IRC channel logs


back to list of logs

<ng0>oh... okay. that'suseful to know that multiple zero? need (and .. i'vejust read the email
<sapientech>janneke: okay sweet, on the site im looking at, i need lib32-stdc
<sapientech>so is there a way to specify the architecture i want when i do guix package -i libstdc++ ?
<mordocai>sapientech: I'm not using GUIX atm but it looks like --system=i686-linux might work. That may be only for guix build though
<ng0>3995 packages :)
<ng0>I think I won't package+service write this ( in addition to working on gogs, kallithea and all other packages I work on at the same time.. looks pretty bug.
<OrangeShark>does guix have a package for git send-email? I don't see any.
<lostcoffee>I have a (home-directory "/home/me") but my home directory didn't get created
<sapientech>mordocai: thanks :)O
<bavier1>OrangeShark: install the "send-email" output of the git package
<bavier1>OrangeShark: e.g. with 'guix package -i git:send-email'
<dvc>mark_weaver: Sorry about the build failures.
<dvc>mark_weaver: Do you think it's necessary to revert the commits?
<alezost>ACTION thinks the best would be to fix the package to make it build successfully
<dvc>alezost: The problem is that it depends on patches from core-updates and they'll continue to fail until core-updates is merged :/
<dvc>I should have pushed to core-updates, but I'm in a habit of prepending and removing my selection of core-updates packages before pushing...
<alezost>dvc: oh, got it, then I would revert it and push it to core-updates, but I would wait for mark_weaver
<efraim>the wayland packages?
<efraim>i haven't seen what broke, but if the patches didn't break anything that was there before I'd consider just leaving them
<civodul>Hello Guix!
<sneek>civodul, you have 1 message.
<sneek>civodul, paroneayea says: see for some "emacs of distros" type discussion :)
<civodul>paroneayea: i had never read that, interesting!
<civodul>very wise
<dvc>efraim: yes the wayland packages. It didn't break anything that was already there
<efraim>so my 2007 macbookpro wouldn't recognize my guixsd usb stick so it looks like I'm going to need to bootstrap the installation from another distro install disk
<efraim>also I'm thinking of waiting on manolis's guix-daemon patch before I try more with aarch64, but I really should just pull out a board and my keyboard, everything else is already on this desk
<dvc>it didn't recognize your usb stick or the file system?
<efraim>the usb stick
<dvc>that sounds like a faulty usb stick or port
<efraim>i know its good because I used it on another dying laptop
<dvc>and the port is good too?
<efraim>the light on the usb stick lit up so it should be good
<efraim>i think it has to do with the efi loader on the macbookpro
<efraim>and we don't have efi support yet, so it didn't recognize it
<efraim>I bet if I burned it to a CD and used that I could get it to boot up
<efraim>before I package refind I figured I should check if it would even work
<dvc>or we can use the EFI stub in the kernel
<dvc>it's probably already enabled
<efraim>the kernel is outside of my area of knowledge
<dvc>civodul: Regarding the rustc package, is it a problem that it uses a binary to bootstrap?
<efraim>yeah, it's a problem
<efraim>I just looked at the patch too, it also only covers x86_64 and i686
<dvc>it involves adding a FAT32 partition and copying the kernel to it.
<civodul`>dvc: it's a problem, but it's a problem that we might have to live with :-/
<civodul`>see the discussion on the list
<efraim>does rust take a long time to compile?
<dvc>that's not a problem, arm and mips can be added
<efraim>go at least builds quickly, its the tests that take time
<dvc>it takes a while because it compiles rust 3 times
<dvc>and it has an extensive test suite
<efraim>also they're removing java support from gcc in 7
<dvc>java is dead since oracle touched it =P
<efraim>the bootstrap process is really 300+ builds?
<efraim>we could also turn off tests for all but the last iteration
<dvc>efraim: are you talking about rust?
<dvc>I think that the rustc people know what they are doing
<dvc>I don't know if it reruns the tests every time
<dvc>and a long time is about 1-2h depending on the machine...
<civodul>FWIW we discussed it at
<dvc>but it's getting quicker and quicker
<civodul>dvc: actually are you the person who sent the initial patch?
<dvc>civodul: Nope, I think ng0?
<dvc>I remember the discussion, but it trailed off. I don't know if it was a won't merge or a we'll live with it...
<efraim>I think people were hoping for a better bootstrap solution
<efraim>sneek: later tell lfam I remember you said before that borg is faster than attic. Over the same data set on this machine it's 3-5 times faster
<sneek>Will do.
<efraim>sneek: botsnack
<civodul>dvc: i think it was more "let's see what we can do, and maybe live with it", but it stopped here:
<civodul>we should definitely restart the discussion
<efraim>mmm, I have to review wlc and sway later today
<efraim>i also feel bad about fabian's mate patches from back in february, but he seems to have dropped them
<efraim>Currently we don't have support for swapfiles declared in the OS config, do we have support for automounting them at boot-up?
<efraim>also, does mcron have something similar to vixie-cron's @reboot?
<rekado>efraim: <efraim> also they're removing java support from gcc in 7
<rekado>efraim: where do they say this?
