<Blackbeard>How can I know the way to start something with sheperd <mbakke>also, the CUPS service definition is near unintelligible, so having a configuration there will help when I need it in the future... <dftxbs3e>Blackbeard, `herd` command I think. You can otherwise have a read to the manual, it's well explained there <mbakke>Blackbeard: you'll first need to add the appropriate service to your OS configuration <Blackbeard>How can I know that '$ herd start privoxy' and not '$ herd start privoxy-daemon' <mbakke>Blackbeard: 'herd status' will list all available services <dftxbs3e>samplet, oh hey samplet!! I finally succeeded getting GNU Guix to work on powerpc64-linux-gnu!! <samplet>dftxbs3e: I saw that on the mailing list. :) Congrats! <dftxbs3e>samplet, thanks a lot for the help early on :D <samplet>It was a while, for sure. Success in big porting projects like that gives me hope for more cool stuff in the future (the Hurd for one). <dftxbs3e>samplet, yes! I havent studied Hurd extensively yet but I think it's got to be better at splitting privileges to users on a system <dftxbs3e>whew started building all packages with 56 cores and 64GB of RAM hah - so satisfying seeing this beast at work. <dftxbs3e>do you think it's better to use "-c 8 -M 8" or "-c 64 -M 1" ? <bavier`>which will allow better network/build overlap, and defense against '#:parallel-build? #f' or packages that otherwise don't parallelize the build well <dftxbs3e>bavier`, alright. It would be nice to be able to label builds for build performance tuning. Like be explicit when the build system doesnt support parallel builds or tests and automatically distribute the load across cores <dftxbs3e>I think the #:parallel-build? and #:parallel-tests? parameters do this well. Now there should be a mechanism to optimize performance from that in GNU Guix maybe? <dftxbs3e>All the Perl packages are quite trivial to build and install so they should be massively parallelized, at at amount even higher than core count. <dftxbs3e>Only I/O limits it, and I have quite fast network and disk so <dftxbs3e>I think it would even be possible to automatically label specific packages with build performance information, by running builds for all packages and make something collect metrics and offer code changes suggestions then <dftxbs3e>and even collect metrics on specific build phases! <dftxbs3e>bavier`, "-c 2 -M 64" seems to work even better for 56 cores 64GB of RAM <dftxbs3e>not sure what the bottleneck is since nothing is maxed out really <Blackbeard>What's the difference between using sudo to do guix system reconfigure and doing it as a root user, in terms of disk space or utility <guix-vits>Blackbeard: i think sudo without -i is action performed on behalf of sudoer, not root. idk about space. <guix-vits>Looks like the files are shared in store, for the same pull version, and can be freely reused by any user. <apteryx>Blackbeard: sudo guix system reconfigure is likely to minimize the space used, because you are sure the Guix revision used for the system profile is the same as for your user, which means sharing lots of items from /gnu/store <apteryx>(whilst otherwise you'd have to make sure the Guix version used from the root user (sudo -i) is matching the Guix version used by your normal user to achieve the same) ***sturm2 is now known as stur4
<Blackbeard>apteryx: when I ran it with sudo it warned me that I haven't done guix pull as root <nckx>Blackbeard: Use ‘sudo -E’ instead, although that should be the default on Guix System nowadays. <Blackbeard>I am hoping that I will be able to send my first patch this week <apteryx>can't seem to have 'make check-system TESTS=btrfs-root-os' pass anymore, on current master. <apteryx>there's a warning about disk space, I wonder if something made it use more and then the instrumented VM runs out of it? <noeu>Does anyone know what to do to install the dependent package? <noeu>Sorry, what does mean "htns"? <KE0VVT>noeu: Sorry. I thought your nick meant you used the Dvorak keboard layout: aoeu id htns <nckx>noeu: inputs are never installed, they are made available in the build environment. <nckx>Ideally your package then retains absolute (/gnu/store/…) to them. How that's done depends on the build system and I have no idea how CL does that. <noeu>That's right. The name is a slightly modified version of my family name. I understand that inputs are used when building. <nckx>noeu: Then why are you surprised that they are not ‘installed’? <nckx>Looking at lisp-xyz.scm, the average Lisp packages builds fine with the required inputs and no extra magic. <noeu>I thought that the dependent packages specified in "inputs" would be installed at the same time. <nckx>No, inputs aren't a synonym for dependency and are never installed. <nckx>Dependencies are the store items that remain in the output, i.e. literal /gnu/store/<hash>… strings in the output files of your package. <nckx>noeu: Are you having any trouble actually running the game? If so, share the error(s) you're getting & maybe someone can help you. Not me, though. <noeu>I want to replace Quicklisp with GNU Guix's package management system. I wanted to define a package that is not currently included in gnu packages, such as cl-sdl2. <drakonis>gotta package that one if there's no importer available <noeu>If you define your own package definition, for