IRC channel logs


back to list of logs

<mekeor>what does "guix lint" mean with "the source file name should contain the package name"?
<luser1>haha lfam
<lfam>mekeor: The file returned by `guix build --source` should include the name of the package in its filename
<lfam>You can change it with the file-name procedure
<mekeor>lfam: ah, i guess it's referring to the file-name ... yeah, okay, ait
<mekeor>thanks :)
<luser1>Huh, I'm surprised by the amount of people in who say GuixSD is difficult to install or to quote chris (dont know his nick) "gentoo for adults".
<luser1>I found it incredibly easy.
<luser1>Though I couldn't get luks to work on my ssd :/
<jonsger>luser1: installing GuixSD wasn't really complicated (I'm Mint user^^) and it is well documented. but the "configuration" didn't worked for me :(
<luser1>jonsger: What do you mean by the last sentence?
<luser1>Like, your config.scm isn't/didn't do what you expected? or?
<jonsger>no the things which are described in config.scm works fine but the "rest"
<luser1>Could you give an example?
<jonsger>it's two months ago. I used it directly on a thinkpad. i think I had no wlan and I didn't got the keyboard to german
<mekeor>jonsger: keyboard layout is just (console-keymap-service "de")
<mekeor>jonsger: many wifi drivers are not free software. that's why i bought a wifi dongle which uses ath9k_htc which is a free driver
<luser1>It's great primarily being on ethernet
<mekeor>luser1: is your config.scm public? i ask because i like reading other peoples configs :)
<luser1>mekeor: Mine is so trivial it's not worth sharing.
<jonsger>yes. I know. I don't want to discuss this now. I wanted only to say that the "installation process" of GuixSD is quite easy, so that you have a "working" system with a shell
<luser1>Essentially default with the tor service.
<luser1>I'm waiting to get some reasons to start hacking it.
<mekeor>not only a working system with a shell but as well desktop environments like gnome are super easy to configure. it's just a single service: (gnome-desktop-service)
<mekeor>luser1: haha i see you :)
<luser1>mekeor: Why make gnome system wide?
<mekeor>luser1: i don't know. how to make it only for 1 user?
<luser1>I suppose if you're on a system with actual multiple users, it makes sense.
<luser1>mekeor: guix -s install gnome ?
<luser1>Or whatever it is?
<mekeor>guix package -i gnome
<mekeor>uhm, yeah
<mekeor>i think there are no user-specific services at the moment, AFAIK? i'm not sure
<mekeor>installing gnome different than using the service, i think
<luser1>It's like, why install i3wm for every user if I'm the only user on the system
<mekeor>s/different/is different/
<mekeor>yeah, i see you
<jonsger>oh yes. I forget that I runned xfce :)
<mekeor>yeah, i think there are (xfce-desktop-service) and (gnome-desktop-service) but no other service for the big DE yet
<lfam>Yeah, but you can add %desktop-services and the package of i3, and i3 will appear at the login manager's list of environments to choose from
<luser1>For some reason, I never see any choices.
<mekeor>luser1: press F1
<luser1>Even though I have i3 and that other default
<lfam>i3 is not really a desktop environment with a complicated web of service dependencies so I didn't see a need to make a service for it
<luser1>mekeor: I do, and nothing happens.
<lfam>luser1: Is i3 in the global package list?
<luser1>So, I had to add i3 to my xsession
<luser1>lfam: It's installed on my user account.
<lfam>luser1: To make i3 available in the graphical login manager, you need to make it available globally in the OS configuration file
<luser1>What about a mechanism for when you type a users name, then you get a list of environment available to that user before entering in the password.
<luser1>Would that be possible to implement?
<lfam>Perhaps? I don't use the big DEs because I don't understand them, so personally I don't know the answer. You might ask on <>
<luser1>I don't either, I use i3wm.
<mekeor>my issue is that the login manager (slim?) always just launches my ~/.xsession no matter which DE i chose...
<luser1>Yeah, I was talking about slim.
<luser1>I just put exec /home/luser/.guix-profile/bin/i3
<luser1>In xsession.
<mekeor>luser1: ah, okay, i see, so you have the same issue. got you :)
<mekeor>luser1: so, i've got gnome and xmonad installed globally. i can switch between them using F1 (i.e. the currently active choice is shown on the bottom of the screen) but the switch doesn't take effect.
