<quiliro>i get an error when using gnome-calculator <quiliro>(gnome-calculator:8823): Gtk-WARNING **: Locale not supported by C library. <mekeor>quiliro: it's just a warning, right? or does the warning have an annoying effect? <quiliro>mekeor: yes...it does not let gnome-calculator operate <quiliro>that is when using it on the command line <quiliro>but when it is just started (gui), it will report: <quiliro>(gnome-calculator:8755): GLib-GIO-ERROR **: Settings schema 'org.gnome.system.proxy' is not installed <quiliro>even if the first message is a warning, it gives me problems with non-us characters <sturm>What do people do in the case where a Python package attempts to download `test_require` requirements during the build process? Patch it to remove these from setup.py and put them in the Guix package definition instead? <quiliro>do i need to guix pull as user when i have done it as root? <catonano>quiliro: yes. As root you update packages available to all users, suchhe as thheh desktop environment <catonano>you update the packages that you have installed in your rofile <quiliro>i don't understand...so, i need to run guix pull as user or not? <quiliro>guix pull is like aptitude update and guix package -u is like aptitude upgrade...right? <quiliro>but as user you cannot do that in debian <catonano>quiliro: yes, you have to, every now and then <catonano>guix package -u updates the deinitions of the packages to the most recent available <sturm>catonano, quiliro: I think this is different between Guix (on foreign distro) and GuixSD right? On Guix the root user's packages include the desktop environment, but on GuixSD the desktop infrastructure is managed by the "system" account. Have I got that right? <catonano>oh yes I was assuming quiliiro is on GuixSD <catonano>on Guix there's no desktop env ad ar as I know <sturm>quiliro: so if I understand right (and I probably don't heh), you need to `sudo guix pull` to update root's package list, `sudo guix package -u` to update any of root's packages (probably not many) and `sudo guix system reconfigure [config file]` to upgrade the system packages. Then separately if you want updates to locally installed packages, you need to `guix pull` and `guix package -u` (as your normal user). <sturm>I think that's the GuixSD equivalent of `sudo apt-get update`, `sudo apt-get dist-upgrade`. <quiliro>what would be the difference between 'guix package -i gnome-calculator' and 'guix package -u gnome-calculator' if gnome-calculator is already installed? <sturm>quiliro: `guix package -i` and `guix package -u` should both upgrade the already installed package - so no difference that I'm aware of <rekado_>Can't locate warnings.pm: No such file or directory at /gnu/store/my4aagbxzxx545624jlzm2z29wdlq63v-git-2.12.2/share/gitweb/.gitweb.cgi-real line 12. <rekado>Hi Guix. I have severe problems on i686 after system reconfigure. <rekado>“ps aux” segfaults, “znc” randomly fails with different errors each time I start it, “gitweb” crashes, etc <rekado>“guix environment guix” fails with either “guix environment: error: fport_fill_input: Connection reset by peer” <rekado>or with “guix environment: error: build failed: illegal option name/value” <rekado>hmm, I had to go back to a version of my i686 system that predates the switch to glibc 2.25 <rekado>as soon as I’m in “guix environment guix” and I’m using new binaries not even “./configure” runs fine. <rekado>is anyone else here using i686 with GuixSD? <civodul>rekado: i saw a patch for libc 2.25 on i686 <civodul>rekado: could you email bug-guix or guix-devel so we can decide what to do? <civodul>it seems we didn't hit the bug while building packages, but i'm not sure if a graft is enough <andijh92>hello. i'm installing guix on ubuntu 16.04. i run "sudo systemctl enable guix-daemon" and i get the error "Too many levels of symbolic links". had someone the same issue? <rekado>andijh92: copy the systemd unit file instead of linking it. ***kelsoo1 is now known as kelsoo
