<mbakke>ArneBab, lfam: no particular reason gst-plugins-base has GL disabled; I just could not get it working easily but did not spend a lot of time on it ... perhaps things have changed in the mean time :)
<jpoiret>well if that's the case: i cannot reproduce your slim issue, I can login fine in a vm
<wingo>cbaines: is guile 3.0.7 being hard to build a guile problem or a bordeaux problem ?
<cbaines>wingo, I think it's a Guile problem, bugs #43521 and #45788 might be related (or I could be confusing different issues)
<vivien>I migrated my server to core-updates-frozen, it was rather painless.
<wingo>who are the guile maintainers, they are very irresponsible
<utis>isn't full-disk encryption with encrypted swap a common desire? might be nice to include that in the installer. also, i after making an encrypted root from the installer, i had to type my password twice.
<abrenon>utis: I guess it was once in grub and once during early boot ?
<pseudonymous>Tried installing Geary (mail client, GTK3 and Vala). Got an error. On Ubuntu 20.04 LTS, Guix 1.3.0 sourcing /etc/profile.d/guix.sh in my shell. ( https://pastebin.com/tneYw7YT -- too long for debian paste). Any ideas ?
<pseudonymous>And on a more general note - I'm getting the sense that it's early days for GUIX, is it ? Is it normal for builds to break etc in a foreign distro setup ?
<guixnewcomer>Hello Everyone, I hope you all are doing great! I really love the work you guys are doing with functional and reproducible package management. I wish to to be part of it and want to try out guix. I am having a month long holiday from the Second Week of December 2021 to First Week of January 2022. So, My laptop is not needed for productivity work
<guixnewcomer>and can be left unattended except for a few hours in night. I wish to utilize this time to setup a guix system for me. I am fairly determined and have done a Linux From Scratch build and a gentoo linux install. So I can say I can read and follow the document. I am not a developer or a programmer though having taken a weekend python workshop in
<guixnewcomer>college. I have a few questions before i jump to guix: 1.) How much storage will it consume. (I have a 250GB SSD, an External HDD and USB drive to backup my data). I can spare 70-80GB for guix system easily. 2.) How much time does cyclic compilations will take if I am using Pandoc, LibreOffice, firefox, chromium, transmission and sway or i3, mpv on
<guixnewcomer>the base install(Packages from Arch Base). I can spare maybe a week and some hours on later days for compilation to finish. My system is LG GRAM 14Z990 Ultrabook with core i5 8th gen CPU and 8 GB of RAM. I am currently running Arch Linux on it. Thanks in advance to everyone.
<rekado>guixnewcomer: if you’re using binary substitutes you generally won’t have to compile things at all.
<rekado>Guix will automatically download binaries from ci.guix.gnu.org (unless you didn’t authorize it)
<guixnewcomer>Maybe i am confused with void and arch packaging policy. They say void splits the package whereas Arch do not so arch pull extra packages. I do not want that. I want to have packages i need.
<pseudonymous>guixnewcomer: it sounds like the difference between a binary distro (Arch, Ubuntu etc) and a source-distro like Gentoo. When you compile packages, they typically probe for libraries to figure out whether a feature should be enabled or not. Typically, you can explicitly request a feature be disabled if you want to. But all of this supposes you compile yourself. Binary distros have to make those choices for you, and so tend to enable
<pseudonymous>most features, meaning packages are a bit larger and have more dependencies. I don't know how to avoid that outside of Gentoo, for instance.
<nckx>Another niggle is that inferior packages seem to be built at evaluation time (before the <operating-system> is returned). I haven't checked yet if that's accurate (they are definitely built sooner than you'd expect) and if there's a way to avoid it.
<ss2>So I can only place a definition to a package?
<jpoiret>utis: the store is where guix stores all built derivations. The problem would be that any application running on your computer would be able to read your luks key, not the ones running as root
<apteryx>zimoun: good! it'll work in our case since we set JULIA_CPU_THREADS
<zimoun>apteryx, the patch seems to tweak something but I miss why. For instance sprint(show, Dates.Date(1, 1, 1)) return the string "Date(\"0001-01-01\")" instead of "Dates.Date(\"0001-01-01\")". Somehow, it comes from «sprint(show». Hum!?
<zimoun>apteryx, if instead of ’using Dates’, I do ’import Dates’, it works. Double hum! :-)
<nckx>If a good case could be made against including it they would have by now managed to do so in far fewer messages, IMO.
<apteryx>well, when you know the history of the zfs licensing story (designed to be incompatible with the GPL), it's only natural that people would question the move of adding it to a GNU-endorsed distribution.
<zimoun>apteryx, rebuilding to check everything is fine :-)
<nckx>Sure, but it doesn't need a seasonal reboot from scratch.