<mekeor>luser1: i'd prefer to be able to successfully switch between (1) gnome, (2) (plain) xmonad, and (3) xsession
<luser1>That'd be nice.
<lfam>mekeor: Selecting the DE doesn't work?
<mekeor>luser1: so you have the same issue, right?
<lfam>If so, please file a bug by sending a message to <>. Please include your system configuration file so we can try to reproduce
<mekeor>luser1: do you also have a custom call of slim-service in your system configuration file?
<mekeor>i do so: -- and i wonder if that's the reason
<luser1>mekeor: My issue is that pressing F1 doesn't give me any options.
<luser1>So, I have to edit my xsession to do what I want (use i3).
<mekeor>luser1: that's probably because there are no options?
<luser1>I also don't have any calls to slime in my config
<mekeor>luser1: well, as lfam said you have to install i3 globally then
<luser1>So, I don't think my case is a bug or anything.
<lfam>Is there a reason you are using %base-services instead of %desktop-services?
<mekeor>lfam: because i wanted to replace wicd with network-manager
<lfam>I see
<lfam>mekeor: I'm confused about the version field in your dzen patch. I don't see any version tags (or any tags at all) in the dzen Git repo
<lfam>I guess it's defined internally
<mekeor>so this version is 0.9.5-svn but i thought the 'svn' part is redundant
<lfam>mekeor: What text editor are you using?
<lfam>I'm wondering if we can find a tool to help with indentation
<lfam>Too bad about the hand-written Makefile that has to be patched :(
<mekeor>lfam: i use emacs, too, but i guess i have a special taste of indentation, so i do it manually
<mekeor>i'm sorry for this. i should probably just have sticked to the commonly used style
<lfam>mekeor: There is a script 'etc/indent-code.el' that will indent the code to our standard.
<lfam>That's in our Git repo
<mekeor>ah cool
<mekeor>lfam: will it also make the newlines correct?
<lfam>I'm not sure, I haven't tried it
<mekeor>i'll try it
<lfam>I replied with some other comments
<mekeor>thank you, lfam
<lfam>mekeor: Maybe I should have explained more about using (assoc-ref). Feel free to ask for guidance
<lfam>I'm trying to install GuixSD on this thing:
<lfam>A headless board with a serial console
<lfam>I made an installer image with a GRUB configured to use the console and passed a few extra kernel-arguments ("console=ttyS0,115200n8 gfxpayload=text").
<lfam>I see the kernel messages, and the Shepherd output. But for each of the console-font-tty* services, there are some errors, and I never see the login prompt:
<lfam>Actually, those errors proceed from the term-tty* services, not the console-font-tty* services
<lfam>I guess those errors are not the problem. I can work around by sleeping during the initialization as shown here:
<lfam>But, I still don't see the MOTD or the login prompt
<lfam>Ah, I don't think that mingetty supports the serial TTYs
<civodul>Hello Guix!
<CharlieBrown>Hello, civodul.
<CharlieBrown>And I'm not a Guique...
<roelj>Hello civodul!
<roelj>Which package contains the gnutls module that GNU Guix uses? That would be 'gnutls', right?
<slyfox>roelj: yes
<slyfox>$ find /gnu/store/ -name gnutls.scm
<roelj>Then I seriously don't understand why my guix cannot download from https websites. 'guix download ...' works fine, but 'guix package -i ...' does not. So I assume guix-daemon handles the download of the install command.. So I run then both right after each other (./pre-inst-env guix-daemon ... ; ./pre-inst-env guix package -i ...).
<roelj>Still no luck
<civodul>roelj: guix-daemon now performs downloads on behalf of users
<civodul>so (gnutls) must be visible to guix-daemon, not just to the client tools etc.
<civodul>roughly that means you must set GUILE_LOAD_PATH appropriately in the environment *of guix-daemon*
<civodul>that happens automatically on GuixSD and with the binary install, but otherwise it's something to pay attention to
<roelj>civodul: Right, that's what I suspected already, so I set those environment variables
<roelj>Ah, I so the first entry in the GUILE_LOAD_PATH is set to /root/.guix-profile/...
<roelj>Does it download as a guixbuilder user?