<SovereignBleak>Where do people host their GuixSD servers here? Looks like AWS is presently unsupported. <SovereignBleak>I’m trying to convince my employer to go with GuixSD over NixOS but missing features like this are more than a mild annoyance to business deployment. <SovereignBleak>Ah it looks like someone on the mailing list got it to function but there’s no premade images. <wingo>what is the guix ssl certificate story? <wingo>isn't letsencrypt a service? <wingo>also how do i make an https server with guile, hum. i guess behind nginx <wingo>tho that is sad, maybe i need to hack a thing <civodul>wingo: letsencrypt isn't a service, but it could/should <civodul>and for https i use nginx, which is admittedly not entirely satisfactory ;-) <civodul>although nginx does other things that i'm happy we don't need to duplicate <wingo>like gzip encoding, etags/last-modified-since, etc <wingo>still for long-poll type things i would rather be able to serve directly <wingo>and range requests for video etc <andijh92>when will be the guix pack command be available in the standard binary? <wingo>i think if guile's http server did that then i wouldn't want nginx <wingo>andijh92: it's there already in git <wingo>i guess a release is coming soonish <andijh92>is ./bootstrap ; ./configure ; make ; make install enough on ubuntu? <andijh92>i'll wait for the nex release. safety first :) <wingo>being able to make reliable binary releases is so wonderful <wingo>this nginx configuration stuff is not so nice <wingo>civodul: who makes sure to run guix publish on bayfront? <Mathieu`>wingo: there's a guix publish service (guix-publish-service #:port 3000) on bayfront <wingo>i was getting confused between bayfront.scm and my local file <wingo>it used to be that guixsd.org redirected to gnu.org/s/guix <civodul>andijh92: if you've run "guix pull" recently you already have 'guix pack' :-) <civodul>wingo: i think the redirect is broken :-/ <civodul>i'll check with Andreas, who's still currently in control of this <civodul>Mathieu`: regarding test-lock, did you try --max-silent-time with something greater? <wingo>oddly i had problems updating my nginx configuration -- herd wouldn't reload the new nginx service definition <wingo>as you can see, mostly stolen from bayfront <wingo>ACTION hoping someone is on the guix channel thing :) <quiliro>i am on i686. 'gnome-calculator -s 2+2' works. but 'gnome-calculator' will not <quiliro>(gnome-calculator:16031): GLib-GIO-ERROR **: Settings schema 'org.gnome.system.proxy' is not installed <civodul>wingo: i haven't reviewed the guix channel proposal yet! <wingo>yeah there's a lot there, have to find something minimal that works <wingo>from the potluck side i hope we will have a git repo with guix packages, shortly. <wingo>i got nginx to terminate the https tho and proxy to git, so that's a thing :) <quiliro>methalo_: to reproduce just run gnome-calculator <quiliro>methalo_: that is probably why you cannot reproduce it <quiliro>but it should have been installed if gnome-calculator needed it <quiliro>i guess it is gnome-desktop that is missing then <wingo>ACTION has an fcgiwrap service, whee <quiliro>'gnome-calculator' needed 'gnome-destop' to work correctly. it was giving an error: '(gnome-calculator:16031): GLib-GIO-ERROR **: Settings schema 'org.gnome.system.proxy' is not installed Trace/breakpoint trap' until i installed 'gnome-desktop' <rekado>wingo: hah, good timing. I was going to do this today, but delayed it. <wingo>rekado: fastcgi you mean? perhaps you would be interested in testing :) <wingo>i guess i can just post my config again <wingo>wtf my wip commits need to be signed, blah :) <mekeor>wingo: so, gnome-desktop should be a dependency of gnome-calculator then? <wingo>mekeor: perhaps quiliro was the one you wanted <wingo>i don't recall how these things work. sometimes i have to remember but i forget quickly :) <wingo>note the ~~ in the nginx config is really just ~; it's a format string <wingo>we need a scheme nginx configuration facility :) <mekeor>quiliro: so, gnome-desktop should be a dependency of gnome-calculator then? would you like to add the dependency into the guix-package of gnome-calculator? <quiliro>mekeor: i don't think it is exactly that dependency <quiliro>maybe it is some package that gnome-desktop installed the one that fixed gnome-calculator <quiliro>maybe i should remove gnome-desktop and test <thomasd>quiliro mekeor: running gnome-calculator through strace, I see that it uses gsettings-desktop-schemas <thomasd>that package contains org.gnome.system.proxy.gschema.xml <quiliro>thomasd: how did you run strace on gnome-calculator? <thomasd>quiliro: from terminal 'strace gnome-calculator > strace.log 2>&1' (this generates a lot of output ... there are probably smarter ways to do it) <quiliro>thomasd: and how did you know about org.gnome.system.proxy.gschema.xml? <thomasd>from the error message you gave: “Settings schema 'org.gnome.system.proxy' is not installed” <quiliro>davexunit: is that what is needed for riot? is it good for freedom? (i do not mean to imply it is not good) <jonsger>davexunit: we should package diaspora :P <thomasd>quiliro: I haven't tested if this is all that's needed to make gnome-calculator work correctly. You could try removing gnome-desktop, adding gsettings-desktop-schemas, and see if that solves it. <davexunit>jonsger: go for it if you want. I haven't used diaspora in years. <thomasd>and/or try adding it to gnome-calculator's inputs (or probably propagated-inputs) and submit a patch :) <jonsger>davexunit: it became quite stable. and more mature over the years. running a server for 2,5 years now :) <quiliro>is there a way to make something like freedombone.net on guixsd? <jonsger>but it has also a huge ruby dep tree. and I failed two package a super small piece of software :( <davexunit>I like twitter more than facebook, conceptually, so gnu social/mastodon appeals more than diaspora <quiliro>i removed gnome-desktop and gnome-calculator give no error after that <jonsger>davexunit: and I would say diaspora seems to be more like twitter then like facebook. <quiliro>i am removing 'gsettings-desktop-schemas' in order to test <thomasd>quiliro: probably you don't have gsettings-desktop-schemas installed then? <quiliro>thomasd: so what was gnome-calculator's problem then? <quiliro>it is still running after removing gnome-desktop <thomasd>you mean when you close it and start again, there's no problem? <quiliro>perhaps it was a dependency of gnome-desktop which was not uninstalled when i removed gnome-destop <quiliro>how can i know what else was installed as dependency when i installed 'gnome-desktop'? <quiliro>that is the answer to what was needed for gnome-calculator <thomasd>maybe there's some caching going on, and gnome(-calculator) somehow remembers the location of the schemas, (I'm just guessing) <quiliro>thomasd: so, i should test by exiting the desktop or restarting the machine? <ng0>Hi! I'm trying to init a system, but it times out: <ng0>building of `/gnu/store/8vq4jg43cyxjxlkzmq73igzwnwd3dcw2-glib-2.50.3.drv' timed out after 3600 seconds of silence <quiliro>but gnome is installed as per desktop.scm <ng0>that's on a minimal debian 9 root <ng0>any idea what could cause this timeout? <thomasd>quiliro: oh, but if gnome was installed, gsettings-desktop-schemas was there all the time (same as on my system then) <ng0>the last line before the fail is this: <ng0>PASS: 1bit-mutex 2 /glib/1bit-mutex/pointer <quiliro>thomasd: but it is not because i could not uninstall it <thomasd>quiliro: now I don't have any idea what caused the problem :) <thomasd>quiliro: it's not in your user profile, but in your system profile <ng0>maybe it works when i restart the init again <quiliro>but i did not do anything to my system profile <thomasd>quiliro: that's what confuses me too <thomasd>is your system older than your current guix version? maybe it should be upgraded (reconfigured with recent guix?) <quiliro>how can i see what packages were installed as dependencies of a package i installed? <bavier>quiliro: we don't have a direct interface to that, AFAIK <quiliro>bavier: is there not a log of the operations that were done when invoking 'guix package -i some-package'? <bavier>quiliro: you can diff profile generations, but propagated inputs are not accounted for there, only implicitly through the dependency graph <mbakke>quiliro: you can see the runtime dependencies of a package with `guix gc --references package-name` <quiliro>$ guix gc --references gnome-desktop <quiliro>guix gc: error: build failed: path `gnome-desktop' is not in the Nix store <civodul>quiliro: 'guix gc --references $(guix build gnome-desktop)' or 'guix size gnome-desktop' <civodul>for the 2nd question :-), 'guix package --list-generations' <quiliro>civodul: i did that. but it would not show me what dependencies were installed...it