IRC channel logs
2024-04-05.log
back to list of logs
<lispmacs[work]>input latency under wayland is very bad on my system, curious what other people are seeing <lispmacs[work]>I switched gdm-configuration back to (wayland #f). That helped a lot, though the newer version of gnome-shell is still somewhat more laggy <lispmacs[work]>I wonder if there is some gnome-shell eye candy that can be turned off <oriansj>ok, this doesn't make sense. Guix build is failing on ghc-8.10.7 ; so I go a machine where it doesn't fail and guix archive --export -r ghc@8.10.7 >| ghc-8.10.7.nar and do guix archive --import < ghc-8.10.7.nar (it imports without issue) and guix doesn't use the ghc-8.10.7 we just imported and instead tries to build again and fails again. <rekado>looks like mu4e changed reply keybindings <rekado>it used to be that "R" replied to all (or asked for whether you want a wide reply); now "R" is just a direct reply and "W" is a wide reply. <rekado>oriansj: could it be due to a graft? <freakingpenguin>If I wanted to mirror a Guix repo for tracking my WIP patches, what's the best way to manage that? Two remotes, one for my mirror and one for upstream? <freakingpenguin>Idea being push/pull wip patch branches to/from my guix mirror to track them on different machines, while still being able pull from upstream. <oriansj>rekado: no grafts, pure vanilla guix <dumbmf>Hey all, I got a question about guix containers. <dumbmf>Right now, I'm able to sandbox/containerize Emacs under wayland with a little oneliner that I wrote: `guix shell --container --share=$XDG_RUNTIME_DIR --preserve=XDG_RUNTIME_DIR --preserve="(WAYLAND_)?DISPLAY" emacs-pgtk -- emacs`. <dumbmf>The problem is that the entirety of the $XDG_RUNTIME_DIR is rw, and I'm not sure how to only share the portion needed for wayland instead of sharing the entire directory. <Ashy>guix pull: error: You found a bug: the program '/gnu/store/kn6cnl8x30bx1zb27ickskrz9fb4npzf-compute-guix-derivation' <Ashy>i've just installed guix on a fresh popos 22.04 install and `guix pull` crashes <Ashy>i tried with the version of guix packaged in apt and then uninstalled that and tried a fresh install with the guix-install.sh script <Ashy>is guix broken at the moment? <dumbmf>Ashy: I did have issues with substitutes earlier today, as well. Try again in a bit. <dumbmf>There were also 3 reports of this that I saw today: 70200, 70201, 70192 <rekado>oriansj: what is the exact build command? <rekado>Ashy, dumbmf: I can't reproduce this. I just ran "guix pull" successfully, 1 minute ago. <lnnk>Is icecat based on Firefox ESR? <nutcase>Hello Guix. What is the easiest way, to locally test an updated upstream version of a package (in my case: aqbanking, I'd like to test 6.5.8beta, while 6.3.2 is in Guix)? Is it via git clone of guix, modifying the package definition and then ./pre-inst-env etc? Or is there an easier way? <futurile>nutcase: probably the easiest way is to create a local package, of aqbanking and the package you're using it with (gnucash?), and to inherit the package definition of both. You can build those with guix build --load-path /your/directory aq-banking@6.5.8beta etc etc <ShalokShalom>I wanted to leave a quick review here. I am in the process of developing a small app, that will utilize a server like Nix or Guix.. <ShalokShalom>Guix disqualified by not only coming with an outdated database - that could be easily updated - but because the update process with guix pull takes an eternity. <ShalokShalom>Performance seems to be of low priority, and having quick access to the system is key for my application. <ShalokShalom>I also noted, that the web ui for the package search is not fast either. Hope you get that handled. <rekado>"guix pull" is in fact slow. It's not trivial to speed it up, but there's ongoing work to address that. <ShalokShalom>It is even slower than the other package search and doesnt find the package I am looking for <ShalokShalom>In the first attempt of using it, it simply directed me to the old search, even? <rekado>that's because it redirects to the other one in case it doesn't return in time <ShalokShalom>why wouldnt you ship an up to date database with the binary? <rekado>it will always be relevant to the version of Guix invoking "guix search" <ShalokShalom>when I install an app without guix pull, it still is capable of installing very old