<roelj>That's it..!
<civodul>roelj: cool!
<civodul>search paths are difficult
<roelj>civodul: Well, I've tangled myself in a very restrictive Guix environment.. (just for testing purposes).
<rekado_>Hi Guix!
<civodul>hi rekado_!
<roelj>This shows that Guix is pretty safe in this sense too.. Downloads are not downloaded as the root user..
<roelj>hi rekado_!
<wingo>greets guix :)
<roelj>look who's there! The great wingo :)
<phant0mas>so there is an interested bug in findutils
<phant0mas>in linux x86_64 test "test-lock" freezes indefinetely while being built from guix
<phant0mas>if I run the same test as a user it doesn't freeze
<civodul>hey phant0mas
<civodul>phant0mas: if the test froze on GNU/Linux, we'd know it, no? :-)
<civodul>or are you testing in a different setting?
<phant0mas>I am on a 64 bit linux machine, has it happened on anyone else? :/
<phant0mas>civodul: I am testing on the same build environment
<phant0mas>the only difference is the user running the test
<phant0mas>can someone try building findutils from core-updates?
<civodul>rekado_, roelj: on combining Nix and EasyBuild, as Kenneth mentioned in the HPC track:
<civodul>phant0mas: so this is core-updates unmodified?
<civodul>hydra built it, though no the latest commit:
<clacke[m]>re: CVMFS-Nix-... "Ensures all paths are rpath’ed (technically: runpath, so LD_LIBRARY_PATH takes precedence)" This can't be true, can it?
<clacke[m]>Or is this a local quirk in their setup?
<rekado_>I'm so annoyed by "academic only" software :(
<CharlieBrown>rekado_: What's that?
<rekado_>CharlieBrown: software that comes with a license that only permits academic use.
<rekado_>I encounter this regularly
<rekado_>there are also many free software projects that depend on tools that have these restrictive licenses, which means that I can't add them to Guix upstream.
<CharlieBrown>rekado_: Ouch.
<rekado_>some are really hairy because the license declaration is spread across different locations
<rekado_>e.g. this is academic-only by default unless it's listed at this URL as free for all uses.
<CharlieBrown>Can't get
<phant0mas>civodul: yes
<phant0mas>it the latest core-updates unmodified
<CharlieBrown>rekado_: Could not --verify.
<CharlieBrown>rekado_: nvm, dl'd old version because i didnt see old ver in ftp listing
<phant0mas>civodul: the machine which the problem appears is running on a amd cpu. I reproduced the build on a intel based system and the error doesn't appear
<phant0mas>maybe there is something wrong with my system
<roelj>civodul: Re EasyBuild/Nix mix: I don't see what the point of EasyBuild is here.. it looks like they think EasyBuild software runs faster
<roelj>civodul: The only thing that can differ there is the compiler options.. right?
<rekado_>roelj: Initially, I thought they use Nix to provide some sort of "runtime profile" on top of which they use EasyBuild, but the slide "Nix or EasyBuild?" suggests otherwise.
<rekado_>(happy to see that guile is installed there)
<civodul>yeah, i'm unclear about this
<civodul>there's also this project mentioned at the end:
<roelj>When I tried EasyBuild, it seemed to have a lot of different versions for a single program (samtools)
<civodul>and the README is also light on details
<roelj>All differing in architecture-specifics
<civodul>hey, nckx!
<civodul>yeah, dunno
<roelj>So, I think they do (a lot of) performance tuning on those recipes. Maybe we could investigate whether there are going to be huge differences..
<civodul>there's also the talk that goes with the slides:
<civodul>maybe i should watch it to understand
<roelj>I recall that Pjotr observed a 50% speed-up of "sambamba" when compiled with -O2, so the potential gain can be quite big
<nckx>civodul: Yo.
<thomasd>civodul, roelj: i'm listening to that talk now. so far, sounds like they use Nix mainly just as a portable package repository.
<thomasd>as to why use easybuilds: so far the speaker has remarked that Nix is not geared towards intel compilers, and that many "scientific packages" are not part of it.
<rekado_>(Guix has many scientific packages.)
<thomasd>yes, I didn't get what was meant by that
<thomasd>but that's what he said.
<thomasd>ah, sorry, I misread. So does Nix not have these packages?
