IRC channel logs
2023-08-17.log
back to list of logs
<nckx>I don't think I'm reading between the lines (much :-) to think you don't like this package (in a maintenance/security sense), and you're probably right, too. The code is the worst example of Perl. <ulfvonbelow>FWIW I've been using the ddclient service all this time and still am <ulfvonbelow>a while back I had to specify a particular commit to get it working though <nckx>I haven't noticed any problems either, but my IP changes only very infrequently. <mirai>nckx: I did try to get the service up to speed at some point <nckx>The dangers of productivity! <ulfvonbelow>although to be fair the api for the service I use is dead simple and hasn't really changed, so even if I had to switch from ddclient, it'd be pretty easy to just use curl <mirai>but got fed up with the config format that essentially is a sequence of commands (so something like 'daemon','foo','nodaemon','daemon') is “perfectly valid” <mirai>coupled with the rather “hopeless” codebase (in kinder terms, you'd rather do a rewrite than maintain it) and the (now reverted) archival you can see what my enthusiasm is for it :-) <piptown>when trying to switch the login manager from GDM to another login manager. How do I accomplish this? <sneek>Welcome back piptown, you have 1 message! <sneek>piptown, nckx says: Hm... 'source' implies to me that you perhaps forgot to make it executable? <apteryx>has there been CUPS regressions recently? <apteryx>someone complains their printer is not working (again) <apteryx>maybe since PAM support has been reinstated for CUPS; cups-minimal can't work (as it lacks PAM) <viaken>nckx: I'm trying to work out how best to keep my Guix configs tidy. Do you keep your home/system/packages all in the same repo? <viaken>I understand you don't want to share your channel, but I figured asking about organization was ok? <gnucode>hmmm...shh-ing into my VPS with 1 GB of ram...guix pull seems to fail. I wonder if the low memory is the problem. <nckx>viaken: I keep everything in a single channel, with nckx/packages, nckx/system, etc. Pretty straightforward. I don't use Guix Home (yet?). <nckx>gnucode: Seems exceedingly likely, especially if you don't have swap. Using -{M,c}1 might mitigate it at the expense of speed if this thing has multiple cores. <gnucode>I might not have swap enabled actually... <piptown>Upon initial installation of guix, I chose to go with Substitutes for packages. How can I change this build things locally? <gnucode>I would upgrade to max amount of RAM. <gnucode>8GB is the minimum I like to have for any laptop/desktop. <gnucode>I have a thinkpad t400, which is pretty close to the same thing as the X200. <piptown>gnucode: ok I'll do that at some point. How can I change my system to not use substitutes? <gnucode>piptown May I ask why you do not want substitues? <gnucode>I personally would not recommend that. You will have a really hard time compiling firefox or libreoffice. <gnucode>I tried disabling substitutes once...never again. <gnucode>I think you need at least 16 Gigs of RAM to compile firefox... <viaken>piptown: You can check builds against the substitutes with `guix challenge`, skip them for a specific build/install with the `--no-substitutes` flag. I don't know the correct way to disable them system wide, but moving /etc/guix/acl does the trick. <aarcov>piptown: if you're wanting to verify packages and build from source as a matter of principle, you might be able to hunt through ebay or your local equivalent and find a desktop/tower system with more ram and use it to build the packages from source and host your own private substitute server. I can say that I have had issues with some VM's when I didn't give them enough ram <piptown>hey how do add bspwm to my config.scm? <PotentialUser-58>I installed guix in ubuntu 22.04 by running installation script in regular user ($ sudo /PATH/TO/guix-install.sh) but directory ".guix-profile" locate at /root/.guix-profile and I could not use guix in current user without sudo. what wrong with my installation? <aldum>if you installed it with sudo, you did not install as regular user <PotentialUser-58>But it should install with root which mention in manual. How to install as regular user? I have another archlinux machine and I ran with same command ($ sudo /PATH/TO/guix-install.sh) in