<vincelevi>Hello, I'm getting "Too many heap sections: Increase MAXHINCR or MAX_HEAP_SECTS" trying to build packages (from a local git checkout) since a few days, despite make {,dist}clean, anyone got that too ? <roptat>Tests run: 1, Failures: 0, Errors: 0, Skipped: 0 <roptat>it's a bit of a mess, but I was able to build a hello world package, and test "assert(true)" :D <jonsger>public class congratulations { roptat } :P <leoprikler>By the way, do we have a solution for GI_TYPELIB_PATH-related problems? To recap, those occur when trying to refer to some typelib from an interpreted language (e.g. Python, GJS, Guile) when the library is not installed in the profile. <leoprikler>wrapping the path is a solution, but IIRC it's a rather sparsely used pattern <leoprikler>perhaps something that we could do as part of glib-or-gtk-build-system is what I'm trying to get at <brendyyn>atw: thats great. will you make a service to run the server? <atw>brendyyn: That'd be ideal but I'm focusing on making the package first <atw>hi pkill9, how are you and yours? <atw>follow-up question to the sentry thing: can I apply the same logic jaeger-client and opentracing, or are those things worth packaging? <leoprikler>if not, feel free to skip them and perhaps add them in later <vagrantc>anyone else have issues with "./bootstrap && ./configure --localstatedir=/var && make" failing due to some .po files being broken? <atw>vagrantc: I think I've had that, probably I deleted the po files that were untracked by git and tried again <jonsger>vagrantc: i usually do a `rm -rf doc/` and then a git checkout doc/ <wheeler>Anybody know of a good example package that combines 2 or more source tarballs in a single package definition? <brendyyn>wheeler: you can define an origin for it <brendyyn>goldenshimmer wanted to do the same thing too. i pointed them to adainisgpl and its data package <lfam>wheeler: Search for ',(origin' for examples <vagrantc>atw, jonsger ... still getting issues with a totally clean checkout <vagrantc>commit 97b1366d49179f38a365bdfa383a8ccba779f8b8 <jonsger>vagrantc: had this issues before: go to en@quot.po:4573 and adapt msgstr[0] to msgid_plural. this fixed the issue for me, I think <vagrantc>i don't really even understand what to fix <vagrantc>it's not even in git ... it's a generated file ... <vagrantc>and for some reason i'm the only one able to reproduce this <alextee[m]>having problems building an autotools-based package <alextee[m]>for some reason it generates a configure file with /bin/sh at the top <alextee[m]>i see no /bin/sh anywhere, idk where it comes from, but when guix runs the bootstrap phase it puts /usr/bin at the top of "configure" <jonsger>vagrantc: you're not the only one with this problem. I just have no more clue about .po files then you :( *jonsger wonders if there is a guide to "over" install Guix system to an existing linux distro... ***Parra2 is now known as Parra3
***Parra3 is now known as Parra
<vagrantc>i guess i'll try to debug this issue with guix pull then ... ***zalox_ is now known as zalox
<webstrand>roptat: looks like `(bootloader #f)` isn't the right syntax to disable installing the bootloader; is there another way? <webstrand>simply removing (bootloader ...) from the definition doesn't work either; it's a required field initializer <bavier`>webstrand: I have little experience with this, but I think Guix cannot function with an "external" bootloader. <bavier`>webstrand: but you can tell the bootloader about other entries it should have <jackhill>webstrand: you'll still get the guix-written grub.cfg <jackhill>bavier`: er, I guess the answer depends on what problem webstrand has, and I don't have the full context, but ran into one bootloader problem before, so wanted to surface that. <Gamma02>Do I have to use "sudo -i guix pull" for root also ? <jackhill>Gamma02: it depends on what you want to do, you might run them both depending on the circumstances <jackhill>Gamma02: so, in general, on Guix System, you don't need to run sudo -i guix. <jackhill>Gamma02: let's start with your regular user's guix (remember that each user can have their own guix command (the guix command comes bundled with all the current package definitions)) <Gamma02>And why "guix system reconfigure config.scm" download packages every time when I run it. Even though I didn't change anything in "packages" definition. <bavier`>Gamma02: if you ran 'guix pull' before, it'll probably require new packages to create the system <jackhill>Gamma02: `guix pull` updates a users copy of guix. Once it is update, you can do this with this guix like install package with the `package` command, or enter one-off environements with `environment`. Modifying the system profile (essentially the operating system config, services, what's in /run/current-system, etc) with the `system command` <jackhill>Gamma02: since the system command needs root since it affects more than one user, but you still want to use your user's copy of guix (which you just upated with pull), so that sudo with no -i <jackhill>Gamma02: now for why you'd run it with -i: The root user can have it's own copy of guix (and managed profiles and environments) just like any other user. Running sudo with -i sets up the environment so guix operates with root's local user copy of guix. <Gamma02>I wanted to rename my host name and username. I made