<roelj>I start to believe it's not about the packages, but more about the 'optimization' of those packages.. Using the Intel compiler would make the program run faster (but I haven't seen anything about how much, or whether it's faster at all)..
<thomasd>having arrived at the end, my impression is they like to use Nix to provide a homogenous environment (automake, glibc, etc) across all their machines (it's a network of hpc clusters), to isolate themselves from the OS (which may be a different version on different sites)
<thomasd>and then indeed, they like to use easybuild because it provides all the packages they need, for various toolchains (like intel)
<civodul>thomasd: so essentially, EB is used as a way to get proprietary software, right?
<civodul>i mean, i'd guess that one would generally prefer to use a single tool instead of 3, when possible
<civodul>also i think Nix does not do the job of modules (environment vars, etc.)
<civodul>whereas Guix does it to some extent
<thomasd>ah, didnt' know that
<thomasd>they do talk about "nix profiles"?
<civodul>right, but Nix doesn't have anything like our "search paths"
<civodul>so it cannot generate ~/.guix-profile/etc/profile, for instance
<thomasd>also, I think theoretically they could wrap intel etc in Nix (?), just that it's probably a lot of work, which is alredy done for them in easybuild
<civodul>what it has is "setup hooks", which are Bash snippets run in the build env, usually to define env. vars
<civodul>"beating ICC" may be a good candidate for the FSF's high-priority list
<thomasd>do you know any relevant benchmarks?
<roelj>Could we extend 'guix environment' for: guix environment --profile=/path/to/profile
<roelj>The --ad-hoc thing is going to be confusing then though
<civodul>thomasd: no
<civodul>thomasd: but i wouldn't be surprised if Intel prohibited benchmark publication from its customers
<civodul>roelj: what would that do? are you aware of --root?
<ng0>I'm not sure wether vim-scheme, the one I linked to last week or so, has bad indentation by default or if there's something I missed.. how do you solve the indent, efraim (or anyone else using vim)?
<roelj>civodul: It would do the equivalent of `guix package --search-paths -p /path/to/profile`
<roelj>civodul: And additionally unset any other environment variables when passing --pure?
<civodul>maybe, that could help
<civodul>it's funny, i think we're reaching a state where there are many ways to achieve the same thing
<civodul>many of which are little-known
<civodul>a bit like Emacs :-)
<civodul>i think at some point we need blog posts or videos showing specific workflows
<civodul>(or sections in the manual)
<ng0>is openssl 1.0.2k or higher in coreupdates?
<thomasd>(if anyone knows: do we currently have a graphical ftp client in guix?)
<ng0>wasn't filezila added recently?
<efraim>ng0: I'm not in front of my computer ATM, but I think I solved it by having the default indent be two spaces and having tabs autochange into spaces
<ng0>I run into this same problem the default scheme mode in emacs has without our guix-emacs workaround
<ng0>thanks :)
<thomasd>ng0: I think filezilla is still in the works
<ng0>that did it, thanks efraim!
<ng0>now the problem is with (lambda* (#:key foo #:allow-other-keys) <ENTER> lands 12 spaces into the next line, below the second "(" .. but I can figure that one out
<rekado_>civodul: I'd like to have a section in the manual to showcase some workflows.
<rekado_>is ICC really competing with GCC, though? I thought it was mostly the Intel BLAS libraries that have a foothold in scientific computing.
<[df]>I've heard anecdotally that ICC is better at vectorising
<braunr>gcc 5 and 6 apparently made great progress at that
<ng0>argh. I've sent one patch in the vim series where I still wanted to check the license if I was correct about it :/
<ng0>well this won't stop people from adjusting my reading of the license
<ng0> this one. my guess was expat, but it could be bsd-3 as well?
<rekado_>ng0: what makes you think it's BSD 3?
<ng0>some irritation with github.
<ng0>i think expat is more like it
<rekado_>I always compare with
***octe_ is now known as octe
<ng0>okay. yes, it is expat then. my "; check again" comment just needs to be removed then
<civodul>rekado_: i've heard it is better, though i tend to take it with a grain of salt because people don't necessarily know whether it really is
<ng0>I don't use neovim currently, but it would be cool to avoid vim + neovim plugin collision and use different rtp folders.
