IRC channel logs
2017-01-02.log
back to list of logs
<davidl>I added it in the list (use-service-modules networking .. ) <jmi2k_>davidl: And you get that dhcp-client-service is unbound? It's weird... <davidl>jmi2k well, recently I get that remove is unbound after trying the method in that link. I also tried adding guix gexp guix store and guix monads because they seemed related to "remove". <davidl>jmi2k_ in the link he had removed wicd-service-type from desktop-services with a remove statement but I get unbound variable remove when trying that. <jmi2k_>davidl: I don't know. Leave a link to a paste of your config. <davidl>jmi2k_ I think I found it - I need srfi-filtering. <davidl>just need to add (srfi srfi-1) probably. <ng0>I have 49 crude (3/4 finished, mostly in need of synopsis+desciption) patches... I know the number will grow to at least 70, should I sent them now to reduce the workload? <lfam>ng0: Wait until everything is ready. It's difficult to review incomplete things. <ng0>all in all they are complete, the thing is, I know someone was working on at least one package in that chain, and I'm no rust user, just happen to need to package lots of packages :) <ng0>the completeness is for a package I can't make public at the moment, as it's for a prototype <lfam>These packages are complete, and they are a dependency of some incomplete package? <ng0>Some packages need a second pair of eyes to control, as I might have missed something (description, synopsis, very rarely: python). The reason why I package them is the full dependency graph of a package which can soon be accessed (but not yet included in any distribution), domain access reasons. <ng0>I think you are right, I should wait some days until I arrived at tokio-core and can rebase again to correct eventual mistakes. It would just feel safer to get a quick review by someone who is more rust wise <ng0>there are some zero dependency packages, but it's difficult to filter them out I think <lfam>If you these dependency packages are done, then go ahead and send them in. <ng0>we could use libc for example.. that's a source for many rust packages <ng0>it was the first package I worked on <ng0>the problem is, rust does neither compile nor complain when something doesn't work after the package is build <ng0>I can add lib to qml, those are safe I think <ile>root@ubuntu-xenial:~# systemctl enable guix-daemon Failed to execute operation: Too many levels of symbolic links <ile>that should have been 2 lines. the command and the error message <ile>also when i try this, it doesn't return to prompt at all: root@ubuntu-xenial:~# ~root/.guix-profile/bin/guix-daemon --build-users-group=guixbuild <ile>also same problems on debian 7, although a little different error message in the first case <lfam>ile: Copy the service file into /etc/systemd/system, and change the value of ExecStart so it executes '/var/guix/profiles/per-user/root/guix-profile/bin/guix-daemon' <lfam>Your version of systemd is too old to support symbolically linked service files <lfam>The guix-daemon is a daemon, so it never returns <davidl>What are the package modules for emacs-paredit? I tried emacs based on the location output of searching for it which works for emacs-wget etc. <ile>thanks, it doesn't quite work yet, says something about locales, but installing "guix package -i glibc-locales" doesn't seem to work either (same error message). have to continue tomorrow. <lfam>ile: That message is just a warning. Guix will still work <lfam>You see it because the proper locales are not available in the environment that the guix-daemon runs in <lfam>It should work when you add the following line to the systemd service file: <lfam>Environment=GUIX_LOCPATH=/var/guix/profiles/per-user/root/guix-profile/lib/locale <lfam>In fact, an equivalent line already exists in the service file we provide <vtomole>"ping -c 3 gnu.org" guves me "unknown host", "ping -c 3 8.8.8.8" works fine. Whats going on? <lfam>vtomole: Can you ping any other named domains? How about google.com? <davidl>vtomole very much sounds like dns-issue which I have had quite some of. You can set resolvers manually in /etc/resolv.conf if you want, globally in wicd-curses and in other ways. <lfam>vtomole: Try restarting the name service cache daemon like this: `herd restart nscd` <vtomole>Still not working, My /etc/resolv.conf has "search HOME nameserver 127.0.1.1" <davidl>vtomole: as a quick fix you can set nameserver 185.121.177.53 in resolv.conf which will probably fix it. it's opennic nameservers. <vtomole>I've been having trouble running guix init, hopefully this solves it. <bldtg>I'm about ready to submit my patch. quick question, does it matter what email I use? should I make a new one for this? do I need to stick with the one I use the first time for future patches? <bldtg>that was multiple questions. oh well <lfam>bldtg: You can use whatever email address you prefer. It will be recorded in the Git log forever <lfam>You can always use a new email address, but the old commits with the old address will not change <bldtg>so, I followed 8.1: Building from Git, and 8.2: Running Guix before it is installed, I modified my package, and built it. Does that sound right? <bldtg>I ran "git status" and it