those changes in users config.scm but instead created new user. <Gamma02>Can you tell me the right way to rename user ? <Gamma02>And when its okay to edit /etc/config.scm ? <webstrand>jackhill: well, guix is building successfully now. <webstrand>I added (target "/dev/null"), so hopefully it'll barf if it tries to write anything <webstrand>I have an existing linux installed on the same disk that I don't want to mangle <jackhill>webstrand: cool! heh, well hopefully that gives you the result you want. <jackhill>webstrand: yeah, I understand that's tricky. I haven't had to do that. <jackhill>Gamma02: you might be albe to rename the user by changing the name, and also specifying the uid to be your current numeric id (the fist number printed by the `id` command) <jackhill>Gamma02: as to config.scm, it's okay to edit it any time. Guix System doesn't consult it. It is only used when running the guix system command. In fact, there is nothing special about /etc/config.scm, it could be at any path. <Gamma02>jackhill: Small douby, when run 'sudo -i guix system reconfigure config.scm' warning says that /root/.config/guix/current not fount and guix pull never run. But it did run it. <jackhill>Gamma02: and is there a /root/.config/guix/current? <jackhill>Gamma02: I'm afraid that I don't know then, as I haven't run into that. <Gamma02>jackhill: Where does the %base-packages is defined ? <Gamma02>Do you have any config to window manager setups ? <jackhill>webstrand: That doesn't look related to the bootloader to me <jackhill>that seems like a bug. Is this from when you ran `guix system {re}configure …` <jackhill>webstrand: also, are you compiling everything locally instead of using substitutes on purpose? <jackhill>well, it's still a bug if it doens't build, but if you can use substitutes you can avoid building this package. <jackhill>webstrand: where are you running guix system init, from the installation image? <webstrand>great, I'll get them set up. I don't have any media to install the image to, so I've installed guix onto gentoo <webstrand>yep, that does indeed make it go much faster, thanks! <hpfr[m]>Just saw janneke’s talk about the full source bootstrap. He mentioned in passing that he wants gash and gash coreutils to be a good experience or something to that effect. Is there info anywhere on how that will manifest? Intended extensions beyond POSIX compliance, etc? <guix-vits>hpfr[m]: it's good to uderstand English without subtitles, btw. <hpfr[m]>guix-vits: I’m confused what you mean <guix-vits>hpfr[m]: i'd heard janneke's talk as something like "Wa Wi Well, well, do woo dis", see? <hpfr[m]>It is handy to speak English natively in the software world <hpfr[m]>But hopefully the subtitles were good <janneke>hpfr[m]: there is nothing concrete or even anything written about that afaik <hpfr[m]>Did you have any ideas behind it though? <janneke>we will probably draw inspiration from rash, scsh and guix' (guix build utils) <janneke>i imagine using scheme on the command-line, mixing it with regular sh <janneke>i am not sure about rash's feature of mixing those, i would like to work on the shell library and give it a friendly interface that you'd want to use directly ***apteryx is now known as Guest28000
***apteryx_ is now known as apteryx
<nagamalli>I am planning to write package description for GUI Kate Editor under KDE. Can anyone suggest me how i can get package description. <guix-vits>nagamalli: you can try see the other distributions descriptions <rekado_>nagamalli: have you looked at the tutorial in the Cookbook? <guix-vits>Arch's is "Advanced Text Editor" -- descriptive, hehe. <nagamalli>rekado_:I am new to guix and just done couple of cran package descriptions and document corrections. Can you suggest me which GUI application could be a good start. <nagamalli>I started with packaging GNU parted. but later understood its not a GUI. GUI version is GPARTED <nagamalli>guix import gnu parted --key-download=always <guix-vits>nagamalli: "definitions"; description is a field in definition. <nagamalli>guix-vits: shall i start with this "Advanced Text Editor" <rekado_>nagamalli: we already have packages for both parted and gparted <guix-vits>nagamalli: Arch's description and Fedora both states "Advanced Text Editor", let's see debian.. <rekado_>nagamalli: I recommend not to work on Kate, because we may not yet have all KDE libraries that it needs. <nagamalli>rekado_:I am very begineer in this ,can u suggest me a simple GUI to start with? <nagamalli>Can i start with GUI for GNU Packages. Will the cran importer help me in this regard <guix-vits>it has a good description for you package definition, but rekado_, as i see, do not recommend you to package the Kate for some reason. <guix-vits>nagamalli: are you need to package some GUI for a tool, or a game will be good too? <guix-vits>i see there is no SevenKingdoms yet; it's old game, and probably not hard to package (idk, honestly) <rekado_>nagamalli: I can’t recommend any GUI programs. Try to find a simple program with a web search. <nagamalli>i tried guix import gnu parted --key-download=always <rekado_>we already have a package for parted. <nagamalli>rekado_:yes i understood from what you said earlier. Just correcting my typo error :) <guix-vits>i'm was banned (hope not forever) from Debian's paste. <jonsger>is there somewhere a description/how-to/blog