<buenouanq>what's the proper guix way of dealing with scripts that have explicit paths (such as /usr/bin/perl)? <davexunit>buenouanq: patch them to refer to a binary in /gnu/store instead <S1ohy>qemu-system-x86_64 -m 1024 -smp 1 -net default -net nic,model=virtio -boot menu=on -drive file=guixsd -drive file=guixsd-usb-install-0.12.0.x86_64-linux <S1ohy>qemu-system-x86_64: -net default: Invalid parameter 'default' <S1ohy>there are two -net switches in the command. <S1ohy>without -net default, it seems wot work <S1ohy>so far anyway, the vm is booting up now. <S1ohy>I had to hit escape though, not f12 <Apteryx>Does anyone understand how dconf works? I have this package, GnuCash, which doesn't save settings unless I install dconf in my user profile. <Apteryx>Can I do something about that in its guix package definition? <Apteryx>And even then; I get the error: WARN <dconf> Unable to open /usr/local/share/dconf/profile/user: Permission denied <Apteryx>Followed by: WARN <dconf> failed to commit changes to dconf: Failed to execute child process "dbus-launch" (No such file or directory) <Apteryx>Are propagated-inputs == run-time dependencies? <cbaines>Apteryx, not quite, propagated inputs are runtime dependencies that are not referenced directly by the package, whereas inputs are also runtime dependencies, but the package has to directly reference them <cbaines>That is how I understand it at least <Apteryx>cbaines: OK. So propagated-inputs are for when I can't figure out how to make substitute* rules work? <cbaines>Looking at a few gnome packages, they all propagate dconf <Apteryx>I got GnuCash prefs working by adding dconf & dbus to my user profile. I guess propagated-inputs would work. <clacke[m]>propagated inputs will be in your profile if you install the package, that's in concrete terms what they do. installing yours will imply installing them. <Apteryx>clacke[m]: Kind of how it works on regular binary distribution such as Debian? <clacke[m]>so you get the potential drawbacks of deps hell etc <Apteryx>OK! I think I understand it better now. And also I see how using substitute* & inputs fits the functional model much better. <clacke[m]>where guix goes "arbitrarily picking one" on the console <Apteryx>I've got lots of those, caused by by mime-xdg something package and another info-xdg or something. Is there a command to print the list of collisions in my profile? <clacke[m]>not that I'm aware. glad I could help. now afk. <Apteryx>Any idea how I can fix a locale problem? The tests of GnuCash apparently fail because of missing locales (en_US, en_GB, and fr_FR). <Apteryx>The funny thing is that when using "guix environment --pure gnucash" and running the test I cannot reproduce. It only happen when using "guix build gnucash". <EvilPython>I want full disk encryption with three partitions. <marusich>EvilPython, support for LVM is currently missing. <marusich>See the manual, section titled "Limitations" <marusich>It describes that and other current limitations. <marusich>EvilPython, I'm pretty sure everyone would love to have LVM! <marusich>Since LVM support is currently missing in GuixSD, I don't think that's possible at this time. <marusich>It's easy to make an encrypted partition, though. You could encrypt your entire file system in a single partition, for instance. That's what I do. But using three partitions with LVM is not possible, since LVM is not currently supported. Hopefully someone will add it soon. <marusich>If you put / and /home in the same partition, which is LUKS-encrypted, and if you use a swap file contained in that file system, then you will get that. <marusich>Perhaps that is good enough for your use case for now? In the future, I hope we get LVM support, too, but today is isn't there. <marusich>I'm not sure why it hasn't been added yet. If you search the email list archive, I'm sure you'll find interesting discussions. <marusich>One reason I can think of is that, I believe, we cannot currently express many possible configurations for LVM very easily in an operating system configuration file. <marusich>If you search the email list archive for guix-devel, you'll find some posts about hurd. Someone is actively working on getting GNU/Hurd GuixSD working. <marusich>Going back to LVM, basically it's rather complicated to express the many ways in which you can configure the devices: when device mapping (e.g. LUKS