<mbakke>ooh, ci.guix.gnu.org is responsive again <mbakke>Sharlatan: it was struggling to load the web interface for a while, but now it's very snappy :) <Sharlatan>I've noticed that bag tracker is neraly uslese when I try to search anything, it's just hangs. <mbakke>Sharlatan: are you talking about issues.guix.gnu.org ? <buhman>I'm wanting to use xorg-xinit, but I get unusual permission issues for /dev/tty0, etc.. what am I missing from my system definition? <buhman>I have (elogind-service), but I think my session is still broken somehow <kori>buhman: you cant use xorg-xinit basically <kori>because X is installed setuid <kori>you *have* to use something like slim or lightdm or gdm <kori>that's going to be changed <buhman>hmm, that's unfortunate; I rather like login(1) compared to the other DMs ***modula is now known as defaultxr
<rekado_>mbakke: I didn’t change anything on ci.guix.gnu.org :-/ <rekado_>I have a patch here to add a “--web” option to Cuirass. When it’s passed the web interface (and nothing else) is started. <rekado_>a second patch creates a separate Shepherd service “cuirass-web”. <nly>(native-inputs '(("gcc-toolchain"))) no ,gcc-toolchain works now? <rekado_>that thing after the “comma” is the important part <rekado_>what follows is the *value* of the “gcc-toolchain” variable. <rekado_>the string is merely a label; could be anything. <nly>thanks, that makes sense <nly>i see that gnu-build-system must be the one adding gcc to guix env <nly>and my typo without ,gcc-toolchain "worked" in that way <rekado_>nly: yes, it’s not going to give you an error, but it’s not respected in any way. <rekado_>it’s just some value that happens to be on that list. <rekado_>nly: the gnu-build-system calls the GCC just “gcc”. You don’t need to pass gcc-toolchain. Just “gcc” will be fine. <efraim>I got a personal bug report about our rakudo/perl6-zef packages, looks like we might have someone who knows something about perl6 finally <eiro>efraim: we're just talking about this but right now in #perl6 <efraim>eiro: so you sent the report? I have a nice long email so far with my notes <eiro>i hope it was ok to fill the bug directly on you (and guess there is a better way to do it) <eiro>#perl6 people want to show the scm file. can you point it ? ( i don't know better than guix edit) <eiro>ohh and chloekek comes from #perl6 <efraim>I think I uncovered another bug in guix at the same time, but it looks like the short version is zef in guix doesn't actually work <efraim>so I got an interesting case where 'guix package -f perl6-package.scm -i rakudo', perl6 (from rakudo) fails to know about the perl6-package, but insert it into the guix source and 'guix package -i perl6-package rakudo' works without problems <rekado_>g_bor[m]: hey, I wonder if you had some success with writing Guile bindings for that networking library. <rekado_>g_bor[m]: I connected berlin to two 10G ports … and right now I can only use one of them, because we can’t do link aggregation or other forms of bonding. <rekado_>I suppose we could run some network manager commands, but doing the network configuration without that intermediary would be sweet. <numerobis>Hi #guix! I'm thinking of getting guixsd on a VPS. Is 2GB enough? <roptat>I have a 512MB VPS running with Guix, so 2GB should be enough <roptat>(although I need to stop some services on each update, because updates need that much RAM ^^) <xavierm02>Does anyone know a fast-to-compile package in Guix that uses Qt and a wrapper to set QT_PLUGIN_PATH? <numerobis>roptat: thank you! Yeah I remember some updates not working some time ago because of a lack of RAM. <roptat>numerobis, as long as you update to a commit that was built on berlin, you don't need as much RAM as before <roptat>because updates can now be substituted <roptat>(but you still need a bit of RAM to compute derivations, so you know what to substitute) <roptat>before, I used to have a local git checkout on another machine that I sent over to the VPS <roptat>but I don't need to do that anymore :) <numerobis>roptat: Oh, that's interesting. Is there a way to update only the packages for which substitutes are available? Or is it just that the Berlin servers are generally more up to date than the other ones? <rekado_>numerobis: there is no