IRC channel logs


back to list of logs

<iskarian>bost, is the host OS or guest OS Guix?
<iskarian>(and if the host OS is not Guix, what is it?)
<bost>Under Ubuntu I start Guix using qemu-system-x86_64
<iskarian>are you able to successfully connect to the VM using spice?
<iskarian>(spice or virt-viewer)
<bost>iskarian: my /run/current-system/configuration.scm - contains (service spice-vdagent-service-type) right from the start
<bost>iskarian: no. Neither spice nor virt-viewer works.
<iskarian>Ah, I see the steps you took in now
<bost>iskarian: yep. BTW I also found the and I'm trying to find something usefull there.
<iskarian>Can you verify the service is running in the VM? `sudo herd status | grep spice` or similar
<bost>That shows '+ spice-vdagentd'
<iskarian>It also looks like you have a space before "name=com.redhat.spice.0"
<iskarian>that's why qemu complains "No such file or directory"; it's trying to parse it as a file to open rather than an option
<bost>iskarian: well the "name=com.redhat.spice.0" is written on a new line in the
<iskarian>Yes, but it should be connected: "-device virtserialport,nr=1,bus=virtio-serial0.0,chardev=vdagent,
<bost>iskarian: leme change it pls.
<iskarian>even IRC broke it up
<iskarian>but yes, it's supposed to be "chardev=vdagent,name=com.redhat.spice.0"
<bost>iskarian: Ok, except that I have multiple '-device' and '-chardev' switches is the qemu start command, and honestly I have no clue which are wrong and which not.
<iskarian>The rest of #3 looks fine to me
<iskarian>Is there still an error with the command?
<bost>iskarian: yes. still no shared clipboard.
<iskarian>It starts with the name= argument though! That's a step forward
<iskarian>So what is the command you are using to connect to the VM?
<bost> qemu-system-x86_64 \
<bost> -nic user,model=virtio-net-pci \
<bost> -enable-kvm -m $guixRAM \
<bost> -device virtio-blk,drive=myhd \
<bost> -drive if=none,file=$guixFile,id=myhd \
<bost> -device virtio-serial-pci,id=virtio-serial0,max_ports=16,bus=pci.0,addr=0x5 \
<bost> -chardev spicevmc,name=vdagent,id=vdagent \
<bost> -device virtserialport,nr=1,bus=virtio-serial0.0,chardev=vdagent,name=com.redhat.spice.0 \
<bost> & disown
<bost>Oh. Sorry.
<bost>Formatting. Hmm.
<iskarian>No, that's to start the vm; there should be a separate one to connect to it with virt-viewer
<bost>iskarian: adding '-spice port=5930,disable-ticketing' doesn't help
<bost>iskarian: ???
<iskarian>yes, that means the VM is now listening for spice connections. You still have to connect with e.g. `remote-viewer spice://localhost:4930`
<bost>iskarian: AAAAAH!!! Now it works!!!
<bost>Thanx iskarian!!!
<iskarian>Glad I could help :)
<iskarian>I suppose the manual could be a little clearer about it
<bost>iskarian: just for the record
<bost>iskarian: yes the manual could be a little clearer about it. Definitely.
<iskarian>Also consider using QXL graphics acceleration, if it's not using it by default
<iskarian>(with "-vga qxl")
<iskarian>However, I'm not sure if the guest drivers for that are loaded by default, or even packaged in Guix
<iskarian>oh cool, we do have a xf86-video-qxl
<bost>iskarian: the "-vga qxl" - that's a qemu parameter right? That doesn't produce any observable difference.
<bost>iskarian: '-vga xf86-video-qxl' returns 'qemu-system-x86_64: unknown vga type: xf86-video-qxl'.
<iskarian>Yes it is; the first one is correct; the xf86-video-qxl is an X11 module to load in your xorg-configuration, like so:
<iskarian>You can test if qxl is being used by running `lsmod | grep qxl` in the guest.
