IRC channel logs

2017-01-03.log

back to list of logs

<ile>Ok, I think I will just use the simplest way for now (homemade scripts to install software), and wait till I get the GuixSD installed natively. Either it will happen soon, or it won't happen at all, so this is the easiest way for now...
<ile>If it's possible to create a Vagrant box (Virtualbox) for GuixSD, it might help a lot of people for setting it up/learning how to do it. So, it would be nice to see such a box on vagrantcloud.com - a suggestion.
<lfam>Now I've got the basics of python-gpg working, but the Python libraries lack the proper runpath references with these patches: http://paste.lisp.org/+76WV/1
<lfam>Any ideas about the Python-idiomatic way to approach that?
<buenouanq>lfam: python-idiomatic? google for a lib that already does what you want
<lfam>I realized that they distribute this 'gpg' library separately from the GPGME source code on PyPi. MY current strategy is to hope that just does what I want without.
<lfam>s/without//
<geemili>How do I tell guix to use python2 with an application?
<lfam>geemili: In the package definition? Put '#:python ,python-2' in the arguments field
<geemili>lfam: Thanks.
<lfam>commit 0b96fb2ac02dc5f0bcebf45a31b4682de4c96a24 is strange
<vtomole>Anybody had success on installing guixsd on vitrualbox? I'm new to this kind of stuff and i give up trying to do it :(. I had success with nixos
<lfam>vtomole: No, I've only used GuixSD in QEMU
<vtomole>I tried qemu too, had network issues....
<vtomole>If you can post what you did i would be very grateful :)
<vtomole>I mean i read the manual but still no luck.
<lfam>I always use it like this: qemu-system-x86_64 -net user -net nic,model=virtio -enable-kvm
<lfam>If you use that technique to set up the network, ICMP (ping) doesn't work
<lfam>Did that method not work for you?
<vtomole>give me a sec let me switch back to my other system.
<vtomole> https://www.gnu.org/software/guix/manual/html_node/Running-GuixSD-in-a-VM.html should the "guix system vm-image" be sun on my host system?
<vtomole>*run
<vtomole>or should it be run on this download://www.gnu.org/software/guix/
<vtomole>lfam: booting from vm was not hard, it's just installing it that's been a pain most recent error has been "can't install GRUB"
<lfam>If you just want to get GuixSD running in QEMU, `guix system vm-image` will generate a new QEMU image with GuixSD in it for you
<lfam>It's probably easier than installing GuixSD in QEMU "by hand", although that does work too
<foobar123>I'm considering guixsd as a nix user, and after going through the manual, i have a few questions. First, the regular channel that guix uses to serve packages is a rolling release or moves with releases only?
<davexunit>master is always rolling
<foobar123>cool
<foobar123>thanks
<foobar123>is there any channel that follows a more stable release model?
<lfam>foobar123: No, nobody has created a stable channel yet
<foobar123>lfam: fine, thanks.
<vtomole>lfam: Trying to install from source "Gnu libcrypt does not appear to be usable"
<foobar123>how do you manage emacs and python packages? pjotrp suggests, in case of python, to set the prefix to something that contains the hash of the installed python: https://github.com/pjotrp/guix-notes/blob/master/PYTHON.md
<lfam>vtomole: I'm not experienced with building Guix from source without already having Guix available so that I can use `guix environment guix`
<foobar123>or do you create a guix package for everything, perhaps with guix import?
<lfam>Ideally there's a Guix package for everything
<foobar123>ok
<lfam>But pjotrp has a lot of experience with Guix. His approach for working with Python software not-yet-packaged for Guix is probably based on that experience
<foobar123>yes, i guess its a quick way to do it
<foobar123>that also preserves isolation
<lfam>*Most* Python software is easily packaged
<lfam>We also have Emacs packages
<foobar123>yes, i know
<foobar123>but some deeplearning packages are quite nasty as they pull binaries from here and there
<lfam>Scientific computing tends to have "nasty" packages from what I've seen
<lfam>We do our best with them :)
<lfam>And then there are 3rd party collections of bioinformatics Guix packages as well. Not all of it is appropriate for GNU Guix
<foobar123>yes i've seen some of those collections. In fact, they are a big motivation to switch to Guix.
<foobar123>Reproducibility is very important for us when we write papers, and many of those packages output very different results with upgrades or compilation flags
<lfam>I'm glad to hear that Guix can perhaps provide a way for you to improve your experimental practice. I suggest it to all the researchers I meet :)
<foobar123>It's also for my own sanity! :D
<lfam>And for humanity ;)
<foobar123>yeah!
<foobar123>i was using arch's AUR for very long
<lfam>What is the point of the experiment if nobody can reproduce it?
<foobar123>but after a while it's a no go
<foobar123>yep
<foobar123>well, you can use other approaches like dockers
<foobar123>but it's a bit ugly IMHO
<lfam>Yeah, we think that Guix can solve the problem more effectively
<foobar123>guix is very elegant, no doubt about that
<foobar123>gonna sleep, thanks for all those insights
<vtomole>Difference between nix/guix vs pacman/apt?
<lfam>Like, how is the packaging idea different?
<vtomole>yeah
<Apteryx>I'm trying to understand where the setuptools v20 is "bundled" or "input" to the python build system, if anyone has a clue.
<lfam>vtomole: I'm not familiar with pacman. Apt / dpkg work by changing the contents of the /usr directory, destroying what was there before. Also, they are binary package managers. Guix and Nix, on the other hand, and build-from-source package managers. And, they put all packages in unique immutable directories and then use symlinks to make a subset of them available to users. These unique directories are named for all their inputs:
<lfam>dependencies, build scripts, etc. This allows the system to transparently substitute pre-built binaries instead of building from source.
<lfam>Generally, I recommend these two papers:
<lfam>Nix: A Safe and Policy-Free System for Software Deployment by Eelco Dolstra et al
<lfam>Functional Package Management with Guix by Ludovic Courtes
<lfam>Those papers aren't too long and they explain the ideas of functional package management and how it is implemented
<vtomole>Thx. This stuff is so different.
<lfam>Apteryx: My understanding is that Setuptools is bundled with Python itself, but I'm not sure
<vtomole>lfam: For the vm, you had to make a config.scm first right? I did this "guix system vm my-config.scm \\ --expose=$HOME --share=$HOME/tmp=/exchange"
<lfam>vtomole: Right, you have supply a configuration file on the command line. I also recommend setting --image-size to some large value. The image file uses a "balloon" so it won't take up that much space until you use it.
<lfam>Oh, perhaps that's only necessary for vm-image.
<lfam>No, it seems that `guix system vm` does take that option. I use 100G
<lfam>I'm not sure what the default is
<vtomole>noo i already ran that command, i also changed /dev/sdX to /dev/sda, hopefully that's right..
<lfam>I guess it doesn't matter much for things created with `guix system vm` since they are meant to be immutable, so everything fits when you build it, or not.
<lfam>vm-image is what creates mutable VM images that might need more space later
<Apteryx>I answered my above question about where setuptools v20 comes from. Apparently it happens at the time the python interpreter is built, with configure flag: --with-ensurepip.
<Apteryx>That explains why it is a bit dated (we are using Python 3.5).
<lfam>Help wanted on the 3.6 upgrade :)
<lfam>I looked at it last night but I didn't start
<lfam>The inheritance chain from 3.4 -> 3.5 -> 2.7 felt like a great weight on my shoulders...
<lfam>I wasn't sure where to insert 3.6
<lfam>Perhaps we can retire 3.4 and 3.5; I don't know
<vtomole>lfam: I'm getting this https://ipfs.io/ipfs/QmPjdqYLvjjRByEmoDd3LFCwpzdLqSd5XAMyog2BNfMzp4
<vtomole>My kvm is fine i think..