encryption), LVM, and RAID are all in play, what is the best way to represent all the possible configurations you can make? <marusich>My guess is that when we add support for LVM, we will have a decent answer to questions like that. <EvilPython>And I don't want Guix because it's GNU But I want newest packages. I want to build my own distro. <EvilPython>I meant I want only newest packages (any free distro). <marusich>EvilPython, I can understand that. If you rely on the package definitions that come with Guix, then (like with any distro) you rely on the maintainers and contributers to keep the packages up to date. <marusich>For a lot of cases, that's fine. However, Guix makes it very easy to keep your own package definitions, too - so if you care very much about a package, you could either submit a patch to guix-devel@ to update it from time to time (usually this is just a 1- or 2- line change), or you can even store your own package definition in a directory and put that directory on the GUIX_PACKAGE_PATH (see "Package Modules" in the manual for more). <marusich>Apteryx, are you familiar with the locale issues mentioned in the "Locales" section of the manual? If not, it's worth a look. Chances are your locale woes are explained there. You might also search the email list for similar discussions; many email threads exist discussing locale issues and fixes. <rekado>EvilPython: what do you mean by “I don't want Guix because it's GNU” <csanchezdll>how can I test cross compiled guile on my PPC hardware? <csanchezdll>is there a way to make a tarball of all the target-required derivations? <rekado>csanchezdll: if this was built with Guix, could you copy all store items indicated by “guix gc --references”? <andrss>is there a way to forcibly reinstall a package? <andrss>libice seems to be broken, anytime i try to install something new or reconfigure system guix complains about '/gnu/store/...libice../' not being found ***ShalokShalom_ is now known as ShalokShalom
<adfeno>andrss: You could go/reverto/roll-back to a previous package generation (or system generation, if you're using GuixSD instead) where the package wasn't present... <adfeno>... Make sure that no other generation includes the package, then do guix gc, and try reinstalling. ***kelsoo1 is now known as kelsoo
<rekado>andrss: hav you modified /gnu/store manually? <andrss>rekado: i didn't re-format root partition before doing `system init`, (i hoped it would reuse already downloaded packages and save some traffic) <andrss>adfeno: i have only 1 generation for some reason <ng0>adfeno: I just realized.. we potentially need the qtwebengine to run kde in the future, if I understood you correctly (I still have to catch up with the thread on libreplanet or what it was). Current version of KDE does not reqire it (at least last version of KDE Plasma I ran, 1 year ago). <ng0>so it's a big thing.. I don't know if I want to take on the process of packaging yet another big thing which takes a long time. It's not super important for my own tasks, just a "nice to have". <ng0>I did some initial writing process though <erliphant>Hi - is there a way to export a package and it's dependencies. guix archive -r doesn't actually seems to do quite the right thing? i.e it doesn't include unlinked dependencies? <ng0>the docker thing and what Petter (?) will talk about at fossdem seems to be comparable to what you mean? <ng0>i call it the "docker thing" because I only read the thread briefly, not in detail. there's a way to (or being developed to) export to docker images <erliphant>@ng0 thanks.. I'll try to google for that. That's pretty much exactly what I want. <ng0>search on "guix export docker" should bring you some right threads <ng0>guessing by subjects: <ng0>guix archive -f docker ... in case that's merged already <ng0>there's no -f here, so I guess that's just to be discussed <ng0>or was it --format=docker , not just -f <ng0>i don't use archive yet, that's why my completions aren't complete there and I would have to read the documentation to tell you more <ng0>i know i just have sent it tuesday, but this is trivial and requires next to no QA: can someone merge: Tue. 11:51 [2/2] Mailmap: Update to new address <nckx>ng0: OK. You've quite the collection of burnt e-mail addresses :-) <ng0>first I thought domains are good. now I've fallen back to just having this old account where