how to convert a foreign distro to Guixi System via `guix system init`? <jonsger>Blackbeard: difficult on a VPS when you can not mount an ISO... <guix-vits>jonsger: try ask nckx, but as i'm understood, He ran it on-top of some other distro. <andydarcyjewell>Hey, I've got a first "successfully built" message for Factor! So the binary is built, but I now have to do the "bootstrap" step, which means I have to run the fresh binary with a flag and the previously downloaded boot image as a parameter. <Blackbeard>That you could make a partition boot a live distro from a vps <andydarcyjewell>I guess there's no guarantee I'm out of the woods yet, but it's looking hopeful. <andydarcyjewell>So could anyone give me a hint about how I run the new binary with the required flags? <Blackbeard>There used to be a way to bootrstrap I remember but then I think it stopped working <rekado_>jonsger: you can generally just run “guix system init /”, but you will need to make sure to delete conflicting files in /etc before a reboot. <rekado_>I’ve done this a few dozen times on servers running Ubuntu and CentOS. <rekado_>if you forget to delete conflicting files that Guix System will want to checkout during booting you’ll be dropped into a rescue REPL where you can remove the files (but it’s not very convenient, even with bournish) <jonsger>rekado_: oke, I were in this situation after boot <civodul>jonsger: is your daemon running in a UTF-8 locale? <jonsger>ah maybe not. I pulled from guix-0.16 on openSUSE Leap, so I guess there is something missing <jonsger>rekado_: how do I find those conflicting files? Or which files are important? <raghavgururajan>rekado rekado_ Sorry, I missed your ping yesterday. I fixed the gsm issue. It was a typo. :-) <jonsger>do we have a tutorial or a documentation how to use bournish? <jonsger>only reboot and which do work. The rest gives unbound variable <rekado_>jonsger: it has echo, cd, pwd, rm, cp, help, ls, which, cat, wc, and reboot. <jonsger>rekado_: yeah but only reboot and which are working. And stuff form guix/build/utils... <rekado_>jonsger: but I can reproduce your problems in a REPL <rekado_>I wonder if that’s due to the use of @@ <rekado_>hmm, also looks like we can’t switch back to scheme from the bournish REPL <rekado_>I’m currently debugging something else, but I’ll try to take a look later <jonsger>I just wonder which files I have to delete addionally <rekado_>jonsger: those that are mentioned during a failed boot <guix-vits>is that ok that guix remove TAB show packages that not installed (in emacs M-x shell)? <rekado_>jonsger: possibly. Do you still need your old /etc directory? <guix-vits>jonsger: so you'd replaced other OS with Guix via init? <efraim>I wonder if we can replace bournish with gash-utils <janneke>i made an initial try towards that, but it made for a large initrd and i left it there *janneke goes to have a look *janneke finds that half of their branches are called gash or bootstrap, hehe *rekado_ almost fixed the nfs tests (and services) <guix-vits>looks similar as if you was attacked by unit of spearmens. <NieDzejkob>huh, have you ever observed an attack of spearpersons through IRC? <rekado_>civodul: nfs-server tests are fixed now ***ericst_ is now known as ericst
<civodul>we must be close to 100% of successful system tests now <jonsger>I finally got Guix System installed on my VPS via rescue booted debian <rekado_>jonsger: what broke after you removed /etc? <jonsger>rekado_: it was fine. I booted into Guix System and did `guix pull` and everything broke <dongcarl>Hi, is LC_ALL set in the build containers? <civodul>jonsger: yay! perhaps a recipe to share on the ML? <civodul>dongcarl: you mean in the build environment set up by guix-daemon? <civodul>but gnu-build-system has it point to a UTF-8 locale <civodul>you can see it at the beginning of build logs <dongcarl>civodul: Yes! Okay cool. UTF-8. Got it! UTF-8 locale from the glibc locale package? <jonsger>civodul: the breakage or my experiences in general? <civodul>jonsger: your experience setting it up <civodul>i think there are quite a few people trying to get Guix System on their VPS <civodul>and perhaps that'll include bug reports, too :-) <rekado_>we should fix the Environment line in etc/guix-daemon.service.in <rekado_>yesterday we got confirmation that changing the locale to en_US.UTF-8 (from en_US.utf8) fixes locale warnings. <guix-vits>rekado_: not worked for leoprikler, probably. <leoprikler>I actuall didn't encounter the bug yesterday after talking about it. <leoprikler>But I'm not quite sure when to expect it either tbh. <civodul>rekado_: really? let's do that, then <civodul>that's weird because glibc-utf8-locales provides both <leoprikler>It also provides de_DE.UTF-8, which is the value I use. <leoprikler>But I'm somewhat unsure about this daemon stuff. <leoprikler>The service file is also for foreign distros, whereas I use Guix System. <vagrantc>it produces a source tarball without the patch on x86_64 (good) ... but when i specify --target=aarch64-linux-libre it still doesn't have the patch ... <vagrantc>trying on aarch64 natively to see if there's a difference <civodul>vagrantc: '((search-patch "...")) is just a list <civodul>it does not lead to a 'search-patch' call because it's quoted <roelj>I'm trying to compile a package that does not work with glibc >2.28. Can