example, cl-sdl2 needs alexandria, but I was looking for a way to install them at the same time with "guix package -i". <drakonis>is there any way to acquire build artifacts off cuirass? <noeu>Sorry for the slow response because I am not good at English. I will check the existing gnu packages lisp-xyz again to see how it works. <noeu>Blackbeard: I have set $ HOME/.guix-profile in GUIX_PROFILE <ramajd>Hi, I want to modify default channels for Guix, and make a custom iso image. but changes that I applied on `guix/channels.scm` in mirror of guix repository that I checked out, wouldn't be applied on target iso. Could anyone help me on this matter? <nly>have you run `guix pull` after updating your `~/.config/guix/channels.scm` file? <Blackbeard>when I run git verify-commit I have a lot of "Can't check signature: No public key" <Blackbeard>should I just try to fetch the key or is this a bug? <ramajd>nly: things that I performed: 1. make a mirror from guix repository. 2. change the default channels. 3. create channels.scm in my home directory 4. perform guix pull and guix system reconfigure 5. checkout the guix source from the mirror 6. run the guix system disk-image /path/to/source/gnu/system/install.scm <ramajd>nly: just didn't perform the `guix describe` part <ramajd>nly: I had checked the `guix describe` and it points to my mirror repository <nly>Blackbeard: that makes sense, you may not have keys of each commiter in your keyring <nly>ramajd: what do you use to check that your channels are not applied in the disk-image? <Blackbeard>make: *** No rule to make target 'authenticate'. Stop <ramajd>nly: first I installed the iso on another machine and it points to official guix repository. second I mounted the iso and check the contents of `channels.scm` file. both are pointing to the official guix repositories <nly>ramajd: ah i see, guix will *not* install your channels you used to build the image in the new installations *nly is running make authenticate <nly>may take a while Blackbeard: make authenticate -> "Authenticating d68de95 to eab07e7 (11632 commits)" <ramajd>nly: in another attempt, I built the guix from git, and run the `./pre-inst-env guix system disk-image ...` but this also didn't helped me. <Blackbeard>i just did git clone for the first time, then git verify-commit and then make authenticate <nly>Blackbeard: make: *** [Makefile:5800: authenticate] Error 1 <nly>for what it's worth i can reproduce the error <nly>Blackbeard: error: message: "commit 7dc0f1d5a8c248addfcaadf76e45bf601bc51411 not signed by an authorized key: 3D2CDA58819C08C2A649D43D5C3B064C724A5726" <nly>ramajd: if you use guix with channels foo and bar to build disk-image Z, then i Z won't reference foo nor bar, 1 sec, maybe i can find something in git repo <nly>what's the guix official channel url? (for grepping) <nly>ramajd: in guix/channels.scm: <nly>(define %default-channels <nly> ;; Default list of channels. <nly>you want to change that in your local git checkout and or channel :) <nly>ah, sorry for wasting your time, haha. <nly>let me try to build an image myself. will this change show up in guix system vm? <ramajd>nly: sorry, didn't get what you mean. when I built the disk image default channels reverted back to original <nly>i changed the default channel and running guix system vm gnu/system/install.scm <ramajd>nly: ah, now I understand. creating a VM instead of disk image. thanks *janneke just hacked their way through python-boot0 for hurd <janneke>hopefully also have a working glibc patch to remove the python patch later <nly>ramajd: i'm having issues with guix system vm haha. <nly>`guix describe` just sits there with high usage on all cpus <ramajd>nly: so, the vm image can't be created? <nly>vm was created. /gnu/store/...-run-vm.sh runs very slow <ramajd>nly: that means the channels are modified, yes? <nly>sorry, i cannot test this for sure with `guix describe` :/ <nly>ramajd: based off of your channels file, i think you're modifying %default-channels in your additions channel. is that right? <Gooberpatrol_66>what is the correct syntax to specify the python version for the waf-build-system? I put '(#:python python2)) and it doesn't like that. <ramajd>nly: I modified both the 'additions and 'guix <Blackbeard>my first package is installing and compiling \o/ <obaej>[Global Notice] Hello, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <obaej>[Global Notice] Hello, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <obaej>[Global Notice] Hello, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff ***lxsameer is now known as Guest83537
<Blackbeard>View build log at '/var/log/guix/drvs/97/6ybx91nf5rs372l0r4k31czbdj2fwr-widelands-d226dac-checkout.drv.bz2'. <Blackbeard>cbaines: indeed hahaha, it failed again but now in a different way :) <g_bor[m]>I would like to publish the draft blogpost about Outreachy and GSoC soon. <g_bor[m]>The Outreachy contribution period starts on Thursday, it would be great to publish this tomorrow. Wdyt? <g_bor[m]>I am on mobile, so can't pass the exact link. <Blackbeard>Right now I am trying to make a package for widelands <ramajd>nly: is it possible that the channels issue, which we talked before, is related to ENV variables? since default channels in running instance is pointing to official ones ***emyles` is now known as emyles