simple way to do that, but all the pieces that are needed already exist. <roptat>you can see what commits have been built by berlin, and using guix pull --commit=..., guix pull will use substitutes <roptat>but then, you don't know if you're going to need to build stuff <roptat>from my experience, using the latest commit that has 4 successes out of for in guix-modular-master will work <xavierm02>I'm going to try to replace the wrapper setting QT_PLUGIN_PATH by some patch to qt.conf. <numerobis>rekado,roptat: thanks, that's very useful information. I think I'll go for 2GB for the first month and see how it goes. <roptat>I mean, berlin would have been quick enough to have built server stuff <roptat>if it hadn't, then your updates will be longer, but 2GB will probably be enough <tune>ungoogled-chromium-wayland seems to crash for me when I right click in the address bar <tune>also something about copy-pasting between chromium and emacs is super broken and annoying <tune>hm okay this type I just did ctrl-v and it crashed. I'm trying to install ublock origin using a url template from the ungoogled-chromium faq <desttinghim>Is there a way to tell guix to fail if a substitute is not available? <rekado_>desttinghim: you can check the build plan with “--dry-run” and then decide whether to move on or not. ***Server sets mode: +cnt
<brutelumpen>is there way to clone guixsd from virtualbox to usb mbr hdd? <brutelumpen>now im try to install on it from vm booted from install img <rekado_>hmm, I can no longer reach arbitrary ports on bayfront <pkill9>why does a basic guix installation require subsititutes to download? shouldn't they come in the ISO? <roptat>no, because the ISO is just a base system: its store only contains items that constitute the system <roptat>also the guix running inside the ISO is different from the guix that was used to create the ISO iiuc, so some of these items will be different in the installed system <pkill9>hmm, i want to create an ISO that can be used to install Guix System without an internet connection <roelj>What do I need to get Guix to be able to verify a certificate when doing 'guix import cran ...'? <rekado_>roelj: probably just nss-certs and the variables as specified in Application Setup <roelj>rekado_: Ah, I was missing the environment variables. Thanks! <rekado_>I’d like to offer rsync access to the nar/narinfo files on berlin. My plan is to create an unprivileged user account for SSH login, but restrict that account to just the directory containing the files. <rekado_>do you know how a user account could be jailed? <roelj>rekado_: It's possible with openssh <rekado_>just configuring it in the sshd_config? <rekado_>I wonder if that’s enough, because I remember that chroot isn’t designed for security <rekado_>perhaps I could slap on some unshare command…? <roelj>rekado_: Yes, in sshd_config. I use: "SubSystem sftp iternal-sftp" and "Match Group sftp_users ... ChrootDirectory %h" <roelj>rekado_: If you don't give users a shell, it should be fine. <roelj>rekado_: But if you do, then I don't know what a secure way is. Sorry. <rekado_>I want to let them run rsync; not interactively, of course, but perhaps that requires a shell…? <rekado_>thanks for your input! I’ll play with this a little. <pkill9>does `guix deploy` handle running `guix pull` on the remote machine? <pkill9>so generally you would run `guix deploy` on a remote machine, to handle other remote machines? <rekado_>you run it on one machine to handle other machines. <rekado_>the machine where you run it doesn’t have to be remote. <pkill9>wouldn't it be desirable to have each machine generate the system for itself, rather than generating it locally and then sending it to each machine? <rekado_>in my case that wouldn’t be desirable; but I think that “guix deploy” also supports that. <rekado_>for ci.guix.gnu.org, for example, it wouldn’t make sense. We have dozens of identical build nodes, none of which accept substitutes. <rekado_>so building the system once (by offloading the work to all of the build nodes) and then pushing it to each of the nodes is much better than building the system without substitutes on each of the nodes repeatedly. <efraim>i just broke master, searching the previous commits for a stray "name" tag <rekado_>I