***j is now known as jess
<bost>lsmod | grep qxl
<bost>qxl 61440 1
<bost>drm_ttm_helper 16384 1 qxl
<bost>ttm 73728 2 qxl,drm_ttm_helper
<bost>drm_kms_helper 229376 3 qxl
<bost>drm 495616 6 drm_kms_helper,qxl,drm_ttm_helper,ttm
<bost>iskarian: seems to be working :)
<bost>Ok. I need to go to bed now. Thank you again iskarian!!
<iskarian>Welcome bost :) night!
***jess is now known as sandcat
<iskarian>What causes substituting to happen in "chunks", where it says "X MB will be downloaded" several times instead of adding it all together and doing it at once?
<the_tubular>I think they are just part of different packages no ?
<iskarian>Hmm, perhaps. I wonder what it would take to batch them together.
<cehteh>btw my laptop with xss-lock does still not lock the screen when one closes the lid, pressing the 'hibernate' button works inck locking the screen. Dunno what the issue is, but i am thinking about adding SIGUSR1 to lock the screen in xss-lock.
<cehteh>that should be pretty useful in other cases as well
<monkwitdafunk>Hello. I was waiting for gnu guix to release an iso
<monkwitdafunk>I am sure my own recipes can be added to the cookbook if it is outstanding in terms of ease of use and minimal maintainence
<monkwitdafunk>I wish to add a recipe for a person who uses guix for less than 1 hour on average
<monkwitdafunk>Or only 20 minutes per day
<monkwitdafunk>I spent half a day reviewing standard practise of info tech stewardship recommended by canadian universities on public domain and a hardware firewall is one piece but the obstacle is using libreboot
<monkwitdafunk>I can have access to support on weekends but i, am still unaware to what pertains to libreboot
<apteryx>monkwitdafunk: I do not understand what is meant by "a hardware firewall is one piece but the obstacle is using libreboot"
<drakonis>monkwitdafunk: but there are isos
<drakonis>if you want a upt o date iso
<monkwitdafunk>I dont have the luxury of my own modem anymore
<monkwitdafunk>apteryx: i only have 3 years in computer recycling. Never had i used proteanOS or libreCMC
<bricewge> cehteh If you are using elogind with default settings, the lid button suspend the system so xss-lock should receive an event
<bricewge>Unfortunately I don't have a laptop on hand to test it
<muradm>hi guix
<muradm>what is the next version for mesa for guix?
***char_ is now known as char
<monkwitdafunk>Is the list of modules run automatic?
<monkwitdafunk>The software is called kmod correct
***char_ is now known as char
<muradm>answer to me: should be 21.1.somoething
<cehteh>bricewge: yes strange, it doesnt seem to do so, but anyway i could have more use in other cases with the SIGUSR thing
<cehteh>(also thats not solving the lid thing directly, but hooking a 'killall -USR1 xss-lock' somewhere in would be useful in many cases)
***eichin_ is now known as eichin
<irfus>bricewge: wasn't the default setting *not to suspend on lid switch*, iirc
<bricewge>That wasn't the case when I read the code and locally the lid suspend “cat $(sudo cut -d= -f2 /proc/$(pgrep elogind)/environ)”
<abrenon>good morning
<cehteh>irfus / bricewge: i changed it to suspend on lid, that works but it doesnt lock
<cehteh>which is a bit odd because when suspending per keypress it suspends and locks
<bdju>if I want to play crawl with the tiles, do I need to install both crawl and crawl-tiles, or just crawl-tiles?
<bdju>ah I see in the crawl-tiles package definition that it inherits crawl. so probably just crawl-tiles is needed then
<irfus>cehteh: sorry if you've already mentioned this, but have you set the screen-locker-service?
<cehteh>hah prolly not, i check later (not at the laptop now)
<cehteh>guix need really document the neccesisties between services, thats very much lacking, leaving things that work somewhat but one dont know whats missing
<irfus>ah, then check the X Window services section in the manual, maybe that helps
<cehteh>also offloading is a works/works not/works/works not thing here, yesterday i tried package update and it failed, have to check today evening again
<cehteh>screen-locker-service is not part of %desktop-services?
