<jgibbons[m]>For (easily guessable) reasons I am embarassed to disclose I need to use the exFAT file system. I see two packages claiming to provide exfat: exfat-utils and fuse-exfat. Which of them do I need to install, and should they be systemwide packages if I want to use gnome-disk-utility? <joshuaBPMan>Hello, I'm thinking about buying a Lenovo T400 laptop and flashing it with libreboot...but doesn't libreboot not ship microcode? doesn't that make it vunerable to meltdown and specter...? <vagrantc>some of it is mitigated by disabling hyper-threading <joshuaBPMan>what is hyper threading? And I'm assuming that would slow down the computer a bit.... <vagrantc>it's a cpu feature that sort of pretends you have more cpu cores than you actually have... sort of. <vagrantc>all the mitigations for spectre/meltdown take a significant performance hit ***apteryx_ is now known as apteryx
<nckx>sneek: joshuaBPMan: Linux takes care of disabling that for you (you can pass ‘mds=…’ on the kernel command line if you like to live slightly faster & more dangerously). ***MinceR_ is now known as MinceR
<xavierm02_>Hey. I'm using my Guix laptop on battery with a usb mouse plugged in, and whenever I stop moving the mouse for 1s or so, then it won't move until I click. I tried setting (usb-autosuspend? #f) in the tlp service and restarting it with herd but that did not help. <xavierm02_>The keyboard seems to also sleep a bit but unless I press the key really fast, it was up before my finger leaves the key so it's fine. <xavierm02_>for i in /sys/bus/usb/devices/*/power/control; do echo on > $i; done <xavierm02_>But if anyone knows a less brutal way of fixing the problem, I'm interested <civodul>anyone knows if we have a Matrix client that can do video? <rain1>i think riot in the browser can do it <sisyphe_>hi, if package A lists package B as input and package C lists package A as cargo-input, then package B should be available during package C build, right ? <civodul>not sure if those hacking on Cargo/Rust are reading here <efraim>i'm not sure how far down the rabbit hole it's supposed to go, we might go too many levels in guix <sisyphe_>efraim: thanks, I should be missing something else so... <efraim>i was going to play with actually installing the libraries that are built and seeing if that can make it a bit better <sisyphe_>not in this specific issue i'm afraid. package B is pkg-config in my case, so not a rust package (package A and package C are) <efraim>only took 9 hours! finally got ndarray to build :/ <efraim>rust-pkg-config automatically finds pkg-config, so I've added pkg-config anytime rust-pkg-config gets added to a package <efraim>some of the other ones need an environmental variable set to find them <roptat>civodul, I took a few weeks to rest and relax, and now I'm back with all my energy :) <sisyphe_>package A (libdbus-sys) also lists rust-pkg-config as cargo-input and it does find pkg-config correctly but package C only compiles when I add pkg-config manually as an input (if you want to take a look: https://paste.debian.net/1097421/) <efraim>rust-dbus will pull in rust-pkg-config, and i assume without pkg-config it'll try to compile a vedored copy of dbus, so I would add pkg-config to rust-network-manager <ng0>efraim: did you know some of what you shared with me as the "rust with nix" articles is an open pr or rfc? and it gets even more fun when you dive down the way of getting to a reliable way to pass through various arguments to rustc etc via cargo <sneek>ng0, efraim says: I have inconsistent results with some crates using 'extern crate' on some deps and not others, and shared deps freaking out the compiler <sneek>ng0, efraim says: libgit2-sys can almost link to libz which can link to libc, but libgit2-sys doesn't like that libz isn't linked to the vendored libc source it's using, so that's where I'm stuck for now <efraim>I didn't dive as deep as the article to hack it together on guix, I didn't really want to parse Cargo.toml files <ng0>i can't recall but have bookmarked them, as I encountered them reading through this <ng0>maybe at some point I'll realize the only good progress will be getting involved in cargo and rust. <efraim>I was told on mastodon it should work, but people don't actually do that, and the --extern flag is what I'm looking for <ng0>it's quiet weird to pass through arguments to the linker, there are ways, but hey please read our environment <efraim>but it sounds like it wants the sources anyway <sisyphe_>efraim: is there a way to write libdbus-sys such as all packages depending on it will also pull