<nckx>nefix: That would be a quick fix to make it work with the current go (I don't know if go is staging/core-updates and haven't checked). <nckx>Propagated inputs ‘pollute’ profiles: if a propagates b, installing a into your profile will also pull in all of b's commands. Like a classic distribution, but not very Guixy. <nckx>It can couse conflicts and is generally regarded as a bad move. <nefix>so, what would be the Guixy way to do it? <nckx>‘wrap [go's] …/link. Or patch its source to refer to /gnu/store/…/bin/ld directly’ <nckx>vpnc-scripts looks relevant. <nckx>Note the required ("guile" ,guile-3.0) ; for the wrapper scripts <nckx>The error you get without it is not clear. <nefix>how can you read a file to strings? <nefix>I've found local-file, but it returns a file type, not a list of strings <Noclip>Does the main substitue server contain known builds that aren't reproduceable? <nckx>‘No data’ is because there's no other substitute that also builds for that architecture to compare against. <Noclip>Why are there so many packages not reproducable? <roptat>the irreproducibility of a package can propagate to dependents ***nckx is now known as ^
***^ is now known as nckx
<Noclip>How about emulating 100% of the build environment? *nckx shrugs, you can't emulate the passage of time and the drop in performance is mostly a waste. Much better to spend your scarce resources actually making builds reproducible at the source. *Noclip thinks that nckx is right :D <nckx>E.g., enjoy building ungoogled-chromium on an emulated single-core CPU. Not an option. *nckx … is right? Might I have that in writing to show my friend? 😮 <nckx>‘friend’ was a typo but makes it extra sad. *Noclip thinks that nckx is wrong about that: "you can't emulate the passage of time" <nckx>Every clever ‘why don't you just’ strategy you come up to ‘reproducibly’ emulate time passing with will break something. Test suites don't just test the programme, they test your assumptions 😉 <Noclip>Maybe you would have to create a new emulator for that but in theory that should be possible. <nckx>I don't think so, but I also think the point is moot. <roptat>I remember someone posting a research paper on this subject to the r-b mailing list <nckx>Your emulator is going to be slow, not fast. Fast good. Slow no good! <nckx>Our build server are underutilised but still a limited resource. <Noclip>Your emulator would be required to always emulate the exakt same system and the emulated system would have to be 100% independent of affects from the outside. <roptat>yay, I fixed java-bsh reproducibility issue :) <Noclip>So it would basically have to be a "closed system". A bit like a little universe. <nckx>‘Theoretically possible.’ <nckx>Perhaps we agree after all 🙂 <roptat>not too hard, it just recorded the date in its MANIFEST.MF <roptat>I wonder how both ci and tobias got the same build, since they recorded the time to the second <nckx>roptat: Don't wonder. g.t.gr is currently tainted until the next c-u merge. I had to ‘restore’ from ci. after a disaster. <nckx>I'm going to bootstrap from scratch whenever c-u gets merged. <Noclip>How long did it take them to build? Compared to short building time a second might not be a very detailed info. <nckx>But I needed subs for my ‘clients’. <roptat>java-bsh take ~10s on my machine <roptat>but the time at which the build server will actually to the build is more or less random <roptat>because it might have many other builds scheduled before it goes to this one <roptat>so to have two independent builds get the same result when the result records the time is unlikely to happen <Noclip>But sometimes unlikely things happen ... <nckx>roptat: Interesting choice, I probably would've hard-coded something date-looking just in case something parses it. *nckx wonders why packages do this by default. Does it really help anyone when debugging? ‘Oh yeah, built at 16:32:06, I remember that exact change.’ <roptat>it's a free form field in the MANIFEST.MF, so I don't think it will cause issue <roptat>and it's nicer than say 1970-01-01 00:00:01 <lfam>It was probably a feature added by $tooling_developer in 1996 <lfam>The power of being first <Noclip>nckx: I assume back then noone thought about something like reproducability but maybe tought more information would be better than less. <nckx>I think that's exactly it. <roptat>ok java-jakarta-oro had the exact same issue, fixed too :) <nckx>We'll hit 100% by the end of the night. <Noclip>Can a channel be also a fs-path or does it has to be a url? <nckx>Noclip: Both. A file:///… URL. <nckx>It must be a git repository, and don't forget to commit before pulling. Guix will only see committed changes. <Noclip>The manual says: "The difference between --roll-back and --switch-generation=-1 is that --switch-generation will not make a zeroth generation, so if a specified generation does not exist, the current generation will not be changed." <Noclip>But when I typed -S 0 or -S -1 it just took my to generation zero. <apteryx>mbakke: weird; plotutils fails to be built on aarch64-linux with for sole error in its log file: while setting up the build environment: executing `/gnu/store/x3gq648qnfnla7nppyfjvj62s2i8y7rl-guile-3.0.2/bin/guile': No such file or directory <apteryx>builder for `/gnu/store/qcilzch61vsm9y4c3k4vbrl89g0fn8a3-plotutils-2.6.tar.xz.drv' failed with exit code 1 <apteryx>build of /gnu/store/qcilzch61vsm9y4c3k4vbrl89g0fn8a3-plotutils-2.6.tar.xz.drv failed <apteryx>seems to be failing to build the source derivation <mbakke>apteryx: are you using qemu-binfmt? I have seen a similar error and had to restart guix-daemon IIRC. <apteryx>the one which is building using qemu-binfmt <apteryx>by the way, it's super cool that 'guix environment' is able to emulate other architectures using --system too! <apteryx>the file that doesn't exist seems to be the builder: /gnu/store/2hibdxsj9f0zx3zilwh8phk1knk6y2fl-plotutils-2.6.tar.xz-builder <Noclip>"Both. A file:///… URL." nckx I just noticed now that filesystem path can also be urls. Good to know! <nckx>apteryx: So it really doesn't exist? That's different from what I've seen with the same message. <apteryx>in that case yes, and retrying it succeeded (!) <apteryx>in another case (lib2geom) with the same error message, the derivation didn't even existed on the offload machine <apteryx>I guix copy'd it, and ran guix build on the .drv manually, and it's building now. <apteryx>Seems like some interaction between offloading and binfmt-qemu is broken <apteryx>nckx: