IRC channel logs


back to list of logs

<castilma>can someone help me with adding a file to xorg.conf.d? I don't know, how this is supposed to work. with #:extra-config in (xorg-configuration-file) i can append a string to xorg.conf, but how do I create a file for xorg.conf.d?
<castilma>well, its a bit late. I will come back.
<catonano>sneek later tell ofosos you're not supposed to explicitly ask for a review of the patches you send to guix-patches. You might be particulary enthusiast or in a hurry; then I guess it's ok to prod people here on the IRC channel ;-)
<sneek>Will do.
<efraim>newick-utils test suite is taking forever on aarch64
<efraim>its been more than 12 hours, time to stop it and try again
<rekado_>Hi guix!
<rekado_>same situation as before: lots of offload processes, cuirass seems idle, all of the build nodes are idle.
<rekado_>next week: make cuirass more verbose.
<m-o>hi rekado_ , I also get cases where cuirass is frozen after evaluation failure, can't find why
<civodul>Hello Guix!
<m-o>hey civodul
<civodul>is Cuirass still hanging on berlin?
<castilma>does man xorg.conf work for somebody? i have xfce desktop installed, so I'd expect this to work.
<wingo>not for me
<wingo>ACTION runs gnome fwiw
<castilma>well, that's strange, right? and I don't see a doc output for xorg...
<wigust>castilma: Works for me with xorg-server package installed.
<castilma>wigust: in your personal profile or sytemwide?
<wigust>castilma: in system profile
<wingo>shouldn't matter tho whether it's system-wide or not
<ofosos>Anybody willing to build python2-pyicu? It fails for me and I'd like to know whether this is deplorable fate on my side or it's borken everywhere.
<sneek>ofosos, you have 1 message.
<sneek>ofosos, catonano says: you're not supposed to explicitly ask for a review of the patches you send to guix-patches. You might be particulary enthusiast or in a hurry; then I guess it's ok to prod people here on the IRC channel ;-)
<rekado_>civodul: no, it looks fine now
<rekado_>but cuirass is still silent.
<castilma>the -M option tells the daemon how many build jobs it should do in parallel at max, right? I passed it -M 8, but the second build job I started is waiting for free slots. what am i doing wrong? I have 10 build users set up.
<castilma>correction: it says waiting for locks or build slots.
<mb[m]1>castilma: That typically means the second build job depends on a currently running job.
<castilma>k, thanks
<g_bor>Hello guix!
<g_bor>rekado: are you here?
<g_bor>rekado_: I've found why hamcrest-core does not build on java8.
<g_bor>rekado_: it is a problem with jarjar. It is referenced here:
<g_bor>rekado_: it seems, that here is a well maintaned fork
<ng0>hey. I've never really used emacs-guix. Could it do things like M-x guix-hash filename and insert that at cursor point? because that's one of my most frequent annoyance I have to open 2 terminals for
<g_bor>rekado_: the problem is that jarjar uses a bundled copy of asm, which is too old. We can deal with this two ways: upgrade the bundled asm, either by patching or by using some newer version of jarjar.
<g_bor>rekado_: however I think it would be better to use the system-asm, like we do it in hamcrest with qbox. WDYT?
<g_bor>rekado_:it seems, that we need java-hamcrest-core for java-asm.
<roptat>g_bor: isn't hamcrest requered only to test java-asm?
<ng0>I'm doing the neomutt update. This update of neomutt will take some minutes more than usual, they've switched the build system.
<roptat>g_bor: so you could build java-asm-bootstrap (without tests), to build jarjar, then hamcrest, then asm (with tests)
<roptat>by the way, I think java-asm can be upgraded to 6.0
<jlicht>hello guix
<castilma>jlicht: hi
<castilma>does guile compilation take ages to compile on your systems, too?
<jlicht>castilma: guile compilation in general? Or a specific package using guile?
