IRC channel logs
back to list of logs
<bovid-19>lilyp: thanks, i'll try to figure out how to do that. <Haider>Thank you so much! Wow, It may not be ideal, but it is amazing. Never knew about the issue tracker!
***lukedashjr is now known as luke-jr
***luke-jr- is now known as luke-jr
***luke-jr| is now known as luke-jr
***luke-jr- is now known as luke-jr
***lukedashjr is now known as luke-jr
<acrow>sneek: later tell blake2b Thanks for recommending the guile compiler tower docs. I've been referencing that documentation for many years now and I am always left with the feeling that I'm in the the best of hands and inspired to imagine doing more. Great stuff. Great work and it just gets better and better. Wow. <VS4mwaLnRSM>Hello. I had a SHA256 mismatch from a package I just defined. Whenever I try to re-build it, I get the same mismatch error despite having fixed it. What am I ought to do to resolve this and use the fixed (corrected), or just simply the latest package changed/fixed in foo.scm? <VS4mwaLnRSM>I tried "guix gc" then re-building the package, and I also tried "guix refresh" just because it seemed to make sense, but I am quite new to Guix, so I am not quite sure. I suppose it is still using the old version and I am supposed to run a command to use the latest, just not quite sure what that command is. <VS4mwaLnRSM>So if anyone could provide me the information necessary to resolve this issue, it would be greatly appreciated. <vagrantc>VS4mwaLnRSM: how are you building the package? <vagrantc>VS4mwaLnRSM: and maybe paste your package definition somewhere (e.g. paste.debian.net) <vagrantc>VS4mwaLnRSM: are you using "./pre-inst-env guix build PACKAGE" ... or ? <vagrantc>unfortunately, "nonguix" is off-topic here <vagrantc>which reminds me, need to post to the list about ... where the heck that's documented <VS4mwaLnRSM>So, assuming it is some other .scm file, is there a way to "refresh" it? <VS4mwaLnRSM>Like, the same thing may happen with my own .scm, unrelated to nonguix. <vagrantc>there are different ways to build packages... <vagrantc>that's not going to use any customized packages <vagrantc>i don't have a lot of familiarity with GUIX_PACKAGE_PATH, it's somewhat deprecated in favor of channels and such <vagrantc>if there are .go files in that directory, try deleting them :) <VS4mwaLnRSM>Oof, thank you. :( I might be able to somehow delete everything and start over, although "refreshing" this single .scm file under GUIX_PACKAGE_PATH would have been better. <vagrantc>e.g. foo.scm and foo.go ... the .go file might cache the results <VS4mwaLnRSM>I think I will just change the name and append -1 to it and keep doing that until it works properly, then rename it to the original name and keep that. :D <VS4mwaLnRSM>But yeah I am pretty sure the package is stored/cached somewhere. <VS4mwaLnRSM>"guix gc" seemed to get rid of its dependencies at least, thought it would get rid of the package and its information (incorrect hash) as well. <vagrantc>guix gc just removes files in /gnu/store (and maybe some databases) <vagrantc>packages that failed to build don't typically end up in the store <vagrantc>but maybe there's some optimization in the way GUIX_PACKAGE_PATH is handled that caches more things ... <acrow>VS4mwaLnRSM: GUIX_PACKAGE_PATH works. <acrow>Without channels and independent of revision control you need to tell guix build where to find your package definition. <acrow>The easy way is to say guix build -L . -f my_file.scm where my_file.scm must evaluate to a scheme package object. <acrow>You can do the same with package to install what you create this way and, maybe, you are done. <acrow>If you want to fiddle with the same package again and again you can define GUIX_PACKAGE_PATH just like the system PATH variable (delimited with colons) and then you don't need the '-L <dir>' entry to run guix build on the scheme files in those directories. <acrow>In all cases you have to identify your scheme file using '-f <my_scheme_file>' and that file must evaluate to exactly the package you are trying to build. <acrow>It's quick, easy and good; but if you want to contribute what you've done you will soon want to read more of the excellent guix documentation. <VS4mwaLnRSM>acrow: I am indeed using GUIX_PACKAGE_PATH and it uses the .scm file that I want, my main issue is that even after doing "guix gc" and doing a "guix build X" still uses the old .scm file with the wrong hash. I suppose it is cached somewhere, and I need to get rid of that cache somehow. Perhaps "guix build -L . -f pkg.scm" may work, I dunno. <VS4mwaLnRSM>Like, I changed the hash in the .scm file that it seems to use, but it is not using the new version of