IRC channel logs

2025-07-09.log

back to list of logs

<sturm>I asked a while back about blank GTK applications windows under Wayland. Seems to be related to the (possibly new) Vulkan renderer. Applications render well with `GSK_RENDERER=ngl`. Bug report here: https://codeberg.org/guix/guix/issues/1119
<sturm>(and related to the Ivy Bridge chipset on my Lenovo X230)
<futurile>Morning all
<jlicht>o/
<efraim>o/
<gabber>\o
<efraim>any touching guix/build/gnu-build-system.scm, even a newline, leads to a world rebuild
<j8>Hello, In Emacs (~emacs --no-init~), ~C-h v find-program~ shows: find-program is a variable defined in ‘grep.el’. Its value is "/gnu/store/6i0xi5kf38jk54sw64akd3iv8scw5g6x-findutils-4.9.0/bin/find". However, in `grep.el`, it’s defined as: (defvar find-program (purecopy "/gnu/store/w4p913kmnga3xh80vx8vnkri42dcmqi3-findutils-4.9.0/bin/find"). How can the value differ if `--no-init` loads no configuration? I’m not sure if its guix bug or emacs bug. emacs-version: 30.
<tazjin>cbaines: fyi, we're now syncing from bordeaux (only the last month or so actively, the rest will get fetched lazily if someone requests it and then cached)
<tazjin>and further every 30 minutes (can tune it) from the recent-changes endpoint
<ekaitz>efraim: who is in charge of banning people from the ml?
<ekaitz>we have some coc violation, maybe you already managed it
<efraim>is it about aviva? IIRC I missed most of that
<efraim>the mailing list is currently in emergency moderation mode, someone needs to manually approve all emails
<efraim>that only helps with emails to the mailing list though
<ekaitz>i think aviva was the name, yes
<ekaitz>i think for us as a project is more important to have a rapid response to that kind of idiocy than anything else (like checking if a software project is done by people we agree or disagree with)
<ekaitz>efraim: in any case, i just wanted to know if that was taking care of, and it was so thank you :)
<ekaitz>efraim: also, now you are here, I opened a pr for the wip-riscv-bootstrap branch
<efraim>I haven't had a chance to look at it just yet, life is a bit hectic :(
<ekaitz>yeah, mine too, just had a kid and...
<ekaitz>it's just a plain pr, but I'd like to start working on it someday and merge it to guix eventually
<efraim>of course, makes sense
<ekaitz>with your help
<ekaitz>(when you are available)
<tazjin>alright, new mirror up & running: https://mirror.yandex.ru/mirrors/guix/
<tazjin>should serve CIS, Central Asia, Arabic countries and parts of Africa pretty well
<wingo>so. i am trying to build firefox from git. it would like for my glibc to expose gnu/stubs-32.h. however there is only stubs-32.h. i assume that this is because the installed glibc doesn't have the sort of cross-compiling support
<wingo>this is all in a container with a guix env made from a manifest.scm.
<wingo>does anyone know an easy way to add "the i686-linux version of glibc" to a manifest?
<Rutherther>wingo: I think it's going to be easiest if you just nest two shells, one of them --system=i686-linux with the glibc, then it should support 32bit
<Rutherther>imo guix doesn't really play too well with multiple systems in one manifest, I am trying to make that different: https://git.ditigal.xyz/~ruther/guix-cross-shells. I think this could be used as well, but nested shell is probably just going to work, so will be easier
<wingo>ah good idea, thanks!