pkg-config automatically ? <ng0>all I wanted is to package alacritty and a couple more and go back to writing my applications <efraim>sisyphe_: I don't think so, in essence cargo-inputs grabs the sources and the transitive rust sources <civodul>roptat: good, good! :-) i hope you had a good time! *civodul relaxed for a few weeks as well <roptat>I see there was progress on the website translation :) <ng0>fun fact: did you know that xplanet release at or around the same time as leap years (if I remembered it correctly)? I have to ask the author some questions and noticed this pattern <civodul>roptat: what i meant is that it's more optimal storage-wise to have a single clone for related branches <civodul>rather than one complete clone per branch of the same repo <roptat>I understand, but I don't know how to do that <efraim>sisyphe_, ng0 : I remember what it was! `crate build foo` and `crate test foo` require all the sources and dev dependencies, even if optional. `crate install foo` will work with the bare minimum (and .rlib files IIRC) and whatever else you decide to put in the environment <roptat>you seemed to suggest something that I didn't understand <roptat>if we share the same directory, we get the race condition <civodul>roptat: well, the name of the cache directory is currently *not* a function of the branch name for that reason <civodul>and you were suggesting to compute that directory name as a function of the branch name <roptat>and to fix the race, we used two separate directories, using two separate names for the same repo <roptat>except one of these names can't be used to clone the repo <roptat>so I was suggesting to use the same repo, different branches, and a different name <roptat>which is just as wasteful as before, but works ^^ <civodul>and then i said this sounds good, *but* this should be done specifically for (sysadmin web), rather than being the default of (guix git) <roptat>your "Ok." was a bit mysterious :p <roptat>so should I push that patch to maintenance, or is there anything I should change in it? (like computing a hash instead of using a name directly?) <civodul>roptat: yes, actually the patch LGTM <roptat>civodul, ok, I'll push this evening then. Hopefully someone will be able to reconfigure and tell me what's wrong with the patch :p <civodul>oh right we'll have to remember to reconfigure <efraim>sisyphe_: rust-xz has a cargo-input of rust-lzma-sys, which has inputs of pkg-config and xz. rust-xz needs pkg-config and xz as inputs when building because rust-lzma-sys needs them available <efraim>similar to your problem with rust-dbus <gnu_srs>Hello, I'm trying to change the keyboard in /etc/config.scm: (keyboard-layout (keyboard-layout "sv" "altgr-intl")) but <gnu_srs>sudo guix system reconfigure /etc/config.scm fails with build of /gnu/store/...-console-keymap.sv.drv failed. <gnu_srs>Build log shows: /gnu/store/...-console-setup-1.191/bin/ckbcomp: Can not find file "symbols/sv" in any known directory <gnu_srs>Changing sv to se does not have any effect, even since there is a file /gnu/store/...-xkeyboard-config-2.26/share/X11/xkb/symbols/se <gnu_srs>And /gnu/store/...-console-keymap.se <sisyphe_>efraim: ok, I did the same and it worked. It's unfortunate that we need to specify system dependency for each reverse dependency of a package <efraim>48 -sys dev dependencies for windows <efraim>i skipped the build and test phase, it installs just fine as is <sisyphe_>yes, I did the same (without noticing) since it's also an indirect dependency of rust-network-manager <sisyphe_>how should I handle a package needing two source repositories ? Guix doesn't seem to allow having multiple source <ngz>sisyphe_: Could you use one source as the main one and the other in an input? <ngz>See, for example, drascula in "games.scm". <efraim>gcc exit status 4 is the C version of exit status 9? <rekado_>civodul: congratulations on fixing ghostscript! I wonder if the fix also resolved other ghostscript bug reports (e.g. that one segfault). <roptat>gnu_srs, then it did build for "se", or am I missing something? <roptat>gnu_srs, "sv" doesn't seem to exist, even on a foreign distro <gnu_srs>roptat: Yes, it built OK for "se". I thought e.g. locales being sv_SE.UTF-8 should use "sv", not "se". But the keyboard is still "us" in the xfce4 terminals. <roptat>gnu_srs, you need to reboot for the change to take effect <roptat>(it comes from the setxkbmap package) <gnu_srs>roptat: Thanks, installing and using setxkbmap solved the problem :) <civodul>rekado_: it fixes pdf2ps, DocView, and (presumably) printing; i don't know about segfaults? <civodul>i feel bad that it remained broken for this long, and for such a silly issue <roptat>gnu_srs, well, only temporarily ;) <rekado_>it was also font related, so there’s a chance it’s fixed now too. <rekado_>with some luck our new servers will arrive on Wednesday. <benderst`>civodul: works for fine from what I've seen using it lightly ***benderst` is now known as bendersteed