<drol_>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <drol_>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <drol_>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <asiece>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <asiece>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <asiece>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <Proika>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <Proika>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <Proika>[Global Notice] Hello chatters, freenode will be undergoing maintenance 3 hours from now. Please disconnect from this network, connect to irc.quakenet.org, and join #Quakenet for updates, otherwise, you may be k-lined permanently from freenode. Thank you. -freenode staff <nly>ramajd: you should also update the guix package in gnu/packages/package-management.scm, otherwise you'll get a vanilla `guix'. <ramajd>nly: it seems reasonable. I'll go for test, thanks <nly>I'm also building to see if that makes the difference ***leoprikler_ is now known as leoprikler
***ChanServ sets mode: +o nckx
***nckx sets mode: +q $~a
***nckx changes topic to 'GNU Guix | get it at https://guix.gnu.org | you must register your nick to talk | 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'
***ChanServ sets mode: -o nckx
<nckx>Berlin has less than 1T remaining again 😒 ***deesix_ is now known as deesix
<zimoun>nckx, rekado_: I am a bit late to the discussion but the Bioconductor <raghavgururajan>> I've been so obsessed with GNU Guix that I even forget to sleep at times :S <nckx>zimoun: That is excellent news. <zimoun>nckx, rekado_: the question is: should the "data" package as 'annotation' be pushed to SWH? <nckx>Nor does /3.4.{7,x}/ and directory indexing is disabled :-/ *raghavgururajan have shipped his laptop to Technoethical for unbricking libreboot. Feels empty without it. Like part of his is missing. 😣 <nckx>raghavgururajan: Aw ☹ I hope you're reunited soon. <zimoun>nckx: ah yes, not all old releases, sadly. <nckx>raghavgururajan: I still think you made the right decision vs. soldering wires to an SMD under pressure. <nckx>But maybe others don't hate soldering modern elektronix as much as I do. <zimoun>nckx: but are these packages differ from Bioconductor version? <nckx>zimoun: I'm not sure what you mean. The hash for 3.4.0 .tgz is different from 3.4.7 .tgz, yes. <zimoun>I am not sure what you are asking. :-) <zimoun>The package does not exist at the version 3.4.7 in the Bioconductor 3.4. <raghavgururajan>nckx: I thought of flashing it myself, so that I can learn. But if I did something wrong, that would just add fuel to the fire. Since I have only one device and it's my daily-driver, I didn't wanna risk it. <nckx>3.4.7 had vanished, so I needed to update it. <raghavgururajan>nckx: I thought of flashing it myself, so that I can learn. But if I did something else wrong, that would just add fuel to the fire. Since I have only one device and it's my daily-driver, I didn't wanna risk it. <nckx>raghavgururajan: I'd learn flashing a non-T model first, and indeed one you can afford to leave disassembled on a table for a while if things do go south. <zimoun>nckx: no, 3.4.7 has not vanished. It has been moved. Because /release/ has been updated from 3.9 to 3.10 so now 3.4.7 live under /3. <nckx>raghavgururajan: Just so you know, your previous message had arrived. <nckx>zimoun: Moved to where? As I said, 3.4.0 works using your scheme, 3.4.7 doesn't. <zimoun>nckx: we have a misunderstanding issue. :-) Let me try to reword. :-) <nckx>zimoun: Keep in mind that this is my knowledge of bioconductor release policy: '(). <nckx>Use short and simple words. <nckx>raghavgururajan: Almost the same message. (You added ‘else’.) <nckx>raghavgururajan: What kind of machine are you using now? <raghavgururajan>nckx: Oh, I edited the message after it's been sent. Technically, the same message should change with pencil icon at the end. But technical translation got lost between XMPP<--->IRC. <nckx>Yeah, IRC is very stup^Wrobust. <nckx>(Not actually robust at all, it doesn't even guarantee message ordering *between participants*.) <raghavgururajan>nckx: I am messaging from Phone with Conversations App. I do not have any other device. That's I am losing my shit. *nckx used a phone once, it was awful. <raghavgururajan>nckx: Yeah, I was/am so worried until my darling comes back to me. Because, OutReachy is nearing and also have to finish my CompTIA ITF+ preparation. <nckx>(…*why* would I want a ‘gnu scented candle’ DDG.) <zimoun>nckx: the packages of the current release of Bioconductor live under /release/. Today, it 3.10 and before the October, 30th it was 3.9. At each Bioconductor release, the packages are in one particular version, say now 3.4.7 and before it was 3.4.7 (yet but imagine Bioconductor version 3.4 and so the package was 3.4.0). The package version and the Bioconductor version are not correlated. *raghavgururajan thinks touch interfaces are any dev's nightmare. <zimoun>nckx: however, our scheme is problematic for "guix time-machine" and the /release/ in URL should be replaced by the real release number. There is a patch floating around but I am not able to put my hand on. <nckx>zimoun: Ah, so /3.4/ working was a coincidence? Indeed, it seems so, since /3.9/ returns the same file for 3.4.0. <zimoun>the version of the package is 2.18.0 <nckx>I'm more interested