<dstolfa>apteryx: that's also a rumor from 1 person, which is disputed by everyone else involved
<nckx>dstolfa: It's not really a rumor, or at least not one concocted by one person.
<dstolfa>apteryx: it being designed to be incompatible with the GPL. from people who were directly involved, everyone but that one person claims that the CDDL is the way it is due to contributed vendor code
<dstolfa>the statement is parroted by others that weren't involved, but i've yet to hear it from anyone else that's ex-Sun and was involved in opening up solaris
<nckx>Like, this is also the position of the SFC, they are quite explicit about it, and their argument was not, IIRC, ‘Danese said so’. I'm not saying it's uncontroversial (hah) but this seems like an oversimplification also.
<mjw>I do believe there are others who truly believe the CDDL wasn't explicitly designed to be GPL incompatible.
<mjw>But it doesn't really matter if it was deliberately designed to be GPL incompatible or that it was just a happy/sad "coincident" that it turned out to be.
<nckx>https://sfconservancy.org/blog/2016/feb/25/zfs-and-linux : ‘We believe Sun was aware when drafting CDDLv1 of the incompatibilities; in fact, our research into its history indicates the GPLv2-incompatibility was Sun's design choice. At the time, Sun's apparent goal was to draw developers away from GNU and Linux development into Solaris.’
<zimoun>dstolfa, yes. Maybe. I miss the final point. ZFS is free software so it possible to distribute source code. It is not possible to distribute binaries. That’s all.
<dstolfa>nckx: it doesn't. i never said it's one person saying it. i said it's only one person out of those involved in creating the license saying it :)
<M6piz7wk[m]>i wish more ppl in FLOSS realized that or had background in economy to adjust GPLv3 for those things as without it they just use full proprietary and are not influences to respect FLOSS values
<dstolfa>there are many people reiterating that statement, but i've only ever heard it from 1 person actually involved in the process
<rekado>hmpf. The fancy apt progress bar keeps messing up my serial connection.
<nckx>No, you said ‘a single person’ without qualifiers.
<zimoun>ouch it is so boring to push code to GitHub because their TOKEN authentication. Just for sending a PR. Arf!
<nckx>We could make the case that Oracle is so hostile that we don't package their thing out of spite^Wprinciple, but then that has to be made explicit and we can discuss it. Now it's ‘there might be legal issues! beware!’ which talks like FUD and has done very little walking so far.
<apteryx>dstolfa: that's just my opinion, but a FSDG distro shouldn't go out of its way to make life easy, trading political points in exchange
<M6piz7wk[m]>does anyone even care of economy in this scenario here.. If you want company to use FLOSS license then enable them economically to do so.. Companies just want more profit so if they gain more from being FLOSS they will jump on it
<M6piz7wk[m]>they won't switch on FLOSS if they can't afford it economically
<dstolfa>apteryx: it's free software. what exactly is the problem with distributing sources of free software under a copyleft license?
<dstolfa>we could then make the argument to remove rust because it's a single implementation with a trademark owned by mozilla. suppose i have a problem with that and keep spamming emails about it, does that make my point any more valid?
<M6piz7wk[m]>and it's very possible to earn more profit on FLOSS license e.g. gitpod case
<mjw>dstolfa, working on that (gccrs) but it will take a bit.
<jpoiret>M6piz7wk[m]: the CDDL licensed ZFS code doesn't make them any money any way
<nckx>dstolfa: <talk to Simon Phipps> Thanks! About what exactly? The first quotes I find are about the original intent of choosing CDDL, which I don't really care about, it's not relevant to (not) redistributing ZFS today.
<dstolfa>M6piz7wk[m]: oracle can't use CDDL licensed changes in their ZFS implementation, so OpenZFS and Oracle ZFS are not compatible IIRC
<apteryx>dstolfa: I was just stating a general opinion I hold (don't navigate licensing grey areas, especially if they don't align with the overall vision of offering a GPL free software system)
<M6piz7wk[m]>also looking at the CDDL license it seems to me that oracle is committed to FLOSS, but they aren't given the opportunity to be a FLOSS company like from what i was reseaching in that when there was this whole issue with it in linux kernel they would have a significant losses on GPLv3
<M6piz7wk[m]>also case for FSF to be more politically and economically competent to do investments to enable such innovations
<apteryx>M6piz7wk[m]: you are right, it's "optimized" to protect your user freedoms
<M6piz7wk[m]>Like take scenario: you need 50K USD to develop a software, it's very complicated to make that FLOSS and lot of people just give up even trying to make it FLOSS.. that oracle even went through and made CDDL-1.0 is a step in the right direction in this area that FSF evidently don't care about
<mjw>M6piz7wk[m] I think you are discussing in the wrong channel.