I've also opened issue #42569 with some information therein that you could use to try to reproduce if you are so inclined. <apteryx>I somehow doubt it to be reproducible :-/. <nckx>Mine was not, and never came back :-/ ♥ that feeling when you put on your bug-huntin' suit and the bug just flies out of the open window. <nckx>But I'm pretty sure it's not dead so thanks. <apteryx>and the strangest thing is that after building the very same derivation on the remote (after guix copy'ing it), it still thinks that it needs to build it and fails the same on my local machine... <apteryx>oh, restarting qemu-binfmt seems to do it, locally. It was not related to offlodaing <nckx>Oh, right: brettgilio: I do not blog, have never blogged, and you cannot prove otherwise. <nckx>I made damn sure of that. <nckx>brettgilio: That's actually a cool genre of introduction I wouldn't mind reading more often. Thanks! <apteryx>ah ah, my machine file didn't contain an aarch64-linux entry for the offload machine, so of course it wouldn't have to do with it. <nckx>But wait, I thought you said it *was* offloaded. <apteryx>I thought it was, but I guess it wasn't. <nckx>(Not ‘related’ but done.) <apteryx>so the problem was a local only problem, and had nothing to do with files not existing (of course they wouldn't exist on the remote since it wasn't offloading there... :-o) <apteryx>it probably looked exactly like the one you had before and couldn't reproduce <nckx>…/bin/guile: No such file or directory. <nckx>But it did exist, was ARM, and did run (no bogus linker nonsense). <nckx>Anyway, not your bug, it seems. <apteryx>I checked the guile binary, it ran fine. <Noclip>How can one export/import the profile? <nckx>apteryx: Through transparent qemu-binfmt? <apteryx>Noclip: I think you can 'guix export path/to/your/etc/profile', no? <apteryx>nckx: yes, I setup a 'guix environment -s aarch64-linux lib2geom' then ran that exact Guile binary with 'guile --version' <nckx>Depends on what exactly you mean by export. ‘guix export’ isn't a thing though. <apteryx>ah, I must have been thinking of 'guix copy' <bavier[m]1>oh, our aarch64 kernel config builds the Panfrost module! Have we had success running on mali-g** hardware? <nckx>apteryx: Right. I didn't use an environment but sounds similar. <apteryx>Noclip try: guix archive --export $(realpath ~/.guix-profile) <apteryx>so, not the etc/profile file, but the profile root itself. <nckx>…and pipe it somewhere 😉 <nckx>TIL realpath. Although I don't see what it offers over readlink -m. <Noclip>How about those manifest files inside the profile? <nckx>An implementation detail. They can't be used as ‘manifests’ by guix. <Noclip>So a profile is only the content of one generation and not the sum of all your "active" generations? <nckx>Hence the ‘diff-line’ -/+ output of ‘guix package --list-generations’. <nckx>Sorry for all the typoes, I'm exhausted. *brettgilio zzz goes to bed <apteryx>seems not only the tests expect exact numerical precision <peanutbutterandc>nckx, Please forgive me for the delay. I have been away. The bot just showed me your message. Yes, sir, the patch does make sense. But I am afraid I do not have any idea regarding the other question... I should perhaps look into it... <Noclip>cyberwolf: Awesome, finally a popular linux YTer makes a video about Guix! And it looks like he might make another video about it in the future. Thanks for sharing that cyberwolf! <peanutbutterandc>mroh, whoa is this invidious thing a front end to youtube? o.O Looks neat... <peanutbutterandc>mroh, Thank you. This invidious thing is super cool! And for the video too! <peanutbutterandc>mroh, Also, is there any chance that you might know how the guile function (trace-calls-to-procedure) from (system vm trace) is supposed to work? I tried (trace-calls-to-procedure reverse) but (rever 'asdf) does not produce a trace <mroh>peanutbutterandc: does it work with other calls? you can also ,trace in the repl maybe that helps? <peanutbutterandc>mroh, ,trace gives a rather ugly output (the procedure depends on an initialization file supplied with a book I'm reading) of all the other instructions... It doesn't work with other calls, no. But I'll try one with factorials right away <peanutbutterandc>I was hoping (trace-calls-to-procedure fac) and then (fac 20) would produce a trace much like ,trace (fac 20) but no <peanutbutterandc>Noclip, Thank you. Is there any chance you might happen to know about the correct usage of (trace-calls-to-procedure) in guile? (from (system vm trace)? <Noclip>peanutbutterandc: No sorry, I have no idea. <nefix>so, yesterday I asked a question but was too tired and went to sleep xD. How can I read a file and get its content as a list of strings? I've found (local-file), but it returns a file type and I couldn't figure the way to make it a list of strings <Noclip>How does a guix pypi-uri look like? Can you give an example? <Noclip>Did that already, but didn't notice that it actually printed the url in the process of downloading the package. <Noclip>(It exactly the one I had assumed above.) <Noclip>Is there a better way to "refresh" packages without messing with files than "guix package/build --with-source"? ("guix refresh" seems to require messing with files and doesn't support all sources.) <Noclip>Strange, --with-sources works with build but not with package. <efraim>for some of the updaters 'guix refresh -u' will bump the version and the hash <Noclip>The following works: Use "guix build --with-source=PACKAGE@VERSION=SOURCE PACKAGE". If it builds successfully it will print out the path where the new package is located in the guix store. Copy that path and run "guix install PATH". <Noclip>So in case you want bleeding edge glances 3.1.4.1 instead of the "old" 3.1.4 (don't take this serious xD) you can run <peanutbutterandc>civodul, Mr. Courtes, hello there. I have just been advised by someone to file a possible bug report regarding guile. But I am not really sure if it is a bug. May I message the issue to you sir? <civodul>peanutbutterandc: please email bug-guile directly <peanutbutterandc>civodul, I see. Well you are the maintainer. I am quite positive I am doing something stupid and that this is not a bug <civodul>(i'm co-maintainer, with wingo doing most of the work) <civodul>also it could be that people on #guile could provide guidance <civodul>don't hesitate to ask there if in doubt *brettgilio is going to refer to Ludo as "sir" from here on out <peanutbutterandc>civodul, I am currently