versions <rekado>"guix pull" installs a new version of Guix itself <rekado>there is no separation of Guix and some database of packages <ShalokShalom>And why not shipping the binary with an updated version in the first place? <nutcase>futurile: yes, I'm using it in conjunction with gnucash. I need to give my test package a different name then, right? How does gnucash know, which aqbanking to use then? <rekado>ShalokShalom: if you're using Guix 1.4.0 from long ago you'll get packages that were recent at the time of the 1.4.0 release. <rekado>this enables "guix time-machine", for example <rekado>you get an up-to-date Guix with "guix pull" <rekado>I agree that a new release ought to be made. <rekado>the release is only there to get you started; users are expected to regularly run "guix pull" <rekado>we're continuously building parts of "guix pull" on ci.guix.gnu.org <rekado>but if you're unlucky you might have to build things when updates were published in the meantime <ShalokShalom>just do a binary release every couple of months, not after over 2 years ^^ <rekado>you can force guix to only ever pull a version of Guix that has already been built on ci.guix.gnu.org <rekado>it only takes more than ~1 minute when your machine needs to compile Guix (and/or dependencies of it) from source <rekado>that source/binary transparency is also by design, because this way you are independent of a centralized build farm. <rekado>delete /gnu, /var/guix, /etc/guix. <rekado>I don't have much experience in removing Guix as I've never done it. <ShalokShalom>rekado: It wont let me delete /gnu, since its a read only file system <rekado>ah, the read only thing is to protect /gnu. It's done via the gnu-store.mount unit. <ShalokShalom>Like, why would you not make an uninstallation script, and the only way to find out how to remove this awful software, is by consulting a third party wiki? <mfg>which module would be most fitting for the intel compute runtime and it's dependencies? <futurile>he kinda has a point that it's weird we don't have a uninstall script - but what a rude human - oh well that's the internet for you <rekado>Robert Sapolsky says it's better (for the individual) to give ulcers than to get them, and that explains a lot of human behavior. <Ashy>rekado: it was just `sudo apt install -y guix && guix pull` <rekado>Ashy: does "guix pull" still fail for you? <futurile>nutcase: how committed are you to trying aqbanking / gnucash out - I have patches to update to 5.5 - I could send them bugs.gnu.org and you could test them? <rekado>his book "Behave" is full of quotable statements. Strongly recommended. <futurile>nutcase: it has aqbanking 6.5.4 and gwenfar 5.10.1 - but I don't use either so couldn't figure out how to run the online quote thing <nutcase>futurile: I use gnucash productively (since 2002 or so) and since today I can't (once again) get transactions from my bank (switched iTAN method 921->923?). Since GNU Guix System is my main (only) OS, I am more or less busted to help where I can. <rekado>ACTION restarts a seemingly never-ending chain of rust builds for the python-team spec <Ashy>rekado: i'll have to try again on monday, forgot to turn off sleep on that machine and it's in the office <futurile>nutcase: OK, I'll export/send what I have and you can see if the newer version works - and we can try and collaborate if it doesn't <rekado>civodul: don't know if you know this already, but I fixed all of the known JavaScript problems that I introduced in cuirass a few weeks ago. Just in case you want to upgrade the Inria instance to remove those bugs. <rekado>Ashy: okay. There's a chance that something is in fact broken when pulling from Guix 1.4.0 from Debian, but I cannot reproduce this. If this is still the case when you try this again, please do report back and we'll try to investigate further. <Ashy>rekado: no worries thanks for the help, i did remove the apt version of guix and then installed from the guix-install.sh script in the doco, and had the same issue running guix pull with that too <rekado>mfg: that's a GPU thing, right? If it's more than a single package (or more related packages are expected to be added in the future) perhaps it could get its own module? <dariqq>the org.gnome.shell.Screencast crash is driving me crazy. <civodul>rekado: i upgraded it a couple of days ago (i think i pinged