<efraim>(@ (gnu packages foo) bar) is bar, but what about bar:lib? <nckx>efraim: Depends on the context. <efraim>i wanted to pull in gfortran:lib without importing gcc, '(@ (gnu packages gcc) gfortran "lib")' didn't do it <efraim>don't remember what the error was now, I just imported gcc in the end <nckx>efraim: (@ (gnu packages gcc) gfortran "lib") doesn't look like valid Scheme to me but I would expect (inputs `("gf:lib" ,(@ (gnu packages gcc) gfortran) "lib")) to work. <gnu_srs>Hi again, I'm a newbie here. How do I find a package/a file in a package, like apt-cache search package/file. Conversely how do I find which package a file belongs to, like dpkg -S /bin/ls <nckx>#~(string-append #$(@ (g p g) gfortran):lib "/lib/foo") would work too. <nckx>gnu_srs: ‘guix package -A <package>’ will print the file name. There's no functionality to map package contents to package names yet. <efraim>gnu_srs: if you have the binary installed then `readlink $(which readlink)` will show you readlink is from coreutils, `guix package -A coreutils` shows you it's in gnu/packages/base.scm <nckx>‘ls /gnu/store/*-*-*/…’ is at that point also a desperate option. <lurkki__>I'm getting a "Cannot make/remove an entry for the specified session" error after doing TUI install <lurkki__>that error happens when trying to log in i should add <nckx>Never seen that message before. When, exactly, does it happen (‘after’ sounds like ‘not during’). <lurkki__>after that it throws me back to the login prompt <gnu_srs>Tank you all for your replies: Further questions: How to know different architecture status, especially Hurd? <gnu_srs>And is e.g. Hurd support cross-compiled from e.g. Linux/amd64? <nckx>lurkki__: It's a PAM error (libpam.so), that's all I can say. Might be worth checking your system logs (from a rescue system, can be the installer) for any interesting messages, and add them to a bug report to bug-guix@ if they're not clear enough to let you fix the problem yourself. <gnu_srs>I'd like to get into make progress of the Hurd port when I've learned more about Guix/Shepherd/... <nckx>I thought all syslog messages were spammed to the (VT) console too but maybe not. <lurkki__>i got the filesystem i installed it on mounted <nckx>lurkki__: Shouldn't be Guix-specific (although anything not using journald is ‘specific’ nowadays, I guess); /var/log/{messages,secure} I guess? I'm not sure what's going on so I'm not sure where to look. <lurkki__>Aug 26 15:14:04 localhost login[203]: pam_elogind(login:session): Failed to connect to system bus: No such file or directory <lurkki__>Aug 26 15:14:04 localhost login[203]: pam_unix(login:session): session opened for user jussi by LOGIN(uid=0) <lurkki__>Aug 26 15:14:04 localhost login[203]: Cannot make/remove an entry for the specified session <nckx>gnu_srs: If you don't get an answer here consider (one of!) the help-guix@ or guix-devel@ mailing lists, the intersection ‘people familiar with Guix/Hurd’ and ‘people who happen to be on #guix’ might well be 0. <nckx>lurkki__: That sounds like dbus isn't running. Strange. <lurkki__>yeah, shpeherd log says it couldn't be started <lurkki__>i think this is enough info for a bug report <roptat>gnu_srs, I think our hurd port works (worked) from a debian hurd system, and there was an image of a guix hurd system at some point, but I don't remember how it went <gnu_srs>roptat: So Hurd packages are developed natively within a Debian GNU/Hurd image, not cross-compiled. <roptat>I didn't really follow the development of the hurd part of guix, but I think so <roptat>basically, it's just i586-gnu instead of i686-linux <roptat>maybe the bootstrap binaries were cross-compiled, but the rest is compiled natively <roptat>because a cross-compiled package is not the same as a natively-compiled package <g_bor>I am now back from vacation, so I can get involved in guix a bit more again. <g_bor>I posted two issues for guix-home-manager for discussion. <roptat>but I'm getting more and more issues with that install... libreoffice just doesn't want to start, audio is failing randomly... <roptat>and most software actually suppose they have write access to their configuration directory <roptat>or store