<jlicht>sneek: later tell lfam: even with your shepherd config snippet, I keep having the issue that services stopped using eg `herd stop cron' leads to a defunct mcron process. Am I misunderstanding some fundamental thing here, or is this a bug?
<reepca>sneek: later tell civodul: reading through run-container, I notice that nothing is done regarding the propagation type of any of the mount points, which the C++ environment setup code does. If I understand correctly, that means that it's possible that the bind-mounts mounted in the cloned process's namespace will also be mounted in the original namespace (making deleting the temp directory rather more difficult) and also possible that
<reepca>anything mounted under the bind-mounts in a different namespace might become visible inside the container depending on whether the temporary directory is in a MS_SHARED mount point. Is that an issue, and if so, should it be changed in run-container?
<reepca>sneek: later tell civodul: anything mounted under the bind-mounts in a different namespace might become visible inside the container depending on whether the temporary directory is in a MS_SHARED mount point. Is that an issue, and if so, should it be changed in run-container?
<ng0>When I don't know if something will exist but within the possibilities of occuring folders within a given source it can exists, how do I deal with them in the install phase for a build-system I write? Obviously they can not all be used in (let), or can I arrange it with (unless (null)) ?
<sneek>lfam, jlicht says: even with your shepherd config snippet, I keep having the issue that services stopped using eg `herd stop cron' leads to a defunct mcron process. Am I misunderstanding some fundamental thing here, or is this a bug?
<ng0>vim is mostly like: here are a bunch of folders, now I copy this to the destination - you will not always see the same folders but the ones which can exist are predefined, so you will not be able to guess what exists and what doesnÄt
<lfam>They always get defensive when things like this crop up. I'm sympathetic, considering how controversial their software is. But, it seems as if you really need a thick skin to be a publicly known free software developer
<lfam>Once your software gets popular, there is a never-ending stream of complaints, mostly uninformed. Sounds emotionally exhausting.
<lfam>The idea behind that bug is that if root tries to create a service for some user whose name begins with a numeral, systemd will run it as root. So, you could trick your administrator into elevating your privileges through this bug. That's bad.
<davidl>I'm having issues gett git service to work. I installed it and the service is running with default values. There is a repo called /srv/testgit.git with the files test.txt and empty file git-daemon-export-ok but when I run git clone git://<ip-address/testgit.git I get the error fatal: remote error: access denied or repository not exported: /testgit.git
<sneek>fps, lfam says: The upstream repo URL changed, so you won't be able to build guile-git without providing the source code "manually". You'll need to get the right Git checkout, check that it has the correct hash with `guix hash --recursive --exclude-vcs`, and then use `guix download` to put it in your store
<lfam>fps: Yeah, according to the maintainer it's not possible to redirect to the new URL.
<ng0>as far as I know you can't just point to a repository you work from. if you can, I forgot how I ever did it when we worked in an onion based project. it depends on how you want to work. there's gitolite in guix.
<davidl>ng0: alright. I assumed I just had to create /srv/git/myproject && cd /srv/git/myproject && git init . && touch git-daemon-export-ok && echo "hola" >> text.txt && git add . && git commit -am "inital commit" and that would be all.
<balduin>I saw a couple conference talks about Guix. Emacs was mentioned during those talks quite often. I would like to explore Emacs + Guix more but do not know where to find resources/material for that. Do you know some could resources websites/tutorials/books?
<cehteh>and i see that a lot ghc stuff is packaged
<oriansj>cehteh: GHC is written in haskell and requires a haskell to compile it
<cehteh>well that binary blob is free software .. now we rule out self hosted languages as well?
<oriansj>now alot of very useful programs are written in haskell (like xmonad)
<cehteh>i mean yes it sux on some language which cant be bootstrapped easily with C only
<oriansj>cehteh: well it isn't that we forbid such things, but rather we want to be able to bootstrap them from a trusted C compiler to eliminate the Trusting Trust attack
<cehteh>which is the trusted c compiler anyway? :D ..