I somehow override which glibc is used to build a package in its recipe? <roelj>I see the package “jamvm-1-bootstrap” puts glibc-2.28 in its native-inputs. That does not seem to work for the recipe I'm trying to get to work. <civodul>roelj: it's not that simple unfortunately <civodul>because gcc has a hard-coded libc file name <roelj>civodul: Like “("gcc-toolchain" ,(make-gcc-toolchain gcc glibc-2.28))”? <webstrand>Is there any documentation on how to boot guix system without an initramfs? i.e. inside of a container? <leoprikler>what kind of container? If you need a docker image, you can create on through `guix system` <civodul>rekado_: there are two unbound variable warnings in (guix import cran), related to hg *civodul shares a cool hack + accompanying wall of text <webstrand>leoprikler: It's not technically a container, I'm trying to boot an installed guix system from a generic initramfs <roelj>Hm building a custom gcc-toolchain seems to not work here. I get the following error a couple of times: xgcc: error trying to exec 'cc1': execvp: No such file or directory. Have I missed something? *dongcarl is glad someone's using my procedures <dongcarl>roelj: Perhaps you can post full logs on ML? <andydarcyjewell>Afternoon. Can anyone tell me how to refer to a file downloaded in the inputs stanza, when invoking a call to the generated binary during a build phase? <leoprikler>(string-append (assoc-ref inputs "in") "/foo/bar/baz") <andydarcyjewell>leoprikler: thanks, but how do I know *where* the file was downloaded to? <leoprikler>For what purpose do you care? Is it a file, that is not included in the packages output? If so, you'll have to include the source. Do you need the source of your own package? That's (assoc-ref inputs "source"). <hulten>Is anyone using ~/.xinitrc or ~/.xsession? I am getting "Failed to execute login command" at login. <hulten>I use standard xorg-server. In ~/.xinitrc is only the full path to the window manager. <hulten>The goal: run a program before the window manager starts. <andydarcyjewell>leoprikler: It's the bootstrap image file for the Factor system, and you have to run the newly generated binary with it as input, so that you can get factor to work. <andydarcyjewell>leoprikler: You have to download it separately from the source code; it's a memory image (I think) of the factor REPL. <leoprikler>Okay, apart from the thing, that adding more bootstrap blobs is unguixy, here's what you'd do normally: <leoprikler>Inside whatever phase, refer to it as (assoc-ref inputs "factor-image") <leoprikler>If that's a directory with multiple files do the string-append stuff from above <andydarcyjewell>Can't do much about the way Factor works (and my understanding of Guix is still pretty basic). <andydarcyjewell>As I understand both better, I will hopefully be able to improve on that. <leoprikler>instead of "-i=...", do (string-append "-i=" (assoc-ref inputs "factor-image") "/...") <guix-vits>hulten: did you tried GUI tools for that ("autostart" things)? <andydarcyjewell>Sorry, just me being thick. I thought it was some scheme syntax/convention that I hadn't seen before. Duh! <civodul>dongcarl: we build 'ar' (Binutils) such that the ordering is deterministic by default <leoprikler>Also instead of hardcoding the arch, you should probably do some target-system stuff <dongcarl>civodul: You mean --enable-deterministic-archives? <andydarcyjewell>>Also instead of hardcoding the arch, you should probably do some target-system stuff <hulten>guix-vits: Wat GUI tool would you suggest? I am running spectrwm. <dongcarl>Right, but I think --enable-deterministic-archives only sets uid/gid/timestamp no? <hulten>Sorry, the last remark should not be relevant, because I'd like to run things before spectrwm starts. <dongcarl>Also, I've found that for .dll.a files, even --enable-deterministic-archives is not enough to get uid/gid/timestamp reproducibility... I had to patch dlltool (which I will upstream) <civodul>dongcarl: ah oh! i thought it took care of file ordering, but maybe not <civodul>it probably takes things in the order they appear on the command line <civodul>so how the shell expands wildcards is key here <dongcarl>civodul: Right, and I'm guessing some script is globbing somewhere <civodul>yeah i saw that comment and initially thought we were not affected <hulten>In the end, I want to synchronise a subset of my $HOME before and after logging in. I am used to using Unison for this. <hulten>Any other approaches than "unison; spectrwm; unison" in my .xinitrc? <dongcarl>civodul: Yeah I'm gunna give it another go myself today, and probably post on repro-builds ML if I don't get anywhere with it <leoprikler>create a service for unison and then launch it whenever <andydarcyjewell>Ok, looking better, but I think the Factor binary is now complaining that it can't find the image file, and the build instructions say it should be in the same dir as the source, so is there a way to move it there? <hulten>Leo, can I use shepherd to run the unison service just before (and after) I run the window manager (as the same user)? <hulten>It sounds very neet if this can be done with Shepherd. <leoprikler>It requires some setup, but it can definitely be done. <leoprikler>First, let's assume you have a unison oneshot service and a service for spectrwm. <leoprikler>Then you can write a service, which