<castilma>jlicht: compilation of guile
<castilma>for some reason berlin.guixsd and hydra don't seem to have a substitute, so guix system build builds guile itself.
<jlicht>castilma: which commit are you on?
<jlicht>I pulled yesterday from master, and let my machine build for about ~7 hours while I was out. It was finished by the time I was done
<castilma>guix version gives me 9c2d2c13ce1f080dd8cff6a451625883fe6abab3. shouldn't be too new.
<castilma>how can i get more info for: # herd start user-homes
<castilma>Service user-homes could not be started.
<castilma>herd: failed to start service user-homes
<castilma>a previos system reconfigure gave me a bit more info: guix system: unloading service 'user-homes'...
<castilma>shepherd: Removing service 'user-homes'...
<castilma>shepherd: Done.
<castilma>guix system: loading new services: user-homes...
<castilma>shepherd: Evaluating user expression (register-services (primitive-load "/gnu/sto?")).
<castilma>shepherd: Service user-homes could not be started.
<jlicht>castilma: I am no expert, but afaik user-homes is responsible for creating home directories (which I am assuming you already have, so it doesn't need to be activated again)
<jlicht>I could be wrong though :-)
<castilma>m), thanks.
<castilma>but it would be cool if one could get more verbose errors for services.
<jlicht>castilma: I think a lot of recent additions to guix have improved the readability of error messages already, but services might still need some love indeed
***propumpkin is now known as contrapumpkin
<civodul>castilma: 'user-homes' is a service that's never marked as "started", on purpose
<civodul>that's confusing though
<ng0>*yep, it is confusing at first
<roptat>rekado_: could you continue your review of my java patches?
<castilma>how should one delete old system generations?
<civodul>castilma: currently it has to be done by hand: sudo rm /var/guix/profiles/system-XX-link
<civodul>one has to be careful not to remove the current one!
<ng0>civodul: is something like switch GuixSD generation (activate what you've just build after reconfigure) a goal for Guix 1.0 or after 1.0? Someone complained to me about this
<ng0>I don't know if someone is working on this or some work already happened
<castilma>ng0: isn't that already the state?
<ng0>to be more clear whatr I mean:
<ng0>the person complained about the requirement to reboot to switch into the new configuration build
<castilma>hmm, I had the impression this was already the case. probably because shepherd does restart some services.
<ng0>prove me wrong, but to my best knowledge a generation switch isn't happening. if you add a new service (like tor) it doesn't become available until reboot
<castilma>can someone check this in a vm: if you reconfigure into an os with a second file-system referring to e.g /dev/sda42 (a partition, that does not exist), the reconfigure hangs and the kernel complains. and rebooting is not possible anymore.
<ng0>hmm.. thinkpads are nice. just another one that survived a fall of 1 meter without much damage (at least from what I can see)
<ng0>damn electric wires
<ng0>the only thing that happened was a considerable amount of dirt came to light :)
<shiranaihito>i'm working on a system configuration .. do all "services" need a corresponding "package"? for example if i have an nginx "service", does it depend on an nginx "package"?
<civodul>shiranaihito: no, the nginx service will automatically pull everything it needs, including the nginx package
<civodul>so you don't need to add nginx to the 'packages' field, for instance
<shiranaihito>civodul ok, cool
<shiranaihito>civodul so the "packages" field would only contain stuff that doesn't have "services"?
<civodul>shiranaihito: yes, it's the set of packages that you want to have in $PATH for every user
<shiranaihito>well, it seems a bit unclear how it's supposed to be used
<shiranaihito>btw, The Last Jedi sucked :P
<thomassgn>so, herd throws me to a guile repl on boot because of an unresolved variable (I'm thought it was resolved) in the configuration of a service. Can I make shepherd ignore that service or variable and go on so i get into the full system?
