utility for conversion <mirai>hexchat goes crazy with the hebrew text <lechner>that's not good. that's a really good web page. what happens? <mirai>mouse over the script part and the "text buffer" gets replaced with the link <mirai>happens with the arabic text as well <nckx>Hebrew looks perfect (to someone who speaks not a word of Hebrew) here on Revolution IRC. Yay HexChat. <lechner>mirai / the bibtex thing is still coming, btw. i hope to do it the right way™ <nckx>Can't wait for the bot to be banned by the other bot for spam when the next spambot spams a spam link :) <lechner>nckx / alternatively, my bot would be willing to assume the ban functionality---potentially after being donated to your patronage <nckx>Even if it were that simple (and it's not), I still wouldn't share our secret sauce 😉 I didn't mean you needed to mitigate this in your bot. Just that it's a funny eventuality. Nothing more. <nckx>*secrecy. My passwords aren't, generally, obscure. <patched[m]>How is portability of a guix home configuration? I am interested in letting it take over more of my dotfile management, but afraid I'll be in trouble if I ever need to use another distribution. Anybody use guix home while also using another distro? <apteryx>something drives me nuts when I use GNOME with its Terminal app in a VM; it keeps spawning a browser due to misdetecting the presence of URLs (?) <apteryx>has anyone encountered this before? any solution? perhaps I should install xterm and forget about it <Wurt>patched[m]: I do not use any other distro, but guix home does not have many services so on almost all programs you need to write their configurations files on their own languages. On the other hand guix home generates the configuration files, so you can copy them if you commit the heresy of changing your distro. <mvnx>determine provenance for current system" then it tries reaching out to update guix channel and spits out some non-recoverable error (long backtrace and I can't copy/paste from my VM). Any ideas? <apteryx>If someone wants to try generating some RPMs using Guix, it's ready! #61255 <rekado>apteryx: re jami and python2: that probably comes from lash, which we should remove. <rekado>I reported this a few weeks / months ago <rekado>needs to be done on core updates, though