IRC channel logs

2019-05-08.log

back to list of logs

<rekado>another problem on the Hurd:
<rekado>this segfaults: /gnu/store/dqlhjyvg0n8v1kdvwfpliqy46n7kpjqb-bootstrap-binaries-0/bin/tar cvfa /gnu/store/p278mqb3aa0xrkrrw4rg1fxbb19hbdyh-make-4.2.1.tar.xz --mtime=@0 --owner=root --group=root --sort=name make-4.2.1
<rekado>but when “--mtime…” is removed it will work fine.
<buenouan1>wait wait, so real Guix/Hurd work is happening for real right now finally?
<rekado>buenouan1: I’m just trying to cut out the bitrot.
<str1ngs>rekado: what vs of tar is that. that seems maybe a tar issue?
<str1ngs>err version*
<str1ngs>also what format is @0 ?
<nckx>str1ngs: That's the format taken by ‘date’ & C° for ‘0 seconds after the epoch’.
<str1ngs>so @0 is 1972 basically?
<buenouan1>rekado: godspeed
<str1ngs>or whatever epoch is
<nckx>str1ngs: date --date=@0 → Thu 1 Jan 01:00:00 CET 1970
<nckx>A Thursday, apparently.
<str1ngs>right, hmmm intresting
<str1ngs>what does $ date --date=@0 output on hurd?
<rekado>Wed Dec 31 19:00:00 EST 1969
<rekado>I should note that the Debian GNU/Hurd tar works just fine.
<rekado>it’s just our bootstrap tar that fails
<rekado>the Debian tar is version 1.30
<rekado>same for our bootstrap tar
<switchy>so I'm running a profile with gfortran@8, and I've been installing packages with a manual input substitution: --with-input=gfortran=gfortran@8. this works in most cases, but trying to install gmsh I'm getting a profile conflict that says gfortran@5.5.0 is being propagated from gmsh -- how can I debug this?
<str1ngs>rekado: I'm thinking it's an issue with the bootstrap build then. have you tried an strace?
<rekado>str1ngs: there’s no strace on the Hurd
<rekado>there’s rpctrace
<str1ngs>hmm or possibly an issue with glibc that the bootstrap tar ball uses. I case you can also try that none bootstrap guix tar works? if you have even reached that part of the bootstap?
<rekado>but it’s much harder to read
<str1ngs>err not tar ball. I just meant tar there
<rekado>it’s statically linked and cross-compiled
<rekado>I don’t have anything beyond the bootstrap binaries because you need the bootstrap binaries to build others.
<str1ngs>right that's what I thought, was just a guess
<str1ngs>based on the input's can you deduce the glibc version that's used for static linking?
<rekado>yes, I know the glibc version
<rekado>it’s 2.28, same as for Linux
<rekado>it’s the first version of the glibc that comes with all the Hurd goodies
<rekado>previously we had to use a heavily patched 2.23
<str1ngs>and local hurd glibc is 2.28 I assume?
<rekado>yes
<rekado>I think it may be a timezone problem
<rekado>using --mtime=/tmp is fine, --mtime=@0 is not
<str1ngs>ahh that would makes sense
<rekado>it’s reading timezone data (?) right before segfaulting
<pkill9>is there a downloader for bazaar repositories?
<ober>gpg --keyserver pool.sks-keyservers.net --recv-keys 3CE464558A84FDC69DB40CFB090B11993D9AEBB5; required by installation script hangs forever
<marlin1113[m]1>Try with root
<ober>am uid0
<ober>strace shows it waiting on reply after syn-ack
<dftxbs3e>that's your network then!
<ober>yeap. all of amazon's network then
<dftxbs3e>I had to execute the same command an hour or so ago, and it worked OK
<ober>awesome, works for you, must be a non-issue now :P
<ober>was able to pull it fine from other keyserver. soo. unless this is magically blocked
<dftxbs3e>Don't get me wrong, I'm not trying to deny that you have an issue, I'm just saying it probably has to do with your machine (network) rather than Guix or GPG
<dftxbs3e>or it could be sks-keyservers.net .. but it's a quite big keyserver so it must be under high availability maintenance
<vagrantc>fwiw, there have been lots of problems with keyserver networks lately...
<pkill9>anbox requires systemd :<
<ober>every vm I have access to gets a hang. so unless three major networks are blocked. :P
<dftxbs3e>ober: I just ran the command on my machine and it worked
<dftxbs3e>so I'm guessing it's a geo-related issue
<ober>cool give me a tmate :P
<dftxbs3e>?
<vagrantc>ober: you could also download the key from https://savannah.gnu.org/users/civodul
<ober>tshark shows it connects, I get the header of the keyserver version, then poll on socket awaiting reply from server
<ober>thanks vagrantc
<vagrantc>and manually import it, etc.
<vagrantc>make sure the fingerprint matches
<ober>easier to just change the trust model to all
<vagrantc>well, if you're downloading a specific key from a keyserver, then no matter where you get that key from, if it's the same fingerprint, it's the same key.
<vagrantc>so it doesn't matter where you get it from (other than who gets to see you're downloading it)
<vagrantc>the trust-model is irrelevent here; you need the key to verify the data. weather you trust that key is a separate issue.
<vagrantc>good luck!
*vagrantc waves
<ober>(= (bool? luck) #f)
*ober hunts for the process docs for submitting a pr for new packages
<sirgazil>What does the % in variable names mean?
<ober>in guile?
<ober>it's not special afaik.
<sirgazil>In scheme code used in guix and system configuration, for example.
<ober>got an example? I've not yet looked at the guix code. perhaps they reserve it for internal use
<ober>or do you mean in format strings?
<ober>e.g. (format #t "hello ~a~%" name)
<ober>looks like it's used for internal stuff. but has no special meaning in and of itself in guile/scheme
<sirgazil>ober: No, not for formatting. See the second code snippet here: https://www.gnu.org/software/guix/manual/en/html_node/Channels.html#Channels
<ober>yeah just saw some
<sirgazil>But I want to know the reason of using it :)
<ober>how hard is it to get new packages submitted via pr? as hard as brew?
<ober>like in c/c++ we have __foo(int bar;)
<reepca>in my experience it usually indicates a global variable
<ober>for private methods/vars etc
<ober>in scheme?
<ober>ear muffs in some lisps
<sirgazil>reepca: It makes sense...
<sirgazil>thanks
<reepca>for example: if you look at the (guix config), every single exported symbol is prefixed with a %, because it's all global variables.
<ober>is everything global scope?
<reepca>No, there's lexical scope. Global variables bound to procedures are usually not prefixed with a %.
<samplet>I think the “%” thing is pretty specific to Guix. Many other Schemers write “*global*” instead of “%global”.
<ober>yeah like CL
<ober>so if you have a new srfi for example, do you do a PR against the guix repo in git? or do you just host a srfi-xx.scm somewhere?
<reepca>*global* could be confusing for those used to common lisp, where they expect to be able to (let ((*global* x)) (proc-using-*global*)) and have proc-using-*global* use x. Doesn't work like that in scheme.
<ober>isn't that more a lisp1 vs lisp2 issue?
<reepca>no, it's a dynamic vs lexical scope issue
<dftxbs3e>samplet: hi!
<ober>oh but lets throw in lexical-let from elisp as the proper way to do it :P
<dftxbs3e>does guix have selinux support?
<samplet>dftxbs3e: You are here! I keep forgetting your nick. :p
<dftxbs3e>when ran on Fedora, guix-daemon wont start with selinux enforcing mode enabled
<dftxbs3e>samplet: aha sorry it's a random string..
<dftxbs3e>glad you're forgetting about it, that's what it's made fir
<dftxbs3e>for*
<samplet>I believe that SICP (or maybe the Lambda papers?) uses “*global*”, so maybe it’s a tradition at least as old as MacLisp.
<dftxbs3e>so you can't look me up from the top of your head :P
<samplet>dftxbs3e: That makes me feel much better about it. :)
<samplet>dftxbs3e: Since you’re here, I should give you a patch. However, I’m still not sure how I want to do it. I guess we should start by patching the Glibc build, and see what happens.