says I modified my package, and a bunch of "po/guix/xx.po" <lfam>bldtg: The .po modifications show up the first time you build the Git checkout. You can commit your package and then discard the .po changes <lfam>vtomole: Re-run the command with --keep-failed, and then share the log of the failing test <bldtg>I think I have to redo the patch. I ran "git diff" and it showed a bunch of weird changes to the file. <bldtg>actually, they're all the same change in several parts. <bldtg>wait hm. maybe that's normal. <lfam>The output of the command should tell you where the build directory was kept. The error messages in your picture show the path of the log relative the build directory: 'tests/cpio.log' <vtomole>lfam:hmm i ran the guix init command in /home, but tests/ is not there <lfam>vtomole: Did you run it with --keep-failed? <vtomole>Yup i'll do it again for good measure <lfam>You don't need to do it again <lfam>When the build fails, a message will be printed telling you where the failed build directory has been kept <lfam>It's probably somewhere in /tmp <lfam>It won't be in /home unless you specifically told Guix to use /home as the location to build in <lfam>vtomole: What about the rest? :) <buenouanq>lfam: do you do php with your nginx too? I'm having trouble thinking about how I'm supposed to approach this on GuixSD. <buenouanq>because I can't install anything as the nginx user, but installed under someone else or root would mean it's not accessible to the nginx user <buenouanq>maybe my issue is that a php service doesn't yet exist? <lfam>Your VM is out of memory. You need to allocate more memory for it <vtomole>lfam: "Can't install grub device on /dev/sdb" <vtomole>my bootflag is on /dev/sda1 and I tried it with that and got the same error except it was "can't install on /dev/sda1" <vtomole>maybe i can change my bootflag to /dev/sdb.. can you do that with fdisk? <buenouanq>(bootloader (grub-configuration (device "/dev/sda"))) <buenouanq>you can't have the partition declared here, just the device <buenouanq>and there has to be a partition on sda that has the bootflag on <vtomole>It worked! I've tried for 3 days to get it to compile and i absolutely do not want to mess it up now. Can i just run reboot? <buenouanq>you can do whatever you want, it's free software :3 <buenouanq>yes, you should be able to reboot without issue <buenouanq>I am curious as to what/why this took 3 days though. <vtomole>The virtual had disk was out of space and i also struggled with making a new partition <buenouanq>partitioning was maybe the least clear part of the install guide. <vtomole>You are not a programmer? what are you then? <buenouanq>to support the greatest modern thing to happen to GNU and *nix in general <buenouanq>A computer using human being who cares about freedom and awesome things. <vtomole>Ahh so you just use free software you don't write it :) <buenouanq>I write little personal things for fun sometimes, nothing worth sharing. <buenouanq>So I lied a bit - But the point still stands that you don't have to be a programmer to use or install GuixSD. <vtomole>Now i'm getting "failed to resolve partition my-root" <buenouanq>my guixsd install was more painless than even debian <buenouanq>and everything just works and i've never encountered any problems <vtomole>Install has to be easier than this. I don't think a lot of people try to run this on VM <buenouanq>yeah, I was shocked at how simple and slick it was <vtomole>from the manual "mkfs.ext4 -L my-root /dev/sda1" but it says that it already contains a file system "gnu-disk-image" <vtomole>Could i just just use "gnu-dik-image" instead of "my-root"? <buenouanq>I assume that the vm is giving your entire partition to the install image. when loaded it spits you in to a root promt right? <vtomole>I'm begging you to try it! I need help! <buenouanq>oh, I'm sorry if you thought I was doing that right now and waiting for me <buenouanq>why not just install it as your fulltime os though, abandon ubuntu or whatever that was <vtomole>In not sure if it willl be able to work, gnu doesn't support proprietary drivers. <jmd>vtomole: Fortunately there is hardware available which works without proprietary drivers. <vtomole>jmd: I need to take guix for a test drive before i buy hardware, that's why i need this vm to work. So frustrating!!! haha <buenouanq>what hardware do you have right now that you need that won't work? <buenouanq>And you seem to have your priorities mixed up. <buenouanq>Freedom is more important than capability, comfort, and convenience. <buenouanq>luckily, as jmd pointed out, we can have it all <rekado>vtomole: it is important that the label you give to the disk matches the label in the config.scm <rekado>it doesn’t have to be “my-root” but it has to be consistent. <buenouanq>I'd actually suggest forgoing the label entierly. <rekado>when you get an error about my-root not being found, then you probably have a disk partition with a different label. <rekado>if you forgo using a label you have to adjust the config.scm to be sure that the partition is uniquely identified. <buenouanq>(file-system (device "/dev/sda2") (mount-point "/") (type "ext4")) <vtomole>I did that earlier mounted on /dev/sdb but it could compile, so i