just showed the packages i ordered being installed <quiliro>civodul: i had problems with gnome-calculator...so i installed several packages (including gnome-calculator itself) in order to repair the problem....the problem was solved when i installed gnome-desktop...but it did not return when uninstalling gnome-desktop <quiliro>so i assume it is a package that was installed as dependency to gnome-desktop which solved gnome-calculator's problem <quiliro>'(gnome-calculator:16031): GLib-GIO-ERROR **: Settings schema 'org.gnome.system.proxy' is not installed Trace/breakpoint trap' <quiliro>when running gnome-desktop from the terminal <quiliro>but it is not any more...after removing gnome-desktop <ng0>on GuixSD you don't install gnome-desktop directly <civodul>quiliro: FWIW "guix environment --pure --ad-hoc gnome-calculator -- gnome-calculator" works for me on x86_64 <ng0>you put (gnome-desktop-service) in the (services). the gnome-desktop is just a meta package <ng0>I hope I can fix the borg'ified debian with just booting from GuixSD medium and throwing out files from /etc/ which are no longer required and get in the way of GuixSD boot <quiliro>civodul: for me it works too on i686 <ng0>i think the bournish shell should be documented somewhere more obvious in case you get stuck or something. if it is documented at all besides ,help <ng0>uh... okay. so debian 9 (snapshot from today) creates ext4 partitions with a feature where the e2fsck of guix (recent pull) børks out <ng0> /dev/sda1 has unsupported feature(s): metadata_csum <ng0>e2fsck: Get a newer version of e2fsck! <ng0>archwiki says "To enable support for ext4 metadata checksums on a new file system make sure that you have e2fsprogs 1.43 or newer and simply do a: " <ng0>does someone know if I can disable this? <ng0>hm.. everything is a file. <ng0>would mtab just hold the mount options? <ng0>because we don't have it <ng0>because I will go ahead and package the upgrade as soon as I have figured out wether this can be disabled or not <ng0>some things are just not updated because no one noticed :) <ng0>so according to some Gentoo users this is nothing new but there's a strict check in place or somethin <quiliro>why there is no "fancy" plugin in claws? <catonano>quiliro: there's no "fancy" plugin for claws because no one packaged it ;-) <quiliro>why were other plugins packaged and not that one? is it somewhere else? <ng0>ACTION updates efsprogs <catonano>quiliro: in thhe claws reciipe I see only 2 pluugins: pgpmime and enchant <ng0>do we have some kind of testsuite for e2fsprogs? <quiliro>but a lot of them are already included in guixsd <lfam>ng0: Does it build for you? I'm looking into a runpath validation failure <catonano>Maybe they belong to the claws code base ? <buenouanq>are we talking about how to get spellcheck working in Claws? cause that would be awesome <catonano>buenouanq: if the spellchecking on claws is based on aspell, there could be issues <catonano>Ludo found a problem in how some env varr is not honored in building packages that use aspell lib <lfam>ng0: It seems that the runpath thing in e2fsprogs's #:make-flags is not honored <ng0>e2fsprogs? I'm still building the checkout <ng0>it has seen no update from out side in over one year <ng0>i think at some point I tried to work on it though <lfam>I tried it in the past too. Let's make it work today :) <ng0>and tune2fs -O ^metadata_csum /dev/sda1 (rootfs) gives me bad magic numbers <lfam>There's a new --enable-hardening flag in e2fsprogs 1.43 that we should enable <ng0>it is correct that I run tune2fs on UNmounted partitions.. right <ng0>I used this command never before <bavier>speaking of upgrades, I've almost completely verified dependent builds for an upgrade of our clang package to 3.9.1 <lfam>Thanks for checking so carefully :) <bavier>lfam: I get less nervous that way <lfam>I do the same thing for a few packages <bavier>especially now that mesa uses llvm <ng0>I think 1.43 is not backwards compatible to 1.42 <ng0>at least e2fsck when the metadata_thing is set I get e2fsck complaining that it is set <lfam>Hm... it would be very suprising if the new e2fsprogs did not support old ext* filesystems <ng0>and I'm triyng to figure out how to remove the option AND get rid of the bad magic number <ng0>I think there