depends on spectrwm-service and launches unison-service. <dongcarl>Is there a way to re-build a derivation you've already built? <dongcarl>Oh right, thanks! And that works with `--rounds` too? <lispmacs[work]>should we submit desired packages as "wishlist" bugs, or would that just annoy the distro maintainers? <lispmacs[work]>there is a few packages I'd love to see but don't have time to work on packaging all of them <janneke>grrr help2man: can't get `--help' info from bin/dzn check <janneke>of course, no further information or log file, and "it works" when running it manually in /tmp/guix-build* *janneke goes to instrument the build <hulten>leoprikler: The spectrwm and unison services (and learning Shepherd) are on my desirata list, but I decided that for now I'll manually call the text interface of Unison from a vc before/after logging in. <hulten>There is a time for automation, and there is a time for work (what I'm payed for to do) :-/ <leoprikler>in that case, you can just manually do the after call on your own, can't you? <leoprikler>and IIRC it should be exec spectrwm inside xinitrc <hulten>.xinitrc does not work properly for me. <hulten>I have "exec path/to/spectrwm", but it cannot execute the command, it says. <hulten>That was the problem to start with. <hulten>"Failed to execute login command" I get when logging in from the desktop manager. <hulten>I should try startx from a vc, maybe. <hulten>.xinitrc is simply in my homedir, $HOME == /home/`whoami`/. <hulten>This should be standard also on Guix System, right? <hulten>Then I see the directories to the users. <hulten>But ls -la /home/`whoami` shows that .xinitrc is a symlink; I'll remove that complexity.. see what happens. <hulten>Good question! The standard xorg-server service, I think slim? <leoprikler>gdm configuration has a xsession field, perhaps that's what you want <hulten>Let me double check an my config.scm. <hulten>Æsj... I now see that I am using slim and forcably removing gdm-service-type. <hulten>I must've had a reason, but I'll (1) see if slim also has an xsession field, and else (2) try with gdm. <allana>Hi Guix. Anyone have experience setting up VPNs using NetworkManager on GuixSD? I recently saw that network-manager-openconnect plugin was added to guix. I use openconnect in a hacky script to establish a VPN connection, but I was hoping for a more streamlined experience using network manager. Now that should be possible but I am having difficulty finding out how to do this in gnome. <dongcarl>The gnu build system does not automatically autoreconf does it? <mbakke>dongcarl: it does if it determines that it is needed (see the 'bootstrap' phase) <mbakke>i.e. if configure.ac exists, but no configure <kondor[m]>allana: what's wrong with a hacky openconnect script? i have one too :) <kondor[m]>did anyone try running a VNC server from Guix? <kondor[m]>i made a tigervnc package ages ago, but never a service to go with it <dongcarl>What is the `mirror://` scheme I see for the gcc package? I've never seen this elsewhere... <allana>kondor[m]: the problem with my hacky script is that when I want off my vpn, I kill openconnect and my network connection seems buggy <kondor[m]>i can do authentication as a normal user and then <allana>guix-vits: When I go to add VPN I only have an option to load from file... I'm not sure what to do with that and my searches so far have nothing promising <kondor[m]>never had issues with it, unlike with cisco client <allana>kondor[m]: yeah I use a similar approach as normal user but sudo for openconnect. I also integrate it with pass *apteryx gets bitten by the CPATH / CMake system problem on master again <janneke>mbakke: yes, i hoped that touching configure.ac would do that -- but gnu-build-system.scm has: (if (not (file-exists? "configure")) <auto stuff> (format #t "GNU build system bootstrapping not needed~%")) <raghavgururajan>I am kinda close to finish packaging of Linphone. But I am blocked by an *error*. <mbakke>janneke: there are a few places that delete the configure script to force a bootstrap <mbakke>apteryx: you mean packages that add -isystem /gnu/store/...-glibc/include ? <mbakke>raghavgururajan: didn't you fix that already by building bctoolbox with cmake-build-system ? <apteryx>When the CMakeLists.txt does: target_include_directories(egit2 SYSTEM PRIVATE "${libgit2_SOURCE_DIR}/include"), it just wouldn't find a header that should have been included system wise (git2.h). <raghavgururajan>mbakke I was about to you ping you. That was different. I fixed it. This is another one. But the trick you mentioned last time didn't work. <apteryx>or it would, but some symbols that this git2.h imports itself wouldn't. <apteryx>removing the SYSTEM from the above CMakeLists.txt works around the issue. <apteryx>It has to do with -isystem not picking up CPATH, IIRC. <raghavgururajan>mbakke The previous one was building bcunit with cmake-build-system. For this one, I already used cmake-build-system for bctoolbox. Also, I tried "-DBcToolBox_DIR=" (assoc-ref %build-inputs "bctoolbox"), it did not work. <mbakke>apteryx: right ... we really should get that branch going! :-) <mbakke>raghavgururajan: it should be "BcToolbox_DIR", not "-DBcToolBox_DIR" <mbakke>raghavgururajan: Oops, I mean "-DBcToolbox_DIR" <mbakke>raghavgururajan: and building bctoolbox with cmake-build-system did not fix