writing an email. I hope this won't cause too much trouble. I am sure it is some foolishness on my part. :) <peanutbutterandc>Haha. I come from a culture that does not refer to seniors by name. And since everyone here is a senior (i'm a n00b, still), I tend to default to 'sir' or Mr. LastName. :) *jonsger was greated today by the Guix website in German :) <Noclip>How long does it take to build ghc? I'm currently building ghc@8.10.1 xD <Noclip>It was supposed to be a short test but it already runs for a few minutes now. <Noclip>There were some errors but it's still running. <civodul>Noclip: i think it takes quite a while <Noclip>Looks like it will soon hit the max /tmp size. (/tmp is a 2G ram-disk.) <NieDzejkob>I tried building './pre-inst-env guix build -v1 --system=armhf-linux hello --no-grafts -K' on core-updates with a bumped binutils, and got 'output (`/gnu/store/06him7a803772zam2xwx830ar7c6nzqk-glibc-2.31-debug') is not allowed to refer to path `/gnu/store/h8yl12yx7nh5yiln35djzrasyakvyjhm-glibc-bootstrap-0'. I tried the same on pure c-u, but every time the build hangs at a random place and processes like sed, bash, xgcc or as rack up hours of CPU time ***lxbarbosa is now known as elxbarbosa
<civodul>Noclip: yes, these are fast machines <civodul>NieDzejkob: that's with qemu-binfmt? <Noclip>It hit the 2G cap and died instantly. ☹️ <Noclip>Looks like ramdisks aren't a good coice for compiling. <civodul>NieDzejkob: qemu-arm is extremely slow, so i wouldn't try building all this on it <civodul>now, the "not allowed" error shows a real issue <civodul>presumably the same should happen on other architectures, no? <civodul>oh maybe not due to the different bootstrap paths <efraim>649191c3e956a1381d58e7d37e8c37c485e5b684 on wip-ppc says FIXME: allow reference to bootstrap-binaries in binutils-final <efraim>NieDzejkob: which package were you building when that came up? <nefix>I know I've asked the same question already twice, but I would really like to avoid having configuration files embedded inside Scheme code <nefix>how can I read a file and get its content as a list of strings? I've found (local-file), but it returns a file type and I couldn't figure the way to make it a list of strings <efraim>it may be we don't need a debug output on that package <civodul>nefix: it depends, but there are many places that accept a "file-like object" <civodul>in those places, you can use (local-file "my/local/file.conf") <jonsger>civodul, NieDzejkob: maybe I should send my findings on proper ARM build machines to the mailing list. A proper machine definitly could help... <civodul>note that the build farm does have real ARM machines, just too few of them <nefix>civodul: the thing is that the function modifies the content, so it can't be just a load-file (I guess) <civodul>nefix: oh i see, this one expects a string <civodul>it could still accept a file and later concatenate them <civodul>but yeah, you can read the file contents in the meantime <civodul>like so: (call-with-input-file file get-string-all) <civodul>and first: (use-modules (rnrs io ports)) <nefix>civodul: so, (call-with-input-file ...) would return the list of strings? <civodul>nefix: no, it would return the file content as a single string <nefix>civodul: with the newlines, right? <nefix>is there any way to read the scheme definitions docs without reading the source code? <efraim>i've started building out to hello on core-updates on aarch64 <mroh>nefix: remove the local-file call: (call-with-input-file "init.vm" get-string-all). you can read documentation without src: `info guile` <NieDzejkob_>mroh: how would I use info to read the docs for suubstitute*? <nefix>mroh: thanks! Now I'm getting an unbound variable :/ <nefix>mroh: fixed it with ` and , :D *kmicu still awaits the flood of new users but currently only saw flood of links to the same video 🤔 <mroh>NieDzejkob_: oh, im sorry, I thought he wants doc about scheme modules <nefix>mroh: that's what I was looking for, yep! Also, it's not he, it's they :) <PurpleSym>What’s the reason for using origin/keyring instead of putting Guix signing keys into master as well? <kmicu>civodul: “Guix Is An Advanced GNU Operating System For Freedom Lovers” on YT. <kmicu>The one thing I like in it is that graphical installation process just works. It feels like a mainstream distro 😺 <Noclip>I'm excited for the follow up video (I hope he does one) after he has tested guix for around a week. He didn't seem to know that much about guix yet so the next video might be much more informative and less "let's get a first impression". <kmicu>Judging a distro based on an installation process or a desktop environment is mainstream. *kmicu forks Guix and rewrittes it in OCaml to make it more elite. <rovanion>Is there a way to disable te postage calculator addon in icecat? I don't live in the US and don't see the point in it. <nckx>rovanion: Menu → Add-ons → disable. <nckx>(Under the ‘…’ button for each add-on.) <nckx>jonsger: Which recent discussions about our Softiron OverDrive in IRC? What was the conclusion? <rovanion>nckx: I already have it disabled, I think... It's the one labeled LibreJS/USPS compatability right? <efraim>we decided we needed more build machines for aarch64/armhf <nckx>That I don't know, I rip 'em out at the source. That ☝ sounds like it just makes the Web site work with LibreJS though. <nckx>rovanion: I think you'll find a separate calculator one if that's the case. <rovanion>Problem is that I don't :/ Seems like a strange thing to be built in to begin with, as if someones personal preferences got included. <nckx>rovanion: It's not built in, strange you can't find it though. <rovanion>When i rightlclick the icon and press Manage Extension I'm directed to the addod named "Workarounds for nonfree JS". <nckx>…that provides a postage calculator? Wow. <nckx>3 of the 5 aarch64 build machines are commented out. Nobody pinged me :-/ <kmicu>BREAKING NEWZ: A prominent GNU delevloper calls arbitrary bundle of gnu-extensions ‘that nonsense’. Outrage on social media. *nckx checks under their bed for prominent GNU developers… nope, just me. *kmicu patiently awaits GuixFox—a FSDG following Firefox. <nckx>That's exactly what I wish IceCat to be. But they put in the (tremendous) effort, they get to choose. 