unrelated stuff in there (I'm looking at you .config/guix/current :p) <g_bor>I also have noted that upstreams usually don't threat this as a problem. <g_bor>One of the most problematic ones that I encountered in more relaxed circumstances is cups writing status information into configurations files in /etc... <g_bor>roptat: also, do you have any indication how many actually writes to the config directory during normal use? <g_bor>Is that possible that some only check for write access, but usually does not write there? <roptat>don't know... hexchat will continuously write its window dimension, opened channels and networks in its config <roptat>minetest writes its game saves there <nckx>roptat: Saving session recovery information, cookies, other site-specific preferences & data, etc. <g_bor>roptat: would it make sense to map this, and provide symlinks to the invloved files/directories on an as needed basis? Also, I assume that the software involved does not fail nicely. Is that so? <g_bor>I mean that if for example icecat fails nicely to write history, (i.e. by simply not doing it), the it might make sense to make that a configurable option. Wdyt? <roptat>g_bor, it does make sense, I already do that for parts of hexchat <roptat>it writes its history to a separate directory in its .config <roptat>the history is configured as a symlink to a read-write directory outside of the home <g_bor>roptat: yes, that's what I was thinking about... <g_bor>roptat: do you think that upstream would be open to add an option to silence that warning? <roptat>they are hostile to moving the scrollback and log directories outside of the config directory <roptat>at least it's harmless, contrary to libreoffice which crashes, even with a symlink to a read-write directory <g_bor>I thought so, otherwise that would be a much better option :) <roptat>I have a lot of symlinks in my home configuration already <roptat>from ~/.config/foo to /data/roptat/.config/foo <g_bor>roptat: could you find out what libreoffice tries to do actually? <roptat>anything I use that I haven't written a home wrapper for ***lio is now known as lenn
<roptat>g_bor, there's absolutely no log, so I'll have to look at their huge code base :/ <g_bor>roptat: I will try to strace it later, to see if anything comes up. Does this happen on startup? <roptat>I had to start it with HOME=/tmp/lo libreoffice for it to work <g_bor>roptat: ok, I will try to have a look at that today, and come back with the results tomorrow. <g_bor>Do you think that we should open issues on the tracker as we encounter them? I believe that would spare us some duplicate work. <rekado_>gnu_srs: the bootstrap packages are cross-built. <rekado_>gnu_srs: they are then copied to a Debian GNU/Hurd system and used at the roots of a Guix installation there. <rekado_>unfortunately, at least one of the bootstrap binaries segfault. <rekado_>I tried to get help with diagnosing this on #hurd, but got no replies back then. <rekado_>so … right now I’d say that the Hurd port isn’t usable at all. <rekado_>I have a Debian GNU/Hurd VM with Guix on ci.guix.gnu.org, but since tar crashes it can’t build anything for us. <rekado_>my last message about this is <87d0ktewzg.fsf@elephly.net> on guix-devel@gnu.org. <rekado_>that was in May. No progress since then. <rekado_>the bootstrap binaries are statically linked; the next step is really to figure out why they segfault, fix that, rebuild them, and then move on. <rekado_>as soon as this problem is fixed we can have ci.guix.gnu.org build binaries for a Hurd system, so that people can use Debian GNU/Hurd with Guix as a package manager. <rekado_>When that works we can add the missing bits for the Guix System on the Hurd so that no Debian GNU/Hurd VM is needed any more. <lenn>Hey rekado_: I'm trying to write a pkg definition that downloads tar archive from github, extracts it and copies file into /usr/bin. I've figured out that building from source would be to hard for me... Do you know of any packages that already do that and I could take them as example? <jonsger>lenn: such packages won't end up in guix as we build everything from source (almost) <rekado_>lenn: note that the binaries might not work on Guix without patching them with patchelf. <lenn>ok, then I guess I'll have to learn the build process... <rekado_>lenn: you can use the gnu-build-system, remove the 'configure, 'build, and 'install phases. <rekado_>and then add a phase that creates the target directory and moves the files there. <rekado_>arguably this is more work than just building from source if the project uses