you?), thanks! <dariqq>the only thing that i have after spending way to much time is not very helpful: Error: GDBus.Error:org.freedesktop.DBus.Error.Spawn.ChildExited: Process org.gnome.Shell.Screencast exited with status 1 <rekado>Ashy: odd. If this still happens on Monday please send the full output to bug-guix@gnu.org and ping us here. <Ashy>rekado: no worries, will do <rekado>dariqq: can you give me instructions on how to reproduce this? I'd like to see if I can find more details. <dariqq>rekado: i've been looking at the service in d-spy <rekado>civodul: oh, good. I have some more WIP patches for cuirass, but they need more time. (And they are a little more ... invasive and opinionated, so I'll take my sweet time to make them nice.) <mfg>rekado: yes it is. So if i create a new module would "intel" be to generic? <dariqq>and then fiound something to trigger the dbus service manually with "gdbus call --session --dest org.gnome.Shell.Screencast --object-path /org/gnome/Shell/Screencast --method org.gnome.Shell.Screencast.Screencast" which gives an invalid args error on my pc running fedora <dariqq>and on guix the error from above <dariqq>i am guessing that the screencast needs other packages to be installed and crashes after it does not find them <rekado>mfg: "intel" would be a bit too generic <rekado>dariqq: how do you launch the screencast thing? <dariqq>rekado: gdbus call --session --dest org.gnome.Shell.Screencast --object-path /org/gnome/Shell/Screencast --method org.gnome.Shell.Screencast.Screencast <dariqq>though I have no idea what i am doing <rekado>is there a "screencast" program to run? <mfg>rekado: would i tmake sense to rename the exisitng rocm module, since rocm fits the same prupose just for amd gpus? <dariqq>that's the problem: it should be available as an option for the gnome shell inbuilt screenshot (i.e. gnome quick-settings > thing next to settings) tool but the option is not there <dariqq>i presume because the dbus service has errors <oriansj>rekado: the exact build command was: guix build ghc@8.10.7 <mfg>is it possible to select specific package version in new-style input lists? <rekado>oriansj: does it also insists on building (on the machine where you imported the archive) when you do "guix build --no-grafts ghc@8.10.7"? <rekado>dariqq: here's what I did: cat $(guix build gnome-shell)/share/dbus-1/services/org.gnome.Shell.Screencast.service <rekado>that shows me $gjs/bin/gjs $gnome-shell/share/gnome-shell/org.gnome.Shell.Screencast <rekado>when I run that I get an error telling me that it doesn't know where the Gst typelib files are <dariqq>weird, i thought i ran that and it exited without errors <rekado>then it complains about gtk's typelib <rekado>so I try it in this shell: guix shell gstreamer gobject-introspection gtk <rekado>this no longer fails and strace shows me that it accesses a bunch of gst plugins that I have installed in ~/.guix-home <dariqq>now i am even more confused. If i run the $gjs/bin/gjs $gnome-shell/share/gnome-shell/org.gnome.Shell.Screencast nothing happens and i am back at a new bash prompt <rekado>d-feet has not been helpful. Just gives me the error message you got from d-spy. <rekado>I also use dbus-monitor on my user's dbus socket <dariqq>thanks, I'll see if I can find out something new in the meantime <civodul>rekado: you’re welcome to do whatever you consider good in the Cuirass UI :-) <civodul>so in the interest of moving forward, i’m tempted to just apply it <futurile>civodul: it 'works for me' - on a VM - you have to umount the store but after that it removes everything - it's better than no option in my opinion <civodul>futurile: do you mean it’s missing something like “systemctl stop gnu-store.mount”? <futurile>civodul: that part didn't seem to work for me - I can try it out again later? (my recollection might be a bit hazy) <nutcase>futurile: Ok, I'm looking forward to your input and I am happy, if I can help <bostx>Hi. `guix package --profile=/tmp/foo-profile --manifest=/tmp/manifest.scm` returns `guix package: error: rename-file: Is a directory`. Any idea why? <bostx>/tmp/foo-profile is an empty directory and /tmp/manifest.scm contains just `(specifications->manifest '("bash"))` and I'm in the shell created by `guix shell --container --nesting coreutils bash`. <bostx>Aaargh! The `mkdir -p "$GUIX_EXTRA_PROFILES"/my-project # if it does not