<pkill9>why do you need it to start in order for it to stop? <pkill9>oh you probably want it disabled <apteryx>the services are always started automatically (not configurable?) AFAICT <apteryx>I've read in the shepherd manual that enabled? #f prevents the service from being started at all, so this is not what I want either. <ArneBab>is there a way to get ffmpeg 4.1 (with AV1 suppert)? <ArneBab>… but before that: my Guix system is operational! The basics I need are in place. Thank you! <apteryx>ArneBab: it's already at 4.1 in latest guix <ArneBab>I have a new box with 8 times the speed, so I want to test the compression ratio of a codec with 1000x the encoding time :-) <apteryx>ArneBab: it seems we'd need to build it with dav1d as input <ArneBab>(need is a too strong word here: I’d like to test it is the better wording) <ArneBab>time to sleep — thank you for your help in finally making the jump to Guix! <apteryx>my system is building *a lot*. Is something wrong with the build farms? I'm on Guix from yesterday morning. <apteryx>it's GuixSD -- I should be on by default (and i do get some from ci.guix.info or mirror.hydra.gnu.org or berlin.guixsd.org <apteryx>right now it's building /gnu/store/v9p37plzihh92pvfbpm1jz7npd2r7szf-subversion-1.10.2.drv <Formbi>maybe it's building derivations? <cnmne>to modify configs for packages like st or xmonad (which require recompilation on every edit, even outside of guixsd), is it possible to define the config files short of making a new package definition? <cnmne>for st, there is a config.h header file which you edit; `make clean install' recompiles the program with your new settings <pkill9>instead of the clipboard patch you'd want to use your own that modfies config.h <bavier`>or add a phase that patches the existing config.h <cnmne>i think that's exactly what I'm looking for <cnmne>this would be used with `guix package -f st.scm', and it would replace the previous st? <pkill9>actually that file won't work with that, it needs at the bottom "(package (inherit st-with-patches))" <cnmne>`guix build -f xyz' outputs a package (?). would I want that instead of `guix package -f xyz', which i think installs as well ? <cnmne>anyway, thanks for the snippet! I'm going to tinker and read a lot more before I ask any more questions :] <pkill9>if you want to install then use `guix package` yeah <pkill9>i meant the `-f` flag to load froma file <tune>seems that herbstluftwm is not packaged <bavier`>davexunit: when you created the patch for MINETEST_SUBGAME_PATH, was there any discussion about handling mods similarly? <apteryx>I'm trying to add network-manager-openvpn to my network-manager service but I'm failing it seems <apteryx>it fails on my contains->polkit? predicate <reepca>apteryx: that would make sense, since you're trying to string-append a <package> to a string. ***Sleep_Walker is now known as Noctambulist
<reepca>you could do package->derivation and then look up the derivation-output associated with "out", and the predicate would work, but its results wouldn't be guaranteed to be accurate, since there's no guarantee the output path even exists yet. <reepca>the only way to verify what's in a package's output is to build it, y'know? <apteryx>but it should be an improvement over the network-manager service. The goal is to make the dbus system files of the network-manager *plugins* available. <apteryx>because currently attempting an openvpn connection fails with an error like: Error: Connection activation failed: The VPN service <apteryx> 'org.freedesktop.NetworkManager.openvpn' was not installed. <apteryx>it builds my config, but sadly doesn't improve the situation :-( <apteryx>and I ran out of gas trying to make this work <roelj>Why do I get "guix environment: error: clone: 1006764049: Invalid argument", when the daemon runs as root? Do we need user namespaces AND the daemon running as root? <efraim>kodi@18.0rc1 randomly failed test 439, kodi@18.0rc3 its test 449 <efraim>roelj: I'm pretty sure yes on both <efraim>unless you disable chroots for the daemon <reepca>roelj: do you mean you only get that error when the daemon runs as root, or you get that error "even" when the daemon runs as root? <roelj>reepca: "even" when the daemon runs as root. <reepca>does your system support user namespaces? ***Guest46067 is now known as Sleep_Walker
<reepca>well there's your problem, guix environment unconditionally attempts to call clone with CLONE_NEWUSER. <roelj>Alright, so the client-side calls "clone". <efraim>i'm testing merging master into staging <nly>I have a vague question, how do you go about hacking guix? ***rekado_ is now known as rekado