<rekado>we need Java in GCC in order to cleanly bootstrap OpenJDK.
<jmd>efraim: That is not decided. Yet. Some of the devs want to ...
<jmd>Woha!! We've just hit the 4000 packages mark!
<efraim>oh, I thought it was already decided
<efraim>I saw it on reddit, which was probably someone summarizing phoronix's coverage
<efraim>is there a licensing issue for why we haven't packaged x265 yet? Or just no one got around to it yet
<rekado>how do you use Emacs+Geiser with "guix environment"?
<rekado>I have an environment specified in guix.scm and would like the environment to be available when I open Geiser on a project file.
<civodul>dvc: oh there was a v2 of the Rust patch posted just today :-)
<jlicht>hello guix
<ng0>where can come from? I do miss something like pfl's e-file in guix.
<ng0>thanks for answering my question, ng0
<ng0>no problem ng0, yw :)
<ng0>can someone explain to me why there is guix-build-ghc-8.0.1.drv-0/ghc-8.0.1/ghc-bin/usr/lib/ghc-8.0.1/terminfo- and guix-build-ghc-8.0.1.drv-0/ghc-8.0.1/ghc-bin/usr/lib/ghc-8.0.1/terminfo- where only includes the symlink which is created in the 'configure-bin phase? Why is the libtinfo link not in the termininfo- ? I almost suspect that was with the previous ghc version ?
<ng0>but it is also mkdir-p in the phase, so it can't be that...
<brendyn>ng0: You're welcome.
<ng0>sometimes it's easier to ask, but I guess I'll do what I always do try until it no longer fails.
<ng0> /me afk
<brendyn>I think I'm starting to suffer the consequences of having two different package managers
<brendyn>Emacs now locks up for 30s until some system call times out on startup
<rekado>brendyn: have you tried bisecting your init.el?
<brendyn>rekado: spacemacs
<brendyn>Normal emacs works just fine
<Sleep_Walker>efraim: thanks for review
<janneke>civodul: thanks for the small rewrites and pushes! *happy*
<efraim>Sleep_Walker: np :)
<civodul>janneke: yw!
<civodul>rottlog was much easier than the MinGW series ;-)
<janneke>haha...yes. mingw is ready for master, anything i need to or can do?
<civodul>given we had already reviewed it, it's probably fine
<civodul>i just need to take time to double-check and push i suppose :-)
<janneke>next up for me is full lilypond cross build and ehrrm npm importer
***vagrant` is now known as cordawyn
<davexunit>rekado: not the most elegant but I run 'guile --listen' in the environment shell and then use M-x connect-to-guile to connect to the REPL and start hacking.
<rekado>davexunit: that's not a bad idea, thanks!
<rekado>how can I tell libcurl (not the command line tool "curl") where to find certs?
<rekado>there's an R package that ends up using libcurl and I'd like it to find the system's certs.
<rekado>Tried CURL_CA_BUNDLE already, but that doesn't seem to work.
<ng0>rekado, same for my not posted progress on pbpst.. it just doesn't work.
<ng0>also, i might have to revert back to yet another email address, waiting on the reply of support of my new SaaSs provider wether cnames for the mailhosting work or not.
<rekado>just checked the sources for r-curl: CURL_CA_BUNDLE is only checked when on Windows.
<rekado>I think I should patch the platform check away.
<jlicht>rekado: if you find out, please do tell me: I had quite some issues with some of my own pet projects + guix' libcurl :-)
<rekado>jlicht: for now I'm just looking at r-curl (because this looks easy to fix). Ultimately, we may want to patch libcurl.
<rekado>might be worth checking how Nix does this.
<rekado>as expected they patch libcurl
<rekado>oh, used to.
<rekado>now they just pass "--with-ca-bundle=/etc/ssl/certs/ca-certificates.crt"
<rekado>previously they defined CURL_CA_BUNDLE as a macro that checks the env vars CURL_CA_BUNDLE and SSL_CERT_FILE
<rekado>I fixed my problem in r-curl by uncommenting special treatment for Windows.
<jlicht>rekado: thanks, I'll have a look at that
<rekado>I think it's bad to configure curl with "--with-ca-bundle=/etc/ssl/certs/ca-certificates.crt"
<rekado>1) it won't work on Fedora (unless root decides to add a symlink)
<rekado>2) it makes per-user certs impossible
<janneke>after rebasing on master my VMs with postgres-service throw an error during startup
<janneke>something with user-add
<ng0>rekado, what could we do instead for curl? do what nixos used to do? or what they do now?
<civodul>rekado: probably the reason for the hardcoded /etc/ssl/certs is that cURL does not honor any env var, nor does GnuTLS
<civodul>i agree that it sucks
<ng0>yeah.. great. I'll open a bug report at runbox.. they can't cope with cnames. meanwhile all my email will now come from/to and I'll change the copyrights and mailmap and leave it like that because chnaging back and forth sucks. sorry
<brendyn>Ah, so is that why I get TLS errors when updating emacs packages
<ng0>most likely.
<brendyn>Hopefully someone else will solve all these problems for me.
<ng0>so it sorta kinda works. I'm not sure. But I already changed 3/4 of my mailing lists and subscriptions to the new model.
