<Jookia>calher: You could just secure it and not hog the bandwidth? <pizzaiolo>calher: why not use your own wifi connection? <calher>Jookia, people in need of WiFi won't be able to get on if they don't live here but are passing by. <calher>pizzaiolo, I can't partition the network right now. <NiAsterisk>you don't have something like a local/statewide meshnet like freifunk? <calher>there's a hyperboria node 30 min away <NiAsterisk>i am confused. okay. so I checked into a branch I had on an external disk with guix sources and it doesn't have problems building in guix environment. I know I sometimes can't access how to do things, temporarily, but this is bad :/ normaly it would go like: git clone $guix-src ; git checkout -b $branchto-workon ; do some changes ; and then the guix environment part, right? <Jookia>I'm pretty proud of my router setup- fully encrypted running librecmc, with the modem bridged with the actual connection done with free software on the router <Jookia>It'd be interesting to see if we could get Guix on a router <pizzaiolo>Jookia: how's your guixSD install coming along? <calher>pizzaiolo, did you add me on tox? <Jookia>I'm considering switching to Riochet from Tox <Jookia>Guix has prsented me with a somewhat useless bootup <Jookia>pizzaiolo: either i'm stuck in GRUB still or there's no graphics initialization <Jookia>oh this would be much better if i could see what i was doing <Jookia>Not sure, I was thinking of trying Ring but Riochet uses Tor <Jookia1>Has anyone managed to boot GuixSD in qemu? <Jookia1>Yes, but it's freezing. Not sure if that's because it's unaccelerated or it's just a bad idea <mark_weaver>Jookia1: it's done all the time. we have a special commands for building qemu images: "guix system vm" and "guix system vmimage" <mark_weaver>I confess I have no experience with this, but civodul and other do this a lot. <Jookia1>Booting Guix using Libreboot seems to not work or at the very least not init my framebuffer <mark_weaver>Jookia1: what kind of machine are you running this on? <Jookia1>mark_weaver: T400 with a ~bad screen~ <mark_weaver>well, fwiw, I'm typing this on GuixSD running on a Libreboot X60, and have used GuixSD on a Libreboot X200 extensively. <mark_weaver>My X200 currently has hardware problems, but paroneayea runs GuixSD on an X200 iirc. <Jookia1>No, the screen works when it's inited by Linux and I don't think Libreboot supports an external monitor <mark_weaver>Jookia1: the Libreboot X200 certainly supports an external monitor, but I don't know about the T400. <Jookia1>It has a VGA port but I don't think GRUB cares <mark_weaver>and I don't know about using *only* an external monitor. I've used an external monitor as a secondary screen, e.g. for watching video on a TV. <mark_weaver>Jookia1: can you access a low-level serial port on the machine? <Jookia1>I can, but that's currently broken in Libreboot- this is more a Libreboot issue :\\ <lfam>Jookia1: There are new instructions for booting GuixSD using QEMU in docs/guix.texi, in HEAD. I don't think they are on the website yet. <kristofer>emacs tells me "the local variables list in /home/kristofer/guix/ contains values that may not be safe (*). <mark_weaver>Jookia1: have you tried booting our USB installer, or are you trying to using our pre-built stuff? <Jookia1>mark_weaver: I'll try the USB installer to see if it's a kernel issue <pizzaiolo>why the heck am I downloading inkscape when installing a barebones guixSD? <xd1le>pizzaiolo: why is it downloading inkscape <xd1le>oh i thought oic stood for "oh i see" <kristofer>pizzaiolo: inkscape is used to make a fancy guix image for grub I think <pizzaiolo>in my case it's useless because grub isn't even installed due to the encrypted disk + libreboot <suitsmeveryfine>pizzaiolo: are you downloading any other weird packages apart from inkscape? <pizzaiolo>suitsmeveryfine: I only did guix system init /mnt/etc/config.scm /mnt <pizzaiolo>I glanced at the screen and saw inkscape, taking a long time to download ***ringst_ is now known as ringst
<CompanionCube>ACTION often tries a guix install of anything with --dry-run first <CompanionCube>since unlike pacman, it does not ask for confirmation so you can be surprised by weird deps <xd1le>yeah just do read -p after the dry-run <mark_weaver>CompanionCube: you can always just Ctrl-C if it looks like more than you want to do <xd1le>CompanionCube: the difference with guix is that it doesn't leave you in a stuffed up state <xd1le>as most other things probably would <CompanionCube>xd1le, shouldn't most package managers have sane handling for ctrl-c <mark_weaver>right, with guix you aren't modifying anything, just making a new profile, and the only thing that affects your environment is the symlink switch at the very end. <xd1le>they could but it's complicated because they usually modify /usr itself as <mark_weaver>and even that symlink switch can be undone with "guix package --roll-back" if needed. <xd1le>so they have to 'unmodify' it all at ctrl-c <CompanionCube>xd1le, yeah but you can just add a signal handler to SIGINT ahead of time <xd1le>CompanionCube: but wouldn't it still need to unmodified? i don't get what adding a sigint handler does <xd1le>sigint handler just fires when you receive sigint <mark_weaver>CompanionCube: sounds like you are making guesses in theory that are unlikely true for real distros. <mark_weaver>with almost all distros, package updates are made in place, affect the entire system, and are generally not reliably reversible. updating the wrong thing could break other things. <xd1le>but like, maybe it's easier to read the dry run output while it stays still (if you make a small script/function) <mark_weaver>with guix, you are just adding more immutable items to the store, and the only modification that happens in a symlink switch as the last step. <mark_weaver>with most distros, it's really important that they ask you if you really want to upgrade all of this stuff, because it could cause damage that is hard to fix. <myglc2>(require 'gnus-article-treat-patch) <myglc2>(setq gnus-article-patch-conditions) <myglc2>(require 'gnus-article-treat-patch) <myglc2>(setq gnus-article-patch-conditions) <myglc2>This keeps up, mama going to take away the keyboard! <mark_weaver>pizzaiolo: our fora are (1) this channel, and (2) the mailing lists. <pizzaiolo>mark_weaver: are there any plans to change the server hardware? <myglc2>mark_weaver, pizzaiolo: as a truly clueless user I can attest these are working really well <mark_weaver>but no, we haven't purchased the hardware yet. we're working on spec'ing it out, and deciding where to host it. <mark_weaver>it's made more complicated by the lack of high-end pre-built libreboot server hardware. <pizzaiolo>do you have the funding for it? I head the FSF is helping you out financially <mark_weaver>yes, we have the money we need now. the FSF set up a donation system for us. <mark_weaver>_`_: sounds good. francis7 was telling me about this a while ago. sounds promising. <_`_>mark_weaver: I believe tpearson first needs enough interest shown. But he already has some produced for demonstration purposes. <mark_weaver>iyzsong: I've already done it locally. waiting on a test build. <mark_weaver>but I suppose I could just push it and hope it builds. <pizzaiolo>mark_weaver: what was that packaging guide you sent suitsmeveryfine earlier? <iyzsong>yes, I have got it build and epiphany work fine :3 <mark_weaver>iyzsong: our webkitgtk-2.4.9 