<wingo>hm, i wonder if that will work given that i am running with -CNF (the fhs mapping)
<Rutherther>wingo: yeah, I am not sure if you can nest with container, you can try --nesting, and maybe also --share the guix-daemon socket
<Rutherther>wingo: I just quickly added support for native compilation to my guix-cross-shells utility and with it, to make a package that will have glibc 32bit, you would do this: https://paste.debian.net/1385039/. Note that you won't get the include in search paths with this, and easiest way to get them would be to change glibc with gcc-toolchain. That way you will get full 32bit gcc toolchain in the cross profile
<Rutherther>wingo: the resulting package should be fine to be added to a manifest
<Rutherther>unfortunately it's not even a channel now as it's just a proof of concept
<Rutherther>okay, I made it into a channel, but I am not on an internet connection where I could test it
<csantosb>Hi emacs team people, a few days ago I remember an email related to including or not built-in emacs code as package dependencies (which-key, org, etc.); did I dream ? I completely lost it in the wild. Ring a bell ?
<podiki>csantosb: i remember some subject line like that too but not seeing it...
<podiki>csantosb: maybe this issue? https://codeberg.org/guix/guix/issues/1055
<ekaitz>podiki: btw congrats for the launch
<ekaitz>csantosb: why don't you apply for commit access?
<csantosb>ekaitz: I appreciate guix; I'm too dumb for that ;-)
<ekaitz>csantosb: i am a commiter, don't underestimate how permissive guix is
<ekaitz>:)
<ieure>csantosb, Hey don't be mean to csantosb! You contribute more than some committers.
<csantosb>podiki: not that one; maybe an email sent amont the emacs team members
<ekaitz>ieure: he just commited a CoC violation
<ekaitz>haha
<podiki>ekaitz: thanks!
<ekaitz>csantosb: in fact, you landed more commits in guix than me
<csantosb>ekaitz: it's not about quantity, but about the quality
<ekaitz>(way more)
<ekaitz>half of mine are just minor changes
<ekaitz>don't be silly
<ekaitz>it's just for making your process of contributing more comfortable and agile
<ekaitz>you could still send patches or open prs for changes you need some review
<ekaitz>just consider it, you are doing a great job
<ekaitz>(i'm not trying to push you though, you decide)
<csantosb>ok, thanks; the day I manage to decrypt what on hell #$@ means, I'll apply, I promise ;-)
<ekaitz>csantosb: ungexp splicing?
<ekaitz># is the character for the reader macros
<ekaitz>$ is the gexp-y thing
<ekaitz>and the @ is usually the splice
<ekaitz>> gexps, consist essentially of three syntactic forms: gexp, ungexp, and ungexp-splicing (or simply: #~, #$, and #$@), which are comparable to quasiquote, unquote, and unquote-splicing
<attila_lendvai_>heh. clang -m32 generates an elf binary that has a /gnu/store/...ld-linux... reference to a nonexistent store item...
<futurile-afk>me laughs and the gexp'y joke - so true
<futurile-afk>gah!
<umanwizard>Is anyone familiar with using private channels in channels.scm and authenticating via ssh? I tried what I thought was the most obvious approach, but it failed with an authentication error:
<umanwizard>(channel
<umanwizard> (name 'mine)
<umanwizard> (url "git@github.com:/umanwizard/guix-channel"))
<umanwizard>(the SSH key I hope/want it to use is the one at ~/.ssh/id_rsa)
<luca>Isn't that a slash too many? I don't think ssh urls look like that on github
<identity>#$@ is what is say when my code is not working...
<identity>i say*
<umanwizard>luca: no, this is the correct URL, as confirmed with `git clone`.
<umanwizard>which slash do you think should be removed?
<umanwizard>presuming you mean the one before the username:
<umanwizard>brennan@eugene /tmp$ git clone ssh://git@github.com:umanwizard/guix-channel
<umanwizard>Cloning into 'guix-channel'...
<umanwizard>ssh: Could not resolve hostname github.com:umanwizard: Name or service not known
<umanwizard>fatal: Could not read from remote repository.
<avalenn>identity: and #$@#~#$`,' when you debugged for too much time ?
<umanwizard>Please make sure you have the correct access rights
<umanwizard>and the repository exists.
<identity>avalenn: (in a very frustrated voice) ungexp-splicing gexp ungexp!