I'm moving every back to and just donate to the collective running the servers instead of paying for names <ng0>you can be sure I will this time definitely stick with this one, as this account is like, 10 years old or more <nckx>ng0: I don't mind. I was just curious why you were ‘getting rid of your domains’, whether political or practical. <ng0>1) poor end of income, didn't expect that with some of the more expensive ones (n0.is), 2) I don't think paying for names is okay, but I did for a couple of years, 3) I don't run anything for some time now which would require a name <rekado>erliphant: you can find the Docker stuff in the manual by looking up “docker” in the index. <ng0>4) and the one meta-collection of future purposes which require names I have can happily sit on a subdomain of a shared community-financed server <nckx>ng0: Thanks for the explanation. I'm ambivalent about (2) myself. Don't know anyone who's subdomain I'd trust more than my current TLD, unfortunately. <ng0>hm.. an application with lib64 instead of lib. should I simply disable the 64bit build if possible? I see not many applications with lib64 instead of lib here, or am I wrong? <nckx>ng0: I'd just install everything into /lib, if that works. <nckx>That's what several other packages do. <ng0>there's a slight chance that this is also the reason why I have to wrap the lib path of mbedtls for this application (hiawatha) <davexunit>lib64 is the "multilib" thing that other distros do <nckx> /lib64 is just... ugly. <jmd>After having done guix packages --upgrade, texlive seems to be broken :( <ng0>exactly. because lvm is there. <ng0>for GuixSD.. this is still work in progress and not natively supported so far <ng0>what I mean is, booting GuixSD from an lvm+luks device <ng0>feel free to search the open bugs, I have no time to answer now <ng0>the first progress last year was to get plain cryptdisk support <rekado>GNUnymous_IRC: lvm works. You just can’t boot from it yet. <andrss`>rekado: i booted from usb image and did `system init` again, it resolved the issue with libICE <ng0>that's something else. <ng0>haha. davexunit the problem was I forgot to set -DUSE_SYSTEM_MBEDTLS <GNUnymous_IRC>Can I setup GuixSD if I will use LVM for root and separate partition for /boot? <jmd>GNUnymous_IRC: I think GPT partitions are not supported at present. <adfeno>GuixSD is making fast progress, but still has long way ahead :D <adfeno>What you could do is: use Guix ontop of Trisquel for the momment. <jmd>echo '^G' will give you a bell. <jmd>Does your terminal support it? <GNUnymous_IRC>I can beep in my non-free BIOS. I don't know how to reverse-engineer. :( <jmd>I didnt know we have a Live USB ***jonsger1 is now known as jonsger
<jmd>what does echo $TERM tell you? <adfeno>GNUnymous_IRC: Try this command: printf '\\a' <jmd>Maybe your speaker is broken / switched off. <adfeno>If I recall correctly, the GuixSD image comes with sound disabled, unless you have BIOS beep for you. <jmd>Well if there's no speaker then you wont hear anything. <jmd>You want to emulate a speaker in software ? <jmd>I think xterm has a "visual bell feature". <nckx>Random musing after civodul's wrapped in my editor: will Guix's comma-separated copyright years expand bravely into the future, or can we legally use ranges instead? <jmd>nckx: There was a edict from the FSF some time ago. <ng0>I think at some point ranges can be used.. ? <ng0>at least gnunet with the new copyright assignment switched headers to $year – $year <nckx>‘Why bother’ being a perfectly valid one :-) <ng0>there are mixed lines. <ng0>Copyright (C) 2011-2014, 2016, 2017 GNUnet e.V. and sometimes 2007-2017, but mostly year,year,year <nckx>ng0: As in 2012-2014, 2016-2017? Yes, that's required if 2015 saw no changes. <ng0>no, as in exactly what I wrote <ng0>anyway it will be legally ok <nckx>After I sent that message, yes. Anyway, IA so very much NAL but am pretty sure mine is (purely by coincidence) equivalent. <nckx>It's OK if ‘you make an explicit statement in a README file about this usage’, according to the above gnu.org link. <nckx>My mother wouldn't let me. <ng0>oh. mbedtls should be renamed to mbedtls-apache. I have no idea how it differs to the -gpl one, but hiawatha (probably) needs the one I packaged. <ng0>I have hiawatha ready for review and will send the update now <ng0>now/in a few minutes <GNUnymous_IRC>Can I install it