packages are a problem :-( <mark_weaver>wingo said that he thought there was a patch for yelp to use the webkit2 api <iyzsong>I think we can't do much thing about 2.4.9, the applications using it need to be ported. the new version only support gtk3 and webkit2 api. <mark_weaver>wingo seemed to think the new webkitgtk would work with gtk3, but I guess he wasn't sure and maybe you are :) <mark_weaver>but in any case, it definitely doesn't support the old webkit api <iyzsong>will update yelp. yes, i'm sure gtk2 support is droped. <myglc2`>./pre-inst-env guix system reconfigure ../config.d.scm <myglc2`>guix system: error: failed to connect to `/usr/local/var/guix/daemon-socket/socket': No such file or directory <myglc2`> |-guix-daemon,271 --build-users-group guixbuild --substitute-urlshttp: <myglc2`> | `-guix-daemon,2449 2440 guixbuild --substitute-urlshttp: <myglc2`>Sorry that didn't wrap very readably <mark_weaver>myglc2`: you need to pass --localstatedir=/var to Guix's ./configure script. <myglc2`>guix system: error: symlink: Permission denied: "/var/guix/profiles/system-15-link" <mark_weaver>myglc2`: you need to run "guix system reconfigure" as root <myglc2`>sudo ./pre-inst-env guix system reconfigure ../config.d.scm <myglc2`>mark_weaver: YEAH! Ludo has fixed the "?" that should have been "'"s ***Digitteknohippie is now known as Digit
<kristofer>mark_weaver: you nentioned that you symlink ~/.config/guix/latest a symlink to your local guix build.. does that happen automatically with a make install? and can you elaborate on what to do with root? <mark_weaver>kristofer: I don't run "make install", and I recommend against it. <mark_weaver>after you've built the local guix and make sure it works, just make ~USER/.config/guix/latest a symlink to the git checkout directory, and then 'guix', when run as USER, will use the copy in that source directory. <mark_weaver>the 'guix' command is rigged to look in $HOME/.config/guix/latest and use the copy of guix in that directory. <kristofer>mark_weaver: is make test enough to confirm that my local copy is functioning properly? <mark_weaver>kristofer: no, you should try building something with it, e.g. "./pre-inst-env guix build emacs" <mark_weaver>./pre-inst-env is the way to run 'guix' from a built source directory. after the ~/.config/guix/latest symlink is in place, you can omit it. <kristofer>my installation from last night won't install guile-sdl currently :-/ <mark_weaver>does it fail to build, or just fail to run properly? <mark_weaver>I guess whoever contributed the package was on x86_64 and failed to notice the problem. <kristofer>mark_weaver: I don't already have ~/.config/guix <kristofer>do I need to restart any services after making the symlinks? <mark_weaver>kristofer: "guix pull" is what creates it. if you've never run "guix pull" as that user, then it won't exist. I've never run "guix pull", so I just created it manually. <mark_weaver>no, you don't have to restart anything after creating the symlinks. <kristofer>I keep getting this bizarre warning with emacs <kristofer>The localvariables list in /home/kristofer/guix contains values that may not be safe (*). <mark_weaver>yes, it's safe, it's just that emacs can't tell that automatically <rekado>mark_weaver: shotwell, guitarix, and gnucash were added by me. I'll check if a later version of webkit can be used. <rekado>guitarix only added webkit as a dependency with the last release. <rekado>sneek: later tell davexunit At FOSDEM I discussed with Manolis about our avr-gcc and he suggested we try building a second stage with the avr-libc. This should at least fix the problems with paths (finding things like crtm328p.o). <civodul>ACTION posts python2-with-package fix <iyzsong>anyone can start a job to build the 'gstreamer-update' branch on hydra? <civodul>iyzsong: i think we'll wait for security-updates to complete <francis7>oh wow, guix has a lot of people in the channel <francis7>did it quadruple or something last time I checked? <Jookia>I wonder when my patch will be reviewed <Jookia>:o It magically got reviewd and patched just now <Jookia>Oh, I see now. I suppose that also answers a few more questions- Even for one line changes you need copyright headers to be updated <Jookia>How much work would it be to ensure every package is downloadable over Tor? Could this become a policy? <efraim>you can do `torify guix command`, I find my connection to hydra is sometimes better that way <Jookia>efraim: Oh, I mean building packages- lots of things that use FTP are unreliable <Jookia>I suppose I could work on a lint feature that checks if the URLs of sources work through Tor <Jookia>Is Guix documentation in a separate repo? <efraim>grep says there are 69 occurance of "ftp in gnu/packages/*scm <efraim>the documentation is in the main repo <Jookia>I mean the documentation on things like configuration <efraim>everything should be in the git repo or in the mailinglist <cajg>So, I got an error message in the installer to the effect that GRUB couldn't install <cajg>but I gather that that might not mean that GRUB hasn't installed <cajg>I've rebooted the installer and re-mounted the ppartitions <cajg>and there's nothing on the /boot partition <efraim>unfortunately I have nothing, I haven't gone through the install process yet <cajg>does anyone have anything I could try from the command line? <cajg>Path `/mnt/boot/grub' is not readable by GRUB on boot. Installation is impossible. Aborting. <cajg>guix system: error: failed to install GRUB on device '/dev/sda' <mark_weaver>what kind of filesystem is /mnt/boot on? is it encrypted? LVM? <cajg>no, just ext4 with a bios boot flag as I remember <mark_weaver>well, I don't know then, and now I have to go afk. maybe someone else has ideas, or you could email bug-guix@gnu.org about it. <iyzsong>cajg: hi! can you run it manually with debug as '/gnu/store/*grub*/sbin/grub-install --debug --boot-directory /mnt/boot /dev/sda", hope the log tell more. <wingo>i want to autoregen with an updated autotools etc to cut the guile 2.1.2 tarball but so slo <wingo>i was a little spooked that a guile-2.1.2.tar.xz that i cut on one machine differed in sha256sum from one cut on another machine <wingo>i guess it's timestamps, but it could be autogen too, or something related to the texinfo pipeline <wingo>i wonder if there's a "best practices" for cutting reproducible tarballs <mark_weaver>wingo: I guess it's because of the .go files, and the fact that our macro expander produces non-deterministic output. <wingo>mark_weaver: ah yeah, that too. grr. <cajg>ah, iyzsong, I think you've given me a clue, perhaps it should be installed to /dev/sda1 if that is the partition I have prepared /boot? <cajg>so I should change my congig.scm <iyzsong>cajg: well, /dev/sda is the right value for 'grub-device'. ***Digitteknohippie is now known as Digit