<brendyn>How can I list the build dependencies of a package? I tried building guix but It complained that I didn't even have `which'
<ng0>this is silly. web interface says the domain is deactivated while I can clearly use it. y u so strange runbox.
<jmd>brendyn: All of them or just the immediate ones?
<brendyn>jmd: I don't know. All I want to do is build guix from git
<ng0>see dependencies for the daemon :)
<davexunit>brendyn: just run 'guix environment guix'
<davexunit>that will get everything you need
<brendyn>Currently I get to make[2]: *** [Makefile:4956: emacs/guix-autoloads.el] Error 255
<jmd>davexunit: But he wants to build without guix. (I think).
<brendyn>davexunit: I tried that with --pure
<brendyn>jmd: I was trying to build without any non-guix stuff actually
<jmd>But you mentioned git.
<brendyn>Well how else am I supposed to be able to hack on Guix? I've read the manual and 2 or 3 different blog posts on it but I'm still having trouble
<davexunit>brendyn: 'guix environment guix', lots of use it for development
<davexunit>it works
<davexunit>brendyn: the manual is quite clear about what needs to be done:
<brendyn>davexunit: I did that then ./bootstrap; ./configure --localstatedir=/var; make; and I get to make[2]: *** [Makefile:4956: emacs/guix-autoloads.el] Error 255
<davexunit>time to debug, then.
<davexunit>perhaps someone made a bad commit to the emacs stuff?
<brendyn>Ok so it's maybe not 100% my fault?
<alezost>emacs stuff wasn't touch for a long time
<brendyn>But how is anyone still hacking on guix today if this is a bug?
<jmd>brendyn: Try having a look at line 4956 and seeing what it says.
<brendyn>... that prevents building
<davexunit>brendyn: you're not giving us much information to go by
<brendyn>ok 1 sec
<brendyn>$(AM_V_EMACS)$(EMACS) --batch --eval\\
<brendyn>plus some lines. 1 more sec
<alezost>brendyn: paste the backtrace please
<brendyn>(The make output is appended further down)
<brendyn>It looks like tee discarded the error output, let me try again
<brendyn>Ok try this alezost davexunit
<ng0> greylists new adresses / address changes for a while.. right?
<alezost>brendyn: can you show line 4956 (and below lines) in that Makefile
<brendyn>Oh right, the makefile is generated isn't it
<brendyn>Sorry, my hands are cold so I'm tying really slowly. Takes ages do use Emacs
<jmd>Ahh those Tassie winters!
<ng0>winter? oh.. really tasmania or somewhere close?
<brendyn>It was so sunny today so I left the balcony door open while I was out, but I got back at midnight and now my room is cold :p
<jmd>ng0: Not that I'm aware of.
<jmd>ng0: There isn't anywhere close to Tasmania.
<brendyn>I'm at Hobart, Australia (42 degrees south (latitude))
<brendyn>Next Stop Antarctica!
<brendyn>Atleast we have fairy penguins. penguins are cool
<brendyn>Thousands of little Linux mascots running around in the night
<jmd>Sometimes, they're freezing!
<ng0> [Whois] Rejected (Invalid DNS settings) [Details| Reactivate] .... still I can send emails. schroedingers email.
<jmd>Anyone from Botswana here who hacks guix while watching GNUs gamboling in the bush?
<brendyn>I actually discarded my mail server for a public one in order to hack on Guix, because the requirement of having a mail server that can actually send and receive mail most of the time was too much to handle.
<brendyn>Currently I can receive mail but can't send it
<jmd>ng0: Guix currently sets /etc/hosts up in a really stupid way, so I wouldn't be surprised if some servers reject it.
<ng0>no. that's not even related :)
<davexunit>jmd: could you elaborate?
<jmd>On what?
<ng0>I used to host this myself on debian, i gave up and returned to runbox because administration sucks and takes your time and I have none.
<jmd>Getting it setup does take time. But once it's set up it's up.
<ng0>opensmtpd is almost packaged.. some tweaking.. and then the service needs to be written.
<efraim>i still host my mail, but i've considered using a third party for sending
<ng0>i mean it is package, -extras is packaged, libasr is packaged, i just have no real usecase (needs service)
<alezost>brendyn: hm, strange; can you look for a line starting with "EMACS =" in the makefile? is it value "emacs"?
<jmd>brendyn: My guess is that for some reason the AUTOLOADS variable is not set.
<brendyn>EMACS = emacs
<efraim>civodul: I'm looking at your recent commits with guile-2.2 for some guile- packages, wouldn't replacing guile with guile-2.2 go all the way down? Like to near bootstrapping?
<ng0>the only selfhosting thing i can't drop is git... I don't know a service which could satisfy the git+https + federation + no webinterface for some repositories I have.
<brendyn>Or maybe crap in my .bashrc sets env variables that break stuff?
<brendyn>I would have thought guix environment guix would fix that
<jmd>brendyn: I would have too.
<davexunit>brendyn: you should never set environment variables in bashrc
<davexunit>that is certainly a problem
<efraim>guix environment guix adds to your environment
<davexunit>and then runs $SHELL
<davexunit>if $SHELL is bash... then it reads .bashrc like normal
<davexunit>and if you clobber environment variables...
