<two[m]>it prints `guix refresh: error: mkstemp: Denied in access` <two[m]>i am trying to refresh a package <ardon>Hi Guix! Is there an equivalent of "Using the Offload Facility" (guix deploy) for Guix Home? I'd like to build my home in my desktop machine and pass it to a resource-limited machine. I've been experimenting with `guix copy' but I'm not sure how to go about after copying the generated home profile <rekado>offloading work for any and all builds <rekado>it is not restricted to a limited set of commands <antipode>nckhexen: It's building a chain of rusts, looks 'good' <nckhexen>two[m]: Using ‘guix refresh -u’ requires a writable Guix git checkout (in practice: ./pre-inst-env guix refresh -u). It will never work without that, but that's not made very clear. Could certainly use some UX improvement. <antipode>Because its building rust to build a newer rust to build ... <nckhexen>But we like those too. Keeps the house warm. <ardon>rekado: Yeah, I botched that. I more-so meant if there's an easy way to build a home profile in one machine and install it in another. From what I read in the manual, I'm not sure what I command I would need to call after copying a profile or a package. <antipode>It works, I just wish the 'Official Rust People (tm)' kept a C implementation around <nckhexen>Bootstrap bottlenecks aside, worker utilisation still ‘feels’ low… <antipode>I looked around at core-updates, staging and master, and it seems that almost stuff is built or failed (on x86_64). <antipode>There's at least 6154 of unbuilt stuff on master though <nckhexen>Beefy boiz. But now the big PPC64LE node has dropped off the map again ☹ <two[m]>how do i get the current guix git path so that i can clone it? <nckhexen>I'd just clone it from Savannah: ‘guix describe’. <antipode>I don't think any additional x86_64 nodes will be needed anytime soon. <antipode>two[m]: WDYM with 'current guix git path'? <antipode>two[m]: "guix pull" downloads from savannah. <nckhexen>antipode: Absolutely not. If there is an issue it will be due to bugs, not a lack of hardware. <antipode>Eventually, yes, though not a git repository, just a checkout. <nckhexen>You can find the checkout in ~/.cache/guix/checkouts but really, unless you're surfing through a 3G 'phone tied to a local tree, why bother. <lilyp>in fact, it is not even a shallow checkout, because guix has to check all the signatures <nckhexen>…oof. guix-daemon ran out of OOM on the builder… <two[m]>there isn't a pre-inst-env there? <nckhexen>That's weird. Out of memory: Killed process 1716703 (cc1plus) total-vm:3556992kB, anon-rss:3282496kB, file-rss:0kB, shmem-rss:0kB, UID:994 pgtables:452kB oom_score_adj:0 <antipode>two[m]: pre-inst-env is a generated file, you need to run ./configure --localstatedir=/var first <nckhexen>This better not be some systemd/cgroup ‘feature’. <antipode>(outside ~/.cache/guix/checkouts, otherwise you'll confuse "guix pull") <antipode>'22.1 Building from Git' has a detailed list of instructions <two[m]>./bootstrap: line 8: find: command not found <two[m]>./bootstrap: line 8: sed: command not found <antipode>two[m]: Don't forget the "guix shell -D ..." listed in the instructions <two[m]>i mixed up --container and --development <two[m]>`--development --container guix` somehow successfully built a wrong shell <two[m]>./pre-inst-env: line 55: exec: guix: cannot execute: Is a directory <nckhexen>…apparently this is a systemd ‘feature’ (OOMPolicy=stop). Good lord. <nckhexen>If any child process runs out of memory, systemd will shoot your service in the face. <nckhexen>It in no imaginable way makes any sense. <nckhexen>If a process runs out of memory, the build fails, the daemon logs the failure, marks the build as failed, and happily continues building. systemd *kills the daemon*. <two[m]>e.g. a shell script may check if the exit status of `grep 'pattern' "${file}"`is 0, and in case of grep ooming on its own, it would assume the file not to contain the given pattern and cause errors <pkill9>guix git says some packages fixed so their tests pass on gnu/hurd, does this mean guix is running on the hurd? <two[m]>it doesn't for this daemon in particular, but i think it would for most programs <pkill9>(able to do so i mean, not by defualt lol) <two[m]>may even potentially cause nondet builds if the exit status of a memory intensive program is used in a build script <nckhexen>That's no excuse for killing the script. <nckhexen>Anyway, I've added ‘OOMPolicy=continue’ on that one node to work around this bug, but we should probably add it to the upstream service.in. <rekado>pkill9: Guix works on the Hurd for quite some time. <two[m]>(but it's a lot of text instead of sleep and it's killed by oom instead of me) <Andronikos>guix gc: freed 19,858.86299 MiBs that is a weird notation. I assume this is US notation but why so many digits after the period? <Andronikos>Also it says it freed 19 Gibibyte but my df -h shows available disk space went from 11M to 14G. <rekado>note that there’s also deduplication <two[m]>> note: currently hard linking saves 26101.80 MiB <two[m]>that guix doesn't use native number format <lilyp>i think the weird thing is that it's using floating points instead of fixed points <Andronikos>Yea sure. I have this behaviour since something like three months. <two[m]>bash: printf: 1.0: неправильне число <Andronikos>Thought maybe something got changed since it worked. <nckhexen>…why does the systemd unit specify ‘RemainAfterExit=yes’ for guix daemon/publish? <two[m]>at least that's what standard printf does? <two[m]>but guix somehow inserts a comma into a sexp does it? <nckhexen>It's just a thousands separator. (Ehm, ‘1.000000’ to mean ‘1’ is not ‘fine’, no…) <Andronikos>How do I need to format logs if I send an email to guix bugs? <two[m]>i mean standard printf uses a comma <nckhexen>Andronikos: If it's a significant chunk o' log, attach it. Otherwise, in-line is fine. <nckhexen>No need for ‘formatting’, mail is plain text. <Andronikos>But some people have this weird -----cut here---- syntax. That's why I am asking. <nckhexen>Some people like to get fancy with their emacs -------8<---cut-here----8<---nom-nom-i-am-scisor--- stuff though. <Andronikos>Yea right I mean that. This is not a requirement? <Andronikos>Ah. Always thought why I can't find any documentation for it. <two[m]>Andronikos: i think three backtics and begin and end do it <vagrantc>i pretty much have formatted emails in just about whatever way i like and haven't gotten too many complaints :) <nckhexen>But OTOH, if it makes you happy, use whatever you want. No right answer means no wrong answer, as long as it's not too noisy. <vagrantc>helps that most guix devs are at least one giant pond away. <nckhexen>vagrantc: We can't throw that far with our puny hacker arms. <vagrantc>besides, better to compost all that rotten fruit locally for the environment <two[m]>where does 'currently hard linking saves' even come from? <two[m]>i only get binary file matches when grepping <nckhexen>This is also where the number formatting above comes from. So feel free to submit patches, but don't expect too many people to jump at the chance of fixing C++ just for cosmetics. <nckhexen>(Back story: we don't like C++. End back story.) <nckhexen>I'm not sure where you're grepping. It's in the git repository. <two[m]>i mean ,/. is important because a user might just assume a wrong amount <podiki[m]>i only used that cut here stuff for the first time yesterday <nckhexen>The code *is* maintained, but only when strictly needed. <podiki[m]>since the formatting shows up nice in emacs email (nice color background, easier to parse from main text) <nckhexen>two[m]: We, collectively, know little C++ as well :) <podiki[m]>it is "message-mark-inserted-region" command on a region to surround it with that (customizable), as I learned <nckhexen>Or if you mean the next line: the nix daemon code is used, and not exactly bitrotted, but it's only touched when necessary, not refactored (or localised with pretty number formatting) for fun. <Andronikos>podiki[m]: Thanks for that information. Will be useful for the next time. <nckhexen>‘"Face holding back tears" emoji looks like a concerned face with big 👀 Eyes that are about to shed 💧 Tears. The facial expression conveys a moment of emotional vulnerability, an inner struggle, and tension on the verge of breakdown.’ <Andronikos>Is Nvidia offtopic? It seems that Vulkan does not detect my GPU in a "guix shell -FC". <nckhexen>Andronikos: Specific vendors aren't off-topic, only encouraging or providing tech support for non-free software is. <two[m]>i mean gettexting this would also be very appropriate since one might interpret "saving" as downloading <two[m]>downloading 20 gigs no cancel prompt hehehe <vagrantc>then again, every once in a while someone pipes up about a guile reimplementation of guix-daemon <podiki[m]>Andronikos: with -C you need to expose/share from the host, things like /dev/dri probably <two[m]>podiki[m]: doesn't mounting /dev fail with error? <podiki[m]>you can't do all of /dev I don't think (some is already set up? I forget why) <podiki[m]>this should get you most of what you want typically: <podiki[m]>--preserve='^DISPLAY$' --preserve='^XAUTHORITY$' --preserve='^DBUS_' --share=$XAUTHORITY --share=/sys/dev --share=/sys/devices --share=/tmp --expose=/dev/dri --expose=/var/run/dbus <two[m]>you can bind individual devices into it? <podiki[m]>little tricky to figure out sometimes, but yes, it is pretty great <podiki[m]>e.g. what /dev/hidraw* you want for certain hardware <two[m]>--preserve=XDG_RUNTIME_DIR --preserve=WAYLAND_DISPLAY --expose="$XDG_RUNTIME_DIR/$WAYLAND_DISPLAY" <clever>podiki[m]: evtest can help with input devices <two[m]>the two dollars above caused it to be latex math <podiki[m]>(my example was from the world of X, indeed) <nckhexen>You need a better client two[m]. You deserve it. <podiki[m]>clever: thanks, will check it out! usually I just share all of hidraw, but if hardware is un/replugged or reset the numbers change and have to share the new numbers <two[m]>so ` --preserve=XDG_RUNTIME_DIR --preserve=WAYLAND_DISPLAY --expose="$XDG_RUNTIME_DIR/$WAYLAND_DISPLAY" ` but with ^$ also <nckhexen>Searching, I find some references to other /dev nodes (like nvidia0), but honestly a lot of the results are barely literate and don't look very reliable. ‘nvidia won't work in my dockers’ OK <Andronikos>podiki[m]: I have already that. The card is shown with lspci. <clever>lrwxrwxrwx 1 root root 10 Oct 29 20:26 /dev/input/by-id/usb-0430_Sun_USB_Keyboard-event-kbd -> ../event19 <Andronikos>Debug of vulkaninfo tells me that it can't find a valid GPU. <two[m]>no, like, i do want to see latex math, but i've never used it tho <clever>podiki[m]: symlinks like this are also auto-created usually <podiki[m]>Andronikos: others: /dev/nvidiactl /dev/nvidia-modeset /dev/nvidia0 /sys/class/drm and so on <nckhexen>two[m]: Latex math is just very rare here, whilst $s aren't, so you might be optimising for the wrong #content. That's all I mean. <clever>Andronikos: strace vulkaninfo, and see what files its checking for <nckhexen>‘Just’ showing up in lspci doesn't imply you've exposed enough. <podiki[m]>there are some vulkan patches btw, I believe some files need to be patched with absolute paths, mainly for vulkan developing I think (not for using, I use