changed to /dev/sda and it did compile, but I did not mount my-root on /dev/sda, but it's the one that compiled <vtomole>Anyway i can go back to where I was? Now every-time I boot i get a guile REPL. When i quit it, I"kernel panic" <buenouanq>/dev/sdb isn't a partition, it should have a number <rekado>vtomole: I replied to your bug report. <buenouanq>the bootloader declaration is different though, it can't be given a partition, just the drive location <rekado>the guile REPL is there to allow you to diagnose the problem <rekado>otherwise you’d get a kernel panic directly. <rekado>the panic is because the root file system cannot be found <rekado>that’s expected behaviour for Linux. <rekado>It doesn’t matter if the target partition is sdb or sda or anything else. <rekado>but you need to be sure which is which <rekado>(the kernel assigns these names randomly) <jmd>If you use the graphical installer, it should prevent you from making those kind of mistakes. <vtomole>rekado: Thanks for the email. Somebody helped me with the network issue by telling me to change the name server in /etc/resolv.conf. I'm rebuilding now. <vtomole>raekado: What did you mean "started from scratch"? <rekado>vtomole: I wouldn’t have copied around files from one partition to the next. <vtomole>The other option was resizing guix1.vdi , but that couldn't boot after resize. <rekado>vtomole: the disk image that you boot from is different from the target disk. I'm not familiar with Virtual Box, but I don't think you are restricted to just one virtual disk. <divansantana>Hello all. Trying to test out guixsd. How do I start sshd in guixsd install image so I can work remotely? <buenouanq>the way the openssh service is described here is kind of confusing <buenouanq>first at the top to your `use-service-modules' add `networking' and `ssh' <buenouanq>then to the services declaration you'll add something like: (service openssh-service-type (openssh-configuration (x11-forwarding? #t))) <buenouanq>how new to this are you I guess I should ask first? have you ever `guix system reconfigure'd before? <buenouanq>what I described above should be changed/added to your os config.scm <buenouanq>then you tell the os to rebuild it self by running (as root): <divansantana>buenouanq: very new and clueless thanks. lol. Hmm so what you describe sounds like it will be applied to the "installed" system. I'm after starting sshd in the "live install image". <divansantana>I'm hoping to try setup the VM as I would my laptop and then later make the switch once I'm more familiar with it. <buenouanq>I'm inclined to think that system reconfigure would still work. <buenouanq>I'm sort of confused though - It is or isn't successfully installed in the VM yet? <divansantana>buenouanq: I'll try that. It's not yet installed. It's just booted the install image. I then was hoping to fire up sshd in the install image and use my emacs editor on my laptop to edit to the config of the future guixsd system. And to back it up. But I think I'm over complicating it. <buenouanq>You should be able to reconfigure the install instance before install in the same way as it's itself just a normal guixsd - But I've never tried nor heard of anyone else doing this. Yeah, I would just install it first adding what I said about openssh from the beginning and you'll be good to go from there. <buenouanq>have you looked through the example os configs yet? <divansantana>buenouanq: great. Thanks I'll do that. Reading them now. From /etc/configuration. <divansantana>buenouanq:it's very different to other linux systems I've tried. But keen to learn and switch and one day get involved. Now on parabola. <buenouanq>I suggest using the barebones one and building it up. <buenouanq>it is very different and I'm still trying to wrap my head around this new way of thinking. <buenouanq>I think you'll find that the install is super slick though. <buenouanq>you declare things in Guile, and then run a command that tells it to build itself based on that file <buenouanq>wow, what I just said is true of almost anything I guess <divansantana>it's great. I'm a bit familiar with puppet. So that helps. <divansantana>I'd love for my laptop to be fully configured via a "manifest" <buenouanq>well, guixsd has it"s rollbacks saved, accessible through grub if you ever really break it <buenouanq>you can just fall back to an os config you know worked <buenouanq>someone in here earlier has been having problems installing it in a vm <buenouanq>if you succeed, stick around and talk to them maybe <divansantana>oh dear. Lol. That's the only place I can install it now.Would love to wipe my laptop but need to work with it tom <buenouanq>systemdebian byorked me over for the last time and I just dove right in <buenouanq>skipped the whole `I'll try it in a vm first' phase <divansantana>buenouanq: the systemd virus is part of the reason to leave arch/parabola which is otherwise awesome. <buenouanq>you'll give up on that idea as soon as you get this installed <buenouanq>guixsd is the future of GNU and everything *nix <buenouanq>in 5 years, everyone will be [poorly] copying what they're doing here <divansantana>buenouanq: I think so. But one still might require another distro for certain purposes. So