exist yet` in the cookbook is rather misleading! :-( <bostx>because the profile specification `--profile="$GUIX_EXTRA_PROFILES"/my-project/my-project` then contains `my-project` twice. <rekado>bostx: my own convention is to always name the profiles .guix-profile. In the above command the directory you created is not actually the profile directory (that's a link created by Guix) but only a place for all those links to be contained. <henk_guix>Hi guix, this must be easy, but I can't find it. How can I set environment for use during the build in a package defintion? <henk_guix>(I mean environment variables), specifically I need to set the CONFIGURE_SHELL variable to be run by config. I've tried to (invoke "CONFIGURE_SHELL=$(which bash)") but then I get a 'file not found', so I'm doing something wrong <jpoiret>henk_guix: you need to use (setenv ...) <jpoiret>invoke just launches a subprocess, which 1) would need to point to an actual executable 2) wouldn't be able to modify the env variables of the parent process <jpoiret>SOME_ENV=value is some bash syntax, you can't just invoke that <jpoiret>(it's the bash equivalent of setenv actually, modulo exporting) <henk_guix>jpoiret, I'm trying this now, but can't quite get it to work - I assume it has to be in a lambda, so now I'm setting it in a build phase right after 'unpack <henk_guix> (lambda _ (setenv "CONFIG_SHELL" (which "bash"))))))) <rekado>henk_guix: change it to (arguments (list #:tests? #f #:phases #~(modify...))) <rekado>henk_guix: when you use quasiquote the whole expression becomes data <rekado>this means that #~ becomes just line noise instead of a G-expression <rekado>by using (list ...) it all remains code <rekado>in this particular case you're not making use of the G-expression so alternatively you could keep the quasiquote and remove the #~ <henk_guix>rekado and jpoiret, that's amazing thanks! Building has commenced <Guest89>Hey, is there a reason developers should use Guix over NixOS? <Guest89>Heard about Guix few times in the past, but honestly don't know much about it <civodul>we lost an opportunity to convince Guest89 <zamfofex>What would you have said if they were still around? <civodul>i’d have argued that package transformations and ‘guix shell’ rock <zamfofex>Are package transformations not available in some form in Nix? (That’s a very important Guix feature, I’m surprised people can live without it in other distros.) <civodul>ultimately, Nix is certainly good as well, maybe a bit less approachable, it’s a matter of seeing which one works best for them <civodul>in Nix one can programmatically customize packages, it’s just not as easy as transformation options, AFAIK <civodul>(that’s in part because there’s no “package” abstraction in the Nix language, to begin with) <zamfofex>I feel like Guix is really well designed. <zamfofex>Maybe I’m biased, but whenever I hear some difference between Guix and Nix, it’s always something that favors Guix. <rdrg109>[Question] I'm running the installation image on a laptop. I want the installation image to retrieve packages from a substitute URL. How to make the Guix daemon, which is running in the installation image, aware of the subtitute URL? <civodul>rdrg109: hi! i’m not sure the “GUI” in the installation image lets you select substitute servers <civodul>if you run the manual installation process, you can always pass ‘--substitute-urls=…’ to ‘guix system init’ <civodul>(provided you first authorized any additional servers) <rdrg109>civodul: Thanks! I forgot the fact that I could pass '--substitute-urls=' to 'guix system'. I am going to use that. <RavenJoad>I just did a guix pull after quite some time, and kdelibs4support is failing test 27 - kmimetypetest. Should I report this somewhere? <zamfofex>RavenJoad: It sounds liek a bug, so I imagine yes. <IcePic1984>Hi! Can someone tell me if it is possible to "cross deploy" using guix deploy from x86 (localhost) to aarch64 (remote)? <IcePic1984>I managed to cross compile an arm image and boot it. However `guix deploy` fails with `building /gnu/store/784zwc1cp3y9nnx077iywbnagbn2v1jm-remote-assertion.scm.drv... <IcePic1984>\builder for `/gnu/store/784zwc1cp3y9nnx077iywbnagbn2v1jm-remote-assertion.scm.drv' failed with exit code 1 <IcePic1984>build of /gnu/store/784zwc1cp3y9nnx077iywbnagbn2v1jm-remote-assertion.scm.drv failed` <brisk-pm>Hi, I have