in fixing such failures when they happen than maintaining bioconductor going forward. You & Ricardo are quite competent in that area. <nckx>raghavgururajan: We both have a touch ThinkPad. <nckx>I bust out the touchiness exactly twice a year when I have to prove that something is ‘responsive’ and ‘mobile friendly’. Which I *love* because it means I don't have to buy a slablet for that. <nckx>Then I go back to typing like a human. <zimoun>nckx: I will not express my feelings about Bioconductor and it is maintained because it is ... ;-) Your point is they also updated the package version and in this case, it corresponds to the Bioconductor version. <nckx>zimoun: OK, but say I wanted to put a /3.x/ URL in the place of /release/ for mmusculus-whatsisname 3.4.7. *Which* URL would that have been? <zimoun>nckx: what I wrote before is wrong: bioconductor version 3.9 and TxDb.Mmusculus.UCSC.mm10.knownGene version 3.4.7 then 3.10 and 3.10 (and not 3.4.7 as I wrote). Well, everything is ok and you did the correct thing. :-) <nckx>raghavgururajan: I stand corrected, it seems you can only digipen your way around a smudgeless ungreasy screen. <zimoun>nckx: in fact, it is the same than in Guix. The git commits are the Bioconductor versions and the package versions are well the package versions. <zimoun>nckx: You cannot know which version of the package correspond to which Bioconductor release. I mean, you need to parse their ChangeLog and their list of package+version. <zimoun>Bioconductor is the same than LTS in the Haskell world. <nckx>zimoun: Thank you, everything makes much more sense now. <zimoun>The problem is that our scheme is /release/ instead of /3.x/ so each time Bioconductor releases (and so updates packages), we have broken URLs <nckx>zimoun: M-hm. So how about making BIOCONDUCTOR-URI take a bioconductor version as well? Would that work? <nckx>s/version/release/ for clarity. <zimoun>nckx: yes, I guess. I do not know what rekado_ thinks. <zimoun>or we could also add a fall back mechanism. <nckx>Less enthousiastic about that. To what? Known previous releases? <zimoun>nckx: the good point with /release/ is that it forces "us" to upgrade because packages are "broken". It is "easier" to find broken packages than non-upgraded ones, IMHO. <nckx>zimoun: I love your optimistic view of broken packages as opportunities for change, but that doesn't work in practice. Our mirror/SWH fallbacks aren't perfect but they're too reliable to make that work. Package with broken source URIs have often been broken for months and it takes someone not using/getting substitutes to find them. <nckx>More personally, I don't like normalising upgrading with a gun to our head. <zimoun>we could just expand the 'list' in 'bioconductor-uri' adding all the known releases and it should do the job. At least for the future and the broken URIs. (But it will fix the past, sadly.) <nckx>Since it won't fix the past I'd rather solve this differently than adding and trying an ever-growing list of slightly different URIs. We know the bioc release at packaging time. I look forward to rekado_'s opinion (if they have one) on (bioconductor-uri "3.9" "foo" "3.4.7") or so. I don't think it should be optional if the scheme always works, but not sure about breaking channels. Thanks for taking the time to explain. <nckx>We can do a one-time sweep for broken URIs and fix them, no need to add code to make them keep working IMO. <raghavgururajan>nckx: Yeah, we do. It's good for mobile tablet-like use. But for high input task. Nah. <zimoun>nckx: see bug#36805 it is the patch that I was talking about <MtotheM>how come fish is a dependency for guix? <nckx>MtotheM: It's… not? What's telling you that? <dutchie>it looks like it uses it to generate fish completion files? <nckx>MtotheM: Eh. No opininion, but it doesn't use fish during the build, it just calls ‘pkg-config --variable=completionsdir fish’ to find out where to put the completions. <nckx>It looks a bit strange to me but I'm not an Arch packager. Maybe people configuring a custom completion directory in their fish.pc is a real plague on that system. <MtotheM>ah, I'm installing the guix distribution right now too, but figured it would be interesting to see how it works in other places. since it's made to be portable. <nckx>dutchie: Not even generate, just ‘where shall I put these here static files’. <nckx>MtotheM: Mm. It tries its best. Integrating with other distributions (settings! fonts! environments! stuff!) is an immensely complex task. It's not as seamless as just using Guix System, though. <nckx>And don't worry, Guix System still has plenty of fun bugs to play with. 🙂 <MtotheM>I should probably just build it myself and learn that way though. <MtotheM>Yeah, it seems complex. was difficult to understand at first when i looked into it. but after watching some talks about it from FOSDEM and what not. it seems really appealing. <MtotheM>Let's see how much trouble i have with the libre kernel <NieDzejkob>rekado_: is there any reason to keep a separate asciidoc-with-python-2 package? (currently called just asciidoc ;) <MtotheM>if i'm not mistaken this can keep several versions of the same program right? <nckx>raghavgururajan: i7 (implies IPS multi-touch screen) X230T. Unfortunately not Librebootable. <MtotheM>Ah, that's really nice. could potentially even help me bug test my own programs if I can store all versions I've ever released in the same 'tree' <raghavgururajan>nckx Also, I think next librebootable devices in X series are going to be X220(s/t) and X230(s/t). Stay tuned. 