🙂 By the way, there are more phases patching out Firefox defaults that I dislike than IceCat defaults (only that one). <rovanion>Do you keep a package list for your user profile separate from the system profile? I think I would like to carry around such a package list between all the different machines I use, both those that run Guix System and those that does not. <kmicu>Personally I seperate them, but that’s nothing more than my preference to keep things modular and split them on an arbitrary line defined by me. <nckx>rovanion: Yes, in a manifest (see the manual). <kmicu>More maintanence but easier to keep system essentials up‑to‑date. <nckx>☝ and much faster. No need to rebuild GuixFox just to update a service configuration option. <nckx>(One could stay on an old commit, but that has its own risks.) <nckx>rovanion: Or locally, ‘info guix’. (Other fine info readers are available, but I don't know of any that use fancier pixels.) <nckx>‘Help’( us!) can be read as a synonym of ‘Contribute’(!), I can see the ambiguity. Oh English. <nckx>‘Get Help’? Is that too colloquial? <rovanion>Ah, I literally did not understand that kmicu's comment was a pointer towards navigating the website. <rovanion>If the help menu contained a dropdown with entries 1. Manual 2. Videos ... and so on the intent would be clearer when navigating the site by mouse. But as I navigate by keyboard I would most like if the words "documentation" or "manual" appeared as links on the front page, but that is a smaller audience than the mouse driven audience. I note that nixos.org uses "Learn" for the same semantic as "Help" <nckx>I don't want this to turn into a bike-shed poll but I like ‘Learn’. <nckx>I like that it's broader than ‘Documentation’ or ‘Manual’ and presents them as one part of a community support package. <nckx>(My bias is that I loathe collapsing navbars into ≡, so anything that keeps it from expanding further is good.) <jherrlin>Hey, trying to install Guix System via the graphical wizard. When selecting an option in the partition view the installer restarts and jumps back to locale view. Anyone know about this issue? <jherrlin>nckx: It's NVME disks in that machine! Thank you! I used the stable one, will try the latest! <jonsger>nckx: recent was when I initially wrote the mail, some weeks/months ago. There wasn't a real conclusion *raghavgururajan says "When it comes to Guix, the line between stable and latest is blurry." <civodul>kmicu: found it, thanks! (yay straw-viewer) <nckx>civodul: I was wondering whether we should deprecate youtube-viewer. I don't think it's hard-non-FSDG, but it can only be used by users who create an API key that can be (and is, and whose main intention is to be) used for tracking. <nckx>Maybe not deprecate but at least point out straw-viewer when someone simply searches ‘youtube’. <civodul>nckx: yeah to me youtube-viewer has become useless <civodul>now, it's still FSDG-compliant strictly speaking, only not privacy-respecting <civodul>the throughput in the video is terrible :-) <Noclip>"@ unsupported-platform /gnu/store/nymrf7yj208cavl5fs4x4h3glwbawr5y-coreutils-minimal-8.30.drv armhf-linux <Noclip>while setting up the build environment: a `armhf-linux' is required to build `/gnu/store/nymrf7yj208cavl5fs4x4h3glwbawr5y-coreutils-minimal-8.30.drv', but I am a `x86_64-linux'" <Noclip>I thought that would use qemu for emulation. Do I need to install qemu manually to make it work? <Noclip>"Judging a distro based on an installation process or a desktop environment is mainstream." <apteryx>Noclip: do you have the qemu-binfmt service running? With the armhf-linux platform configured? <apteryx>Noclip: look in the info manual under node 14.6 Submitting Patches <Noclip>kmicu: That's okay for a linux distro that aims to be very user-friendly and easy to use out of the box but I don't think that's a main goal for guix. At least not yet. <kmicu>Guix brings quality computing to the people. Guix can be user friendly and that’s why graphical installer exists, translations, i18n, only a11y is lacking, but soon Gnome should be in a good shape too. <raghavgururajan>nckx,civodul: Does YT require us to have unique API-Key for each youtube-viewer installation? If is not, then we can create an API-Key for just Guix, that is compiled for all installations of youtube-viewer. In this way, the key is not unique to a particular user, so no tracking issues. <nckx>raghavgururajan: Yes to your lead question. <kmicu>(and that’s not exclusive thing, thanks to Guix design principles Guix can be user‑friendly and user‑unfriendly at the same time on the same computer) <hendursa1>"Guix is user-friendly, it's just choosey about its users!" *kmicu uses ‘mpv ytdl://ytsearch3:"gnu guix freedom loving"’ [jokin’] ’cuz that’s not user friendly and elite. <branjam>Thought I'd be clever and and bring out the ol' roadmap... <branjam>but sure enough "user-friendliness" was not a post-1.0 goal <nckx>raghavgururajan: As long as that's allowed by their ToS I think it's fine by ours 🙂 <nckx>It allows tracking ‘Guix users’ as a whole, akin to adding /Ubuntu to a browser's User-Agent by default. The latter is not something we'd ever do but I don't think it's *prohibited* if it's required… <nckx>‘I think’ being the operative word here. <nckx>raghavgururajan: Oh, I don't. <nckx>That's why I use it, it's a nice little ‘CLI’ programme that launches MPV. At no point in time do I see a single button 😉 <raghavgururajan>> nckx: It allows tracking ‘Guix users’ as a whole, akin to adding /Ubuntu to a browser's User-Agent by default. The latter is not something we'd ever do but I don't think it's *prohibited* if it's required… <nckx>No, I meant your OWM key. <Noclip>Is Electron compatible with the FSDG? <nckx>Everyone who uses that key is obviously a Guix user (or, fine, someone who copied your key). <jherrlin>nckx: latest release worked fine and my system is up and running :) <nckx>Pff. Outfancied by kmicu. <nckx>Using Unicode is cheating. <raghavgururajan>nckx: I don't think peresonalised tracking is possible here. Even if OWM attempts to track, then they are tracking me. <nckx>raghavgururajan: Oh. Our straw-viewer package installs a gtk-straw-viewer.desktop but it refers to a ‘gtk-straw-viewer’ that we don't build or install. I didn't know there was a GUI. <nckx>raghavgururajan: Not personalised, just (like a ‘/Ubuntu’ or whatever browser user agent) collecting a tiny bit of info. <raghavgururajan>And I will be seens as being at *N* number of places at the same time. <nckx>You're getting anonymised by the vast mass of Guix users 😉 <nckx>Where we go one we go legion, forget us. <raghavgururajan>nckx: I was thinking about creating an account with email guix-maintainers@gnu.org? WDYT? <raghavgururajan>> nckx: raghavgururajan: Not personalised, just (like a ‘/Ubuntu’ or whatever browser user agent) collecting a tiny bit of info. <nckx>raghavgururajan: I don't think ‘we’ care about who owns the account. If you'd go rogue or win the lottery we could just replace the key. <nckx>raghavgururajan: Like that, but in this case ‘uses Guix, or is at least very likely to, since they use the default Guix key’. <raghavgururajan>That is unavoidable on any internet services, unless Tor comes to the rescue. <nckx>By modern standards that's benign but I just wanted to point it out. <nckx>Well that's not true, if I contact your services without providing an API key you have less info. You still get my IP but you don't know my distro. <kmicu>^^ that’s not a libre issue tho so not a stopper from an inclusion into Guix. <raghavgururajan>> nckx: raghavgururajan: I don't think ‘we’ care about who owns the account. If you'd go rogue or win the lottery we could just replace the key. <raghavgururajan>Haha! Btw, as many api-keys can be generated per account. I currently have two: one is for my personal use and one for guix. <nckx>kmicu: We're not obligated to ship anything simply because we can, either. <kmicu>Oh sure but if you plan to block some *libre* FSDG following software then you should have a clear and transparent policy about it. <kmicu>It’s ok to exclude non ethical software, not privacy respecting software and so on but Guix should be clear about it. <raghavgururajan>nckx: Regarding straw-viewer, yeah, I just noticed gtk-straw-viewer is missing under /bin <nckx>raghavgururajan: Broken desktop files are extremely obvious if you use them, but many people (and packagers) never do, so thanks for reporting that. I wasn't aware of the gtk- variant at all. <Noclip>Can guix system boot directly into different generations? <Noclip>"do you have the qemu-binfmt service running? With the armhf-linux platform configured?" <Noclip>apteryx I'm not using Guix system, only the Guix package manager so I assume I can't cross compile? <nckx>Noclip: You can cross-compile (--target=) but not simulate a native build (--system=) and the identical hashes that come with it. *nckx lol, claims not to know about a GTK front-end, wrote ‘Add #:module-build-flags '("--gtk") dependencies’ when packaging s-v in 2019. *raghavgururajan is listening to his favourite album 'Nothing But The Beat'. <nckx>I hope there's a bit more. <nckx>raghavgururajan: OK, it's coming back to me now. The gtk- version requires packaging perl-gtk3, that was a pain then, I didn't care, so I gave up. Feel free to add it yourself but I won't, sorry. <nckx>& sorry for being so demented that I forgot. <apteryx>Noclip: I think mothacehe had a way to make --system work on foreign distros <apteryx>Supposed to document how at some point :-) if you see them around here, feel free to ask, interested users always provide a motivation boost ;-) <Noclip>apteryx: I assume one could also manually set up a guix vm with qemu. <nckx>Or use ‘sneek: later ask mothacehe foo’ to leave a message. <nckx>Although that works only if they talk and they don't seem very chatty. <Noclip>apteryx: That thread doesn't really sound like he had any success with -s on other distros. <apteryx>Mathieu had success on Ubuntu 19.04 using the static versino of qemu-user, or on 18.04 when enabling the F flag for all binfmt installed entries <apteryx>I think it's also necessary to provide the --chroot-directory arguments to the guix-daemon so that the chroot can access the needed QEMU components. <apteryx>in other words, it's much simpler on a Guix System ;-) <Noclip>"I assume one could also manually set up a guix vm with qemu." <Noclip>Installing alpine linux in the vm and then guix on alpine linux might be a better idea as there is no offical install image for arm. <Noclip>And I assume Guix System is needed to create an arm image, right? <apteryx>Noclip: so your host machine is ARM based? <NieDzejkob>Noclip: I use native build emulation on my Debian machine, it's not impossible <Noclip>I'm just trying to understand guix and play around with it. <apteryx>OK, sorry, I got lost somewhere :-). If you are on a x86_64 machine, you can install Guix System on it, then use --system=armhf-linux or --system=aarch64-linux provided you configured the qemu-binfmt service provided on the Guix System. <Noclip>Can "guix system" be used on non-guix-system installs? <Noclip>"And I assume Guix System is needed to create an arm image, right?" <Noclip>Seems like the answer has to be yes as "guix system disk-image --system=armhf-linux -e '((@ (gnu system install) os-with-u-boot) (@ (gnu system install) installation-os) "A20-OLinuXino-Lime2")'" (manual 3.10) also uses --system. <raghavgururajan>> nckx: raghavgururajan: OK, it's coming back to me now. The gtk- version requires packaging perl-gtk3, that was a pain then, I didn't care, so I gave up. Feel free to add it yourself but I won't, sorry. <NieDzejkob>Hmm, now I'm confused as to what I did to make Guix put the QEMU binary in the chroot <str1ngs>NieDzejkob: are you using qemu-binfmt-service-type ? <Noclip>Nixpkgs and Guix channels are much simpler to maintain than traditional repos, right? <NieDzejkob>str1ngs: that's the thing, it's a Debian machine. I have the binfmt packages installed, but there's no evidence of me modifying guix-daemon.service to add it to the chroot or anything <branjam>finally understand what the manual meant wrt guix system docker-image <Noclip>Do you think the Guix repo will get also really big at some point? <str1ngs>NieDzejkob: if you have qemu-user-binfmt it will populate /proc/sys/fs/binfmt_misc. <str1ngs>NieDzejkob: I think the guix-daemon is just smart about exposing the interpreter <Noclip>A few minutes later: "guix pull" --> Downloads a lot of stuff 😂 <kmicu>Noclip: it depends what do you have in mind. Facebooks’s repo is 800GiB but I think Guix folks are wise enough to stick to text so we should be good for this century. <str1ngs>NieDzejkob: also if you install qemu-user-static binfmt will use the static binaries. which is preferred for the chroot <nckx>kmicu: Facebook's repo of… what, ffs? <nckx>Facial recognition data? <nckx>branjam: Suggestions to improve wording/clarity are always very welcome on one of the mailing lists. <kmicu>Monorepos, microservices, you know, the fashionable stuff today. <Noclip>kmicu: I mean the amount of included packages / projects. <kmicu>Nix and Nixpkgs are in seperate repos, Guix and Guix Pacakges sit in