<rekado>nly: what problem are you trying to overcome? <rekado>when I work on Guix I mostly just package things; for that task “./pre-inst-env” is usually sufficient. <rekado>nly: if I need to update search paths I simply “source $profile/etc/profile”. <nly>i want to update guix-search-paths every time i install somethinng <jlicht>nly: maybe direnv integrations could solve your problem in a different way <jlicht>I know there is some direnv+guix+emacs solution somewhere that allows you to adapt the environment when changing project directories <nly>rekado do i need to pay attention to guix path recommendations when i install something? <nly>the usual output export xyx etc after a transaction <rekado>nly: usually not, because etc/profile contains all of the environment variable definitions for you. <rekado>(I’m actually inclined to suggest that we remove those hints and recommend the use of etc/profile instead) <nly>but you say "usually not" <rekado>if you use a shell other than Bash you may need to update the environment variables manually. *civodul sends patches for a useful --verbosity flag <g_bor>you now have access to the video makefile repo on gitlab. <lsl88>I am having an issue with my Makefile :) <lsl88>it is the only part that when testing everything everything everything does not behave as I would like to, and I don't know why :/ <rekado>lsl88: could you give us a bit more detail? <rekado>hey, any GNOME users here who could test the wip-gnome3.30 branch? <lsl88>rekado: yes! I was waiting for someone to answer :) <rekado>Rene wrote on the mailing list that GNOME 3.30 works fine, but I haven’t been able to make it work on my laptop, so I’m interested if other people can test this. <rekado>Don’t want wip-gnome3.30 to go stale (again). <lsl88>I have different make cleans for each "part : ie: one for deleting videos, one for deleting translation files, and so on. In the one for cleaning the "slides"(the svgs for the given locale sent to the makefile) I check with an ifeq the value of the variable LOCALE_LANG (it was your original LANG) so that it is not en_US, since then my bare .svg files are deleted, and I only want to delete the svgs that are created for any other locale. <lsl88>rekado: i have just figured out that instead of a ifneq I used ifeq <civodul>rekado: agreed, that would be sad; what's wrong when you test on your laptop? <rekado>civodul: GDM starts up and I input my credentials, but after hitting “Sign in” the mouse cursor skips around and then … I’m back in GDM. <lsl88>it was deleting the files because I did not put the n in between <rekado>I can’t find any logs that would explain what happens in the background. <rekado>lsl88: good you figured it out :) <rekado>lsl88: BTW when you write git commit messages, please leave one empty line after the first line. <lsl88>rekado: I am cursing myself for that n <rekado>lsl88: the first line is supposed to be a short summary; the explanations can follow on the third line. The second line should always be empty. <lsl88>yes, I was wondering if I should follow the commit standars we used for the contributions or not. <rekado>lsl88: I’ve made similar mistakes in the past: using “and” when I meant “or”, or using “unless” when I meant “when”, etc <rekado>lsl88: no need to follow the strict format, but leaving an empty second line is pretty much universally accepted for git commit messages. <lsl88>rekado: the ifeq and ifneq is horrible, it is just a typo! I'd rather have only ifeq and then a not, ! or whatever to make false my condition. Thank you, will take into account the commit stuff, did not know that. <jonsger>I'll try to build bootstrap-tarballs for powerpc64le. For that you need gcc >= 6.2. While building libstdc++-6.5.0 it fails. I wonder know which definition of libstdc++ is built here: one from gcc.scm or from commencement.scm? ***dddddd_ is now known as dddddd
<civodul>rekado: that could be a GDM issue not specific to the branch, no? <rekado>it could be, but I would find that surprising given that Timothy made changes to the GDM service and has been using this successfully for some time now. <rekado>besides, even without GDM (using .xsession) I can’t get GNOME to start. <rekado>I tried, but … it’s incredibly slow! <rekado>It takes minutes before I can even try to log in. <rekado>it’s all snappy until GDM starts up. <rekado>yes, it takes a long time before I can even begin to input the user name. <civodul>without GDM it works at a reasonable speed for me (with KVM) <rekado>and then it writes 2 characters per 30 seconds or so. <rekado>I think this might be a graphics acceleration problem. <civodul>but anyway, can you reproduce the issue in the VM? <rekado>well, it’s so frustratingly slow and it switches VTs after a