is no new -> old issue, just old -> new <ng0>which is what I have here <ng0>lfam: yeah, runpath issue here <bavier>has anyone else noticed segfaults from grep while building packages on/for i686? <ng0>lfam: do you have an idea what could be done to fix the runpath issue? <lfam>ng0: I'm reading the e2fsprogs Makefiles <lfam>bavier: There was discussion of serious problems with i686 recently <lfam>I guess that nobody tested it before the merge :/ <bavier>lfam: ok, I haven't thoroughly read backlogs/guix-devel yet <bavier>yeah, I was busy with other projects while that was going on <ng0>okay, Debian 8 jessie still has 1.42 <ng0>that's what most hosters will give you, and that was my intention for testing anyway <bavier>lfam: thanks for the link; fun bug :P <lfam>Nix has 1.43.4 in their tree; I assume they validate runpaths but I don't know <ng0>they have --enable-relative-symlinks <ng0>it's an option we don't use <lfam>It doesn't make a difference for this problem. <lfam>For some reason, the rpath setting isn't being honored by the libraries <ng0>i wonder how they get away with it ***mo is now known as Guest790
<lfam>I don't know nix well enough to say <ng0>hacky-dihack is sometimes a solution I see in nix packages. like, I approach torproject and ask for trademark use in an not so grave altered torbrowser, Nix just combines two things and elfpatches them :) <ng0>it's a hack and works. just like a CPU is just a rock we teached to think <lfam>I think of it as a beach (silicon = sand), but a rock is cool too <ng0>has someone tried out https://redox-os.org/ ? I'm curious about it, but the documentation lacks the points where I have question how they want to solve certain features. It's interesting, but also full of potential issues <ng0>an OS + microkernel + later package manager build in rust <ng0>everything is crates <ng0>"omg it's full of crates" <amz3`>ng0: how does work the rust package manager? <ng0>and they are also writing an fs (or reimplementing it in rust?) and drivers, etc <ng0>makes you wonder.. would it be logical to take the steps and implement a micro kernel in scheme. okay, we have the hurd coming up, and kernel development depending on how you approach it can be rather slow or fast <ng0>redox even writes a bootloader in rust <ng0>so basically re-implement everything > rust <lfam>Next they'll reimplement Python in Rust <ng0>but honestly, I'm really curious about this system <lfam>e2fsprogs 1.42 uses LDFLAGS when building the shared libraries, whereas 1.43 uses LDFLAGS_SHLIB. So we need to adjust our package definition to account for this <cbaines>When running system tests through Guix/QEMU, does anyone know any tricks to get a shell in the VM to debug things? <ng0>it would make sense if you do the update, because you can immediately push. update looks straight forward, but we should test it <lfam>And the Nixos e2fsprogs 1.43 package's libraries do have the proper runpath <lfam>I have to go away for a while. Maybe I'll finish the work tonight, but hopefully somebody else beats me to it :) <ng0>I'll build the change, but how do we check? <ng0>I just guess lower -> upper has no issues <ng0>i mean the logical way, downwards compability exists, but not upwards (how would you predict the future) <ng0>my issue was just limited to this one option set <ng0>it's not so easy as just setting the LDFLAGS_SHLIB <ng0>but when you use LDFLAGS_SHLIBS you see what's going on and can add the missing things to the path <ng0>I'll do it tomorrow if you haven't done it then, lfam <rekado>wingo: we do have a way to configure nginx in a more descriptive fashion; it’s just not comprehensive enough. <rekado>I tried to use it for my server’s configuration, but had to revert to a plain text configuration file, because our config bindings aren’t expressive enough. <quiliro>This is what claws-mail developer says: "you probably do not have Fancy because your OS does not ship webkitgtk library, which is what Fancy uses to render html" <quiliro>and "all current plugins are included with the main source code" ***ng0_ is now known as ng0