<brendyn>Really? I thought that's what .bashrc was for
<davexunit>the 'guix environment' docs in the manual talk about this
<efraim>civodul: nevermind, guix build guile2.2-minikanren shows just downloading minikanren source and guile-next
<davexunit>brendyn: .bash_profile is the correct place
<brendyn>I mean, just following internet tips I've got lots of crap in there
<davexunit>because .bash_profile is only eval'd with login shells
<davexunit>see the bash manual for more detail
<efraim>i put mine in .profile, and then .bash_profile sources .profile
<brendyn>My bash_profile contains [[ -f ~/.bashrc ]] && . ~/.bashrc
<davexunit>that is fine
<jmd>davexunit: But sometimes you *want* a variable to be set in all/most shells.
<davexunit>jmd: but usually you don't want that set when you make a non-login shell
<alezost>brendyn: the thing is: if you set some env vars in your .bashrc, then "guix environment" may be influenced
<davexunit>I set my $PS1 in bashrc, but I never set search paths.
<jmd>Well I can think of many instances when one does.
<brendyn>Ok Let my try move things to bash_profile and see how many other things this breaks
<jmd>Setting EDITOR for example.
<jmd>But, like you say, that only gets sourced from login shells.
<davexunit>which is exactly what you want
<brendyn>ACTION goes to that bathroom to run warm water of his hands before using Emacs
<jmd>... most people doe that AFTER using vi!
<davexunit>but anyway, those variables don't really matter so I don't want to argue about them. just know that setting *search paths* in .bashrc is bad practice.
<brendyn>jmd: You're real funny today
<jmd>davexunit: Well I want my EDITOR variable to be set in all/most interactive shells. Not just login shells.
<davexunit>'guix environment' is just one of the programs where such practice can cause problems
<davexunit>jmd: but it would already be set!
<davexunit>because your interactive shell would be a child process of your login shell!
<jmd>ACTION experiments.
<ng0>i have everything in .bashrc and I start it all with my.* .. worked for over 15 years
<brendyn>and guix is a child of my interactive shell, that is ok?
<jmd>No. Not necessarily. What about starting from a Xsession for example?
<davexunit>jmd: X creates a login shell.
<jmd>davexunit: Sorry! Not on my system it doesn't.
<davexunit>when you login at your display manager, it's creating a login shell for that user.
<jmd>Possibly. But what if you open a terminal?
<davexunit>it's a child process of your login shell
<davexunit>just like any other program
<jmd>I'll have to log out and in again to do check that.
<davexunit>some terminals have an option to create new login shells, too
<jmd>and I'm building something right now so I don't want to.
<brendyn>So does guix erase all your environment variables, set up the required ones, then run bash, loading .bashrc ?
<davexunit>gnome-terminal can do that, for example.
<davexunit>brendyn: if you use --pure, the environment is cleared sans some "precious" variables
<ng0>civodul: this video is gone says wget:
<davexunit>brendyn: 'guix environment' will run any program you ask it to run.
<davexunit>if you don't specify what to run, it will run $SHELL
<davexunit>please see the manual for all the details.
<jmd>But I do remember years ago, when I first started in unix being perplexed why sometimes variables were not getting set. Then I realised I had them in profile instead of rc Ever since then I have always used rc.
<davexunit>it contains several examples of 'guix environment' usage
<brendyn>But how is putting settings in .bash_profile going to change that behaviour? The variables will be there regardless?
<davexunit>brendyn: a non-login shell does not evaluate .bash_profile
<brendyn>Ok, It looks like I still do require quite a few things in .bashrc
<brendyn>davexunit: But it will still have those variables due to being the child of a login shell that did evaluate it?
<davexunit>when you do something like set $PATH in .bashrc, what happens is that 'guix environment' sets $PATH to be what it needs to be to access the binaries you asked for, and then starts up $SHELL (bash in your case), and bash reads .bashrc, cloberring the $PATH that 'guix environment' configured.
<davexunit>brendyn: yes, but it happens *before*
<brendyn>Ok, I understand now davexunit, thanks.
<davexunit>it doesn't clobber what 'guix environment' tried to set
<brendyn>I just need to make sure to use `export VAR=...' instead of just `VAR=...', right/
<OrangeShark>you are usually appending some new paths which might overshadow the binaries references in the $PATH guix environment sets
<efraim>lightweight-desktop.tmpl needs a fix: should I change "alice" to "alex" or change "bob's brother" to "bob's sister"?
<ng0>bob's sister.
<brendyn>jmd: Maybe 15 years ago you should have used export VAR instead of moving stuff to .bashrc
<ng0>isn't that what everyone still does? export $var=stuff ?
<jmd>brendyn: Maybe. Anyway based on what davexunit said, I'll see if I can move some stuff back to .profile
<brendyn>I've got some crazy things here like PROMPT_COMMAND="history -a; $PROMPT_COMMAND"
<brendyn>Presumably I need to add export to his when I move it
<brendyn>But what about shopt -s histappend; that seems to be a special bash command, so I guess it needs to be ran every time
<brendyn>I have my bash history set to stay around for eternity
<efraim>do you only have 1 terminal open at a time?