it. <VS4mwaLnRSM>I am going to give it another try since I rebooted (had to), and see what's up. <VS4mwaLnRSM>Except I already changed it to the correct one in the .scm file... :/ <mroh>try to rename your package. <VS4mwaLnRSM>Yeah, thought about that as a last resort. I think that will work. <sneek>Welcome back dcunit3d, you have 1 message! <sneek>dcunit3d, jaft says: Regarding fonts not rendering in QT WebBrowser app.s, someone pointed out a patch that was supposed to backport a fix and I manually applied said patch to QTWebEngine; you can find that here: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=53011#40. No one's responded, yet, on their thoughts regarding that approach but it's worked for me, with QuteBrowser. <VS4mwaLnRSM>After "The following derivations will be built:", the XXX in /gnu/store/XXX-package is supposed to be the same all the time? <char[m]>jgart: I found how to make cargo find the crates. export CARGO_HOME=$GUIX_ENVIRONMENT/share/cargo/registry/ <char[m]>. problem is, cargo tries to create something there and gets a read-only file system error <dcunit3d>VS4mwaLnRSM: yes, if all the packages your derivation depends on are also deterministic <VS4mwaLnRSM>I was wondering if it should have been different if I modify it (change the hash, for example). <dcunit3d>but according to the manual, new guix packages are tested before they are merged in. determinism in the build is very important. <VS4mwaLnRSM>Thanks. Once I get my GPU driver and plain dm-crypt working the way I want, I will make Guix System my desktop OS. :) <VS4mwaLnRSM>My plain dm-crypt setup might not be supported, so I may have to add it or something, we'll see. <dcunit3d>if you are using grub to boot from a LUKS drive, then that is supported <VS4mwaLnRSM>The boot partition is on an USB stick, along with the kernel and everything that goes inside /boot in general. <VS4mwaLnRSM>Anyways, it does work under Arch, I'll make it work under Guix System somehow. <dcunit3d>you should look here. this generates a USB installer, but a system like that would be similar. <char[m]>I found that If I set CARGO_HOME to ~/.local/share/cargo/registry and put symbolic links to all the creates in $GUIX_ENVIRONMENT/share/cargo/registry in there, it works. <dcunit3d>one problem is that the /gnu/store is not on the device with /boot/efi and / <char[m]>It is not scalable at all, there has to be a better way <dcunit3d>but i don't know what other problems you may encounter. <VS4mwaLnRSM>dcunit3d: Although as long as I manage to get to the booting stage or post-boot, it should be fine, right? <VS4mwaLnRSM>Because I have a feeling I could manage to do that, but we'll see. <Michal_Atlas[m]>Hello, I gave a problem that guile doesn't seem to accept keyboard interrupts, nor in the repl, nor when I run a script, can I enable that somewhere? Or do I have to run the blocking task on another thread, or? <VS4mwaLnRSM>Jeez, the issue was that I misunderstood what "expected hash" and "actual hash" meant. I assumed "expected hash" to be the hash that Guix expected, not the hash that I have given that was incorrect. It did not help that the only difference between the two was one character, and I thought the actual hash was the hash that I have provided. Facepalm... <jpoiret>VS4mwaLnRSM: your /boot is unencrypted on a USB right? <VS4mwaLnRSM>can I somehow avoid steps when building? the "build" phase too much time and I need to only test the "install" phase <roptat>VS4mwaLnRSM, if you don't build, how can you install, though? <jpoiret>VS4mwaLnRSM: wdym about skipping build? <jpoiret>oh, you've already done that phase previously and don't want to re-do it? It's currently impossible unfortunately <VS4mwaLnRSM>I don't want to rebuild for the 100th time, it takes too long. <VS4mwaLnRSM>I want to skip that phase (after having already done it) and just perform the "install-copy" phase to see if it copied everything it had to, created the proper symlinks, etc. <tplaten>When I try to build pulseaudio on my Talos II, I get #error Define either WEBRTC_ARCH_LITTLE_ENDIAN or WEBRTC_ARCH_BIG_ENDIAN <brendyn>VS4mwaLnRSM, you can delete the build phase with (delete 'build) <brendyn>VS4mwaLnRSM, maybe its possible to build with lower optimisation level <VS4mwaLnRSM>oh thaat could work! and since it's always the same XXX in /gnu/store/XXX, it should work as intended <VS4mwaLnRSM>if it doesnt work, I'll ask my friend to give me ssh access to his PC, it's high-end lol <Michal_Atlas[m]>I never tried this, but what I presumed would be the case (based on nothing at all), was that when you ran build with -K that there would be a guile script or something in there that allowed you to run the different