certain amount of time has passed… <rekado>ah, no such problem with -m 1024… <civodul>i think the default run-vm.sh uses 256M of RAM <rekado>ugh, should have tweaked that first before reading about PCI passthrough and whatnot <rekado>hmm, so why doesn’t it work on my laptop? *rekado shakes fist at the state *rekado is now on a watch list <pkill9>when you use a manifest and/or a system configuration with an inferior, does the channel that that inferior is built from get garbage collected? (assuming that the profile/manifest that package is in is pinned like all profile generations are) <rekado>only 558 CRAN packages to add before we can add all of Bioconductor! <civodul>pkill9: no, the channel doesn't get collected <civodul>it's cached in ~/.cache/guix/inferiors even <pkill9>ok, so if i removed ~/.cache/guix/inferiors it would garbage collect the inferiors? <civodul>push at 15:44 and eval #2613 completed at 16:03 <civodul>pkill9: they could be gc'd (if you run "guix gc"), so next time you'd effectively rebuild them <pkill9>oh yeah i mean the next time i run `guix gc` <divansantana>civodul: seems not so. Or I'm going crazy. Only .guix-profile/bin/go .guix-profile/bin/gofmt <Formbi>I'm having errors like this all the time <divansantana>perhaps not packaged. I used go get -v golang.org/x/tools/cmd/godoc for now. <jonsger>Formbi: are you on a foreign distro or on GuixSD? <Formbi>(i replaced the AUR guix-daemon with the one from Guix) <jonsger>do you use guile-2.2? like /usr/share/guile/site/2.2/... ? <Formbi>it is there, but I don't think it's in use <jonsger>is there a usr/share/guile/site/2.0 folder? <jonsger>oke, then I don't know. I'm not a guile expert <rekado>Formbi: you may need to use a different variant of Guix to run “guix pull” again. <civodul>Formbi: which version of Guile does the AUR package use? <civodul>i'd recommend running 'guix pull' and then using the ~/.config/guix/current/bin/guix <Formbi>I removed the Guix I got from AUR <civodul>yeah it's easier for us if you install Guix from the binary tarball we provide <jonsger>civodul: is there are a reason why guix doesn't use libgit2 0.27.7? It's still using 0.26.8 <apteryx>has anyone bothered making their trackpad click when using GuixSD ? <apteryx>I seem to recall it'd be something to do with the synaptic driver in a Xorg config or such. <pkill9>what do you mean making the trackpad click, apteryx? <civodul>jonsger: no particular reason, i guess we should just upgrade <civodul>jonsger: the upgrade seems to work so i'll probably push it if it passes on ARM as well <asterope>I thought about writing some service now, to learn more about guix and guile <asterope>a compton service would be good? With some configuration capability... <asterope>I don't really know yet what services can do, and if they can do per-user configuration <efraim>civodul: there's more to a libgit2 update than just libgit2 <efraim>after git-grep'ing the source it looks like guile-git and python-pygit2 <efraim>2609, my master->staging merge, didn't get evaluated on berlin <efraim>... i take it back, python[2]-pygit2 isn't used by anything <apteryx>pkill9: re the trackpad, I meant click by touching on the touch surface rather than using the physical switches <pkill9>i've set some settings for my touchpad <pkill9>i set it using xorg configurations yeah <pkill9>also i set it to use libinput which is more updated i think <davidl>how can I get the source code of a build on hydra using the cached source code instead of downloading from the package-defined urls? <apteryx>pkill9: OK! could you share an example of such config? <pkill9>possible options are in the libinput manpage <rekado>the crash came from a stale ~/.local/share/gnome-shell/ <bgardner>Hey all, I have a VM install of 0.16 (my first) that I'm stepping through the instructions on, and it is hung(?) on "building /gnu/store/x8f<bunch of digits>910n-mutter-3.24.4.drv..." <bgardner>Except not really hung as the little spinner is still spinning. Pinned at 100% CPU but no movement for hours. <apteryx>pkill9: this looks like it! ClickMethod clickfinger :-) <lfam>jahb: I haven't seen that, and I'm using VLC from Guix on Debian <rekado>jahb: can you try a different rendering backend? <lfam>bgardner: How did you start the VM? <lfam>bgardner: It's possible that it's working but very slowly due to a lack of RAM <bgardner>lfam: I allocated 2G to it, with a rootfs of 5G <lfam>Okay, I think that should be enough <jahb>lfam: that's good to know, thank you. <jahb>rekado: will try that, thank you. <lfam>bgardner: It will probably be a pain now but I would make strace available (`guix environment --ad-hoc strace`) and attach to the guix-daemon process to see