<lfam>You don't have permission to use the /dev/kvm device. I just change permissions on the /dev/kvm file...
<lfam>Is it the right way? I don't know :)
<vtomole>I'm new to linux so sorry for the stupid questions, how do you change permissions? do you use chown?
<lfam>You'll want to use chmod, and just change the permission bits, rather than changing ownership with chown. I've been using `sudo chmod 666 /dev/kvm` but, like I said, no promises that this is a good idea! The permissions are reset after reboot
<Apteryx>lfam: Just saw your answer, thanks. Yes it is bundled with Python. I'm trying to test a newer setuptools (required to run pip's tests). I've extracted pip's sources and tried to use the environment: "guix environment guix python-pip --ad-hoc python-setuptools". However a "pip list" in that environment shows that I'm still using setuptools 20 rather than 31. Am I creating the environment wrong?
<lfam>Apteryx: You probably can remove the 'guix', unless you need Guix's dependencies in the environment. I would also use, at least, --pure, if not --container
<lfam>Beyond that, Hartmut is the expert on guix-devel
<Apteryx>lfam: Right, the guix is only so that I can rebuild the package files I'm editing.
<lfam>Ah
<lfam>If you are making your changes in the package files, then the python-pip and python-setuptools won't have an impact when you rebuild the packages in the package files
<lfam>You have to make all your changes in the package definitions. Sorry if I misunderstand your approach
<lfam>I meant to say that Hartmut is the expert on Pip and Python on guix-devel. He recently revamped the python-build-system
<Apteryx>lfam: I guess my flow is a bit confusing, I kind of doing two things in parallel (adapting the package definition and testing manually (independent of the package definition) in an environment).
<lfam>Ah
<lfam>Then I'm not sure why the wrong Pip is being used
<Apteryx>It looks like the python build system is doing something wrt the PYTHONPATH which I haven't understood yet.
<Apteryx>I guess I'll have to ping Harmut :)
<vtomole>lfam:Finally some good news:https://ipfs.io/ipfs/QmV4LyeW7japoRFseZ2VhJNcoaY2ywNxFhw5BHR9w2fLss. What do i do next?
<lfam>The last line is a shell script. Just run that script
<lfam>You might also want to protect the script from being garbage collected by symlinking it into /var/guix/gcroots: `ln -s /gnu/store/dl2xk2qnqaai3703hgmas0hbc76q542q-run-vm.sh /var/guix/gcroots/my-vm`
<lfam>Otherwise it will disappear when you run `guix gc` to free up space
<lfam>That depends on your use case
<vtomole>lfam: hmm "./gnu/store/dl2xk2qnqaai3703hgmas0hbc76q542q-run-vm.sh bash: ./gnu/store/dl2xk2qnqaai3703hgmas0hbc76q542q-run-vm.sh: No such file or directory"
<lfam>The leading '.' means "the current directory". Try leaving it off, so the command starts with the '/'
<vtomole>"qemu-system-x86_64: -virtfs local,path=/home/vtomole/tmp,security_model=none,mount_tag=TAG_home_vtomole_tmp: 9pfs Failed to initialize fs-driver with id:TAG_home_vtomole_tmp and export path:/home/vtomole/tmp"
<lfam>Hm... I've actually never tried those --expose and --share options in this context
<lfam>That appears to be what's failing
<lfam>Are you sure the $HOME/tmp directory exists?
<vtomole>ahh I thought it did. but it doesn't.
<vtomole>lfam: Okay it started up, thanks for being patient with me :)
<lfam>No problem, I'm glad it's working now :)
<vtomole>lfam:Idk if it's hanging or loading:https://ipfs.io/ipfs/QmSRpVs5nK9QjT31g69ej5QguPbaAkW9nbbsRmtLAkBhsr
<lfam>vtomole: You need to bash on the keyboard for a minute or so to provide entropy to the Linux kernel's random number generator
<lfam>"Please type some random data"
<lfam>Actually, that's not for the Linux RNG. That's for the LSH ssh daemon seed
<lfam>Same thing
<lfam>Something demands entropy
<lfam>You'll need to do this each time you boot the VM, and it's annoying. You can try the OpenSSH or Dropbear SSH servers instead of Lsh to avoid this
<vtomole>Wait what? HAHA omg i didn't know linux was THIS interesting lol
<lfam>It doesn't have to be this "interesting". There are some very "boring" distros :)
<Apteryx>lfam: When using "guix environment --pure --ad-hoc python", the "python" executable is not even found. Any idea why?
<lfam>Apteryx: I suspect it has to do with the wrap-python3 procedure in (gnu packages python)
<lfam>I bet that python3 is available in the environment
<Apteryx>lfam: Actually, without --pure I'm missing python as well. Hm.
<Apteryx>It is!
<vtomole>lfam: Regarding the config file. Is the "comment" variable the password?
<enoeht>hello
<Apteryx>hi there
<lfam>vtomole: Which file and which password? :)
<enoeht>I'm trying to make an alias for guix and I'm having issues
<enoeht>here is my bash code
<enoeht>#!/bin/sh
<enoeht>sourceVars(){
<enoeht> source $HOME/.bash_profile
<enoeht>}
<enoeht>guixPackageInstall(){
<enoeht> sourceVars
<enoeht> echo "guix package -i "$@""
<enoeht> guix package -i "$@"
<enoeht>}
<enoeht>alias gpi=guixPackageInstall
<lfam>enoeht: Please use a paste site such as paste.lisp.org. Don't paste code directly into the chat
<enoeht>ok ok
<enoeht>anyway this is what i get
<lfam>In the future, please do that. This script fits on my screen
<enoeht>gpi iproute \\ guix package -i iproute
<enoeht>guix package: error: failed to connect to `/var/guix/daemon-socket/socket': Connection refused
<enoeht>thats my output
<lfam>enoeht: Is the guix-daemon running?
<vtomole> lfam:my-config.scm:https://ipfs.io/ipfs/QmfXWud7M9Ct6JPqH9jucuq2u45tay7pHuiCnW6KrGM9GS, what would be the password to alice?
<Apteryx>enoeht: Looks like your problem has nothing to do with your bash alias.
<lfam>enoeht: Unrelated, I don't believe that shell aliases take arguments. I'd use a shell function declared as gpi()
<enoeht>omg
<enoeht>nvm im an idiot
<enoeht>its systemctl thats messing up
<lfam>vtomole: By default, user passwords are set by logging in as root and using `passwd`, but they can be set in the OS configuration file, as described in the section on User Accounts
<Apteryx>lfam: I think assigning a shell function to an alias is the usual way to work around the fact that aliases don't support args. I've used it before at least, and it was working.
<lfam>Apteryx: Right, I just use functions too
<enoeht>no the args work fine
<lfam>Okay, I didn't realize they could be chained like that :)
<enoeht>thats what $@ is for
<lfam>vtomole: If you want to set a user password, you should really follow the instructions in the section Running GuixSD in a VM. The method you used created an immutable (unchangeable) VM image. Maybe that's what you want, I'm not sure
<vtomole>lfam: You ever have this?https://ipfs.io/ipfs/QmXJr2DSWbutD3YuRMUpHazhTnYCCLzw1CNi4ofNCtTTjj
<lfam>vtomole: Yeah, but I'm not sure what's going on.
<lfam>vtomole: Did you notice the note in section 7.2.14 Running GuixSD in a Virtual Machine about adding the -net user flag?
