IRC channel logs
2025-09-29.log
back to list of logs
<ieure>FuncProgLinux, Seems fine to me. <robin>kde's having trouble with my keyboard layout because of...libxml2 apparently, not what i was expecting :) (maybe i should take a look at libxcbcommon) <mange>These libxml2 issues sure are popping up everywhere. <FuncProgLinux>mange: I think the best course if you are a user is to report the issues on Codeberg so the Guix magicians track the issue down as well :) <trev>please someone tell me how to test a code change of a service type before I cry <trev>i tried a system reconfigure with -L [path to guix codebase with change] but i don't think it worked <mange>Have you tried ./pre-inst-env? Or do you need some other channels to make it work? <trev>mange: did try that, but i need other channels for my system config <untrusem>> please someone tell me how to test a code change of a service type before I cry <trev>untrusem: it's my regular system <mange>Certainly, using a container/vm is great if you can do it. <mange>But that doesn't solve the problem of "how do I construct the operating-system object I need to build the system". <trev>yeah i think that will be my only choice if i can't figure out how to hot load the modified service <trev>i figured with the beauty of repl-driven dev there might be a secret way <mange>One heavyweight way to do it is to use guix time-machine (or guix pull) to pull from your local git repo and build a full guix. <trev>yeah, my next step was to add a channel for my local repo <mange>You can't just add it as a new channel, it would need to replace the guix channel. <mange>That's why -L doesn't work, I expect. <trev>mange: yeah, i would comment it out <mange>Didn't you say using -L didn't work? <trev>i don't believe it did. perhaps the code change i made is just ineffective <mange>Right. Well, you should answer that question first before we go any further. <mange>Easy way to check: make a change that would definitely break, and if using -L doesn't cause it to fail then it doesn't work. <trev>adding an (error "foo") and using -L in the reconfigure doesn't work. nor does `guix system edit [modified service]` and checking for that line in the code <mange>Yep, okay, so -L won't help. I swear I read something recently on a mailing list about exactly this, but I'm struggling to find it now. <trev>a bit surprising. i'm willing to test it however everyone else does it though, so if that means making a VM i guess i should start <trev>i searched for a futurile-afk tutorial as well D: <Rutherther>trev: you need to use pre-inst-env. Guix command will prepend its own modules to the load path, so it is non-replaceable. Since you are using multiple channels, add those channels to load path with -L <trev>Rutherther: thanks, trying now. was in the middle of attempting to add a test but this will give me more instant feedback (until i learn how the testing framework works) <apteryx>ACTION attempts to build their system/package collection with mesa-updates <apteryx>the cmake build system enabled color output, but that seems to make things worse for me; anyone else? Am I missing something? less --use-color still displays things like: player.c:ESC[mESC[K In function ‘ESC[01mESC[Kvoid ESC[01;32mESC[K on_eofESC[mESC[K(void*[...] <efraim>ugh, I'm looking at mesa-updates too, we have to cut down on the number of compilers. on aarch64 I have gcc-14 x2, llvm-18, gcc-13, llvm-13, llvm-for-mesa <apteryx>efraim: make sure to hard reset, I just force pushed <efraim>I should look at getting more RPI5s and netbooting them from the same image. Those things are fast and solid with the added fan <apteryx>currently I got a build failure in liblinphone (linphone-desktop) <apteryx>guess I need to provide it with an older ffmpeg <efraim>I just told berlin to build mesa on mesa-updates, so we'll have some substitutes there soonish <apteryx>hm, already on ffmpeg-4 for mediastreamer2, not sure what changed there <identity>fnat-xmpp: i think it is for making the closed port object unreachable so it can be garbage collected by Guile? <identity>i assume that when you close a port the underlying resources (like a file descriptor) are freed, but the object itself stays behind so you can ask ‘output-port-open?’ and such <identity>(let ((port (open-output-file "test"))) (close-port port) (output-port-open? port)) is #false, for example <fnat-xmpp>identity: Thanks. The way this is structured (including opening the file as part of the pre-activate lambda instead of before the dynamic-wind) is a bit obscure to me. (It's me, not the way this is written.) I guess there are advantages in doing things this way. <identity>fnat-xmpp: dynamic-wind is used when you want to free and regain resources in the presence of non-local exits with call/cc and whatnot, just to make sure that the continuations created by call/cc are definitely not lightweight. :) if you are going to escape the dynamic scope of the dynamic-wind, the third thunk is going to be invoked, closing the port and set!ing it to #false, while if you are going to enter the dynamic scope of the <identity>dynamic-wind, the first thunk is going to be invoked and the port is going to be re-opened. for example, if an exception