what it's doing <lfam>Something like this (untested): `strace -f -p $(pidof guix-daemon | head -n1)` <lfam>bgardner: It's a QEMU VM? With KVM? <bgardner>lfam: It is, exactly that - on a Core i7 that I confirmed ahead of time had extensions <lfam>All in one: `$(guix environment --ad-hoc strace)/bin/strace -f -p $(pidof guix-daemon | head -n1)` <lfam>bgardner: The other question is ... why is it building mutter? I would expect a substitute to be available if you had authorized the substitute signing key for berlin.guixsd.org <bgardner>So break this 'guix system init' call or open a new tty? <lfam>I would do it in another TTY in case it's almost done :) <bgardner>lfam: No idea, like I said, my first time with guix and following the installation steps to boot the install in a VM <lfam>But, I think it's stuck or something, because it shouldn't take hours to build mutter on an i7 VM with KVM <bgardner>lfam: Hmm, any advice on how to get a new tty to do that? I'm trying to migrate away from Virtualbox so I'm starting new and not familiary with KVM <bgardner>lfam: Ah-ha, Ctrl-Alt-F3 does it, stand by while I try <jahb>rekado: it works fine with XVideo (XCB) and X11 (XCB) output modes. <jahb>lfam: what's your video output mode set to? does it work ok in OpenGL mode? <bgardner>lfam: I think it's dead, Jim. Other consoles are all accessible, but won't respond to keystrokes. Info docs on tty2 also locked and cannot be navigated. <lfam>jahb: I'm not sure :) How do I find out? <bgardner>lfam: Anything useful I can get for you here before I wipe the partition and restart? <lfam>bgardner: No, I don't know enough to go much deeper. If it was my thing I would probably try to spend some time figuring out if the VM is really stuck or just "paused" or something, but it would probably be faster to start over. Sorry <jahb>lfam: In preferences, video tab, display section, output drop down. <bgardner>lfam: No worries at all, just trying to contribute any info I can - I'll report back later :) <lfam>jahb: "Automatic". Any way to figure out what it picked? <jahb>lfam: 'fraid I don't know. Can you try it in OpenGL mode though? <lfam>jahb: Works fine in that mode <apteryx>hello; would someone know of a CVE scanner? Does such a tool exist freely? I'd be interested. <jahb>lfam: aha! thank you, it must be something unique-ish to my machine then. <lfam>apteryx: `guix lint` checks package versions against the CVE database and tries to report vulnerable packages <lfam>metasploit (and to some degree nmap) can probe for the actual vulnerabilities <apteryx>lfam: this is nice, but I was talking more about some source code scanner that finds pattern and matches them to CVEs. <lfam>I haven't heard of a tool like that <lfam>I don't think the CVE database includes code samples <apteryx>yeah, I don't think either. It's just version strings <apteryx>so someone would have to go and build that version -> source snippet map <apteryx>they do it for viruses though, so it should be possible ;-) <apteryx>(except instead of sources the signatures are blobs) <lfam>I think you might run into the same problem as the virus scanners which is that it doesn't really work :/ <civodul>efraim: i built guile-git on all 4 arches after upgrading libgit2 and it went well <efraim>i updated python[2]-pygit2, but only the python2 version built <civodul>(i checked guile-git because obviously it's a dependency of Guix) <bgardner>lfam: Restart going better - I didn't provide swap first time around so this time I added a token 200MiB swap and it is zipping along. <apteryx>ouch; I've caused by system to loose sudo because of an error in the setuid-programs service <apteryx>it crashed in the middle of populating it because I used (file-append mtr "/bin/mtr") instead of (file-append mtr "/sbin/mtr"), and now I only have one binary which has setuid (no longer have sudo there). <apteryx>maybe we should protect against this? <apteryx>time to make use of my btrfs journaling <civodul>apteryx: could you report the issue to bug-guix, explaining how you ended up "losing sudo"? <civodul>oh that's because the thing is copied at activation time, right? <lfam>Challenge to get root with that one setuid binary ;) <Formbi>my bash completions from Guix disappeared <civodul>how did you get them in the first place? :-) <civodul>you need to source the file again, i gues *rekado attempts to merge staging into core-updates… <jackhill>Hi, it looks like my experiments with GuixSD have been noticed at work, and I've been asked to produced a list of installed packages and versions. <jlicht>jackhill: you make it sound like you are in trouble <jackhill>(for "security" auditing). I don't think anything like this exists, but