<lfam>That's changed since the last time I tried this
<enoeht>I keep getting this error "substitute: warning: failed to install locale: Invalid argument"
<enoeht>ive tried guix pull but no effect
<lfam>enoeht: It's just a warning, not a fatal error
<enoeht>yeah but its annoying
<lfam>Basically, you need to have glibc-locales available in the environment where the daemon runs
<vtomole>lfam: I didn't get the chance to run that since this worked: https://ipfs.io/ipfs/QmYUCKSYkfhkgYo7Eho8fcCUC5wwU9LaE9ZiD3vjnwXVd4
<lfam>Typically, this means installing them as root and they adding 'GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale' in the envrionment where the daemon runs
<enoeht>ah, is that for root or guixbuild user
<lfam>enoeht: See https://www.gnu.org/software/guix/manual/guix.html#Locales-1
<lfam>enoeht: Root, who runs the guix-daemon
<lfam>You're using sytemd, right? Add that variable declaration to the service file
<lfam>vtomole: Great :)
<enoeht>cool it works, the docs just said nothing about the root user needing GUIX_LOCPATH
<vtomole>lfam: The net flag is in the sh file:https://ipfs.io/ipfs/QmUP2SLxXF3TC1yPvtrYXPwBJhBuD3HeqwoGntqUuvuaea, how do you connect to the internet in guixsd vm?
<enoeht>I'm trying to install kodi through guix but the build fails, is there a link to a page detailing how to manually fix packages
<enoeht>here is the error
<enoeht>guix package: error: build failed: build of `/gnu/store/p0d7wwzlqkimqnk7fcnlhxn7x9b7v76f-kodi-16.0.drv' failed
<lfam>There must be some more messages printed :)
<enoeht> http://paste.lisp.org/+76Y7
<enoeht>i know how to code and fix compile errors I just needed to know the standard process, do i use "guix enviornment kodi" ?
<lfam>That will update your shell environment to make Kodi's dependencies available. You can add --pure to unset existing variables, and --container to enter a container, for extra isolation
<lfam>You can use `tar xf $(guix build --source kodi)` to make the source code available
<enoeht>ok cool
<lfam>I usually make changes in the package definition of the failing package itself, because that way I don't have to worry about isolating the test builds from the rest of my system, but others do it with `guix environment`
<lfam>I wonder if they turned deprecation warnings into fatal errors
<vtomole>lfam, you said you've gotten this error too, have you ever fixed it? I mean you just don't use guixsd on vm without being connected to the internet, do you?
<lfam>vtomole: I mean, in general, I've seen dhclient fail. It usually means that the network interface was not actually activated, or there is some problem with the dhcp server I'm querying
<lfam>vtomole: Did you start the VM with '-net user' like the manual suggests?
<lfam>Our kodi package could be updated to 16.1.
<Apteryx>Any way to vizualize the output of "guix graph python | dot -Tpng" directly into emacs?
<Apteryx>Without piping it to a file and then opening that file?
<lfam>If Emacs can read dot format, I'm sure it can be done. But I don't use Emacs
<lfam>I'd be surprised if Emacs can't do this
<lfam>enoeht: Hm, looks like Kodi has not built for a while: https://hydra.gnu.org/job/gnu/master/kodi-16.0.x86_64-linux
<enoeht>i may have had internet issues while downloading, how do you clean the build and start over
<lfam>Wow, not since March?!
<Apteryx>lfam: Oh, I should have looked in the direction of emacs-guix. Seems there's an option to visualize the graph directly.
<lfam>enoeht: Here's the first failing build. It has a link to display the code diff between the failure and the last successful build: https://hydra.gnu.org/build/1115934
<lfam>enoeht: I'd just delete the source code and start over
<lfam>Maybe you could use Git to keep the initial state on disk
<enoeht>whats the easiest way to get the source location besides using locate and find?
<Apteryx>enoeht: guix build "yourpackage"
<lfam>the source location?
<Apteryx>Well, this will give you the location of the build. Pass the --source option to get the source.
<enoeht>nice
<enoeht>uh oh guix build: error: build failed: getting attributes of path `/gnu/store/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz': No such file or directory
<lfam>Did you set a custom TMPDIR for the guix-daemon? There's a bug that breaks this at the moment
<enoeht>i dont think so
<lfam>Otherwise, what command did you run to get that error?
<geemili>Is it normal for Guix to keep rebuilding these packages? http://pastebin.com/fm1JsP0b
<lfam>geemili: Yes, are not really "packages", but rather the things that make up your user's profile. Everytime you create a new profile they must be rebuilt
<enoeht>guix build --source kodi\\
<enoeht>sudo rm /gnu/store/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz
<lfam>enoeht: You can't make changes like that in /gnu/store
<enoeht>i didnt realize it was an archive
<lfam>The store is only meant to be altered by the guix-daemon
<enoeht>:O
<lfam>Files in /gnu/store are treated as immutable
<enoeht>so guix build -source shouldn't be used to locate the source?
<lfam>To delete a file in the store, use `guix gc --delete`. But in this case, there's probably no need to recreate that file
<lfam>If you want to use the source code, use `tar xf $(guix build --source kodi)` in a directory outside of /gnu/store
<enoeht>yeah im compiling kodi now but I wanted to try to reinstall from scratch, but it seems this is a known issue
<lfam>Here, the store is explained: https://www.gnu.org/software/guix/manual/html_node/The-Store.html
<enoeht>the store is the place where derivations that have been built successfully are stored ???
<lfam>Okay, maybe it's too low-level :)
<lfam>Just note this part: Users must never modify files under /gnu/store directly."
<enoeht>so how do i just install the last successful build?
<enoeht>sorry I know I could read the entire docs but I'm lazy
<lfam>I'm sure we don't have it anymore. March is too long to still have the binary available.
<enoeht>ok I'll just to try to fix it
<enoeht>thanks
<lfam>I bet the problem has to do with the ffmpeg update that happened between the last good build and new failures
<lfam>We still have the old ffmpeg-2.8 package in our tree
<Apteryx>Where is all the information for a package DAG kept? In a database?
<Apteryx>Just looking at the graph of Python; it's really interesting to be able to visualize the dependencies so easily :)
<enoeht>Hi, I've downloaded the source to a guix package that was broken, now I want to test to see if it works, I'm thinking I just run guix environment kodi --pure -- make install... is this correct
<enoeht>or is there a link to where I can read how to make changes to existing packages?
<Apteryx>enoeht: Not really yet... I like to be working directly in guix from git with "guix environment guix --ad-hoc your_package", then I can edit the package definition in-place. In emacs I then run "C-c C-z" to fire Geiser, then 'M-x guix RET e "your_package"'
<Apteryx>the emacs guix command will bring you to your package definition.
<Apteryx>From there there is a Guix mode in emacs which allows you to lint & build your package definition using C-c . l and C-c . b
<enoeht>cool thanx
<Apteryx>To set your system so that you can use the flow described above you'll want to refer to the manual at chapter 8, it gives useful directions to work directly from Git.
<enoeht>I accidentally deleted the source archive, now I'm getting guix environment: error: build failed: getting attributes of path `/gnu/store/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz': No such file or directory... is there any way to fix this
<Apteryx>enoeht: Oh, did you delete it directly from its store location?
<enoeht>yeah ;(
<enoeht>i wont do it again
<Apteryx>enoeht: OK. I guess you can try to uninstall it, then garbage collect it.
<Apteryx>Not sure if guix will complain about it being missing in the first place.
<enoeht>ok that worked
<Apteryx>ok nice!
<rekado>bavier: 0b96fb2ac02dc5f0bcebf45a31b4682de4c96a24 looks incomplete. Did you forget to add more hunks?
<Apteryx>If a package is distributed without its tests through pypi, but available with tests from github, do we want to use github with tests?
<Apteryx>I'm looking at enabling tests for pip and noticed the tests are not found in the archive downloaded from pypi.