<jlicht>shiranaihito: no spoilers please :-)
<thomassgn>if I feed ',q' to the repl it gives me a kernel panic because init (pid 1) fails
<shiranaihito>jlicht right, ofc not, unless knowing that it sucked spoils it for you :P
<shiranaihito>it just sucked so much that i felt compelled to comment :)
<shiranaihito>i think it's the worst SW movie ever, even worse than the first prequel
<M283750649[m]>ACTION sent a long message: M283750649[m]_2017-12-15_15:55:13.txt <>
<castilma>M283750649[m]: the uuid is most probably generated on disk/iso formatting. you can check every partitions uuid with fdisk. see, if any partition in the iso has that uuid.
<M283750649[m]>ACTION sent a long message: M283750649[m]_2017-12-15_16:19:18.txt <>
<M283750649[m]>fdisk give the same uuids
<castilma>if you look into the grub.cfg, you see that grub loads the debian iso from /boot/iso/*.iso. see which one fails to start and look into that iso for the uuid
<castilma>i.e. guixsd-install*.iso
<castilma>side note: the uuid has a lot of 3s. looks like that didn't came from a random source. might be a hint. maybe it was chosen by guix's deterministic iso creator?
<M283750649[m]>ACTION sent a long message: M283750649[m]_2017-12-15_16:53:30.txt <>
<M283750649[m]>ACTION sent a long message: M283750649[m]_2017-12-15_17:10:56.txt <>
<castilma>hmm. --root does not look like a linux bootparam (root without -- is one, though). this option seems to be passed on to the boot script in the initrd. or linux accepts them with and without --. and how would it boot without any params(they all start with --)
<M283750649[m]>my thoughts exactly... I really hit a wall here... My knowledge on boot process beyond grub is quite limited.
<castilma>mine too
<castilma>wild idea: maybe one fs on the install iso has another fs saved as iso file. and somewhere that image is put on a loop device and contains --root
<castilma>that would explain, why we don't find any fs with that uuid
<castilma>but not, why it doesn't find it
<castilma>ACTION asks why he didn't learn any of this in school
<M283750649[m]>from the backtrace, I know it fails at /gnu/store/dpg5vf8lw7qix0ywscj6xn5ddck7gbds-linux-boot.scm:389
<M283750649[m]>Maybe there's a clue here somewhere, but I'm very new to scheme and can hardly decipher it...
<castilma>canonicalize-device-spec sounds interesting
<adfeno>castilma: I would be very glad if people learn how to use revision/version control systems, even CVS or Subversion would be OK for me, just seeing people being taught some RCS/VCS is enough.... but this doesn't happen, and it should happen, even in non-programmer courses.
<castilma>adfeno: what subject would be responsible for teaching revision control, if not IT?
<adfeno>"Collaborative work'? or "Team work"?
<castilma>is that a new one or does that exist? never heard of anything similar in germany
<adfeno>In brazil we do have that, but unfortunatelly it's only seen in business/management courses.
<adfeno>... and currently focused on subject not related to computers.
<castilma>M283750649[m]: if you can solve the issue, please tell me how(via irc or mail). if not, you might want to ask on guix-help.
<quiliro>castilma: what issue?
<quiliro>why do i get command not found with recsel?
<quiliro>$ guix package -s emacs | recsel -p name
<quiliro>bash: recsel: no se encontró la orden
<quiliro>i want to find all emacs packages
<quiliro>be right back
<M283750649[m]>castilma: sure ;) and thank you for trying to help me out :D btw, found some docs on --root etc.:
<castilma>quick note: have you seen the warning on the bottom of the debian-multiboot page? it says you need another usb drive for installation. not sure if that's relevant here.
<castilma>quiliro: not my issue, just something im interested in, now. see here: . starts at[15:55:13]
<M283750649[m]>castilma: no I haven't.. hmm, it kinda defeats the purpose, I guess :P but yeah, maybe when you actually need to install. Nonetheless, I guess it should be possible to boot into the install environment -- it is at least with the debian-live*.iso
<apteryx>sneek: later tell quiliro Do you have recutils installed?