😉 <nckx>raghavgururajan: Never used a T, only X and L (and L makes no sense to me, but my partner has one). Libreboot: I hope so, but I've been waiting for years. Or have you heard of a specific recent development? That would be wonderful. <raghavgururajan>nckx I see. I wavering between X and T, then decided on X for the portability. About that, yeah I kinda asked Technoethical, if there is going to be another X-Tablet model any time soon. I got a reply to wait till this summer. <raghavgururajan>nckx When you are back. let me what chipset you have on your X230-T. <NieDzejkob>when guix lint says "source not archived on Software Heritage", is that something that should be taken care of? <sneek>Welcome back Blackbeard, you have 1 message. <sneek>Blackbeard, nckx says: ~/.guix-profile is correct. ~/.config/guix/current is a profile containing only the latest Guix obtained through Guix pull. You should have both in path (and never install guix into your ~/.guix-profile) but the first is your ‘proper’ GUIX_PROFILE. <Blackbeard>I used the whole commit number and the abbreviated <bavier`>Blackbeard: the package version should typically use the 'git-version' procedure <Blackbeard>bavier`: ahhh hahahaha dammit. And now that you say that I realized it will also fail <Blackbeard>Because git changed but I didn't change the hash <lfam>Specifically the part about version control snapshots <nckx>Blackbeard: If it's still relevant, interrobang in (gnu packages admin) is a very random example of a good idiomatic package. Use let, git-version, and (commit commit). However, if the projects uses git tags like "foo-0.3" or "v0.3" or…, you don't need any of that and can use something like (commit (string-append "v" version)) directly. <sneek>Welcome back nckx, you have 1 message. <sneek>nckx, Blackbeard says: thanks! :) <nckx>raghavgururajan: That's great, if mysterious, news. My chipset is a QM77 Express, w/ HD 4000 graphics, i7-3520M, 16 GiB of RAM. Libreboot would make it perfect. <Blackbeard>nckx: they recently changed from launchapd to github <nckx>(The tab next to commits in the upper left corner IIRC.) <nckx>Which project we talkin'bout? <nckx>(let ((commit …) (revision …)) …) it is 🙂 <Blackbeard>Well I wouldn't say that. Flathub is still using it <Blackbeard>What do you think is closer to the guix philosophy <nckx>Pfffrt. We do prefer releases, but git commits are often acceptable if there's a valid reason. ‘Arch does it’ isn't one, but e.g. ‘git fixes a bug that really ruins the experience’ is. <nckx>If there are important improvements in git that haven't seen a release in a while, I usually open a polite ‘how about a new release?’ issue. If that doesn't get an immediate reaction I upgrade the Guix package to use git, with a comment pointing to the issue URI. <Blackbeard>I have only played the flatpak version and I haven't noticed any major bugs <nckx>Blackbeard: I can't decide, I don't know the game. You presumably do. If the launchpad version is horribly out of date, use github, otherwise stick to a release. I guess that means ‘build 20’, which is a confusing name for a source release. <nckx>Build 20 is from May 2019 by the way. <nckx>I can only say ‘it depends, what do you think’ in so many different ways 🙂 <Blackbeard>nckx: the flatpak version has been stable so far <Blackbeard>But now I don't know how to unpack and use cmake <nckx>Blackbeard: Are you using the cmake-build-system? <vagrantc>kind of depends on what changes are in git ... <nckx>Blackbeard: Sigh. Sorry for not giving you a clear yes/no answer, but it really depends™. I usually try to get a feeling for the kind of project it is (and yes, the ‘3 years apart’ thing is a good indicator). Here's a list of unreleased changes: https://www.widelands.org/changelog/ <raghavgururajan>nckx There is X201-T on the line before X220-T and X230-T. But there is no technical difference between X200-T and X201-T. Since we already have librebooted X200-T, spending time on X201-T would make no sense. <nckx>What I'd do is package the release, but if I hit a build failure that looks like it's fixed in git (e.g. the GCC 9 thing, no idea what default GCC is now), I'd just switch to git and not bother fixing it. <leoprikler>Is there a way to launch a Guix container with a non-existing user? <nckx>raghavgururajan: I thought the 201 was just a minor rev of the 200 but I don't really know much about those models. I did consider buying a 200 for the extra freedums but I really need that extra CPU & RAM ☹ <dftxbs3e>raghavgururajan, :-) - what would make X220 libreboot-able? <dftxbs3e>isnt the boot firmware signed on those..? <leoprikler>e.g. if I want to vsftpd from a container but don't have an ftp user in my /etc/config.scm? <raghavgururajan>dftxbs3e The whole boot firmware will be removed and replaced with libreboot. But one thing that stopping us from making X220 model libre-bootable is their chipset. <dftxbs3e>unless someone cracked the RSA private key, I'm curious what has changed <dftxbs3e>raghavgururajan, AFAIK it's not possible to replace because it's signed and verified by the hardware directly <Blackbeard>Ok. For now I will try the git version