<cajg>Huh, installation finished, no error reported <iyzsong>pizzaiolo: this look strange, can you get the guile REPL? or find some way to look /gnu/store/c4..-etc/localtime .. <cajg> /boot/grub 'n' stuff all there <pizzaiolo>iyzsong: thanks for the help. how can I get the guile REPL? <iyzsong>pizzaiolo: oh, before the kernel panic, 'new prompt' is the REPL. but I guess it's broken. this is the first reboot after install right? then, I guess you have to reinstall. sorry :( <iyzsong>pizzaiolo: you can check the '/gnu/store/*-etc/localtime' during install to see it's actually there. but I have no idea what's going wrong.. <mark_weaver>wingo: I don't feel strongly about it, but I would prefer that the pre-built .go files be packaged in a separate tarball. that would also solve the reproducibility problem for the main source tarball at least. <suitsmeveryfine>pizzaiolo: you could also try to use the desktop config instead of bare-bones <pizzaiolo>iyzsong: this is the prompt just before the kernel panic: ^ <mark_weaver>suitsmeveryfine: the order of the fields in the OS definition certainly doesn't matter. <iyzsong>pizzaiolo: oh, so after you enter the right password, it panic? <suitsmeveryfine>mark_weaver: yes, I didn't think it would but I see no differences in the pizzaiolos file compared to mine <petter>pizzaiolo: what did your cryptsetup command look like? <wingo>mark_weaver: ok, noted. i think i prefer them in the tarball because we want to give a good default experience, and we should make sure .go files are reproducible in any case. but we can iterate on this. <pizzaiolo># cryptsetup -v --cipher serpent-xts-plain64 --key-size 512 --hash whirlpool --use-random --verify-passphrase luksFormat /dev/<your-encrypted-root-partition> <wingo>mark_weaver: it's probably fine to use a deterministic syntax-session-id for prebuilt files, given that they will not be used at all once bootstrap is built. but maybe i misunderstand how session ids propagate. <wingo>or we could use the version as the session id ;) <mark_weaver>wingo: it has to be a different session id for each "guild compile" command <pizzaiolo>iyzsong: how can I check the localtime file? <wingo>in practice each session id will be distinct across "guild compile" invocations because they will point to different ELF files i think <iyzsong>I haven't get familiar with disk encryption stuff :- <wingo>and they are strings, not symbols. but i could be forgetting :) <civodul>wingo: 'make dist' produces things non-deterministically because of user names, time stamps on the archive's files, and the fact that it doesn't use gzip's '-n' flag by default <petter>suitsmeveryfine: i missed most of it because there were problems with the train signals :/ <iyzsong>pizzaiolo: after boot the usb install media, you can manually mount and de-crypt the root partition, and look the file. <civodul>mark_weaver, wingo: it'd be awesome if we could fix the syntax-session-id thing for 2.0.12 <wingo>civodul: there's probably a way to fix that. i'll just make sure to bootstrap with latest autotools. <wingo>for 2.0.12 i don't know, i am focused on 2.1.x prereleases as you know <pizzaiolo>iyzsong: 'nano /gnu/store/*-etc/localtime' ? <wingo>never edit anything in the store :) <wingo>it will only lead to sadness <pizzaiolo>wingo: how can I look at it without editing? <iyzsong>pizzaiolo: it's ok. but 'cat' will do. make sure look in the mounted 'root' partition. <iyzsong>pizzaiolo: I guess you just look the store in the usb install image, not the root partition.. <wingo>i like the warmer look that you get from building texlive from hand; these modern binary packages just can't produce that depth of typography <iyzsong>pizzaiolo: I suggest you can try first install without LUKS, I think it will make debug more easy :) <wingo>you get better ligatures if you fetch your texlive tarball from tug.org, but only over ftp ***kelsoo1 is now known as kelsoo
<pizzaiolo>anyone has any other ideas before I reinstall with an unencypted partition? petter suitsmeveryfine Jookia mark_weaver <suitsmeveryfine>pizzaiolo: considering that your setup is identical to mine it should actually work for you <suitsmeveryfine>pizzaiolo: is it possible that you made something wrong when first creating the file system? <petter>yeah, i don't know either. Maybe try a different cryptsetup command sometime. (I don't know if your issues are related to disk encryption though) <pizzaiolo>suitsmeveryfine: I followed petter's guide to the letter <petter>pizzaiolo: what computer do you have? <suitsmeveryfine>pizzaiolo: the only difference is that I used the desktop condig file and not bare bones <suitsmeveryfine>...and the order of modules in initial RAM which mark_weaver says doesn't make any difference <petter>suitsmeveryfine: can you show your config? <suitsmeveryfine>petter: no now unfortunately; I don't have access to the macbook at the moment <Jookia>suitsmeveryfine: what Guix version are you on <petter>suitsmeveryfine: that would be good, thanks. I don't have a initrd field in my config, so i'm not sure how this should be. <Jookia>pizzaiolo: are you using the git version of guix? ie, git pull? <Jookia>suitsmeveryfine: when did you last guix pull <Jookia>If your configs are identical, then the problem must be in your Guix <Jookia>a network failure at boot killing the system? <pizzaiolo>Jookia: no no, when I do guix system init, the connection is frequently severed <pizzaiolo>cajg: did your install go smoothly or did you experience connection failures? <Jookia>suitsmeveryfine: sounds like you two need to try and sync your configuration + guix versions + libreboot versions some more <cajg>pizzaiolo: install went ok, but that was yesterday evening (London time) <suitsmeveryfine>Jookia: My configuration should be somewhat different at least since I used the desktop version as a basis <Jookia>suitsmeveryfine: either way, this would rule out errors <mark_weaver><wingo> in practice each session id will be distinct across "guild compile" invocations because they will point to different ELF files i think <pizzaiolo>suitsmeveryfine: I'm using Europe/Paris in the config <mark_weaver>if the session id is simply the version, then it won't be distinct across invocations. <wingo>mark_weaver: distinct in an eq? sense <wingo>i don't recall how some of these things work tho <pizzaiolo>suitsmeveryfine: wed feb 3 14:42:17 CET (whereas time in brazil is 11:42) <mark_weaver>wingo: ah, well, that doesn't matter because the session id is used to create marks, which are symbols. <mark_weaver>and we can't guarantee that if we use the same session id for two different invocations of guile that generate code. <mark_weaver>see 'new-mark' in psyntax.scm. the names in marks are composed of the session-id and a gensym number. <wingo>there are other ways of predictably generating session ids of course, e.g. version+path <mark_weaver>the gensym numbers created by two invocations will overlap, so the session numbers must be different. <pizzaiolo>you know when I used less to see localtime? well now my unicode characters are all messed up <wingo>but i dunno, i can't page this in right now <Jookia>pizzaiolo: use 'strings' to see your localtime <pizzaiolo>suitsmeveryfine: whoa I just got an unknown key 0xff bug in libreboot <mark_weaver>wingo: I wonder if we should switch to representing marks as strings. that might be the thing. <pizzaiolo>I just rebooted, kept pressing the downward arrow and the bug didn't happen again <pizzaiolo>I just love when bugs can't be reproduced :P <suitsmeveryfine>pizzaiolo: I flashed like 20 libreboot versions to track down the bugs <pizzaiolo>Jookia suitsmeveryfine my guile version is 2.0.11 <Jookia>pizzaiolo: You could try doing an unencrypted install, but most likely you could get in contact with suitsmeveryfine and see how you can synchronize systems to rule out software error <NiAsterisk>hi! which one of those would pass in a package definition? 