IRC channel logs
2018-01-27.log
back to list of logs
<civodul>if we were to have 10s of 1000s of packages, that'd probably be an issue <civodul>but then there are always tricks we could do to speed things up <buenouanq>idk, guix pull took days on the beagle bone... <buenouanq>and then system init still failed because of `hardware limitations' I don't understand <OriansJ>Out of curiousity does the guix track at FOSDEM support remote talks? <mbakke>Phew. 200 emails down and still 317 to go. <bavier>has anyone else has issues with the ntpd service constantly respawning the daemon? <atw>or at least it used to happen for me. Looks like it's no longer happening after I used tlsdate to set my clock properly <bavier>it also hasn't managed to update my system time for several weeks now <atw>bavier: let me find the tlsdate invocation for you <atw>I used the above to set the clock on a new computer (hwclock was wrong) <bavier>had to add a "nobody" user for the SSL handshake for some reason <bavier>finally, tor can work again too :) <wigust->bavier: Don't you use %base-user-accounts? <catonano>it's probably a bit too cold, but I love it nonetheless <amz3>yeah, i prefer the first one too <amz3>the guix logo is very good! <catonano>I remember there was a guile module aimed at descriptive statistics and it used to produce warnings when buuilding Guix. Where is it ? <amz3>sorry, midly off topic again, what's the status of l10n / i18n website? <orelozno3>Hello :) Free software end user, I discovered the existence of GuixSD. Happy to have reached here... <orelozno3>I hope to be able to contribute as a guixsd beta tester and maybe to the translation. <adfeno>orelozno3: You're very welcom to do so! ;) <adfeno>If you have any questions, just ask ;) <orelozno3>For now, I would like to install guixsd in qemu but I can't configure the internet connection. <orelozno3>$ qemu-system-x86_64 -m 1024 -smp 1 -net user -net nic,model=virtio -boot menu=on -drive format=raw,file=guixsd-install-0.14.0.x86_64-linux.iso -drive file=guixsd.img <orelozno3>Everything seems to work, I get to the graphical screen then guixsd starts, starts a list of services and gives me the hand in admin root <orelozno3>Hmmm, I don't know very well irc, maybe too much questions as the same time :)! <adfeno>orelozno3: After booting the image, does "ip addr show" or "ifconfig -a" display any device besides "lo"? <orelozno3>If I remember there were 2 devices, the other is named ens3 (a 'eth0' like) <adfeno>And do you expect to work with wired connection? Or do you want wireless instead? <g_bor>I would like to do something I have not done yet. <orelozno3>I just want wired connection. Today ifconfig give me : eth0 <adfeno>orelozno3: ifconfig gave it to you in GuixSD or in your host machine? <g_bor>I would like to get the version string of an input in the package definition. How can I do that? <orelozno3>This qemu command contains connections options <orelozno3>I seem to have read somewhere that by launching qemu with connection options, we had ping problems... <g_bor>orelozno3: So once you got eth0, and once ens3? <orelozno3>It's depend, sometimes it's ens3, other times eth0 <orelozno3>I tried to launch guixsd in differents ways (differents options). It may explain this thing <g_bor>I actually don't know if any problem arises, if you try to configure both of those interfaces... <orelozno3>I bookmarked the adress, I'm going to look at it <g_bor>You are welcome! Should you have any further problems, we're trying to help you:) <g_bor>Also, it might be possible that you are better of managing your vms trough libvirt, but I don't know your environment. <orelozno3>I'm running debian 10 (Testing) and I 've installed qemu but encoutered problems to install libvirt (problems of dependancies <apteryx1>orelozno3: If you use the QEMU option -net user than you are using the SLiRP or whatever way to do networking. It should work from inside the machine without you having anything more to do. Interface name is irrelevant. There are limitations with this way: ICMP won't work (e.g. cannot ping the guest) and you can't reach the VM from the outside network. <orelozno3>Normally I think there is n-othing to do with the options, but it isn-'t working, so I tried without the connection options <orelozno3>apteryx1: $ qemu-system-x86_64 -m 1024 -boot menu=on -drive format=raw,file=guixsd-install-0.14.0.x86_64-linux.iso -drive file=guixsd.img <orelozno3>apteryx1: I just launched qemu with only the -net user option. ifconfig shows only the lo interface <ng0>mbakke: dunno if it was you, but I'm just about to release a new version of gnurl, so I'll send the revert for my last commit together with the update patch <dieggsy>oh, lol. i think i have a parentheses issue <ng0>just waiting for the ftp to catch up so I can send the patch <mbakke>Building evolution-data-server with -std=gnu++11 makes it succeed (on core-updates). <dieggsy>wigust: almost, after some changes. needed to specify jansson package comes from (gnu packages web), and forgot to add the flags to #:configure-flags, not just arguments lol <dieggsy>wigust: well, we'll see how the build goes heh <dieggsy>does the git-fetch method use shallow clone by default? if not, can i make it do so? <ng0>it needs to be modified <ng0>wingo proposed something 2 years ago but it never made its way into master.. so we still clone the full git checkout afaik <ng0>except when you clone something with 2+ GB it will be ok <wigust>dieggsy: You could use autogenerated tarballs with Emacs case, couldn't you? <dieggsy>wigust: not if i want to build from latest git <dieggsy>wigust: been doing it wrong all this time <wigust>dieggsy: How is it differ from changing git hashs? :-) <dustyweb>may be a good thing that go packaging has recently improved if ipfs can become available to icecat users :) <civodul>mbakke: i suspect it's a non-deterministic test failure <ng0>hey.. would it be okay if I provided a "normal" (dynamic) build of a file system for a starter package? I'm making no good progress with xfs since last spring <mbakke>Derp, it succeeded on the sixth attempt. <ng0>I mean, it should be. this application is also for managing XFS filesystems.. we just need the static output for the initrd. <mbakke>ng0: You mean, not providing a static package? Sounds good, I didn't bother for f2fs-progs either. It's not necessary until we support it in the initrd. <mbakke>Also, xfsprogs, excellent. I tried packaging it before but failed. <civodul>mbakke: it's almost deterministic then :-) <ng0>no good progress since january 2017, I think I had it building before that but at some point I had to move my branches <dieggsy>where does substitute-keyword-arguments come from? <dieggsy>and is there any documentation for it <ng0>mbakke: I think I started with xfsprogs 4.08 or something and it's now at 4.14 <mbakke>Cool, maybe they've fixed whatever the problem was. My branch is for 4.9.0. <ng0>it's one of those long optional packages I have.. technically I haven't spent a year with it :) <wigust>Hydra's failed jobs should be restarted manually? <ng0>mbakke: what did you configure it with? anything I could publicly read or fetch from? <mbakke>wigust: Yes, for nondeterministic build failures they have to be manually restarted. If the derivation changes Hydra will pick it up. <ng0>I'm thinking that 2 pairs of eyes and hands have more time to work than just one pair :) xfsprogs shouldn't be that hard <mbakke>ng0: My package is super simple and does not have [arguments], and the only inputs are gettext and util-linux. <mbakke>Happy to help out though, if you get stuck :) <mbakke>ng0: Any way to peruse with a web browser? <ng0>mbakke: yes, that is the webbrowser URL already <wigust>ACTION thinks gource will be build only when some of his input will changes <mbakke>wigust: Oh, right. Do you know if the latest 'master' evaluation includes the commit that fixed it? <mbakke>ng0: clicking that link gives "no repositories found" <mbakke>wigust: Right. I see that no new build job was created on Hydra, so restarting the latest job would not include the commit that fixed it. Tricky. <mbakke>ng0: Wow, that xfsprogs definition is monstrous. Let me know when you've updated for 4.14 and I'll give it a go :) <ng0>last time I worked on it a year ago.. wished I had gitea so I could note the problems of all those branches. right now it's all on paper :) <wigust>mbakke: Because commit didn't change a source derivation, I guess. <mbakke>Yes indeed, since the hash is the same. <ng0>well, ja.. maybe I could drop some of those inputs. really idk where they came from <wigust>ACTION wonders why does restarting the latest job will not include a commit? <mbakke>wigust: Each evaluation is tied to a git commit, and each job is tied to the evaluation. So hydra would check out the old commit and attempt to build that IIUC. <ng0>well the only problem right now is libuuid.a is not found <wigust>so, a broken url will break a package forever, because if you change broken url nothing happens <ng0>while both libuuid and libuuid:static are in the inputs <ng0>if you want to give it a try, I'll push it <mbakke>wigust: Normally that's not that big of a problem, since the source is likely already on Hydras content-adressed mirror. But yes. <mbakke>ng0: Great. I'm currently working on core-updates, but ping me one of these days :) <ng0>okay. I'll pick up work on it probably after 15th <wigust>mbakke: Thank you for explanation! <Apteryx>hello! does anyone has the gource-0.47.tar.gz archive in their /gnu/store, with sha256 of 1llqwdnfa1pff8bxk27qsqff1fcg0a9kfdib0rn7p28vl21n1cgj? <ng0>"hello, I'm a github source and I broke again" ? <Apteryx>*i've reran the fiasco script have have against all our github packages, and only gource and ocaml-jbuilder have problems. <Apteryx>I don't have the previous gource we were using to diff against; in teh case of ocaml-jbuilder a folder was renamed inside the tar. <mbakke>Apteryx: That's odd, the github "release" archives are fetched from Amazon S3 and should never change. <wigust>Apteryx: 1llqwdnfa1pff8bxk27qsqff1fcg0a9kfdib0rn7p28vl21n1cgj should be a good one, it's from GitHub's release <Apteryx>gource Guix hash: "1llqwdnfa1pff8bxk27qsqff1fcg0a9kfdib0rn7p28vl21n1cgj", gource upstream hash: "11zw0cpjv45ls1vj0hj7bn4yp9n7fa0j1d1d26s1zmvfzgp614k2" <mbakke>`guix build -S gource` works for me. <Apteryx>aha, I'm one commit behind 45b48698, that explains it. <wigust>Shouldn't fiasco script return the hash of the release url tarball even for behind 45b48698? <Apteryx>there's also a small API regression I'll send a patch for in Guix that broke the fiasco script. <Apteryx>url-fetch returning #t instead of FILE. <atw>I have an environment file with emacs-ess in it. emacs-ess has emacs-minimal in its inputs. What's the best way to get full emacs into the resulting environment? <mbakke>atw: I think the easiest way is to add `--ad-hoc emacs` to the `guix environment` command. <mbakke>I guess you can add the regular emacs to propagated-inputs as well. <atw>I think I'll (inherit emacs-ess) and propagate emacs. I'd like to do everyting from a file. I had tried something like (map specification->package '("emacs-ess" "emacs")) but it looks like emacs-minimal was arbitrarily chosen in the collision <mbakke>atw: Do you need emacs-minimal still, if you propagate the "regular" emacs? <atw>I shouldn't. I don't know enough scheme to (inherit emacs-ess) and replace an input though. Does the manual have an example of that? <Apteryx>atw: inheritance by default leaves you with the exact same package definition as the parent. <Apteryx>Then you can choose to override some fields of it <Apteryx>or extend/manipulate the inherited ones through explicit referencing of the parent (propagated-inputs (some-alist-operation (package-propagated-inputs emacs-nss)) <atw>Apteryx: OK, just have to find the right alist opertation. Thanks! <Apteryx>I'm trying to find an example I wrote when fixing manpages generation of GCC <Apteryx>ok, found it: 19d27131968355a1da962e52163df51e0e03aba1 <Apteryx>is there a git command to find which branch this commit is on? <Apteryx>git branch --contains <commit> (thanks, stackoverflow) <Apteryx>apparently no branch contains the commit... hm. <atw>I was able to find it via cgit. Also, magit can show which branches contain a commit <Apteryx>I often one to see a specific commit in magit, a la 'git show <commit>'. How do we do this in magit? <atw>I usually pull open the log, then type RET on a specified commit <Apteryx>yeah, me too. That's not as direct though. <atw>does this look right? (inputs (alist-cons "emacs" emacs (alist-delete "emacs" (package-inputs emacs-ess)))) <atw>judging by an error I'm seeing and the package reference in the manual, inputs is a list of pairs, not an alist. What's the best way to manipulate a list of pairs? <Apteryx>example of delete: (native-inputs (alist-delete "texinfo" (package-native-inputs gcc))) <bavier>atw: quoting is used in other package recipes <Apteryx>atw: so I think your delete part is fine. It's the alist-cons which probably needs some tweaking <bavier>atw: e.g. (inputs `(("emacs" ,emacs) ,@(alist-delete "emacs (package-inputs emacs-ess))))) <Apteryx>^ splicing is probably cleaner syntax <atw>Thanks! let me try that out <atw>OK, looks like that's what I want <mbakke>Huh, "python-xapian-bindings" on core-updates somehow has a python3.5 sitedir. <mbakke>Oh, it's set in configure-flags. <mbakke>civodul: Maybe we should disable the test for i686? <civodul>in test_pty, with just a '\\n' that differs <mbakke>python 2.7.14 failed on one of Hydras build slaves a couple of times as well IIIRC. <mbakke>Updating sbcl to the latest (1.4.3) makes the build succeed, but then a bunch of dependent packages fail :/ <atw>How can I launch octave with a GUI? Specifying --force-gui still launches with the CLI <ng0>you need to apply the patch that is up for merge on the guix-patches list <atw>OK, known issue. Thanks! <ng0>it's due to be changed (the Qt gui should be in its own package because of the huge depndency chain).. <ng0>if you trust my remote and want to have it, you could run it from a branch I published Kei's patch in <ng0>I also have the pending Chromium patch in there <ng0>no idea if my substitutes server catched up already building it <atw>Eh, it's not critical for me. I was preparing a little Guix demo featuring R and Octave, but I've got the R part together and I'll leave it at that <atw>I should contribute by testing packages...but I also need to migrate a bunch of stuff from my Debian machine too <groffer>I have guix installed on a foreign distro and want to move /gnu to it's own partition, can I just use "rsync -aH"? <mbakke>groffer: I believe that will work fine. <wigust>ACTION finished with licensecheck, prepares for sending a patch <mbakke>Ugh, for some reason the django 1.11 test suite cannot find pytz even though it's on PYTHONPATH. <civodul>mbakke: oh thanks for evolution-data-server and other fixes <civodul>i didn't even think about C++11 when i looked at it <mbakke>civodul: np. Still a long way to go though :/ <civodul>although Cuirass is not 100% ok it seems <civodul>but you can monitor M-x guix-hydra-latest-builds <civodul>ACTION has been watching cuirass logs for some time <wigust>Is it OK to have a (gnu packages license) module? <wigust>I have at least 3 packages for this. <wigust>perl-regexp-pattern-license, perl-string-copyright, licensecheck