<samplet>Gimme a sec.
<dftxbs3e>samplet: some folks told me that only gcc 8.2 + supported float stuff correctly
<dftxbs3e>on powerpc
<dftxbs3e>usually you shouldnt have to set options for supporting it properly
<dftxbs3e>defaults arent correct
<dftxbs3e>so we have to set additional options
<samplet>dftxbs3e: Oh? That could be true. We aren’t really using much of it though, so maybe we can skirt by with this option.
<samplet>The 7.4 manual says the same thing about setting “-mfloat128”.
<terpri>ober, guix uses emailed patches for contributions, rather than PRs. https://www.gnu.org/software/guix/manual/en/html_node/Contributing.html#Contributing has the full details
<dftxbs3e>I'm getting a CONFIG_HEADER not found for sed tests about selinux, it tries to see if selinux is enabled for the sed build and fails because it can't find CONFIG_HEADER
<dftxbs3e>samplet: ^
<dftxbs3e>I set up a new system for trying builds and it's Fedora, it has SELinux
<dftxbs3e>I had to set it to permissive to run guix-daemon, otherwise it wouldnt work
<samplet>dftxbs3e: I know very little about SELinux. I know that Ricardo has worked on it, so there’s a bit of information in the list archives.
<dftxbs3e>samplet: okay then, I'll switch back to my old setup for building :P
<samplet>dftxbs3e: There is a policy for the daemon in the repo, but what you do with it or how it works is beyond me!
<samplet>Here’s where I think we should stick “-mfloat128”: <https://paste.debian.net/1082037/>.
<dftxbs3e>ohh I found SELinux in Guix manual!
<samplet>Even better!
<dftxbs3e>samplet: that's OK, and thanks for the patch! will try shorty
<brendyyn>I think in racket they write global%, which is consistent with other things like set!
<terpri>i have a lot of standalone package files (managed with guix package -f outside of the guix repo), maybe i should write a guide on getting those kinds of additions into guix proper, if one doesn't already exist
<dftxbs3e>samplet: finished re-compiling with patch in! Now trying patch
*samplet crosses fingers.
***anonnn is now known as anon
***anon is now known as Guest34371
<dftxbs3e>samplet: I think the change triggered a world rebuild
<samplet>dftxbs3e: That’s probably true. I should have conditioned on the target. Let me fix that. It will be much faster than waiting. :)
<dftxbs3e>samplet: aha thank you for that
<dftxbs3e>I'm learning about Scheme and Guile
<ober>is there a way to manually start the guix daemon?
<dftxbs3e>ober: without a service you mean?
<ober>guix pull: error: failed to connect to `/var/guix/daemon-socket/socket': No such file or directory
<dftxbs3e>what is your system?
<ober>yeah in docker, it failed to find init system, and so is not launching daemon
<ober>gnu system in docker
<ober>a distro has no name :P
<dftxbs3e>okay well you can just run the guix-daemon
<dftxbs3e>it will block so use a new screen session, or bash jobs
<dftxbs3e>run it as root
<ober>or job control?
<dftxbs3e>yes, jobs
<dftxbs3e>guix-daemon &
<ober>yeah enofile
<ober>binary install script does not place a guix-daemon it seems.
<ober>found it, unlinked
***rEnr3n2 is now known as rEnr3n
<samplet>dftxbs3e: This should be better: <https://paste.debian.net/1082040/>.
<dftxbs3e>samplet: thanks
<samplet>There may be a typo, since I can’t easily test it.
<dftxbs3e>it compiled, running now, are you sure it's CFLAGS=-mfloat128? doesnt that replace the whole CFLAGS? - samplet
<samplet>dftxbs3e: I’m not sure – I followed the example of other packages. (My intuition is the same as yours.)
<dftxbs3e>samplet: can we somehow tell guix to print a build strategy? Like tell what it's planning to do
<nly>--dry-run
<samplet>dftxbs3e: I think “guix build -n” is the best we can do.
<dftxbs3e>ok thanks will try both
<dftxbs3e>same output
<dftxbs3e> http://dpaste.com/3PKAJYT
<dftxbs3e>quite a lot of packages
<samplet>FYI, “-n” and “--dry-run” are the same thing.
<samplet>dftxbs3e: Yeah, we are kind of starting over here.
<samplet>We are not rebuilding the whole world now, just the PowerPC world.
<ober>the docs reference pre-inst-env from the git repo. however on fresh checkout it's not to be found.
<dftxbs3e>ober: you must build for it to appear
<dftxbs3e>./bootstrap && ./configure --localstatedir=/var && make
<dftxbs3e>samplet: it's going to take days :S
<ober>ahh. not mentioned in the docs. thanks.
<ober>m4 errors ftw
<dftxbs3e>ober: I think it is, under "Building from git" or something
<ober> https://www.gnu.org/software/guix/blog/2018/a-packaging-tutorial-for-guix/
<dftxbs3e>you need to get build dependencies; I usually just run guix environment guix
<ober>has it as the very next step after a git checkout
<dftxbs3e>ober: well then that blog post is outdated; refer to https://www.gnu.org/software/guix/manual/en/html_node/Building-from-Git.html
<ober>ok. Thanks. have a binary version already installed. just need to test some new packages I'm adding and another one I'm fixing.
<dftxbs3e>pre-inst-env probably never was in the git directly so the tutorial was probably wrong from day one
<samplet>dftxbs3e: Yeah – Glibc is the worst possible package to have to adjust in terms of rebuilding. :/
<ober>so `guix environment guix` setups up a new context?
<ober>rtfm i will
<dftxbs3e>ober: In the tutorial it says "Follow the instruction from the "Contributing" chapter in the manual to set up the repository environment." - and that includes what I sent you, which includes running bootstrap etc; so the tutorial is correct
<dftxbs3e>ober: think of guix environment guix as; "I
<ober>ahh I see
<ober>yeah I was following that, but m4 errors struck and since I've already got it installed, and just need to test a modified package...
<dftxbs3e>'m a developer, I want to hack on guix, so prepare the environment for me to build guix, it will download and install all necessary dependencies and then spawn a shell with environment variables set proper for you to start working"
<ober>pkgsrc it is
<ober>thanks
<dftxbs3e>samplet: heh, would be nice to be able to test non-reproductably
<samplet>dftxbs3e: For sure! Maybe we should just patch the static Glibc for now, and see what happens.
<dftxbs3e>samplet: what does the bootstrap-tarballs contain exactly?
<dftxbs3e> https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/bootstrap.scm#n590 ?
<dftxbs3e>libc, gcc, binutils, coreutils, bash
<dftxbs3e>that's all?
<samplet>Guile, GCC, Binutils, Glibc, and “Coreutils and company”, which is...
<samplet>...something like Tar, Gzip, Bzip2, Xz, Patch, Coreutils, Sed, Grep, Gawk, and Bash.
<samplet>Hmm.... I think Make should be in there, too.
<samplet>I don’t see it, though.
<ober>warning: 'https_proxy' is ignored.
<samplet>Maybe Make is built with a bootstrap Shell script.
<dftxbs3e>samplet: I'll try constructing a bootstrap tarball from a powerpc64le machine directly
<dftxbs3e>I have an idea
<dftxbs3e>I could compile guix
<samplet>dftxbs3e: The source file that defines the bootstrap binaries is “gnu/packages/make-bootstrap.scm”.
<PotentialUsurper>samplet: the first thing that gets build with the bootstrap bins is GNU Make
<PotentialUsurper> https://www.gnu.org/software/guix/manual/en/guix.html#Bootstrapping
<PotentialUsurper>^ see right below the graph
<samplet>PotentialUsurper: Thanks! For some reason I thought it could not be built without itself, but I guess I was wrong.
<PotentialUsurper>samplet: np I've been digging through the docs trying to wrap my head around this thing
<PotentialUsurper>speaking of which, can anyone point me to any additional documentation or complete examples demonstrating the construction of manifest objects?
<samplet>dftxbs3e: This is possible in theory, but I worry you are trading machine time for human time. Let Guix and your computer do the work for you! ;)
<dftxbs3e>samplet: it's building in parallel so I'm just having more fun :P
<samplet>PotentialUsurper: See the manual section on “Invoking guix package”.