<umanwizard>How can I create a profile that's the same as the one `guix shell` would have created, without actually shelling into it? I.e. I want to make it so ~/profiles/foo is a symlink to the profile that would have been generated by `guix shell foo bar -D quux`
<ekaitz>umanwizard: guix install -p $PROFILE_NAME?
<ekaitz>umanwizard: in your case it would be something like `guix install -p $PROFILE_NAME foo bar -D quux` if i'm not mistaken
<ekaitz>hm! maybe you cannot -D in install
<ekaitz>you could list the inputs and install them instead
<umanwizard>yeah, I tried guix package -i and ran into the same issue (no -D)
<umanwizard>is there a command to list the inputs (i.e. the same thing guix shell -D does) ?
<ekaitz>hmmm there are tricks for sure
<ekaitz>umanwizard: maybe guix shell -P?
<ekaitz>or -r
<ekaitz>hmmm
<csantosb>sneek, later tell umanwizard, what about creating the shell, then `guix package --export-manifest > manifest.scm`, and using this manifest to create a profile ?
<sneek>Got it.
<AidenIsik>Hi, does anyone know how to create a guix shell with a specific glibc version used for gcc-toolchain?
<AidenIsik>I am building an old-ish Android/Linux kernel, so I need older versions of both GCC and glibc.
<AidenIsik>I am already good on the older GCC side of things, I can just use gcc-toolchain-5 for that, but I can't seem to figure out how to build a custom toolchain using glibc-2.29 (which I am making a guess will work fine based on the kernel source release date I have).
<AidenIsik>I found a StackOverflow answer outlining how to use make-gcc-toolchain to do this (https://stackoverflow.com/questions/66063337/build-against-an-old-glibc-with-guix), and I also found the original thread which lead to the libc being added as a parameter to that function (https://issues.guix.gnu.org/36346), where I can see that at the top of the file the make-gcc-toolchain function was exported.
<AidenIsik>Looking at the top of that file on my system now (via guix edit gcc-toolchain), I can see that it is no longer exported, so I cannot use that function. Is there another function that is preferred to be used now instead? Was it's export removal a mistake?
<AidenIsik>Thanks in advance.
<csantosb>AidenIsik: You know 9.1.2 Package Transformation Options ?
<AidenIsik>Reading that page now
<csantosb>guix build gcc-toolchain --with-source=glibc=SOURCE’
<AidenIsik>I guess there are no version-specific patches which need to be applied to glibc in that case?
<AidenIsik>I just assumed there would be because it's quite a large project
<csantosb>Another option, look at here: https://codeberg.org/guix/guix/src/emacs-vunit-mode/gnu/packages/commencement.scm#L3701
<csantosb>Then, `make-gcc-toolchain` accepts a libc parameter, taking glibc-final by default
<csantosb> https://codeberg.org/guix/guix/src/emacs-vunit-mode/gnu/packages/commencement.scm#L3600
<AidenIsik_>Problem is I can't access that variable in my own .scm file as it's not exported. I'll try to inherit the package and override the source that way, hopefully that works. Thanks for the help.
<AidenIsik_>Wow my connection is flaky today
<csantosb>AidenIsik: one last idea, using time machine with this commit: `5f3f703918 gnu: glibc: Update to 2.29`
<AidenIsik_>Ohh that could be an option too
<csantosb>Well, anything before `0b3df5c913 gnu: glibc: Update to 2.30.` would do it, I guess
<AidenIsik_>I think I'll do that actually because then it'll bring all the packages down to the version they were back then
<AidenIsik_>Avoid any further headaches
<csantosb>6 years ago, so not sure you'll have substitutes, 🤞
<AidenIsik_>Ah, that's fine. I can compile GCC in maybe 10 minutes or so
<AidenIsik_>Everything else should (hopefully?) be faster
<csantosb>Good luck !
<AidenIsik_>Thanks, currently seeing if I can find an older commit with ncurses 5 lol
<AidenIsik>Currently I'm getting that from another repo but it'd be nice to get it from just the one
<csantosb>Guix git log are quite good to that purpose