regular user account and it installed as regular user. What's the difference? Shouldn't be the same with same? <PotentialUser-58>[1692247633.360]: [ PASS ] activated root profile at /root/.config/guix/current <aldum>on arch, you can install from AUR <nckx>The script doesn't install .guix-profile, does it? <nckx>It provides a bootstrap guix in /usr that does so when you first install something. <nckx>s@guix-profile@config/guix/current@ <nckx>The /root link is created because it's how the daemon is kept up to date (on a foreign distro, you are expected to pull as root once in a while, in addition to regular users). That's the only reason it exists and it's an implementation detail you shouldn't worry too much about. <PotentialUser-58>It seems to be the problem of sway wm, it works the same as arch machine in tty and gnome. However, my arch machine also running sway, I still not know what's wrong about that. (Archlinux .config/guix both existed in root and regular user but ubuntu only exist in root) <hulten>I have a static IP address configured, but now the service 'networking' cannot load. <hulten>I do not mind 'networking' is not loaded, but CUPS minds. <hulten>Can I tell the service 'cups' that it does not need 'networking'? <nckx>The static-networking service should provision 'networking though? <jab>friendly reminder: I have access to a Talos II through my friend. So if there is some guix package that you might me to try to build on power, then please let me know! <minima>hi, i have a gexp where a command expects some env variables to be set; this minimum example doesn't work https://bpa.st/7VRA is it because i'm using bash for the example or am i using setenv wrong? <minima>the actual gexp would use another executable but i've started thinking that my env variable doesn't get set - so i thought of that minimal example with bash <gcarlos>where I can download a prebuilt database for use with guix locate? <civodul>gcarlos: hi! for now there's no such database; the goal is to eventually distribute one from ci.guix for example <gcarlos>how can I use things like nconfig and menuconfig from inside a container? It seems that I've no terminfo at all available, so I get this: "error opening terminal: unknow" <dthompson>civodul: hey it would be "real nice" if guile-next could be updated sometime soon now that everything that guile-hoot needs is in guile main. what do we need to consider before updating it? <efraim>gcarlos: try adding ncurses to the list of packages <gcarlos>efraim: it already is in my manifest, but this error is for menuconfig. For nconfig i can't even build, I get a missing library error "ld: cannot find -ltinfow: No such file or directory", this is why I asked before about guix locate, I don't know what package provide it (ncurses-with-tinfo seems to provide libtinfo, not libtinfow) <minima>hi, i'm (naively) trying this command to play with nmcli `guix shell --no-cwd --container --network bash-minimal network-manager -- bash' - but when i then lunch nmcli from the shell i'm told `Error: Could not create NMClient object: Could not connect: No such file or directory.' <jpoiret>minima: network manager communicates via dbus, you'll have to share some stuff with the container <minima>jpoiret: ah right, i see, thanks <jpoiret>ah, and in general, network-manager needs a system daemon to be running <jpoiret>so if you don't have nm configured on your computer already, most likely it won't work <minima>qemu and a full OS definition might make more sense for this bit of experimenting that i'm working on then <razlix77>totally not getting why (local-file "../../some.file") works in most places but not for keepalived :( <cbaines>dthompson, I think guile-next can be updated easily, there's nothing to consider <dthompson>(I forget these sorts of things all the time) <andreas-e>apteryx, roptat: Same for lieserl, actually. <civodul>apteryx: i've restarted cuirass-remote-worker on overdrive1, in case it helps (it was up and running according to the log, but who knows) <apteryx>this machine is using wireguard to stay connected, right? <minima>hi i'm trying to inherit from an operating system definition, like this: https://bpa.st/O3YQ but when i get to build the image i get an error `inherit: unbound variable' <minima>ah obviously, i'm overwriting `operating-system', which i guess is bad, i'll try with more sensible var names <civodul>apteryx: yes, it's runs WireGuard; it could be that the remote-worker/server got out of sync for some reason <apteryx>(otherwise I'd have suggested to use the recently added (monitor-ip? #t) option to guard against loosing the link) <apteryx>ACTION notices monitor-ips-internal, which should read monitor-ips-interVal <apteryx>is there a 115.X release for both icecat and thunderbird we could update to? <civodul>though... how long this can be sustained is an open question... <apteryx>we'll need bots, like NixOS, and efforts on the tooling <apteryx>civodul: overdrive1 is building stuff now, thanks <jab>apteryx: wow! That's incredible that we beat fedora! <apteryx>jab: yes! makes we wonder if the data of repology.org is really accurate w.r.t. fedora <civodul>i think Fedora is moving high-level things to Flatpak <civodul>also, they certainly don't have zillions of Haskell/TeX/R packages <jab>and they don't have a lot of package varients do they? <mirai>fedora flatpaks usually still have a rpm counterpart <andreas-e>I have been reviewing a package that puts libraries into lib64/. Is that okay, or should we manually move them to lib/? (It is based on cmake, and I do not know if there is an easy way to force lib/) <apteryx>it may be; Red Hat is quite selective in general to keep their scope limited <apteryx>andreas-e: I was able to restart lieserl's cuirass-remote-worker: ssh 10.0.0.14, then sudo herd restart cuirass-remote-worker <apteryx>it'd be useful if we could see which architectures each machine supports on the cuirass web pages <andreas-e>apteryx: Good to hear this! And it now builds packages. I will remember that it is that easy. <andreas-e>apteryx, civodul: But overdrive1 has now disappeared from the workers page. <cbaines>apteryx, yep, I think QA hasn't even submitted builds yet (you can see that by following the "View comparison" link) <cbaines>bordeaux.guix.gnu.org has 98% substitute availability for some recent revisions (just prior to the elogind update), so the ~8% missing substitute availability for the qt-updates branch looks to correspond with those 1935 changed packages. <civodul>andreas-e: i think it's a bug (now fixed) in Cuirass, whereby remote-worker wouldn't notice worker pings in a timely fashion, and thus mark them as unavailable <apteryx>ah, so it just needs a reconfigure/reboot? <andreas-e>It is back on the workers page (without showing package builds, but this is sometimes a bit behind). <andreas-e>apteryx: I agree it would be nice to show the architectures, but they are also easy to memorise: hydra-* is x86, *p9 is powerpc-9, and all the weird names are ARM. <elpogo>where does guix pull store the clone of the guix repo? <andreas-e>You can also click on the machine names and see their architectures (plus a zabbix error message). <civodul>andreas-e: i'm skeptical about the "texlivebin" trick (removing hyphens), looks like it could lead to interesting issues down the road <andreas-e>elpogo: I think in $HOME/.cache/guix/checkouts <andreas-e>civodul: Well, one should ever only install either "texlive" or a mixture of packages from the modular "texlive-*". Hm, ngz just modified things so that texlive and texlive-biber are compatible. Maybe they modified the profile hook, and the trick is not needed any more? <elpogo>ty andreas-e . you saved me a lot of bandwidth <andreas-e>didi: There is emacs-next, which is at 29.0.92. <podiki>building, so you could locally use these patches and get (some) substitutes for instance <apteryx>cbaines: well done with communicating the status of QA via user messages such as "Builds for new patch series suspended as master branch substitute availability is low for: armhf-linux i686-linux " <andreas-e>cbaines: What is it based on? i686 is green on ci by being just above 80%. <cbaines>the idea is to have QA wait until the master branch is in an OK state before submitting more builds for patches <cbaines>indeed, although given how few we currently have going, even one or two more machines would make so much difference <apteryx>answering a previous question, yes there's a 115.x ESR for both firefox and thunderbird, so we could try upgrading to them <didi>andreas-e, podiki: Thank you. <ngz>andreas-e: yes, I modified the profile hook, which used