<irfus>not per the manual, no
<cehteh>where in the manual is that mentioned? i am missing that
<irfus>cehteh: System Configuration->Services->X Window
<cehteh>i mean whats part of %base/%desktop -services
<irfus>there's a separate section for Desktop services that lists what's included there
<cehteh>missed that
<cehteh>would be nice if each services had some tage (%desktop) or (%base)
<jahor>Hello everyone! How do I pass a string with some special characters in the link like & to a url fetcher. It argues that & is not allowed :( Also I can't use parens when specifying name for keyboard layout like "by(latin)". Is it about Guix or Guile string in general?
<leoprikler>jahor example?
<leoprikler>keyboard-layout should probably use @, e.g. by@latin, IIRC
<leoprikler>either that or it's (keyboard-layout "by" "latin")
<jahor>leoprikler, thank! I'll try! Nvm about the uri. It's not about the link, but about the store. So my bad was that I didn't give it a proper name.
<Las[m]>Reading, it seems like the commits are supposed to be signed, but looking at the commits, I don't see any signature. How is the integrity of the source code verified?
<sneek>Welcome back Las[m], you have 1 message!
<sneek>Las[m], ng0 says: We have no docker package so far
<bost>Hi. Is there a command to merge my system configuration /run/current-system/configuration.scm with manually installed packages (e.g. `guix install vim`) and creates an scm file?
<bost>I've been googling for a while and found nothing, hmm.
<rekado>bost: what is the desired behavior?
<bost>rekado: my long term goal is to have a scm file containing all the packages (and their configuration) so that I can easily move between different machines.
<rekado>you can of course have a manifest and an os-config in the same file; you’d still need to use two commands to instantiate either of them.
<bost>rekado: well at first I need to learn what is a guix manifest...
<abrenon>is there something special to do to use the `leave` function ?
<abrenon>I see it's defined as a syntax rule, what does that entail ?
<abrenon>I get "source expression failed to match any pattern" messages trying to compile code that uses `leave` to print a message and exit
<rekado>bost: a manifest is a declaration of what packages you want to have installed to your profile
<rekado>it can look like this: (specifications->manifest (list "python" "emacs" "whatever@3.2"))
<rekado>then you instantiate it with “guix package -m manifest.scm”, which will install these packages (and *only* these) into a new generation of your default profile
<bost>rekado: Thx. I also found some docs about it
<bost>rekado: I see there's is the `guix package --list-installed` to see what I installed. That helps me to create the specifiction->manifest. I just wonder if there's some kind of `--format=channels` switch to generate the code for me.
<rekado>you can also generate the manifest
<rekado>use --export-manifest
<rekado>guix package --export-manifest
<bost>Oh. That was easy. Then sorry for my lame question.
<vivien>Hi! Is there a procedure to install guix on a librebooted thinkpad?
<vivien>I’ll try the normal way :)
<vivien>Thank you :)
<sculp-ns>hey all, new to guix + the declarative paradigm; I am trying to initialize an environment with a given python3 version (3.6.4), but I can't find a package entry for ["python@3.6.4" "python-3.6.4
<sculp-ns>when searching packages with `guix package --list-available='python-[0-9]'`, or similar registry search commands
<abrenon>if it's an older version maybe with guix time-machine to retrieve a state when this was the latest ?
<sculp-ns>from the 'version' docs, it seems as though there are just no legacy entries for python versions, and I will have to specify a package definition for each prior version of the lang
<sculp-ns>version docs:
<rekado>that’s right, we don’t keep older versions unless there’s a very good reason. You can still access them via the time machine.
<sculp-ns>abrenon; thanks for the suggestion - I assumed time-machine will only help if i have some record of the historic packages in my system history
<sculp-ns>I will try something like `guix time-machine -- build python-3.6.4`
<abrenon>no, I think it allows you to retrieve everything that's needed to build a previous state (prepare for some time though since I don't expect there will be substitutes available)
<abrenon>I have no idea how to invoke it but that seems optimistic, is it that simple ?
<sculp-ns>so is there no canonical protocol for referencing versions? I know in spec we have foo@1.a.3
<sculp-ns>abrenon; no, that command yielded `guix build: error: python-3.6.4: unknown package`
<sculp-ns>i will go deeper into time-machine docs
<abrenon>I would have loved it though but I had no real hope : )
<abrenon>good luck
<sculp-ns>haha thanks for the time anyway
<abrenon>you're welcome
<abrenon>I've got a general development question regarding the store: how much should it be used ?