<PotentialUsurper>yeah the second example there isn't clear to me
<PotentialUsurper>what is "specifications" in that context, and how is it defined
<samplet>PotentialUsurper: A specification is what you would write on the command-line, like “foo”, “foo@1.2.3” (specific version), or “foo:gui” (other output).
<samplet>It’s the shortest way to write a manifest file.
<dftxbs3e>samplet: would be so much easier if guix had a switch "--bootstrap-with-system"
<samplet>dftxbs3e: Yeah, but that kind of flies in the face of everything Guix stands for.
<PotentialUsurper>samplet: does that work? it looks like the -m flag is expecting a scheme program, not just a bunch of strings
<samplet>PotentialUsurper: Here’s what the complete Scheme program looks like: <https://paste.debian.net/1082042/>.
<dftxbs3e>samplet: Right; I get how. I'm compiling guix from source on powerpc64le right now, and then I'll try running the compilation for bootstrap binaries there
<dftxbs3e>On the other end, guix is still cross compiling bootstrap binaries on my low performance x86 laptop
<dftxbs3e>my Talos II has 64 threads
<dftxbs3e>which would make build 16x faster
<dftxbs3e>even more, because per thread perf is better
<brendyyn>some builds dont really go much faster on a faster machine
<brendyyn>not sure why
<dftxbs3e>bootstrapping guile for example is single-threaded
<samplet>Although I’m sure with enough work to do, 64 build users, and “make -j” it would help quite a bit.
<PotentialUsurper>samplet: thank you! that module import is exactly what I was looking for
<brendyyn>i wish i had a tallos ii
<samplet>PotentialUsurper: It should probably be in the manual (for both examples). I will make a note and try and remember to send in a patch.
<samplet>PotentialUsurper: Of course, if you wanted to contribute it, it would be even better!
<dftxbs3e>brendyyn: they're getting cheaper
<PotentialUsurper>samplet: yeah I'm going to see if guix supports my use case and then if it make sense I'll go through my notes and add whatever makes sense -- or blog about it -- or both
<brendyyn>dftxbs3e: i think ill wait until software is better ported. maybe there will be a talos iii
<dftxbs3e>brendyyn: sure, it's a more comfortable situation, but to me, all the fun is right now, porting software and all :D
<dftxbs3e>brendyyn: https://raptorcs.com/content/BK1B01/intro.html - $1,269.99
<brendyyn>dftxbs3e: i think im not very helpful there. sounds like id have to know how to program in C
<dftxbs3e>brendyyn: not necessarily, lots of work isnt C, look; we're porting Guix without writing any C code
<dftxbs3e>most of the stuff that would require writing C code is already ported or was portable
<brendyyn>dftxbs3e: what kinda changes to guix do you need to make?
<dftxbs3e>brendyyn: compilation flags
<dftxbs3e>Guile package definitions with more compilation flags for support
<dftxbs3e>samplet: ah well; I just compiled guix on ppc64le but it still requires bootstrap binaries to try to do anything so..
<ober>how does one make use of a caching squid setup if https_proxy is ignored?
<samplet>dftxbs3e: You would have to carefully build statically-linked versions of all the bootstrap binaries to make it work. :|
<PotentialUsurper>anyone aware of any work on using svn URIs for source origins?
<samplet>PotentialUsurper: We have an “svn-download” module. I’ve never used it, though.
<PotentialUsurper>oh heythere
<samplet>PotentialUsurper: See “guix edit libsmpeg” for an example.
<PotentialUsurper>yeah just dug it up in the guix source
<PotentialUsurper>another thing to add to the reference manual
<crab2>Hey, I'm trying to write a package and I need to clone one git repo inside another before building, does anyone know how to do this?
<crab2>as in normally to build this you run "git clone subproject subprojects/subproject" "meson build" "cd build" "ninja install"
<brendyyn>you want to download stuff during build?
<samplet>crab2: You could download the supbroject in another package and then copy the files over.
<samplet>You can even include a second “origin” record in a package’s inputs.
<samplet>E.g., (inputs `(("subproject" ,(origin ...)) ...))
<crab2>copy the files over how?
<samplet>crab2: Add a phase using “modify-phases”, then use the “copy-files-recursively” procedure.
<samplet>Look at the “editorconfig-core-c” package for an example.
<samplet>It downloads tests from another repo.
<samplet>crab2: ^
<dftxbs3e>samplet: oh, by the way, you said you couldnt easily test the changes, but you can? I am currently cross compiling from an x86_64 machine, so it's just whether you do or not want to occupy computing resources to building this, what machine do you have?
<samplet>dftxbs3e: Mostly it’s because I haven’t built Guix on the core-updates branch, which would take a while.
<dftxbs3e>samplet: okayy, well how does GuixSD CI works?
<brendyyn>crab2: you can refer the the locations of all the inputs in a package using assoc-ref. look through packages for the many examples
<samplet>dftxbs3e: Can you be more specific? I can tell you it’s called “Cuirass” and that it basically just builds all of the Guix packages and records their statuses and logs in a database.
<dftxbs3e>samplet: does it somehow build things any faster? that is my question
<samplet>dftxbs3e: Yes! It distributes builds across multiple computers. So it can be as parallel as the build graph allows.
<dftxbs3e>also; I'll have to open a ppc64le substitute server when its bootstrapped
<dftxbs3e>samplet: this could be used as a substitute right? http://ci.guix.gnu.org/jobset/core-updates-core-updates
<samplet>dftxbs3e: Yes, but it is incomplete, and IIRC it was missing some important substitutes.
<dftxbs3e>samplet: I wonder how much it would cost to run CI infrastructure against an arbitrary branch of Guix
<samplet>dftxbs3e: As in, how much money would I have to give, say, Rackspace to build all the Guix packages quickly for me?
<dftxbs3e>samplet: yes
<samplet>I would like to know that, too. It would be a fun experiment.
<dftxbs3e>samplet: I use https://bithost.io/ some times, it is a reseller for Digital Ocean that accepts Bitcoin
<samplet>dftxbs3e: That’s pretty cool.
<wr>trying the guix install again, last time logins didn't work and i could not run install on MATE
<wr>oh no https://imgur.com/96YbJjf
<nly>sometimes `--cores=1 --max-jobs=1` can help *some* builds go through
<wr>nly, only used 1
<nly>which commit are you on? `guix describe`
<nly>gdm is not broken on 952afb6f8c209692e52f9561965ee39e143e1d88
<nly>`guix pull --commit=<commit>` on that one, and it should work
<nly>wait that's a wrong commit
<crab2>Is it easy to send patches for guix without git cloning the whole source? My usual process is git clone, git add, git commit, git send-email but that kinda needs you to clone the whole big thing
<nly>952afb6f8c209692e52f9561965ee39e143e1d88 looks right
<nly>clone? do you delete it afterwards? (900mb right?)
<crab2>I've never tried to contribute to such a large project
<crab2>cloning is normally fine for little ones
<crab2>I'm not sure what workflow to use for something too big to clone
<nly>you clone once, then pulling is cheaper
<nly>git pull right?
<crab2>can you clone a portion of a repo? like just one file?
<nly>nope
<nly>git clone --depth=1?
<nly>having the history has many advantages tho
<nly>where do I send patches for a golang-importer?
<reg[m]>> nckx: Thanks for letting me know. I think I'll try a reinstall with US QWERTY just to rule this out (or not).
<reg[m]>It didn't help. Still stuck at boot. Also when pressing e or c the countdown disappears but then I'm stuck on the grub menu.
*reg[m] uploaded an image: IMG_20190507_215837.jpg (83KB) < https://matrix.org/_matrix/media/v1/download/disroot.org/AJJajLeYXXjMPWLIpfFrJtfJ >
<reg[m]>I also got some errors at the end of the installation process, as you can see on the screenshot.
<reepca>I don't suppose anyone would know why loading a dynamic module from emacs would return t, but not provide any of the procedures that are supposed to be available?