on LVM with separate /boot partition outside LVM? <clacke[m]>exactly what I wanted to ask. thank you. :-) <nliadm>ng0: I think it's just dual-licensed? not very sure, though <ng0>they have to tarballs <ng0>maybe one has an openssl exception or something similar <ng0>do we have "mutt" users? I need someone who tests notmuch-neo or whatever the name was beyond just launching it (which is what I did) <ng0>wrong again. neomutt <ng0> January 19 [2/2] contact.ng0@crypt... Add: Neomutt. (guix guix::patch) <phant0mas>efraim: my aarch64 machine is currently building your branch :) <ng0>addition to the neomutt thing: according to past threads a problem was that it segfaults when new emails are added, but you have to keep in mind that back then I used a config which only worked with the frankenmutt from Gentoo. <Apteryx>What's the closest thing I can get to a "build" environment? "guix environment --pure" ? <Apteryx>Tests are failing when building the package, but passing inside a guix environment... So I'm a bit puzzled. <Apteryx>Hmm, I used "guix environment --container gnucash", and now running "make check" I get errors such as "libguile.h: No such file or directory". Ideas? <davexunit>you'll need to inspect environment variables and stuff <Apteryx>(My CWD is set to that of the failed gnucash build -- guix build -K gnucash) <GNUnymous_IRC>Can I install GuixSD on LVM with separate /boot partition outside LVM? <Apteryx>GNUnymous_IRC: rekado has said that he is using such a setup. And he pointed you to an archived message on the mailing list where he detailed how he went to do that. <ng0>GNUnymous_IRC: see topic, url of this irc log <Apteryx>GNUnymous_IRC: Sorry, can't find it anymore. Maybe I dreamed it. <ng0>Well I gave a search result link for LVM earlier with not that many messages. <Apteryx>davexunit: OK (regarding inspecting environment variables and stuff). I've already at least discovered that in the container I only have the C & POSIX locales listed (locale -a). <Apteryx>While in a pure environment I see the full list of %default-locale-definitions. <Apteryx>That probably explains why the locale sensitive tests pass in the environment but not at build time. <Apteryx>But looking at $GUIX_LOCPATH, it points to the same store location in both the pure environment and the container, so I'm not sure what differs. <andrss>I've installed GuixSD inside VM to a physical drive. But it's bootable only from VM, not from the psycical machine. Why it could be so? <andrss>Also, GRUB installation always fails for some reason, but succeds with --force option. So far it works from QEMU and VirtualBox, on physical machine it just skips to next drive druing boot. <andrss>hm, I didn't mean guix vm image. I basically attached physical disk to VM and installed GuixSD on it. <andrss>I haven't used `build vm-image` for that. <andrss>I've tried to re-install grub from my physical machine, and VM crashes during boot after that <Apteryx>davexunit: I could reproduce bug in the container :) Silly me wasn't using the same drv-* in my /tmp, which was not built as far as the one I tried in the environment. <jmi2k_>I can't run glutin examples, it fails with 'thread 'main' panicked at 'called `Option::unwrap()` on a `None` value', src/libcore/option.rs:323'. Running strace reveals the problem is that it can't find libX11.so. Any suggestions? <Apteryx>So now I'm sure it's about adding locales to the package for tests purposes... Not sure how to do that yet. <Apteryx>jmi2k_: There is a libx11 package... Might help? <nliadm>random question: is it safe to run a gc while I've got a guix environment shell running? <davexunit>jmi2k_: sounds the package needs to rebuilt and patched to refer to the absolute path of libx11 in /gnu/store <davexunit>the things in the environment will be protected from GC for the duration of the sessions <jmi2k_>davexunit: it's not a package, it's just a cloned glutin repository <davexunit>you need to setup an environment where libx11 can be found <jmi2k_>guix environment --ad-hoc libx11 is enough? <davexunit>if you run that, you could try setting LD_LIBRARY_PATH=$GUIX_ENVIRONMENT/lib and see if it works <jmi2k_>Ok, I'm trying... What is the right way to setup an environment for a project, if there is any? <davexunit>it just boils down to installing the right software and setting the right environment variables <davexunit>sometimes