IRC channel logs
2025-10-08.log
back to list of logs
<csantosb>sneek: later tell apteryx, I'll take care of old open pr by the end of the week <hugohugo>ieure: next to xvfb-run there is also xorg-server-for-tests, which I think is the preferred way to run tests that require X, but not sure <csantosb>Any idea of which package provides `getopt` ? <sneek>kestrelwx, you have 1 message! <sneek>kestrelwx, bdju says: I do not have a decent machine. I think the build would take over 24 hours. <csantosb>`guix search getopt` returns quite a few answers <efraim>futurile: no problem. It's definately project related <efraim>kestrelwx: I just told berlin to build your patch ontop of my two patches <kestrelwx>efraim: Thanks! I think i should've patched more `dlopens` just in case. And the loading in ANGLE code. <efraim>I think we might need the angle bit specifically. I previously patched all the dlopen spots but without the patch and that wasn't enough <efraim>if this doesn't work then my next attempt will probably be the 3 patches we have now plus all the other dlopen spots <efraim>somewhat related, in that it is new, when swayidle kicks in I get dumped back to the login screen <kestrelwx>On master, if you select text on a Web page and then hold LMB and try to move it around, you can actually see it renders fine. <efraim>I've killed swayidle on my machine for now <kestrelwx>So it's something about rendering to the window itself. <efraim>I don't remember if opengl is the default or if vulkan is, but I saw the answer earlier this week <efraim>I wonder if the LMB "trick" is using vulkan or opengl, and if opengl also needs adjusting <efraim>I guess once we have it fixed I should break out my pinebook pro and see if it works there too <kestrelwx>I think the upstream patch is the only thing that really matters and that Vulkan rendering were broken for all of the time, including when QtWebEngine views were working. <efraim>kestrelwx: 11:08:40 WARNING: GBM is not supported with the current configuration. Fallback to Vulkan rendering in Chromium., followed by a segfault <kestrelwx>Is there anything with `qutebrowser --qt-flag enable-logging --qt-flag v=1`? <kestrelwx>Would trying to access a bad path in `seccomp` sandbox segfault? <apteryx>should we consider it a bug that we can't use 'guix shell -C' in a 'guix system container' ? <sneek>Welcome back apteryx, you have 1 message! <sneek>apteryx, csantosb says: I'll take care of old open pr by the end of the week <apteryx>it prints e.g.: guix shell: error: remounting /gnu/store writable: Operation not permitted <jlicht>apteryx: not sure if you saw my message, but I think your core files also were whisked off to some random place w.r.t. mount namespaces <apteryx>jlicht: thanks, yes I saw it :-) that explains why the mysterious cores files were not produced. I thought I had tried to put the segfault-causing fragment and still couldn't see one, so I had ruled that out. <efraim>kestrelwx: no change from before with more dlopens patched <kestrelwx>efraim: Does it crash if you do `qutebrowser --qt-flag disable-gpu-sandbox`? <kestrelwx>Like this? `qutebrowser --temp-basedir --qt-flag disable-gpu-compositing` <efraim>qutebrowser --temp-basedir --qt-flag disable-gpu-sandbox, i'll try the other one <kestrelwx>This actually makes the page render on master. <kestrelwx>efraim: Does it also crash if you have "QT_QPA_PLATFORM=xcb", it probably should. <efraim>however, with the `qutebrowser --temp-basedir --qt-flag disable-gpu-compositing` it pasted incorrectly before, it doesn't crash with the patches <kestrelwx>Is `--qt-flag disable-seccomp-filter-sandbox` also not enough? <kestrelwx>Could you see if there's a `dlopen` before segfault in strace? <kestrelwx>We should probably just ask Qt people if they know what'd be enough. <kestrelwx>It'd be `strace -f 2>out ...` since it's probably in a child process. <kestrelwx>I think if you do `qutebrowser -d` it should print the stacktrace in the console. <kestrelwx>The paths being truncated somewhere mid hash is like reality show preview. :D <kestrelwx>Ah, you would have to rebuild it again with symbols. <ekaitz>fatal: mismatched algorithms: client sha1; server sha256 anyone? <ekaitz>i have this issue when using git-fetch from codeberg <ekaitz>i'm packaging something that uses objectformat=sha256 in git <ekaitz>and it doesn't clone it, and reports that issue <futurile>if you'd like to help out posting news about Guix and engaging with people on interesting topics let me know - be great to have a group! <ekaitz>prepare for people complaining at you <ekaitz>so i don't see how to configure the git client to use sha256 instead of sha1 <ente`>after a recent system upgrade, I've been having problems with having my X session freeze when I close the lid <ente`>it seemingly has nothing to do with suspend/resume, which works fine <ente`>when I open the lid again, I get a black screen, from which I can vt-switch <ente`>does this sound familiar to anyone here? <kestrelwx>lilyp: I'm actually building Julia once again real quick :D. The tests are complaining about missing `pcre2.6` shared library. <kestrelwx>`phase `check' failed after 9186.8 seconds`. lol. <lipklim>Hi all, I'm trying to package tree-sitter-fortran but it builds non-deterministically: for some rounds output dirs in the store have 'write' mode. Could please somebody take a look? https://paste.debian.net/1400076/ <FuncProgLinux>Thunderbird died, but ERC look s nice, the doom-modeline tells you if you have unread messages :) <ieure>FuncProgLinux, I like ERC a lot, I've been using it a very long time. <ieure>FuncProgLinux, IMO its best feature is erc-track, which lets you quickly skim channels with unread messages, dropping you right back to where