<crab2>I'm noticing a couple things that are bothering me. First, I have to add pkg-config as an input to anything that uses the meson-build-system, should that be made so it's just there without adding it since all meson packages will need it?
<efraim>If all the packages using the meson-build-system use it then it might be worth making it a default native-input
<crab2>Second, it doesnt seem that meson can ever see the current wlroots even if it's added as an input and I can't figure how to circumvent the issue. I know swaywm does it somehow, but I can't figure out how and have no idea if it's a problem with the wlroots package or if that's just how it's gotta go
<rekado_>wr: make sure to “guix pull” first.
<rekado_>wr: I fixed mate after the release.
<crab2>and I believe meson uses pkg-config to check for the presence of dependencies so it should be a necessity that every package using meson has it
<wr>rekado_, elucidate me more on that “guix pull” first.
<wr>rekado_, how much time it takes to install guix on real hardware?
<rekado_>crab2: you may have to look at the package’s build system files. Some build systems assume invalid locations of dependency files.
<rekado_>dftxbs3e: we have an SELinux policy for the guix-daemon (it’s in etc/guix-daemon.cil in the git repo).
<rekado_>dftxbs3e: you need to use “semanage -i etc/guix-daemon.cil” to install it, then relabel your file system, and then turn on permissive mode to check if any changes are needed.
<rekado_>I wrote the policy file last year on Fedora; I don’t know if it will work with newer versions of Fedora.
<dftxbs3e>rekado_: hi, I did apply the policy and relabled but then systemd wasnt allowed to run the guix-daemon - so I set back the policy to permissive to use guix, and set to enforcing when I'm done using guix.
<rekado_>ober: % just means “special”.
<rekado_>dftxbs3e: does the audit log tell you why it’s not working?
<rekado_>wr: please take a look at the manual. “guix pull” is how you update Guix.
<wr>rekado_, guix mate fail is on install
<wr>rekado_, on the installer
<rekado_>wr: you need a more recent version of Guix is what I’m saying.
<rekado_>dftxbs3e: someone sent an email to the mailing list this week about changes to the policy.
<wr>rekado_, what do you mean by more recent? i just download iso on site
<wr>*downloaded
<rekado_>wr: unfortunately, Mate was broken on the 1.0.0 release, so you can’t install it with the installer.
<wr>rekado_, ah ok, so why is it on the installer?
<dftxbs3e>rekado_: yes it says that it does not have permission to execute guix-daemon (systemd)
<rekado_>more recent = something newer than the last release. You can get it with “guix pull”
<crab2>where do I go to look at the definitions of the build systems?
<rekado_>dftxbs3e: hmm. Anything else?
<rekado_>crab2: in guix/build-system/
<rekado_>wr: because this wasn’t caught during testing.
<crab2>thanks
<dftxbs3e>rekado_: that's all - audit says AVC denied { search }
<dftxbs3e>search == execute
<dftxbs3e>when relabeling I had few conflicts
<wr>rekado_, so i run the installer to install MATE and the install of GUIX fails because MATE fails on install and MATE is broken, sweet..
<wr>rekado_, next time i install GUIX with nothing, it is best way....
<rekado_>dftxbs3e: yes, the conflicts are because of deduplication. /gnu/store/.links contains hardlinks to other files in /gnu/store and the policy can’t provide rules for stuff in /gnu/store/.links
<rekado_>wr: ?
<wr>it's taking more than two hours too
<rekado_>wr: I’m telling you that MATE won’t be installable with the installer image.
<wr>rekado_, no DE
<rekado_>wr: what takes two hours?
<wr>rekado_, i got that at first time
<rekado_>?
<wr>rekado_, GUIX install on my vm takes bit time
<rekado_>how much memory does the VM have?
<wr>1024mb ram
<rekado_>I’d recommend at least 2G
<rekado_>1G is difficult, unfortunately
*rekado_ –> afk
<wr>with 2GB will make it faster?
<dftxbs3e>rekado_: I don't really care much for SELinux compatibility so I'll just run with permissive if I really need it, I have a VM with Guix instead. Thanks for the SELinux policy and if it ever starts working again on later versions I'll apply it.
<wr>rekado_, never failed on 1GB just takes a while
<wr>i mean, it failed, but it runs on 1GB not memory trouble :/
<wr>i installed GUIX like 50 times taking 3 hours each install but i'm sure this time is gonna work
<wr>150 hours on GUIX install was less time that probably took to make the distro
<wr>maybe not...
<dftxbs3e>am I getting this right that you are mocking GNU Guix?
<dftxbs3e>If so; it's inappropriate, it's a 1.0 release, but the developers are doing their best to make it work, sorry for the issues but if you have problems, please provide the developers with debugging information so they can solve them and then; please be patient until they get solved. Thank you.
<buenouanq>I unironically think the [now `manual'] guix install is one of the slickest things ever
<wr>rekado_, i think i3 is broken too after install
<dftxbs3e>wr: are you using VirtualBox?
<wr>dftxbs3e, this time yes
<dftxbs3e>wr: I was having issues previously and I was told that VirtualBox graphics drivers are not supported. Did you try Gnome Boxes?
<crab2>I've never had that many problems, installing GuixSD has always been a breeze
<lisbeths>So I have gotten fairly good at installing guix now though I still have a few issues
<lisbeths>First of all I continually get this error that says guile: warning: failed to install locale
<lisbeths>and also I have trouble rendering gnu icecat correctly
<dftxbs3e>lisbeths: for the locale, did you install glibc-locales?
<dftxbs3e>are you running Guix on another distribution or you are running the Guix GNU System?
<lisbeths>I am on xubuntu
<wr>dftxbs3e, by the miracle of the free softwares i installed this and it works on GNOME!
<dftxbs3e>wr: Works on GNOME Boxes? Wonderful!
<lisbeths>I think there should be an environment variable that lets guix know if it is running inside of a guix distribution and if it is not it should automatically instlal glibc-locales
<dftxbs3e>lisbeths: Alright, so for the locales problem, you somehow need to pass an environment variable that should be given to you when you install glibc-locales, to the guix-daemon, Xubuntu should be systemd, so you would have to edit the service file for that.
<lisbeths>well right now I am doing a guix install glibc-locales
<dftxbs3e>There's a directive, "Environment"
<dftxbs3e>Okay, when you're done, it will tell you what to do
<wr>dftxbs3e, but it fails on xfce
<lisbeths>ok I am ready
<wr>dftxbs3e, it works on awesome too
<dftxbs3e>wr: How does it fail? What DEs have you tested with GNOME Boxes? XFCE: Fail - Awesome: Works, what else?
<dftxbs3e>lisbeths: so now, do you see that it's telling you to set an environment variable?
<wr>dftxbs3e, all
<dftxbs3e>wr: All DEs fail?
<lisbeths>I got a export PATH="/home/lisbeths/.config/guix/current/bin${PATH:+:}$PATH" and I did that already
<wr>no
<wr>have tested all distros on GNOME
<dftxbs3e>lisbeths: That's okay, isnt there something else?
<wr>dftxbs3e, ratpoison also fails here on the GUIX
<dftxbs3e>lisbeths: https://www.gnu.org/software/guix/manual/en/html_node/Application-Setup.html - Here it says about GUIX_LOCPATH, this is the variable you should pass through the systemd service unit
<dftxbs3e>wr: okay, so XFCE and ratpoison fail. Any others?
<lisbeths>How do I tell if it worked?
<dftxbs3e>wr: Could you run these commands and give out the .log files please? dmesg > dmesg-guix-DE-fail.log; cp /var/log/gdm/greeter.log greeter-guix-DE-fail.log
<crab2>I'm really getting stuck here. Meson's pkg-config step can't find wlroots and I can't fathom why. Can anyone take a glance? https://0x0.st/zTOe.scm
<dftxbs3e>lisbeths: You need to restart the service and then you can run a command that used to print "guile: warning: failed to install locale"
<dftxbs3e>"sudo systemctl restart guix-daemon" - lisbeths
<lisbeths>They need to add that command to the guide
<lisbeths>and then those three commands needed to be added to the installer.sh
<dftxbs3e>lisbeths: which command?