<vtomole>Having problems with pinging again "100% data loss"
<buenouanq>I realize fully why no one wants to talk with me about about php,
<buenouanq>but someone packaged it, so I know that at least one of you can answer my questions.
<rekado>vtomole: try “sudo herd restart nscd”
<vtomole>rekado: no luck
<rekado>vtomole: are you pinging an IP address or a domain name?
<vtomole>I pinged local host, gnu.org, and 8.8.8.8 all are bust
<vtomole>ifconfig eth0 up && ifconfig eth0 up didn't work
<buenouanq>and you're sure it's eth0?
<buenouanq>what does `ip a' output?
<vtomole_> https://ipfs.io/ipfs/QmZqLuMaxzvW3fy5joyufKcoqNTA6d9G5n3URep7XVpfTh
<buenouanq>but you're connected to the wifi network, not over eth0...
<vtomole_>My host system is connect to wifi, but vm is on eth0, that's how qemu is configured
<vtomole_>I think, right rekado?
<bldtg>You know what would be neat, is a guixsd GTK theme.
<rekado>turns out that removing the preprocessing step from Hugs (and manually preprocessing the Nhc compiler sources) makes Nhc on Hugs *much* faster.
<rekado>takes only a few seconds now to compile a Haskell source file to C.
<rekado>I’ll soon have a patch set on top of the latest Nhc release that allows bootstrapping Nhc from Hugs.
<civodul>rekado: woohoo!
<civodul>crazy stuff that you're doing
<rekado>this is fun! Crazy but fun.
<jmi2k_>How do I delete old system generations?
<amz3>meh guix pull compiles everything from scratch on x86
<adfeno>amz3: Did you add the substitutes server's public key?
<adfeno>Oh... You're referring to `guix pull`, not `guix package`.... Sorry.
<adfeno>Being a x86 user myself, I'm getting used to `guix pull` compiling the .scm recipes into .go files.
<adfeno>It doesn't take much time, atl east for now.
<adfeno>Although I would prefer a `-c [Number]` option being available for `guix pull` since it seems to be the only step in which my computer fries most.
<amz3>adfeno: so does mine
<amz3>adfeno: it not compiling scm to go but compiling guix dependencies AFAICT
<adfeno>Hm....
<adfeno>... Isn't this supposed to be done when doing `guix package -u` (or some command that requires Guix installation or upgrade)?
<amz3>I don't understand
<rekado>guix might need to build things to build the latest version of guix
<rekado>so it is possible that it first downloads substitutes
<jmd>How is the staging merge progressing?
<amz3>except in my case I doesn't download substitutes but compile from scratch
<amz3>was the hydra key changed during the last 6 months?
<civodul>jmd: good question, it's about time!
<civodul>we should use call-with-container & co. for our services to achieve protection like that of systemd: https://lwn.net/SubscriberLink/709755/90323f967f792219/
<roelj>civodul: That would be really cool if we could do that.
<amz3>so guix pull finished, and now i am running 'guix system reconfigure'
<amz3>and substitutes are fetched
***jonsger1 is now known as jonsger
<alezost>jmi2k_: to remove system generations, you need to remove system symlinks in /var/guix/profiles/ manually
<jmi2k_>alezost: Ok, thanks
<enoeht>I accidentally deleted an archive from /gnu/store and now I am getting this error "guix package: error: build failed: getting attributes of path `/gnu/store/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz': No such file or directory", is there any way to get this file back? I've tried removing the package and calling gc with no luck
<rekado>enoeht: deleting items from /gnu/store is very bad
<enoeht>I know
<roelj>enoeht: Possibly explicitly doing: `guix gc -d /gnu/store/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz'.
<enoeht>nope
<rekado>you could try to put it back after downloading it:
<rekado>wget hydra.gnu.org/nar/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz
<rekado>I'm not sure if this is correct, though
<roelj>Or maybe guix build -S kodi@16.0
<roelj>Which would also download that file
<rekado>ah, that's better
<roelj>Or would it skip the download, because it thinks it is already in the store?
<rekado>BTW: I successfully built a docker image archive with just Guix.
<roelj>rekado: Wow!
<jmi2k_>Help, please! Xorg doesn't detect any input: keyboard or mouse. I read that it should be related with xserver.conf, but I don't have that file. I'm not using a login manager, just installed xorg-server, xinit and some drivers.
<davexunit>rekado: wow!~
<davexunit>now *that* is awesome
<rekado>and it includes the annoying step of linking /gnu/store/.../bin to /bin
<ng0>I've packaged some more dependencies for libextractor, but it's difficult to get more replies (mainly: I need a maintainer to reply) from upstream libmp4v2, where it went etc... as code.google.com will disapear soon, is there a place where I can propose to GNU to fork it to continue development, or if I comment it enough will a fork at gnunet.org be enough? the situation out there of the forks is wild chaos..
<ng0>those are the two last options I would want to decide on if I get no reply this month
<rekado>a fork is a committment
<rekado>I wouldn't feel comfortable with an unmaintained video library
<ng0>the thing is, everyone forks and bundles and extends mp4v2 these days... my search for the real upstream can be long
<ng0>so it would be a temporary location, until the most popular fork becmes the new upstream
<ng0>how long will the archives of googlecode be accesible?
<rekado>when I download a thing from hydra, how can I import it? "guix archive --import < the-thing" says "input doesn't look like something created by `nix-store --export'"
<alezost>jmi2k_: I'm pretty sure the problem is that X server doesn't know where drivers are placed, do you just run "xinit -- .../Xorg" ?
<civodul>rekado: check the different between 'export-paths' (plural) and 'export-path' in (guix store)
<civodul>rekado: the plural form adds a 1 before each item, and a 0 at the end
<civodul>a pretty ad-hoc format
<alezost>jmi2k_: you need to run X server with "-modulepath /run/current-system/profile/lib/xorg/modules" (if you install X server and the drivers in a system profile)
<enoeht>when I try sudo wget hydra.gnu.org/nar/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz
<enoeht>it downloads then turns it into wget hydra.gnu.org/nar/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz.lock
<enoeht>i mean it turns it into /gnu/store/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz.lock
<rekado>enoeht: stop changing the /gnu/store
<rekado>enoeht: there's no need to use sudo
<enoeht>im trying to fix it
<rekado>don't write directly to the store
<enoeht>im following advice given
<alezost>jmi2k_: I have to go now, ttyl
<roelj>enoeht: rekado: wouldn't `guix gc --verify=repair' repair the missing tarball?
<civodul>it should repair it or remove it
<civodul>but it takes quite a bit of time
<enoeht>guix gc --verify=repair
<enoeht>guix gc: error: build failed: you are not privileged to repair paths
<enoeht>sudo guix gc --verify=repair
<enoeht>warning: failed to install locale: Invalid argument
<enoeht>reading the Nix store...
<enoeht>checking path existence...
<enoeht>didn't really do anything
<roelj>Did it finish running already?
<enoeht>oh you mean its running in the background?
<roelj>No, if you get a new prompt then it should be finished.
<enoeht>it seems to just quit after checking path existence...
<civodul>weird
<roelj>Aha. Same here.
<civodul>rekado, roelj: looks like i didn't get a slot in the "testing" track at FOSDEM: https://fosdem.org/2017/schedule/track/testing_and_automation/
<enoeht>its crazy that there is no way of coming back from deleting an archive from the store
<roelj>civodul: Oh :(
<civodul>enoeht: maybe it's "guix gc --verify=contents,repair"
<civodul>yeah
<civodul>well, less work for me, not so bad ;-)
<roelj>civodul: Hehe :) Did you have to conclude this from seeing the schedule? (didn't they send an e-mail?)
<civodul>right, i didn't get an email (yet?)