<abrenon>I'm trying to retrieve .tar.gz archives in an importer, I can cache the archive itself with http-fetch/cached
<abrenon>but I don't know where I should extract them ? does their extracted content belong to the store ?
<abrenon>or to another cache folder next to ~/.cache/guix/http ?
<abrenon>I don't really know where to find hints about the possible changes of the archive so I guess I can't really generate unique hashes
<rekado>sculp-ns: time-machine takes a commit hash of the Guix repository or a channel definition file. It then looks up your packages in that alternative version of Guix.
<rekado>the version string itself is not important to us, because there are so many other things that affect a binary (the versions of all of its inputs, recursively)
<roptat>abrenon, I'd extract it to ~/.cache/guix/opam/<repo-name>
<rekado>you’d need to map Guix commit hashes to versions of interest.
<abrenon>thank you so much for you're advice !
<rekado>there was an Outreachy project to give you command line access to the history of Guix. I don’t know if it ever got merged.
<sculp-ns>rekado; thanks for the info - I understand the perspective from the repro goals of transparent inputs/outputs - the 'api' to getting a given build of a package by the channel's commit SHA is just very procedural
<abrenon>btw, the developer answered and it's not a deliberate choice, they didn't know about this structure recommendation and don't have strong opinions about it
<abrenon>anyway my latest-version finder works now so : )
<sculp-ns>I will play around with this mapping between package->version->[channel] and hit the mailing list if anything interesting comes up
<roptat>abrenon, that's awesome :)
<abrenon>do you know how to "pipe" the content of a port into a call to system* ? I'm passing the port-filename to the command but that's ugly considering the file is already open in reading
<roptat>why use system?
<roptat>it's best if you can do it in guile
<abrenon>cause I have no idea how to call tar either
<abrenon>probably, so module-ref and all ?
<abrenon>I've found a couple examples, which were doing gexpressions and all, hence my previous question about the store ^^
<roptat>use invoke, but why do you need to pipe a port? don't you have the filename in a variable somewhere?
<abrenon>hmmm more or less, using http-fetch/cached, I get an open port
<abrenon>so I'd rather use that
<abrenon>like I said, I can always port-filename it, but that seems clumsy
<roptat>oh, then I don't know
<abrenon>considering the port is already open and ready to be read
<abrenon>ohhh : (
<roptat>you mean you want to extract immediately, without writing to a file
<abrenon>invoke only works on filenames ?
<roptat>no, it's similar to system*, but we use it instead
<abrenon>(btw thanks ! I had seen invoke but I thought they were the same thing as calls to system*)
<abrenon>no, it's cached, so it gets written anyway
<roptat>I was thinking you could use (invoke "tar" "xf" filename)
<abrenon>sounds very similar
<abrenon>I'm sure there must be a way to pass a port as current-input-port or something
<abrenon>I'm pretty sure I've seen that in the past 72h
<roptat>I think you can't do that directly with system or invoke, so you have to fork and exec in the child
<abrenon>awww : c
***sandcat is now known as jess
<roptat>I tried "(define (cat) (set-current-input-port (open-input-file "test")) (execl "/bin/cat" "/bin/cat"))" but (cat) doesn't print the content of "test"
<roptat>so there must be another way to set the standard input...
<abrenon>I've checked and invoke seems to be a mere wrapper around system*, were you referring to something else when you said to do it in guile ?
<abrenon>perfect : )
<roptat>ah: (define (cat) (let ((port (open-input-file "test"))) (dup->fdes port 0) (execl "/bin/cat" "/bin/cat")))
<abrenon>dup->fdes is the new >&
<roptat>or with primitive-move->fdes
<abrenon>how did you find it ?
<roptat>not sure what's the difference
<roptat>you'd use dup2 in C for that, so I was looking for it in the manual
<abrenon>damn, I don't know enough C
<roptat>it's on the page you've sent though, so you'd have found it
<abrenon>only the first paragraph is sooo long
<abrenon>and I obviously can't skip it for now
<abrenon>anyway it confirms my first uneasiness when I started to discover guile: ports are not file descriptors, guile have more or less reimplemented their own interface for this ?