<lisbeths>sudo systemctl restart guix-daemon
<crab2>They can't because those assume systemd
<wr>dftxbs3e, after login?
<lisbeths>They can check for systemd and if it is not present then ask the user to do it manually
<dftxbs3e>wr: yes, you'll need to run these as root probably because the log files must be only readable by root
<crab2>That's still favoring systemd though
<dftxbs3e>lisbeths: Yes, there's probably room for improvement here; especially for locales, automatically manage it upon install, but it's not easy, you use Xubuntu, but others use various other distributions that would need individual support for them.
<dftxbs3e>lisbeths: If you want the best support and painless installation and usage, use the Guix system instead of installing on top of your distribution
<dftxbs3e>wr: when you're done, please upload the log files to https://paste.debian.net - Make sure to inspect these log files so you can make sure you're not sharing any personal information, and if so, please strip it from the log file and then send it. Thanks.
<wr>dftxbs3e, it's a vm, i don't care, you guys can see anything
<dftxbs3e>wr: Only making sure it's fine with you :)
<wr>dftxbs3e, will have to run one by one?
<dftxbs3e>wr: You can copy and run the whole thing, the ";" character will separate both commands
<dftxbs3e>wr: Then in the current directory, you should have both files with their names terminating with guix-DE-fail.log, send me those on https://paste.debian.net
<wr>dftxbs3e, i get an error that bash could not find command cp, weird
<dftxbs3e>wr: ah well; run: guix install coreutils
<dftxbs3e>then re-run the commands
<dftxbs3e>it's a known bug in the 1.0 version
<wr>lol, right
<wr>dftxbs3e, failed with coreutils too, what do i need more?
<wr>nevermind did the export path
<rekado_>wr: please see https://www.gnu.org/software/guix/manual/en/html_node/Guided-Graphical-Installation.html
<wr>dftxbs3e, https://paste.debian.net/1082052 https://paste.debian.net/1082054
<dftxbs3e>wr: thanks
<dftxbs3e>wr: that's still on VirtualBox?
<wr>dftxbs3e, rekado_ are those DE's gonna be removed on the installer?
<wr>dftxbs3e, yes, vm's rock
<wr>dftxbs3e, a native hardware install would be too easy
<dftxbs3e>wr: okay, the VirtualBox graphics driver is unsupported, that's why it's not working. If you used GNOME Boxes, it should work
<wr>dftxbs3e, i am a poor guy i don't have a pi
<wr>dftxbs3e, i can put another driver i guess
<wr>i won't pay
<wr>it's a vm ;)
<wr>dftxbs3e, should i report that to the vbox team?
<rekado_>wr: why would they be removed when they have been fixed already…?
<dftxbs3e>wr: no, VirtualBox is fine; I'll send the logs to the mailing list, or you can do it too
<wr>dftxbs3e, you can send
<wr>those 150 hours of guix install already killed me
<wr>and i give you full permission to use my dmesg with any scenario
<rekado_>it should not take 150 hours.
<rekado_>I suggest asking for help before the first hour passes.
<dftxbs3e>rekado_: they saying this ironically
<wr>this time it took the DE's was failing a bit...
<dftxbs3e>they're*
<wr>next time i put 2gb ram
<wr>rekado_, 150 hours of 50 installs because it did one on each DE
<wr>i lost the count how times i installed it
<wr>*of
<rekado_>wr: if you just want a VM with the Guix system you can download a pre-built image.
<wr>i love the fact that i do a command guix pull on the savannah on the gnu's lol
<wr>rekado_, ah i already did the qcow2
<wr>i did them all
<wr>qemu works flawlessly
<wr>but is slower than vbox
<dftxbs3e>wr: if you can get Spice graphics acceleration working under QEMU it will be as fast as VirtualBox
<dftxbs3e>It needs guest VM cooperation AFAIK
<wr>dftxbs3e, would have to see instructions on that
<wr> https://www.linux-kvm.org/page/SPICE ?
<dftxbs3e>wr: this is too specific to QEMU; Try looking in the GNOME Boxes user interface, there may be some settings, also, a package may need to be installed in GuixSD, it may even not be packaged yet, so you'd need to create it.
<efraim>sneek_: botsnack
<sneek_>:)
<efraim>sneek_: later tell civodul I ran `guix gc -R /gnu/store/x54z4746npss80i663ij5mgpvyyyq9sg-gobject-introspection-1.56.1` on bayfront and got no references, which seems to be the root of all of the build failure problems
<sneek_>Will do.
<efraim>sneek_: botsnack
<sneek_>:)
<boogerlad>How does Guix deal with packages with the same name and version?
<roptat>if it has different input, guix will create a different derivation wich leads to a different store path
<roptat>so in the store there's no confusion
<roptat>if you have two packages with the same name and version simulteanously available, I think guix will choose an arbitrary one
<roptat>when you use its name to refer to it, that is
<roptat>another possibility is to refer to the package by its variable value in guile, for instance in a manifest, so you're sure which variant guix will choose
<boogerlad>when you say arbitrary, is it simply the newest one?
<rekado_>packages are Scheme variables.
<rekado_>these package values have names that you can use to refer to them on the command line
<civodul>Hello Guix!
<sneek_>civodul, you have 1 message.
<sneek_>civodul, efraim says: I ran `guix gc -R /gnu/store/x54z4746npss80i663ij5mgpvyyyq9sg-gobject-introspection-1.56.1` on bayfront and got no references, which seems to be the root of all of the build failure problems
<rekado_>when package name and version are identical for two different variables you obviously cannot disambiguate them using just names and version strings, but you can use the variable name, which is uniquely identified.
<boogerlad>you are able to use the variable name in the command line like guix install program --with-input=dependency=guilevariablename ?
<civodul>efraim: that's a serious issue; problem is we cannot delete it because it's still alive somehow
*rekado_ –> afk
<civodul>boogerlad: when you type --with-input=A=B, both A and B must be "package specs"
<civodul>i.e., a package name, or name@version
<civodul>see https://www.gnu.org/software/guix/manual/en/html_node/Package-Transformation-Options.html
<boogerlad>that's what I thought.
<boogerlad>let's say I had a package like this
<boogerlad> https://pastebin.com/t2jJX1BP
<boogerlad>if I wanted to install mpv with that "march=native" ffmpeg dependency, how would I do it? the name and the version are the same as the original. If I give it a guile variable name, that only helps if I make a "mpv-optimized" package
<civodul>could you use say https://paste.debian.net? pastebin.com blocks Tor users
<boogerlad>ok
<boogerlad> https://paste.debian.net/1082060/
<civodul>so you would need to bind a top-level variable to that package, and to give it a different name
<civodul>so (define-public ffmpeg/optimized (package (inherit ffmpeg) (name "ffmpeg-optimized") ...))
<civodul>then you can do: --with-input=ffmpeg=ffmpeg-optimized
<boogerlad>ah I see. I was afraid of that. This would get out of hand quite quickly for a large number of packages
<boogerlad>thank you very much! if you have another suggestion, I'm all ears :)
<civodul>yes
<civodul>now, i'm not sure an -march=native variant is necessary for ffmpeg
<boogerlad>it's not, just an example
<civodul>see https://guix-hpc.bordeaux.inria.fr/blog/2018/01/pre-built-binaries-vs-performance/
<numerobis>Hi #guix! This is going to be a yes/no question. After using guix as a secondary package manager and trying it a couple of times in a VM, I am thinking of using guix system as my main distro. To smoothen the transition, I would first like to install guix system alongside my current distro first. To save space, it would be convenient to use the same /gnu/store and /home for both my current distro and guix
<numerobis>system. Assuming that /gnu and /home already have their own partitions, is this possible?
<ArneBab>this sounds interesting: guix package -i octave --with-input=fftw@3.3.5=fftw-avx
<PotentialUser-68>Hi guys, trying to test out guixsd, but not sure how i go about creating a custom xsession (would like to use exwm + emacs)
<ArneBab>PotentialUser-93: did you ask about exwm? If yes: I can’t solve that for you, but I would like to know how you solve it
<brendyyn>civodul: I notice that guix has (guix hash). Does guix not use your guile-gcrypt even? Also I have a question... If I add foo, open-foo-input-port, open-foo-port, port-foo, and file-foo for every hash algorithm, it comes to about 230 symbols. Is it reasonable exporting that many just when someone probably just wants to use sha256.
