<PotentialUser-61>weirdly enough, I can't get the good time out of the "date" command. In my operating system declaration, I tried "(timezone "America/New_York")" and "(timezone "America/Montreal)", but both give me the following output when using the "date" command: Sat 31 Oct 2020 03:12:07 PM EDT. The timezone seems right, but not the time
<yarnton>Hi, I'm test driving GuixSD as I'm thinking about switching from NixOS. Why do I see Julia version 1.3 (even after I do guix pull) if I have the default channel (git master?) and there there's a newer julia (1.5)?
<yarnton>is the text installer for GuixSD not showing a limited amount of WMs? ie StumpWM seems packaged and supported, but is not there.
<ryanprior>cbaines: that looks useful! Are we interested in moving towards something like that can display patches, commentary, and checks like these all in one place similar to how the various git forges do it?
<maav>yarnton: as far as I know the current ones are there because somebody thought they were interesting and sent a patch... maybe it's a bit late for this release as we're on freeze, but I don't think currently there are other reasons to be picky with this issue, as the screen should have room enough for more desktops :)
<yarnton>maav: against my argument u could always argue that most stumpwm and other wm users would do a manual install, so its fine :)
<maav>yarnton: of course, but some also could want to share their favorite desktop to the newcomers :)
<yarnton>maav: anyway, im glad to see guix is progressing so much. i've been checking every year or so and its pretty mature now to switch from nixos
<yarnton>i maintain some nixos packages, but imho the distribution is becoming too chaotic, there are lots of broken things
<maav>yarnton: i started using guix as my os a couple of years ago and i cannot think of using anything else anymore :)
<helaoban>hey guix, I'm getting an error all of a sudden when trying to run './etc/indent-code.el' over a package definition: 'Symbol's function definition is void: cl-pushnew'.
<helaoban>I don't use emacs so I'm in the dark here.
<morgansmith>I don't quite understand why we prefer release tarballs over using the source repository. For a lot of the software I use, I want to play with the source and make changes. If the Guix package is set to use source, then all I have to do is clone the repo, make changes, and then point Guix to my new repo. This workflow isn't possible when we point to release tarballs. So what is the benefit of using a release tarball?
<roptat>user_oreloznog, hi! I've sent my changes to the TP, so you can download the file and continue from there. If you don't have time, please tell me, I'll continue the translation where you left it :)
<roptat>morgansmith, I don't think there's any benefit, except that it's sometimes easier to find maybe
<cbaines>morgansmith, I'd expect for most packages, you'd be able to use the source repository in place of a release tarball. Do you know of one where this doesn't work?
<morgansmith>also I've sumbitted emacs packages before using the source and been specifically told to use elpa instead
<morgansmith>mind you emacs packages are simple enough (usually) that it shouldn't matter too much. But if you build emacs-modus-vivendi-theme from source you actually install both emacs-modus-vivendi-theme and emacs-modus-operandi-theme since they are in the same repo. I'm not sure what would happen if you tried to install both from source so you had conflicting file names. I again submitted a patch to move these to source (since
<morgansmith>having trouble updating elpa) but I was told just to wait for elpa to update :/
<cbaines>morgansmith, I think things like ELPA (PyPI, Rubygems.org, ...) are a slightly different issue to release tarballs
<cbaines>Maybe the artefacts from those sites aren't ideal for using for Guix packages, I know for PyPI and Rubygems at least, often it's not possible to run the testsuite from those artefacts
<morgansmith>ok, well emacs-org (not emacs-org-mode, I goofed) does use a release tarball and it is currently preventing me from developing org stuff :/. I have it half packaged from source at this point but now I actually have to try to get the tests to pass and also I'm a little scared it'll just get rejected
<cbaines>From my perspective, there's a difference between a "release tarball" and the tar file ELPA is providing
<morgansmith>Is the elpa thing just a stability thing then? Like some dude's gitlab could go down but elpa probably won't?