steps, doesn't that sound more convenient? <brendyn>Michal_Atlas[m], It doesn't exist but I've always wished one could go inside the build enviroment and interact with it more <VS4mwaLnRSM>Well, it successfully builds... hooray. I just have to test it, but I have never even had Guix System so... first thing's first. :D <VS4mwaLnRSM>If this nvidia-390 package works with my GPU and all that, then I can make it my desktop OS! I might even be able to fix the plain dm-crypt thing. I would really love to. <roptat>ss2, it's you're time to shine ;) <VS4mwaLnRSM>(Ideally I would not have an NVIDIA GPU, but costs...) <ss2>Just about to set off. :) <VS4mwaLnRSM>I would discuss some of the issues I ran into as a complete newbie! Hah. <tplaten>Yes, NVIDIA GPUs, are bad, as they need non-free software. <VS4mwaLnRSM>I don't think I have open hardware in my desktop PC... motherboard, CPU, GPU, etc. are all closed. :c <VS4mwaLnRSM>That said, I would rather have AMD Radeon than my current NVIDIA card. <VS4mwaLnRSM>Maybe I can get a used one that has a similar performance and whatnot without having to spend too much. Then I will probably start with updating Dm-crypt / LUKS on DragonflyBSD when I get the time. <lilyp>Radeon is also pretty bad w.r.t. needing blobs. <VS4mwaLnRSM>I'm out of touch with GPUs, and I was never in touch with AMD GPUs. <lilyp>old nvidia GPUs work but with quirks (talking something like 710 here) <lilyp>see the nouveau feature matrix for that <lilyp>intel gpus also work pretty well from personal experience (the ones you'd find in laptops) <VS4mwaLnRSM>I am not sure if nouveau is even a thing on OpenBSD or DragonflyBSD, but nouveau on Linux may be good enough for my case. I will give it a shot. It is almost as old as GT 710. <tplaten>I have tested an old NVIDIA GPU with my Talos II, it only works in petitboot, not in the main OS. I have heard that raptorcs wants to use the libre-soc as a GPU for their Talos II. If there is funding, it may actually happen. <efraim>I've been driving my local computer store crazy by picking up 710 and 730 when I need another GPU <VS4mwaLnRSM>So... new GT 730 costs 103 EUR, and I bought GT 630 a really long time ago... many many years ago, at the price of ~56 EUR. Jeez. :P <jpoiret>looks like connman doesn't report state=online anymore even though there's obviously some connectivity... i wonder if this has to do with the version bump to 1.41 <acrow>VS4mwaLnRSM: glad to see that you got your build working. :) <VS4mwaLnRSM>Thank you! ^^ I spent way too much time on a silly problem; I misunderstood "expected hash" and "actual hash". :| After that it was pretty quick. <VS4mwaLnRSM>Still, it was a good learning experience, and more to come! <VS4mwaLnRSM>I just started reading about Guix 2 days ago, so pretty new to it. I dunno if I could test the package myself as of yet, because I am not sure how it is even supposed to work, but someone might... at least from the other channel, as it is related to NVIDIA. <VS4mwaLnRSM>I've never installed Guix System before, never made "install/config.scm", etc. <acrow>I think there is a talk about the installer on Guix-Days, in a little while. I recall it does a pretty good job. <VS4mwaLnRSM>I came up with my own pretty quickly, I think the trouble will be using an USB stick for the unencrypted /boot partition, while the hard drive has full disk encryption without any LUKS headers. <VS4mwaLnRSM>I am not even sure if it possible, so I might have to ditch it for now. <VS4mwaLnRSM>At least based on the stuff I read it does not seem like it is possible. <acrow>Yeah, you're aiming high for a first install. You may want to just keep it very basic to start. It will deliver you a valid config.scm. Then you can change the config.scm and do iterative reconfigures to get to the end result you want while taking advantage of the system rollbacks that guix offers. <VS4mwaLnRSM>That's true. :) The package I made/am making is nvidia-390 for my GPU, so I can at least have that right off the bat. <VS4mwaLnRSM>I have ideas as to how to make plain dm-crypt work with Guix, but I will have to install it first and have a better idea of it. <VS4mwaLnRSM>I'm still too much of a newbie anyways. I want to pick up Guile as well. I have some experience with developmenmt in Common Lisp, so I don't mind working with parentheses. :D <acrow>Guix will change your world view. :) In many ways. All good. <VS4mwaLnRSM>I was thinking if I should go with NixOS or Guix System but Guix System seemed more to my taste/like. <Michal_Atlas[m]>I tried Nix but it really lacked the friendliness and community that Guix has, and I couldn't understand their language too well. Tbh, you can just add