<g_bor[m]>hello guix!
<g_bor[m]>I have found a reproducer for the shepherd bug I found.
<g_bor[m]>I will file a bug report.
<g_bor[m]>I have seen the bug report address, to be bug-guix. Is that correct? Should I tag my bug report with shephed?
<amz3>o/
<amz3>brendyyn: what file are you refering to?
<amz3>numerobis: good question
<brendyyn>amz3: im talking about completing the bindings guile-gcrypt at https://notabug.org/cwebber/guile-gcrypt/src/master/gcrypt/hash.scm
<amz3>numerobis: I don't see where their would be a problem with that.
<brendyyn>amz3: of course, if you have advice on good ways to make guile modules, im happy to hear
<amz3>brendyyn: I read that code recently, my take on it is to follow an approach *similar* to gcrypt and pass the hash algorithm as a symbol
<brendyyn>amz3: so basically deprecate the bindings that are currently provided?
<amz3>numerobis: you should make sure /gnu partition is big enough, I am not sure how big is good. I guess it depends on how you use guix
<brendyyn>and do it like https://github.com/gpg-rs/libgcrypt/blob/master/src/digest.rs
<numerobis>amz3: Many thanks for your response, I'm looking forward to trying this. I'll see how big my /gnu folder is the moment and decide what size to allocate to the partition accordingly.
<amz3>brendyyn: yes something like that. You'd better keep the exposed procedures for backward compatibility but re-implement them in terms of e.g. (open-hash-port algorithm)
<brendyyn>I already made an open-hash-port so thats good ;)
<amz3>numerobis: you need to make sure to mount /gnu partition in the configuration file.
<amz3>numerobis: better experiment in a vm first. Ask around for the min /gnu store size
<amz3>numerobis: fwiw, my /gnu is 5G big and it is on top a foreign distro. I think 20G is a minimum
<pmikkelsen>hey guix
<amz3>pmikkelsen: o/
<brendyyn>amz3: and what i also mentioned is that these same bindings are also in the guix source tree.
<amz3>brendyyn: I guess it will be replaced with guile-gcrypt
<amz3>at some point. I looked up https://issues.guix.info/ but I found not bug related to this.
<brendyyn>amz3: thats alright people are busy with other things
<numerobis>amz3: Good idea thanks, I'll experiment in a VM first :)
<dutchie>I keep getting the "failed to install locale" warning, but I've done `guix package -i glibc-locales` and I have GUIX_LOCPATH set properly
<dutchie>I'm running on top of Arch, just installed the 1.0 binary distribution following the instructions in the manual
<runejuhl>dutchie: do you have a special locale set?
<runejuhl>dutchie: I got the same errors because I set LC_TIME=en_DK.UTF-8
<dutchie>it's en_GB.UTF-8
<runejuhl>I'd assume that would be supported, but you can try unsetting it and see if it makes a difference
*brendyyn tries to write scheme without dot notation
<dutchie>`unset LANG LC_CTYPE; guix pull` gives same warning
<amz3>dutchie: fwiw I have the same error x)
<runejuhl>dutchie: sorry, then I'm of no use -- unsetting my weird locale made the warning go away for me :)
<amz3>dutchie: it is gone
<amz3>dutchie: Here is what I have done. As root user, I installed glibc-locales and restarted the guix-daemon service
<amz3>dutchie: as a user I install the locales too and exported the GUIX_LOCPATH
<amz3>(actually it was already exported in my ~/.guix.sh file but...)
<dutchie>that's done it
<amz3>dutchie: no bug then?
<dutchie>so i guess you need the locales installed both for root and normal user
<dutchie>docs bug maybe
<civodul>brendyyn: Guix had (guix hash) until we switched to Guile-Gcrypt
<brendyyn>oh maybe its just because im on wip-build-systems-gexp its all old stuff
<bgardner>g_bor[m]: If you still wanted to check out my Postfix submission, the bug # is 35619
<g_bor[m]>thanks, I will have look at it time permitting. It's on my todo list, I will add the bug number.
<bgardner>Sounds good
<M4R10zM0113R>manual method speaks about using cryptsetup... What about headless crypt partitions?
<civodul>hello M4R10zM0113R
<civodul>M4R10zM0113R: that's a good question, i guess this is not really possible so far, unless you have IPMI-style remote console access
<nckx>M4R10zM0113R: If you mean something like dropbear in the initrd, that's not (at all) implemented yet.
<g_bor[m]>civodul: I've sent a quite small reporoducer to the shepherd problem I noticed. http://issues.guix.info/issue/35631
<M4R10zM0113R>:(
<amz3>M4R10zM0113R: do it :)
<M4R10zM0113R>amz3: pretty sure it could be done easily if I can make the kernel have an initrd and changing the mounts as mappers instead of regular scsi disks
<M4R10zM0113R>currently I have a setup where an usb is in charge of the booting, not that fast I know, but it certainly removes any clues how to decrypt from the computer itself
<rvgn>Hello Guix! Folks who are available or will be available, please share your thoughts about this bug (https://debbugs.gnu.org/cgi/bugreport.cgi?bug=35586) as a post at the same link. Thank you!
<M4R10zM0113R>(Not guix itself, though... But I wanna achieve the same)
<amz3>learn guile and diy :)
<M4R10zM0113R>Trying, but I can't even get a hold of where's the closing parenthesis lol
<amz3>in my editor, I have setup matching parenthesis highlighting along rainbow delimiters
<amz3>when you get used to it, you don't count parens, it becomes an instinct
<pmikkelsen>Do you have any recommendation as to which VPS providers offer a Guix VPS?
<nckx>M4R10zM0113R: How do you enter the passphrase?
<wr>guix repos server seems a bit faulty
<M4R10zM0113R>nckx: it's a key file, not a passphrase
<M4R10zM0113R>Usually appended it in the initramfs
<M4R10zM0113R>and had the boot partition and grub outside of the machine, in an USB
<nckx>What security does that provide?
<M4R10zM0113R>outsourcing the headers off the machine
<M4R10zM0113R>provide no clues in it
<M4R10zM0113R>Since the machines could run without rebooting, no need to carry the USB after initial boot
<nckx>That's a misleading definition of headless. Might as well hook up a USB keyboard if you can unplug a storage device 😛
<M4R10zM0113R>Headless as in no headers on the encrypted partition
<nckx>*headerless.
<M4R10zM0113R>oh
<M4R10zM0113R>Headerless
<nckx>Headless means something quite different (no ‘head’ = no screen & input devices) ☺
<nckx>Usually used to describe servers only accessible over a network connection (whether a remote serial port is headless is even debatable).
<nckx>M4R10zM0113R: Now, that's *also* not implemented yet, but at least now we know what you want. ☺
<civodul>g_bor[m]: thanks!
<M4R10zM0113R>Would I have to redefine linux.scm
<nckx>M4R10zM0113R: You'd probably start in (gnu system mapped-devices), which is where the LUKS decryption magic happens.
<nckx>(That's gnu/system/mapped-devices.scm.)
<M4R10zM0113R>Where's that?
<PotentialUser-65>Hello, I tried to install GuixSD, but during the installation I get this https://fs33.fex.net/preview/2268711505/0x0 what should I do?
<M4R10zM0113R>there's no /gnu/system in the fs
<brendyyn>M4R10zM0113R: In the git repo off gguiix
<M4R10zM0113R>o
<nckx>M4R10zM0113R: It's in the Guix source code, https://git.savannah.gnu.org/git/guix.git.
<wxie>Hi, if my disk has a system install and a free space, I want to install guix system in the free space, will it recognize my installed system and update grub automatically?
<nckx>PotentialUser-65: First, I'd retry in case that was just a network error (no way to tell from that output and retrying is cheap), then I'd take a look at that .bz2 log file (‘bzless <file name>’).