questions about finding procedure and module information from within `guix repl` (maybe this is more guile related than guix): <brisk-pm>1) Is it possible to inspect the source of a given procedure or generate more detail than say `,describe <procedure>`? <brisk-pm>2) Can I pull up the guix reference information found in the info guix page or do I have to do that in a separate window? <brisk-pm>3) When loading a scm file into the repl, can I make live changes to the file? Or is this only something that's possible with a gieser/emacs combo? I tend to gravitate towards vim and currently edit file, close, open in repl, and test run. Repeat. <futurile>brisk-pm: 1) AFAIK there is no way to get the 'source' of a function in the repl (e.g. guix repl) -> maybe you can do it as an 'emacs user' which the manual covers <futurile>brisk-pm: 3) is what I'm doing - for the reason you mentioned - I just use tmux and a window <brisk-pm>Gotcha, I think I had an expectation about a "live coding" experience in a repl that doesn't necessarily match reality. Thanks. <Guest36>How do I disable resolveconf to stop touching resolv.conf? <Guest36>I have NetworkManager installed though <podiki>anyone with issues with v4l2loopback-linux-module building for 6.8? <podiki>/tmp/guix-build-v4l2loopback-linux-module-0.12.7.drv-0/source/v4l2loopback.c:717:9: error: implicit declaration of function 'strlcpy'; did you mean 'strscpy'? [-Werror=implicit-function-declaration] <brisk-pm>Guest36: Are you using `network-manager` for your configuration? Something like <brisk-pm> (dns "none")) can stop it from managing resolv.conf <brisk-pm>Guest36: Are you using `network-manager` for your configuration? Something like <brisk-pm> (dns "none")) can stop it from managing resolv.conf``` <brisk-pm>Guest36: Are you using `network-manager` for your configuration? Something like <brisk-pm> (dns "none")) can stop it from managing resolv.conf` <brisk-pm>Guest36: Are you using `network-manager` for your configuration? Something like <brisk-pm>`(network-manager-configuration (dns "none"))` can stop it from managing resolv.conf <brisk-pm>oops, slack muscle memory made me "edit" the post <podiki>re v4l2loopback: i think solved with update, will do that <Guest36>I don't have configured network. I use simply desktop-services. NetworkManager isn't configured either. I only have the package installed and ran sudo nmtui to set DNS servers, though resolveconf always adds a search domain that conflicts with the set dns server <brisk-pm>If nmtui runs and there is not a network-manager configuration declared, then the default will add nameservers from all active connections to resolv.conf <ieure>Q1: Is there any way to avoid compiling the whole Linux kernel *during* a Nonguix install? After it's installed, it'll use substitues, but the Guix that runs during installation doesn't know about them. <ieure>Q2: I got this failure trying to install last night: https://paste.debian.net/1313231/ It smells like a Guix bug, but I don't know if I can get any help with it, since the installer wasn't running linux-libre. <ieure>I get the same failure with both the stock linux-libre kernel and nonguix linux (but it takes like an hour to compile in the latter case, see Q1). <ieure>Both the nonguix and system crafters installers failed the same way. <ieure>I've successfully installed other machines with the exact same image before, so not sure what's wrong here. They `guix pull' at the start, and there's a Guix bug, I guess? <Guest36>ieure: Do you use the installer ISO? This should have substitutes already declared and just downlods linux <ieure>Guest36, As I said, I tried both the nonguix installer and the system crafters one. Neither seemed to have substitutes. <ieure>Well, they had *guix* substitutes. But not *nonguix* ones. <Guest36>ieure: System crafter have nonguix substitutes as well. If they are not working it means there is something on with the server. Just try guix weather linux, if down report it <ieure>Guest36, Okay, I can try that. Any thoughts on the installer failure? That's a separate issue. <Guest36>ieure: Nope, sadly this is above my skills. I only have experiences with the ISO installers, because sometimes the subistutes did not work for me, too. Sometimes they have problems with the build farm just like guix has <ieure>Guest36, It's not a build issue, this