1) (arguments `(#:make-flags '("-f Makefile.unx"))) 2) (arguments `(#:make-flags '("-f" "Makefile.unx"))) or 3) (arguments `(#:make-flags "-f Makefile.unx")) I am having trouble with (3) currently, which gives something like Odd number of arguments. <taylan>NiAsterisk: "-f" "Makefile.unx" is the right way <taylan>make-flags has to be a list of strings, which will become the strings in the ARGV of the make process <NiAsterisk>okay. it's a bit of uncharted knowledge for me packaging this way, learning while writing. I started taking notes, hopefully to improve the manual or an beginners guide at some point. <NiAsterisk>after poking around and confusing error outputs with a bit preoccupied mind because of illness, I realized I need notes :) <mark_weaver>pizzaiolo: the relevant error is this: No such file or directory: "/etc/static/localtime" <NiAsterisk>I also catched up with older threads on the devel list yesterday.. packagename-0.0-$guixversion.'commit trimmed to 7' is acceptable, right? as in lispf4-0.0-1.sasaqsa <mark_weaver>civodul: pizzaiolo's fresh install is failing to boot (panic due to PID 1 exiting) with the error message above. <mark_weaver>pizzaiolo: when you ran "guix system init", it ended with the error "Path `/mnt/boot/grub' is not readable by GRUB on boot. Installation is impossible. Aborting.", right? <pizzaiolo>mark_weaver: pidgin forces OTR, I can't reply in PM <pizzaiolo>mark_weaver: I don't make vegan pizzas, but I am vegan :P <pizzaiolo>if there's any mushrooms involved, I bet they're awesome :P <NiAsterisk>no output (with guix lint) is good output, right? <rekado>(I make veggie pizzas too; if anyone has a known-good vegan pizza recipe I'd love to get it!) <NiAsterisk>me. but I'd have to translate it first. I could ask for permission and put it online later on. iirc it's just for pizza drough + an optional cheese-like sauce <civodul>(much better than those frozen pizzas we talk about ;-)) <sneek>Welcome back davexunit, you have 1 message. <sneek>davexunit, rekado says: At FOSDEM I discussed with Manolis about our avr-gcc and he suggested we try building a second stage with the avr-libc. This should at least fix the problems with paths (finding things like crtm328p.o). <efraim>second stage might be like the intermediate stage in bootstrapping <NiAsterisk>:/ damn. doesn't "git clone https://domain.bla/bla.git" guix hash --recursive bla/ give the right hash? environment, what I put into the package definition AND another run afterwards with guix hash all differ ._. <pizzaiolo>civodul: any ideas about mark_weaver's comment above? I'm having installation issues <NiAsterisk>and in the environment running guix hash --recursive bla/ also produces something else. <civodul>pizzaiolo: could you access this partition from some removable media (possibly the USB install)? <pizzaiolo>civodul: that's what I'm doing at the moment <civodul>cool; could you inspect /etc/static and /etc? <cajg>so the installer is installing grub to the mapped device, sda2 on /mnt/boot <cajg>payin no attention to my config declarations that /boot is on /dev/sda1 <cajg>I'd be very grateful if someone could point me to the installer scripts <cajg>this really isn't ging well <civodul>pizzaiolo: first "ls -ld /etc/static /etc /etc/static/localtime /etc/localtime" <rekado>davexunit: I don't really know how to put it (or how to achieve it), but we need a real cross-compiler for AVR, not just the sans-libc cross-compiler as it uses the wrong paths. <rekado>I tried rebuilding the cross-compiler with the avr-libc on the train, but I haven't yet tested it on actual code. <NiAsterisk>forget what I just said. might be the wrong hash <civodul>pizzaiolo: i mean, do all these files exist? <pizzaiolo>I'm not sure /etc/localtime exists, because it's not light-blue colored like the other two <civodul>pizzaiolo: what about "ls /gnu/store/c4az*/etc/"? is there a 'localtime' file there? <civodul>pizzaiolo: if 'ls' doesn't report "No such file or directory", then it exists <civodul>pizzaiolo: did you ever modify stuff in /gnu/store on that machine? <civodul>pizzaiolo: "no such file" for which one? <davexunit>rekado: this is stressing my compiler knowledge. :) <davexunit>rekado: so how did you rebuild the cross-compiler? <Dezponia>Buildservers for Guix. What sort of need are we talking about here? <pizzaiolo>ls: cannot access /gnu/store/c4az*/etc/: no such file or directory <Dezponia>Around 3000 pacakges according the FSF article linked in the channel title. Thats modest but I guess it deppends on how often they're rebuilt etc <davexunit>Dezponia: our biggest need right now is a new frontend for the build farm. <civodul>pizzaiolo: i guess on your system it's actually /mnt/gnu/store/c4az* because you mounted it, right? <civodul>pizzaiolo: i'm going afk for an hour or so <Dezponia>davexunit: Interesting. What would be needed in terms of hardware/bandwidht for that, roughly? <pizzaiolo>civodul: with /mnt/ it still does not find the file/dir <davexunit>Dezponia: I actually don't have those figures, because I don't maintain the machine, but someone else here might be able to chime in about that. <Dezponia>davexunit: I'll stick around for a bit incase someone knows :) <davexunit>AFAIK, we're planning on using a recently-Librebooted server motherboard for the front-end <Dezponia>davexunit: I'm planning to build a server using that board to migrate my own stuff to. I run a few buildservers for GNU/Linux distros on my current hypervisor but would rather do so on the Libreboot-D16 with a FSF approved distro on the host (already run that on the current host but want more power and libereated firmware). Not sure if that setup is sufficent for your needs but cant hurt to offer :) <davexunit>Dezponia: it would surely be sufficient as a worker in the build farm. <Dezponia>davexunit: Then its still interesting at least. Planning to load it up with some pretty beefy specs so it should work fine :) <suitsmeveryfine>pizzaiolo: you can find my working OS definition here: paste.lisp.org/+6KgK <pizzaiolo>the first system works, the second is broken <davexunit>pizzaiolo: the order of the fields in the 'operating-system' form doesn't matter <pizzaiolo>davexunit: maybe it's a problem with the barebones install? <pizzaiolo>suitsmeveryfine: if I install with xfce maybe it would work? <pizzaiolo>that's another difference between our installs <davexunit>pizzaiolo: you can be running different versions of guix <davexunit>it seems that our migration from dmd -> shepherd has caused some hiccups <davexunit>pizzaiolo: yes, but the version of guix could be different. <davexunit>if you are using something built from git that won't tell you enough information <davexunit>if the 2 configs are really the exact same sans transposing some fields, then that's the most likely culprit. <pizzaiolo>davexunit: I haven't built anything from git, I grabbed guix from the download section of the website <suitsmeveryfine>pizzaiolo: but you grabbed newer packages over the Internet when installing <suitsmeveryfine>pizzaiolo: why not just copy my config file exactly and try to install <fhmgufs>How do you use git send-email? I mean, with postfix? <sneek>fhmgufs, you have 1 message. <sneek>fhmgufs, davexunit says: I agree that 'guix pull' is too slow. we need to find ways to make it faster. <fhmgufs>davexunit: Couldn't 'guix pull' leave the packages and so on out, so that they are built when they are first used by guile? <davexunit>we improved the way we compile the guile modules. <davexunit>fhmgufs: if we didn't compile them, the first time you ran something like 'guix