vulkan but not nvidia) <Andronikos>guix shell --container --emulate-fhs --network --manifest=manifest.scm --preserve='^DISPLAY$' --preserve='^XAUTHORITY$' --preserve='^DBUS_' --share=$XAUTHORITY --share=/sys/dev --share=/sys/devices --share=/home/chris --share=/tmp --share=/dev/dri --share=/dev/nvidia0 --share=/dev/nvidiactl --share=/dev/nvidia-modeset --share=/dev/bus --share=/dev/shm --share=/proc/self --share=/var/run/dbus --expose=/etc/machine-id <Andronikos>and "VK_LOADER_DEBUG=all vulkaninfo &> vulkaninfo.txt" says: Failed to detect any valid GPUs in the current config <nckhexen>I agree with the strace suggestion. Find out *what* it's ‘detecting’. <clever>Andronikos: try `VK_LOADER_DEBUG=all strace -e open,openat -ff -o logfiles vulkaninfo` and then grep thru the logfiles to see what returned ENOENT <Andronikos>Also my container is stupid. It goes through all kinds of paths to find libraries. How can I fix that? <Andronikos>Do you use AMD? Are those drivers actually better? <two[m]>guix shell mesa-utils strace -- strace -o >(grep -E -o '"/(dev|sys)/.*"' | sort | uniq) glxinfo >/dev/null <two[m]>someone has already sent an strace <podiki[m]>put the -F option last in the list after all packages, as a workaround; or use LD_LIBRARY_PATH=/lib:/lib/nss <Andronikos>Well it looks like it can find all libs it needs. <nckhexen>Andronikos: The notation bug report doesn't really explain what the problem is. ‘7,294.56603 MiBs’ should be ‘7,294.56’ MiBs? Or…? <Andronikos>error: XDG_RUNTIME_DIR not set in the environment. Is this critical? <Andronikos>Well I assumed this is already clear. If US notation is used yes. <nckhexen>It said ‘broken’, so I assumed it wasn't only ‘too accurate’. <nckhexen>I think an ‘expected output’ would help, yes, but maybe I'm just dim and it's obvious to everyone else :) <Andronikos>The only thing that helps me with vulkaninfo is the following: ERROR: setupLoaderTermPhysDevs: Failed to detect any valid GPUs in the current config and ERROR: setupLoaderTrampPhysDevs: Failed during dispatch call of 'vkEnumeratePhysicalDevices' to lower layers or loader to get count. as well as ERROR at /tmp/guix-build-vulkan-tools-1.2.162.drv-0/source/vulkaninfo/vulkaninfo.h:248:vkEnumerateInstanceExtensionProperties failed with <Andronikos>Oh I maybe should mention that I use Guix on a foreing distro. <podiki[m]>you'll need to share with container drivers (libraries) I think, but I'm not sure if it will work exactly <podiki[m]>there's the whole mesa/libglvnd that guix doesn't do right now (mesa dispatching to vendor gl libraries) but I don't know if that is relevant for vulkan here <podiki[m]>yeah...I would try without the container, but I bet it needs what we don't do right now <Andronikos>I think I know why. There is no ICD for Nvidia in /usr/share/vulkan/icd.d but exposing it will still result in missing the file. How can I copy it to the container? <podiki[m]>you can expose/share something and specify what path it should appear in the container; or there's I think a vulkan icd path env variable you can set; still I would try without the complications of the container first, unless you got that to work ***lukedashjr is now known as luke-jr
*podiki[m] really ought to work now <Andronikos>You also mentioned that bug and I should put -F after packages. If I use the -m arg does that mean -m -F? <podiki[m]>I would just put it last just in case for now <nckhexen>Andronikos: Why will exposing it still result in missing the file? <Andronikos>podiki[m]: Does not work I still need to set it manually. Guess I need to wait for the patch. <Andronikos>nckhexen: Good question. I don't know why. I even exposed just this specific file but it won't work. <podiki[m]>what does not work? the -F bug I mentioned is just for glibc (e.g. a version that uses /etc/ld.so.cache should be in the container but may not); nothing to do with vulkan exactly <podiki[m]>again, on a foreign distro I would try without a container first, does that work? <Andronikos>podiki[m]: You mean just guix shell? Well I can't really test that since it requires FHS <Andronikos>nckhexen: No but it is owned by root. Maybe that is the cause. <podiki[m]>I thought you were on a foreign distro? presumably with typical FHS setup? <podiki[m]>I don't know what you are trying to do, so I'm just guessing <Andronikos>podiki[m]: Oh you mean if the software runs on my system? Yes on my system everything works as expected. <podiki[m]>so you just want to run the same thing in a guix shell? this thing being e.g. some binary? <Andronikos>I am trying to run it in a guix container since it requires gnome-keyring and I don't want that on my main system. I need Vulkan since it is a launcher for a video game. <podiki[m]>(as in want to run it in an isolated container rather than on your host system, and want to use a guix container) <podiki[m]>well, again, no idea if nvidia vulkan will work like this; first step would be to use something like vulkaninfo/tools to see that you get vulkan when running that from guix <podiki[m]>so just "guix shell vulkan-tools -- vulkaninfo" and figuring out what that needs to work <podiki[m]>but again, I don't know if that will work with guix as things are currently (and even if it does, likely you need your host's GL drivers to be used, and that won't as far as I know) <Andronikos>Currently I want the icd to be copied since without that Vulkan will not the driver. I tried exposing it but it does not work (guess because it is owned as root). Therefore I copied the file and tried "--expose=nvidia_icd.json=/usr/share/vulkan/icd.d/nvidia_icd.json" which gives permission denied. <Andronikos>Running " guix shell vulkan-tools -- vulkaninfo " gives "ERROR: [Loader Message] Code 0 : libGLX_nvidia.so.0: cannot open shared object file: No such file or directory" <podiki[m]>the icd also points to some libraries that will need to be accessible <Andronikos>I mean on my container setup but not on the cmd from you example. <podiki[m]>interesting that it does try to load the glx from nvidia; would need to strace probably to find what it fails at <podiki[m]>anyway, have to go, good luck (though sadly think it isn't currently supported, but I don't know for sure) <podiki[m]>could try posting to help mailing list as well for visibility and details of what you are doing <Andronikos>Well it is some proprietary game launcher as well as proprietary Nvidia drivers. Doubt that this is okay. <podiki[m]>the general question of using drivers from a foreign distro should be fine <Andronikos>Don't understand why you example loads the nvidia icd. I have vulkan-tools installed on the container as well.. <Obikawa>Hello. I have managed to write a package definition for a simple Gnome Shell Extension which successfully executes all phases of $ guix package -f package.scm . Now I wish to include it in the package list of home-configuration.scm . Is it possible to do so without having to modify a local copy of the GUIX repository? ***jpoiret9 is now known as jpoiret
<lilyp>Obikawa: sure, just (define my-guix-extension (package ...)) in your home.scm and refer to it <rekado>sneek: later tell Andronikos the scissors lines are made by selecting a region and then hitting C-c M-m (message-mark-inserted-region). ***dragestil_ is now known as dragestil
<vivien>Hello guix! I made my /etc/hosts on guix system with both IPv4 and IPv6 for localhost, but when I run: (getaddrinfo "localhost" "http") I get only 2 entries, both IPv4 addresses (presumably, one entry for TCP and one entry for UDP). How do I get both IPv4 and IPv6? <vivien>Oh I just noticed that it depends which comes before in the hosts file, the IPv6 or the IPv4. So I guess it’s a limitation of the libc ***jonsger1 is now known as jonsger
<vivien>So, there should be a file at /etc/host.conf that contains "multi on" to let getaddrinfo return both IPv4 and IPv6, not just the first entry it finds in /etc/hosts. <vivien>Maybe I will have to reboot first. ***Dynom_ is now known as Guest7625
<cbaines>the test suite seems to fail, but I'm not sure why (there aren't any failing tests?)? <PurpleSym>cbaines: I see a stray `../build-aux/test-driver: line 107: 2519 Aborted (core dumped) "$@" > $log_file 2>&1` in there – not sure if relevant. <cbaines>those lines look like they could be related to the XFAIL tests <PurpleSym>Yeah, a few tests fail with an assertion failure, which will probably abort(). <cbaines>I do wonder if the XPASS could be causing the build to fail <cbaines>it looks like the test-tls is marked as XFAIL, but it doesn't fail <unmatched-paren>Could we add a #:environment-variables argument to all gnu-build-system-based build-systems? It would make it much easier to <PurpleSym>cbaines: Looking at the Makefile XPASS must be zero indeed. <cbaines>cool, I'll look at removing test-tls from the XFAIL_TESTS then :) <xd1le>Hi, I seem to be stuck on something simple with using package modules via -L but am not sure what I'm doing wrong. <xd1le>Scary thing is I'm pretty sure this used to work and I don't think I did a guix pull or anything since then. <Korven[m]><xd1le> "What am I missing?" <- Yeah that happens a ton, unfortunately the errors aren't the most accurate <Korven[m]>and then you can do `,use (my-packages emacs-xyz)` and figure out the actual error <xd1le>unmatched-paren, Korven[m], ty both <xd1le>I really should be using this new guix repl command more <xd1le>turns out emacs-build-system was unbound, found it out using guix repl like Korven suggested <Korven[m]>Anyone managing emacs packages with guix got tree sitter working? <Korven[m]>it's just trying to write out stuff to the GNU store and that ain't flying <xd1le>oh yeah guix repl isn't new.. I just forgot about it.. what was I thinking about.. <xd1le>Korvarx[m], actually the above is me starting to switch to using guix for my emacs packages, so I could try out tree sitter tomorrow probably (now I need to sleep) <xd1le>are you having issue with tree sitter or the emacs tree sitter package <xd1le>you mean the emacs package yes <Korven[m]>it is mainly it trying to write to the GNU store <unmatched-paren>Korven[m]: there is a (probably bit-rotting) patchset adding emacs-tree-sitter and several tree-sitter grammars <Korven[m]>well the emacs-tree-sitter patch/package? is installed, if that's what you're saying <Korven[m]>i forgot I added it from the `guix-emacs` channel, which builds stuff from melpa <xd1le>it's probably trying to compile grammars into the store? <Korven[m]>i think so, it tries to create a bin directory in there <xd1le>that's my guess, but the package looks a bit involved <unmatched-paren>that's why we need to package tree-sitter grammars for guix to make it work <Korven[m]>plus the tsc-dyn.so module relies on fhs type library deps <unmatched-paren>but doing so is nontrivial as it requires modifications to the node-build-system <Korven[m]>i literally just started using Guix the day before yesterday, i would have no clue where to start lmao <xd1le>I remember glancing the source to this when the author gave his talk <Korven[m]>Also, ,I could use emacs stuff on my foreign distro IG <unmatched-paren>Korven[m]: That would just be written with something even worse like M4. <Korven[m]>that's fair, I'm just trying to use some lisp to configure my system XD <xd1le>yeah I think he decided to use javascript as a way to