<ng0>I'm away for a while :)
<janneke>phant0mas: is `./pre-inst-env guix package -i hello' supposed to work on hurd;
<janneke>your wip-hurd-native branch?
<phant0mas>with some tests disabled, yes
<phant0mas>I need to update the branch
<janneke>phant0mas: are these kind of errors: ice-9/boot-9.scm:109:20: Throw to key `encoding-error' with args `("scm_to_stringn" "cannot convert narrow string to output locale" 1073741930 #f #f)'.
<janneke>something you recognise?
<phant0mas>you should'nt have this type of errors
<phant0mas>let me check
<janneke>i tried to play with with-fluids ((%default-port-encoding #f)) etc
<janneke>and patches got get rid of utf-8 and latin1 ... just to get thing to work and make a good fix later...but now i'm kinda stuck
<phant0mas>janneke: please pull the github repo
<phant0mas>I forgot to push the locale patch for glibc-hurd
<ng0>would someone of you with an github account be interested in joining the neomutt packaging team group, so we can make changes to our distro page of their website directly?
<ng0>otherwise it's okay if changes come in via suggestion on their devel mailinglist as far as I understand them
<ng0>I will forward (CC) our list with the next reply to the thread, then you'll see what I mean
<ng0>I'm away again
<ng0>bye o/
***jonsger1 is now known as jonsger
<rekado_>ugh, the HN crowd is so off-putting when it comes to licenses...
<thomasd>any GNUrus here who know if/how you can configure gnus to use another mail client when replying?
<thomasd>(I'm asking for the guix-patches debbugs)
<rekado_>ha, I have the same desire
<rekado_>I see that debbugs has support for rmail and gnus
<rekado_>using a different email client seems to require small changes to the debbugs.el code
<rekado_>thomasd: what mail client would you like to use?
<thomasd>mew (not to be confused with mu :0) )
<efraim>Can I 'mutt -f http://...debbug...'?
<thomasd>btw I had to make a small change to paroneayea's emacs config code. sent a message about that to the list but it seems to have gotten stuck somewhere again
<rekado_>thomasd: I've seen a notification of your email (because I'm one of the list admins)
<rekado_>I don't know if this is normal.
<rekado_>I got a notification about my own email as well, right after I had subscribed.
<rekado_>maybe that's the usual graylisting.
<thomasd>does the notification say anything about why it gets stuck? There's also an obscure bug report of mine that never made it to bug-guix (2 attempts) or guix-devel (1 attempt)
<rekado_>I don't have access to my mails right now, so I can't check.
<thomasd>well, if you ever see something I should be worried about, let me know :)
<nliadm>is there a way to run things on an interactive bash shell exit? or to bust out some SAT skills, bash_profile:bashrc::bash_logout:?
<nliadm>will traps just work?
<nliadm>for context, I want to start some things in the new environment when I do `guix environment ...` but then also kill them when I leave
<civodul>thomasd: if it's the first post with a given address, it takes some time
<thomasd>strange, I don't think ti was my first post with that address (talking about guix-devel)
<thomasd>anyway, I have to go now. see you later!
<nliadm>update: yeah, traps work fine
<ng0>There was this question about django, which is why the postorius packages haven't been pushed (includes defusedxml), can someone comment on those questions?
<ng0>or maybe I'm wrong and they have been pushed
<ng0>git says no.
<takside>hello, is there any info about running a web server with guix?
<lfam>takside: Yes, check the manual section 'Web Services':
<rekado_>takside: with Guix or with GuixSD?
<takside>lfam: thanks very much
<takside>rekado_: GuixSD, i'll check the manual
<lfam>I recommend looking through all of the 'GNU Distribution' section to learn how to add and configure services like the web server
<rekado_>I'm running a web server on GuixSD. Feel free to ask here for clarification if anything in the manual isn't easy to understand.