the Nix service in Guix, and I've found that to fulfill the extent to which I need to use it. <VS4mwaLnRSM>Michal_Atlas[m]: I did not like the fact that Nix uses their own language either... that was one of the deal-breakers. <florhizome[m]>VS4mwaLnRSM: There is a url for each patch-series under a single issue ;) <florhizome[m]>Nothing in the UI or in the doc really tells you how to get to them. <florhizome[m]>I'm looking for an url i can just curl into a patch tool to apply a patch-series to a local checkout <GNUtoo>hi, adding '(if #t (display #t)(display #f))' in a system.scm displays #t, but if I add it inside file-systems somehow I get "invalid field specifier" <GNUtoo>Is it possible to use if inside fields somehow? <GNUtoo>Or do I need to use it outside and have some code duplication <GNUtoo>In my case I've something like that: (file-systems (append (list [...] (parabola32-bind-mount "/home/gnutoo") [...]) %base-file-systems)) and parabola-32-bind-mount is defined as something like that: (file-system [...]) and I added the if inside the file-system inside parabola32-bind-mount <GNUtoo>Here I probably miss some lisp concepts <VS4mwaLnRSM>I am new, too. I found out a lot today by looking at example files. For example that made me realize what what "," does, which was pretty useful. <GNUtoo>for things like ',' the guile manual has a good introductions on quote, quasiquote and so on <GNUtoo>The hardest thing is to find ',' without knowing its name <VS4mwaLnRSM>I figured out what it does by looking at example .scm files. ^^
***julm is now known as Guest3504
<jlicht>Does anyone have experience setting up Guix on a Fedora machine? I've never used one, but it seems the desktop sessions don't source /etc/profile.d/<things>.sh, for some reason, and instead use systemd-specific shenanigans that can probably do many interesting things, but not source /etc/profile.d/guix.sh :-) <roptat>jlicht, ah that's why I kept having issues and had to source from my .bashrc :/ *GNUtoo added source where needed to source files <roptat>oh, our cross-gcc is missing a few libraries... the "lib" output only has libcc1.so, no libstdc++ or libgcc_s :/ *GNUtoo wonder how that works with selinux though <roptat>it doesn't, my fedora systems didn't have it enabled, but I know other users have to disable selinux <AwesomeAdam54321>jlicht: I don't know about Fedora, but I changed my graphical terminal to use --login with bash <jlicht>Hmm, it might also have something to do /w Wayland, but I'll try it out AwesomeAdam54321, thanks for the pointer! <jlicht>I ran Guix for years on Ubuntu without issues as well, it really seems to be related to the fact that GDM-started Sway is not a login shell (so does not do login shell things, naturally) <odiferousmint[m]>If I understand you correctly: if your `/etc/profile`, `/etc/profile.d/*.sh`, and/or `~/.profile` does not get sourced by the login shell, then perhaps use `~/.bash_profile` or `~/.bashrc`. The last one gets executed each time you call `bash`. You could use the code I pasted to source `$GUIX_PROFILE/etc/profile`. <AwesomeAdam54321>odiferousmint[m]: Thanks, I probably need to set LC_ALL so that I don't need to use glibc-locales and use glibc-utf8-locales <odiferousmint[m]>You can run `locale` and see what is missing. For me `LANGUAGE` is the only missing one, the rest is all `en_US.utf8`. <odiferousmint[m]>I installed `glibc-utf8-locales` and it still kept nagging me, that is when I thought I should do this, and realized `LC_ALL` was unset. <aeka>Hi Guix, unexpected results when trying to apply transformations to emacs-next <aeka>$ guix shell emacs-next --with-latest=emacs-next --with-branch=emacs-next=master <aeka>guix shell: error: the source of firstname.lastname@example.org is not a Git reference <aeka>any ideas as to why? guix edit emacs-next reveals the package origin is using (method git-fetch) <jlicht>sorry, aeka: did you try with only the with-branch transformation option? <aeka>I have not jlicht, let me try. <aeka>well it's updating the checkout, so I assume that this will work <aeka>do the two options simply compose poorly? <maximed>akea: '--with-latest' uses (guix upstream) to look for the latest release. <avp>BTW, I'm using guix 45995ea (according to 'guix describe') <maximed>It does not look for the latest commit. <maximed>I think "--with-branch=...' looks for the latest commit on that branch? <maximed>Though recently, there's is a not-yet-merged latest-git updater which could be useful for emacs-next and the like, so perhaps in the future you could do "guix shell emacs-next --with-latest=emacs-next" <mroh>avp: try to change the python-2 input(s) to python-3. <avp>mroh: Okay, will do.