<roptat>now a fellow guile hacker will come up and tell us how stupid we are to use such low-level procedures when there is this other handy procedure we haven't seen before
<roptat>most languages do
<abrenon>I bet they will
<abrenon>but I have to produce something that at least works to receive such positive critics
<abrenon>ohhh right, execl, meaning my guile repl becomes cat, and then closes… : (
<abrenon>at least I could check it worked
<roptat>oops, sorry ^^'
<abrenon>hmmmm but that would also close the guix importer : /
<abrenon>hmmm I have to fork before
<abrenon>like you said
<roptat>that's why you need to fork
<roptat>(let ((pid (primitive-fork))) (if (= pid 0) (display "I'm the child\n") (display "I'm the parent\n")))
<abrenon>yeah, I expected as much
<abrenon>I haven't done much system programming but I remember that
<roptat>(again, looking for "fork" in the manual)
<abrenon>just didn't thought I'd need it to extract a simple tar.gz ^^
<roptat>you could close the port and extract the file it created, too
<roptat>mh... if you can find it, that is
<roptat>but the initial version of the importer used the tar gz too, how did we do it back then?
<abrenon>yeah, looking back it seems like the clever thing to do
<abrenon>hmmm I'm gonna check
<roptat>oh, we didn't
<roptat>(http-fetch/cached (string->uri ""))
<abrenon>ahhhh, text files, so easy ; )
<roptat>I know, right?
<abrenon>and url-fetc/tarbomb imports an ungrafted tar from the %store-monad
<abrenon>and invokes /gnu/store/…/tar on the retrieved tar.gz which is also a derivation in the store if I understand correctly #$drv
<abrenon>what are these scary %repository-cache-directory ?! they appear to be stateful Oo
<apteryx>caches are state
<maximed>abrenon: there is open-pipe*
<abrenon>yeah I've seen it and currently am struggling a bit
<maximed>it's less low-level than forking, redirectig and execing
<abrenon>I managed to open an open port
<abrenon>I get an input port from http-fetch
<abrenon>now I only have to understand how to plug them together : )
<maximed>Keep in mind that 'current-input-port' is a Scheme-only thing. It won't interact with file descriptors
<abrenon>ohh, I see
<abrenon>so that's why it was cruelly failing with all my attempts to dup->fdes and such ^^
<maximed>The port from 'http-fetch' is most likely a ‘custom port’, constructed by a procedure like 'make-custom-binary-input-port'
<abrenon>but is standard-input-port the same thing ?
<maximed>the port from 'http-fetch' probably does not have a file descriptor ...
<maximed>abrenon: is standard-input-port the same thing as what?
<abrenon>hmm it's created from a good ol' open-input-file
<abrenon>sorry, I meant the same thing as "current-input-port"
<abrenon>(I know, "the difference between a parrot", etc. ^^)
<maximed>abrenon: You're asking if 'standard-input-port' is the same thing as X, but you're neglecting to say what X is here ...
<abrenon>I tried to fixed this at 16:32:45, my sincerest apologies for it again
<maximed>FWIW, when Guile starts, it creates a port backed by file descriptor 0. (current-input-port) is initially set to that port
<maximed>likewise for (current-output-port) and fd 1
<rekado>this is a bit odd. I’m using “guix environment --profile=/gnu/store/…-profile --container” and Guix first builds bash even though it’s part of that profile.
<podiki[m]>hi guix! had fun playing on my new guix system yesterday, really enjoy using scheme for everything. finally a system I can understand (insert gif of jurassic park unix)
<podiki[m]>and of course Mesa releases a new version, but will probably wait for the first bugfix release to submit a patch. might try it on my own anyway (at least building X with it)
<vagrantc>since a lot of things depend on mesa, there might be a newer mesa in core-updates or staging
<roptat>there is a newer mesa in core-updates-frozen already
<abrenon>how does one modifies mtime
<roptat>touch -mt
<abrenon>I found set-file-time in guix build utils since
<roptat>oh, cool
<roptat>but why do you need that?