see that when compiling bioinformatics. <rekado_>commit d84e9b75b2aba98583531f2876d9298871389585 looks fine <rekado_>the commit message also isn’t correct :-/ <efraim>i'm going to revert the last two and fix them in a bit <str1ngs>sneek: later tell nly. yes to reproduce the abort to unknown prompt use message. <xavierm02>I couldn't find any tutorial showing qt.conf accepting several paths. If it doesn't, would it make sense to use a package union of all the qt plugins, so that the path to this union can be added to the qt.conf? <xavierm02>And more generally, when is it fine to use union packages? <str1ngs>xavierm02: qt.conf and well qt in general work better when referring to a single prefix I found. for example locations of locales etc. <str1ngs>xavierm02: in short I think a union would help the Qt packages. if not then some work on the current qt.conf inheritance would help <str1ngs>xavierm02: or maybe a qt-build-system with help with qt.conf creation for modular qt packages. <davexunit>is it just ld-wrapper that hardcodes runpaths? I tried removing it from the build container as an experiment and the shared lib I built seems to still have a hardcoded runpath. <apteryx>do gexp obscure guile backtraces even more? <apteryx>also couldn't figure out how to make that backtrace extend wider than 80 chars, which didn't help <pkill9>rekado_: how do you use `guix deploy` in the way i described? <apteryx>I see lots of 'make check' failures on master. Is it just me? <rekado_>apteryx: you need to set COLUMNS in the daemon’s environment. <rekado_>apteryx: it’s not just you, but the error is not with Guix <rekado_>apteryx: we switched to guile-json@3, but when you do “guix environment guix” you probably still get a Guix with guile-json@1 <rekado_>appending “--ad-hoc guile-json” for now should fix it. <rekado_>the error with gexps is due to the fact that you’re trying to use a plain value where a monadic value is required. <rekado_>I think that it might be better for us to have a type for monadic values instead of using plain procedures. <apteryx>rekado_: about Guile backtrace... ah! The backtrace is actually printed thrown by the daemon... I didn't think about that. <rekado_>monadic values *are* procedures, but it would help us throw errors when someone passes something that’s not a monadic value. <apteryx>rekado_: I think you misunderstood the error; it was trying to eval a string like ("ext4" fs) due to a variable name clash. <rekado_>davexunit: have you tried setting GUIX_LD_WRAPPER_DISABLE_RPATH? <rekado_>apteryx: yeah, I stopped reading at run-with-store… sorry <rekado_>trying to apply something that’s in procedure position is a very common mistake, so I whipped out my default response. <apteryx>rekado_: eh, that's pretty much where the error becomes unreadable, so I can understand ;-) <davexunit>rekado_: no I didn't know about it. I will try it out! <davexunit>if it can save me from building a custom gcc that is great <apteryx>rekado_: the "In unknown file:" without any line information was the problematic part of it, mostly, followed by some cryptic expansion like: (_ #(#(#(#<directory (gnu system vm) 3c22820>) "disk-image" "x86_…" …) …)) <rekado_>there’s a discussion on the guile list about whether to change the default truncation <rekado_>I think Guile shouldn’t truncate anything by default. <apteryx>there's a bug about this on bug-guile <apteryx>where someone maintaining guile-cv has a much different opinion <apteryx>and as noted by Mrak this opinion seems to be biased by Emacs terrible performance with long lines <apteryx>but yes it should be easily configurable <rekado_>yeah, this Emacs problem still bites me more often than I’m comfortable with. <rekado_>but Emacs has since gained a blessed workaround for long lines. <apteryx>I'm killing the REPL continuously sometimes to keep it at managable speed <rekado_>it’s a hack that disables features known to impact performance. <rekado_>I only read about it on emacs-devel. <apteryx>OK. I'll look it up tomorrow. Maybe try it and report back. <minall>How can I unrar? I don't see the package on guix... <str1ngs>xavierm02: I'm not sure how much that help. but I guess if I were to give feedback. my suggetion for QT package improvement is to maybe start by creating a