for that it will be devuan if that app is supported well on there. <buenouanq>it's so obviously the way to go, that we ever didn't do it this way retrospectively just seems crazy <divansantana>buenouanq: yeah agree. Want to learn and get involved as soon as I can with time I have :) <rekado>guix environment --ad-hoc emacs-no-x-toolkit -- sh -c 'tar -c $(guix gc --requisites $GUIX_ENVIRONMENT) | docker import -c "WORKDIR $GUIX_ENVIRONMENT" -c "CMD [\\"bin/emacs\\"]" - emacs' <rekado>unfortunately, this doesn't fully work because I don't know how to run "ln -s $GUIX_ENVIRONMENT /" during the import <rekado>this means that /bin/sh isn't present, which causes quite a few problems. <rekado>would work with a dockerfile, though. <vtomole>buenouanq: "Can't install grub on /dev/sdb1" <rekado>vtomole: grub should be installed on a disk, not a partition <buenouanq>vtomole: the grub line cannot have a partition number, just the disk (/dev/sdb not /dev/sdb1) AND the disk must have a bootable flag <buenouanq>rekado: I think that the partition part of the install guide is glossed over too quickly. Would be great if it had some example partition schemes for me (like myself) who really had no idea what they were doing. <divansantana>a website with links to various ppls system configuration would be great. <buenouanq>the part you should not is what I explained about the openssh service at the bottom <buenouanq>for grub I made a small empty partition on /dev/sda1 and flagged it as bootable - But notice how the bootloader declaration only says /dev/sda <vtomole>buenouanq: I did /dev/sda earlier, it but i didn't add those file systems. my file system was just "my-root" I followed the manual letter by letter... <buenouanq>I find labels annoying and confusing, so I like to just use the explicit paths as you can see. <rekado>that's not safe, because the order of device names is arbitrary <rekado>it could change on any boot depending on which device the kernel sees first <buenouanq>I've worried about that, but they've always been the same on my machine... <rekado>if you don't want to use labels you can use uuids <vtomole>ahh that mean i did exactly what you did. mkfs.ext4 -L myroot /dev/sdb/ and then mount LABEL=my-root /mnt <vtomole>The in boot configuration I did /dev/sdb <vtomole>and just put "my-root" as file system <vtomole>It mounts fine, jst when i run guix system initn /mnt/etc/config.scm it has fun doing what's it's oing for 45 mins or so before stopping with "cannot install GRUB" <buenouanq>what does your bootloader line in your config look like? <buenouanq>oh neat, y'all added a luks thing to the guide <vtomole>(bootloader (grub-configuration (device "/dev/sdb"))) <rekado>okay, here's a full command to build a functioning emacs docker image: <rekado>guix environment --ad-hoc coreutils bash emacs-no-x-toolkit -- sh -c 'tar -c $(guix gc --requisites $GUIX_ENVIRONMENT) | docker import -c "ONBUILD RUN [\\"$GUIX_ENVIRONMENT/bin/ln\\", \\"-s\\", \\"$GUIX_ENVIRONMENT/bin\\", \\"/bin\\"]" - emacs-base' && echo -e "FROM emacs-base\\nCMD [\\"/bin/emacs\\"]" | docker build - <buenouanq>vtomole: and does sdb have a partition on it flagged as bootable? <buenouanq>and why would you want to put your boot there <vtomole>My main disk ran out of memory. The biggest i can mount on was 8MB so i had to create an empty citrual machine disk <vtomole>yes i ran fdisk and the gnu-disk image was in /dev/sda1 <ryanwatkins>Hey guys, I am trying to run GuixSD on my laptop but I am unable to get the USB drive install to work, I just can't seem to boot from the thing. Do I need some specific bootloader or something? <buenouanq>vtomole: what is the gnu-disk image? the install image? <ryanwatkins>I used dd to put the usb drive install onto my usb but it does not work :( <vtomole>buenouanq: I encourage you to run guixsd on vm, it will be easier to know what types of problem i've been running into. <buenouanq>ryanwatkins: have you changed your bios to boot to the usb? maybe it requires you to hold a function key or something? <ryanwatkins>buenouanq: I set my USB to highest priority if that is what you mean? I am usually able to boot, for instance right now I have a archlinux live usb plugged in that works but this is an ISO file <buenouanq>ryanwatkins: then it's prolly something weird and specific to your laptop <vtomole>buenouanq: Yes and it takes up most of the disk. You can't write a "my-root" on it and it runs out of space anyway if you use it for guix-init. <ryanwatkins>buenouanq: Nono, to verify I could boot something from that same USB, I used 'dd' to put an ArchLinux bootable ISO on that same usb instead. <ryanwatkins>buenouanq: I thought maybe my UEFI bootloader perhaps only read bootable ISO images? <ryanwatkins>buenouanq: perhaps there is a way I can use GRUB? I saw some people mentioning chainloading but I did not understand <vtomole>buenouanq: Okay /dev/sdb doesn't work either, /dev/sda works but the system crashes when i try to do anything. The last step is making /dev/sdb bootable? <buenouanq>for you it will be /dev/sdb and the sizes will be different <buenouanq>my bootloader line looks like this: (bootloader (grub-configuration (device "/dev/sda"))) <buenouanq>ryanwatkins: check out rEFInd - I've used it to liberate uefi macbooks before and it works. <buenouanq>vtomole: have you tried doing just a straight qemu vm? <ryanwatkins>buenouanq: I just installed rEFInd now via the arch live CD, hopefully it will help, just dd'ing the GuixSD usb install now again <ryanwatkins>does USB stick installation require the USB to be a specific fs type? <ryanwatkins>I think the issue was that I was trying to Multi-boot. I am erasing my main HDD now and hoping for the system now to detect the USB <quigonjinn>ryanwatkins: if you use 'dd' to write the installation image to a USB, the old filesystem will be overwritten anyway <ng0>i'm a bit stuck in my thought process and need someones opinion <ng0>Rust crates do not complain if something is missing. right now I have interdependent packages to solve, I only saw this after drawing the relations out on paper and my board. Should I skip the bootstrap we normally do, or should I do it? <ryanwatkins>quigonjinn: I am able to boot from USB now but I am getting an error: In procedure fport_fill_input: Input/output error ... any ideas? <ng0>this is just one graph level: tokio-core -> env-logger -> regex -> quickcheck -> aho-corasick -> quickcheck -> env-logger (for feature: use_logging) <ng0>quickcheck cargo.toml: <ng0>env_logger = { version = "0.3", optional = true } <ng0>so in our one dimensional packaging for hydra, should I bootstrap? <ng0>I think it's best if I assume the worst and bootstrap <ng0>is unlicense equal to public domain? <ng0>"in other words, it is a template for dedicating your software to the public domain. " <ryanwatkins>How long does it normally take to boot during GuixSD install, my TTY has hung at "Serivce syslogd has been started" <civodul>ryanwatkins: did it really hang, or is it just that the prompt was printed before that line? <ng0>do we have aliases for licenses? so that I can add unlicense and it will be public-domain? <ryanwatkins>civodul: but I assume something went wrong because I received the line stty: 'standard input': Input/output error previously <ng0>or do we already have unlicense? <civodul>ryanwatkins: yeah the initial screen is a bit messy right now, we'll fix it <ng0>I have a work in progress for this fancy boot screen fedora and others use.. in case anyone wants to take on that, it's on my pagure.io repository <davidl>Im trying to start xmonad but get error saying that /root/.Xauthority doesn't exist and that /root/.xmonad/xmonad-x86_64-linux doesn't exist but both of these exist. <civodul>you're working on lots of things in parallel :-) <ng0>it improved my lack of keeping track of work :) <ng0>what's not on there is the live-system related work, at least some parts, as we're still waiting for the new server <ng0>rust.scm is already at almost 2000 lines <ng0>I hope to submit it before the end of the week, so I can deal with personal/job stuff afterwards <ng0>civodul: what's up with the hidden-service fields in the bayfront machine, is that just for sysadmins or will it be exposed for public access (substitutes) later? <civodul>ng0: it is accessible already to anyone who knows the onion name :-) <civodul>however 'guix substitutes' cannot deal with them <rekado>ng0: re rust.scm exceeding 2000 lines: there are two separate concerns I have <rekado>ng0: if you have packages that are for e.g. web stuff, please place them in web.scm <rekado>ng0: we don't really want to have the python situation <rekado>ng0: the other concern is: how do you intent to keep it all up-to-date after submission? <ng0>the other concern: as long as our applications needs the rust packages (and therefore the chain leading up to them), I will keep them up to date. For web etc, I will move them around but as someone who isn't into rust programming, I can only judge as per descriptions <ng0>as I'm packaging from 0, I need to add some very often used crates anyway <rekado>civodul: do you think we could run a simple application on bayfront that checks for updates and notifies certain people? <ng0>i would submit way less packages if the base wouldn't be flatout zero rust crates. the crate "tokio-core" has a very long chain and 80% are basic crates reused everywhere on crate.io (which is good) <civodul>rekado: something that would run 'guix refresh', roughly? <civodul>i suspect people would quickly be annoyed by the messages and ignore them, though :-) <ng0>I'm okay with messages <ng0>better would be a web view with colors.. this still requires you follow it, but you have it in one place <ng0>for the moment, messages would work I guess <rekado>civodul: yes, nothing much more complicated than "guix refresh" <rekado>a web view would be fine as well, but I wouldn't want it to be public <rekado>another set back in the Haskell bootstrap story: although Hugs seems to be able to run Nhc, Nhc can only generate code for 32 bit architectures. <rekado>so the bootstrapping idea is now: use Hugs to run Nhc on i386, build Nhc native, use Nhc to build old GHC, cross-compile to other architectures, then build later GHCs. <rekado>it's not clear yet whether Nhc can build any version of GHC, though. <rekado>Nhc could be built without Nhc on Hugs, but that would be bootstrapping from generated C files. <rekado>(older versions of GHC can