<oriansj>cehteh: my work is reducing our trusted bootstrap down to a single 280byte hex monitor, which is then used to bootstrap a macro assembler, a lisp and a FORTH, which will be used to create our trusted C compiler
<cehteh>i mean seriously gcc is huge and complex, just define it as trusted isnt really promising
<rain1>but i am working on chroot type setup with guix, not guixsd instead
<balduin>I currently get the following error message while trying to build a package: /bin/sh: bad interpreter: No such file or directory
<rain1>i feel like that might be easier to get by with
<cehteh>it works, but hydra's troubles to keep up, this broken master problem and my unfamilarity with scheme makes this all a slow mess to me, i still try but everything takes 50 times longer than it should
<cehteh>aka guix has a rather high entry barrier still, thats untrue about the ideas behind, writing package descriptions ought to be much simpler than on any other distro
<cehteh>but in practice there are so much other things around to learn
<ng0>Okay I just sent the patches for vim-build-system
<rain1>another trouble i have is no instructions how to set up my own hydra
<cehteh>search the manual for 'publish' server and 'offloading'
<rain1>writing package descriptions ought to be much simpler than on any other distro <- I made a tool that makes it easy to create packages
<catonano>cehteh: I sent some emacs modules as guix packages. Also some python and perl libraries. Packaging stuff is not that hard. But I agree that te entry barrier is hhigh and there's a lot to learn.
<cehteh>that never troubled me. but system configuration is completely new, updates stall, lots things are build locally etc
<efraim>Hydra hasn't built in guix for a long time now
<ng0>what's left to make the switch to cuirass? the "build your own branch" posts I've read made it seem like it is rather functional already
<efraim>I just have a command that I should turn into a script to build packages at random, and I run guix publish on that machine
<balduin>I have to agree with rain. The broken hydra substitute situation is annoying.
<cehteh>i'd wish that publish and offloading could be cascaded, and an offload server could send a signed promise to some server on the upper hierarchy, that would give a complete distributed build farm with maybe one central server (optional)
<sneek>civodul, reepca says: reading through run-container, I notice that nothing is done regarding the propagation type of any of the mount points, which the C++ environment setup code does. If I understand correctly, that means that it's possible that the bind-mounts mounted in the cloned process's namespace will also be mounted in the original namespace (making deleting the temp directory rather more difficult) and also possible that
<sneek>civodul, reepca says: anything mounted under the bind-mounts in a different namespace might become visible inside the container depending on whether the temporary directory is in a MS_SHARED mount point. Is that an issue, and if so, should it be changed in run-container?
<efraim>If I understand correctly, the drv includes a signature from the machine presenting it, if the signature were detached then we could have multiple signers
<janneke>rain1: getting it to run was enough inspiration for me to start contritubing to cuirass
<ng0>coming back to my question: what's left for cuirass to really replace hydra?
<janneke>rain1: depending on your needs it may need work, but it's nice and small
<lfam>The primary issue with cuirass is that its developer quit Guix
<ng0>my use case is once I have enough packages to start with building all the different GUIX_PACKAGE_PATH repositories of infotropique OS and LIVE and provide binary substitutes for them. that should work.
<adfeno>I can't wait for my current packages upgrade to finish so that I can test upgrading the GNUnet package using the package recipe provided by the GNUnet repository itself. I might have found a bug with gnunet-pubish.
<adfeno>Gess.... Building WebKitGTK takes around 6h here. ;)
<adfeno>Hm, I wonder if installing openssl solves the problem. I'm using Guix in a foreign system distribution, so my copy of OpenSSL comes from the foreign system distribution, but I do have a copy of mpv that comes from Guix. Although I n longer use youtube-dl because of a uncofirmed GNU FSDG issue that is being discussed in gnu-linux-libre mailing list --- the workgroup for free/libre system distributions, not just "GNU Linux-libre" as the
<davidl>quiliro: you should be able to boot into an old system configuration regardless of running guix package -u. That's just a system change which I don't think creates a new entry in grub but Im unsure. the guix system reconfigure will definitely include an entry and a rollback.
<rain1>does anybody know why building guix compiles these gnu/packages files?
<rain1>so hmm maybe i should stop it but then I'm not sure how id get them back
<quiliro>guix system: error: build failed: some substitutes for the outputs of derivation `/gnu/store/3hp49xsxwy5mw6jyrq2lncn1nhdfb2z8-linux-libre-4.11.8.drv' failed (usually happens due to networking issues); try `--fallback' to build derivation from source