fires and it is not handled until the code unwinds to the dynamic-wind, the port is going to be closed before continuing further <fnat-xmpp>identity: There was definitely something that I was missing, thanks! This makes it a bit more clear. :pray: <ColdSideOfPillow>All I did was add (keyboard-layout (keyboard-layout "en" #:options '("ctrl:hyper_capscontrol" "compose:ralt"))) to my system <Rutherther>ColdSideOfPillow: I dont think such layout exists, did you mean en_US? <efraim>that was going to be my suggestion <efraim>there's a lot of options, I use en_IL <ColdSideOfPillow>kestrelwx: I've already tried "us"; that was what I was originally doing. <kestrelwx>'ls "$(guix build xkeyboard-config)/share/X11/xkb/symbols"' <attila_lendvai_>ColdSideOfPillow, my custom keyboard layout disappeared due to the ongoing libxml2 issue in the guix repo. if you have pulled recently, then it may be due to that. i'm running with these commits that are before the breakage: guix: "7f86f1eae0", nonguix: e8b5109 <ColdSideOfPillow>attila_lendvai: I've heard about the libxml2 issue, but I've never actually configured my keyboard layout on the kernel level before (on my old system, I did it on the compositor settings) so that wasn't the issue. <efraim>grr, just had another aarch64 board fail to boot, down to 1 now <kestrelwx>Not visible as in there's none or there's 1 instead of 2? <kestrelwx>You have to escape backslashes with backslashes. <kestrelwx>I guess one more backlash is eaten by the shell. <efraim>perhaps one gets eaten by guile also? <efraim>perhaps 4 would do it then, although I'd test in a VM so you don't have to constantly reboot <ColdSideOfPillow>I thought that perhaps /etc/issue interprets backslashes, so I tried using `\\\\`, so that it would translate to two backslashes, but that didn't work as well. <efraim>you could cheat with some UTF-8 backslash look-a-like symbol <kestrelwx>Could you try wrapping that in single quotes? <efraim>my only one online right now is an RPI5. I have a pine64 and a rock64 that only boot once, and a couple more in my drawer <kestrelwx>I'm trying to flash an old Android phone, though I seem to have to use a blobby kernel. <efraim>actually now that I have my (newish and fancy to me) POE network switch I can probably pull some of them out and use them <efraim>I used to have a large mess of power strips and an 8 port switch <kestrelwx>I understand I would have to flash a ROM bootloader that would pick up the one in 'current-system' on an SD card. <lipklim>Hi all, I've just switched my local guix repo from old Savannah remote to Codeberg and in the end got an error "guix git: error: unknown introductory commit and signer", but the master branch is updated anyway. Is it expected during migration or should I worry? <Rutherther>lipklim: what operation ended up with this error? <lipklim>after "git remote set-url upstream ssh://git@codeberg.org/guix/guix.git" <Rutherther>I can see that happening only if you have your own commits and the pull strategy as merge. And at that point, if you arent a committer it makes sense you have unauthenticated commits <Rutherther>If you dont have your own commits, then it could also happen if codeberg and savannah were diverging. That was the case few months ago. And if it is caused by that you should probably git reset to origin/master rather than creating merge commits every time you would pull <lipklim>Rutherther I have my own commits, but only on the other branches. master branch is clean. <lipklim>git config --get pull.rebase gives true <Rutherther>I am talking only about the branch you pull, yes <Rutherther>Do you have merge commits from yourself in master branch? <ColdSideOfPillow>lipklim: fwiw I get that error regularly as well, but it doesn't seem to affect anything... <Rutherther>It cannot affect anything, it is in post merge hook, so it is only information <ColdSideOfPillow>Rutherther: I could have worded it better, I meant that it doesn't seem critical. <lipklim>Rutherther, I don't have [guix "master"] or something like this in .git/config. <lipklim>I guess I need to run guix git authenticate? <Rutherther>lipklim: that was already ran and produced the error. I dont understand your message about the git config, why is that relevant? <lipklim>Rutherther, I mean to run as "guix git authenticate COMMIT SIGNER", because post-merge runs without COMMIT SIGNER <lipklim>Rutherther, I've just read (guix) Invoking guix git authenticate <lipklim>On your first successful run, the introduction is recorded in the <lipklim>‘.git/config’ file of your checkout, allowing you to omit them from <lipklim>Oh, I think I know a source of this issue. <lipklim>My repo is shallow, missing initial commit, mentioned on (guix) Building from Git. <lipklim>Maybe I didn't notice this error before, because I was pulling using vc-update and didn't pay attention to the output. <lipklim>I reverted remote url and did 'git pull' and got the same error <efraim>it might be your remote isn't deep enough, IIRC it needs to go back to the 1.4.0 commit <lee-thomp>I've stumbled across this info page port of SICP on MELPA (https://melpa.org/#/sicp) and was wondering if there's a more appropriate package file to add it to than