<efraim>when I close terminology I lose all but one tab's history
<brendyn>I usually have lots
<brendyn>I use terminology too
<jmd>Then which history do you keep?
<efraim>I think its whichever one closes last
<brendyn>I never use terminology tabs though.
<efraim>i have 4 tabs, and one has a screen session in it
<efraim>which also messes up my history
<brendyn>efraim: Maybe try that setting I'm using to save history after every command?
<efraim>i don't have a lot of interesting history so it doesn't really bother me that much
<brendyn>Most of my terminal used is pressing the up arrow to get commands starting with a certain string
<ng0>you could like zsh or fish..
<brendyn>Never heard of fish
<efraim>i tried fish once, i couldn't get used to seeing the commands show up on their own
<brendyn>They are not all that different, so I'd rather just keep using the copyleft bash, all else being equal
<jmd>brendyn: What do you think ! is for?
<ng0>though it's all somehow bash improved extended.. there's csh vs ksh vs bash vs the rest.
<brendyn>jmd: Getting the last arguments?
<jmd>!s gets the last command starting with s
<ng0>fish is very different, even in scripting syntax. too different for me to get used to.
<brendyn>jmd: I don't want the last command, I usually type ssh UP to scroll through previous ssh commands
<jmd>But if you use ! etc it saves you all that bother.
<brendyn>For example, all my backup commands are in my bash history and I just go find them like that
<brendyn>jmd: What? That runs the last command without me even getting a chance to read it
<jmd>You can read it if you want to.
<efraim>i sometime use sudo !!
<jmd>But if you're on cmd 1000 and you want to use cmd 2 do you really want to hist UP 998 times??
<brendyn>What?? I don't need to?
<alezost>brendyn: I only export env vars in my .bash_profile, the rest (aliases, shopt, etc.) I keep in .bashrc
<jmd>Well if you're a masochist you still can.... nobody is stopping you.
<brendyn>Up doesn't scroll just scroll up for me, I can type ssh and then scroll up through ssh commands
<brendyn>I think it is bash completion or something like that
<jmd>And what do you do on a terminal with no arrow keys?
<brendyn>ACTION notices jmd is trolling
<jmd>No. I'm trying to save you a lot of bother.
<brendyn>But I don't understand? Doesn't ! just return the last command?
<jmd>Possibly this is a discussion for #bash
<jmd>But no. !4 gives you command 4
<brendyn>But I have *no idea* which number the command I want is
<jmd>that's what the "history" command is for.
<jmd>!e gives you the last command begining with e
<jmd>Anyway, If you want to continue the discussion, lets do it on #bash
<brendyn>Can you give me an example of this being in anyway faster than my method?
<jmd>do you think pressing UP 10000 times is faster than pressing "!2"
<brendyn>Never once have I needed to run the second oldest command in my bash history
<jmd>#bash if you want to continue!!
<ng0>can someone notice me when my perl-mhonarc patch hits the list? it seems has host greylisted while i changed my address.
<ng0>i like inbox: 0 but it's weird.
<ng0>:me afk
<jmd>Somebody seems to have recently pushed something that is causing a lot of rebuilds!
<jmd>Hey. I'm getting "suspicious ownership or permission on `/gnu/store/g43gwmjw5g3r199m5br24c8gnsaa9mbp-harfbuzz-1.2.4'; rejecting this build output" when doing guix evironment guix
<jmd>Is anyone else getting wierd errors or is it just me?
<janneke>not sure how to debug postgres, created bug report :-(
<jmd>janneke: What's the issue?
<davexunit>janneke: #t got passed to system* which expects strings
<davexunit>should be easy to debug from there
<janneke>great, i hoped so...i'm just lost there
<davexunit>take a look at the generated shepherd configuration
<davexunit>or whatever module has that system* call
<davexunit>you're comfortable with guile so it should be fairly straightforward to debug
<davexunit>wrap a 'pk' around 'args' in the 'add-user' procedure in (gnu build activation)
<davexunit>then you can see the full arg list and find out where the #t is
<jmd>We have a service for postgres?
<janneke>davexunit: TIL: pk; i'll have a go with this info, thanks
<janneke>jmd: yes, (postgresql-service)
<ng0>finally has my new host.
<rekado>what do you think about removing the line “Found valid signature for /gnu/store/…”?
<rekado>each item that is being downloaded gets four lines.
<rekado>Found valid signature for /gnu/store/xyl43fnjp37p2jwv8kfb69p8cakqda0r-virglrenderer-0.5.0
<bavier>rekado: I would be for that
<rekado>Downloading xyl43f...-virglrenderer-0.5.0 (428KiB installed)...
<rekado> virglrenderer-0.5.0 1.4MiB/s 00:00 | 155KiB transferred
<rekado>I think that’s a bit much
<rekado>and repetitive
<bavier>I'd be more interested in seeing a message for the "invalid signature" case
<davexunit>rekado: sounds like a good idea to me.
<brendyn>Ok so moving my .bashrc to .bash_profile didn't fix anything, but using --pure after installing `which' did.
<ng0>now if only my mail to the list would arrive at my end... guess I need to wait on something else now.