one… monorepo. <kmicu>[conspiracy theory] if Guix sits in a monorepo, then Guix is a corporation and Ludo is our CEO. <nckx>kmicu: Only if there's a gold fringe around the flag^Wlogo. <nckx>Hm, our logo basically is a gold fringe. Not sure what that means. <kmicu>It’s an illuminati symbolism. Horns of the beast. *kmicu went reading The Da Ludo Code; a 2020 mystery thriller novel by Dan Brown in which the central plot is that Guix is a corporation setup to increase compilation time via mass rebuilds and increase global temperature to create an environment friendly to reptilians. <dustyweb>does someone mind walking me through the current way to do git-version stuff? <dustyweb>for the (git-version version revision commit) <dustyweb>should version be "the last version released" or "the next version *to* be released"? <str1ngs>dustyweb: personally I like to use git describe for version. see the emacsy-minimal package. <str1ngs>dustyweb: you can use git describe to get release revision and git here hash. <dustyweb>fatal: No annotated tags can describe '1b26af2c1a701481ac5d90928fe0200e389756c3'. <dustyweb>However, there were unannotated tags: try --tags. <apteryx>dustyweb: the last version used, typically (that's what git describe would do, to, use the last tag as a base). <str1ngs>dustyweb: with the v prefix for version you'll need (version (string-drop commit 1)) . see the emacsy-minimal package. <nckx>And with that, I'm leaving work. o/ <apteryx>that's a refreshingly high level of ridiculous, thanks for the laugh <apteryx>civodul: ah, those mirror-list and hook packages... ;-) <civodul>kmicu: uh if Gephi is Java, it's prolly not a challenge for me :-) <ternary>Does anyone know how to get GlusterFS working? I've been fighting it for a bit, but I assume somebody has since it's part of the default channel. As far as I can tell, glusterd needs a patch because it has datadir + /var/lib that it tries to write to, and that's obviously read-only with guix <nefix>how can I use a local copy of guix as the channel? If I don't add %default-channels it complains, but if I add it (with my guix channel, it doesnt end the guix pull) <apteryx>trying to use gnome-boxes: GLib-GIO-ERROR **: 11:54:59.810: Settings schema 'org.gnome.desktop.input-sources' is not installed <apteryx>raghavgururajan: ^^ perhaps you'd have an idea? <apteryx>I'm guessing it needs the gsettings-desktop-schemas package ***terpri_ is now known as terpri
<apteryx>should this be propagated, or is there a nicer way to add support for it without propagationÉ <nckx>apteryx: That should not be necessary. Try adding it as a regular input. <raghavgururajan>apteryx, if you do not have gnome-desktop-service (which adds g-d-s to profile), then you have to install it in your profile. <kmicu>nckx: you have some mad gimp skills, that graphic looks like a real thing from a fringe subreddit. I WANT TO BELIEVE. <nckx>raghavgururajan: I don't have either. I hope we can make it work without that, as gitg, terminator, gthumb, evolution, … seem to do. Just looking at inputs, didn't test. <nckx>I should be explicit: ‘I don't have either’ → and it works fine, albeit without icons, in a --pure environment. <apteryx>ah... then I must be doing something wrong, like not resourcing my profile after installing gnome-boxes <nckx>Or I have some unidentified pollutant. <apteryx>(after sourcing ~/.guix-profile/etc/profile) <apteryx>I'm using a minimal desktop operating system. <apteryx>the problem is when adding a virtual machine, and trying to select the installation image *nckx adds professional watermark before kmicu shares it on 4chan. <nckx>apteryx: Ah, I'll try that. <nckx>That'll probably fail, solving the mystery 🙂 <apteryx>eh, I hope so (?). I don't like mysteries much. *NieDzejkob just realized that their core-updates patch doesn't actually do anything because defaults for build-system arguments are defined in 3 separate places <apteryx>nckx: could reproduce it in a pure env <nckx>Sloowly downloading the Guix iso… <nckx>apteryx: At what point, exactly, does it fail? I get as far as the (optional) ‘Customise’ window but can't ‘Create’ without installing (I presume) the libvirt service. <apteryx>nckx: top left '+' -> 'Create a Virtual Machine...' -> Click on the 'Operating System Image File' button -> select file, OK -> Boom <nckx>guix environment --pure --ad-hoc gnome-boxes -- gnome-boxes <apteryx>Could it be that something on your system propagates gsettings-desktop-schemas? <nckx>I'm almost certain of it. <apteryx>are you using GDM for example? I'm not. <apteryx>GDM pulls lots of GNOME components, so I wouldn't be suprised that'd be one of them. <nckx>No, slim → .xsession → i3. <nckx>But I use simple-scan for example. Maybe that's how it sneeks in. <nckx>Oh, scratch that, it's not installed yet. *nckx keeps eyeing the list suspiciously. <apteryx>the funny thing is that `guix gc -R $(realpath ~/.guix-profile) | grep gsettings` returns gsettings-desktop-schemas! *apteryx has a bright idea: reboot *raghavgururajan hasn't got to gnome-core yet. Still at gnome-core-deps. <apteryx>raghavgururajan: yes, when doing the steps described above <apteryx>after building it with g-d-s as inputs, it works as expected <nefix>can you change the python command to point python3? <nckx>Sounds like apteryx already did. <nckx>nefix: In packages? ("python" ,python-wrapper) <nckx>I presume you can install it too, but never tried. <apteryx>nckx: do you think adding gnome-desktop-schemas as an input is a valid fix? <nckx>Oh, wait. The danger of abbreviations. g-d-s to me meant gsettings-desktop-schemas. <nckx>Otherwise: yes, I do, although I'm no GNOME expert. <nckx>I like some of its applications but don't use The Platform. <nefix>I get a weird exception when trying to import concurrent.futures in python3. It should import it's own library (the one from python3), but instead it tries to import the backport for 2.7, which has another syntax and makes it break. Does anybody have the same issue? Or am I doing something wrong? <nckx>Yeah, futures means ‘Python 3 stuff for Python 2’. <nefix>nckx: I've found the issue: ttps://dpaste.com/5PJNDBGCQ <nckx>Erm… I thought you were writing a Python 3 package (hence the -wrapper), I see it's the opposite. <nckx>I don't understand the connexion between your Python 3 questions (like why it should import its own [Python 3] futures) and the paste you posted. <nefix>the thing is that the trollius package requires the futures pacakge, and thus it adds it to the python3 