***EndOfAnEra is now known as you
<aeka>maximed: ahhh, understood <aeka>I thought --with-latest was with latest commit
***you is now known as simplex
<rekado_>I’m building llvm on aarch64 since two days. <rekado_>this is the fourth time, because previously the build would repeatedly time out. <rekado_>it’s probably going to take another day to finish. <rekado_>either because they’ve been silent for too long <rekado_>or because they have been going on for too long <GNUtoo>oh ok that's in guix itself or is it in cuirass or similar software? <GNUtoo>ah there is even a --timeout=SECONDS in guix build <GNUtoo>As I've not tried that I've no idea if it can help or not <GNUtoo>ok, I didn't know that one either <jgart>Better documentation for using the minetest packages/mods in the context of using minetest with guix was one I forgot to mention
***lukedashjr is now known as luke-jr
<roptat>oh I thought cross-gcc did not have a libstdc++, but it's in i686-unknown-linux-gnu/lib/libstdc++.so, which is not in CROSS_LIBRARY_PATH <roptat>also I see a comment when defining CROSS_LIBRARY_PATH: "Only search target inputs, not host inputs." is it related? what does it mean? <roptat>I think fixing that would allow me to cross-compile android stuff <VS4mwaLnRSM>I need more context, but I suppose it just means that we should not search the host's, only the target's inputs... <VS4mwaLnRSM>Seems like it could be a starting point if you haven't came across them yet. <roptat>yeah, looks like I could special-case my cross-clang <VS4mwaLnRSM>android 13's kvm support, allows you to run windows and linux, and doom on either <roptat>well, not really related to what I'm trying to do right now, but will have a look <VS4mwaLnRSM>he's compiling linux 5.17 allnoconfig for arm64 on arch on android <roptat>I'm just trying to build the host tools, like fastboot, adb etc <roptat>that works, but I wanted to also add cross-compilation support, which requires a cross-clang <roptat>and cross-compiled libcxx and libcxxabi which don't build because the cross-clang doesn't find libgcc_s <mrw>Hello folks. I'm a new guixSD user. I'm trying to fig doeure out if it's possible to build and/or use cross toolchains for embedded development (linux or otherwise). It's turning out to be a bit tricker than I expected. crosstool-ng complains that C_INCLUDE_PATH and LIBRARY_PATH being set will mess up the build (but no details given). <mrw>Buildroot complains that /usr/bin/file (with exactly that path) is a pre-requisite (again no details given). Using a pre-build toolchain from, say, bootlin has trouble because of interpreter paths. Perhaps I will have to finally learn some toolchain arcana I have been avoiding for years, but before I dive in, are there any comments / advice / <mrw>pointers anyone could give me on this subect? <drakonis>there was a fosdem talk regarding that a while back <mrw>Thanks drakonis, just went seareching and found it as well. I'll give it a listen. Yocto is a bloody nightmare I avoid at all costs.
***_ctb_ is now known as _ctb
***lukedashjr is now known as luke-jr
***simplex is now known as EndOfAnEra
<roptat>oh I might have found the issue. I was trying to build for i686-unknown-linux-gnu, but clang doesn't know about it. It lists i686-pc-linux-gnu though, so I'm trying that now <roptat>ah no, it's added in a phase anyway, so that's not it <kitzman>is it okay if I submit a single patch with multiple packages? a total of three packages, created by the same person, two being dependencies for the last one <kitzman>or a patch set would be more appropiate? <VS4mwaLnRSM>are 2 of em of any use by themselves or with another package? if it's a "whole" whole, then I'd say single, otherwise it sounds like a set, but what do I know :P <kitzman>theoretically yes - but I checked the AUR and no other package there uses them <kitzman>i'll go with a single one for now ^^ <jlicht>kitzman: One patch serie with multiple patches would be The Way ;-) <kitzman>is it okay if i close the current one sending an email to NNN-done? <kitzman>(plus i'd also need to amend some things there) <tribals>Is anyone using racket installed by guix? <roptat>but I don't think you need to close, you can send a v2 to the same patch number <roptat>tribals, hi! I don't use racket unfortunately
***Guest3504 is now known as julm
***julm is now known as Guest660
***Guest660 is now known as julm