<morgansmith>From my perspective, any service that provides the source in a form different from the development source is bad. I guess my beef isn't directly with release tarballs, but it is related
<morgansmith>Ok maybe not "bad". I'm just trying to understand why we would want to accept anything besides the source in its development form
<cbaines>There's probably a range of different factors, that vary depending on the exact nature of the project
<cbaines>I'd perhaps ignore the general "release tarball" issue, and send a message to guix-devel about using other sources for things that are available via ELPA (like org-mode)
<rekado>wxwidgets does not contain any cmake files nor does it contain pkg-config files
<mfg>okay, those aren't all dependencies, i have to search a bit :D
<rekado>so you may need to manually specify the cmake variables for libraries and includes
<mfg>at least slic3r is a perl program, so it needs perl-wxwidgets
<mfg>i will have to lookup the differences between prusa slicer though
<ecbrown>hello guix! i am following the advancements with the hurd with great interest. one question, python is noted to not compile. i noticed that python 3.9 has some hurd augmentations--would this cure the python/hurd build issue? (just planning ahead, want to ask those who may be in the know)
<morgansmith>Is perl-local-lib not a dependency of slic3r? I packaged that but haven't submitted it
<pinoaffe>morgansmith: prusaslicer seems to have replaced a lot of the perl code in slic3r with c++, so it would make sense if some perl dependencies are no longer necessary
<mfg>morgansmith: i think i don't saw such a package, but it might as well be missing :D
<morgansmith>Maybe I'll attempt again. I forget what I was stuck on
<mfg>pinoaffe: Ah that's great!, the official slic3r repo also contains a lot c++ bits, but it's deemed not ready for production there... I guess it would be way easier to package without those perl things :D
<nalaginrut>I've installed linux-libre 5.9 from guix, how can I boot it?
<morgansmith>welp, can't compile guix right now due to the documentation have nonexistent nodes... Is it just me?
<morgansmith>nalaginrut: if you installed it using a guix system reconfigure then it should be the default in the grub menu
<nalaginrut>morgansmith: thanks but what's the correct command? "guix system reconfigure" requires an argument
<janneke>nalaginrut: are you using guix on a foreign distro?
<morgansmith>you have to pass it a system configuration. The path to your current configuration can be seen by running: guix system describe
<roptat>something's wrong, my computer feels very slow, but I'm far from 100% use of my CPU, memory or IO, what could it be?
<user_oreloznog>roptat: If you can continue the translation where I left it, it's nice! I think I'll have my guix environment ready tonight. And from there, I will be able to take my turn when I'm ready again... Thank you!
<morgansmith>ok, a git clean -xfd seemed to fix it. I hate our documentation build system though. It's gotten in my way a few times. I don't actually know if it's problem with the build system or me, but I'm going to blame to build system
<cbaines>morgansmith, I'm glad you found a way around the issue. I'm not sure what was going on though, as I'm not sure documentation issues have blocked me running make successfully
<morgansmith>so I use magit as my git interface, but I had it freeze once or twice due to massive diffs in po/guix/ta.po po/packages/es.po po/packages/zh_CN.po so I told git to always assume they are unchanged. And that's worked for months. But then for this issue I told git to worry about them again. I don't know what those files are but maybe that's it?
<cbaines>My approach has just been to discard the changes in those files when they appear
<mroh>The Installer doesn't have this module, afaik.
<morgansmith>mfg: I think the commented out inputs are copy pasted from the slic3r make files. Like they are inputs but we don't have them packaged yet so I was trying without them. or maybe I was trying to minimize inputs? idk
<nckx>The installer should be patched to load/include as many modules as possible, at least for anything remotely network-enabling.
<nckx>I don't get it. Upstream says distro PCI ASPM (power management) can conflict with the driver, then suggests disabling AER (error reporting). If that even works (and isn't merely a typo), it sounds like risky advice.
<mfg>pinoaffe: not sure what's going on with wxwidgets here, i mean cmake should be able to find that ... what happens if you use either wxwidgets or wxwidgets-gtk2, not both?
<pinoaffe>and adding "glew" as a dependency also didn't change anything
<mfg>pinoaffe: adding glew should make cmake use the system version instead of the bundled one, because it's a policy in guix that packages should not use bundled versions of libraries that are available as system libraries.
<mfg>morgansmith: perl-alien-wxwidgets also has no error for me :-(
<nalaginrut>if I'm on debian, then guix can't install a kernel for me?
<mfg>pinoaffe: i'm looking deeper into it later, it might be possible to manually set /all/ variables cmake would set, but i'm pretty sure there has to be a way to let it find everything on it's own...
<mfg>Might be better to look for version 3.16 as this is what guix uses atm
<HypoSampled>Ok, how do I add 'pci=noaer' to the kernel command line?
<OriansJ>hmm looks like email@example.com has a build issue when using 8 cores but not when built with -c 1 or -c 2 or -c 4; possible race condition for the build? Is there a way in guix to specific limits (like limit to only 4 cores or less) or prompt for requirements (like is tmpfs big enough to even do the build); so that users have less cryptic problems when building and that builds that will not work can fail faster?
<HypoSampled>I guess I must have touched the dark side a little before
<morgansmith>I bought a Qualcomm Atheros QCA9596 QCNFA335 last april for $14.49 CAD and I'm quite happy with it
<mroh>HypoSampled: not sure, if the 8192ce driver works with a 8188ce chipset. I guess, 8188ce works with vanilla drivers, but you need realtek-firmware.