to be triggered by a gross condition. <podiki>build machines meaning everything but x86-64 or that included? <ngz>andreas-e: However, I (!) still frown upon anyone mixing monolithic and modular packages ;) <andreas-e>ngz: Yes, we should not mix them. But maybe the texlive-biber package is a useful exception. Since I am not actually using it currently, I do not really know. <cbaines>podiki, that included, the bordeaux build farm only has 3 build machines for x86_64-linux/i686-linux, so it struggles with keeping up with master and building packages for patches and branches <andreas-e>The safest bet will still be to install *either* texlive *or* a collection of texlive-* packages. <andreas-e>ngz: I did not know about the texlive "systems" and "collections". Are the collections a complete partition of the packages? And the systems a partition of the collections? <ngz>andreas-e: If we were to keep `texlive' around, I think a better fix would be to compile the "biber" executable right from the source (this is what texlive-biber does). But I think we should get rid of `texlive' as soon as possible. <ngz>andreas-e: schemes are bundles of collections. collections are bundles of packages. I think collections are complete partitions of the packages, although I'm cannot certify it there's no intersection between them. <andreas-e>texlive is about the amount of complexity I can muster... I do not feel comfortable to work on the modular thing. Indeed as soon as we have all the packages, the monolithic texlive could go away. <andreas-e>Sorry, I meant "schemes", not "systems", indeed. <ngz>Right now our modular TeX Live covers between 50% (pessimistically) and 51% (optimistically) of the packages. <andreas-e>So the plan would be to package all packages, and then to create metapackages with the collections that pull in all the packages as propagated-inputs? And then schemes with collections as propagated inputs? <ngz>It's not difficult to add packages, the importer does a good job. It's just tedious. <ngz>Actually, I import recursively collections. I add the packages, with the collection at the end. This way, I can try to build the collection and check if every package in it builds before pushing. <ngz>At the moment, there are only 4 collections missing: collection-bibtexextra (I'm on it), collection-fontsextra, collection-latexextra (this one is a beast with 1000+ packages), and collection-publishers. <ngz>I'm not counting collection-texworks, which is about a Windows-only binary. <ngz>Once these collections are complete, we can add `texlive-scheme-full', which propagates all collections, and declare TeX Live complete. <ngz>(I already added all other schemes) <apteryx>ngz: thanks for you work toward that goal! <apteryx>you should consider joining the tex team :-) <ngz>There are a lot of things that I do not know about TeX, for example, I'm still clueless about LuaTeX. I don't know if it works properly (I expect it to work as well as in the monolithic TeX Live, but how well?) <andreas-e>Currently only rekado is mentioned as a tex team member. <apteryx>i'm sure you already know more than most of us <ngz>Then we're doomed ;) <andreas-e>If the framework is correct, then there is no reason it should not work in the modular setting also. <apteryx>can I cause a build that fails post install phase to preserve its incomplete /gnu/store/ output? that'd make debugging much easier <apteryx>otherwise I have to rebuild the thing using a custom prefix, and tediously rewrite many of the build flags to point there (qtbase) <ngz>andreas-e: True, but there is a suspicious file, namely "texmfcnf.lua" that should probably be tweaked but isn't. There are also reports of strange behaviour from LuaTeX. <andreas-e>cbaines: harbourfront will remain in operation until towards the end of the year. Whatever this means, but in any case beyond end of August. <cbaines>andreas-e, that's good :) although we should still get on with working out what to do next regarding build machines <dthompson>anyone else getting errors running make on master? for me `MAKEINFO doc/guix-cookbook.de.info` fails <andreas-e>cbaines: We shall have to try setting up guix and the builders on our openshift instance. <apteryx>dthompson: time to 'guix pull', I think <apteryx>I've recently patched po4a to support partial menu entries <andreas-e>dthompson: I get the following error on aarch64:HELP2MAN doc/guix-copy.1 <andreas-e>help2man: can't get `--help' info from guix copy <andreas-e>Annoying, but everything else still compiles. <apteryx>or you could try within ./pre-inst-env <cbaines>andreas-e, cool! I have used the dynamic authentication I added to more easily support that use case for hurd builds, so it should still work which is good :) <apteryx>(with your local git checkout having the patched po4a) <andreas-e>Does anyone have a clue on my lib64/ question above? Is it okay to keep it, or should it be renamed to lib/? <apteryx>there's no lib64 in guix as each derivation/outputs are already architecture specific. It should be lib, me thinks. <apteryx>it's probably configurable, you can refer to 'info cmake' <apteryx>also tooling such as relocatable guix pack wrappers will expect libraries under /lib, not /lib64 <andreas-e>apteryx: Thanks, this is also what I thought. Concerning cmake, I will try to have a look. I find it much more intimidating than the autotools... But maybe this is just a question of being used to one or the other. <viaken>I'm a little puzzled about putting a small shell script on the filesystem. I can splat it down with (extra-special-files (plain-file)), but then it's not executable? Is there a better way to go about this? <cbaines>I'm not sure that'll work for a shell script hako <hako>okay, then... invoking bash in program-file to run the script? <viaken>Yeah, that might work. Or I can try actually using Guile. :) It's not a complicated thing. <andreas-e>apteryx: I just saw that we are about a year behind with calibre releases. They went from 5.44 to 6.0, and one of the changes is that support for 32 bit architectures has been dropped, supposedly because Qt@6 does not support them. This is weird, I do not see this restriction in our qtbase package, for instance. <andreas-e>Hm, Qt says they do not ship prebuilt binaries for 32 bit, but that they can still be compiled by hand. <mirai>what package provides epstopdf and pdflatex? <ngraves>Hi guix ! Is there a way to reload the load-path - if the underlying files have changed - in a guile script ? Thanks ! <andreas-e>mirai: pdflatex is in texlive (the big monolithic package), and should be in some other texlive-... package as well. <mirai> /gnu/store/534c6kvw8kp5d5hqbp1qxqq7mjva46fv-texlive-scheme-basic-66594 is empty to me <mirai>`tree /gnu/store/z98df6svjzrym5hka2nijyfzb1wxxsk4-texlive-updmap.cfg-59745/` doesn't contain any bin/ directory <mirai>this is from the wrapped dblatex script <mirai>so dblatex isn't really seeing pdflatex & co. <andreas-e>mirai: This looks like an error in the dblatex wrapping. pdflatex comes from the texlive-bin package, which is not a direct input of the dblatex package. <apteryx>andreas-e: I think they mean by "official" support. <apteryx>I'd be surprised if debian didn't build Qt 6 for i686 <andreas-e>I just tried to update calibre to 6.0. It would require us to add python-pyqt@6, and to carefully juggle the variable names if we also need it @5. <apteryx>as Qt 6 gains traction, it'd be nice to standardize all Qt 5 packages *variables* to end with '-5' <andreas-e>Indeed. PyQt is bit special. We have python-pyqt5-sip, for instance; the 5 comes from the PyPi name. <apteryx>I wonder if with the HTML updater ability to rewrite versions in URLs (#65230), if we could drop the limitation of not updating packages whose release is on github and other forges (the '(not (member host hosting-sites)' code in (guix gnu-maintenance)) <pret7>heya! is there any good way to test a substitute server? <pret7>I have a substitute server, and I think it works, and guix weather works <pret7>but when I try and do a build of a package I know it has it still builds locally <attila_lendvai_>pret7, last time i had this problem the signing key of the server was not installed <pret7>that's in my guix-service-type <pret7>is there any way to check that its working? <pret7>I will say, running guix repl, then ,use (guix) still shows only the default ones in %default-substitute-urls, but I think that's expected <pret7>how can I check what that variable is in the guile that's running guix? <pret7>and the %default-authorized-guix-keys variable <pret7>interesting. guix weather only checks ci.guix.gnu.org and bordeaux.guix.gnu.org