<takside>lfam: will do, ta
<takside>rekado_: okie dokie, thanks
<paroneayea>did I do something to break debbugs-gnu
<paroneayea>C-u M-x debbugs-gnu <RET> <RET> guix-patches <RET> n y
<paroneayea>and I only see one bug
<janneke>i see the same
<paroneayea>I sent a control message to one bug to mark it with patch
<paroneayea>and then refreshed
<paroneayea>and then bam, only one package showed
<janneke>not sure what archive and `unwanted' bugs are
<paroneayea>all I know is if I turn on archived bugs, it never loads :)
<janneke>oh wait, i said N to both achived and unwanted
<janneke>with "unwanted" I get 13 bugs
<paroneayea>did I somehow mark as unwanted? :\\
<janneke>well, you said 'y' to show that should not matter?
<paroneayea>I said y to surpress unwanted
<janneke>ah, my bad: say no <RET> n n
<paroneayea>it shows then
<paroneayea>but it was showing before
<janneke>inverting negatives is not my biggest strength
<janneke>ACTION wonders what `wanted bugs' are :-)
<paroneayea>oh now they show up again
<janneke>yes, here too
<paroneayea>debbugs-org is out of this world
<CharlieBrown>Before I went to bed, I put all the commands in the binary installation guide into a file and ran it as root in a script(1) session. I edited the lines to fit my system. I ran the script, pressed enter, quickly did Ctrl+Alt+F4, and ran away from the laptop to make sure I didn't keep myself preoccupied with it.
<CharlieBrown>Does this look OK? $ cat ~/data/guix/asroot-*.sh | sprunge
<efraim>about the archived bugs, it could be a bug that since there are currently no archived bugs it doesn't know how to show nothing
<CharlieBrown>Oh, and here's the output. $ cat ~/data/guix/asroot-*.out | sprunge
<CharlieBrown>I'd like to see a human's evaluation before I see the damage in TTY2.
<nliadm>'go install' is broken for me, it keeps trying to rebuild the stdlib
<lfam>CharlieBrown: What version of systemd are you using?
<lfam>CharlieBrown: Older versions of systemd don't support symlinked service unit files. I guess that if the daemon is running, then you're fine
<CharlieBrown>lfam: I just installed Parabola. It should be the newest version of systemd, yeah?
<lfam>CharlieBrown: I don't know. You can check with `systemd --version`. Like I said, if the daemon is running, then it worked.
<CharlieBrown>What, did you see something that said it was running?
<CharlieBrown>lfam: -bash: systemd: command not found
<CharlieBrown>$ systemctl --version
<CharlieBrown>systemd 232
<lfam>That should be fine
<efraim>glibc@2.25 broke my bootstrapping by relying on falloc.h from the linux headers
<efraim>LFS to the rescue!
<nliadm>> go install runtime/internal/sys: open /gnu/store/59sg6rc7y5xs0g2jdk1ka495la0p60y4-go-1.7.4/pkg/linux_amd64/runtime/internal/sys.a: permission denied
<nliadm>anyone else care to confirm 'go install' barfs?
<CharlieBrown>What's go?
<jmi2k>CharlieBrown: A programming language
<CharlieBrown>jmi2k: Oh god. Why do programming languages have their own package managers? :-( Can Guix wrap around them like Chocolatey?
<CharlieBrown>It's bad when you need a package manager manager.
<nliadm>go doesn't really have a pacakge manager
<CharlieBrown>>"go install ..."
<lfam>ACTION slowly bashes an agetty-service into existence
<lfam>ACTION hopes he can finish bashing out the service before he bashes in his head ;0
<nliadm>go install will build and cache things
<lfam>nliadm: Does that file 'sys.a' exist?
<nliadm>^W yes
<lfam>Hm. Maybe we need to create a go-toolchain package, analogous to the gcc-toolchain package.
<nliadm>I'm looking to see how the go tool figures out what to build
<lfam>nliadm: Also, are you interested in testing the go 1.8 pre-release?
<nliadm>I'm assuming it has something to do with the 0 timestamp
<nliadm>and yeah, I'd give it a spin
<lfam>It's probably too late to report bugs, but at least we could get a head start on working around them ourselves
<efraim>well, i hacked in linux-libre-headers-boot0 as an input to gnu-make-boot0, we'll see if that works, especially with a suprise missing perl input
<civodul>hey lfam!
<civodul>lfam: i'm looking into that CVE issue you reported
<lfam>Hi civodul!
<civodul>there are quite a few CVEs that do not specify the version of the affected "product"
<civodul>like this:
<lfam>Yeah, a real shame :(
<civodul>but then, that means they'll always be reported
<civodul>kind of a problem
<lfam>Always reported, or never reported?