as I already wrote it like that. <Blackbeard>I want to see if the cmake procedure works or if I need to change stuff <Blackbeard>When it works I will change to the release version <raghavgururajan>dftxbs3e That's correct. If we could reverse-engineer the chipset, then we can disable and replace the boot-firmware. <dftxbs3e>raghavgururajan, isnt the verification routine just burned into the chipset's silicon..? So that no matter how you reverse engineer it, it's still not possible to replace. <nckx>What do we mean by ‘boot firmware’ here? The EC stuff? <nckx>(Considering coreboot already boots the thing.) <dftxbs3e>nckx, coreboot doesnt replace firmware AFAIK. I just uses a proprietary blob from Intel. <nckx>Hence my question what you mean by ‘firmware’. <dftxbs3e>nckx, well the chipset isnt the EC AFAIK. The chipset is responsible for booting the main CPU <nckx>And that same part's actually libre on the X200s? Wow. <dftxbs3e>nckx, the firmware wasnt signed on these models, I think. Then they became signed and non-replacable by third parties <raghavgururajan>raghavgururajan It can be. That's how even Intel ME is removed in current librebooted models. <raghavgururajan>dftxbs3e coreboot/libreboot is the complete boot-firmware (BIOS/UEFI) replacement. <raghavgururajan>dftxbs3e No,chipset is like combination many chips. One of them has boot-firmware. <Blackbeard>ok I am getting this error " Could not find ICU include directory" <nckx>Blackbeard: And ("icu4c" ,icu4c) is an input? <Blackbeard>yeah that's a cute emoji nckx.. I miss those emojis my phone uses some ugly ones :( <nckx>Oh, thanks I guess, it's from an old revision of Google noto. The newer ones suck. <Blackbeard>So in theory when it finishes I can just do 'widelands' from a terminal <nckx>Blackbeard: Depends on what you're running. ‘guix install’? Yes. <nckx>I look forward to testing it. <nckx>Knock on wood & all that, but that was quite painless. <dftxbs3e>raghavgururajan, is there any chipset/cpu model on which the firmware can't be replaced at all? <raghavgururajan>dftxbs3e Theoretically, you can manage to replace boot-firmware on any chipset. The problem is, compatibility with other parts of the chipset. Currently, libreboot only support chipsets designed for Core 2 Duo CPU models. <MtotheM>I noticed 'clear' doesn't come with the GUIX distro. it's part of ncurses right? <dftxbs3e>raghavgururajan, so with enough effort and without tampering with the hardware, one could even use libreboot on those latest Thinkpad T490? <raghavgururajan>dftxbs3e You could install, but not use. Since libreboot does not come with binary blobs to support some parts of the T490's chipset. <dftxbs3e>raghavgururajan, but could those blobs be free-ed? <raghavgururajan>dftxbs3e You need code-level modification to libreboot to make it usable on T490. <raghavgururajan>dftxbs3e Yeah, that's what coreboot and libreboot has been doing. But coreboot still includes blobs to support more hardware. But since libreboot is free software, the hardware support is limited. <dftxbs3e>raghavgururajan, are you absolutely sure? Because AFAIK the latest models have the chipset built in inside the CPU itself and there's a silicon hardware routine to ensure the firmware is signed by Intel. <dftxbs3e>Until now, I didnt know that it was an effort question, but that libreboot just froze to older CPUs/chipsets because future ones are just not possible to support without Intel's agreement. <raghavgururajan>dftxbs3e I have been telling you based on what I currently know. Things could have changed. You can refer to coreboot/libreboot project. <dftxbs3e>raghavgururajan, Alright. I have informed myself before but maybe not from Libreboot directly. Thanks. <NieDzejkob>MtotheM: yeah, you need to add ncurses to your profile to get `clear' <raghavgururajan><not possible to support without Intel's agreement.> Yes, a.k.a not possible to replace the code that require this, <NieDzejkob>(though C-l (that's lowercase L) does the job too :P) <MtotheM>interesting, never know about that binding NieDzejkob <raghavgururajan>dftxbs3e Almost forgot! In a net shell, coreboot is like vanilla-linux and libreboot is like linux-libre. The former supports most of the hardware, whereas, the latter supports only few. <dftxbs3e>raghavgururajan, I figured that out yes, been a Free Software supporter for a long while :-) <mehlon>I did it again. I forgot the slash <jackhill>MtotheM: yes, clear is in ncurses. Also, I think cntl+l is a commoon terminal emulator combination for clearning the screen <jackhill>NieDzejkob: oops, sorry I didn't see your reply to MtotheM. At least we said the same thing :) <allana>Hi everyone. Just want to share a short story. Over a week ago I believe that my employer's IT department pushed some kind of change that disabled my ability to run GuixSD on a virtualmachine. I am now running GuixSD on the actual hardware and it's going great! The install was a little stressful, mostly due to my own gaff. I had done a practice install on an older spare laptop using the graphical installer. That was new to me and it <allana>worked great. For my newer laptop, a Lenovo T470s, it did not go as smoothly. The graphical installer crashed just after partitioning my hard drive. So at that point I decided to try to continue with the advanced install option. I had done this in the past on my