<adfeno>About recipe for abbaye (that is currently depending on Google Code): I received an email from nevat, and according to him, the text font which is bundled is no longer used because they decided to write the game text using images.
<civodul>kinda weird
<enoeht>Isn't there some config file I can edit that will tell guix that kodi was never installed
<roelj>enoeht: Well, you could remove it from your profile (and any other profile if it has been installed there), and remove all generations that had this kodi version installed.
<roelj>enoeht: But there has to be a better way..
<enoeht>+1 for guix package --purge
<rekado>enoeht: nobody should *ever* edit /gnu/store by hand
<rekado>enoeht: it's like saying that it's crazy that there's no way to recover from "sudo rm -rf /"
<roelj>Well, I think we should make `guix gc --verify=repair' so that it is able to fix this problem.
<enoeht>no its not
<enoeht>i should be able to uninstall and reinstall a package
<civodul>it's --verify=contents,repair as i wrote: https://www.gnu.org/software/guix/manual/html_node/Invoking-guix-gc.html
<rekado>enoeht: sure, but you don't edit the store for that
<rekado>enoeht: guix has commands to uninstall and reinstall packages.
<roelj>civodul: Cool! Thanks for checking.
<rekado>civodul: looks like I didn't get a slot in the distros track (according to https://fosdem.org/2017/schedule/track/distributions/)
<rekado>but there's going to be a Nix talk about NPM
<civodul>bah!
<civodul>oh well
<enoeht>"guix gc --verify=contents,repair" doesn't detect it
<jmd>enoeht: You install a package with "guix package -i <package-name>"
<enoeht>i know that
<enoeht>but I get this
<enoeht>guix package: error: build failed: some substitutes for the outputs of derivation `/gnu/store/s14xmy0liyr93cgybf16a40mlkyqlrb9-kodi-16.0.tar.xz.drv' failed (usually happens due to networking issues); try `--fallback' to build derivation from source
<roelj>enoeht: Have you tried gc'ing the derivation? `guix gc -d /gnu/store/s14xmy0liyr93cgybf16a40mlkyqlrb9-kodi-16.0.tar.xz.drv'?
<roelj>And then let it rebuild (or substitute)?
<jmd>enoeht: Well perhaps you do indeed have network issues. Did you try the --fallback option as syggested?
<rekado>jmd: enoeht deleted an item in the store.
<jmd>rekado: How did he manage that?
<enoeht>try it for yourself create a vm, install guix then delete an archive file and try removing the package and reinstalling it
<jmd>Oh in a VM. I've never played with those.
<enoeht>i said a vm so you dont screw your machin up, you could use a container or an overlay filesystem as well
<roelj>You can also do that with your regular Guix store, but that might (read: WILL) break things ;)
<roelj>Aha! What about this:
<roelj>sqlite3 /var/guix/db/db.sqlite "DELETE FROM ValidPaths WHERE path='<the-store-path-you-deleted>';"
<roelj>(please make a back up of db.sqlite before doing this)
<roelj>And possibly stop the daemon first.
<enoeht>lol, I think I'm gunna just reinstall guix
<enoeht>it shouldn't be this hard
<roelj>Good idea! Don't remove paths from the store directly next time ;)
<enoeht>believe me I won't
<enoeht>I challenge someone to try it though
<enoeht>and tell me if I'm crazy or not
<civodul>:-)
<rekado>enoeht: did you install just Guix or GuixSD in a VM?
<civodul>there are warnings against it in https://www.gnu.org/software/guix/manual/html_node/The-Store.html and https://www.gnu.org/software/guix/manual/html_node/Invoking-guix-gc.html and
<rekado>on GuixSD the store directory is re-mounted as read-only.
<enoeht>im using guix on top of ubuntu (I'm well aware what I did was bad, it was a mistake I'm just trying to fix that mistake but so far no one has given a solution)
<civodul>enoeht: sure, no problem :-)
<civodul> https://sfconservancy.org/blog/2016/dec/26/reproducible-builds-summit-report/
<roelj>civodul: I thought there were representatives of GNU Guix too ;)
<civodul>good point! ;-)
<civodul>50% more than Nix representatives, even
<roelj>Well well well! So, next time wear a t-shirt with a massive GNU Guix logo including the text "GNU Guix". Maybe you'll get noticed ;)
<quiliro>hello
<quiliro>i cannot remember my root password
<quiliro>how can i recover it?
<rekado>quiliro: does your user have sudo privileges?
<rekado>quiliro: then you can do "sudo su -" and run "passwd" as root
<quiliro>rekado: i will check
<cbaines>So, I'm trying to use substitute* from (guix build utils) in a service start script, but it doesn't seem to work (I get an error that looks like its not using that special syntax), I imagine this is a Guile thing, but I'm wondering if there is a way to make it work?
<rekado>I think I have a fix for enoeht
<quiliro>i can sudo with other commands but not with su -
<quiliro>i can "sudo nano /etc/passwd"
<quiliro>but authentication error on:
<quiliro>"sudo su -"
<rekado>download the substitute with wget, then: bzcat p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz | guix archive -x /tmp/p142amjzp1zplf6c659fmgrilhzdl86n-kodi-16.0.tar.xz; then sudo mv the file back into the store.
<quiliro>or "sudo su"
<rekado>quiliro: does it ask for your password once or twice?
<quiliro>twice
<rekado>that's weird, but it happens to me too
<rekado>(it shouldn't happen because after sudo authentication you are root, so su should not ask for a password)
<ng0>what's ECM in context of qt?
<quiliro>rekado: what do you suggest i do...an update makes network manager not start
<ng0>ah
<rekado>quiliro: there's another trick that might work for you.
<rekado>quiliro: I haven't tried it myself on GuixSD, but it used to work on Redhat.
<rekado>boot into single user mode by adding a "1" to the end of the kernel line in Grub.
<rekado>with some luck you won't be asked for a password and become root
<civodul>cbaines: what's the error you're getting?
<rekado>then you can run passwd as root
<rekado>quiliro: another way is to boot from a different medium and chroot into the GuixSD system, but I haven't tried that.
<jmd>The graphical installer is now at the stage where it needs field testing. Any volunteers?
<rekado>quiliro: it's a bug that "sudo su -" doesn't work :-/
<rekado>jmd: me! me!
<ng0>ecm ... extra-cmake-modules ?
<ng0>description: The Extra CMake Modules package, or ECM, adds to the m
<ng0>ah..
<quiliro>rekado: good idea..thanks...will try
<quiliro>rekado: i will file a bug
<rekado>quiliro: thank you!
<quiliro>rekado: thank YOU for the idea!
<cbaines>civodul, If I do something like (substitute* "foo" (("bar") "baz")) I get the error, ERROR: Wrong type to apply: "bar"
<quiliro>i booted with guix boot usb but it does not contain passwd command
<rekado>quiliro: does it contain "chroot"?
<quiliro>jmd: i will test
<quiliro>rekado: i will check now
<jmd>rekado, quiliro: Great. Thanks.
<rekado>quiliro: the idea is to run passwd inside the chroot.
<rekado>jmd: to test we need to build the image from your branch?
<jmd>Yeah. Or I suppose I could upload an image somewhere.
<ng0>I think our chroot can do at least that, but some guix daemon related features will not work in a chroot. sorry, forgot to open a bug about that a month ago
<quiliro>isn't herd cow-start something like chroot?
<jmd>quiliro: It's a wrapper around unionfs
<quiliro>jmd: please tell me when you upload the image
<quiliro>jmd what does a wrapper do?
<jmd>civodul: Is there anywhere I can upload 2 x 1GB disk-image files for testing?
<civodul>jmd: hmm not really
<civodul>i usually upload things like that to my ISP/NPO web site
<civodul>if you could send it to me somehow, i can put it there :-)
<jmd>I'll put them on my website (which doesn't have much bandwidth).