a conventional build system. <jonsger>lenn: what are you trying to package? This community is usually quite helpful :) <lenn>I guess it uses go build system <lenn>Ok, tnx for now, once I have more written, I'll post it and ask for more help. <lenn>I think my next step is studyng the existing pkg definitions <lenn>there are also some holes in my knowledge of build systems (or maybe that's a hole in my knowledge of unix systems :) <lenn>sry, I wanted to say, that I'm willing to learn <lenn>Ok, gtg for now, have a nice day #guix <bandali>hi guix, what’s a good setup for using a different version of emacs (e.g. emacs 27 master) with guix? <bandali>i know there’s emacs-minimal which inherits from emacs, and is used for byte-compiling <bandali>i guess for best results i should redefine emacs (is that possible?) in a channel of mine? <poet>Hello, I'm trying to compile guix from source, and I'm getting the following error, "configure: error: The Guile bindings of GnuTLS are missing; please install them" <bandali>that probably means you’re missing the guile-tls dependency <nly>bandali you must have tried building emacs from source? <nly>git clone emacs, guix env emacs, make <poet>bandali: You're probably right, but how do I find and install this? <bandali>nly, i have, e.g. on debian. i was wondering what the best practice would be for doing it entirely using guix <bandali>nly, right, but i want to package it and use it instead of guix’s emacs 26.2 <nly>maybe one way would be to define a guix.scm in emacs.git/ <nly>with an emacs package, you'd copy from guix tree then tweak the source, wait let me grab an example <vagrantc>poet: you'll need to rebuild gnutls with the guile bindings enabled, as it was disabled in Debian some years ago <vagrantc>poet: there's a bug report asking for it to be re-enabled, and i filed a patch for it... <nly>bandali see nomad-local <poet>vagrantc: Ahh okay. So, what do I do? Download the gnutls source, the look for the proper config flags for guile? <vagrantc>poet: a patch is in the bug report; not sure if it still applies <nly>bandali this approach has the benefit that you won't be required to enter the hash again <bandali>nly, thanks. but for emacs specifically, i want to use emacs-xyz packages like emacs-org. i assume by default those get byte-compiled using emacs-minimal, which inherits from emacs. so, with that in mind, should i attempt to redefine emacs? <bandali>because i’d like all the emacs-* packages i install to be built with the right emacs version (i.e. emacs 27 master) <poet>Okay. I'll have a look and add a comment, then I'll try the patch. Thank you for your help, vagrantc <vagrantc>poet: working on adding guix to debian; that's the main blocker at this point <nly>ah right, you'll have to replace emacs-minima and emacs <nly>export guix_pkg_path='/path/to/emacs-custom.scm' <vagrantc>that, and i'm not sure gnu/packages/patches have proper licensing information ... :( <poet>vagrantc: You're trying to do that? <bandali>nly, i assume i could use channels instead of guix_pkg_path too, right? <vagrantc>poet: yes, i've uploaded all the other dependencies to debian already <nly>channels will be more work, butsure <poet>vagrantc: Oh wow, that's great! I love guix and I think it coexists quite nicely with apt. <poet>vagrantc: I hope you're successful. <nly>you definitely want to use that (source (local-file ...)) idiom <bandali>ok cool, i ask ’cause i already have a channel set up <bandali>thanks. i know next to nothing about that, i’ll look into it <nly>i am not very experienced with channels, but updating them is a bit too much work, maybe i am doing something wrong? :) <nly>sry, updating packages through channels* *vagrantc should learn how to support channels <vagrantc>i still do "./pre-inst-env guix ..." which often takes a long time to build guix when i start working on something new <bandali>ha, yeah it *is* a bit indirect: bump the hash in definition in channel, commit and push change to channel, update hash in ~/.config/guix/channels.scm (if you’ve pinned a channel to a hash), guix pull, guix upgrade <nly>i like ./pre-inst-env a lot lol <nly>do a make first, and it's pretty much instant i think <nckx>The workflow bandali describes is the minimum required to share your work with others, the point of channels, so it makes sense. For ‘local hacking’ pre-inst-env is great but also hacky. I use both all the time. <nckx>s/hacky/you promise to keep everything straight in your head, yourself/ 🙂 <nly>that's a great explanation, thanks <poet>I went ahead and used the shell script to install