qt-build-system. <davexunit>rekado_: hmm I set that variable in the build container but the resulting libraries still have hardcoded runpaths <str1ngs>xavierm02: also from my example you may not want to install qt.conf. though for this package it was a hack to get qtwebengine to find locales. which is related to the union issue you describe earlier <rekado_>davexunit: does the package itself add -rpath or some other linker flags? <rekado_>the wrapper only makes sure that libraries end up on the rpath, but if the linker is passed these flags by the build system anyway it will not make any differenc. <minall>How can I refer to a linux kernel in my reconfigure.scm? I have to downgrade since the newer kernel freezes my pc... I'm tring to refer to it as (kernel linux-libre-5.2.6), but no luck <davexunit>rekado_: I don't think they do. I'm working with several packages here. but I see one other big issue: things like libc, libm, etc. are also hardcoded <rekado_>minall: is the variable “linux-libre-5.2.6” defined anywhere? <minall>linux-libre-5.2 or linux-libre-4.9 seems to work, so probably is the last term <davexunit>I need to disable all this runpath stuff entirely. <rekado_>minall: the field accepts a variable name. A variable of that name must exist. <davexunit>probably requires a special build of gcc, which I'm happy to do, but that code is... not easy to work with. <rekado_>davexunit: perhaps you could do the opposite that ld-wrapper does and actively scrub -rpath <minall>rekado_: What do you mean? well I have (gnu packages linux) <minall>Or should I add it to use-modes? <rekado_>minall: what variables does (gnu packages linux) export? <minall>rekado_: Not sure... where can I confirm that? <minall>Probably not linux-5.2.6... since is not assigned <rekado_>minall: I suggest opening the file gnu/packages/linux.scm to take a look. <minall>I'll do, but I'll reconfigure with linux-4.9, since I only need to downgrade, it's ok <minall>Yes but, In that case it will give me the latest, which is 5.2.7, the one that gives me the error <minall>So I have to downgrade on any linux kernel or preferably to linux-libre-5.2.6, which works <rekado_>minall: you can define a new variable for 5.2.6 <davexunit>rekado_: I don't think a special ld-wrapper would work. I'm looking over some make output and there are no -rpath flags. <davexunit>or rather things like libc libgcc are implicit so I think I need a special build of gcc <davexunit>oh wait, turning on debug info in ld-wrapper has revealed some rpaths <davexunit>well, I guess it's fine for libc and co? I just ran ldd on a random ubuntu library and I see similar stuff for libc <tmssgn>I get "guix package: error: corrupt input while restoring archive from socket" anyone seen this recently? I did a pull two or three days ago. <rekado_>xavierm02: it’s called %build-inputs <rekado_>tmssgn: could this be a network error? Sounds like a corrupt download. <tmssgn>possibly, yes. I'll just retry then <tmssgn>:) Thanks rekado_, this happened during my build tonight and now again so I thought it was something else. But now it's working away no problem. :) <rekado_>tmssgn: it’s not a great error message. Sorry about that. <rekado_>a program’s test suite often exercises many parts of the built tool, so we can be much more confident that our build is usable. <rekado_>without tests you really don’t know if the program or library will work at all. <rekado_>it makes sense to run the tests especially because Guix installs packages each into their own prefix. <rekado_>that’s something that might trip up less flexible software, so having some way to automatically check the build is very useful. <rekado_>not all test suites are good, of course. <rekado_>some are really expensive to run, others don’t really tell you much at all. <xavierm02>If I have already built a package, and now it's in my profile. Can I access the result of the build? <rekado_>in those cases it’s fine to disable them. <rekado_>a profile is just a link to a link to … a build. <rekado_>in a build phase you can access the built binaries — after the 'install phase. <xavierm02>yes, but I want to see all the intermediate files <rekado_>you cannot access anything that’s outside of the output directories of the current package, though. <rekado_>you