advertise how you can easily write grammars for your favorite language.. because everyone likes javascript right. <xd1le>I might be wrong, just off the top of my memory <xd1le>I guess the bigger culprit is just how difficult it is to package node things, but yeah he definitely could have just used rust. <xd1le>Korvarx[m], "i literally just started" <- Welcome! Yeah unfortunately node ecosystem has been kind of a pain to deal with. <xd1le>Not sure how the situation is now though. <Korven[m]>well i'm having problems with emacs-juypter too (wooo!) so we're off to a great start XD <Korven[m]>You appear to be setting environment variables ("PATH") in your .bashrc or .zshrc: those files are only read by interactive shells, so you should instead set environment variables in startup files like .profile, .bash_profile or .zshenv. Refer to your shell’s man page for more info. Customize ‘exec-path-from-shell-arguments’ to remove "-i" when done, or disable ‘exec-path-from-shell-check-startup-files’ to disable this message. <Korven[m]>I get this, maybe that's why it can't find ZMQ. Any ideas? <xd1le>unmatched-paren, well yeah exactly, that's probably ideal <xd1le>Korvarx[m], are you setting PATH in .bashrc though? <Korven[m]>it's being set by guix home, and I'm adding to it in my .zshrc <xd1le>Korvarx[m], maybe try setting PATH in .zshenv instead and see if it works? <xd1le>I forgot what the filename is for the zsh profile file <xd1le>ah ok it seems to work differently in zsh, doesn't seem to be a profile file. Anyway you could still try setting it in .zshenv <xd1le>Korvarx[m], are you on Guix System? <xd1le>nevermind yeah there is a .zprofile <xd1le>ah so what happens when you try to boot? <Korven[m]>i get to the login screen, key in my password <unmatched-paren>Korven[m]: yeah sometimes host env vars and guix env vars can collide <xd1le>maybe you did an update and things broke and xserver won't start? <xd1le>so you would have to find the xserver logs for your distro <xd1le>that usually used to be the culprit for me but now I use wayland mostly ***darosior7 is now known as darosior
<Korven[m]>it should only GC stuff not being referenced right? <Korven[m]>if i do a Guix home reconfigure after that, it just downloads everything in my profile again <Korven[m]>like why did it have to download it all again? surely it didn't delete actively referenced things <lechner>Korven[m]: I also find gc aggressive at times, but mostly when things need to be built locally <unmatched-paren>Korven[m]: well, there's probably something similar with env vars happening <Korven[m]>the reconfigure is downloading the universe, so I'm waiting <minima>hi, anyone has any recommendation for handling clojure dependencies in guix? <minima>i suppose clojure can fetch and install a project's dependencies via deps.edn but i'm wondering if that can be actually delegated to guix <minima>that's a bit of a rhetorical question probably... of course it can... my real question is: i don't seem to see many clojure libraries packaged under guix - any specific technical reason for that? <unmatched-paren>although, leininegen isn't packaged because of a bootstrapping issue <minima>unmatched-paren: oh ok, so if there's a build system i might be able to create some guix packages by myself (and possibly create a PR) <minima>oh... a build system but not an importer... ***jesopo is now known as jess-o-lantern
<Korven[m]>I added a channel to guix. (guix-emacs, for melpa packages). But my I can't use it from within a `guile` repl or `geiser` repl, it says the module has no code. <Korven[m]>Anyone know how I can add it to the GUILE_LOAD_PATH? <Obikawa>Korven[m] - (home-environment (services (home-bash-configuration (aliases '("GUILE_LOAD_PATH" . "/path"))))) ? <Obikawa>It is (home-environment (services `(,(service home-bash-service-type (home-bash-configuration (environment-variables `("VAR" . "path"))))))) . Sorry for the mixup. <n8henrie>Hello all -- I'm a first-timer with Guix, trying to use the equivalent of nix's `builtins.fetchurl` to import my SSH public key from a server. I'm toying with `guix repl` to see if I'm doing it right, after getting the hash with `guix download`, but I keep getting errors about it wanting symbols instead of strings. Can someone plase point me in the right direction? <n8henrie>example resulting error: In procedure symbol->string: Wrong type argument in position 1 (expecting symbol): "hashhere" <antipode>n8henrie: Looking at the docstring of 'url-fetch', it appears that the second argument of 'url-fetch' must be a symbol. <antipode>However, sha256 doesn't evaluate to anything, as it is undefined. <antipode>Maybe you need 'sha256 instead to get a symbol. <antipode>Also, "hashhere" is not a valid hash; it expects a bytevector representing the hash. <antipode>(base32 "output of guix hash here") may be useful <antipode>(you'll have to replace the hash with the right hash; I copied the wrong hash from the 'hello' package) <nckhexen>n8henrie: Yes, but you still have to quote it with ': (url-fetch "https://github.com/n8henrie.keys" 'sha256 "hashhere") — otherwise you're asking for the variable named sha256. Variable names are of the symbol type. <antipode>n8henrie: multiply-by-2 is a symbol; however, when interpreted (or compiled, whatever) as a Scheme program, it is a variable (de)reference of multiply-by-2 in the current lexical environment. <nckhexen>raghavgururajan: What do you mean? powertop doesn't have a configuration file, only options. <nckhexen>I've used and calibrated it, but there's precious little to configure. <Obikawa>raghavgururajan: You can try declaring a normal configuration file for PowerTOP. <nckhexen>Obikawa: You know more than me. What file? <Obikawa>nckhexen: I know nothing about PowerTOP. However, Scheme - and thus GUIX - lets you write configuration files in the code. <nckhexen>But powertop doesn't have configuration files. <Obikawa>Aha - my bad. I also remembered about (invoke "command" "argument" "argument" ...). <n8henrie>@antipode -- Thanks for your help. I'm getting `error: base32: unbound variable` with your example `,run-in-store` command -- am I missing another import here? <Obikawa>nckhexen: I saw it used in (package) declarations, but I don't see why it couldn't be used elsewhere. <nckhexen>raghavgururajan: Did you mean, how to start powertop at boot? <n8henrie>Got it, think I needed to `use (guix)`. Thanks! <yarl>In a service definition that depend on a package, where is the dependency declared? For example, the php-fpm-service-type extends the shepherd-root-service-type with the php-fpm-shepherd-service. As far as I understand, this last one is knows how to launch php (the (start ...)). Is is the fact that php appears in this Gexp that binds php to the service? <nckhexen>yarl: I guess, I'm not sure what ‘dependency declared’ means. That's how the service knows where to find PHP, yes. It does not express dependencies on other services. <yarl>nckhexen: I mean, if I make use of the php-fpm service in an operating-system declaration, php will be installed. <nckhexen>I didn't check, but it won't be added to your system profile (= you can't run ‘php’ at the prompt) by default. Services have to explicity add packages to the system list for that. I don't like services that do that. <nckhexen>The service will know where PHP is, and will run it, but that's it. <yarl>nckhexen: In my (operating-system ...) declaration, I don't need to add php to (packages ...) I only need php-fpm-service-type in (services ...). <nckhexen>Let me check the service to see if it installs PHP. <yarl>That's what I meant by dependency of a service over a package. <nckhexen>yarl: A service with ‘#$php’ anywhere in its code ‘depends’ on php in the sense that (define (x y) (display y)) ‘depends’ on the variable ‘display’. So, yes, but it's a (to me) strange way of putting it. <yarl>Ok, the choice of the word 'declared' was probably bad. <nckhexen>I was just going to say: your choice of the word ‘knows how’ is very apt. <nckhexen>The service knows where to find PHP, directly in the store. <nckhexen>I didn't check how, but it's going to be a variable reference somewhere. *waves hands* Not a list of ‘oh, I'm going to need these things:’. <nckhexen>Guix generally avoids those, because it can, because magic Gexp magic. <yarl>nckhexen: right, that's this "magic" I am trying to catch on a particular example. <nckhexen>Do you want to know exactly where the binding comes from? Because I'm stuck with a Web browser, and eh. <nckhexen>Plus, you seem to already know that ‘php appears in this Gexp’. <yarl>Yeah, don't you worry, I'll try to find it myself, I'm not that expericenced, I asked go gain some time. Sometimes chatting helps to clarify things. <nckhexen>I'm not finding that Gexp with C-f. Really, the ‘I guess’ above is almost certainly a ‘yes’, with the minor caveat that I might be telling you a lie because I can't find the exact code. <minima>hey, after installing 'clojure-data-xml', anyone knows how to make it available in a repl? <minima>if i try to import it, i'm told "Could not locate clojure/data/xml__init.class, clojure/data/xml.clj or clojure/data/xml.cljc on classpath." <yarl>the gexp is in php-fpm-shepherd-service <nckhexen>yarl: <Sometimes…> Better luck next time (◍•ᴗ•◍) <minima>ok, not sure it's the recommended way of doing this, but i think i have it working now <minima>once clojure-data-xml has been installed, i think a deps.edn file needs to be created that contains: <minima>'{guix/clojure-data-xml {:local/root "~/.guix-profile/share/java/clojure-data-xml.jar"}}' or similar <minima>once that's done, in order to use the library from within a repl, one has to import it with '(require 'clojure.data.xml)' <FareTower>Hi. I'm trying to install guix on a new laptop, but at startup it fails to start user processes, including failing to start getty, so the machine is unusable. How do I debug that? I can't even scroll up to see whether there was another error before. <Obikawa>FareTower: Is the laptop freedom-compatible hardware? <FareTower>maybe not for the builtin wifi, but otherwise yes <FareTower>(I am also having issues with grub-efi + cryptsetup + lvm, but at least I know how to work around them) <nckhexen>Is there anything (visible) before that? For example, Guix doesn't deal with failing mounts very well, and then everything that depends on them (which is often literally everything) fails to start too. There's no ‘best effort’ option yet. <nckhexen>It's tolerable when you have a previous generation to roll back to, but a horrible UX when you just installed. <pkill9>it would be neat to have a gnome-shell extension that lists guix packages when you search in the dash <FareTower>nckhexen, maybe the way I specify swap is wrong? My old configuration on a previous machine used a way that is deprecated, and I may have gotten the new way wrong. <nckhexen>Maybe! Most ways to get it wrong would result in a configure-time, not boot-time, failure, but who can say. <nckhexen>If you have your system.scm handy to share, I could look for obvious errors (that don't depend on your machine). <FareTower>ok, so Guix offers a graphical login... but (with a mostly empty $HOME, I only created a ssh key), it logs me out and I get back to the login without going through an actual session. <nckhexen>Did you use a preset