do the same.) <civodul>rekado: impressive that you got this far! <civodul>rekado: the 'guix refresh' web view might actually be easier to do than email <ng0>too many layers of bootstrap with my pckages.. I hope I made the right choices <ng0>caught in a loop... I will just make the decision to have one define-public with one -bootstrap dependency in it, I can't resolve this <ng0>oh no... quickcheck is a dependency for tests and I could have done away with just adding it later <ng0>can I resolve such a process by just adding packages with not all dependencies and then add them after the dependency is packaged? rust build system doesn't really run any tests <ng0>should be allowed and possible <rekado>civodul: dunno, I feel bad about publicly displaying a long list of outdated packages. <rekado>ng0: do you mean haskell's quickcheck? <rekado>quickcheck is an original haskell thing. <jmd>Do we have a service which runs acpid at all? <rekado>civodul: I remember that some time ago we had cross-compile support for i586-hurd; shouldn't we cross-build packages on Hydra? <davidl>Is anyone successfully using xmonad in Guix? <civodul>rekado: i asked phantomas about this on the ML a while back but never got an answer <rekado>I now have access to darnassus to perform some Hurd + Guix experiments. I'll ping him later. <jmd>rekado, civodul: I recall you both mentioned it would be a good idea to have the graphical installer include a page to choose from a list of standard configurations. What do you think would be a good idea to include in that list? <__red__>'sup sneek, how's NE Ohio this morning? <__red__>rekado: I'm extremely interesting in hurd + guix, that was the main selling point over NixOS for me <__red__>any references / threads / sites / tarballs / repo references welcome <rekado>__red__: I'm also pretty new to this. The best person to talk to about this is a person who is known on IRC as phant0mas. <rekado>I got a simple QSort.hs to compile with Nhc on Hugs. It took 2 minutes(!), but the output is identical to the bundled generated C code. <rekado>unfortunately, a chunk of Nhc is implemented as a shell script, which handles all the command line arguments and options. <rekado>porting this to use the interpreted Nhc will take some time, but it looks like it could actually work. <rekado>I guess now it's time to cheat, build Nhc from the generated C sources, and see if Nhc is at all useful for building GHC. <rekado>jmd: IIRC we thought it would be good to have a config for a minimal desktop installation (as it's used in the manual), an option to enable full disk encryption, and a minimal server installation. <jmd>Is full disk encryption actually working yet? <davidl>jmd: Im using it, seems to be no problems so far though that's only after a few days of usage. <jmd>davidl: That's good to hear. I haven't actually tried it myself. <rekado>I will soon reinstall my system. I wonder if I should just wait a little longer until the installer is ready. <davidl>jmd: yep =) though previously it failed after a week or so but then I had installed with LVM. <jmd>Is there currently any documentation at all for FDE installation? <rekado>jmd: the manual mentions encryption a couple of times. <rekado>once it's about preparing the disk <davidl>jmd: I have documented it on my personal wiki/notepad <rekado>another time it's showing the configuration file needed to use the encrypted disk <jmi2k_>I'm installing xmonad, but it complains about not finding ghc. Shouldn't it be a propagated input? <jmi2k_>For example, I can't do xmomad --reconfigure <lfam>efraim: Interesting platfom. Will you select the option that includes at trip to China, including the hot springs? ;) <sneek>Welcome back lfam, you have 1 message. <lfam>Finally something beyond the Cortex A53 <efraim>All I want is the sticker, so I was thinking of ~200 separate $1 pledges <lfam>It's also got some fast I/O. I think it's the first "development" board I've seen for sale like this. <jmi2k_>Can anyone help me? xmonad can't find libXss.so.1, and it doesn't start. <lfam>jmi2k_: Have you figured out what is expected to provide libxss? <lfam>From there, you can take a look at the xmonad package definition to see if it includes that package as an input. <jmi2k_>lfam: No, it doesn't include it as an input <lfam>jmi2k_: Okay, I recommend adding it as an input to the xmonad package definition, rebuilding and upgrading xmonad, and then seeing if it works. Does that make sense? <jmi2k_>lfam: In fact I found another problem: ghc should be a propagated input because it's needed to run commands like xmonad --recompile <jmi2k_>lfam: I'll apply these changes and see if it solves everything <lfam>Alright, sounds like xmonad has some improvements on the way :) <jmi2k_>lfam: it didn't work because it isn't xmonad itself who fails. When xmonad is recompiled, it generates a binary, and this binary complains because it can't find libXss.so.1 <jmi2k_>lfam: about ghc as a propagated input, it worked, so if I solve both I'll submit a patch with the two fixes together. <lfam>jmi2k_: Are you familiar with strace? My next step would be to run xmonad in strace (sounds a little tricky...) to see where it's looking for that library <lfam>I'd also look at the