<abrenon>I need to mkdir -p in the path
<maximed>abrenon: there is the 'mkdir-p' procedure
<abrenon>does it allow to set mtime only if the folder had to be created ?
<maximed>Why does the mtime matter?
<maximed>presumaby mtime is only modified if a directory is actually created (mtime == modification time)
<abrenon>because I'm checking if the expanded folder is newer than the .tar.gz archive
<abrenon>I need to do the opposite
<abrenon>I need to put it back for when it's created
<maximed>I don't understand what you're trying to do
<roptat>(if (file-exists dir) (set-file-time ...) (mkdir-p dir))
<abrenon>sorry, I'm not making a lot of sense
<abrenon>step a -> retrieve index.tar.gz
<abrenon>that can fail, typically if the input repos isn't actually a repos
<abrenon>step b -> if and only if step a completed with success: extract index.tar.gz to a repos holding the corresponding repos
<abrenon>but we're not going to do it if we haven't really retrieved a new index.tar.gz because it hasn't changed
<elb>roptat: regarding GUIX_LOCPATH: that did it, thank you!
<roptat>elb, great!
<abrenon>so we start by checking which one of index.tar.gz and ~/.cache/opam/repo is newer
<abrenon>trouble ! you forgot to initialize ~/.cache/opam/repo -> no problem, I'll just mkdir-p it before the extraction, in the case when the retrieval actually yielded something
<abrenon>problem: the first time you use a repos: index.tar.gz will be created at t, and its cache repos at t+dt, so the empty dir holding it will be newer
<abrenon>so the first time the archive won't get extracted
<abrenon>we must confuse the system into believing the repo directory has always existed
<roptat>(begin (mkdir-p dir) (set-file-time ...)) :)
<roptat>set it to 0 or something
<roptat>and after you extract, set the mtime again to current time
<abrenon>exactly what I'm writing ! : ) thanks for the file-exist, I hadn't stumbled upon it yet so I was embarrassing myself with calls to stat with #f to check if it existed
<abrenon>won't the extraction do that ?
<roptat>not sure, maybe
<elb>elb: heh it looks like my config actually has that, but somehow my .bashrc and my .profile got out of sync, and so my currently-logged-in-guix is only half-configured. oops, fixed now :-)
<podiki[m]>(yes I know about mesa in core-updates-frozen, I worked on that. I just meant even before that gets done mesa has a new version)
<raghavgururajan>Hmm. system reconfigure warns about "could not do 'eval' on 'root' ..." something and system diesn't boot.
<radvendii>o/ hey all
<radvendii>I'm back to trying to figure out how to compile this opengl application with guix. With `mesa` in `inputs`, I get "opengl not found", with `libglvnd`, it compiles but then just shows a black screen.
<radvendii>looking online, it seems like depending on `libglvnd` isn't really supported, but it's hard to tell
<podiki[m]>mesa right now in guix is not built with libglvnd (not hard to modify it to do that though, there was a version of the current patch to update mesa that did it)
<podiki[m]>but sounds like it should work with just mesa, that provides opengl libraries
<apteryx>I think the eolie example in the manual is a bit wrong; it exposes /etc/ssl/certs only, but the certs can be symbolic links pointing to elesewhere in the store
<apteryx>ah, but I see nss-certs is added to the environment, so that would work. Nevermind.
<radvendii>podiki[m] yeah, that's what I thought. I'm using meson as a build system, which is using pkg-config to find opengl.
<podiki[m]>do you know what specific lib it is looking for? could check versus what is in mesa (since there are different names/versions/vendors for gl)
<podiki[m]> if you want to see versions of mesa with libglvnd, if you wanted to build and see if that helps
<podiki[m]>there is discussion about moving to libglvnd, seems like something comes up every day
<radvendii>I see some things like in `packages/gnu/gnome.scm` in the definition of `cogl`, which has me suspecting that somehow things need to be guided to find opengl
<podiki[m]>yeah, some packages have paths hardcorded to find opengl library
<podiki[m]>likely that's what you are running into, I'd guess
<podiki[m]>(and coincidentally those would need to be changed again if we move to libglvnd)
<radvendii>I don't know what specific lib it's looking for. I have control over the application itself. It's just a `dependency('opengl')` call in meson. Maybe i'll look into jumping over to `libglvnd`
<podiki[m]>might try a snippet like the one you sent, is probably common
<leoprikler>for the record, whatever cogl does, it seems very broken from gstreamer's perspective
<leoprikler>at least my personal experience
<podiki[m]>radvendii: maybe look at others too, but there are a handful of packages that need changes to where they look for libgl in order to compile on guix; a minority, but not unheard of, and quicker to try than building mesa
<bost>I'd like to connect to my Guix running in a VM through SSH. (As described
<bost>I uncommented the `(service openssh-service-type)` in the default /run/current-system/configuration.scm, ran `guix system reconfigure /path/to/configuration.scm` and rebooted.