it? perhaps you need a newer version? <mbakke>weird that the project expects to find that file if bctoolbox does not have it! <mbakke>janneke: how confident are you with moving the Hurd work over to the core-updates branch? :-) <raghavgururajan>mbakke Is there way to pass an argument to build system to skip looking for that file? <mbakke>raghavgururajan: I wouldn't know, you'll have to read the build scripts to see if there are other ways to make bctoolbox available <janneke>mbakke: it would be great to get see a delta with core updates, then it would be less a question of feeling confident :-/ the cuirass build did help me find a silly typo (an else-less if); of course we can always try to blame civudul for being a sloppy reviewer ;) <janneke>mbakke: /me is trying to say, yes let's move wip-hurd onto core-updates :) <mbakke>janneke: the aarch64 and arm failures seem to be because they unsurprisingly fail to build 'mes-boot'; not sure why they even attempt that <pinoaffe1>does anyone else have issues with webrtc on guix (or on the contrary, has anyone managed to set up a mic over webrtc on guix)? <bavier`>pinoaffe1: I've gotten it to work with ungoogled-chromium <pinoaffe1>bavier: anything special you had to do? because for me, the webcam does work but the microphone doesn't <pinoaffe1>if I try to toggle it to "allow" on the settings page, it goes back to "block" automatically, and it does not do this for other options <lfam>Does the microphone work in application pinoaffe1? <lfam>I mean, in any application? <mbakke>pinoaffe1: for ungoogled-chromium in particular, check if ~/.config/chromium/Default/Preferences contains a line such as "audio_capture_enabled: false" and remove it if it exists <mbakke>it was a bug in early revisisions, sorry about that! <pinoaffe1>mbakke: oh, that seems to have done the trick, thanks! <mbakke>janneke: aarch64 and armhf seem to take the normal bootstrap path on wip-hurd still (phew), so I guess there is something odd with the Cuirass specification <mbakke>the other failures seem to be with the cross-compilation infrastructure <janneke>mbakke: good -- yes i did notice some mingw cross builds to fail <mbakke>mingw, powerpc, risc-v... no shortage of things to fix <janneke>how to best test wether wip-hurd introduces a problem in the cross build? <mbakke>though I don't think they are blockers <mbakke>janneke: AFAICT we have the same problems there <mbakke>including the odd mes builds for aarch64 and armhf <janneke>mbakke: "very nice" -- from a wip-hurd perspective <Rovanion>Do I need to do anything special to get WiFi on my Intel 8265 card? <janneke>they would have to be verified for sure <Rovanion>lspci -nnk lists the wifi card with the kernel module iwlwifi loaded, but nm-applet shows no ability to configure it. <mbakke>janneke: good point. I can build them overnight and check if I get the same result. <kmicu>Hi Rovanion your card is not old Atheros9 or old Broadcom B43 so for sure it requires a blob firmware and will not work with kernel Linux-libre. You can start by installing stock kernel; there’s a blog entry about it. <lfam>So, is (gnu packages cran) just a place to put R packages that don't fit anywhere else? <kmicu>Rovanion: If you choose that path that will require kernel recompilation from time to time after guix pull. <janneke>mbakke: thanks, i'll do that too -- i have rebased a couple of times on core-updates and i seem to remember a glibc change recently? <mbakke>janneke: they would have to be built from the commit mentioned in eb63855847230418c05d6cfecbc6dd364d0a2eb8, so any later changes won't matter <Rovanion>kmicu: Huh, is there a way for me to know if the drivers for any hardware requires binary blobs programatically or is it just a matter of memorizing the few pieces of hardware that doesn't require blobs in each category? <janneke>mbakke: yes, if you do that then that's right <dongcarl>Perhaps someone else has more experience with this: are we allowed to just copy over patches from debian to use in Guix? Are the patches free to take? <mbakke>janneke: unless there is a reason to add newer bootstrap tarballs? <janneke>mbakke: that commit lived on wip-hurd a few rebases ago, so it may be pruned some time? <mbakke>we need to ensure that commit is in history, hmm <janneke>mbakke: yeah -- so i could rebase wip-hurd on core-updates right now <janneke>mbakke: and we could do an overnight build -- or mark that commit with a tag? core-updates can be a pretty moving target? other than this, i see no reason to rebuild them <mbakke>janneke: can the Hurd bootstrap tarballs be created from the current 'core-updates' branch? <jonsger>dongcarl: In general yes, I guess. Still it's not 100% sure under which license we "distribute" them <janneke>mbakke: no, not useful ones anyway; we need at least the glibc fixes and the static-guile => 2.0 change <lfam>dongcarl: Yes, it's okay <dongcarl>Okay, I'm going to leave the original author in the patches, just in case <lfam>Like jonsger says, some aspects are not totally clear but given the synergy between our goals we should assume it's okay <janneke>mbakke: i think we pretty much need all commits (or equivalent ones) below the commit that adds them <lfam>dongcarl: For reasons beyond copyright attribution, it's very helpful to leave the authorship info and a link to the source of the patch <lfam>I would