library scope (or something like that), then python3 breaks because of syntax error <nckx>Why is python2-trollius in your Python 3 development environment? Or are you writing a package (same question then, though)? <nefix>I'm trying to add the package <nefix>I see that all the other packages that use python2-futures are build with package-with-python2 <apteryx>nckx: to find the package that already pulls gsettings-desktop-schemas in my profile: guix package -I | cut -f1 | xargs -I@ guix graph --path @ gsettings-desktop-schemas 2> /dev/null | grep -v 'gsettings-desktop-schemas' <nefix>but python2-trollius is a package exclusive for python2, so it doesn't really make sense to write a pacakge and then wrap <nefix>but then, if I install another package with python2-futures, it doesn't break python3 <nckx>Which Python 3 package did it break? <nefix>which is bundled by default by python3 <nckx>Seriously: evince, inkscape, vinagre, gimp, qemu, … <str1ngs>hopefully my pinebook pro wont catch fire building guix temp1: +67.8°C 😬 <apteryx>Which makes it weird that it only finds it for me when it gets build with it <nckx>Luckily mostly obsoleted by Guix for me. <nefix>I might have fixed it, let me see... <apteryx>so, the gnome-boxes requires a libvirt service to run? <nckx>nefix: I'm not a Python programmer so am not going to make general statements, but (proceeds to make general statement) it seems to be as simple as ‘$PYTHONPATH isn't versioned, also kinda clobbers itself?’: https://paste.debian.net/plain/1158103 <nckx>I forget what the other one's called but they're documented in the same chapter. <nefix>nckx: I have no idea what are you talking about, I know python but not that much :) <nckx>Then we're in the same boat. <nefix>what I don't get is that packages such as python2-promise do use python2-futures, but they don't break everything <nefix>So I've wrapped everything with (package-with-python2), but it's still building <leoprikler>if they have the same prefix (either all python or all python2) of course they won't hurt each other <NieDzejkob>perhaps we should patch Python to introduce GUIX_PYTHONPATH{2,3} <nckx>I know that's been proposed before. <nefix>so, wrapping everything with (package-with-python2) doesn't work <nefix>now I don't understand anything <nefix>so, with python2-promise installed everything works as normal <nefix>but if I install python2-trollius it doesn't <nefix>and right now looks like this: <leoprikler>could there a problem with python2-trollius itself? <nefix>leoprikler: yeah, it's the package that I'm trying to build, but failing miserably xD <leoprikler>I think the outer package-with-python2 is a bit overkill <nefix>leoprikler: yeah, it's because all the other working packages that use python2-futures use it, so I figured that it might do something <nefix>also, all the packages have (strip-python2-variant), let's see if it works... <leoprikler>okay, so your trollius package works (without wrapping) in an environment, that has python2 <nefix>the thing is that it then breaks python3's asyncio <leoprikler>you mean in an environment, that has python and python2-trollius <nefix>I have python3 installed (with guix install), also python2 and a package that depends on python2-trollius <nefix>so if I install the package that depends on python2-trollius, it breaks the python3 installation <leoprikler>okay, this really isn't a problem with your package description then <nefix>yes, because if you take python2-promise, it also uses python2-futures and it doesn't break anything <leoprikler>that doesn't sound like a solid argument for anything, really <nefix>that depends on python-pynvim, which has an extra dependency in the python2 version, which is python2-trollius <raghavgururajan>nckx: Not sure if it means 'where to install respective files' or 'where to find respective files'. <leoprikler>The thing is, that guix doesn't really like the fact, that you're trying to mix python2 and python3 in the same "environment" <leoprikler>try to extract neovim plus plugins into a separate environment and run it from there <nefix>I'm not sure what are you suggesting <nckx>raghavgururajan: It's a mix: systemdsystemunitdir is almost certainly a destination, while system_ca_path is probably used at run-time, while iptables is a path to a binary, not directory, etc. <nefix>the thing is that neovim requires both python2 and python3 :/ <nckx>raghavgururajan: I don't think that header is meant to convey anything beyond ‘these are file-namey things’. <nckx>The other things are not. That's about it. <nckx>raghavgururajan: Most build systems don't make a clear distinction between search and destination paths, e.g. autoconf's libdir is often both where libraries are sought and where they are put. <nckx>brettgilio: You get to spam once for each hemisphere. Sounds fair. <nefix>I think I'm going to give up :/// <nefix>I'm really frustrated because it doesn't make sense <nckx>The short story is: PYTHONPATH {sh,c}ould have been PYTHON2PATH and PYTHON3PATH and this probably would have worked. But it's one variable used by both, while the packages aren't in fact compatible, so breakage. <nefix>why would python2-promise work and python2-trollium don't? <nefix>both use the same library that creates the error <nckx>nefix: guix environment --pure python2-promise python-flufl-i18n → echo $PYTHONPATH still gives me only 2.7, by the way, 3.8 is missing. <nckx>So I'm not sure if it really works or just happens to ‘work’. <leoprikler>If you have python, python2 and python2-futures, you'll get this error <leoprikler>python2-promise uses it as native-input, but does not propagate it <nefix>I really thought it was propagated <nefix>yeah, with python2-fasteners it breaks also <nefix>so not my fault, I'm a bit relieved now <nckx>Guixers who care about Python should start lobbying for PYTHON4PATH now. Unless it's actually going to remain compatible this time. <apteryx>I'm trying to use GNOME boxes still, now I have libvirt and virtlog services running, but attempting to create the VM still fails. The message seems to be this one: (gnome-boxes:1734): Boxes-WARNING **: 14:26:50.547: vm-assistant.vala:133: Failed to create machine: Unable to start domain: internal error: qemu unexpectedly closed the monitor: 2020-07-28T18:26:50.341048Z qemu-system-x86_64: -device <apteryx>usb-storage,bus=usb.0,port=2,drive=drive-usb-disk0,id=usb-disk0,removable=off: Could not reopen file: Permission denied. <nefix>anyway, I have to go. ping me if you find a