emacs-xyz.scm? I think the emacs lisp in the package is just there for MELPA purposes. I've seen libretro-core-info which looks like an 'info page only' sort of package, but this makes sense to be in emulators.scm. Any ideas? <untrusem>guix info manual is not opening for me in emacs <untrusem>getting `gzip: stdin: unexpected end of file` <identity>a lot of ways, really. it never hurts to be sure <kestrelwx>lee-thomp: Is that different from 'sicp' in Guix? <untrusem>I am currently reconfiguring my system, lets see if it happens after that too or no <trev>can someone using gnome + flatpak (guix home) try to install org.telegram.desktop and see if it can be launched from the Gnome shell launcher? trying to rule out my own system <identity>kestrelwx: the *.texi seems to be slightly different, it seems. <tusharhero>anyone here know why I keep getting these kinds of messages? <ieure>tusharhero, It's a bug, it'll be fixed when the mesa-updates branch merges. <ieure>tusharhero, There was a libxml2 update for a security issue, the update was grafted, but the update also broke the libxml2 ABI and tons of things depend on it. So stuff's been busted in a variety of ways. <ieure>Is the tl;dr of what's going on. <untrusem>identity, only guix main `~/.config/guix/current/share/info/guix.info-1.gz` is affected, other info files opens fine <untrusem>Error while executing "gzip -c -q -d < /home/untrusem/.config/guix/current/share/info/guix.info-1.gz" <identity>untrusem: it works fine on my machine™, and i am on a fairly recent commit <identity>i doubt those commits have changed the manual <identity>that is too small, for me guix.info.gz is 7,1K and guix.info-1.gz is 89K <untrusem>hmm, how to debug this, what could have caused it :? <identity>you could start with “guix gc --verify=contents,repair” <lee-thomp>kestrelwx: identity: my apologies no idea why I missed the sicp package already in Guix, yeah they seem to be both forked from the same upstream with minor differences <untrusem>but still can't access that guix manual from emacs same error <identity>you probably need to re-login/restart emacs/rebuild the profile <untrusem>yeah, I thought that reconfiguring my home <apteryx>efraim: perl-xml-libxml is broken; making it use libxml2-2.11 causes 2 test failures <apteryx>(had to use libxml2-2.11 there as well) <apteryx>scrollkeeper also broken on mesa-updates, because of libxml2 bump <htgoebel>Hi, I try to build a different version of a package using --with-version. This does not apply the patches defined in the package-source. What is the recommended solution? <ieure>htgoebel, Surprising, are you sure that's the case? <htgoebel>tusharhero: this is what mentions --with-version :-) <tusharhero>maybe you forgot to --with-version=package-name=version <htgoebel>tusharhero: no. tested for you: this would raise an error:error: 7.0.7: invalid upstream version specification <Rutherther>htgoebel: I think easiest will be using guile. You can use procedure package-with-upstream-version from guix transformations module. If you pass it preserve-patches? as #t, it will preserve them. --with-version calls that function and uses the default, false. <tusharhero>htgoebel: I think the error message is pretty clear <htgoebel>Rutherther: Thanks. This is over-challenging my scheme/guile/guix skills :-) Esp. since I need to build other packages depending on this transformed one. <htgoebel>Rutherer: Looks like the easiest way is to temporarily change the package definition. <Rutherther>htgoebel: that's what package-input-rewriting is for, but yeah, it can be challenging to get it right <gabber>is the xdebug license considered free as in freedom and worthy for inclusion in guix? <mrh57>ootl have there been any talks since the codeberg migration about agit? <gabber>CaptainBeyond: thanks for the reassurance! <gabber>mrh57: you mean like discussions or presentations? <gabber>mrh57: i guess YES to the former and also YES to the latter although i doubt the latter was specifically for the Guix project <PotentialUser-88>currently, I'm trying to figure out how to package my custom quickshell-based shell in my personal channel <PotentialUser-88>however, I don't understand how I can install to ~/.config/quickshell/ > <ieure>PotentialUser-88, You can't, you need a service in addition to your package. <ieure>PotentialUser-88, A Guix Home service, specifically. <ieure>PotentialUser-88, Guix packages cannot put files onto the filesystem in the way you want at all. All package contents are contained with in their store item, which is a directory in /gnu/store <PotentialUser-88>ieure: it seems that for something as simple as putting something on the filesystem, I can just use the simple-service helper to extend etc-service or home-files-service ? <ieure>PotentialUser-88, Those are great options if you're only dropping static config into those places. If it's dynamic or configurable by the user, you'll need to be a bit more sophisticated. <ieure>PotentialUser-88, I'm not sure, #guile probably a good place for that question. <ColdSideOfPillow>hmmm I was wondering why my Anki (a flashcard program) flashcards weren't working, and I've realized that Guix packages version 2.1.16 whilst the latest version is 25.19.2