can do that at any point in the build. <rekado_>just inject a build phase where you want it. <rekado_>if you want to do this manually, you can add a build phase that fails and then build the package with “build -K” <rekado_>this will keep the build directory in /tmp/guix-build*, so that you can inspect it. <rekado_>(arguments '(#:phases (modify-phases %standard-phases (add-after 'build 'fail (lambda _ (error "bleh")))))) <minall>What package should I install on guix in order to support unrar? <xavierm02>minall: There is no such free package that works for recent versions of rar <minall>What can I do to extract a rar package ? <pkill9>rekado_: how could i use `guix deploy` such that the system is generated on the remote machine? <rekado_>pkill9: I don’t know. “guix deploy” is currently under development, so it’s gaining features regularly. Don’t know if building on the target has already been added (or even if it’s planned at all). <lispmacs>if I wanted to examine the bootstrap-binaries (like, to dissassemble them) how can I download them or figure out which binaries belong to that package? I tried guix build --source but that did not work of course <lispmacs>or, maybe the question should be, where is the package definition stored? <str1ngs>bootstrap-binaries is kinda subject. but maybe you mean gnu/packages/boostrap <str1ngs>which has some helper boostrap binaries <str1ngs>lispmacs: in which case they are in the guix source tree already. <lispmacs>str1ngs: where is pulled version of guix source tree on my system? <bgardner>Good morning guix, I'm trying to set up authentication against an LDAP server. I have a running server and a client with the nss-pam-ldapd package and an (nslcd-service-type) entry, but I know I need pam entries still - can anyone point me at an example configuration for that? <str1ngs>for this it would be best clone the guix repo <str1ngs>lispmacs though all of this is dependent on if these are the bootstrap binaries you are referring to <rekado_>bgardner: I’ve got a configuration which … almost works. <Adamclisi>Hi. I'm trying to install guix on my laptop but the installation stops with an error when installing the bootloader. <bgardner>rekado_: I prototyped this out on a Debian box (so I could steal the config from there) but it includes a lot of Debian-specific plumbing that isn't present on guix, nor do we want it. Trying to do it the 'right way'. :P <rekado_>bgardner: the nslcd-service-type provides the required PAM services; it extends the pam-root-service-type <bandrami>Is there a package for the moby/docker CLI front-end? My guix package -s fu is failing me here. <rekado_>bgardner: it ensures that for all pam entries “pam_ldap.so” is loaded and marked as “sufficient”. <rekado_>bgardner: you will need to extend the name-service-switch field yourself, though, because that’s no system service. <Adamclisi>The error says Could not prepare Boot variable: No such file or directory... help please... <rekado_>bgardner: unfortunately, though, I haven’t been able to make this work; using “id”, for example, won’t look up the account in LDAP. Only if I set LD_LIBRARY_PATH to point to the pam_ldap.so location. <rekado_>so I still can’t log on with LDAP accounts. <rekado_>looks like something’s wrong, but I haven’t been able to find it. <bgardner>rekado_: Oh boy, I might be too dumb to help, but I'll report if I find anything <rekado_>I’m dumb, too, and I got that far. If we’re dumb together maybe we can crack it. <bgardner>rekado_: First I need to factor out mdns, it conflicts with my network. Give me a bit and I'll get back to you <rekado_>php fails to build on the “master” branch. <mbakke>Adamclisi: that error message sounds like you are using the EFI GRUB on a non-EFI system <Adamclisi>If that's the problem, what can I do? The installation script made the configuration. As far as I know my laptop supports EFI <dftxbs3e>I'm getting: "/gnu/store/vx06vchdp2lwlfd1k9rhxzpcph4km1ph-binutils-cross-boot0-2.31.1/bin/x86_64-guix-linux-gnu-ld: cannot find crt1.o: No such file or directory" on master when changing default gcc with gcc-6, any idea what that is about? <dftxbs3e>I am running the command: guix build -M 2 --target=powerpc64le-linux-gnu bootstrap-tarballs --keep-failed <dftxbs3e>and