<civodul`>rekado: i agree, that's too much
<civodul`>jmd: unfortunately it's a known-but-unfixed bug
<jmd>civodul`: Sorry. Which one?
<brendyn>Ok so I've build guix, now when I run a command like guix package -s emacs; I get guix package: warning: failed to load '(build-aux build-self)':
<bavier>brendyn: are you running under pre-inst-env?
<brendyn>bavier: I symlinked .config/guix/latest to my git directory, so I thought I didn't need that?
<kingcons>Hey there. I'm trying to install a nonfree kernel in Guix SD so my wifi card will be supported. I grabbed this scheme file and was attempting to load it with `guix package -f linux-nonfree.scm` but am getting no output.
<kingcons>Is there a particular directory I need to place that in or a pre-build step I'm not understanding?
<brendyn>kingcons: GUIX_PACKAGE_PATH=dir/with/file/
<kingcons>Aha. Thanks brendyn!
<brendyn>kingcons: then look at the define-public lines in the scm file. those are what you can install with guix package -i foo
<kingcons>Yep, that makes sense. Thank you :)
<brendyn>:) let me know if it works
<brendyn>It's cool how it imports the standard linux-libre definition only requiring small changes
<brendyn>kingcons: What nonfree stuff are you so unfortunately forced to use :p ?
<kingcons>iwlwifi for my intel 2200-n
<kingcons>On a Lenovo X230.
<jmd>kingcons: I thought that box could handle free wifi modules?
<kingcons>I used to do a lot of Common Lisp stuff so I've been excited to follow Guix's progress and looking forward to getting my hands dirty a bit.
<brendyn>Ah ok. Fortunately I have a wifi card from an old think penguin laptop that works with linux-libre
<kingcons>Yeah, I think there's an alternate video chipset it could have been ordered with but I bought it off a friend for cheap :-/
<kingcons>*wifi chipset
<davexunit>jmd: thinkpads (and lots of other laptops) have a whitelist for what cards you can put in the PCIe slot
<brendyn>kingcons: Nice. I'm also a beginner as of a week or two ago, but I've already had 1 package accepted into Guix. Feel free to contribute
<davexunit>so if you try to replace the nonfree wifi chip with a free one, it refuses to boot.
<kingcons>davexunit: that's horrible :-(
<davexunit>I flashed a hacked version of the proprietary BIOS on my thinkpad x220 that removed the whitelist
<jmd>So replace the BIOS
<davexunit>and I installed an athoros chip from thinkpenguin
<davexunit>replacing the BIOS isn't trivial.
<brendyn>Wow I'd never heard of that. I thought SecureBoot was the only thing like that
<brendyn>Is there a list of such computers?
<davexunit>these days you could use coreboot, I think, but you still have some blobs.
<jmd>Perhaps this is a discussion for #libreboot
<davexunit>libreboot will never support the x220 or above
<davexunit>due to the intel "management engine" stuff
<brendyn>Why does the whitelist exist?
<davexunit>to keep people from installing their own hardware into the machine
<brendyn>So there is no excuse other than that?
<ng0>discussion solved. yw.
<brendyn>"But they're just doing it to make money, right" -- everyone I ever bring up such issues with
<kingcons>So, still having a little trouble here:
<kingcons>Any idea what I might be missing?
<kingcons>Gotta run. Back in a few.
<mark_weaver>jmd: I didn't read the entire log, but I saw you mention that on your system there's no login shell in an X session. it's true that's the case on Debian, for example, and probably many other systems. in that case, I recommend sourcing .bash_profile from .xsession
<rekado>kingcons: Guile expects a module to live in a directory that matches the module name.
<brendyn>kingcons: Hmm, I think wingo 's package may be out of date
<rekado>kingcons: a module with the name (gnu packages foo) has to be at $somewhere/gnu/packages/foo.scm
<civodul>$ ./pre-inst-env guix package -A |wc -l
<civodul>we made it!
<rekado>kingcons: $somewhere would be the root of the directory you add to GUIX_PACKAGE_PATH
<jmd>We were up to 4000 this morning!
<brendyn>bavier: btw, I tried with ./pre-inst-env and had the same issue
<jmd>Do we have a party when we get to 2^12 ?
<jmd>mark_weaver: Ok. That would concur with my experience. Thanks for the information.
<mark_weaver>you're welcome
<janneke>davexunit: yay: '(("-g" "postgres" "-c" "PostgresSQL server user" "-d" "/var/empty" "-s" #t "postgres"))
<jmd> is going to need to be redesigned if we get too many more packages. It takes too long to load already.
<janneke>so, shell is #t
<janneke>and probably expected to be "/bin/bash" or something...
<janneke>now where's the `In unknown file' that issues the b0rked (system*
<dvc>davexunit: are you still working on the guix gui project?
<ng0>hihih... register account believing that cname is valid like before. find out that cname causes their system to choke. try to delete and use a different domain in place does not work because the cname is used as the base name for the account. the bugs you run into, yay...
<ng0>so hopefully I can receive mail again tomorrow.