Nix xmonad package, and maybe some other distros, to see what they do <lfam>I'm not sure what you mean by "it isn't xmonad itself who fails. When xmonad is recompiled, it generates a binary, and this binary complains". Sounds like xmonad is the failing thing <lfam>nckx: Thanks for those recent updates to the AWS-related packages <jmi2k_>lfam: xmonad --recompile generates a binary, ~/.xmonad/xmonad-x86_64-linux. I think xmonad looks for it and runs it. I'm not very familiar with the xmonad internals, so I can't give more details. <jmi2k_>lfam: I have to go, I'll look more in depth later. Thanks for the help :) <lfam>Interesting. I wish I knew more. Good luck! <nckx>lfam: heh. I may or may not have triple-checked the signatures for those... (Thanks for the hook, though; I'll try it Real Soonish®) <ng0>Currently our rust bindings for libpsyc are developed and kept inside the same source as libpsyc. Is there a way to start another build system inside a build-system (just system cargo build will not work), or is my assumption that we need to split this off libpsyc, at least in releases, correct? or is there some other idea I can try out? <ng0>where our obviously means psyced/secushare <jmi2k_>Anyone can think a good description for the beep command? It's so simple I can't come up with anything... <civodul>jmd: i'd include the configs that are in the doc: bare-bones, lightweight, GNOME/Xfce <phant0mas>civodul: I didn't forget that email about hydra :-) <phant0mas>we will be able to build binaries for the Hurd on it when we merge core-updates with master <phant0mas>because half of my patches are in the one and half in the other <phant0mas>when we merge master into core-updates we will be able to :-) <Apteryx>Anyone else finding that the git repo's down at Savannah? <Apteryx>Although fetching still doesn't work. The git remote seems to be corrupted. <ng0>as we have a dependency on the crate bitflags with our rust bindings, I wonder if just adding a replacement there (for the local src output of bitflags) would be enough. I don't know. I'm not in favor of keeping everything in one place <ng0>I want to avoid downstream maintenance <ng0>technically this is bundling, right? <davidl>Apteryx: are you doing this from a guix installation? because guix seems to have dns problems at least during install. You can try and ping savannah.gnu.org. I solved it for myself by setting nameservers manually in resolv.conf. <ng0>a solution I have in mind which requires not so much maintenance is to inherit the package and replace the (arguments) and (build-system) with appropriate ones for the crate <Apteryx>davidl: Nope, this is happening from an installed system. DNS seems to be working fine. <davidl>Apteryx: or in the wicd gui on the xfce desktop. <Apteryx>davidl: I had to use static DNS servers in wicd to get started about a month ago too though. I had the same problem as you. <lfam>nckx: The hook works great for me :) <davidl>lfam: yep, that's just how it was for me too. <Apteryx>lfam: I don't think it is the same bug. I didn't experience DNS failures using the installer. The problems appeared after rebooting in my newly installed system; I had no connectivity. <lfam>Apteryx: No connectivity at all? Were you able to get it working? <Apteryx>lfam: Well, no DNS capability. I got it working by finding my ISP's domain name servers addresses and using those as "static DNS" in wicd. <lfam>Hm... not a great first experience <Apteryx>Not sure how DNS is supposed to work by default? I thought most access points (home routers) can be used as a DNS gateway. <bldtg>lfam: about the email you sent about changing the commit message, do I still use git commit? Or, the manual says M-x add-change-log-entry is the easiest way. Do they both do the same thing? <lfam>bldtg: I'm not sure exactly what M-x add-change-log-entry does, because I don't have much experience with magit, the Emacs Git interface. I'd use `git commit --amend` <rekado>add-change-log-entry is not part of magit <rekado>bldtg: use “git commit”. This should open an editor where you can input a commit message. <rekado>CharlieBrown: I haven’t used it but what I heard is that it’s not very usable. <bldtg>lfam: So I did git commit --amend -m "* gnu/packages/gnome.scm (arc-theme): Update to 20161119" <rekado>bldtg: without “-m” your editor will be opened and you can type out a full commit message there <bldtg>rekado: if I do git commit --amend it says to use either the -m or -F option <bldtg>didn't see what it said befor that <lfam>I'm not sure how to handle the combination of the GNU build system and the Python build system <civodul>lfam: you can add a phase taken from one of these build systems <civodul>just make sure you have to two (guix build ...) modules listed in #:modules <ng0>that was exactly the question I had before.. but I hope it's okay that I inherit and separate <ZombieChicken>Interesting. guix: offload: command not found <- Anyone have any ideas? <civodul>ok, but guix-daemon invokes 'guix offload' just in case <civodul>so maybe it's a situation where guix-daemon is more recent than the guix command <civodul>rekado wrote about the issue and workaround recently, rekado? <ZombieChicken>so there is an automagical dep on SSH for