from the graphical installer? <FareTower>no, I didn't, I just tried to cobble a configuration together. <nckhexen>It's been ages since I've installed from scratch, but I believe you need to manually passwd each non-root user. Which can be a problem, because some display managers don't allow you to log in as root IIRC. <FareTower>and the passwd was set, so that's not the issue. Maybe it's not finding any window manager to launch or something? <nckhexen>Possible, that all depends on what you installed or didn't. <nckhexen>Try to find the log file of your display manager (mine logs to ~/.local/share/sddm/wayland-session.log). <nckhexen>I can't offer more specific advice, sorry, and will now vanish for a while. Good luck. <lechner>FareTower: which window manager did you start? <lechner>FareTower: are you in a text terminal? <Andronikos>"guix shell vulkan-tools" does have /usr/share/vulkan/icd.d/nvidia_icd.json but "guix shell -CF vulkan-tools" does not. Why? <sneek>Welcome back Andronikos, you have 1 message! <sneek>Andronikos, rekado says: the scissors lines are made by selecting a region and then hitting C-c M-m (message-mark-inserted-region). <lechner>Hi, why does guix shell --development --file=guix.scm work, but this does not, please? guix shell --development --check --file=guix.scm <nckhexen>FareTower: Which window manager did you install (through ‘(packages …)’).? <FareTower>I hadn't installed any. Now installing ratpoison. <nckhexen>Andronikos: Because of the -C. Your host distribution bleeds through a regular ‘guix shell’, but isn't being exposed in the container. <lechner>with --check, the program is unavailable <Andronikos>nckhexen: If I understand you correctly that means that this file is from my host system right? What I don't understand is why only nvidia and not the other files as well? <nckhexen>I *assumed* the file is from your host system. Is it? <Andronikos>nckhexen: Yea you are completetly right. I am just confused from all the testing. <Andronikos>Well does someone know where I can get nvidia_icd.json? <nckhexen>lechner: If you compare ‘env’ with and without ‘--check’, the latter omits some crucial-looking search paths and other variables. I don't think that's intentional, or if it is, ‘--check’ should just exit reporting its result, not pretend to give you a usable shell. To the bugmobile, I'd say. <lilyp>if it's not included with vulkan, propably some shady blobs are involved <nckhexen>Andronikos: I have no idea how --expose interacts with --emulate-fhs (probably not well-tested), but you could try that. <Andronikos>nckhexen: Tried it but does not work. Can I use expose with a full path to a file? Since I get "guix shell: error: open-fdes: Permission denied" <Andronikos>lilyp: Explains the trouble I have but know I know what needs to be done. <nckhexen>If this is proprietary software, I don't provide support. Sorry. Not that I know the answer and am refusing to give it, though. *nckhexen not really sorry. Sorry :) <podiki[m]>the vulkan icd is likely provided by whatever package your host system uses for the driver/vulkan support; the ICD basically just says where to load the driver library from <podiki[m]>a guix process being able to load a host provided driver seems unlikely <nckhexen>It should have permission to try & fail, though? <podiki[m]>yeah, likely get some error loading a library or something <Andronikos>Sure no problem. I am not going through this hassle with Nvidia all the time again. (Waiting for RDNA3) <podiki[m]>but might also need to tell vulkan where to look for ICDs, I don't know if guix's looks in e.g. /usr/share (would doubt it, but we haven't properly patched vulkan paths as per the recent patch series to fix that in mesa/vulkan) <podiki[m]>nvidia (non-nouveau) is not supported in Guix proper due to freedom issues; sadly AMD isn't by the linux-libre kernel as newer cards need binary firmware to be loaded for full support <podiki[m]>(despite the open, and excellent, mesa drivers) <podiki[m]>(so you'd have to look outside of official channels for such things; which is where this topic ends on this official channel) <Andronikos>But AMDs drivers are overall better or not? At least I hear that all the time. <podiki[m]>depends I guess, compute stuff and video encoding is not as good? but for linux in terms of overall experience I certainly like dealing with just mesa better, yes <nckhexen>I think that's a good distinction. If you still use your GPU for the G, then AMD is probably much better; if you buy into the GP retcon they might not be. <Andronikos>Bind (DNS) does not have any documentation. Does that mean it is configured without Scheme? <nckhexen>Andronikos: There is no BIND service in Guix (yet). The only nameserver available that way is Knot. <Andronikos>I see. Saw that bind as package is available and thought that means the other as well. <Andronikos>Actually good since I could work on that to learn. <nckhexen>I do recommend Knot though. I don't use it merely because there's a Guix service for it. In fact I think my use predates that. <nckhexen>Oh, even better, yes, please contribute a BIND service to Guix and ignore what I just said. Knot suck. bah. <podiki[m]>Andronikos: welcome. it's true we lack recent/powerful GPUs with complete freedom, sadly <Andronikos>nckhexen: You wrote: "if you buy into the GP retcon they might not be. " I don't understand that. Can you elaborate? <nckhexen>I just meant ‘General Purpose’ rather than ‘graphics’. podiki[m]'s ‘compute and video encoding’ vs. real-time graphics. <podiki[m]>if you do use a lot of compute on GPUs you probably know about that already; I don't but that's what I hear <podiki[m]>other than