a binary copy, and now I'm getting these errors. How do I fix them? <FreeUser78>Hello guys; I need a bit help. I am trying add Windows to GRUB on my Guix, but I dont know the enough Guile. Any help? <rekado_>poet: have you installed any package yet? <rekado_>‘--list-generations’ cannot list anything when you haven’t ever installed anything with Guix yet. <quiliro>Hello. Would Guix be able to replace Ansible? <poet>rekado_: I haven't installed anything yet. I only ran --list-generations as a random command to show the errors for the screenshot. <poet>Right now I'm trying guix package -i glibc-utf8-locales and it's just hanging... <quiliro>poet: does that error come up with other commands as well? <rekado_>the error says that no profile exists. <quiliro>poet: maybe it has not hun...it could be working <rekado_>that’s expected when nothing has been installed <rekado_>all commands that act on profiles need profiles. <rekado_>I think the error message could be clearer, though. <poet>rekado_: The "warning: failed to install locale" warning is expected? <rekado_>it could say, for example, that the *default* profile at /var/… is missing. <poet>So how to fix the profile error? <quiliro>poet: that error is about oçyour locales not installed <rekado_>poet: for that see the Application Setup section in the manual. <rekado_>poet: you can’t do anything on a profile when you have no profile. <rekado_>you’ll get one as soon as you install something <poet>rekado_: Okay, thank you for your help. =) <quiliro>Would Guix be able to replace Ansible? <ngz``>Is there a tutorial of sort somewhere on how to debug the internals of Guix? How to use Emacs to that effect, etc.? <ngz``>quiliro: For some reason, I cannot make it work properly. <ngz``>No, but anytime I try to do something with it, I get an error: "geiser-repl--narrow-to-prompt: Wrong type argument: markerp, nil". <ngz``>And Invalid face reference: nil <ngz``>Besides I'm not really talking about emacs-guix, but more on inserting break points, debug functions, etc. <rekado_>quiliro: eventually ‘guix deploy’ might be able to replace Ansible — but only for Guix System. <rekado_>Ansible can be used with any distro. <rekado_>‘guix deploy’ only manages Guix System instances. <rekado_>it would be pretty useless on Fedora or Debian. ***ngz`` is now known as ngz
<poet>rekado_: When I was using the shell script to install guix, towards the end I got a prompt that went something like, "Is it okay to download pre-built binaries." I replied with "yes." Is there a way to change that? <nckx>poet: Not a foreign distro power user, but I think you can remove /etc/guix/acl and edit the guix-daemon service to remove any --substitute-urls= arguments. <nckx>Standard ‘make backups first’ disclaimer applies. <poet>nckx: I'll have to do some reading before I tinker with those, but I saved your suggestion to my notes. Thank you for your help. =) <nckx>poet: Why'd you change your mind? Building everything from source is 100% supported but very time-consuming. <vagrantc>don't you still need substitute servers to be able to download source tarballs in some cases? <quiliro>rekado_: Thank you... Guix does everything Ansible does and more (except being portable across package managers). So, why is Ansible available in Guix when it is not needed. <vagrantc>it can be used to set up non-guix machines <poet>nckx: You're exactly right: it's very time consuming indeed. I'd far prefer to install from source, but the truth is I'm just too lazy to fix or configure something every step of the way. <vagrantc>basically, the reason for anything in guix is typically someone had a need for it, i'm guessing <rekado_>quiliro: note also that I said that ‘guix deploy’ *might* *eventually* replace Ansible. <rekado_>and vagrantc’s point stands that Ansible can still be used to manage sites that don’t use Guix System. <nckx>vagrantc: IME, yes (although we should fix those when we encounter them). <nckx>vagrantc: Source files that are available only as substitutes. <vagrantc>so, if i'm getting a failure to run "make dist" on core-updates should that go to guix-help or guix-devel? or a bug? <nckx>vagrantc: I'd vote devel@. *vagrantc accepts the result of the vote <vagrantc>the short of it is something like... no rule to make target po/doc/guix-manual.pot <civodul>oh, did you run ./bootstrap initially? <tmssgn>how do you manage polkit? my original problem is that I just switched to networkmanager and now I can't add connections as my regular user... <nckx>tmssgn: I ‘manage’ that by running ‘sudo nmtui’, so… not really. <nckx>vagrantc: It would still be a bug, but have you tried just plain ‘make’ first? <tmssgn>hehe, that's what I had to do also. When I think about it, that's what I did with connman... through an alias... <roptat>civodul, would you like to reconfigure berlin? <nckx>My CPU is too busy to attempt reproducing anything on core-updates right now. <FreeUser78>I installed Python on Guix but "python" command not found. Any idea? <mcmx>python-wrapper makes python command point to python3 <mcmx>(if I understand it correctly) <vagrantc>this sounds suspiciously like what was trying to solve in 81824af6a4b30368ed497686976c26f742bc6c8a build: 'guix-manual.