IRC channel logs
2023-07-29.log
back to list of logs
<ncd>What are you talking about, the_tubular/trevdev[m]? <nckx>OK, but what do you mean by ‘redirect’ that doesn't break HTTPS? <ncd>What kind of secrets, the_tubular? <ncd>Yeah, you can just keep them in #guix. <nckx>We'll never tell, and the logs are encrypted with HTTPS. <porcupirate>guix, is the 'drop-game-from-paths phase of renpy necessary? It seems to be the source of a bug where the launcher fails to generate a basic project. <the_tubular>Sorry, I'm on and off, meant like API keys and stuff you have to pass to your app, but should not be stored in plain text <the_tubular>Like giving user password through guix system reconfigure <bdju>my root fs went read-only and I had to force shutdown. can anyone assist? luks encrypted btrfs guix system install. I'm in a mint live session at the moment <porcupirate>I just tried running renpy without drop-game-from-paths. Now there's another bug that's preventing the launcher generating a basic project :( <oriansj>just a thought make mirroring source code easier /gnu/src/${package_basename}/${sha256sum}/${source_code}; then one would only need to backup their /gnu/src/ directory to ensure they always have the source code for their programs and it would be easy to identify <zacchae[m]>I tried installing ntfs-3g, but it complains that it can't find mkfs.ntfs <nckx>ntfs.mkfs exists. Who is complaining? <nckx>It also provides ntfsresize, which I've never tested. <zacchae[m]>Ah. I'm in the ISO, and it doesn't add ~/.guix-profile to root's path if you install something until you start a new login session <nckx>/etc/profile's ‘if [ -f "$profile/etc/profile" ]’ is a bit silly on a Guix System. <nckx>(-f instead of -e is merely icky.) <viaken>Is the ~/.guix-profile thing possibly why I can't get an install to work right with extra channels enabled? <ulfvonbelow>I suspect that the "port-filename" error message in issue 64653 may come from a call to 'read'. There don't seem to be any likely invocations within the service or guile-netlink code, and the very first call made in the definition of 'read' is to port-filename <nckx>viaken: I guess it's possible but at first glance not likely, since that's not where Guix lives. <nckx>Depends on what you're doing and what goes wrong (as always). <ulfvonbelow>however I can't find any invocations of 'read' in guile-netlink. Perhaps it's the interpreter itself calling 'read', and the file it's trying to load gets closed somehow? <ulfvonbelow>also, could we modify shepherd so that unhandled exceptions in service constructors/destructors/actions show backtraces? <hab25[m]>[1] Is there guix-specific terminology for this? In the nix ecossystem, they're called "realisations". <next4th>hab25[m]: inputs are mostly input-addressed, while origin (for sources) are content-addressed. <janneke>"So all we need to do is to move the Nix store from an input-addressed model to a content-addressed model" -- yeah, right ;-) <trevdev>Anyone have a working stumpwm with sbcl-stumpwm-notify? Mine can't find sb-rotate-byte from SBCL libs at runtime <tassosm[m]>Does anyone think having something like a state monad for guix packages seems useful? Each package would be parameterized by some state input and then also produce another output containing a modified version of the state. Then a list of these monadic packages could be traversed to produce a monadic list of packages, which could then just be run with an initial state and installed like a list of normal packages. <tassosm[m]>I feel like this would be useful for some build systems that rely on a global state modified during the installation process (maybe racket), but now I'm hearing of other ideas for the racket build system and I'm wondering if this idea is still useful otherwise. Has anyone already explored something like this? Maybe its too inefficient? <tassosm[m]>s/I feel like this would be useful for some build systems that rely on a global state modified during the installation process (maybe racket), but now I'm hearing of other ideas for the racket build system and I'm wondering if this idea is still useful otherwise. Has anyone already explored something like this? Maybe its too inefficient?/I feel like this would be useful for some build systems that rely on a global state modified during the <tassosm[m]>installation process (maybe racket). Now I'm hearing of other ideas for the racket build system, and I'm wondering if this idea is still useful otherwise. Has anyone already explored something like this? Maybe its too inefficient?