<bost>SSH daemon seems to be running: `sudo herd status | grep ssh` shows '+ ssh-daemon' but I can't connect.
<bost>Is it enough just to have the `(service openssh-service-type)` or do I need to configure it?
<cbaines>The default configuration should be OK, you've probably got a networking issue getting to the VM
<cbaines>Assuming you're trying to ssh from the host, how is the traffic meant to be routed to the VM?
<bost>cbaines: I use `qemu-system-x86_64 -nic user,model=virtio-net-pci,hostfwd=tcp::10022-:22 ...`
<bost>cbaines: and I try to connect with `ssh -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -p 10022 guest@localhost`
<bost>or just `ssh -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -p 10022 localhost`
<bost>without 'guest@'
<cbaines>Do you get any connection at all?
<bost>cbaines: I think so. I get 'Warning: Permanently added '[localhost]:10022' (ECDSA) to the list of known hosts.'
<cbaines>Ok, in what way does ssh fail to connect?
<bost>And then a prompt for password. Which is strange since my 'guest' user has no password
<cbaines>allow-empty-passwords? is #f in the default openssh service configuration, so that may explain the behaviour
<bost>cbaines: So do I need to have `(service openssh-service-type
<bost> (openssh-configuration
<bost> (permit-root-login 'without-password)))`
<radvendii>okay so apparently i needed to change the "opengl" dependency to a "gl" dependency. weird
***lukedashjr is now known as luke-jr
<cbaines>bost, I was pointing out the allow-empty-passwords? field of the configuration, since you said that your guest user has an empty password
<cbaines>but in general, just change the configuration to what you want it to be
<bost>cbaines: I put `(service openssh-service-type (openssh-configuration (permit-root-login #t) (allow-empty-passwords? #t)))` to my configuration (just in case) and now I'm running `guix system reconfigure configuration.scm`...
<bost>cbaines: that takes time... but let's see if it helps.
***lukedashjr is now known as luke-jr
<podiki[m]>radvendii: glad you figured it out! there's a lot of gl/opengl and variations, confusing
<maximed>How does one do ‘positional arguments’ in 'format' strings?
<bost>cbaines: so reconfiguring with `(service openssh-service-type (openssh-configuration
<bost> (permit-root-login #t) (allow-empty-passwords? #t)))` helped. Thank you for that "allow-empty-passwords? #f" hint!
<maximed>To allow "Let ~a do ~a" to be translated as "Let ~a be done (by ~a)" (just an example)
<maximed>The two ~a need to be swapped
<maximed>I ended up with (format #t "Let ~0@*~a do ~1@*~a" 'x 'y), but maybe there's some nicer way to do this
<roptat>I guess you want to swap 0 and 1?
<roptat>you could keep "Let ~a do ~a" as the English, and translate as ""Let ~1@*~a be done (by ~0@*~a)"
<maximed>roptat: Yes
<maximed>I wanted to make it easy for the translators so they don't have to know that ~N@* exists and how it functions
<leoprikler>there's no solution with plain guile strings, but if you want to parse placeholders you can do "Let <0> do <1>"
<leoprikler>plain guile format strings, that is
***lukedashjr is now known as luke-jr
***lukedashjr is now known as luke-jr
<rekado>I’m having another problem in a container.
<rekado>I installed wget and nss-certs, SSL_CERT_DIR and SSL_CERT_FILE are both set, but wget does not use the certificates.