*.po' no longer depends on 'guix-manual.pot'. <nckx>mcmx: Your understanding is correct. <rekado_>roptat: I’m actually reconfiguring berlin right now. <rekado_>roptat: I’m only doing this for the split cuirass service <rekado_>is there something else that requires reconfiguration? <FreeUser78>mcmx: thank you, now works perfectly. (I installed python-wrapper) <civodul>rekado_: changes to guix-maintenance.git regarding web site generation <mcmx>don't thank me, thank the Guix developers dedicating their time to this project :) <civodul>i had almost forgotten about the split Cuirass service, thanks rekado_! <FreeUser78>when I try install new pip from python, it say: "Read-only file system: '/gnu/store/..'. I executed the commands with sudo <rekado_>I’m still waiting for ‘guix pull’ to finish *rekado_ runs ‘guix gc’ on a few build nodes *rekado_ updated the copy of maintenance.git, reconfigures berlin now <FreeUser78>Could not install packages due to an EnvironmentError: [Errno 30] Read-only file system: <rekado_>this is odd: error: static-web-site-service-type: unbound variable <rekado_>hint: Did you forget `(use-modules (sysadmin web))'? <FreeUser78>any idea? sorry for the repost, I write complete message <rekado_>FreeUser78: pip tries to write to /gnu. It shouldn’t because it’s read-only by design. <rekado_>you’d have to tell it to install stuff elsewhere. <vagrantc>so, to be clear... to generate a tarball from clean git checkout: ./bootstrap && ./configure && make dist" <vagrantc>so, to be clear... to generate a tarball from clean git checkout: "./bootstrap && ./configure --loclstatedir=/var && make dist" should generate a tarball? <ngz>Just to be sure : now, when fixing a typo in the manual, I assume we may have to fix it in every translated manual, right? <rekado_>FreeUser78: I don’t know how to use pip, sorry. <FreeUser78>rekado_: then, how I could enable write on /gnu? <rekado_>roptat, civodul: any idea why ‘guix system reconfigure -L modules berlin.scm’ insists that (sysadmin web) isn’t loaded? <rekado_>FreeUser78: it’s read-only by design. <rekado_>nothing but the guix-daemon should ever write to /gnu. <civodul>rekado_: probably something's wrong with (sysadmin web)? <rekado_>I’ll reconfigure tomorrow unless someone beats me to it. <civodul>modules/sysadmin/web.scm:72:40: missing '(' in vector or array literal <ngz>Hmmm, it seems that I was wrong. <FreeUser78>rekado_: but on /gnu exist a pip installed but not updated. For to update it, I need write the new pip <FreeUser78>But it is done with pip instead of Guix package, I think that is the problem. *civodul reconfigures berlin <quiliro>FreeUser78: Why don't you create a new guix package for that python package so you can have an updated verasion? <quiliro>FreeUser78: it is not too difficult....you have to learn to do it though <ngz>FreeUser78: what is your package? <roptat>FreeUser78, you can create a pyenv with a newer pip maybe? <roptat>I think that's the recommended way anyway <roptat>well, from python people who don't know guix (yet) :) <roptat>also, you can probably use --with-source=pip=<a path to a downloaded pip source tarball> <ngz>hmmm "requires pytest-3.8.2, actual pytest-3.8.0". Is core-updates frozen? If so, when should a python-pytest update go? <ngz>1950 dependant packages <pkill9>roptat: i dont think --with-source uses that replacement syntax (=package=source), it just parses the source filename to work out which package to replace, i think it should use the replacement syntax in your example though, as it would be consistent with the syntax in other commands, and would let you use a source url that isnt of the form package-version.ext (which means not having to download it and change the name of <ngz>FreeUser78: I updated pip to 19.2.1. <civodul>ooooops, "herd restart mcron" crashed berlin <quiliro>FreeUser78: remember to use channels...the other method i do not think works any more...search channels on that web page 