upstream software is a pain, like glutin looks to be <nliadm>I do `[ -n "$GUIX_ENVIRONMENT ] && export C_INCLUDE_PATH=$GUIX_ENVIRONMENT/include/ LIBRARY_PATH=$GUIX_ENVIRONMENT/lib ` in my .bashrc <jmi2k_>nliadm: Thanks for that, seems very useful <Apteryx>Does anyone knows which package provides the "locale" util? <nliadm>on my fedora system it's provided by the glibc package <nliadm>maybe if you explicitly pull in glibc you'll have it? <Apteryx>nliadm: OK! I'll investigate. Thanks! <nliadm>are the locale packages just generated locales, then? <andrss>GNUnymous_IRC: it can, look for Mapping devices section of documentation <Apteryx>nliadm: What do you mean by generated locales? <Apteryx>I've tried: "guix environment --container --ad-hoc glibc-utf8-locales", then "export LOCPATH=/gnu/store/bhj5xdwwd2dg770lmlks7hyny8vzjm9x-glibc-utf8-locales-2.24/lib/locale/", then "locale -a": still only C & POSIX locales found. Hmph. <nliadm>Apteryx: the form used by glibc, made be localedef(1) <nliadm>Apteryx: I think read localedef(1), you might be able to make the locales you need? <nliadm>all my machines only have en_US.UTF-8 and C <andrss>when I try to jump to a definition in emacs, it asks for TAG file. I've installed guile, geiser and guix. What is missing? <andrss>GNUnymous_IRC: you will probably need efi-enabled GRUB <alezost>andrss: how do you try to do it (what key?), is it from a .scm buffer? <andrss>alezost: M-.. It's Scheme buffer <andrss>I've figured that I need geiser-mode inside such buffers <andrss>But stiil I cannot jump to definition of a thing, only to a module file <alezost>andrss: at firest you need to "M-x run-guile", then you need to use the current module to be able to jump to definitions <avoine>andrss: I'm not sure but I think jumps only work for imported modules <andrss>run-guile opens a *Guile REPL* buffer, run-geiser does the same. After both of this actions it keeps asking for TAGS file. <andrss>When I do `geiser-mode` it stops asking for TAGS and just says "Cannot find edit location for ..." <andrss>oh, it actulally just worked for `source`, but doesnt for `package` or anything else <andrss>Should I even expect some elisp-level ability to jump between definitions? <andrss>Sometimes it jumps to (define..), sometimes just inside a file. But most of the times it just doesn't work <alezost>andrss: it's not geiser that asks you about TAGS, it's some default emacs command (probably 'xref-find-definitions'). When you enable geiser-mode (it should be enabled by default btw), you begin to use 'geiser-edit-symbol-at-point' command <alezost>you can either compile current buffer with "C-c C-k" or run ",use(foo bar)" in the repl to be able to jump to definitions in the current module <alezost>I'm not sure about "C-c C-k" though as I don't use it <andrss>but only for package definitions <andrss>can I jump to defitions of macros and other things? <alezost>andrss: yes, what macros and other things do you mean? <andrss>like (bootloader ...) seems to be a macro, but i cant jump to it definition <andrss>I see, it cannot jump to data field definition, but can to record type definition. <alezost>right, bootloader is a field of 'operating-system' record; it's not a defininition, so you can't jump to it <nckx>sneek: later tell ng0: I've sent my zpaq patch. Sorry for the delay. <Apteryx_>Emacs (ubuntu system) is not honoring my INFOPATH which contains ~/src/guix/doc. It doesn't know about the guix manual... the Emacs' Info-directory-list seems to contain the correct values of my INFOPATH. Standalone info from the terminal picks it up. <adfeno>It only works if I do: sudo info guix <Apteryx_>adfeno: In my case "info guix" works (I have this in my profile: export INFOPATH="$HOME/src/guix/doc:$HOME/.guix-profile/share/info${INFOPATH:+:}$INFOPATH"). <Apteryx_>What isn't as expected is that the top level directory doesn't contain a Guix entry for the manual. <Apteryx_>I'm no info expert, so that maybe normal. Maybe there needs to be something in those directories to make a menu entry appear in info's top node. <Apteryx_>Oh, apparently info files have to be "installed". We learn something everyday. The folders which INFOPATH points to must contain some "directory node" file to which info files can be added using the "install-info" command. <rekado>Apteryx_ you can open an info document at an unusal location with “C-u C-h i”.