<ss2>is there a way to direct the messages from the application ths shepherd service is trying to spawn? It is is failing, and I never see an error.
<M6piz7wk[m]>apteryx: apteryx: i am not saying that it's impossible to develop something that even needs 200K USD to develop as FLOSS i am saying that it's demanding to develop the business plan and you often get into areas that make it impossible to do so at the time
<f1refly>So I tried installing guix to a partition on my laptop. my laptop has a bios-grub partition, a /boot for my current system (arch) and a /boot for the new system (guix). the fourth partition is a luks volume that contains my logical volume group on which the root filesystems for arch, guix and my /home reside
<f1refly>obviously, with such a setup, grub.cfg cannot contain references to /gnu/store, but after installing it does. I pretty much followed the example setup where they explained how to define the mappers for lvm and luks, but somehow guix didn't figure out it needs to actully put stuff on /boot
<f1refly>is there a special step or entry that needs to be done in order to tell guix it needs to copy the things to /boot?
<podiki[m]>okay got wine to build with `guix build wine -s i686-linux --without-tests=plotutils --without-tests=glib-networking --without-tests=lib2geom --email@example.com`
<podiki[m]>disabling/fixing tests on those would probably help improve the i686 coverage as well as fixing the wine build
<jpoiret>f1refly: there isn't a built-in way for now for guix to put things in /boot
***ns124 is now known as ns12
<jpoiret>the way that it works with LUKS devices for now is that GRUB itself will unlock the root partition
<podiki[m]>I'm not sure if I'll have the chance to disable or fix each of those failing tests, but I'll report this on our tracker with links to the relevant logs to see the failing tests
<rostranj>hello, any clues as to what is causing the issue with GUI apps in a foreign distro?
<f1refly>jpoiret: Oh, I did not know grub could do that. Coming from arch, I'm used to the unencrypted kernel + initramfs -> decrypt root-> mount root
<f1refly>I'll try to reboot then (Honestly I didn't try after seeing nothing in /boot, but I'll give it a shot now)
<jpoiret>GRUB has been able to do that for a bit with LUKS1 partitions, and recently with LUKS2
<darth-cheney>hey all, when I have a package def I'm making for, say, some Guile modules, but I also want to install some guile shell scripts on my path after they are available, how do I go about doing that?
<f1refly>jpoiret: I could fix the german and english version, would that help?
<jpoiret>darth-cheney: there is no post-install scripting support in guix currently (and maybe ever). However, making things available to the user when they use a package is something that is completely doable currently
<jpoiret>if you put them in PATH with proper shebangs it should work out of the box
<darth-cheney>jpoiret: yeah is there a way to, say, copy the script to some location on the users path
<jpoiret>f1refly: well, the english one should be ok with the patch, but the german one will need a translation :) i'm not sure how translations of the manual work though
<civodul>apteryx: yeah but it has "rc" in the version string :-)
<unmatched-paren>right ok i'll just log out, make the issue happen, then reboot and rollback. the log should still be there after rollback, right?
<darth-cheney>jpoiret: I'm using the git method (rather than tarball) amd the scripts would be available there. Also using the guile build system since the modules the scripts use are what is being installed
<zimoun`>civodul, apteryx: could you apply to master https://issues.guix.gnu.org/51870 before merging? It will simplify the janitor work after the merge (python2- broken because upstream or because new sanity check)
<unmatched-paren>i deliberately choose nicknames for my username in case i need to share logs :)
<jackhill>oh, I guess there are tools like wgetpaste as well
<unmatched-paren>kak is able to yank the file but i'm not sure how to put a yank into the clipboard in kak
<zimoun`>apteryx: yeah maybe I will endup with #:tests? #f. It is so fragile. Each time, one thing is fixed, another one raised. Anyway, I am doing a break. The test suite is almost fixed, julia test-suite is parall, and julia-build-system now run test in parallel (if I am correct but not well tested yet)
<unmatched-paren>see, i'm not sure which one was from the session after i reconfigured
<jackhill>lfam: re: mailing list post about gvfs and webkit I think the originally complaint was about the closure size which was made worst by the person not liking the services needed webkit to support. At least that's how I read it.
<unmatched-paren>re. tree-sitter, nix do have it, so i'm going to have to learn nix and figure out how they do it :P
<unmatched-paren>nix feels really weird to me... instead of guix install pkg, you do nix-env -i pkg???
<nckx>jackhill: zstd outperforms bzip2 both in speed and compression. The relatively low level used by most transparent uses like file systems will not reach bzip2 sizes (but it will be close, and *much* faster). If you can set the level by hand, you can get better compression than bzip2 that's still faster.
<nckx>unmatched-paren: No, it would just be a property like superseded is.