virtual machine for pre-1.0 versions of guixSD so it wasn't so intimidating. I poked around to see where the graphical installer presumably left off before crashing, and I <allana>falsely saw that my disk had actually not been partitoned. I then repartitioned what I thought was my hard drive. It was actually the install media! I am traveling for several weeks and it made things quite scary because I need this laptop for doing my job. Anyway, I had one chance to get everything right, and thanks to the great documentation everything did go right. Thanks for that :-) <dftxbs3e>Software Heritage servers ended up telling me '429 Too Many Requests' - Is it a CDN? <bavier`>allana: glad it worked out in the end. <bavier`>allana: do you think you'd have time to put details together for a bug report? <allana>bavier`: Unfortunately I did not save the guile traceback. <allana>bavier`: I don't think this will be useful, but it crashed when scanning for available wifi networks. After I went for the advanced install option, I had to run the command "rfkill unblock all" to get network access. <allana>bavier`: perhaps then the graphicla install may have worked. but at that point I was going to continue along the path that I had chosen. <NieDzejkob>raghavgururajan: the install image from the release is a year old <allana>It is strange also because I did not run into the issue when installing the same GuixSD release on another laptop using the same wifi hardware. <raghavgururajan>sneek later tell bandali: When you are available, please ping me via PM. <nckx>Blackbeard: I'm back. If you paste the error message to the bin in the channel topic (or your favourite freedom-respecting alternative) someone might be able to help you. <Blackbeard>nckx: seems like it is frozen again. I am waiting for it to come back <Blackbeard>I had the feeling that I failed because I ran it without "-c 7" <nckx>Blackbeard: Sounds like your system is out of memory. <nckx>Blackbeard: Wait, your default is more than 7 core—oh. <nckx>How much RAM do you have? <nckx>Not as low as I expected. ***Server sets mode: +cnt
<mehlon>oh, I didn't know libreboot had luks support <mehlon>oh. must be a library issue or something <nckx>dftxbs3e: Not yet, but soon, maybe very soon, once I have time to look into it and make sure everything in the stack now supports it. <nckx>Blackbeard: I usually use ‘guix build’. Could you ‘guix build -f’? That will print a store file name, then we can see what's in its bin/. <nckx>Everything's in place but I was waiting for GRUB. I think it's been merged but I don't know if it's been released. <nckx>OK, now reading the backlog you're not just curious, your LUKS won't decrypt. That's not normal. <nckx>Do you use a non-US keyboard layout? <nckx>Blackbeard: It's the same as --install-from-file, so add your .scm file. <dftxbs3e>nckx, French AZERTY, yes. But I don't even see a prompt <nckx>Blackbeard: I meant dftxbs3e but now I have another weird layout variant to look up later 🙂 <nckx>Sorry, ‘weird’ as in ‘interesting’ I mean. <dftxbs3e>nckx, It's Libreboot that can't decrypt, not GNU Guix <nckx>Sorry, I just got behind my laptop. <nckx>dftxbs3e: Did you use the guided installer or manually cryptsetup? The former should definitely not use LUKS2, but the second might have. *nckx goes back to enjoying food instead of half-following the convo. <nckx>Blackbeard: Briefly. Be back in 8 minutes. <dftxbs3e>nckx, I used the graphical installer, so LUKS1, it seems that the lvm volume group name should match something specific for Libreboot to work <raghavgururajan>dftxbs3e The name mach applies only for lvm-setup. Guix doesn't support lvm. <nckx>Eh, we support booting from LUKS. We don't support booting from LVM. At all. <raghavgururajan>dftxbs3e On start-up, select the option ‘Load Operating System [o]’ in libreboot. It might throw some errors. Just *slowly* keep pressing 'enter'. Once you see "Attempting to decrypt [...]", type the luks password and press enter. Again you will see bunch of errors, just keep pressing 'enter', now you will see Guix's grub screen. That's it. <nckx>Blackbeard: Is there a data directory (find /gnu/store/77mv0sblws9m4l12vl0f7wdcdrkfvwa4-widelands-20-1.d9513d4 -type d -name data)? Sounds like you're close, and all we need to do is tweak some paths (install executables to bin/, make widelands look in [say] share/data, …). If the build system is reasonably sane that should be relatively easy. <nckx>raghavgururajan: The problem is less LUKS or Libreboot than LVM and Guix. I think Guix's GRUB/initrd will still choke on that even if one manages to boot it. <dftxbs3e>raghavgururajan, nothing's appearing.. been pressing enter for a while, I don't see any error <nckx>Damn. I don't think this is something that can be debugged through conversation. You can share your package definition if you want, or try finding out yourself why data/ isn't being installed. <raghavgururajan>dftxbs3e Could you send a picture of what's shown on the screen please? <nckx>raghavgururajan: ‘Choke’ as in ‘issues’ as in ‘die’. <nckx>If LVM support hasn't been added by the next release the installer should probably warn when you install to a pre-existing LVM member. <dftxbs3e>raghavgururajan, nothing is shown on the screen, literally. Just the Libreboot background <Blackbeard>nckx: maybe I should build without guix first