/ <next4th>tassosm[m]: sound like profile hooks, which take a prebuilt profile (eg: ghc with some haskell libraries) and the generate a new one with its package cache. <next4th>split that whole profile-hooks into package level will bring some advantages like avoid useless computing, yes i think that would be useful. <next4th>maybe it can also be used to get rid of various envirnoment variables we are leaking now by wraping every packages with its closing envs from the 'state' <tassosm[m]>Thanks for the input! I think I'll wrap up my implementation with packages and maybe send a link in the future. <geri>hey, virt-manager on sway ignores my gtk theme even though it's set in gtk3's settings file, gtk2's rc file and with gsettings. nicotine+ and icecat use proper theme/font; setting a theme with GTK_THEME variable does work but it doesn't include font settings so not really an option. does anyone know where else should I look? <geri>there's so much abstraction im not sure if its a sway or guix issue.. <jas>hi guix. i just got a ampere altra 128core arm64 machine, are there any documentation on how to get guix installed on this? <geri>doesn't it have an official installer image? <geri>i think it's called aarch64 instead though <jas>geri: :) i meant a bootable ISO <geri>it's the manager, not the os <geri>probably just cross compiling an image with guix and flashing it onto the drive or something <geri>i remember there was some option to do that <geri>guix system image --target=aarch64-linux-gnu /path/to/config? <geri>128 cores though, that's pretty insane <nckx>You can boot from an aarch64 image, then use ‘guix system init’ to install Guix System on the main drive. That initial image doesn't even need to be Guix, it can be any ‘live CD’, where you first ran guix-install.sh . <nckx>(Only because SystemRescueCD, my standard installer medium, didn't exist for ARM.) <hab25[m]><next4th> "hab25: inputs are mostly input-..." <- thank you <hab25[m]><janneke> "hab25: also, guix uses grafting..." <- thank you <grim`>Hello. I'm following the `(guix) Sending a Patch Series` part of the manual to submit some patches to guix but I'm getting the following error after issuing ``: `fatal: base commit shouldn't be in revision list` <grim`>`git send-email -1 -a --base=auto --to=guix-patches@gnu.org` <nckx>I use format.useAutoBase=whenAble but that's more of a work-around than a solution IIUC. <grim`>I would really like to contribute but I'm a bit lost with this topic to be honest. Never used a mailing list before <nckx>Assuming you've told Git about your e-mail account already, the command above should have simply sent the mail for you. So you were almost there :) <attila_lendvai>mailing lists are becoming archaic for the upcoming generations who are now in queue for becoming contributors... :) <grim`>Thanks for the support guys :) <grim`>nckx: the `format.useAutoBase=whenAble` you suggested where should I put it? <nckx>I *assume* in your case it'll just disable the autoBase, so being equilavent to simply dropping the ‘--base=auto’ from your command above, but at least it will be automated and you can get into the habit of using it unconditionally. <nckx>ACTION not a git wizard, just a cargo connaisseur. <grim`>Seems I'm still getting the same error. This is my .gitconfig: paste.debian.net/1287282 <grim`>Perhaps I have to do something special on the git repo? Like checkout a different branch or something? I don't really understand the error <grim`>fatal: base commit shouldn't be in revision list <grim`>format-patch -o /tmp/qOPBqaolrP -1: command returned error: 128 <grim`>I get the same error even when issuing: `git send-email -1` <nckx>As a last-ditch effort, you could remove the .gitconfig lines again and omit --base=auto from the command. It's a rare error, and the obvious reading of that message (you've specified a base commit that falls within the ‘-1’ range) seems like something Git should be clever enough not to do by itself. <nckx>I've encountered it exactly once, when my local checkout's branch structure was… messy, indeed. <nckx>Or manually specify --base=e43cbeafd1b632f3, no idea if that'll work. <nckx>Let's guess randomly together 🤝 <grim`>Removing the lines does not seem to work. As a sanity check. This are the two commits I made on the master branch: paste.debian.net/1287284 <grim`>I have some other things packaged such as Ulauncher but wanted to start with something easier 🤔 <nckx>origin/master → Is there a public repository I could clone? <nckx>ACTION AFK for 5 minutes. <grim`>Other packages I want to add I have them for now on a `GUIX_PACKAGE_PATH`. I'm afraid that was for xfce had a dependency on Meson 0.63 (if I recall) but Guix updated Meson to an incompatible version 🥹 <nckx>Is there no meson-x.y variable that would work for you? <nckx>ACTION should've known better than to trust ‘do you have a minute’; sorry :) <grim`>Well I think they dropped all versions older than "1.1.0" <geri>so i found out virt-manager is affected by gtk-3.0/settings.ini, but not completely - enabling dark mode works, but fonts and themes are ignored <grim`>And If I remember Meson introduces a change in 0.64 or something <grim`>I will check latter if I can port it to "1.1.0". But first I need to be able to send the patches haha <nckx>Oh, you're right. Commit 311255. Radical. <nckx>Some packages are broken on master because of these upgrades :-/ Mostly the ‘positional arguments’ thing, which is easy to patch, but I wonder how these failures went unnoticed. <grim`>It was a bit shocking to me that they dropped all the meson version <grim`>Anyways. Did you see