<quiliro>rekado: chroot booting from guixsd usb worked
<rekado>sweet!
<quiliro>i will report bug now
<quiliro>i rebooted but the password did not work
<rekado>did you chroot into the same system that you just booted?
<ng0>when I package a font, and I have the choice between unhinted and hinted source, what am I to pick?
<ng0> https://github.com/google/roboto/releases about this
<ng0>or should I build from source?
<ng0>I think on their own advice I will take -hinted
<rekado>hinting is a feature.
<ng0>okay. but that does not give a reply on wether we want hinted or unhinted fonts
<quiliro>rekado: it worked now
<quiliro>maybe becaused i sync'ed this time
<rain1>hey guys
<rain1>im still sad that I got bullied out of this community, i really like guix[sd] and i wanted to help contribute to it
<quiliro>rain1: cool
<quiliro>file bugs
<civodul>hello rain1!
<rain1>hiya
<ng0>ah, we take noto-hinted too, so I will just use -hinted for roboto too
<jmd>Is there a log which keeps the stdout of builds ?
<civodul>jmd: yes, see /var/log/guix and "guix build --log-file"
<jmd>That will only have the builds that succeed won't it?
<ng0>okay, added the roboto font and fluid to the big patch bundle coming soon.
<civodul>jmd: all of them
<paroneayea>ACTION really needs to find some time to package git-annex or something
<cbaines>civodul, so I tried putting the exact code "(substitute* "foo" (("bar") "baz"))" in to a service start script, and it gave that exact error: ERROR: In procedure bar:
<cbaines>ERROR: Wrong type to apply: "bar"
<cbaines>Any ideas on why this isn't working? Just putting that code in a file, and running it with guile works, as well as doing it in a REPL.
<rain1>you need ' mark: (substitute* "foo" '(("bar") "baz"))
<civodul>no, you don't need a quote
<civodul>cbaines: looks as if (guix build utils) wasn't loaded, but it should be there
<civodul>specifically, it's added in (gnu services shepherd)
<civodul>could you check the generated /gnu/store/*shepherd*.scm file for that service?
<jmd>Uhh. building guix on i686 fails with g++: Internal compiler error
<rain1>maybe I can raise what happened as code of conduct violation or something? it would be good to have some kind of progress/resolution
<civodul>jmd: fun, which version?
<jmd>of what?
<civodul>of g++
<cbaines>civodul, ah, I'm passing a (program-file ...) as #:start, and the substitute* call is within that. I'm using with-imported-modules and have (use-modules (guix build utils)) within the gexp.
<civodul>ok
<civodul>but #:start must be a procedure anyway, not a file name
<civodul>or am i missing something?
<jmd>civodul: I dunno. Whichever gets installed I suppose.
<cbaines>civodul, yep, I'm using make-forkexec-constructor and passing in the program-file
<jmd>4.9.4 I think.
<civodul>cbaines: (make-forkexec-constructor (program-file ...)), right?
<civodul>in that case, you need to make sure the program in question has (use-modules (guix build utils))
<cbaines>#~(make-forkexec-constructor #$start-script)
<civodul>in addition to with-imported-modules
<civodul>cool
<civodul>yeah i mean (make-forkexec-constructor #$(program-file ...))
<cbaines>I've had a look in the directory in that file which is set as the %load-path, and it does include the (guix build utils) module
<Apteryx>Hi Guix!
<rain1>hi
<Apteryx>Looking at our Python-2.7 package, I was wondering: do the patches need to reference the exact filename to be patched or is there some fuzzy matchig mechanism? I see we are still patching 2.7.10 but we're now using 2.7.12.
<civodul>i just pushed zile-on-guile, you folks should give it a try, it's pretty cool :-)
<Apteryx>civodul: Cool! A precursor to a fully working emacs-on-guile? ;)
<rain1>will anyone help me ?
<civodul>it dates back to 2012 in fact!
<civodul>ACTION has to go
<paroneayea>wow zile-on-guile!
<davexunit>what is this?
<Apteryx>davexunit: the zile text editor running on guile?
<davexunit>somehow I missed civodul announcing this
<Apteryx>rain1: What do you need help with?
<rain1>what i said earlie, can paste to you in pm fyou missed it?
<Apteryx>You are looking for a resolution looking o possible code of conduct violation? I'm not sure how the community handle this.
<rain1>yeah
<Apteryx>I'd be tempted to say that the first thing that must happen is to provide the story of what happened and any available evidence. Have you read submitted such a claim at guix-maintainers@gnu.org?
<Apteryx>The CODE-OF-CONDUCT says: Instances of abusive, harassing, or otherwise unacceptable behavior may
<Apteryx>be reported by contacting a project maintainer at guix-maintainers@gnu.org.
<Apteryx>All complaints will be reviewed and investigated and will
<Apteryx>result in a response that is deemed necessary and appropriate to the
<Apteryx>circumstances. Maintainers are obligated to maintain confidentiality
<Apteryx>with regard to the reporter of an incident.
<rain1>okay thanks, thought it might be easier to just discuss but yeah
<Apteryx>No problem! I hope that your conflict can be resolved and that you can start contributing in high spirits again! :)
<rain1>I hope so
<Apteryx>Any idea why I can't see the build log of python? I'm using: guix build --log-file --no-grafts python
<Apteryx>It says there are no log. If I remove the --no-grafts option I see a log of 2 lines (grafts).
<ng0>Can the cargo build-system currently cope with replacements where something is added, such as: default-features = false because we only have rust stable, not nightly?
<ng0>it should.. i'll give it a try and see what breaks
<jmi2k_>How can I use xorg-configuration-file without slim? I think having this file would solve my problem with Xorg, but I don't know what to do to get it.
***kelsoo1 is now known as kelsoo
<ng0>oh, it's up to the dependending packages to solve that default-features=false thing
<ng0>never mind then
<ng0>s/packages/crates
<alezost>jmi2k_: did you solve your problem with X server?
<jmi2k_>alezost: no, I'm trying a lot of things and none of them work.
<alezost>jmi2k_: do you want to run it with xinit?
<jmi2k_>alezost: yes
<alezost>did you try to specify -modulepath as I wrote?
<ng0>is @code{enum Void { }} valid or do I have to escape the { } ?
<jmi2k_>alezost: no, I think I missed the message let me check...
<jmi2k_>alezost: ok, I can see the messages in the log! I don't know why I didn't receive them...
<ng0>I assume it is and will move on, corrections are best placed in the review then
<ng0>when I'm done in a few hours with polishing drescriptions etc, what's the best way to mass lint check a list of packages?
<alezost>jmi2k_: btw I seem to recall you get problems with making Xorg setuid, did you figure it out?
<jmi2k_>alezost: yes, I was trying to setuid a symlink and it didn't work. It's solved.
<alezost>jmi2k_: great, so it should work in theory; I would try to run it like this: "xinit -- /run/current-system/profile/bin/X :1 vt8 -modulepath /run/current-system/profile/lib/xorg/modules"
<alezost>you can also use "-configdir" if you have some special configuration
<jmi2k_>alezost: it says: The '-modulepath' option cannot be used with elevated privileges
<alezost>jmi2k_: try "/run/setuid-programs/X" instead of "/run/current-system/profile/bin/X"
<jmi2k_>alezost: The same. Let me try without suid
<alezost>it should be inside "Files" section, so...
<alezost>... I would make "/etc/X11/xorg.conf.d/00-path.conf" file with the following contents: http://paste.debian.net/906224
<alezost>jmi2k_: then try to run "xinit -- /run/setuid-programs/X"
<jmi2k_>alezost: It doesn't work
<alezost>jmi2k_: in what way? is there anything interesting in /var/log/Xorg.0.log ?