package -A foo' it would load every single package module and compile them. <davexunit>any operation that wanted to scan the full package set would do this. <fhmgufs>Ok, understood. I'm not so familiar with guile, yet. <fhmgufs>And also not so familiar with how guix does things in detail. <fhmgufs>To come back to my first question: I don't really know how to configure postfix or sth like that for 'git send-email' to use my mail provider. <NiAsterisk>found that github has done an zip package of lispf4, I guess that's more appreciated? what else is needed to unzip this .zip other than the gnu packages unzip as module? I tried grep and found no .zip for sources <jin>hi guixs, i am testing nautilus file manager and i've error when trying to connect via smb <bavier>NiAsterisk: you're wanting to build a package from a zip archive? <jin>this error display: gvfsd-network <NiAsterisk>bavier: I have the choice between git checkout or github thrown together master.zip <bavier>yes, I think we'd prefer the zip <bavier>there are a few other packages that use zip <bavier>faad, faust, microscheme, couger <suitsmeveryfine>I just reconfigured my system and when I try to reboot or halt I get "error: connect: /var/run/shepherd/socket: file or directory does not exist" <bavier>NiAsterisk: I think you really just need to add the "unzip" package to 'native-inputs' <NiAsterisk>I still get: tar: This does not look like a tar archive <bavier>suitsmeveryfine: try to run reboot from /run/booted-system/profile/sbin/reboot instead <bavier>suitsmeveryfine: your system was last booted with dmd, and the reconfigure introduced sheperd <bavier>NiAsterisk: does the filename end in ".zip"? <bavier>NiAsterisk: what's the build-system? <bavier>suitsmeveryfine: you'll need root or sudo <bavier>NiAsterisk: interesting, are there patches or snippets for the source? <bavier>I'm not sure patch-and-repack can handle zip files, but the unpack phase from gnu-build-system should be just fine with zips <bavier>suitsmeveryfine: what's the error? <suitsmeveryfine>error: connect: /var/run/shepherd/socket: file or directory does not exist <bavier>suitsmeveryfine: if you're in /run/booted-system/profile/sbin, make sure you run halt via `./halt' so that you aren't getting the one from PATH <bavier>NiAsterisk: I see, include the ".zip" in the origin file-name field, then it should work <suitsmeveryfine>pizzaiolo: I get a kernel panic also after having reconfigured guix sd <bavier>I didn't follow the conversations recently regarding kernel panics, so I don't think I'll be able to help here <suitsmeveryfine>his installation failed and it was suggested that perhaps the latest version of guixsd was causing it <efraim>NiAsterisk: instead of master.zip, you can change the zip to .tar.gz <NiAsterisk>it's not the issue anymore. now it's up to stuff failing to compile. but: can I? is it interchangeable? <bavier>NiAsterisk: actually, in this case, a git checkout might be preferable, since we can specify a commit to checkout, rather than getting whatever is the current HEAD. <bavier>but feel free to fix the rest of things before revisiting that <NiAsterisk>it's not like this is ever getting worked on again.. or at least not very likely. it has been ported from the 80s version, and some stuff added and that's it. I had it using a git checkout before I discovered the zip <bavier>by specifying a commit we can avoid the possibility of future breakage completely <NiAsterisk>and if commits happen again, I can test and change the commit <efraim>the git commit tag is near the "download zip" button on github <NiAsterisk>I know the git commit, it's commented out in the uploaded file <NiAsterisk>whoa, 78.9 MB documentation because of scans and inclusion of old documentation.. I'll split that later <pizzaiolo>suitsmeveryfine: kernel panic? what did you do? <paroneayea>I need to do something to make guix system vm work... <suitsmeveryfine>pizzaiolo: so the problem might actually lie with the latest guixSD version <pizzaiolo>suitsmeveryfine: in any case, I've copied your configs <pizzaiolo>suitsmeveryfine: still, it's good because I want xfce <pizzaiolo>suitsmeveryfine: so you updated guixSD and when you rebooted had a kernel panic? <hoijui>i just went to the download section of the Guix homepage, and i was very confused, until i read the whole page (read the first, unseeming paragraph last) <hoijui>i think it would make sense to write "Operating System" as title for GuixSD, and "Package Manager" as title for Guix binary <hoijui>that is really the first thing a user has to know when reaching this page <taylan>hoijui: below the title it says "As of version 0.9.0, the Guix System Distribution can be installed on an i686 or x86_64 machine. It uses the Linux-Libre kernel and the GNU dmd init system. Alternately, its package manager, GNU Guix, can be installed as an additional package manager on top of an installed Linux-based system." <taylan>maybe it could be made more apparent <wingo>ACTION locale issues again, somehow :/ <NiAsterisk>does gnu-build-system imply ./configure by default, as this package builds without a configure file on debian (it does not exist) <wingo>mark_weaver: this machine has guix on nixos. it has GUIX_LOCPATH and LOCPATH set. it can't install locales any more though. <mark_weaver>NiAsterisk: gnu-build-system does have a 'configure' phase that runs ./configure, but that phase can be removed or replaced. we tend to use gnu-build-system even for packages without a configure script. <hoijui>taylan, yes exactly, that is how i found out, but that isa text that looks very unimportant by its formatting, while it is the most important text of hte page <wingo>i'm not sure i have time to understand this again <efraim>I'm loving the -M build option, even on my netbook <taylan>hoijui: feel free to file a bug report by mailing bug-guix@gnu.org <NiAsterisk>mark_weaver: I mean, there is an override for this in (arguments)? (related error in the build process: /gnu/store/l7px210li6zviymgvp3cps6n48x7fgpl-bash-4.3.42/bin/bash: ./configure: No such file or directory , there's no configure file <taylan>hoijui: thanks for the report :-) <hoijui>taylan, i dont have to subscribe first? <taylan>hoijui: pretty sure not for the bug MLs. on the other ones, mails by unsubscribed addresses have to be reviewed by the admin AFAIK. <mark_weaver>overview of the way we dealt with the messy glibc 2.22 -> 2.23 transition which broke compatibility with locales. <wingo>somehow i had it fixed before and it's unfixed now <wingo>also, every time i try to download coreutils from hydra it carps about a truncated bz file!! <mark_weaver>gnu/packages/patches/glibc-versioned-locpath.patch describes the current handling of GUIX_LOCPATH <wingo>/hydra.gnu.org/nar/jdk9y20lq7yll763mycv52jpq7yx342r-coreutils-8.24 <wingo>does that not always give you an invalid bzip2 file? <GChriss>h-node.org is experiencing issues, might be related <mark_weaver>wingo: you're right, it seems to be corrupted on hydra's cache. I'll see about clearing it. <wingo>i get the same size file each time fwiw <civodul>mark_weaver: do we have to remove all of nginx's cache, or can we do better? <pizzaiolo>civodul mark_weaver maybe this glibc transition is causing the kernel panics? <pizzaiolo>wingo: well, it came about due to some change in guix, that's for sure <civodul>pizzaiolo: regarding the /etc/localtime issue, i don't see how this can happen, i think something is corrupt on the way <civodul>we'd need to find out what exactly, by mounting the partition and inspecting as we discussed before <pizzaiolo>civodul: suitsmeveryfine said he updated guixSD and also had a kernel panic <civodul>pizzaiolo: could you paste your config? <pizzaiolo>civodul: suitsmeveryfine said he'd write a bug report about it <mark_weaver>pizzaiolo: please listen to civodul and do as he suggests. your problem is very different from suitsmeveryfine's <mark_weaver>pizzaiolo: the kernel panic happens when the init process quits, and it can quit for many different reasons. in your case, it's quiting because /etc/static/localtime isn't there, or is a broken symlink. <mark_weaver>and that's something we've not seen before. it's not about configs or kernel versions of locales. <mark_weaver>pizzaiolo: civodul is very good at debugging these sorts of problems, but you haven't responded to his queries with adequate information. <mark_weaver><civodul> pizzaiolo: first "ls -ld /etc/static /etc /etc/static/localtime /etc/localtime" <paroneayea>guix system: error: failed to connect to `/usr/local/var/guix/daemon-socket/socket': No such file or directory <pizzaiolo>mark_weaver: I tried asking him what I was looking for to avoid typing out the whole output <paroneayea>built with: ./configure --with-localstatedir=/var <mark_weaver>paroneayea: you need to pass --localstatedir=/var to ./configure <mark_weaver>huh, so ./configure silently accepts unknown options like that? bummer. <mark_weaver>NiAsterisk: there's no 'cc' on Guix. you should configure the package to use 'gcc'. <NiAsterisk>I think I leave this be and continue tomorrow when I'm hopefully better <mark_weaver>pizzaiolo: so I guess /etc/static/localtime was a broken symlink? <mark_weaver>actually, this will be a bit tricker, because the symlink points to something in /gnu/store but that needs to be changed to /mnt/gnu/store <mark_weaver>pizzaiolo: forget the command I wrote above, but can you tell me when you've booted into the USB installer and mounted the root partition? <mark_weaver>pizzaiolo: stat $(readlink $(readlink /mnt/etc/static)/localtime) <mark_weaver>stat /mnt$(readlink /mnt$(readlink /mnt/etc/static)/localtime) <kristofer>git master has xf86-input-libinput (version "0.8.0") while the latest version is at 1.1.5.. am I using the wrong branch or something? <mark_weaver>kristofer: we haven't updated our X11 packages in a while. lately, every time I update the X server, a bunch of drivers need to be patched to fix them to work with the newest server, and some other drivers are broken beyond my ability to repair them. <mark_weaver>(and some other distros have simply dropped those unmaintained drivers) <mark_weaver>kristofer: my last update attempt was a year ago, on the wip-xorg-server-1.17 branch <mark_weaver>the current server has been working well enough for most of us, but maybe the newer ones work better on the old MacBooks? dunno. <mark_weaver>kristofer: I'm not sure what's needed to get two-finger scroll working. more than just updating X, I suspect. <kristofer>I read that libinput can handle it beyond version 0.20.0 <mark_weaver>kristofer: would you like to try updating that package? <mark_weaver>and: readlink /mnt$(readlink /mnt/etc/static)/localtime <pizzaiolo>readlink /mnt$(readlink /mnt/etc/static)/localtime also no output <pizzaiolo>readlink: /mnt/etc/static: No such file or directory <pizzaiolo>mark_weaver: bin, boot, etc, gnu, home, lost+found, mnt, root, run, tmp, var <mark_weaver>civodul: does /mnt/etc get populated by "guix system init", or not until first boot? <civodul>mark_weaver: only at activation-time, so on the first boot <pizzaiolo>civodul: "drwxr-xr-x 2 root root 4096 feb 2 18:44 /mnt/etc//" <mark_weaver>pizzaiolo: did you type what civodul wrote, or did you add "/" to the end of it? <pizzaiolo>civodul: "drwxr-xr-x 2 root root 4096 feb 2 18:44 /mnt/etc/" <mark_weaver>hmm, that's not how "ls -ld" behaves for me. strange. <mark_weaver>for me, "ls -ld /etc" -> "[...] /etc", "ls -ld /etc/" -> "[...] /etc/", etc. <pizzaiolo>I was thinking of trying to reinstall with suitsmeveryfine's exact config.scm <pizzaiolo>mark_weaver: do you think that would do anything? <mark_weaver>civodul: is 'ls' aliased to 'ls -F' in the USB installer? <paroneayea>is there any sort of environment variable or shell command I can call to tell that I'm booted into guixsd as opposed to debian? Probably not, I guess... <mark_weaver>pizzaiolo: in the debugging you just did, were you using the Guix USB installer, or something else? <mark_weaver>it's not important, but now I'm curious where that ls -F behavior is coming from <mark_weaver>paroneayea: it's a hack, but /run/current-system will exist on a GuixSD system and not on Debian. <mark_weaver>I hope such a test will never end up in any published software. <paroneayea>mark_weaver: nah, conditionally loading something from guix on guixsd vs on debian.... <pizzaiolo>I'm waiting for civodul to maybe come up with a debugging tip, or else I'll try to reinstall <pizzaiolo>mark_weaver: cannot stat, no such file or directory <mark_weaver>it seems to be a broken symlink. maybe your timezone is somehow not a valid one. <mark_weaver>pizzaiolo: what is the output of: readlink -v localtime <pizzaiolo>mark_weaver: /gnu/store/*blablabla*-tzdata-2015g/share/zoneinfo/Americas/Brasilia <mark_weaver>that's the problem. Americas/Brasilia is not in tzdata-2015g <mark_weaver>I specifically looked at the timezone in there, because it seemed a likely problem, but obviously Europe/Paris is valid. <pizzaiolo>I wasn't sure if the _ would cause me problems <mark_weaver>pizzaiolo: America/Sao_Paulo is there. note "America" instead of "Americas" <pizzaiolo>mark_weaver: so if I reinstall shall I be ok? <mark_weaver>civodul: I guess we should check that the timezone is valid at an earlier stage <davexunit>kristofer: it's a list of 2-tuples in the form (name package) <davexunit>kristofer: have a look at any of our package recipes for examples. <mark_weaver>pizzaiolo: how is it that you posted a config that wasn't the one you used? <pizzaiolo>mark_weaver: the install might have borked along the way <suitsmeveryfine>pizzaiolo: there seems to be an error with GuixSD and not your computer <suitsmeveryfine>as I said before I got a kernel panic after simply running guix system reconfigure <mark_weaver>suitsmeveryfine: the problem was an invalid timezone that doesn't exist, and the config he posted wasn't the one he used. <mark_weaver>suitsmeveryfine: your problem is probably different. <kristofer>davexunit: are those tuples supposed to represent build deps though? libinput depends on mtdev <pizzaiolo>suitsmeveryfine: I think the install was borked the second time <pizzaiolo>suitsmeveryfine: let me try again this time and I'll say for sure <mark_weaver>suitsmeveryfine: my 'swap-devices' field pointed to a device that didn't exist, and recently there was a change that made that a fatal error leading to kernel panic, whereas before it was silently ignored. <mark_weaver>in my case, it pointed to something in /dev/disk/... which had stopped working a while ago. <pizzaiolo>is it possible to do guix pul && guix system init ? <suitsmeveryfine>or, rather, it works on my last-working version. Rollbacks are great <mark_weaver>suitsmeveryfine: I would try commenting that out or removing it. <kristofer>okay I've got my changes to compile successfully.. I'm curious now, should I rebuild xorg, or guix system