solution or something, the vim plugin (deoplete) is one of the best vim completion plugins out there and I'd love to have it working <nefix>bye and thanks for all the help!! o/ <apteryx>I probably at least add myself to the libvirt group <nckx>bdju: Help updating Anki would be much welcome. <nckx>It's npm work, so I won't be doing it. <bdju>I've never helped with packaging and today is not the day I'll start! <bdju>there is an amount of guilt about it that eats at me, but it's been so long now I've learned to live with it... <nckx>bdju: Are we missing cool new features? <bdju>also, I sometimes have good luck just posting here and getting someone else to update a package, like with minetest a few weeks ago <bdju>well, the second newest update says "this is a big update", but nothing specifically I know is missing <nckx>bdju: Please don't feel guilty, it was a genuine invitation, nothing more. <bdju>ah, thanks for that. I do sometimes worry I'm not contributing even though it's not like I contributed to other distros either. something about guix has a "small town" vibe where it feels like everyone is helping out <jackhill>bdju: I find Guix to be particularly fun to work on. Both the people and the code. However, I'm in this to build a useful tool for others, so I agree with nckx there should be not guild. Use it, it's what it's for. <jackhill>Pointing out things that could be upgaded sounds helpful. At least we know someone is interested in a new version :) Unfortunately, Anki sounds like it could be challening project. <bdju>I got a popup when I launched anki that there was an update, asking if I wanted to download it, but then it just opened the webpage. <bdju>I wonder if that popup could be disabled <bdju>I guess it was useful in a way, but since it's not trivial to update it, it seems like a tease <nckx>bdju: Since that means that Anki is ‘phoning home’, it should be. I'm not saying Anki is malicious, but you're still sending them your IP for no reason. <bdju>oh wow. I didn't think of it like that. <nckx>It's something we disable out of principle. <Tirifto>Hello! Icecat in Guix still has WebRTC disabled, right? <nckx>I'm almost 101% certain it does. <Tirifto>Sounds like you might be entirely sure! :) <nckx>Tirifto: Well, we still pass --disable-webrtc. <nckx>Unless some weird bug causes it to ignore that flag you should be safe/screwed, depending on why you asked. <apteryx>I can recreate my gnome-boxes error with: virsh start debian10-uni <nckx>Do you think the USB part is relevant? Is it possible to disable USB support/pass-through? <apteryx>the idea of gnome-box is you press a button and it works, it seems. No configurability beyond RAM or disk amount. <nckx>Are you in the ‘kvm’ group? <apteryx>I am. And just added myself to the libvirt one too. <nckx>apteryx: Yeah, great if it works, not so much elsewise. <apteryx>ah, I could untick the "express install" option, then I'm shown the Debian graphical installer. <apteryx>I think I'll add note to the description of the package. <Remiii>I can't run binary file, can you help me? <nckx>Remiii: I'm leaving, but please explain your problem in more detail (don't ask to ask, just ask) and perhaps someone will respond. You can also send mail to help-guix at gnu dot org. <pkill9>i've improved my guix launcher some more <bdju>a new youtube-dl version recently came out *raghavgururajan + zopiclone = zzZ <apteryx>how about this gnome-boxes error: dconf-CRITICAL **: 17:11:07.622: unable to create file '/run/user/1000/dconf/user': Permission denied. dconf will not work properly. Any ideas? <apteryx>we don't seem to have a 'dconf-service-type', what service is used in Guix? <apteryx>is it launched automatically by the applications? I have /gnu/store/jxc1wsa5dvcvlhsfklgivws4bs547qwb-dconf-0.34.0/libexec/dconf-service running for example. <apteryx>looking at package definitions, it seems to require to be propagated <leoprikler>interestingly, I have one with exactly the same hash <apteryx>it's often propagated, but sometimes appear in inputs as well... hmm <leoprikler>i.e. gnome-boxes tries to run a different dconf-service <leoprikler>I've had such problems with glib/gtk quite often sadly <adfeno>Hi there, what's the name of the farstream package need for Gajim audio/video calls to work? <adfeno>What's the status of it in Guix? <nckx>It doesn't seem to be packaged (searching for farsight/farstream and looking at gajim itself). <apteryx>leoprikler: I think I had just ran gnome-boxes as sudo once to troubleshoot something, and it created /run/user/1000/dconf/user as root... <leoprikler>you should kill the daemon running as root and delete the file then <adfeno>kmicu: It seems to be the only depency left. now that eu.siacs.conversations (mobile client) supports DTLS-SRTP (encrypted audio/video calls), and no other desktop XMPP client supports it, Gajim could be a way, and I think Guix would benefit from this… <adfeno>… to spread free/libre software values if we can cativate the people who install Guix to escape outdated software (if farstream is really free/libre, which I don't know, although Trisquel has it so I think it's OK) <adfeno>One thing that strikes me as odd though is why Gajim from Guix can't recognize farstream from Trisquel as installed. <adfeno>I guess Gajim is expecting something under /gnu/store <nckx>adfeno: I'm not kmicu and I'd never heard of farstream until five minutes ago, but if it's ‘gstreamer based’ I guess you need to set some environment variables. It's still not guaranteed to work though, as loading dynamic libraries from one distro into processes of another isn't trivial. <adfeno>kmicu: Sorry for mentioning you, I confused nicknames <leoprikler>you could try packaging farstream for guix though <adfeno>nckx: Yeah, my falt on the wrong mention, I meant to direct the message to you. The typing area currently has tiny font size, and I'm trying to set it atleast to 20pt. <nckx>No problem, I thought perhaps you'd discussed this before. <nckx>Packaging farstream for Guix is the winning answer by the way, I'm just looking at it through gitweb but it doesn't look hard. <nckx>All I see is C. That's good. <adfeno>nckx: OK, I'll investigate that. :D <leoprikler>plus all the dependencies except for libnice are already packaged <nckx>I agree that this is important stuff (now more than ever blah blah). <adfeno>Sorry for that introductory message. It was no means meant to be in aggressive tone. <pkill9>how in guile can i get a list of all available packages? 