anything weird on the two commits I have on that codeberg repo? <nckx>I wonder if all dependents were supposed to be tested on CI but the recent CI trouble hid the failures (apteryx?). <grim`>Quite unfortunate. I had to pin the guix channel since the meson change because it broke packages I'm using in `GUIX_PACKAGE_PATH` <nckx>grim`: So I'm really not that familiar with Git's failure modes, but I *think* it's confused by HEAD (the top commit) being origin/master, not being very clever, and assuming that the ‘remote master’ must be what it should use as base. Do you have a different origin tracking Guix upstream? For example, I have upstream/master pointing to Savannah's repository, and --base=upstream/master seems to work. <nckx>Re: meson: I'd copy the deleted meson package into your own package collection instead of holding back all of Guix. Or I'd use an inferior, even though they're slow. <nckx>OK, then --base=upstream/master should work, and if it doesn't my hypothesis is bunk. <attila_lendvai>i'm surprised that git send-email adds a X-Debbugs-Cc: field with several personal email addresses. maybe that's how i'm sometimes receiving those bug emails that i usually don't understand why are also addressed to me directly? <grim`>Weird. I'm getting now this: git send-email --base=upstream/master --annotate --to="test@email.com" --cc="mailing@list.com" -1 <grim`>fatal: unknown commit upstream/master <grim`>format-patch -o /tmp/UPgSw3IGPz --base=upstream/master -1: command returned error: 128 <grim`>How are you guys getting the new lines in the same message? <nckx>You mean in this channel? <nckx>That depends on your client. <geri>If you use evil mode you can go to normal mode and press <geri>though it for some reason shows at two messages lol <grim`>Yeah. That allows me. In emacs mode you can do with <C-j> but anyways some get split <nckx>geri: And that's called flooding and is punished with a 30-second quiet 😉 <nckx>‘Punished’ because it's usually done by accident when someone pastes a whole file into their text box. It's not really a punishment. <nckx>grim`: Does ‘git log upstream/master’ show anything? Did you ‘git fetch --all’ or so once before? <grim`>Re: meson: For some reason that issue got unnoticed but we commented it here in the IRC and some of you told me that you could replicate it. <nckx>attila_lendvai: Are you sure it's git send-email doing so? <nckx>What exactly is ‘it’ here, grim`? <nckx>The fact that some packages are just broken now? <grim`>I had not fetched upstream 🫨 <attila_lendvai>nckx, well, i'm issuing a git send-email from a terminal. not sure whether something interferes or not, but i'd be surprised if it does. <grim`>nckx: with 'it' I meant that we commented the issue in this IRC and someone else told me that It was not working for him so they asked me If I could file the bug report and thats what I did. Not sure if It's now working for everyone else. Is it? <nckx>grim`: This is not user-friendly advice, but have you tried finding a different commit that has the meson you need? There are probably many, and the bug report says ‘certain commits’. <geri>virt-viewer fails to compile, how to report? <nckx>Uh, I just updated it this week or so. <nckx>The general answer is ‘send mail to bug-guix at gnu dot org’, but hold off and let me test first. <nckx>Builds here. If the one that failed for you was version 11.0, please do say so, some build failures are non-deterministic. <grim`>After fetching upstream now I have a more reasonable error message: <grim`>git send-email --base=upstream/master --annotate --to="test@email.com" --cc="mailing@list.com" -1 <grim`>fatal: base commit should be the ancestor of revision list <grim`>format-patch -o /tmp/6qc8ddDmQS --base=upstream/master -1: command returned error: 128 <nckx>Which happens to be exactly what I was alluding to earlier with grim`. <nckx>grim`: git rebase upstream/master ? <nckx>If any merge commits pop up, you're in the best position to fix them. <nckx>I have a ‘git fetch --all && git rebase upstream/master && other stuff && git push upstream’ one-liner I just C-r yank from my bash history each time. <geri>oh yeah, virt-viewer compiled properly <grim`>Wait. I will try to fix this with some git fu. I will comeback when I figure it out haha <nckx>I have no idea why git is so confused. 9ff1e76 is on upstream master… :-/ Sorry, I'm of little help here. <nckx>Congrats! For posterity's sake: what did you have to do? <nckx>attila_lendvai: #64930 duh, yes, of course, thanks! (No idea why your mails ended up in moderation; won't happen again.) <grim`>Aside from configuring the smtp server. (Which I still haven't tried if it works). I had to: <grim`>1) Fetch all changes from what git considers as base (upstream in this case). <grim`>2) Rebase all commits that affect the branch getting emailed to the base (upsteam/master) in this case. <grim`>So I'm going to send the patches. This are the 2 that I want to send which are in the head of my master branch. <grim`>9aa29f2e29 * master gnu: Add abstract-sddm-theme. <grim`>613f21c0ac * gnu: Add dexy-color-sddm-theme. <grim`>Can