<rekado>it only works when I pass --ca-directory
<rekado>this is inside of an RStudio shell session, so there’s a chance that this is responsible
<cehteh> << anyone bored and want to package that?
<ngz>cehteh: It looks interesting, although a bit young.
<maximed>rekado: apparently 'wget' doesn't have any 'native-search-path'
<cehteh>i havent tested it, a friend just shown me the link
<cehteh>but 'alphapapa' looked familar, he maintains the magit-todos i am using :D
<maximed>rekado: maybe export /etc/ssl/certs in the container? That's what GnuTLS looks in by default.
<ngz>cehteh: In any case, Emacs packages are usually not difficult to package, you may want to give it a whirl.
<ngz>not difficult to write*
<maximed>(Ideally, wget would recogise the relevant environment variables ...)
<abrenon>ok I got regular import to work
<abrenon>only need to fix recursive-import since I renamed one key
<Fare>I'm new to guix. I'm trying to install guix in an encrypted lvm partition. Right now, it's failing very early on, not finding the root filesystem. I don't even see cryptsetup in the initrd.
***lukedashjr is now known as luke-jr
<Fare>Is there a guide somewhere to booting guix on an encrypted lvm ?
<vagrantc>haven't tried, but you presumably need to make a mapped device that might also consist of mapped devices ...
<jackhill>Fare: appologies, I'm not aware of a write-up, but here's a config.scm that works for me:
<Fare>jackhill, thanks!
<jackhill>If you're putting swap in the encrypted lvm like I am, my experience is that it has to be specified with the uuid, using the /dev/mapper/… path will confuse the code that checks device-mapper dependencies.
<jackhill>I don't believe lvm2 need to be installed in the global packages for it to work, but it's convienent to have it available by default on the system.
<monkwitdafunk>Hello. Can i take linux foundation training, or enroll in linux professional institute programs to be able to volunteer for guix?
<iskarian>Is gcc known to be broken on core-updates x86_64?
<iskarian>(I'm getting "collect2: fatal error: cannot find ?ld?")
<Fare>jackhill, thanks
<jackhill>Fare: you're welcome, good luck!
<rekado>iskarian: use gcc-toolchain
<rekado>(is that what you use? Then it sounds broken to me.)
<jackhill>monkwitdafunk: trainings are not required to volunteer with guix. Knowledge is of course alway good, but I'm not familiare enough with those courses to know how relivant they would be to guix (also, I guess it depends on what you already know) :)
<iskarian>I just dug deeper into it; go-1.14 is failing to pass a test on core-updates. The test is attempting to use "-fuse-ld=gold"
***lukedashjr is now known as luke-jr
<Fare>is btrfs the way to go these days, or is it still ext4 ?
<iskarian>rekado: this package is working on master and the only change to it for core-updates was a small change to use search-input-file/search-input-directory
<jackhill>Fare: clearly you see how how feel, but I'm not anti-ext4, I just like some of the btrfs features :)
<iskarian>Ahh, I just searched upstream, this is an issue with gcc 9+; it was fixed in later versions
<iskarian>later versions of go*
<jackhill>*how I feel
<Fare>I'm just worried about stability and have heard horror stories about btrfs
***lukedashjr is now known as luke-jr
<maximed>Fare: fwiw, is functioning just fine on my laptop (1 year so far)
<jackhill>Fare: I haven't done anything to really push it (single disk, nothing too fancy. I'd like ot use bees for de-duplication, but haven't finished writing a Guix service for it yet), but it hasn't been a problem for me. I haven't tracked performance. Great thing about Guix is that it provides up to date kernels
<jackhill>looks like there have been fewer bugs recently, at least in the areas touched by bees:
<iskarian>Okay, why can't I search "spec:core-updates-frozen" on cuirass? (it returns nothing)
<iskarian>Oh, you need other terms...
<Fare>I made some progress, and am now stuck in the early boot Guile. How do I diagnose from there?
<Fare>(system "ls") returns 2, I suppose that means error
*Fare discovers ,bournish
<Fare>cryptsetup: no such file or directory. I suppose that might explain this.
<maximed>Fare: Guix doesn't put things in /usr/bin or /bin
<maximed>"cryptsetup" not being in PATH isn't unusual