<nckx>wxie: No.
<nckx>Guix does no OS detection at the moment and there's no way (or code) to make it do so yet.
<nckx>You'll have to use --no-bootloader and add a ‘configfile …path/to/guix/grub.cfg’ to your existing GRUB installation, or add your existing distro to Guix's GRUB using ‘menu-entries’ in your system configuration (and that's as much detail as I know).
<wxie>nckx: Thank you.
<nckx>Having said that, I'm not even sure if --no-bootloader generates a grub.cfg at all. Not my area of familiarity, sorry.
<nckx>The second option is your safest bet anyway; one day you'd forget ‘--no-bootloader’ and overwrite your GRUB ;-)
<PotentialUser-65>nckx: The error is repeated in the following installation attempts, how can I call the terminal during the installation to enter "bzless <log file>"? The installer only offers the options of "Restart installer" and "Retry system install"
<wxie>nckx: if my guix installation fails, will I still have the other system?
<nckx>wxie: Yesss… but Guix will overwrite its GRUB unless you use --no-bootloader, so you definitely want to add a menu entry in your Guix ‘operating-system’ definition before installing it. It comes down to how familiar you are with GRUB, really. Guix doesn't detect & manage other OSes for you at all right now.
<M4R10zM0113R>on git repos do you get the directory hash or is there a way to retrieve a zip of the clone?
<nckx>PotentialUser-65: Can you use Alt+Fn (where n is 1…12) to switch VTs?
<nckx>Maybe Ctrl+Alt+Fn.
<PotentialUser-65>nckx: Here is the error from the log https://fs17.fex.net/preview/2268739298/0x0
<nckx>M4R10zM0113R: You get the recursive hash of everything in the directory (excluding VCS data like .git), or guix hash -rx . (See the ‘Invoking ‘guix hash’’ section in the manual, near the end.)
<wxie>nckx: Thanks. Is there a dual boot config.scm example for my case?
<M4R10zM0113R>oh
<nckx>PotentialUser-65: I'm afraid I've never used Mate, or the installer (great support here I know), but that looks a legitimate packaging bug. Is there any way you could install the system without mate-applets for now?
<nckx>wxie: Not that I know of. Basically, that area of Guix currently assumes you know how to configure GRUB yourself.
*nckx feels like they're in the disappointin' business today; and business is a-boomin'.
<wednesday>Is there any easy way to restart alsa/pulse? Wanna try do some stuff in my ~/.config/pulse/daemon.conf
*nckx → AF the K.
<dutchie> https://www.gnu.org/software/guix/manual/en/html_node/Building-from-Git.html#Building-from-Git I'm confused about what I should pass ./configure as localstatedir
<dutchie>I tried /var but then some tests failed
<wednesday>./configure --localstatedir=/var works fine for me
<civodul>yes that's usually the correct argument
<civodul>tests must be failing for some other reason
<wednesday>So is there no easy way to restart pulse?
<wxie>nckx: OK. You are very nice. I will figure it out.
<wednesday>well guess I gotta restart again to test pulse stuff
<wednesday>rip me
<PotentialUser-65>nckx: Yeah, I can just refuse to "Mate" with the installation, but I would like to solve the problem with the package, and not to abandon the package. To do this, I need to write here about the problem with the package https://lists.gnu.org/mailman/listinfo/bug-guix ?
<amz3>wednesday: probably using sheperd?
<nckx>PotentialUser-65: I'm not suggesting we ignore the bug, I suggest fixing it from an installed (if incomplete) system where you can ‘guix pull’ a bit more safely and report bugs a bit more conveniently. It's up to you, though, and yes, a bug report to that list (with the output you pasted above) would be very welcome! Thanks ☺
<PotentialUser-65>nckx: Okay, I get it, thanks for your help
<nckx>Eh. ‘Help’. I wonder where all the knowledgeable Mate/Guix users are today, if there even are any…
<PotentialUser-93>Hi guys, is it possible to define a simple xsession inside my config.scm?
<PotentialUser-65>By clicking on the links in the pinned post I noticed that https://bayfront.guixsd.org/.well-known/logs/ does not work, Firefox writes about the expiration of the certificate: SEC_ERROR_EXPIRED_CERTIFICATE , maybe there is someone who is engaged in the maintenance of this site and can update the certificate?
<rekado_>PotentialUser-65: thanks. I had to restart nginx on that host.
<sykloid_>Is there any way to set environment variables for use in `guix environment`? e.g. I'd like to run `guix environment -l foo.scm -- ./test.sh` and have some variables specified in `foo.scm` available to `test.sh`.
<rekado_>sykloid_: I think you could abuse the search-path feature for that by recording a search path to the package that “guix environment” evaluates.
<sykloid_>rekado_: That's what I thought, I see that `libressl` does that for `SSL_CERT_DIR`, etc. However, it isn't working just yet. I guess I'll just keep trying.
<sykloid_>e.g. `(native-search-paths (list (search-path-specification (variable "FOO") (files '("lib")))))` results in no extra line in the profile.
<charadon>Hey, so I'm trying to get Icecat working, but for some reason fonts aren't loading and all I see is squares. I'm using guix on debian stretch. Anyone know whats going on?
<brendyyn>charadon: icecat can't see debians fonts in its current form. fontconfig hard codes the path
<brendyyn>to find fonts
<brendyyn>you can install fonts via guix though
<brendyyn>or perhaps you could find a way to make it find debians fonts
<charadon>I did, I got font-liberation and font-bitstream-vera, but fonts still won't load.
<brendyyn>oh im not sure, maybe you need glibc locales?
<charadon>Got those also, It was throwing locale errors earlier, but after installing glibc-utf8-locales, the errors stopped.
<charadon>Only thing I got to go off of is that a font named pango can't be found or something.
<brendyyn>pango is a program for finding fonts
<charadon>Wow, it turned out to be pango the entire time lol
<charadon>I thought it was a font, so I kept looking for a font, man, I'm dumb.
<brendyyn>did you need to install it?
<charadon>Yep. After I installed it, fonts are loading.
<brendyyn>Sounds like a bug. Guix is a package manager and thus should install everything that is required.. or perhaps there was some environment variable conflict
<charadon>After I installed pango, it did tell me to set some environment variables, but I can say for certain it wasn't installed.
<brendyyn>icecat does provide panga as an input but not as a propagated-input
<charadon>Now I got a second issue. Icecat picked up adobe-source-code-pro fonts, but not liberation or bitstream, so I can't select them. Meaning i'm stuck with a monospaced font.
<brendyyn> https://www.mail-archive.com/help-guix@gnu.org/msg06474.html
<brendyyn>this title looks promising
<brendyyn>indeed it may have set some environment variables that are not yet available until you log out and back in
<charadon>Okay, let me relog.
<charadon>So, I can say for certain it isn't saving environment variables. But just out of curiosity, was I supposed to be using sudo while using guix?
<brendyyn>no sudo
<charadon>Alright, then I can safely say its not saving a single variable.
<brendyyn>id say its an environment variable thing. do you source $GUIX_PROFILE/etc/profile ?
<charadon>No
<charadon>I echo'd $GUIX_PROFILE but its blank.
<brendyyn>GUIX_PROFILE="$HOME/.guix-profile"
<brendyyn>It's just that string, you can set it, or not use it
<brendyyn>id try sourcing that profile and running icecat
<brendyyn>from that terminal
<charadon>Okay, so that did fix the environment variables, and I can just put that in my bashrc file or whatever its called, right?
<brendyyn>.bash_profile
<charadon>But, liberation fonts and bitstream fonts are still non-selectable.
<brendyyn>you don't want to put anything like this in .bash_rc
<charadon>Yeah, I mean't bash_profile, oops.
<brendyyn>anyway i have to go to bed
<charadon>Oh, well, thanks for the help.
<charadon>Atleast I can read icecat lol
<brendyyn>did you read that thread?
<brendyyn>and try things like installing fontconfig and running fc-cache -f
<charadon>Oh, I didn't even notice it was a thread.
<brendyyn>i think your answer is in there :P good night.
