<jake518>when I install a package with the command guix package -f a-python-library.scm Do you I have to source something afterwards in order to start using the package or does it immediately become available to a python repl by doing import my-python-library?
<raghavgururajan>nckx: OOPS! Along with mesa, I also need qtbase, ffmpeg, gstreamer, sdl2 and jack2.
<brown121407>Is it normal for a package (emacs in my case) to appear twice when using `guix package --list-installed`?
<kmicu>Hi folks, I’m testing the installer on the latest iso and it shows me that glib is missing during the installation process. There’s glib in /gnu/store/ but a different hash. Is that a known behaviour?
<dftxbs3e>brown121407, different versions maybe? A package in the GNU Guix sense is a full recipe that builds it including any modifications. A single software can have many recipes building it with many different options!
<brown121407>dftxbs3e: don't think so... I have this exact line output twite: emacs 26.3 out /gnu/store/lza7sccqgrv313gbrb9j8f7lgn49irdn-emacs-26.3
<fredmanglis>Hi everyone. I'm experiencing an issue where after doing a "guix pull" followed by "guix package -u" and "guix package --upgrade --fallback", some graphical applications like drracket and emacs have problems displaying text in menus, and sometimes even input elements.
<fredmanglis>It seems like they cannot find the correct encoding: they display each character as a tiny box with the four numbers in a 2X2 matrix formation
<nckx>Not sure if the first command is still needed but it can't hurt.
<fredmanglis>I tried a rollback, and it seems the update from version 2.29 to 2.31 of "glibc-utf8-locales" and "glibc-locales" might have had something to do with the issue, but re-installing those in my latest generation seems to not have fix the issue
<nckx>fredmanglis: That update was a while ago. Are you running Guix System?
<fredmanglis>The UI works fine on generation 22, but switching to 23, it "breaks" and starts displaying the tiny boxes.
<PotentialUser-29>Hello, I'm trying to package raylib for guix, but am running into some problems where neither the guix cookbook nor the manual seem to be of any help. Would anyone have the time to help?
<nckx>Not saying it's definitely a font problem but if all of your packages are up-to-date (and you haven't been manually setting environment variables somewhere) I don't see much room for glibc locale mismatches.
<fredmanglis>@nckx That's okay. Thanks for your suggestions. They should help eliminate some issues.
<fredmanglis>I do set up some environment variables in ~/.bashrc to make the packages installed via guix available on the shell.
<fredmanglis>I'll investigate that to see if it might be responsible for some of the issues.
<PotentialUser-29>1. How can I check the validity of the definition without running "guix package --install-from-file=raylib.scm"?
<PotentialUser-29>It seems guix lint is meant for this, but it's unclear how it can be run on a file instead of a package already part of the system.
<PotentialUser-29>2. Currently, installing this package by this definition leads to an error as alsa-lib apparently is an unbound variable. I assume I have to somehow grant the definition access by some sort of use-module command?
<PotentialUser-29>guix package --list-available=alsa lists alsa-lib as an available package, so in my naive approach it seems it should be available to the definition.
<fredmanglis>@PotentialUser-29: > as alsa-lib apparently is an unbound variable...
<fredmanglis>Try adding `#:use-module (gnu packages linux)` to your definition file
<mbakke>yes, the syntax highlighting is really nice, added by civodul not too long ago
<mbakke>mfg: perhaps removing the .guix-channel file will work, as I think that adds an extra (guix guix packages nvidia) namespace
<NieDzejkob>maybe your 'guix' directory is conflicting with the one in the main guix channel? perhaps try using something like (lihebi packages nvidia)
<mfg>sry i was afk, i will try it now and thx for the help :)
<kmicu>So the latest iso requires manual nomodset and then breaks on missing glibc version, 1.1.0 version works without nomodset but wifi connection hangs on it 🤔
<mfg>mbakke: NieDzejkob: Both ideas don't change the error message. It even keeps giving the same error if i have nothing in the channel but nvidia.scm at the toplevel with (define-module (nvidia)). Is there any chance of getting better error emssages?
<kmicu>mbakke: I have an old laptop here with libre wifi and an ATI card. I’m testing the graphical installer on the latest iso. Now I see the real error is with installer failing on building guix 1.1.0-15.03deb1e-checkout.drv and after resuming installation process it moves to an error: openning file /gnu/store/3133*-glib-2.62.6.drv.
<PotentialUser-29>Does someone have any idea why this environment variable would change between phases?