<jmi2k_>alezost: the same happens: Xorg starts, the windows appear but no keyboard or mouse input.
<quiliro1>how can i include the character set that I am missing? I am not sure this is my problem. It is just that some characters show as squares with numbers inside them
<alezost>jmi2k_: could you paste "/var/log/Xorg.0.log" ? also please tell what xorg drivers you installed
<jmi2k_>alezost: there are some lines in /var/log/Xorg.0.log that can be interesting. Give me some minutes to upload the paste.
***quiliro1 is now known as quiliro
<quiliro>perhaps the applications i am using are not using utf8
<alezost>quiliro: did you install any fonts?
<quiliro>i don't remember
<jmi2k_>alezost: I have xf86-input-{synaptics,evdev,keyboard,mouse}
<quiliro>alezost: how can I tell which fonts I have and which I need?
<alezost>quiliro: "guix package -I" and "guix package -p /run/current-system/profile/ -I" to see the installed packages (in your user profile and the system one)
<jmi2k_2>alezost: http://paste.lisp.org/display/335658
<alezost>jmi2k_2: hm, do you have "/run/current-system/profile/lib/xorg/modules/input/evdev_drv.so" file?
<jmi2k_2>alezost: to see if I could notice any differences, I enabled the slim service and it worked perfect. Then, I saw the Xorg.0.log (this one was generated at ~/.local/share/xorg/Xorg.0.log) and the last lines were pretty much the same.
<jmi2k_2>alezost: Yes
<alezost>is there «LoadModule: "evdev"» line in "~/.local/share/xorg/Xorg.0.log"?
<alezost>jmi2k_2: could you please also paste "~/.local/share/xorg/Xorg.0.log" file?
<jmi2k_2>alezost: give me a minute
<alezost>sure
<alezost>btw it may be that "xf86-input-keyboard" and "xf86-input-mouse" conflict with "xf86-input-evdev"; I would try to remove those 2 drivers
<alezost>ah, no, it would make no difference :-)
<jmi2k_2>alezost: sorry I made a mistake: the file is saved at /var/log/Xorg.0.log, and they don't look alike at all. Wait a minute...
<jmi2k_2>alezost: oh, xf86-input-libinput is also installed. I forgot (I installed all the drivers just to be sure haha) http://paste.lisp.org/+76ZV
<alezost>jmi2k_2: I think "/gnu/store/n3ragpildb1cyn2kg6swgczlnvswiwlz-xorg.conf.d" may contain something missing, I mean something that you need to put in "/etc/X11/xorg.conf.d"
<alezost>jmi2k_2: I think there should be "10-evdev.conf" file that you need to copy to "/etc/X11/xorg.conf.d"
<alezost>it should "tell" X server to use "evdev" driver for keyboard and mouse
<jmi2k_2>alezost: I copied the files to /etc/X11/xorg.conf.d and it doesn't have any effect
<jmi2k_2>alezost: I have to go, thanks for the help ;) I'll keep investigating...
<alezost>jmi2k_2: what files? I meant only "10-evdev.conf". Does xorg log still contain "No input driver specified, ignoring this device."?
<alezost>ok, bye :-)
<dvc>I assume people have noticed the ocaml patches already?
<buenouanq>ERROR: failed to open configuration file '/gnu/store/...-php-7.0.14/etc/php-fpm.conf': No such file or directory
<buenouanq>anyone here use php on guixsd yet?
<dvc>FYI: I'm currently building them and will probably push them tomorrow after scrutenizing them.
<adfeno>buenouanq: It seems to be a bug....
<adfeno>Please report to the project's bug mailing list (the project: Guix).
<adfeno>It's an error in the recipe (which tells Guix how to provide php-fpm to you).
<adfeno>Because the php-fpm recipe should not write "/etc/" to the Guix store ("/gnu/store").
<buenouanq>do you know who packaged it? I don't feel I should report anything because I have no idea how this is supposed to work. Someone has obviously used php here before (therefor it's packaged), but everything written about it assumes a normal file system hierarchy.
<buenouanq>I think I'm probably approching it wrong. Should I be making it into a service or what, I don't know.
<dvc>adfeno: It's reading not writing...
<dvc>ah can't tell
<adfeno>dvc: You're right.
<dvc>could actually be either
<adfeno>The current php-fpm builds seem to expect the file to be inside the store.
<adfeno>buenouanq: I don't use GuixSD, so I can't tell you how services work, but as far as I can tell, php-fpm itself seems to be expecting the configuration files insde the store.
<buenouanq>that was just from me calling it
<buenouanq>I think I can give flags for other config locations
<buenouanq>I'm more interested in THE GUIXSD WAY to doing this.
<dvc>so are using apache or something?
<buenouanq>nginx
<dvc>so it looks like you can set the config file location with php-fpm -y path-to-config
<buenouanq>right, but
<buenouanq>am I supposed to set up a php user and call it with a cron job? am I supposed to make it a service? what user does it get called under then?
<dvc>it sounds like you want to write a service
<dvc>but I'm not familiar enough with php and how it interacts with nginx
<buenouanq>me either, and that's the other half of this problem (;-___-)
<dvc>so you got a basic php site up and running already?
<buenouanq>no, but I understand how you normally go about it on say, debian
<dvc>so the guix service would create a php-fpm configuration file and add it to the store
<buenouanq>this php-fpm thing runs and nginx talks to it via a specific ip:port or unix socket
<dvc>then it would pass it to php-fpm with the -y flag
<dvc>a service is similar to a systemd configuration file with a lot of added coolness
<dvc>so if you can get the website running from the command line you should be able to automate it and turn it into a service
<dvc>but IMO php is one of the ugliest languages ever invented =P
<buenouanq>agreed
<buenouanq>it's horrific
<buenouanq>I'm really doing this as a learning thing.
<buenouanq>KNOW YOUR ENEMY
<braunr>nah php isn't thta ugly
<braunr>but its greatest achievement is probably the "people hate php/perl/python/programming" joke :)
<dvc>what's wrong with python?
<dvc>=P
<dvc>I only hate php and C#... =P
<braunr>c# isn't that bad either
<dvc>it's windows only
<dvc>nothing against the language per se
<braunr>there is mono
<braunr>i actually deployed commercial stuff on mono written in c# at work :)
<dvc>Why would you try to get something to run somewhere where it clearly doesn't want to run?
<braunr>adoption
<dvc>there are 100's of languages to choose from...
<braunr>not really
<dvc>just saying
<braunr>students coming out of school know c# and java basically
<dvc>but gotta go...
<braunr>so when they get a job .. :)
<braunr>ok see you then
<dvc>to be continued =P
<vtomole>Anyone know how to fix "100% packet loss" when i ping?
<buenouanq>vtomole: have you tried connecting to the internet?
<buenouanq>;3
<buenouanq>this is still in a vm?
<buenouanq>I feel it prolly has more to do with how your vm is set up and works than anything guixsd side.
<vtomole>If it wasn't connected to internet it would be "unknown host"
<braunr>yes, how are you connected ?
<vtomole>Im on qemu, I added the -net user tag like lfam said "qemu-system-x86_64 -m 1024 -smp 1 -net user -net nic,model=virtio -boot menu=on -drive file=guixsd.img -drive file=guixsd-usb-install-0.12.0.x86_64-linux"
<braunr>well that's why
<braunr>-net user doesn't provide icmp
<vtomole>ahh
<vtomole>Wait what is icmp?
<braunr> https://en.wikipedia.org/wiki/Internet_Control_Message_Protocol
<braunr>ping uses icmp echo-request and echo-reply messages
<braunr> http://wiki.qemu.org/Documentation/Networking
<braunr>Note - if you are using the (default) SLiRP user networking, then ping (ICMP) will not work, though TCP and UDP will. Don't try to use ping to test your QEMU network configuration!