right? <raghavgururajan>dftxbs3e Woah! Does that happen right after device start-up? Or after opting ‘Load Operating System [o]'? <mehlon>when you do a manual partitioning it warns you that it will destroy your existing partitions, even when it won't. I think that's a bug <dftxbs3e>raghavgururajan, after 'Load Operating System (incl. fully encrypted disks) [o]' <raghavgururajan>dftxbs3e That's worse than I thought. Is it a new device with libreboot? <nckx>Blackbeard: That's an option. Not one I would take, but then I run Guix System so a ‘manual build’ would be it's own time-consuming misadventure. You do you. <Blackbeard>nckx: haha ok. Anyway I am happy that it is working :) <dftxbs3e>raghavgururajan, Yes I just received it from Technoethical, Trisquel used to boot on it. <dftxbs3e>I just found out something. Trying to read from ahci0 seems to block forever <nckx>Blackbeard: Yeah, it's installing *something* even if that something's broken 😉 You wrote half a package, 2 cheers! <nckx>Oh god GRUBs ahci driver nightmares come flooding back. <Blackbeard>so that means I just gotta make sure all the flags work. <raghavgururajan>dftxbs3e Okay so we can rule-out grub.cfg inside libreboot being corrupted. <Blackbeard>but first I need to go to the dentist. Then I'll look into the flags <raghavgururajan>dftxbs3e Okay, no problem. What partition scheme did you use? MBR or GPT? <raghavgururajan>dftxbs3e Okay, did you install guix alongside trisquel? or was it clean-wipe and install? <nckx>Blackbeard: Aha. We certainly can't expect Guix's cmake-build-system to magically add wideland-specific WL_ flags. Try adding #:configure-flags (list (string-append "-DWL_INSTALL_DATADIR=" (assoc-ref %outputs "out") "/share/widelands") …) to the arguments field. <nckx>Same for _BASEDIR I guess. <nckx>Blackbeard: Erk, dentist. ‘Have fun’ doesn't quite cut it, does it. ‘Good luck’! <raghavgururajan>dftxbs3e perfect! Okay, I assume you are messaging here from different device? Can you do the steps as I tell you here? <drakonis>spinning up a iso with the latest goods just for it <dftxbs3e>raghavgururajan, yes! I use my Talos II. <raghavgururajan>dftxbs3e Awesome. As you start the device, press "C" to enter command line, <dftxbs3e>raghavgururajan, done (btw thanks a lot for the help..) <drakonis>might be a better way to improve my guile chops than running nix <nckx>drakonis: It is. I'd never seen Guile (or Scheme) before I booted the installer and wiped my disc. Now look at me. <nckx>I'm a slightly less shit Guile programmer. <drakonis>nix is sufficiently comfortable that i dont feel any direct desire to do anything to improve things <raghavgururajan>dftxbs3e Hmm. Now I am worried. Can you take and send a short video clip of the sequence? <dftxbs3e>raghavgururajan, ohh.. after a while it did show. Just took time <enderby>Hi, i'm trying to use rust in guix. I see that cargo is installed but what do i have to do so make it available in CLI? <raghavgururajan>dftxbs3e Awesome! So you did `ls` and see (ahci0,msdos1) on the list? <nckx>enderby: How did you install cargo? <dftxbs3e>raghavgururajan, yes! It took some time though <nckx>If you ‘guix install rust:cargo’ the command ‘cargo’ will be available in $PATH. <nckx>And there my knowledge of cargo ends 🙂 <nckx>No idea if it's Guix-compatible. <raghavgururajan>dftxbs3e It will ask to LUKS password (may be in a while). Now type the password and enter. <drakonis>yarrr i'm waiting for the kernel to finish building <drakonis>its probably going to take a couple hours <dftxbs3e>raghavgururajan, sorry, had to double check keyboard layout, it now says 'Slot 0 opened' <raghavgururajan>dftxbs3e Now do the command `configfile /boot/grub/grub.cfg`. And tell me the good news. 😉 <dftxbs3e>raghavgururajan, well it cleared the screen and showed the grub prompt again <raghavgururajan>dftxbs3e Okay then, we found the issue. The grub.cfg file was not generated during guix installation. <raghavgururajan>dftxbs3e Yeah, that's why even the "load operating system" option did not work. Because it could not find the grub.cfg file under /boot/grub <dftxbs3e>raghavgururajan, when I do: 'ls /boot/grub' it says, 'error: disk `crypto0` not found.' <dftxbs3e>raghavgururajan, yes, sorry, the typo was in IRC, not grub :P <dftxbs3e>It indeed says: 'error: disk `crypt0` not found.' <dftxbs3e>are the spaces before and after important, no, right? <raghavgururajan>dftxbs3e If even the above command gives same error, then LUKS partition is locked again. You will have restart what we did. Instead of `configfile [...]` do `ls /boot/grub`. <dftxbs3e>raghavgururajan, well, funnily, using crypto0 worked. <dftxbs3e>I used 'ls' to figure out the exact name <dftxbs3e>raghavgururajan, Maybe I have a special setup, it's possible :P <dftxbs3e>I got the laptop shipped pre-installed from Technoethical <raghavgururajan>dftxbs3e So after ` set root='crypto0' `, do `configfile /boot/grub/grub.cfg` <dftxbs3e>raghavgururajan, yup, I did it! GNU Guix booted! :D <dftxbs3e>raghavgururajan, aha, I had it working from this message: 'raghavgururajan, well, funnily, using crypto0 worked.' <raghavgururajan>dftxbs3e Be careful, first try with grubtest.cfg file and then try with grub.cfg. <dftxbs3e>raghavgururajan, alright. I'll be careful. <allana>Does anyone know why rpm (package manager) is distributed as a package in guix?