I guess the garbage collectors live-paths would ba good place to start <jackhill>the problem with that is it is just store paths, but I also want the package versions and to filter out things that aren't packages. <jackhill>I'm happy to use the scheme API (that why I'm using Guix afterall), but could use some help identifying how I might proceed. <civodul>rekado: we should merge staging in master rather than core-updates no? <civodul>i think that was the goal but we all dropped the ball <rekado>right, but later master would be merged into core-updates, no? <rekado>I wanted to merge wip-gnome3.30, but it causes so many rebuilds that it should rather go to core-updates. <civodul>i though wip-gnome3.30 was also targetting master :-) <civodul>so gdm works for me on master, but i cannot choose a non-GNOME session <rekado>it was, but I forgot about glib :( <apteryx>civodul: I see you were the one who added the vpn-plugins to the network-manager-service; do you recall if you had successfully created/connected a VPN connection with this? <rekado>I can try to build GNOME again, but this time without the glib upgrade. <rekado>then it could go either to staging or even master. <civodul>apteryx: i think it never quite worked but i never figured out why <civodul>i remember discussing it with ng0 and you <ng0>what where how? I just connected after weeks <civodul>apteryx: i saw your message on help-guix and it looks like you're making good progress :-) <civodul>we were discussing the openvpn plugin for NetworkManager <apteryx>civodul: yeah I thought my suspicion might bear fruit, but it didn't :-/ <apteryx>(suspicion being: the plugins themselves must extend the dbus service with their own dbus definitions to be usable) <ng0>I didn't use the vpn plugin. when I was working on this I tried to connect to the other plugin I needed for the university vpn <ng0>s/connect/package but didn't work/ <ng0>which is probably the case here? <apteryx>I see; they (your university) were not using openvpn but another package. <ng0>i didn't try, I succeeded. it's just broken wrt system integration <apteryx>has anyone debugged dbus errors before? Anything I should look at? strace is not very useful <ng0>i mean if you want another usecase, I can zip it up <apteryx>ng0: ah, OK. So you managed to package the vpn tool, but didn't succeed in integrating it with the network-manager service? So you ended up using that tool directly? <ng0>I'm in the middle of weeks of backups >.< no work accessible <ng0>I think you are right with dbus. other systems hinted at that <apteryx>yeah the error message is strictly about the dbus definition of the network manager pluging not being *installed* (in other words, it's not found). <apteryx>so I extended the dbus-system-service as part of our network-manager service hoping it would fix this error, but it didn't <civodul>apteryx: which .service file is not found again? <civodul>it's that one that you should add as an extension of dbus-root-service-type, i think <apteryx>org.freedesktop.NetworkManager.openvpn; it comes from network-manager-openvpn <apteryx>you can see what I try to make it discoverable in the bug 33832 <apteryx>basically instead of just passing the network-manager package itself to our dbus-root-service-type, I give it the list of network-manager <apteryx> + its plugins: `(,network-manager ,@vpn-plugins) <serichsen>When I have created a config.scm which after a seemingly successful `guix system init …' turns out to be unbootable, do I have to delete everything before trying again, or can I do something like `guix reconfigure …'? <apteryx>serichsen: you should be able to guix reconfigure! <jlicht>what was the super url for the fancy web interface to the ML again? <serichsen>apteryx: but it seems that reconfigure cannot take the additional argument to specify the target directory, like init does. <apteryx>serichsen: it should be safe to rerun guix system init otherwise. <apteryx>hopefully the items it had already cached under /mnt will still be there and it will be speedy <civodul>apteryx: right, passing the list of nm + plugins looks good... then again you could strace dbus-daemon to see where it's looking for that .service file <apteryx>civodul: good point! I didn't know what to strace. Thanks. <civodul>it's a sledgehammer but it often works well in situations like this :-) <apteryx>sneek: later tell civodul with my patch, I can see that the dbus-configuration used by my system generation contains a system-local.conf file which has a new entry: <includedir>/gnu/store/gw3ckmw2pihc44d23lc8pipfw7wr16g7-network-manager-openvpn-1.8.0/etc/dbus-1/system.d</includedir> <apteryx>so it should have the necessary dbus file