I found that error inside config.log for libmpx <rekado_>dftxbs3e: this sounds like the error that we had to work around on the core-updates branch. <dftxbs3e>hmm, should I use core-updates instead to change default version to 6? <mbakke>Adamclisi: does /sys/firmware/efi exist in the live USB environment? <bgardner>rekado_: Okay, got that in and reconfigured so now my nsswitch.conf is correct but still nothing in /etc/pam.d referring to ldap - what do you have for your nslcd-service-type config? <rekado_>that’s because I need to talk to ActiveDirectory <nixo_>hi, do you know why cross-compilation to armhf does not work? I get: while setting up the build environment: a `armhf-linux' is required to build `/gnu/store/f21dmq9kdxd097xq1vx6w7m90i3qrry4-arpack-ng-3.3.0.drv', but I am a `x86_64-linux' <nixo_>or maybe I should pass --target instead of --system... <nixo_>Yeah it seems to be better now <nixo_>binutils armhf-linux is broken :( <dongcarl>If I define a procedure in the same file as my package definition, why can't I use it in a phase? <davexunit>dongcarl: that procedure is client-side code, not build-side code. <dongcarl>davexunit: How might I define a procedure that can be used inside a phase? <davexunit>dongcarl: the simplest thing I can think of is to just move that procedure inside the phase itself <davexunit>there's ways to include entire scheme modules in derivations, but that seems overkill for a single procedure. <dongcarl>"definition in expression context, where definitions are not allowed" <davexunit>you have some invalid syntax, it sounds like. <dongcarl>davexunit: So I can't `define` in a phase, I can only do lambdas? <davexunit>it's a scheme syntax rule, not a guix thing. you can only define in certain places. I think you've just misplaced your code <davexunit>I can't see your code, but I assume you are using something like (modify-phases %standard-phases (add-phase 'foo (lambda* (#:key blahblah #:allow-other-keys) ...))) ?? <davexunit>in which case you should define the procedure in the body of the lambda <davexunit>defines are only allowed at the beginning of a lambda <davexunit>like this: (lambda () (define (foo x) (+ x 3)) (foo 1) (foo 2) (foo 3)) <tmssgn>does anyone have cgroups running on guixsd? <tmssgn>hm, guix weather says 100.0% substitutes available for webkitgtk@2.20 but guix package -i webkitgtk@2.20 still starts building from source... anyone know what's happening there? <tmssgn>it seems I hadn't authorized the new substitute servers. guix is now getting substitutes as expected from berlin... <raingloom>hiya! is there an example for unobtrusively modifying one's elogind config? i just wanna set handle-lid-switch to ignore and be done with it >_> <raingloom>and idk enough scheme yet to do it quickly on my own <Tirifto>I wonder: has anyone tried using Firejail with Guix packages on a non-Guix distro? <quiliro>minall: Saluton. Mi fartas bone. Ĉu vi finis vian celojn? <minall>If I finished, celojn? what is celojn? <minall>quiliro: I wanted to add some stuff, but it looks pretty neat! <minall>quiliro: unbounded variable, did you forget use-modules form? <raingloom>also, how do I use geiser in emacs for both a local checkout of Guix and the one I'm running <minall>Did you guys have tested emacs-exwm? <quiliro>Mi estas lernanto de Guix-o, Guile-o, Skimo kaj Esperanto. :-) <quiliro>Mi ne povas uzi Apertium-on en Guix-o. <Tirifto>The Guix installation script tells me to fetch a public key, but doing that has no effect, apparently. :/ <quiliro>Tirifto: Mi ne konas Firejail-on. Kio ĝi estas? <Tirifto>Ah, of course, had to sudo the gpg command. <Tirifto>quiliro: Firejail estas programo por sekurigi aplikaĵojn en sistemo. <Tirifto>quiliro: Pardonu, mi ne agordis chapelojn bone, mi tajpos komforte kun hoj. :P <Tirifto>quiliro: Oni povas ruli programon per firejail, kaj tiel limigi la permesojn de la programo. <Tirifto>quiliro: Ekzemple, per Firejail vi povas limigi la permesojn de via reta foliumilo tiel, ke ghi ne povos vidi dosierojn en via hejma dosierujo – nur la dosierujon por elshutoj. <raingloom>anyone knows how elogind-configuration works? i'm trying it in a repl and i can't evenfigure out how to pass arguments to it <raingloom>(elogind-configuration (handle-lid-switch 'ignore))