reconfigure? or can I just build the input driver and restart x? <kristofer>I'm not sure if xorg needs to rebuild so the symlinks to the input modules are reset <paroneayea>guix system: error: symlink: Permission denied: "/var/guix/profiles/system-8-link" <paroneayea>mark_weaver: do you know? I'm terribly confused as to what could cause this <paroneayea>I'm so used to not having to run guix things as root! <davexunit>yeah it's like the one thing you need to be root for <mark_weaver>paroneayea: is it okay if I remove the linux-libre-4.2.5 package? <paroneayea>error: connect: /var/run/shepherd/socket: No such file or directory <paroneayea>rebooting from pre-shepherd to post-shepherd environment... <fhmgufs>Are dependencies listed in the private section of a .pc file also propagated inputs? <davexunit>ACTION can never remember the answer to this one <bavier>fhmgufs: the private section is used for static linking, so those libraries should probably also be propagated <fhmgufs>And what's the rule if a propagated input propagates another propagated input? <paroneayea>and then I can actually *use* GuixSD most of the time as my main distro :) <paroneayea>davexunit: btw, do you have a list of things left to be packaged for mediagoblin? maybe I can help this weekend. <davexunit>paroneayea: check the wip-mediagoblin branch. <davexunit>for now, I have the mediagoblin package in the (gnu packages python) module <davexunit>I just copy/pasted something out of the setup.py file <bavier>is there any way to make user-installed guix packages accessible via `ssh foo@example 'bar'`? <davexunit>since it doesn't start an interactive shell, I'm not sure. <mark_weaver>bavier: as I recall, the environment variables aren't set correctly by sshd in that case. <mark_weaver>I think it uses some hardcoded PATH that's only right for FHS <bavier>mark_weaver: yes, it only contains /bin:/usr/bin <mark_weaver>I did some hack on one of the build slaves, let me see if I can find it. <bavier>could we patch our sshd to include /run/current-system/profile/bin in PATH? <bavier>at least that way one could install the needed programs system-wide <bavier>I was wanting to ssh to my guixsh machine with mosh, but it needs to start mosh-server on the host <mark_weaver>bavier: On a Guix-on-Debian system, I set "PermitUserEnvironment yes" in /etc/ssh/sshd_config and then populated ~hydra/.ssh/environment with <variable>=<value> pairs. <mark_weaver>that latter file is not a shell script, its syntax is strictly limited to those pairs. <mark_weaver>it would be good to have a better solution for this. <mark_weaver>bavier: the thing is, PATH alone is probably not enough. we need a bunch of environment variables. <bavier>mark_weaver: yes, it would be nice if ~/.guix-profile/etc/profile could be sourced <bavier>but there might be some security concerns I'm not aware of <mark_weaver>I'm not even sure that sshd launches a shell in that case. <mark_weaver>when running ssh like that on the client side, the client side shell will have already done shell expansions and quote interpretation. doing it again on the server side would probably be bad. <mark_weaver>at some point, our 'lynx' stopped supporting https, even though it's configured and linked with gnutls. not sure when it happened. <paroneayea>I figured there would be a bundled service but I don't see one... <paroneayea>hm, I see in my logs that it's implied that pulseaudio will be started itself...! <mark_weaver>paroneayea: I don't know. I remember trying to figure out how that was supposed to work long ago, but never found out. I lost interest because my audio programs seem to just work anyway. maybe civodul knows. <a_e>I would like to take out one of the inputs of a package only on arm. Does anyone remember a package where this is already done? I am getting lost in quotes and unquotes. <paroneayea>okay, well I installed pulseaudio in my system profile <a_e>mark_weaver: Thank you! I almost got it right, but I was missing the "else" branch with the empty list and a few parentheses :-) <a_e>But indeed, the correct solution looks logical. <myglc2>... I saw a discussion on guix-devel & hoped it was sorted, <myglc2>... on guixSD I did 'guix pull ; guix system reconfigure' <myglc2>... to make a new vanilla user, glc4, then <myglc2>... ssh glc4 ; emacs ; 'M-x guix-' which produces ... <myglc2>... '/home/glc3/.guix-profile/share/emacs/site-lisp' does not exist.' <myglc2>Oops, sorry, that shoudl read '/home/glc4/.guix-profile/share/emacs/site-lisp' does not exist.' <civodul>paroneayea: pulseaudio is started automagically <civodul>basically, programs that use the client library launch it via dbus if it's not already there <civodul>ooh, reading the backlog, i realized we had completely forgotten to gracefully handle the shepherd transition regarding "reboot" <kristofer>I updated the xf86-input-libinput package! it builds just fine. How can I update xorg-server to use the new definition for libinput instead of the existing? <kristofer>I have tried to guix package -u xorg-server xf86-input-libinput <davexunit>does anyone here use any free (of course) CAD programs? <davexunit>I have something I'm designing that I'd like to use a CAD program to assist with. <civodul>ACTION had been playing for some time with PID 1 and was surprised he had not crashed/rebooted the thing yet <suitsmeveryfine>kristofer: is the new libinput package going to fix any touchpad-related bugs? <kristofer>suitsmeveryfine: I hope so, I would like to test it.. <fhmgufs>But if I build it it says 'In procedure module-lookup: Unbound variable: %build-inputs'. <kristofer>suitsmeveryfine: I need the xorg-server package to update the symlink to the updated libinput <suitsmeveryfine>either I get the generic touchpad driver in linux-libre or synaptics gets loaded also and messes it up completely <bavier>fhmgufs: no need to unquote the string-append <kristofer>davexunit: I don't use it personally, but librecad may do what you need? <fhmgufs>bavier: Hmm, no. But I did exactly the same as in libgweather. <bavier>fhmgufs: what you pasted is not the same, the libgweather recipe has an additional level of quoting <fhmgufs>Oh, yes. I didn't see that, sorry. Now it's working. <davexunit>kristofer: I should see what it takes to package one of those. thanks. <bavier>davexunit: guix has a librecad package already <pizzaiolo>the trackpad doesn't work properly but hey, one victory at a time :) <petter>pizzaiolo: good you got it working! I've been away for a while now, were there issues you bested that should go in the installation draft? <pizzaiolo>petter: you should probably mention the list of possible locales :P that was my issue, basically <suitsmeveryfine>petter: one thing you need to change: it's impossible to ping gnu.org; change it to www.gnu.org <petter>ok. I made a note of this earlier <petter>since apparently the locale command isn't available, i'll change to this instead: ls /run/current-system/locale/X.Y/ <pizzaiolo>mark_weaver gave me a great way to set up wifi <pizzaiolo>wpa_supplicant -c wpa_supplicant.conf -i wlp2s0 -B <petter>i have mark's setup and kristofer's, also one for wifi without passphrase <suitsmeveryfine>petter: "Make a filesystem, f.ex. ext4." -> "Make a filesystem, e.g. ext4." <pizzaiolo>petter: mention the kernel modules hid-generic, hid-apple, serpent_generic and wp512 <pizzaiolo>petter: also, explain how to creat a swap file <petter>or if we should have one disabled (commented out) in a