<nckx>This should not break anything. This sounds like a bug in phosh.
<guixy>I'll check that it's the problem. But it also bugs me a little that the script generated by guix home tries to source a file that doesn't always exist without checking if it exists. It isn't good form.
<guixy>Replacing the dotfiles with what's in /etc/skel and rebooting fixes phosh. I'll try to send a bug report to the phosh developers.
<unmatched-paren>actually i have a question about (s)mac(k)os; does guix care about it? because the tree-sitter package declaration in nix has an if that checks if darwin is the OS, which i'm assuming is important to make the package work on mac...
<nckx>That's already been fixed on master. Failing when a shell emits something on stderr (or whatever is causing phosh to fail) is far worse form.
<dstolfa>unmatched-paren: the BSDs are already free, or do you mean fully deblobbed?
<unmatched-paren>fully deblobbed. according to hyperbola, some of openbsd itself is technically proprietary since those parts don't have a license. not sure if that's actually true.
<unmatched-paren>plus they're removing bits under the old gpl-incompatible original BSD license
<guixy>If a component isn't fully deblobbed it isn't free. That's why linux-libre exists.
<dstolfa>openbsd, just like any BSD, happily loads firmware blobs into devices. they are probably referring to that, as i'd be really surprised if their actual OS code was missing a license (unless explicitly pulled from something in public domain)
<roptat>any java expert who would know what "Scope org.apache.maven.execution.scope.internal.MojoExecutionScope@1a193b10 is already bound to org.apache.maven.execution.scope.MojoExecutionScoped" means? :)
<easbarbosa>yep, im about to move on to Guix System after reading that they are really after flatpaking EVERYTHING!
<apteryx>civodul: one thing I wasn't sure about was leaving polkit* a public, after seeing a comment in pkg-config.scm about %pkg-config doing the same, supposedly for fold-packages
<jab>So I managed to bork my guix system laptop. Guix is still installed, but I must have messed up grub somehow. When I boot the laptop, grub says that it cannot find some file. Which I'm not certain how that would be happening...I'm using osboot, so it should have grub bundled in there anyway....Anyway and pointers for how I could go about
<f1refly>I just changed my luks2 keyslot to use pbkdf2 instead of argon2i, so it should be supported by grub now. I ran guix system init with the configuration i adapted from the windowmanager sample and it ran through without issues, but after rebooting grub couldn't find the lvm volume my root resides on
<f1refly>is there more to it than to define the luks volume and the lvm volume as mappers and the root and home filesystems in the filesystem declarations to get grub to decrypt the root?
<f1refly>All my experience stems from arch where it's common to have a unencrypted /boot, so I don't really know where to begin debugging this
<jpoiret>jackhill: the grub menu itself is configured in the esp partition, but the assets are in the store. So if you see a blue background on GRUB, that means that it couldn't open the store partition
<utis>all right. i'll come back in a couple of years, then
<jackhill>I just moved resume call, but potentially it could be made smarter to track the dependencies :)
<f1refly>thank you for your hard work anyways jpoiret
<jpoiret>jackhill: oh, no, I had not seen this patch. Tbf though i'd like to rework the initrd into something more modular in the long run
<f1refly>jackhill: can you share your encrypted setup config, for reference?
<jackhill>jpoiret: yeah, that was definitley we talked what would be minimally requiered here and it tried it, so decided to post it.
<nckx>apteryx: Thanks! Especially considering I didn't add [PATCH].
<nckx>jackhill: About what exactly? Although I'm leaving again, but I'll be back later.
<unmatched-paren>anyway, looks like the librem is pretty much the only realistic option, and it isn't even that realistic because of the price
<lilyp>lfam: though interestingly, the same holds for 1.18.2
<lfam>It's not just ARM. It's a first generation chip for their new architecture, optimized for lower power consumption and without many architectural tricks (such as out-of-order execution) meant to speed things up
<lfam>We are at the point where ARM is competitive with x86_64 at the top end
<lilyp>perhaps a change in meson makes it so that typos are no longer corrected
<bbsl>I want to set up an environment with `guix environment --ad-hoc ffmpeg` but I want ffmpeg to have Fraunhofer/libfdk enabled. Reading the source of the package I can see that it is a defined a possible `#:configure-flag`
<bbsl>if this was on nix I would reach for an `overlay` but as I understand it there are no equivalents in guix?
<bbsl>so then I am wondering where to look? :) what is what I want to do called in guix?
<vivien>bbsl, you would define a package that inherits ffmpeg, while passing additional configure flags, and use that one package. See substitute-keyword-arguments in (guix utils), to just change the configure flags.
<ss2>That's a better answer. Just had something else prepared.
<vivien>For instance, see how mygui-gl in (gnu packages development) changes configure flags