assume that a patch on a Debian package would be fall under the license of the package <dongcarl>Okay got it, I will be sure to specify the source as well! <janneke>mbakke: what if we do it in two stages: merge wip-hurd up to just under the bootstrap-binaries right now <mbakke>janneke: that will create new bootstrap binaries because of the Guile 3.0 switch :/ <mbakke>janneke: but I think we can do a git-merge of commit 7d260ede842141ad3405ec484d6265ac266757a0 directly <janneke>mbakke: ow... great! we are so parallel <mbakke>we'll get some useless commits such as the flex 2.6.1 fix, but I guess it's not a big deal <janneke>mbakke: didn't i rewrite -- yeah; and also i changed some annotations on patches -- no that's not so great *janneke needs to slow down a bit <mbakke>a tag works, but it's not great because some mirrors may lack it, or someone could accidentally delete it from savannah, etc <mbakke>so it's best if the commit is reachable from the linear history <mbakke>janneke: is it a lot of trouble to just create new ones? :/ <janneke>yes, so i like the two stages approach best; maybe the guile3 change isn't so bad; we get new hashes only once again and only i need to do an extra rebuild <mbakke>..assuming the bootstrap works from the new Guile 3 binaries ;-) <janneke>mbakke: i think that's what i like best too! <janneke>right...someone would need to do a quick check :) <mbakke>janneke: OK, let's merge the commits up to the bootstrap tarballs and start building! :) <janneke>mbakke: would you like to do that, otherwise i can do that in 1.5hours or so <mbakke>janneke: very happy if you can do it! not in a hurry :-) <mbakke>I promise not to break core-updates in the mean time! :D <mbakke>then we should be ready to start the full rebuild in a day or two \o/ <mbakke>civodul: will you have time to do a Guile 3.0.2 release in the next few days? <Parra>is it possible to change the glibc when compiling with Guix? Let's say by musl. <civodul>lately i was trying to address compiler performance issues <civodul>but i guess that's not going to happen overnight <jonsger>mbakke: how could I add glib:bin to my config.scm? I forgot how to do it <mbakke>jonsger: (cons (list glib "bin") %base-packages) <mbakke>or (append (map (specification->package '("glib:bin"))) %base-packages) <mbakke>civodul: was the GC bug with libgc 8? last time I tried making it the default 'guix' failed to build *janneke pushed the bottom half of wip-hurd to core-updates <janneke>let the bootstrap-tarballs rebuild begin! <mbakke>janneke: wohoo, I'll start './pre-inst-env guix build --target=i586-pc-gnu bootstrap-tarballs' <civodul>mbakke: oh yes, that one was fixed too <jonsger>mbakke: yes it was. resolving my issues on opensuse with libgc 8 was included :P <civodul>i was referring to a bug with weak sets <janneke>mbakke: yeah! me too -- oh boy, i do hope those will work too <mbakke>civodul: OK great, let's promote libgc too when 3.0.2 is in and get rid of libatomic-ops :-) <jonsger>testing mate with glib:bin looks a lot better <civodul>mbakke: woow, you were quick to fix ImageMagick! <civodul>looks like graphicsmagick doesn't have the problem <mbakke>civodul: I had to start it right away so it did not disappear in my queue again ;-) <lfam>That is not supposed to happen <lfam>What's the situation where it crashes? <pkill9>possibly when there's no sound being processed, but other than that it's completley random <lfam>I mean, what kind of system are you running? Is it Guix? Graphical? Architecture <lfam>How do you observe the crash if there is no sound being processed? <pkill9>it did this when i ran gnome too <pkill9>my bluetooth headphones disconnect <pkill9>then i pgrep pulseaudio and it' snot running <pkill9>it did this before using bluetooth headphones too <rekado_>pkill9: do you find any information in the logs? <lfam>I'm not a pulseaudio expert but my understanding is that it's spawned when needed. So maybe there is some way to keep it running even when you don't need it <lfam>I think it's more likely that it's exiting, not crashing <pkill9>really? i thought pulseaudio just continues to run <pkill9>rekado_: not in /var/log/messages, where else would i check? <lfam>pkill9: Why don't you think it is exiting normally? <pkill9>i just don't see why it would exit <pkill9>it's supposed to handle multiple audio applications <pkill9>why would it be designed to just stop running? <lfam>That's my experience with it. When there is nothing running sound, it exits. The application is supposed to start a pulseaudio daemon when required <rekado_>pkill9: your assumption seems to be that it doesn’t handle multiple applications wanting to “do sound” <pkill9>rekado_: my assumption is that it *does* handle multiple applications wanting to "do sound", did you mean lfam? <mbakke>civodul: 'master' sounds fine to me, great to finally fix that issue! *mbakke signs off for the night, gn everyone :) <lfam>sneek: later ask efraim: How do you decide which Qt modules go in pyqt, efraim? Is there a list of required modules? Can we add more? I'm packaging something that expects to be able to run qtcharts from pyqt <lfam>Hm, maybe it's not correct to say that pulseaudio exits, not sure. But it does shut down whatever abstraction is has over the output when nothing is being processed. I experience by a loss of volume control from MPD when it's