<NieDzejkob>PotentialUser-29: maybe the test script is using `cc` by itself, while the compilation is managed by cmake?
<kmicu>mbakke: I’m using defaults in the installer; they worked in the stable iso 1.1.0. Not sure how to check why building that checkout fails. There’s no log under the directory pointed by the error message. 🤷
<PotentialUser-29>I feel like a retard now. Yeah, couldn't find it before, but the test script uses cc.
<NieDzejkob>kmicu: this is because there once was a commit that used the wrong hash
<dftxbs3e>aww too bad! guix system disk-image doesnt have --with-source
<kmicu>NieDzejkob: lol, for real? xD Thank you so much. I will try it and report back.
<kmicu>I can switch to another tty and update guix but how to invoke graphical installer to test it?
<kmicu>Will the original ginstaller on tty1 pick up new guix version?
<kmicu>Or I could prepare config with g. installer but invoke guix init manually at the end.
<mbakke>kmicu: regarding the ATI driver problem, can you file a bug report with hardware details and workaround?
<kmicu>Is it a bug? For me adding nomodeset is a standard thing when installing a distro and screen is black xD
<kmicu>Also funny thing, when mod is set and I jump from another TTY to TTY with graphical installer I can see it for split second.
<NieDzejkob>kmicu: I'd generate a config with the graphical installer and then invoke manually
<kmicu>(It reacts to key presses but buffor only flashes when jumping in from another TTY otherwise is black.)
<kmicu>NieDzejkob: thank you. Is there a way to execute a command in a shell to start a graphical installer? That would help when a g. installer screams with a stacktrace. Otherwise we need to reboot each time.
<NieDzejkob>there probably is one, but I don't know what it is. Let me grep the sources a bit... :P
<raghavgururajan>nckx: Could you also add gtk+ and inkscape to the list. Thanks! Sorry for scattering my requests, I just found out that I needed them as well.
*raghavgururajan has just finished a 12hr shift. Gonna zzZ.
<lle-bout>nckx, hey! I meet a problem with gcc-5: "/gnu/store/hzz6g01dwjqvr5gpr4spm92lg2xb0l4h-gcc-5.5.0/lib64/libatomic.so: error: depends on 'libgcc_s.so.1', which cannot be found in RUNPATH ("/gnu/store/31p3wqzxhi3cjdkn4qf8780v84czp4ag-glibc-2.31/lib" "/gnu/store/hzz6g01dwjqvr5gpr4spm92lg2xb0l4h-gcc-5.5.0/lib" "/gnu/store/8wiy4y38kb5grmn79zkawcgz6lj55fnp-glibc-2.31/lib")" - Basically it looks for /lib but everything is in /lib64 - any
<lle-bout>idea how I can solve it in a central place?
<kmicu>guix weather still shows me 504 at the end and no subs for rav1e. I will wait for Internet to propagate entangled particles of libre subs.
<nckx>kmicu: It's suspicious because that's what we're used to, but it shouldn't be. ci.guix is ~25 verry powerful servers. AFAIK they are still basically unused (<2%) due to a scheduling bug.
<kmicu>Hash for my rav1e starts with rpf but ci shows w5d3iqx and xg4j4v
<kmicu>It’s a fresh install, I did not change anything 🤔
<kmicu>nckx: also checked and those dirs are not under / or /mnt/
<nckx>lle-bout: Patch the packages to install to /lib instead.
<nckx>kmicu: ‘rm -rf’ shouldn't care if they exist or not. And what did ‘guix weather’ return?
<lle-bout>nckx, okay.. is there any reason for not making programs look in both /lib and /lib64 like other distros?
<kmicu>I checked anway before rm anything. Just a habit.
<nckx>kmicu: I just tried to build it and got /gnu/store/w5d3iqx328f4xmaajwyj671gr42grnhb-rav1e-0.3.3. I also forgot… are we solving a real problem here or are you just curious? (Both are fine, but it's nice to know.)
<nckx>lle-bout: I… agree, which is why Guix doesn't do it. Using /lib64 is a distro decision, packages have to conform to the distro, not the other way around. /lib64 is a legacy hack that makes no sense on Guix.
<pkill9>lle-bout: i dunno, but it's simpler just to install all libs to the /lib directory in the package, plus there are other systems that a package could be built for
<nckx>I know you're using some other arch, but just FYI 😉
<nckx>pkill9: Yes, but I think lle-bout is porting Guix and hence deep inside the sausage factory.