example config <petter>what do you guys think of making a simple installation, default timezone and locale etc. just to get a booting system. Then have a post-installation document for further setup? <suitsmeveryfine>I think the user should be able to set up his/her system properly from start <suitsmeveryfine>but it can be mentioned in the guide that locales and stuff always can be changed later <civodul>mark_weaver: BTW, what's the status of security-updates? <suitsmeveryfine>petter: "If you're going to change locale you should check what is available and exactly how it is typed; close the editor or change virtual console (Ctrl-Alt-F#), and run the command "locale -a". This command doesn't work for me. <petter>suitsmeveryfine: right, and as a wrote a few minutes ago i'll change this <petter>further assuming: ls /run/current-system/locale/X.Y/ will work during installation <mark_weaver>civodul: about 100 builds remaining on x86_64, same on i686, about 650 remaining on armhf, and over 1000 on mips. <petter>pizzaiolo: do you think the <your-root-partition> bits could be done in a clearer way? <pizzaiolo>petter: yes, I got that wrong in the first try <pizzaiolo>I used /dev/sda1 when I was supposed to use /dev/mapper/guixsd <petter>i'm thinking they should be in italics at least, or another color, when we do a formatted version. But maybe there're better ways <petter>pizzaiolo: the tricky thing here is balancing this between users doing encrypted installation and those doing unencrypted <a_e>mark_weaver: Oh dear! And there was not only the week on security-updates, but also at least two before on core-updates. <petter>for the user it would be nice with a dedicated installation manual for whatever partitions they want encrypted (if any). But this will be harder to maintain <suitsmeveryfine>petter: something really important to add to the guide is to explain the GRUB install error <mark_weaver>suitsmeveryfine, petter: no, we should just fix the problem. <petter>suitsmeveryfine: right. I'll add this <a_e>We have a separate nettle-2 package, with a comment that this is needed for lsh. <suitsmeveryfine>mark_weaver: OK, but you seem to be very busy with security matters :) <mark_weaver>if someone wants to mention it on a paste site, that's fine, but I'm talking about what will go in the Guix manual. <petter>also, will the crypto modules be added as default at some point? <suitsmeveryfine>then I suggest that we prepare a guide for libreboot.org to begin with <civodul>mark_weaver, pizzaiolo: commit 37dd1e6 will avoid the timezone-induced kernel panic in the future <mark_weaver>I don't want to propagate docs that say "ignore the grub install error", and "make sure not to run the garbage collector", etc. it's terrible. <civodul>mark_weaver: could you push the usb-hid/hid-apple patch? <petter>there should be no need for explaining how to add modules to initrd, right? <mark_weaver>petter: it's already in the manual, and we'll add the modules needed for this. <mark_weaver>ACTION just pushed the hid-generic and hid-apple modules to the default list. <petter>mark_weaver: i'm thinking of for the new step by step draft <paroneayea>mark_weaver: that's one downgrade from the other kernel we had <davexunit>paroneayea: you know what, I may also be without sound. <davexunit>I need to verify, but I had some surprising things happen in that regard recently. <paroneayea>mark_weaver: now mplayer is working on this reboot (it didn't seem to be last reboot <civodul>mark_weaver: ok for the crypto modules, i think i replied to both patches, no? <paroneayea>and does not launch pulseaudio automatically, if it should, and if I run it, I get <paroneayea>E: [pulseaudio] module-alsa-card.c: Failed to find a working profile. <paroneayea>E: [pulseaudio] module.c: Failed to load module "module-alsa-card" (argument: "device_id="29" name="platform-thinkpad_acpi" card_name="alsa_card.platform-thinkpad_acpi" namereg_fail=false tsched=yes fixed_latency_range=no ignore_dB=no deferred_volume=yes use_ucm=yes card_properties="module-udev-detect.discovered=1""): initialization failed. <mark_weaver>petter, suitsmeveryfine, pizzaiolo: I just added the needed modules to the default set. after a "guix pull" you'll have them. but there's no harm in leaving them in your OS config for a while. <pizzaiolo>mark_weaver: if I ever need to remove it, please do tell me <pizzaiolo>petter: are you also using a librebooted macbook? <petter>pizzaiolo: no, i have a Libreboot X200 <petter>so you and suitsmeveryfine had some problems i didn't have <petter>pizzaiolo: also, i used a basic cryptsetup command which worked out of the box. For the installation draft i borrowed the one used in libreboot.org for the encrypted parabola installation <suitsmeveryfine>petter: I used the parabola one and it worked fine with the added modules <petter>suitsmeveryfine: right, just saying you faced problems i didn't <suitsmeveryfine>the parabola guide contains instructions for wiping the disk. I wasn't able to do that in the the GuixSD installer <suitsmeveryfine>the post install instructions should include details for setting up swap <petter>suitsmeveryfine: you're using an unencrypted swap partition? <suitsmeveryfine>but with the latest `guix system reconfigure` I had to remove swap entirely because it was broken <mark_weaver>suitsmeveryfine: did removing the swap fix the problem? <mark_weaver>I've never tried using a swap file (as opposed to partition). <mark_weaver>suitsmeveryfine: did you check to make sure it actually added the swap? <mark_weaver>for me, it was silently failing, and that recently changed to a fatal error <mark_weaver>but we don't yet support hibernation. another item on the TODO list :) <suitsmeveryfine>when I ran "top" it said that the swap was empty but a lot had been cached for it <suitsmeveryfine>mark_weaver: oh, I see; I will try the old OS version with swap and see if it was actually enabled <mark_weaver>suitsmeveryfine: sure, I'd be curious. I suspect it never worked. <suitsmeveryfine>I filed a bug report regarding this and tried to send additional details about the remove-swap fix <suitsmeveryfine>that wasn't published though and Ludo already took the time to test my config in a vm <paroneayea>guix-ui-read-profile: Symbol's value as variable is void: guix-current-profile <paroneayea>why is this happening to me in guixsd but not debian <suitsmeveryfine>mark_weaver: you were right -- swap wasn't enabled in the first place <mark_weaver>paroneayea: that's a questoin for alezost, I think. civodul might also know. <fhmgufs>suitsmeveryfine: Not really, except from the name change. <davexunit>the name was changed to a conflict with the "digital mars d" compiler project. the developer of this software contacted ludo and requested that he change the name. <mark_weaver>I think there were some not-entirely-trivial changes, as I recall. <pizzaiolo>suitsmeveryfine: the shepherd shepherds the daemons <suitsmeveryfine>yes of course, but it sounds a little bit as if though s/he was in control over the whole flock <fhmgufs>What to do if a package wants to install its gir data into the gir directory in the store? <mark_weaver>sneek: later tell fhmgufs: if a package wants to install its gir data into an existing store directory, you need to modify it to install the gir data in its own package output directory. see the grilo and libgee packages in gnome.scm for examples, but the fix might be somewhat different in your case.