<mfg>pinoaffe: if you add "--trace" "-Wdev" to your configure-flags you get more verbose output about what cmake is doing. I see that it actually found wxWidgets but after that there some FPHSA* things going on i don't quite get -- those fail
<morgansmith>I finally built slic3r btw. trying to figure out how to launch it now :)
<Bumblehorse>When defining a new package, use-module is for defining your dependencies correct?
<pinoaffe>Bumblehorse: use-module makes the public variables defined in the specified guile module available to the scope
<pinoaffe>so if you want to refer to a package "blarg" defined in the module (gnu packages blarg), you'll need to "use" that module
<Bumblehorse>pinoaffe: This may be a stupid question but how do I figure out what modules to use? I'm trying to package gtick and following a similar definition to the gnu hello example because it uses the gnu-build-system. When I --install-from-file the example hello.scm it just works but with gtick it doesn't. I think this may be because I don't have the required modules.
<dustyweb>where are channels checkouts stored generally again in one's homedir?
<roptat>I covered almost 3% of the translation until now, I'm at 93% complete :)
<pinoaffe>Bumblehorse: you'll need some modules for the build system and for the dependencies, for dependencies you can run `guix package -A blarg` to find out in which module / file the package blarg is defined
<morgansmith>mfg: I really need to do some school work today so I'm just going to leave you with this: https://paste.debian.net/1169519/ it looks like it works, but once it's done installing, it no longer knows where any of the perl stuff is
<morgansmith>Also feel free to submit any patches I've helped you with. Maybe put my name in there somewhere but I'm not picky :P
<Bumblehorse>pinoaffe: After reading some logs, it complained that it didn't have the command autoreconf. I then did "guix package -A autoconf" as you said and it told me it is found in autotools. I then added (gnu packages autotools) and after reading the logs again, it gives me the same error. Do you have any insights?
<apteryx>civodul: I'm trying my hand at creating a release, going through release.org. So far, I'm stuck where the Makefile uses an old PACKAGE_VERSION = 184.108.40.20636-eaae3-dirty.
<apteryx>I have a 1.2.0 entry in the NEWS file, and a v1.2.0rc1 tag
<OriansJ>anyone have a guess why guix archive --export --recursive bash | guix archive -t would result in: guix archive: error: corrupt input while restoring archive from #<input: file 0>
<OriansJ>So how do I get guix to create .nar files that it will consider valid?
<OriansJ>a bug that goes back to v1.1.0 as far as I can tell
<OriansJ>using /gnu/store/p7jcjqcaw66g05zjim0famggfwpna8k0-guile3.0-guix-1.1.0/bin/guix produces identical results (for those who which to check at home)
<Bumblehorse>When I try and install a package I'm trying to define, no matter how many autotools relate modules I add, I always get the error "./autogen.sh: line 3: autoreconf: command not found" in the logs located in /var/log/guix/drvs/. Does anyone know why this is the case?
<leoprikler>raghavgururajan: not directly, but if you guix time-machine -- environment guix -- sudo guix system reconfigure, it should do what you want in a rather roundabout fashion [remember kids, don't do this at home]
<raghavgururajan>leoprikler: Hmm, `guix time-machine --commit=d9dfa10acb8b543e75efcc7b180fd56d6b3ded60 -- system reconfigure /etc/secondary.scm` is computing something.
<leoprikler>oh, yes, it will do the build, but it won't have permission to change your bootloader
<cbaines>raghavgururajan, given data.guix.gnu.org doesn't know about that derivation, I'm guessing it's new in the last few hours. ci.guix.gnu.org's last successful evaluation of master was ~20 hours ago I think.
<cbaines>lfam, I'm not sure Cuirass supports retries yet, but anyone with SSH access to ci.guix.gnu.org could just request it to be built. Providing guix publish caches the substitute, it would probably stick around for some time.
<civodul>maav: hey, i figured this would provide more context for translators
<maav>one detail maybe you have on the top of your head, ~* caused problems with msgfmt or was only for ngettext? if not, maybe it can be used there, to avoid the or: "Encryption: ~:[Yes (label ~a)~;No~a~]"
<civodul>yes, note that i avoided ~* here because of that
<civodul>i think it's really just ~* that caused problems, not ~:[...~]
<maav>yep, i meant on the #f branch, surely ~:[ should not generate them, and i'm not sure if ~* was only with plural forms when caused problems... i'll take a look after the release to that gettext code to see what can be done
<maav>my contributions there have been quite scarce, i have to keep up with my plan of binding libgettextpo...