<nckx>lle-bout: It's similar to why you won't find /gnu/store/*/local/bin, despite that being common (or even mandatory) on many systems. It's a distribution policy, packages don't have a say. And yes, we patch plenty of packages that hard-code /usr/local or whatever. And yes, I consider those packages naughty.
<nckx>If tomorrow some package starts installing to /bin64, we patch it. We won't be amending $PATH.
<kmicu>Am I correct that 0.0% substitutes available (0 out of 1) says there’s no subs?
<hendursaga>I just downloaded the tarball for the latest Shepherd and ran sloccount on it. It says it has ~2,500 SLOC of Lisp. Does that sound about right?? That's a whole lot less than, say, systemd.
<nckx>kmicu: Not much. It supposedly means there are no substitutes, but I say that's not true.
<nckx>Unless you are requesting something that differs from all the hashes you've posted so far. And I see no reason to think so.
<kmicu>I guess I could download it manually and manually add it to store.
<nckx>hendursaga: It souds plausible. Shepherd does a whole lot less, too.
<hendursaga>nckx: Like what? What does systemd do that Shepherd doesn't/can't?
<kmicu>nckx: it’s a fresh and default installation with guix pull, hash guix just before executing guix system init. Nothing changed by me.
<kmicu>(I even formatted disk with graphical installer.)
<nckx>hendursaga: cgroups managing/tracking for one, socket activation, … Plus running wc -l (or whatever) on the entire systemd repo doesn't mean anything; it contains udev and a whole set of separate but well-integrated daemons.
<kmicu>hendursaga: systemd has many libs that can be opt‑out. I hope you did not count lines on its whole repo.
<nckx>How do you even weigh lines of C against lines o' Lisp.
<hendursaga>nckx: sloccount, and I didn't count systemd myself
<nckx>Then you just need to have a server that runs Guix System & enable guix publish as I explained above. ci.guix.gnu.org is just a Guix System that shares everything in its own /gnu/store. There's no separate ‘substitute directory’.
<rndd[m]>nckx (@freenode_nckx:matrix.org): oh. Thanks, found needed chapter. Didn't know about publish argument
<nckx>If you run ‘guix publish’ on one laptop, set up keys, then run ‘guix install foo --substitute-urls=http://192.168.1.bar:8080, it will just work. It's not that complicated.
<nckx>A channel is a collection of Scheme code that allows other to download things. A substitute server simply serves compiled binaries to anyone who asks for a specific hash.
<nckx>You can set up a substitute server that serves compiled binaries *for* your channel, but they aren't coupled (and you can't, say, advertise a substitute server in your channel's source code).
<jgart[m]><nckx "You can set up a substitute serv"> great to know! that clarifies alot
<jgart[m]>so `ci.guix.gnu.org` is nothing more than a guix system instance running `guix publish`
<nckx>jgart[m]: It's a bit more than that, but not much. The other half is something that ‘fills’ the store: basically constantly calling [the equivalent of] ‘guix build <every package>’. Does that make sense?
<jgart[m]><nckx "You can set up a substitute serv"> but I can link my guix channel to optionally pull substitutes from a guix system that is offering a guix publish service (i.e. substitutes) for the entirety of my guix channel?
<nckx>jgart[m]: It uses fancy software called Cuirass to do so. But my substitute server *is* just, literally, guix build <all the things>, in a bash loop.
<jgart[m]><nckx "jgart: It's a bit more than that"> but guix build does not install the package to my profile. It only builds the derivation to the store, correct?
<nckx>jgart[m]: You can't link the channel. You can tell users about it in your channel README, but they will have to add your key and substitute URL by hand.
<dustyweb>there has been some work I understand, though I don't know the state of it; Dimakakos Dimos has done the most work on it, see "Racket packages: formerly Move DrRacket to a separate output?" threads on guix-devel
<jgart[m]>I want to be able to just do `guix install racket-frog` or `guix install frog`
<jgart[m]>In the same way one can do `guix install haunt`
<jgart[m]><dustyweb "there has been some work I under"> I'll look for this. Thanks!
<jgart[m]>Can the `android-ndk-build-system` be used to build android apk's directly? It would be nice to be able to build android apps from source with guix in order to provide "apk substitutes" or atleast a guix channel. Maybe this could be the fdroid alternative /gnu/store that the replicant project has been looking for all along...
***nckx changes topic to 'GNU Guix | register your nick to talk | get Guix at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs & patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: http://logs.guix.gnu.org | 1.1.0 is out! https://guix.gnu.org/blog/2020/gnu-guix-1.1.0-released/'