<charadon>Night, thanks for the help :)
<dutchie>i'm trying to build a package with (guix build-system gnu) but the tests are failing because they can't find /bin/sh, perl, or env
<dutchie>shouldn't those be pulled in automatically? or do they need to go into one of the `input` fields?
<nckx>dutchie: perl should go into native-inputs, sh and env are provided but the problem is probably not that they're missing but that the test suite tries to find them in /bin and /usr/bin.
<nckx>That'll need some patching.
<dutchie>ok, /bin/sh was definitely hardcoded
<amz3>I just installed icecat on top of ubnutu cosmic and it works.
<dutchie>ok, so trying to work out how to do the patching by looking at other package files. they all use the search-patches function but I can't see how it works or where the patches are saved
<amz3>dutchie: isn it patch-shebangs procedure that you are looking for?
<amz3>dutchie: in that case you don't need a patch, but on-the-fly substitution
<dutchie>yeah, i just found that
<dutchie>last test is failing because it's trying to write to /tmp/test
<dutchie>well, trying to write and then read from it and it doesn't exist
<dutchie>although interestingly it succeeds in one test case and fails in another, weird
<ober>all these .go files can be misleading
<rekado_>ober: how so?
<Apichat>Hi ! I've translate in French the publication announcement of Guix 1.0 maybe some of you could check it : https://linuxfr.org/redaction/news/gnu-guix-version-un-point-zero
<ober>rekado_: as if helloworld.go might infer that it's written in the Golang language
<sirgazil>The manual says: "one can use the ‘specification->package’ procedure of the ‘(gnu packages)’ module". I'm not using that module in my config.scm, but I do call that procedure, and reconfiguring the system didn't show unbound variable errors...
<rekado_>ober: Guile’s use of .go predates Golang.
<ober>sure.
<azdle>Hello from guix system!
<amz3>:)
<amz3>azdle: welcome
<azdle>Thanks, amz3.
<azdle>So, I had a lot of trouble getting guix system installed. Turns out it was because I had enabled the legacy bios boot stuff to install my previous distro. Is that a known issue?
<azdle>Several issues with the automatic partitioning not creating an EFI partition and later with grub-install failing.
<amz3>good to know
<azdle>Interestingly, after I manually created the efi partition once, it the guided setup would create one.
<azdle>(Going through the whole process from scratch again I mean.)
<samplet>dftxbs3e: Any news about “-mfloat128”?
<civodul>azdle: re EFI, it sounds a lot like https://issues.guix.info/issue/30311
<civodul>so you had UEFI turned off, hence the installer thought you were in legacy BIOS, but eventually booting failed, is that correct?
<azdle>It had both available, as far as I understand.
<civodul>ok
<civodul>so it tried to make a "legacy BIOS" install?
<azdle>The config.scm had grub-efi-bootloader set and the grub-install step had efi params, but the grub-install step would fail.
<azdle>IIRC it was complaining about a boot partition?
<efraim>one of these days I'm going to actually try to build enlightenment/wayland
<rekado_>bah, there are so many things broken with Guix on the Hurd, I don’t even know where to start :-/
<rekado_>built a bootstrap Guile 2.0 (because of https://issues.guix.info/issue/34427), but that Guile doesn’t work on the Hurd.
<rekado_>no crash, no output, nothing at all.
<rekado_>rpctrace randomly hangs, so debugging is really painful
<rekado_>looks like my Hurd session just ran out of memory…? Something’s leaky.
<rekado_>the tar segfault is also pretty damning.
***Piece_Maker is now known as Acou_Bass
<civodul>rekado: it's going to be a journey, be patient ;-)
<charadon>Hey all, so i'm having an issue, basically I installed guix on debian, and I installed icecat. It works fine, but I don't know how to make it integrate with my system. Like, for example, I have to load it using an sh script, and while that DOES load it, unfortunately whenever I open a link, it just opens a new window of icecat, but doesn't actually load the url. Basically, how do I get guix packages to integrate with the desktop
<charadon>environment?
<nckx>charadon: How are you passing the new URL?
<nckx>"$@"?
<rekado>civodul: yeah. I was really hoping to just pick up where Manolis left off, but it looks like we’ve regressed a bit more than I expected.
<charadon>The way XFCE normally does it. Which is taking the program (or in this case, script) and then throwing the url at the end.
<nckx>charadon: Could you paste your script to paste.debian.net?
<dmarinoj>civodul: maybe there were some debian patches in older versions and after the upgrade they were removed?
<charadon>Sure, but its an incredibly simple script lol
<nckx>charadon: I've used a simple wrapper script myself (exec env MOZ_USE_XINPUT2=1 icecat "$@") for yonks and have never had any weirdness.
<charadon>omg, I just added "$@" and it works now. =P
<charadon>Okay, just for future's sake, why does that work?
<nckx>charadon: "$@" expands to ‘all command line arguments (if any) of your script, properly quoted’.
<charadon>Ohhh, that's actually really good to know. Thanks
<nckx>If you just put ‘icecat’ in ./foo.sh, then running ‘./foo.sh’ will run ‘icecat’, but so will ‘./foo.sh --ignore all this "stuff"’.
<amz3>TIL
<charadon>Right. Yeah, i'm a complete newb when it comes to bash scripting haha
<charadon>Alright then, thanks dude.
<amz3>btw nice behavior added to 'guix environment guix' which spawns a new shell instead of changing var envs
*nckx is happy to have helped someone with the first answer that isn't ’that's a bug in Guix, sorry’ of the day 😛
<amz3>hey
<nckx>you
<amz3>nckx: better than "wait for someone knowledgeable" reply :)
<nckx>amz3: What else was I to say? I've never even *seen* a Mate desktop :-/
<amz3>nckx: I was talking about me.
<nckx>Oh ☺
<amz3>nckx: yeah it happens, not everybody can know everything.
<nckx>16:41 <nckx> Eh. ‘Help’. I wonder where all the knowledgeable Mate/Guix users are today, if there even are any…
<amz3>also my favorite reply is also "Have you checked 'dmesg'"
<nckx>amz3: I also… literally asked someone that today.
<nckx>We are all but replaceable by bots.
<amz3>:)
<nckx>Good thing sneek_ isn't a threat.
<amz3>sneek_: is dormant.
<amz3>sneek_: botsnack.
<sneek_>:)
<nckx>sneek_: No offence. I love you. You're just a bit special.
<amz3>not that much.
<amz3>nckx: there is bot in #scheme they reply what other have said recently, trying to stay on topic. it is funny.
<amz3>but doesn't always work.
<nckx>amz3: Doesn't that get spammy?
<amz3>nckx: they only reply, they do not speak by themself.
<amz3>nckx: also, there is much less discussion on #scheme so you can fool around sometime.
<nckx>As if to prove your point, turns out I've been in #scheme all this time and forgot I joined.
<amz3>I need to try guile-build-system, it seem straigth forward.
***apteryx_ is now known as apteryx
<amz3>guile 3 will require specific backends for architectures, I *might* on one backend, what is the most popular architecture in guix outside amd64 and 686?
<amz3>might +work+
<amz3>\cc efraim
<vagrantc>there are some people who do armhf-linux and aarch64-linux
<vagrantc>x86_64 i686 armhf aarch64 are all the architectures that currently ship a binary build of guix
<vagrantc>there's some support for mips*, powerpc (or ppc*?) ...
<vagrantc>there's work on i386-hurd i think
<pkill9>is ungoogled-chromium failing to build?
<donofrio>does this project require a kernel (for usermode use only - not standalone project)
<rekado> http://bayfront.guix.info:3333 displays *slightly* nicer logs now
<vagrantc>pkill9: i have rarely ever seen it succeed
<pkill9>i got it a while back, but i can't now
<civodul>rekado: nice!
<nckx>donofrio: I don't think anyone can answer that question without some more explanation.
<amz3>vagrantc: tx
<dutchie>`make check` on building failed, both times with the error "unshare failed: Operation not permitted"
<dutchie>should I just report a bug? or is something else causing the problem?
<pkill9>donofrio: do you mean does the Guix package manager require a kernel when running it on a foreign distribution?