I do `git send-email --base=upstream/master --annotate --to=guix-patches@gnu.org -1`? <grim`>and later `git send-email --base=upstream/master --annotate --to=guix-patches@gnu.org -2`? <grim`>And since this is just adding very minimal packages. Should I annotate with the `-a` flag? <grim`>I've already figured it out. Patches sent! Thanks all for your support and kindness 🥰 <geri>what's the difference between service and simple-service? <geri>might sound silly but when i need to add something to home-environment-variables-service-type, service actually seems simpler than simple-service <geri>because i don't need to specify a name for it or anything <grim`>I'm going to submit a fix which I'm not sure if it should be a patch or a bug-fix. We have a package under games.scm which is `cockatrice`. It's lacking a dependency to work with wayland, `qtwayland-5`. Wayland versions should be a new package called `cockatrice-wayland` or should I just add a fix to `cockatrice` adding the missing dependency? <nckx>If the result works fine on both: the latter. <geri>nckx: was the results thing to me or grim`? :D <porcupirate>Guix, yesterday I sent a patch to fix an issue guix causes with renpy. Please review and apply. <nckx>geri: To grim. I steer clear of Guix Home support :) <geri>just imagine it's any other service <next4th>porcupirate: looks good, seems have an extra 'renpy' at the end? <porcupirate>I sent a reply with a new patch without the extra renpy. <lilyp>Uhm, what's the exact issue you're trying to solve here? By default, you want to point renpy at the 'game' subfolder. <lilyp>Note that upstream renpy works differently. Since It needs the whole python fluff, it points to the root rather than the game subfolder most of the time. <lilyp>For compatibility's sake, I'd like to actually create the game directory, if that makes sense. <attila_lendvai>rekado, issues.guix.gnu.org seems to be lagging behind the debbugs data. at least an hour now. <attila_lendvai>damn. now i'm also getting the "fatal: base commit should be the ancestor of revision list" error. something must have changed somewhere. reading the backlog... <attila_lendvai>i was also making a mistake of using -2 while i only had 1 commit in the branch <attila_lendvai>building texlife-font-maps fails for me with (delete-file "fonts/map/dvipdfmx/updmap/kanjix.map"), "In procedure delete-file: No such file or directory" <apteryx>should elogind be updated on a feature branch? <apteryx>woud be nice to fix this 'guix lint' noise: label 'util-linux' does not match package name 'util-linux:lib' <porcupirate>lilyp: FileNotFoundError: [Errno 2] No such file or directory: '/gnu/store/wkajjqr7x7hxailixvj2y6y4ydpj0q8f-renpy-8.1.0/share/renpy/gui/gui/bubble.png' <porcupirate>In summary it should be looking for gui/game/gui/bubble.png instead. <radio->And what is discussed there is something that really interests me <radio->I was wondering if any progress was made to implement that feature <apteryx>ACTION pushes a elogind-updates feature branch and register a spec for it on ci.guix.gnu.org <jlicht>apteryx: any exciting changes to expect, or more of a *git push & pray* kind of change? <jlicht>nvm, just received the debbugs mail /w delays :) <renngar[m]>For any "guix shell --check <pkgs>" command, I get a "guix shell: error: fport_read: Input/output error". <renngar[m]>Works fine with sudo, but...that's probably not the answer. <efraim>14 hours into the build phase of icedtea@2 on powerpc-linux the build was OOM killed <podiki[m]>i'll merge mesa-updates to master later today I think (by merge I mean I'll just cherry pick the commits probably, seems cleaner to my eyes at least) <apteryx>seems my newly created job spec in cuirass didn't build anything yet... <lilyp>rebasing is almost always superior to cherry-picking <apteryx>the last revision it build for the master branch is c7e45139fa from 3 days ago <apteryx>ACTION tries 'herd restart cuirass' on berlin <GrigoryShepelev[>apteryx: oh. you mean the "ci.guix...". substitutes server seems to be fine for me. it's just I couldn't write a proper specificiation for it few days ago to be able to build non guix channel 🦀 probably just mine "golden hands" <attila_lendvai>ACTION will just wait for the open source fairies to fix it... o/ <minima>attila_lendvai: do you have the feeling this only affects a subset of users, triggered by some specific circumstances? <minima>i haven't had the chance to look at the logs, anything there that hints at something in particular? <nckx>apteryx: Feel free to restart cuirass, but I don't think it helps much. <apteryx>nckx: agreed... cbaines had good ideas with a pointer to the place to try it in the code <apteryx>not a total fix but to limit the damage to the broken spec/repo <apteryx>at least it picked up my new job spec now, elogind-updates <janneke>jpoiret: i've run guix pull in a hurd-team childhurd, but alas, guix build hello says: <janneke>Throw to key `record-abi-mismatch-error' with args `(abi-check "~a: record ABI mismatch; recompilation needed" (#<record-type <build-system>>) ())'.