<civodul>currently they're never reported
<civodul>but if we fix that, they'll always be reported
<civodul>(unless there's a patch with the right name)
<civodul>so that's not great either
<CharlieBrown>I'm thinking about installing everything with Guix, on Parabola.
<lfam>civodul: I think the CVE system is changing right now. I sense a general dissatisfaction with it from upstream software vendors and distribution teams.
<lfam>So, hopefully we can help steer it into something more useful.
<lfam>I've been meaning to look at DWF more closely:
<lfam>It's not as if the oss-security mailing list included all the important bugs, anyways.
<civodul>i just read that DWF is a CNA
<lfam>CharlieBrown: I do something similar on Debian. It's a fun game to keep packaging missing things so one can install them with Guix :)
<lfam>Lots of organizations are CNAs. The big distros, for example.
<lfam>We're already ahead of the game with a tool that parses the NIST NVD feed.
<CharlieBrown>lfam: The commented lines, I want in Guix. Does this config sound reliable?
<efraim> the commit that makes glibc depend on the kernel headers :(
<lfam>CharlieBrown: The Xorg stuff might not work. I think it needs to be setuid, although I'm not sure
<lfam>efraim: At one of the FOSDEM dinners, Jose asked us why the glibc team never gets bug reports from Guix :)
<lfam>Wait, that was the GCC team, not glibc
<lfam>Well, anyways, you get my point ;)
<efraim>we sent that armhf bug with gcc-5.x and 4.9.4
<lfam>Right, I mentioned that at the dinner
<lfam>Also the chunked store references thing
<lfam>I guess he wants more bug reports :)
<efraim>i should let them know about the aarch64 flags not working on 4.9
<efraim>the "target armv8 specifically" and the two "whoops, fix erata xxxxxx and yyyyyy" flags
<slyfox>4.9 was released a while ago
<nliadm>lfam: yeah, the go tool is reporting that all the build ids are wrong for the stdlib
<civodul>lfam: i'm trying to find the spec corresponding to the "" XML namespace used in the CVE/NVD database
<civodul>do you know where that could be?
<civodul> is quite intimidating and not necessarily à propos
<civodul>efraim: so is the problem that falloc.h is only available in recent kernel versions?
<CharlieBrown>Doesn't 'à propos' mean 'about'?
<lfam>civodul: I haven't looked
<efraim>its there at least as far back as 3.16, the problem is its not in our bootstrap process
<CharlieBrown>=> "... not necessarily about."
<lfam>I've never heard 'à propos' used in a way that means "about"
<lfam>It means "appropriate to", "related to", "on the subject of"
<efraim>like "vis a vie"?
<slyfox>i have a silly question about closures: 'guix size ghc' reports gmp 3 times, libffi 3 times and ncurses 2 times.why those libraties are not shared? i tried to run 'gux graph --type=...' but seems all the graphs contain build-time depends
<lfam>Well, they are both French expressions. A native French speaker can explain the connotations more clearly
<civodul>slyfox: try 'guix graph -t references ghc', for the run-time deps
<civodul>lfam: i think you're right about the meaning of "à propos", at least that's what i wanted to convey :-)
<clacke[m]>apropos is most close to "speaking of", I would say
<efraim>__USE_GNU isn't linux specific, right? does that also target hurd?
<clacke[m]>vis-a-vis is "with regards to" or "as compared to"
<slyfox>civodul: yay! that's exactly what i'd like to see. gmp comes as a depend of guile, coreutils and ghc: . shouldn't be shared?
<nliadm>$ go list -f '{{join .Deps "\\n"}}' net/http | xargs -n1 go list -f '{{if .Stale}}{{.ImportPath}}: {{.StaleReason}}{{end}}' # all the build IDs are wrong
<efraim>well, i'm going to bed, i'll find out in the morning if this works.
<efraim>i don't believe I currently have anyhting to push to core-updates
<clacke[m]>ah, civodul used apropos without an object. then it probably means "on topic", as it was used here
<civodul>slyfox: yes, they should, there's room for optimization here
<civodul>slyfox: the problem is that the guile that you see here is "deeper" in the graph than the ncurses ghc directly depends on
<civodul>efraim: __USE_GNU is an internal macro corresponding to _GNU_SOURCE (which users can choose to define or not)
<lfam>I've got this agetty-service:
<lfam>Using it, I can see the login program running on /dev/ttyS0 in a QEMU VM, although I don't know how to connect to the QEMU serial console.