<buenouanq>neat
<braunr>i usually use a bridged tap device with qemu for that reason
<buenouanq>learn something new every day
<braunr>and also because your vm naturally becomes an easy-to-access server
<vtomole>So should another tag be added to that command? The manual doesn't mention it. How do you do the bridged tap.
<braunr>hum it does
<braunr>−net
<braunr>tap[,vlan=n][,name=name][,fd=h][,ifname=name][,script=file][,downscript=dfile]
<braunr>i tend to use my own qemu.sh scripts though
<braunr>i configure my network with a permanent bridge, say br0
<foobar123>newbie trying guix, xmonad seems to be wrongly packaged as no ghc is present? When i invoke xmonad --recompile it crashes
<foobar123>[env]# xmonad --recompile
<foobar123>xmonad: ghc: runProcess: runInteractiveProcess: exec: does not exist (No such file or directory)
<foobar123>if i install ghc manually too this doesn't happen
<braunr>vtomole: then i use tunctl (with sudo) to create a tap device that my unprivileged user can access
<vtomole>braunr: sounds complicated, no other way to connected guixsd to the internet other than that?
<foobar123>but then xmonad calls an arbitrary version of ghc, not the one it was built with, which probably goes against guix ethos
<braunr>vtomole: and then i use the -net tap,ifname=tap0,script=none,downscript=none option to make qemu use that interface
<braunr>vtomole: well -net user works
<braunr>you just don't have icmp
<adfeno>braunr: Does GuixSD image comes wit wget?
<braunr>i'd be surprised it didn't :p
<braunr>ah the image i don't know
<braunr>i'm just an ehtusiast following the project
<braunr>enthusiast*
<adfeno>If so, I suggest vtomole to try using wget to simply download the index page of gnu.org, by doing: wget gnu.org
<braunr>hum what for ?
<adfeno>braunr: vtomole still thinks his virtual machine doesn't have Internet. So I'm telling vtomole a way to test it.
<braunr>oh
<braunr>sure
<civodul>foobar123: re xmonad, you should probably report it or ask for guidance on guix-devel@gnu.org
<civodul>i don't use it myself but there's a bunch of Haskell-savvy people there :-)
<vtomole>adefno: wget is not available, i'm still trying to install guix.
<civodul>vtomole: in the installation image, you can run "guix package -i wget"
<adfeno>vtomole: civodul's last message was helpful: (19:47:42) civodul: vtomole: in the installation image, you can run "guix package -i wget"
<adfeno>This should also test your network connection.
<adfeno>(because guix package will probably try to download something).
<vtomole>guix package -i wget https://gnu.org give "guix package error:https: unknown package"
<adfeno>vtomole: Please, use only this line: guix package -i wget
<adfeno>This tells `guix package` that you want to install wget (hence the `-i` or `--install` option)
<vtomole>It's just hanging, same thing happens when i ping.
<buenouanq>braunr went over this, you can't ping from in qemu
<adfeno>vtomole: Don't worry, just wait until guix package really exits. It also hangs a little here.
<quiliro>i installed font-google-noto
<quiliro>and then ran fc-cache
<quiliro>but i still get strange characters
<vtomole>buenouanq: I was just comparing both commands
<quiliro>they appear as squares with numbers inside them
<vtomole>adfeno: It's downloading!
<adfeno>See? :)
<buenouanq>you've been connected to the internet this whole time vtomole
<buenouanq>what a confusing pain that was
<vtomole>lol sorry, this is all new to me
<buenouanq>well, only one of us even knew
<adfeno>vtomole: Don't worry, it was also new to me.
<buenouanq>I'd no idea.
<buenouanq>thank you braunr
<adfeno>(knowing that one cannot test Internet connectivity in QEMU through ICMP packets).
<braunr>sure thing :)
<civodul>braunr: good to see you here BTW! :-)
<braunr>civodul: ;)
<quiliro>anyone can give me a tip about what to do about my character problem?
<quiliro>that sounded weird :-|
<foobar123>civodul: ok, i will do so, thanks
<quiliro>it sounded weird when translated literally to Spanish, at least
<civodul>quiliro: what's wrong with your character? :-)
<quiliro>civodul: some characters appear as squares with numbers inside them
<quiliro>woops... i forgot to make my bug report regarding double password question when executing "sudo su -"
<civodul>quiliro: probably means you need to install a suitable font
<civodul>see https://www.gnu.org/software/guix/manual/html_node/Application-Setup.html#X11-Fonts
<civodul>HTH!
<quiliro>civodul: i have font-google-noto which was suggested in this chatroom
<quiliro>and then i ran fc-cache
<quiliro>but still have the same problem for a couple of weeks
<quiliro>since installation of GuixSD
<CharlieBrown>Can I run GuixSD from an SD card? I want a system to boot into that I can play around with.
<civodul>quiliro: in what application?
<adfeno>quiliro: When/where does the problem appear?
<adfeno>(that was the same question civodul just asked)
<civodul>quiliro: the page above mentions gs-fonts, font-dejavu, and font-gnu-freefont-ttf
<quiliro>civodul: let me check
<quiliro>civodul: what page?
<civodul> https://www.gnu.org/software/guix/manual/html_node/Application-Setup.html#X11-Fonts
<quiliro>i think it is only in gnome file manager
<quiliro>ĉ
<quiliro>it is correct here (pidgin)
<adfeno>Oh, I see the "c" with accent here in Pidgin too.
<quiliro>and in clawsmail
<adfeno>Hm....
<buenouanq>anyone have any luck with clawsmail spellcheck yet?
<quiliro>not me
<adfeno>... I wonder if I can remove my file manager (from Trisquel) in order to test the one provided by Guix).
<adfeno>ACTION attempts to remove nautilus package and sees lots of unknown packages to be removed.
<adfeno>... And... that's a "no".
<quiliro>i have another problem
<quiliro>totel will not open
<quiliro>since updating guixsd.
<adfeno>Oh... totem?
<adfeno>Let's see...
<buenouanq>adfeno: you should be able to have both on the same system
<quiliro>the error translated....
<adfeno>I have Totem from Guix here.
<buenouanq>they'll be in different places and won't touch
<adfeno>buenouanq: Hm.... Interesting...
<quiliro>it was not possible to start totem. Some needed addon is missing. Make sure the program is installed correctly.
<adfeno>I might install it then, to try to help quiliro.
<adfeno>quiliro: Hm... I'll try to start Totem here.
<quiliro>Element 'scaletempo' is missing
<quiliro>Element 'autoaudiosink' is missing
<quiliro>(totem:17314): Gtk-WARNING **: A floating object was finalized. This means that someone
<quiliro>called g_object_unref() on an object that had only a floating
<quiliro>reference; the initial floating reference is not owned by anyone
<quiliro>and must be removed with g_object_ref_sink().
<quiliro>how to go back to a previous totem installation?
<adfeno>quiliro: Did it break after a GuixSD upgrade, or after a upgrade with Guix?
<adfeno>s/with Guix?/with `guix package`?/
<civodul>quiliro: i don't use Nautilus ("Files") but maybe you could email that to help-guix@gnu.org
<civodul>it Should Work
<civodul>make sure to specify how you installed GNOME (gnome-service i guess?) and what fonts are in your profile
<adfeno>If it was broken after GuixSD upgrade, try using either `guix system roll-back` or `guix system switch-generation [Number]`.
<cbaines>Evening all, does anyone know if you can point GUIX_PACKAGE_PATH at a subdirectory, so only gnu/packages in Guix?
<quiliro>adfeno: upgrade with guix