<sneek>Got it.
<CharlieBrown>AI detected.
<CharlieBrown>snape`: 0110100101100100
<CharlieBrown>sneek: 0110100101100100
<nckx>lfam: Yeah, I kind of forgot to write a proper cover letter for the lzip variations. I hope to use (some of) them in Guix if they perform acceptably.
<lfam>nckx: Cool :)
<alezost>ng0: re "M-x guix-hash" - currently there is no such a feature, but it's a great idea, thank you!
<ng0>ah, nice
<ng0>looking forward to it in future version :)
<adfeno>Hi all, quick question, can Guix+Shepherd services be used as substitute to Ansible?
<adfeno>Or have I misunderstood what these do?
<adfeno>As far as I know, Ansible is used for service configuration, so please correct me if I'm wrong.
<rekado_>adfeno: shepherd doesn’t compare to ansible.
<rekado_>ansible is a system to describe roles for machines and then apply them.
<rekado_>shepherd is just an init system
<rekado_>GuixSD with its operating-system configuration language comes closer to ansible in that you can describe the desired state of a system.
<rekado_>but it’s not close enough that you could just use GuixSD instead of Ansible.
<rekado_>ansible is used to manage a large number of systems. People often say “cattle vs pets” to describe this.
<rekado_>GuixSD doesn’t yet have the tools to manage more than one system easily.
<rekado_>ng0: reconfigure does switch to the new system
<rekado_>it just isn’t complete: some shepherd services need to be stopped before reconfigure, because they won’t be reloaded/restarted.
<rekado_>it also won’t hot-swap the kernel
<rekado_>but /run/current-system changes and will differ fro
<rekado_>*from /run/booted-system
<adfeno>rekado_: Oh, I see.... Thanks ;)
<ng0>thanks for the explanation
<g_bor>rekado_: I've checked why java-hamcrest-core does not build with icedtea8.
<g_bor>rekado_: It is a problem with java-jarjar. It uses a bundled asm, version 4, which is not working with icedtea8, so jarjar fails to relocate qbox, and we got missing classes in hamcrest.
<g_bor>rekado_: it seems, that using java-asm, which is version 5.2 would solve the problem, but it has hamcrest as a dependency, trough junit, but tests in asm are disabled anyway, so I propose to remove junit from asm dependecies.
<g_bor>rekado_: I've sent a patch for that in the mailing thread about switching to icedtea8 as default. Should I also send it to guix-patches?
<g_bor>rekado_: It seems that the patch is ok for inclusion in master and also in core-updates.
<rekado_>g_bor: what about the suggestion of building java-asm-bootstrap first and then java-asm again with junit?
<civodul>(guix man-db) :-)
<rekado_>to reduce build output I’d like to add a build phase that runs “make -n” to count expected work, and change the build phase to run make with a custom shell wrapper that increases a command counter for each executed command.
<rekado_>that might be good enough to get an approximation of remaining work for the build phase – when “make” is used.
<civodul>rekado_: the estimate would be the number of lines emitted by "make -n"?
<civodul>i wonder how well that'd work, we should experiment
<rekado_>yeah, I’ll play with this a little
<civodul>cmake augments each command in the makefile so that it invokes its progress tool
<rekado_>“make -n” would be used to count the number of commands that would be issued during the build.
<rekado_>then all we have to do is count.
<civodul>the question for me is whether we really get the same line count with and without -n
<rekado_>we might not, but it might be good enough to be useful.
<rekado_>I’d have to test this with more complicated builds.
<rekado_>I only tried it on my simple guile-debbugs package.
<civodul>yeah, but it sounds like a good idea :-)
<rekado_>civodul: do you think we could aim for another quick release before FOSDEM?
<rekado_>I’d like to close a couple of bugs and drain guix-patches a little
<rekado_>maybe land a simple implementation of channels, maybe integrate part of the GSoC work.
<civodul>rekado_: sure, we can try!