<amz3`>ng0: what's the point of their package manager there is not code <ng0>i don't quiet understand you <ng0>it's not just a package manager <amz3`>ng0: it's like javascript tried that too, but now they are specialized in doing GUI stuff the primary target <ng0>it's an OS in development <ng0>it's not done by rust <amz3`>ng0: yeah but is cargo their package manager? <ng0>redox-os is not rustlang <ng0>it is *written* in rust and developed by someone who is involved in rustlang <ng0>it requires even the nightly of rustc <amz3`>I also rely on nightly biwascheme, that doesn't make more interesintg... <mekeor>hey. i installed guixSD on my hard disk. but my grub-config was wrong in the config.scm. now, i'm in the live usb-image again. how can i re-apply the config without downloading and compiling all the packages again? i just want guix to apply the new grub-config... <amz3`>mekeor: you know how to chroot into your installed machine via the usb image? <ng0>but chroot wasn't really working last time I checked :/ <ng0>you can do some things, but interaction with guix daemon wasn'Ät possible <ng0>amz3`: there are some open documentation points in the docs of redox-os like future, goals, etc which would cover questions I have <amz3`>I know how to do it but, I can't write it down <amz3`>mekeor: basically, you have to mount the sda of you hardisk on mnt, then you mount /proc and /dev into respectively /mnt/proc to /mnt/dev and execute the command 'chroot /mnt /bin/bash' <amz3`>tada, you by passed the bootloder grub <amz3`>as if you booted on your machine <ng0>mekeor: how long did you spend with compiling? you could just save the config.scm somewhere, start again (with a minimal config) and then configure starting from there <mekeor>ng0: but it's a good idea to use a minimal config instead, yeah <ng0>sneek: later tell civodul: licensing 'team' of FSF came back to me, but they just suggested what you did suggest to me and I already resolved :) ***mo is now known as Guest16430
<mekeor>amz3`: i mean, i successfully did chroot. now: you said, i can do "guix system init" again. but couldn't i rather do "guix system reconfigure"? <paroneayea>guix-devel might be the only mailing list around where gmail isn't even showing up as a user-agent <amz3`>mekeor: not sure which guix command is correct <amz3`>mekeor: yes maybe it's only reconfigure <ng0>paroneayea: but there is gmail used <ng0>there are also people who hide their mailing agents or spoof them <ng0>I wonder how often Alpine is used on all of my lists <ng0>I'm running this for all now.. could take a long time <amz3`>I mostly use gmail to keep in touch with old friends... <amz3`>I try to self host, but old people have an hard time to remember the new email <ng0> /etc/mtab should be a symlink like it is in the current runnign GuixSD system, it is not a file which is changed upon system activation, right? <ng0>out of 848.116+ emails, 534 are Alpine ;D <ng0>ah, I found my way to do the /etc <ng0>do as wingo wrote, mv /etc-old/guix back in (ACL!), run guix system reconfigure <ng0>just mtab is questionable <ng0>it originated from a debian 8 system <ng0>so I think the safe way is to create a symlink to what GuixSD has here <ng0>oportunistic reboot, the second :) <ng0>are we Guix yet? i hope <ng0>I have succesfully assimilated Debian 8 into GuixSD :) <ng0>your last tips helped wingo :) <wingo>now we should make them unnecessary :) <ng0>updating my web log post now <ng0>the issue now is, you need to place a pubkey in the $HOME before you reboot <ng0>because after configure / init you have a root with no password and a user with no password set <reepca>civodul: Just finished reading the 2013 paper. Haven't experimented much beyond that, I'm sort of in the "salvage grades" phase of the semester. <ng0>oops, wrong username <ng0>i set the userpassword, but the password thing was still manually (hands on keyboard at the machine).. <ng0>how did you fix that, wingo ? just copying in the ssh pubkey in advance? <ng0>I think you mentioned it in the email, I'm going to read the thread again <sneek>Welcome back civodul, you have 1 message. <sneek>civodul, ng0 says: licensing 'team' of FSF came back to me, but they just suggested what you did suggest to me and I already resolved :) <ng0>Okay. Debian8 -> GuixSD worked. To have the real server context I need to repeat it and not cheat by using the keyboard but deploy my ssh pubkey in the root profile next time <ng0>but it was entertaining and interesting for today