you started at the end. <ieure>I extended that to slack-mode, it's great for work stuff, I can dip in and out as needed. <FuncProgLinux>ieure: I just started using it, I liked the IceDove interface but I only used it to IRC chat. I now have another emacs package to configure for hours lol <alex_>How can make ssh server in guix <bavier>lipklim: I wonder if there's non-determinism in other tree-sitter packages. I didn't get this error until the fourth or fifth round of check. <lipklim>bavier thanks for confirmation. As a reference, I --check'ed json package multiple times and it never failed <FuncProgLinux>Does that mean those packages are not bit-by-bit reproducible? <peterpolidoro>I am trying to submit a pull request on codeberg and I am getting a 500 internal server error. Is there an issue with codeberg right now or am I doing something wrong? <ieure>peterpolidoro, I did notice some 500s earlier. 500s explicitly mean you're not doing something wrong, the server broke. <peterpolidoro>do we create a pull request to guix:master if there is not an appropriate team branch? <ieure>peterpolidoro, It also says <100% availability in the last 15 minutes. <ieure>peterpolidoro, Most PRs should go to master, unless it's specifically part of a team effort. <ieure>peterpolidoro, You can try again now and it'll probably work. <lipklim>The same node ("none") was mentioned as anonymous and named in different places. <lipklim>It's strange that node.js produces different results though. <lipklim>I'm also checking already packaged packages and so far so good. <lipklim>bavier could you please give an example of the command that you've run. it seems to be diffoscope, but I don't understand why I didn't get the same diff :( <kestrelwx>sneek: tell efraim later the one on my branch works out the box for me, thanks for queueing it. <sneek>efraim, kestrelwx says: later the one on my branch works out the box for me, thanks for queueing it. <kestrelwx>I think Vulkan in ANGLE doesn't work, though. I think code in `vulkan_loader.cpp` is relevant for that, but I'll look into it tomorrow. <trev>it's strict about how you name the package. in that example it must be `emacs-next-pgtk' <kestrelwx>It's kind of exciting I guess, like controlling a puppet. <luca>Hi, I'd like to hack on a C++ project. I ran `guix shell -D biboumi` and build it with `cmake . && make`, but when I try to run it I get an error `cannot open shared object libiconv.so.2`. Am I missing something obvious about how development should work? <lispmacs[work]>I'm trying to do a system reconfigure after my last one around six months ago. But the repository pull keeps dying after about 50 or 60 percent, with a service not available error. Do I need to switch to some mirror...? <ieure>lispmacs[work], Are you pulling from Savannah? <lispmacs[work]>I actually did a successful `guix pull', but when I run `sudo guix reconfigure ...' it for some reason pulls the repository again <Kolev>I haven't updated my Guix machine in ages. I'm scared of the time consuming nature of the task. <FuncProgLinux>I have a laptop I haven't updated since May lol, I also think a clean install would be better <ieure>lispmacs[work], Simplest thing is to `guix describe -f channels > channels.scm', edit it to change the host and commit, then `guix pull -C channels.scm', then reconfigure. <bavier>lipklim: I just ran `diffoscope --text - <previous output> <check output>`. Cool that you found a fix! <Kolev>All it does is run Jellyfin for the house. It should be fine running am ancient version of System. <FuncProgLinux>I wonder how difficult would it be to package OpenLiteSpeed in Guix (it's GPL-3.0) <FuncProgLinux>I also have a server setup with Podman containers, but I'm thinking into moving it to Guix. Idk if rootful podman containers work there <luca>I guess the expected way to do dev work is to make a guix.scm file and run `guix shell -f guix.scm -- biboumi` to run it /shrug <Phosphenius>Hey, new Guix user here. I’m trying to provision a VPS on Hetzner using their API und `guix deploy`, but it runs into a problem while trying to connect via SSH to the rescue system. <redacted>The way guix home manages shell profile files means you can't set PATH for login shells with the zsh service. <redacted>.zshenv gets sourced first and sets the environment variables you gave the zsh service config in `environment-variables', but then .zprofile sources /etc/profile, which unsets PATH <redacted>Maybe /etc/zsh/zprofile should source /etc/profile instead, like Arch does it. <kestrelwx>sneek: later tell efraim I don't know if I should remove the patch for Skia code from the commit(it's not harmful, but also not sufficient, and not related to the rendering problem we see), so I won't rebase tonight, please see for yourself. <bvk>Why does `debootstrap` package depend on these many different keyrings? Am I reading the output correct? dependencies: bash-minimal@5.2.37 debian-archive-keyring@2025.1 dpkg@1.22.21 elxr-archive-keyring@2024.1 gnupg@2.4.7 kali-archive-keyring@2025.1 pardus-archive-keyring@2021.1 perl@5.36.0 + pureos-archive-keyring@2021.11.0 trisquel-keyring@2022.10.19 tzdata@2025a ubuntu-keyring@2023.11.28.1 wget@1.21.4 zstd@1.5.6 <bvk>Shouldn't it just be debian-archive-keyring ? <ieure>Guix has a package for debootstrap? TIL. <FuncProgLinux>Question about the issue I opened on MATE about refreshing it before 1.5.0. Does it really make sense to enable features in pluma? <FuncProgLinux>I mean guix already requires some gnu/linux knowledge and Emacs is by far the best integrated editor here...You could argue pluma starts faster for "quick snippets" but so does the emacsclient. <nomike>I tried adding "nss-certs" to the packages inputs, but that doesn't seem to help.