<lfam>But, when I use it on my device that only has a serial console, I can see the system boot over the serial console, but Shepherd doesn't mention that the agetty-service has started, and the login prompt is not presented
<lfam>In QEMU, Shepherd does report that it's started the service
<civodul>lfam: the service looks good!
<lfam>I'm using it like this, in the installer image:
<lfam>I've tried a few variations of the arguments to the service
<lfam>I also noticed that agetty is not running in QEMU, and the login program is running even when I ask for auto-login. So I wonder if I'm invoking it correctly
<lfam>I'm a bit stuck :)
<lfam>I don't have any way to interact with the booted GuixSD system on the physical machine, nor a way to access the serial console in QEMU
<civodul>i suggest testing as much as possible in QEMU
<civodul>from there you should be able to run "herd status agetty" etc.
<civodul>well you could run mingetty on one of the ttys to make sure you can at least log in somewhere :-)
<lfam>mingetty doesn't support the serial console, AFAICT
<lfam>That's why I'm using agetty
<civodul>but for testing it should be good enough?
<lfam>In QEMU, sure
<lfam>I need to figure out how to make QEMU emulate the serial console, and then try connecting to that
<lfam>So, the term-ttyS0 service is reported as running in QEMU, the login program is running on that tty
<lfam>And bash as well. Presumably due to the auto-login
<lfam>All the instructions I find for other distros involve /etc/inittab, but I think that is totally irrelevant for us
<lfam>Okay, I can kill that bash, and the service is respawned. That's good
<lfam>Aha. Passing '-serial pty' when starting QEMU makes it possible to connect to the QEMU VM using `screen`
<lfam>So, the service does work in QEMU :)
<lfam>But it doesn't seem to start at all on bare-metal!
<civodul>of course debugging on bare metal is harder
<lfam>So, the difference is that for QEMU, I am using os-config-bare-bones.texi and adding agetty-service to %base-services. On bare-metal, I am editing gnu/system/install.scm to include the agetty-service
<civodul>you should test the exact same one in QEMU
<quiliro>hello guixsters!
<lfam>Hi quiliro
<quiliro>would someone please provide information on how to start mariadb?
<quiliro>hi lfam
<quiliro>i don't understand how the filesystem structure is in GuixSD
<quiliro>so i have no idea how to deal with this
<quiliro>do i need to start it as root?
<quiliro>but i installed mariadb as user not as root
<quiliro>i am reading the manual completely
<quiliro>is there a shorter route i should take
<quiliro>i want to collaborate
<quiliro>i was able to start mariadb
<quiliro>i just created /var/lib/mysql as root and changed its owner as the user and group as users
***braunr_ is now known as braunr
<quiliro>but i have other errors
<quiliro>the new error is
<nliadm>oneliner to reproduce the problem I'm seeing: guix environment --ad-hoc go -- bash -c 'export t=$(mktemp -d); cd $t && export GOPATH=$(pwd) GOBIN=$(pwd)/bin && go get; cd && rm -rf $t'
<quiliro>what was the last message that was receive by me?
<quiliro>is there a bot that is up to that task?
<braunr>23:45 < quiliro> the new error is
<braunr>i take it you meant "from me"
<jmi2k>How can I disable --enable-fast-install in gnu build system? I'm trying to build vice and its configure script can't recognize it...
<quiliro>braunr: thank you...yes that is what i meant
<quiliro>guix pull gives errors after update
<CharlieBrown>I have guix on my foreign system.
<CharlieBrown>Installed. Do I do guix pull before installing anything?
<CharlieBrown>Fresh install.
<quiliro>CharlieBrown: are you familiar with debian
<quiliro>guix pull is like apt-get update
<quiliro>you have freshly installed so you have the latest guix
<quiliro>so you do not need guix pull
<lfam>Actually, CharlieBrown won't have the latest Guix.
<quiliro>lfam: really? why?
<lfam>0.12.0 was released on December 20, and there have been many important security updates since then