happens after everything has been downloaded/built (if necessary). <Guest36>ieure: I was referring to the first issue (building Linux, it should never build unless substitutes are disabled or you changed kernel config) <Guest36>or you require aarch64 Linux. They don't have that available, too <ieure>Oh, I put this all in the wrong channel. Sorry :/ <bdju>are guix home services used even if I don't use guix home? I handle config files by hand. but I don't know how to set up pipewire for example. and I used to have a syncthing service copied from an example in guix blog post but I think it stopped working at some point <apteryx>is it possible to control whether some tests are run serially vs in parallel for automake tests? <dariqq>rekado: you can run the gjs command with GST_DEBUG=9 to get a lot of debug output but his was also not really helpful. Also found a lot of reports from others of missing pipewire but at least the thing told them that pipewire was missing (i.e. via dbus monitor) and not just crashing. <apteryx>ACTION asks the same question in #autotools, which is probably a better place <lechner>apteryx / would it be acceptable to run all of them in serial? <lechner>Hi, is there a way to make Guile modules available in Geiser without adding them to the user profile? Emacs is my window manager <graywolf>lechner: you can configure geiser-guile-binary <mekeor>lechner: perhaps geiser-guile-load-path? <mekeor>also consider using emacs buffer-env package, it works with guix' manifest.scm <lechner>mekeor / yeah, i used buffer-env for a while but it gets loaded even in magit <mekeor>lechner: oh right, maybe we should change (add-hook 'hack-local-variables-hook #'buffer-env-update) to (add-hook 'hack-local-variables-hook #'my-buffer-env-update-if-buffer-file) <mekeor>lechner: also, can we make "peanuts" bot less talkative please? it should only talk when told to do so. <mekeor>lechner: i defined my-buffer-env-update-if-buffer-file as (when buffer-file-name (buffer-env-update)) and it seems to work well <oriansj>is there an option in guix build to skip running of tests? <oriansj>I see --without-tests=pkg would that do it? <rekado>dariqq: that just tells me that gstreamer was initialized successfully (in so many words) <rekado>I'm reading the source code of the Screencast service <rekado>it says that it needs the "pipewiresrc" element; that's a gstreamer source. <rekado>libgstpipewire.so exists, but I don't see it in the GST_DEBUG output <rekado>"0:00:00.043734961 7525 0x2440350 LOG GST_ELEMENT_FACTORY gstelementfactory.c:145:gst_element_factory_find: no such element factory "pipewiresrc"" <rekado>"rm -rf ~/.cache/gstreamer-1.0/" got me past that error <dariqq>i have ( after adding pipewire to system profile): GST_REGISTRY gstregistrychunks.c:741:gst_registry_chunks_load_feature: Added feature pipewiresrc, plugin 0x1517400 pipewire <rekado>now it doesn't exit right away but lingers for a few seconds <rekado>and now "dbus-monitor --address unix:path=/run/user/1000/bus" actually shows me something relevant <rekado>now it complains about not finding services that are provided by gnome-shell, so I'm adding that to the "guix shell" command <rekado>I see that it's asking for the screen size and for whether screencasts are supported. I see a positive response on dbus. <rekado>dariqq: the screencast feature works... when I run "dbus-run-session -- gnome-shell --nested --wayland" inside a "guix shell gnome-shell gtk gstreamer gobject-introspection" <rekado>this runs gnome-shell in a window <rekado>(a little difficult to close it then, but I somehow managed to do it anyway) <rekado>so my guess is that gnome-shell may need to run in an environment where more gi services are available <dariqq>rekado: with the extra dbus-run-session (even outside of any additonall shell) it works for me <rekado>as in "dbus-run-session -- gnome-shell --replace"? <dariqq>as in dbus-run-session -- gnome-shell --nested --wayland <dariqq>needs pipewire in the profile though <apteryx>lechner: this will make all tests serial no? <apteryx>the manual mentions it's deprecated by the parallel test harness <dariqq>and without the pipewire home-service the button is still there on the second session but i get errors for missing pipewire running <nutcase>How can I apply a patch series from issues.guix.gnu.org to my local guix repository?