no reason whatsoever? <civodul>not really, you choose at configure time whether you want it or not <civodul>if you have ~root/.config/guix/latest, you probably need to run 'guix pull' as root <civodul>or just rm ~root/.config/guix/latest <ZombieChicken>guix pull --no-build-hook doesn't work, nor does guix --no...., and --help doesn't mention it either <civodul>'guix pull' has reached end of life... <jmd>Is there an easy way to "prove" a config.scm file, without actually initialising/reconfiguring anything? <rekado>jmd: do you mean check its syntax? <bldtg>the patches I'm seeing all have the commit message right below the subject, but when I do "git format-patch", the commit message IS the subject. As in "SUBJECT: [PATCH] * gnu/packages/..." <jmd>rekado: Well the syntax would be a start, but its semantics too would be nice. <ng0>bldtg: that's only half true <lfam>bldtg: That's expected, and it's correct <ng0>two parts of the commit message <lfam>The commit title is the subject, and the rest of commit message is part of the patch body. When applying the patch, the right thing happens <rekado>jmd: I guess you could just load the configuration in a REPL where (gnu system) is loaded. <lfam>bldtg: What I had asked is for your patch to have a commit title and message. <rekado>this will give you either a value or an error. <lfam>bldtg: So, the first line of your commit will be "gnu: arc-theme: Update to 20161119." <lfam>This will end up in the SUBJECT line of the patch when you create the patch <lfam>The third line of the commit will be "* gnu/packages/[...]". <lfam>The second line should be blank <lfam>When applying the patch with `git am`, the right thing will happen <bldtg>So I'm good then. I can safely submit my patch? <lfam>Yes, and it's no problem if the patch has some mistake. We can always fix it :) <bldtg>I would like to thank everyone who's helped me. Sorry if some of the questions were obvious or unrelated to guix, I'm really going from 0 to patch here. <lfam>bldtg: We're happy to help :) <ng0>I think the cargo build-system currently ignore build.rs .. I was just informed that qml-rust needs DOtherSide, and I only found out after reading our dependencies <ng0>which is referenced in the cargo.toml <ng0>I have to rebase anyway, but this is one more thing which I found in the last two days <ile>Hi again. Got Guix running after some time (I think), but the manual suggests running nss (or nscd) but I couldn't find how to do that? Should I do "guix package -i nss" or something else? <lfam>ile: NSS and nscd are totally different. NSS provides the certificates requires for the TLS / HTTPS public key infrastructure. You'll need that package if you want to communicate securely over TLS, including HTTPS web sites. <lfam>nscd is a programs that aids in resolving DNS names <lfam>If you're using Guix on a foreign distro, check the manual section 2.6.4 X.509 Certificates <lfam>If you're using GuixSD, check section 7.2.9 X.509 Certificates <ile>I'm running on a foreign distro yes (ubuntu 16.04) <ile>Trying to make a "hosting server" for web app that's easy to replicate <lfam>Ah, there are two things called NSS. The Name Service Switch, as you've linked to, and Mozilla's Network Security Services, which offer TLS and an X.509 certificate store <lfam>I've never noticed that section of the manual (2.6.2). It is "strongly recommended"... <ile>Yes, "strongly" seemed like I should try to do it. :) <ile>Also I'm looking to try Guix as then I wouldn't have to go Docker/Chef/Puppet way... Guix seemed better in that regard from what I read. <lfam>It's more appropriate to try GuixSD if you're looking for a totally declarative system. <lfam>Guix on another distro will still require you to set up your services the old way <ile>I would do that but on that OVH cheapest server ($ 3.50/month) they just support some of the most basic distros... <ile>yes... Maybe I will put this on hold until I get a server that I can install GuixSD on <ng0>"The Material Design icons by Google are released under an Attribution 4.0 International license. The icons are directly copied from Google's GitHub repository at https://github.com/google/material-design-icons." can I include that, comes with the qml-material source, or is this incompatible with guix? <ryanwatkins>Hey guys, I managed to perform a successful install via USB and ran the reboot command. Upon startup now I just receive ERROR: In procedure scm-error: LUKS partition not found etc.. Any ideas? <civodul>ryanwatkins: could it be that the LUKS UUID provided in config.scm does not correspond to your partition's LUKS UUID? <ryanwatkins>civodul: That sounds like the ticket, I will take a quick check. Is there any way in which I could access the config.scm file again without reinstalling everything all over again? <buenouanq>if you followed the guied, that's /etc/config.scm <civodul>but you could boot from the installation image <davidl>ile: I haven't tried this myself but you could probably install your GuixSD conf to vm with guix system vm-disk command and then look at for example lib-virts website on how to setup NAT between host and guest, though this is a bit complicated and requires that the rented server allows for nested virtualization. 