not playing <lfam>I'd guess there is something configurable here <Blackbeard>civodul: do you think I should write a library specifically to parse systemd unit files <civodul>Blackbeard: there could be a module within the Shepherd dedicated to parsing those files <lfam>It seems like we would have to add a lot of functionality to Shepherd to meaningfully translate systemd unit files <lfam>But, even being able to translate the basics would be useful if it motivates more Shepherd development <Blackbeard>Ok so should the systemd unit files be their own child class of <service> <Blackbeard>lfam: I am excited to work on this. But I know it would be challenging <civodul>Blackbeard: i agree with lfam, parsing is icing on the cake <civodul>the core bits of the work are probably implementing some of the "unit" management functionality found in systemd <pkill9>im running pulseaudio with high verbosity level and this time not in daemon mode, so when it fails agian i will see what it says <pkill9>dunno why i didn't think of that <pkill9>i got so used to running pulseaudio -D <kondor[m]>I cannot export org files to LaTeX. The LaTeX output complains that grffile.sty is missing. Any idea which texlive package do I need for this to work? Btw, I tried installing grffile.sty into texmf/tex/latex and this still did not work. kpsewhich grffile.sty does not return anything. <kondor[m]> * I cannot export org files to LaTeX. The LaTeX output complains that grffile.sty is missing. Any idea which texlive package do I need for this to work? Btw, I tried installing grffile.sty also into ~/texmf/tex/latex and this still did not work. kpsewhich grffile.sty does not return anything. <Blackbeard>kondor: how are you exporting, are you using extra packages? <rekado_>kondor[m]: this should be provided by the oberdiek package. <kondor[m]>rekado_: thanks that got me to the next missing package <kondor[m]>for the reference, after installing texlive-latex- ... wrapfig, oberdiek, ulem, capt-of and hyperref org export stopped complaining about missing packages. now it simply dies ... <kondor[m]>maybe because some fonts are missing; there was no sensible error message <kondor[m]>also above, it is texlive-generic-ulem not texlive-latex-ulem, sorry <rekado_>not all of our texlive packages are actually complete <rekado_>I spent a lot of time fixing some of the most broken texlive packages, but there are many more that are incomplete <rekado_>it’s usually those texlive-{latex,generic}-* packages that are bad <kondor[m]>ugh, sounds like the kind of work that would suck the life out of me <kondor[m]>rekado_: btw, tex importer always dies on me <kondor[m]>guix import texlive never works, not for any package I ever tried ... is this normal? <Blackbeard>rekado_: thanks for your work. I use latex everyday <rekado_>kondor[m]: the importer is … special <rekado_>it helped me a lot to create latex packages initially <rekado_>but it really only works on the texlive SVN repository <rekado_>it’s very sensitive to what arguments you pass it. <rekado_>Pierre wanted to rewrite it so that it uses the texlive.tlpdb but it looks like he lost interest. <kondor[m]>yeah, i tried divining stuff from that svn repo, <rekado_>I suggest looking at texlive.tlpdb instead to figure out what files should be included, and to be able to guess which of these are source files. <kondor[m]>how complicated is it to transpose, say, debian texlive package to guix? <rekado_>kondor[m]: you can’t easily translate that because Debian installs everything to the same prefix <kondor[m]>i'm by no means a guru when it comes to the guts of tex(live), but i'll try <rekado_>this is *much* easier for a package like texlive than to install each part into its own prefix as it’s done in Guix <rekado_>in Guix we try to tie these individual packages together by pouring a lot of configuration glue over them. <rekado_>(in a profile hook and with texlive-union) <kondor[m]>is guix texlive configured so that it does not give a darn about the users ~/texmf, or is it some misconfiguration on my side of things? <kondor[m]>i mean kpsewhich -var-value TEXMFHOME returns ~/texmf correctly <kondor[m]>but running pdflatex on stuff will ignore anything in there <kondor[m]>Blackbeard: would it be too much to ask you for a list of your latex-.* and font-.* packages? <kondor[m]>i am trying to export a very very simple document <Blackbeard>kondor: can you share your document ? Did you modify something un your org export configuration? <kondor[m]>i think i didn't, at least not in the part that concerns latex <Blackbeard>kondor: did you install LaTex as your user or in config.scm <rekado_>texlive in Guix wraps the binaries in TEXMFHOME, so it’s possible that ~/texmf is accidentally ignored. <kondor[m]>ahaha, that sounds like chaos ... reproducible, but chaos :) <rekado_>kondor[m]: you can use the huge “texlive” package if you really need to build the document urgently <rekado_>but it would be better for Guix if you investigated what’s wrong and reported a bug <rekado_>because we know that “something’s wrong” — but fixing it requires a real investigation. <kondor[m]>what does that do? installs the entire texlive? <kondor[m]>rekado_: well i can tell you about one strange thing <kondor[m]>the dir returned by kpsewhich -var-value TEXMFCNF <kondor[m]>though, you did tell me earlier to ignore kpsewhich