some light OpenCL in Darktable for photo editing <nckhexen>lechner: It already has a service in Guix, so I can't use it to trick people into contributing another. <podiki[m]>though there's this new program vkdt that does it all with vulkan and is so so much faster; I have a package I should submit actually, but the program itself is early stages, though usable <nckhexen>I see how ‘I don't use it…’ could be confusing. I didn't mean for it to be. <Andronikos>General question since you may know more as I. Do HW manufactures not open their stuff because of fear? I mean the competitor could just learn to much from it. <podiki[m]>that's the argument I hear, I don't really buy it on the firmware/software front though <nckhexen>I've never met anyone who's bought it. And yet it must be extremely compelling to all these companies. Weird. *nckhexen does not know more than Andronikos. *podiki[m] calls dibs on knowing the least <nckhexen>Thinking your stuff is genuis and your competitors (who make equivalent stuff) at the same time so stupid they couldn't reverse-engineer it if they wanted to makes no sense, but I guess it's a strong delusion. That & just inertia & paranoid risk-aversion & legal CYA. <pkill9>they probably think people who want control over their computers are mud-people <Andronikos>"* nckhexen does not know more than Andronikos." how do you do that and what is that since it is not a normal message? <nckhexen>No, wait, first question. Second answer is: an ACTION, but you really don't need to know that. <Andronikos>What is the purpose? Seems useless to me to be honest. <nckhexen>Uh, to convey actions? And don't use 'em then. I over-use them, so that compensates. *nckhexen grabs Andronikos' quotum. <nckhexen>It was, but it doesn't work. ‘guix shell: error: open-fdes: Permission denied’. Ideas? <Andronikos>unmatched-paren: ls -l from host: "-rw-r--r-- 1 root root 140 Oct 22 22:02 nvidia_icd.json". I thought maybe it is because it is owned as root therefore I did a cp to my local dir with chown -R myuser:myuser. ARG uses full path which results to this error. <Andronikos>On "guix shell -FC" is -FC an argument or a parameter or called something else? <nckhexen>Argument. Actually, two arguments. Or flags (but they are a subset of arguments). <nckhexen>One argument, two options, strictly. God I'm already bored. <Andronikos>Parameter was in functions like void num(int a, int b)? <nckhexen>Andronikos: What do you mean, exactly, by ‘on runnnig the full command’? <nckhexen>Don't overthink the whole parameters/arguments distinction. ‘Arguments’ is the most correct™, but you'll find people calling them parameters and only a pedant will complain if the intention is clear. Same of function parameters/arguments. <Andronikos>It is just my urge to do stuff right. To be that precise is for most people not necessary but I need to be. <nckhexen>You sound like the kind of person who writes capitalised double-spaced sentences on IRC. Must be tiring. *lechner also thinks it's tiring for package descriptions <nckhexen>I guess I'm weird for teaching myself to double-space on a typewriter. I don't remember where I got it from. I actually unlearned it, but then Guix brought the habit back and it was just too hard to switch. *nckhexen is not as old as that makes them sound. <lechner>i had to do it in college and liked it, but it's apparently a holdover from typewriters that is no longer needed in modern typesetting systems. <lechner>i did it on computers though, so i am not that old <podiki[m]>there's the monospace vs proportional aspect too, so i hear <podiki[m]>though I shouldn't get distracted by that site, lots to read there <nckhexen>I don't see a technical need for it on modern typewriters, I assumed it was merely a human convention. <nckhexen>Yes the typographic Web is liquid crack. <lilyp>I read that thing and it makes some quite weird claims. <nckhexen>podiki[m]: And my MUA did *not* like quoting that entire ‘patch’. <nckhexen>Actually, the MUA did fine, I did not like deleting all of it. <podiki[m]>issues front end is usually slow to load for me on the kernel patch series too <lilyp>👆️ this. Communicating intent to your machine is good praxis. <lilyp>Otherwise you end up with bonkers machine learning processes that can't differentiate stuff like e.g. from proper sentence endings. <lechner>The dot/space combo kind of works for me ***kori1 is now known as kori
<Andronikos>But it is hard to read if every char is the same width. <lilyp>Well, the writer of this blog would argue that monospace fonts are hard to read anyways <nckhexen>They seem to understand arguments, and that would be a bad argument. <Andronikos>Do some of you run your own email server? What type of software suite are you using for automatic setup? <lechner>Andronikos: I have my own SMTP server, if that's what you are looking at (vs. an office suite) but it does require a manual type setup <Andronikos>lechner: Kinda yea. Do you have problems with IP beeing blacklisted or generally problems using your own? <lechner>Andronikos: not really. i am blacklisted from time to time, possible due to being part of a larger address block, but very few folks seem to pay attention if it's just one registry ***ChanServ sets mode: +o litharge
***litharge sets mode: +b *!*@2001:470:69fc:105::2:ae8c
***danoneil[m] was kicked by litharge (You are banned from this channel (by nckhexen))
***litharge sets mode: -o litharge
<lechner>Andronikos: The providers have different reputations. For example, I have heard that Linode is better than OVH <lechner>Andronikos: But Linode will also only open port 25 under close supervision. I had to send in part of my resume