<dvc>davexunit: writing a packagekit backend for guix would work with most package manager guis
<dvc>looking at the api that would have to be implemented, it's not made for guix that's for sure, but we could probably get some nice integration
<brendyn>kingcons: Change (gnu packages linux-nonfree) to (linux-nonfree), and then change (version "4.3") to (version "4.7.2"). That might work
<janneke> seems <user-account> got an extra entry
<janneke> (home-directory user-account-home-directory)
<janneke>+ (create-home-directory? user-account-create-home-directory? ;Boolean
<janneke>+ (default #t))
<janneke> (shell user-account-shell ; gexp
<ng0>I should get a third year subscription for free for discovering this bug.
<brendyn>What command can I run to download the source for a package?
<rekado>brendyn: guix build -S <pkg>
<brendyn>Ah ok, it's "build"
<mark_weaver>rekado: I didn't read the entire log, but there _are_ environment variables that openssl consults, iirc.
<mark_weaver>specifically SSL_CERT_DIR and SSL_CERT_FILE
<mark_weaver>however, there are none that are supported by GnuTLS iirc. GnuTLS needs a more complex (though probably better) setup to enable per-user CA certificate stores, which is something that Fedora does iirc. this is all from vague memories of me investigating this issue long ago. there's a thread on guix-devel about it years ago.
<mark_weaver>janneke: you need to "make clean-go && make"
<janneke>mark_weaver: i think i found a fix, but will try this too
<mark_weaver>when that field was added, it changed some macros related to that record in such a way that a rebuild is needed.
<mark_weaver>(I ran into the same issue)
<janneke>ahhh...i was composing a mail saying i don't understand what's going on
<janneke>i just wrote a patch that `fixes' i'm interested to find out what happens after a clean-go...sigh still not done ;-)
<mark_weaver>janneke: I'd be curious to see your patch
<janneke> testing without this patch after a clean-go && make
<brendyn>Hi mark_weaver. I've got guix build from git, set the symlink but when I run guix commands I get guix package: warning: failed to load '(build-aux build-self)':
<mark_weaver>janneke: if that worked, it might have simply been that it forced a rebuilt of the module that was causing you trouble.
<janneke>haha, possibly. it's fine now without the patch anyway
<janneke>mark_weaver: what can/should i do about my uninformed bug#24390: Acknowledgement (VM with postgres service broken in master)
<janneke>i.e.: solution: make clean-go && make
<mark_weaver>brendyn: does ~/.config/guix/latest/build-aux/build-self.scm exist?
<rekado>mark_weaver: yes, the variables are also documented in the manual, but libcurl does not respect them.
<rekado>libcurl users are expected to set up the bundle file by themselves.
<brendyn>mark_weaver: yes
<mark_weaver>rekado: I have noticed that 'git', which uses libcurl, also doesn't consult those variables. for 'git', you must instead set GIT_SSL_CAINFO, so maybe it depends on how the libcurl API is used.
<ng0>civodul: gitlab has a bunch of dependencies if you compile from source.. if guixsd is the target, someone else might want to take over working on go-build-system .. I'm new to the build system stuff but my basic system should be ready soon. we need this to not repeat setting GOPATH etc each time manually, and more. I have to reply via irc to this because my email hoster support has human working condition, no 24/7 support :) .. email is stuck
<ng0>in detail:
<mark_weaver>brendyn: what is the output of: readlink ~/.config/guix/latest
<brendyn>mark_weaver: /home/b/guix
<bavier>a gitlab service would be fun to get working
<mark_weaver>brendyn: and what was the precise command that led to that error?
<ng0>so if one put their mind to it, gitlab is relatively quick to package, but I already package 3 version control related packages and their dependency graph right now, a 4th... well I *can* do that, but I expect someone else to take this work from my shoulders.
<civodul>the mythical Someone Else :-)
<bavier>we all have one of those
<bavier>so there should be plenty to go around :)
<ng0>we have most of the dep. graph packaged. just missing: go build-system (for making it easier to package stuff, optional), and the gitlab specific packages.
<ng0>gitlab-workhorse uses go.
<bavier>I recently started using a sandstorm gitlab grain for a personal guix mirror, but would much rather run my own
<brendyn>mark_weaver: Sorry, I think I may have fixed it. I started I fresh terminal, unset LD_LIBRARY_PATH; unset GUIX_PACKAGE_PATH; ran guix environment guix, then I seem to be able to use guix just fine
<mark_weaver>brendyn: the only place where I see 'build-self' used is 'guix pull', but if you have that symlink pointing to your git repo, then you should avoid 'guix pull' anyway.
<brendyn>I think I had GUIX_PACKGE_PATH also set to /home/b/guix, and that borked everything
<mark_weaver>brendyn: it's best to avoid setting LD_LIBRARY_PATH, and it shouldn't be needed with Guix-built software in most cases iiuc.
<brendyn>I just did a git rebase to my dev branch and all of a sudden guix package -s praat finds my package :)
<mark_weaver>brendyn: ah yes, setting GUIX_PACKAGE_PATH to the 'guix' source tree does indeed lead to badness.
<brendyn>Only guile whines about updated .scm files, but that's not a real problem
<brendyn>LD_LIBRARY_PATH is set to /home/b/GNUstep/Library/Libraries:/usr/lib for some reason. No idea why.
<mark_weaver>guix tries to load every *.scm file in there looking for package definitions, and many of the *.scm files in the guix source tree are not meant to be loaded that way.
<mark_weaver>brendyn: ah, that's probably set via a wrapper for windowmaker. bah.
<mark_weaver>although I'm not sure why the /usr/lib is in there. that's not something we would have done, I think.
<janneke>i found we have haveged...but no haveged service
<mark_weaver>actually, neither of those are things we would have added in a wrapper. sorry for the false starts.
<janneke>who calls/how do you all call haveged upon VM creation or otherwise avoid the ssh keygen for VM hosts?
<brendyn>Ok, now I now I need to make Emacs+guile load things correctly so that I can use guix.el. Currently C . b fails because the guile repl hasn't loaded the requisite modules
<mark_weaver>brendyn: did you read the last bit of section 4.1 (Emacs Initial Setup) in the guix manual? iirc, I pointed that out to you earlier.
<mark_weaver>starting with "If you did not install Guix at all and prefer a hacking way"
<brendyn>Sorry I probably forgot due to being tired. I see that the sun has already come up... again.
<mark_weaver>heh, no worries :)
<ng0>janneke: usually I delete the pre-generated keys, generate my own with stronger sets and set immutable bits
<brendyn>I like programming but it really messes with my health. my sleep cycle goes whack when I spend too much time on the computer
<janneke>ng0: for throwaway test VMs?
<ng0>I'd be sad if I ran into a system which does not allow this
<mark_weaver>brendyn: I can relate to that problem :)
<ng0>no.. not for thiose. sorry, had no context
<janneke>i'm just getting annoyed by typing aseoutth aoeneu satoetuh hssnatothe usnstoaehh eusnsahosonuse haosnoseuh hsaheu haneuh nha all the time
<brendyn>One day, I will start a blog and write my own tutorial for Guix :x
<janneke>and want to take the step to scripted tests using VMs...
<mark_weaver>janneke: have you looked at Guix's "system tests"?
<mark_weaver>janneke: those are scripted tests using VMs.
<janneke>mark_weaver: i was looking for a pointer on those, heard about it. tnx!
<brendyn>mark_weaver: I already have guix.el loaded
<ng0>what.. o_O so i just received one email. what's up with for 7 hours not getting through to my new host but I can get out.
<brendyn>I evaluated the snippet at the bottom, but it didn't change the result
<mark_weaver>brendyn: hmm. I don't know then, sorry. maybe alezost would know?
<mark_weaver>he's our local emacs guru :)
<bavier>aseprite is weird, currently trying to build it with shared libraries. cmake isn't as amenable to command-line/shell fixes to linking problems
<bavier>many other packages failing the validate-runpaths phase with -DBUILD_SHARED_LIBS=TRUE
<rekado>woo, after more than half a year I fixed the arm-none-eabi compiler
<rekado>it now builds the firmware for my Axoloti board without problems.
<dvc>rekado: nice! what was the problem?
<brendyn>rekado: Now could you write a compiler for virtualbox so we can make it fully free software :p ?
<civodul>rekado: congrats!
<dvc>brendyn: what?
<rekado>argh… no, it’s not fixed after all :(
<rekado>I still had some cached stuff lying around
<rekado>bah, that’s sad
<rekado>I’m using the same sources, the same configure flags, … but the binary appears to crash the board.
<rekado>ACTION —> zzzZ
<brendyn>Hmm I'd read that. It says fixed.
<ng0>how do i find out the description libcs want to set/see? I have not finished packaging uclibc-ng, but much of it builds. now building a hellow world, how do i set it to use uclibc-ng instead of gcc?
<dvc>brendyn: interesting
<brendyn>but I don't think it's in the repo?
<dvc>rekado: I guess you've tried to compare the assembly?
<dvc>rekado: a jtag debugger should give a rough idea where it's going wrong...
<brendyn>dvc: It seems like it "only" requires someone to implement that part of the compiler and then virtualbox would be free
<dvc>brendyn: probably easier to port the bios to use a different compiler?
<dvc>and what's wrong with seabios, that's free
<janneke>hmm guix system: error: service 'file-system-/run/systemd' provided more than once
<janneke>i commented-out all services...
<janneke>i commented-out all services...
<mark_weaver>janneke: when building a system, services are automatically added for each file system in 'file-systems'.
<mark_weaver>and now, iirc, that /run/systemd file system is also added by elogind-service
<mark_weaver>and /run/systemd also used to be in %base-file-systems
<janneke>mark_weaver, yes changing %base-file-systems -> '() gives another error
<janneke>something changed
<janneke>elogind-service is missing then...hmm
<mark_weaver>it's probably a bad idea to omit %base-file-systems
<mark_weaver>/run/systemd should no longer be in there. the same commit that added it to elogind-service removed it from %base-file-systems, iirc.
<janneke>i simply have (cons* (file-system (device "guix") (title 'label) (mount-point "/") (type "ext4")) %base-file-systems) ...
<mark_weaver>janneke: maybe you ended up with 'elogind-service' listed twice?
<janneke>possibly, not sure where
<mark_weaver>civodul: any idea why janneke is getting "guix system: error: service 'file-system-/run/systemd' provided more than once" ?