<rekado_>and here’s one by jackhill on the bubblewrap sandbox being incompleet <rekado_>but fonts not showing up in gnome-font-viewer is suspicious already <rekado_>have you tried strace to see what it’s looking at? <KE0VVT>I'm still having issues with my optical drive. <KE0VVT>Even after adding myself to the "cdrom" group. <leoprikler>for the record, I do have fonts working in gedit <leoprikler>they also show up correctly in gnome-character{s,-map} <leoprikler>I think gnome-font-viewer is borked, it never displayed anything for me anyway <leoprikler>Okay, I think it's really related to the Webkit sandbox <leoprikler>Devhelp shows similar behaviour in that it doesn't apply fonts correctly and works for even less fonts <leoprikler>sooo, should we patch this into webkitgtk or what do we do? <calher>I just built and installed libdvdcss on Trisquel. What do I need to do the same on Guix? <cbaines>calher, I'm not quite sure what you're asking, assuming you used the Debian packages on Trisquel, I wouldn't try doing similarly when running Guix as a system. Guix has a libdvdcss package you might want to consider using. <calher>cbaines: Does the package handbrake not pull it in? <calher>cbaines: Handbrake can't see this DVD I put in. <cbaines>calher, the Guix handbrake package does appear to require libdvdcss. Are there any messages you can see that might provide more details? <jackhill>We worked around that for pulseaudio by not using a symlink for the client config. Someone will have to figure out exactly what's going on :) <jackhill>poking at it more has been on my list, but I haven't had a chance yet. <wxie>I have a problem in bootstraping the recent guix. <wxie>configure.ac:249: warning: PKG_PROG_PKG_CONFIG is m4_require'd but not m4_defun'd <wxie>m4/guix.m4:328: GUIX_LIBZ_LIBDIR is expanded from... <wxie>configure.ac:249: the top level <wxie>configure.ac:260: warning: PKG_PROG_PKG_CONFIG is m4_require'd but not m4_defun'd <wxie>m4/guix.m4:339: GUIX_LIBLZ_FILE_NAME is expanded from... <wxie>configure.ac:260: the top level <wxie>configure.ac:90: error: possibly undefined macro: PKG_CHECK_MODULES <wxie> If this token and others are legitimate, please use m4_pattern_allow. <wxie> See the Autoconf documentation. <wxie>autoreconf: /gnu/store/yk3zw1pqbs5wlf1lapf25986yjd0g736-autoconf-2.69/bin/autoconf failed with exit status: 1 <wxie>Where is PKG_CHECK_MODULES? <jackhill>wxie: hmm, sounds strange. What commit of guix are you trying to bootstrap? I was just able to sucessfully bootstrap ec944bdde93c73724f3f3fec2ecd0347a964504c <leoprikler>The thing is, we can't just have "no symlinks" when it comes to fonts. User fonts are taken from ~/.guix-profile, which is already a symlink. <wxie>jackhill: origin/master ec944bdde93c73724f3f3fec2ecd0347a964504c <wxie>leoprikler: Do you have a solution? <mbakke>leoprikler: I think applications look for fonts in /var/cache/fontconfig and/or ~/.cache/fontconfig though, which are absolute, but may not be accessible in the sandbox <mbakke>wxie: did you run 'guix environment guix' first? <wxie>mbakke: No, I did not do that. Just './bootstrap' in bash. <leoprikler>hmm, /var/cache/fontconfig is empty in guix, but I'll try with ~/.cache/fontconfig <leoprikler>although looking at it it doesn't really appear as if that cache contains a copy of all the fonts <wxie>within [env], './bootstrap' is ok, but './confiure' has error. <wxie>checking the current installation's localstatedir... /var <wxie>configure: error: chosen localstatedir '/usr/local/var' does not match that of the existing installation '/var' <wxie>Installing may corrupt /gnu/store! <wxie>Use './configure --localstatedir=/var'. <xelxebar>efraim: Thanks for merging my first patch! <xelxebar>Highly recommend them if you're just getting started. <kamil_>Is the Nix package broken on Guix System? <kamil_>I'm having a several issues with it; I'm experimenting in VM to see if it is possible to use home-manager one way or another. <kamil_>It could be so because I've installed it for my user from CLI <kamil_>leoprikler: It's in very early stage of development. As for Nix on Guix System, I think I've found a bug. No derivation can be built due to "/gnu/store/*-bash-minimal/bin/bash no such file or directory" error <kamil_>even though the binary is in place <ptibedo>I created a recipe widelands and it's a success <ptibedo>I created a recipe widelands and it a success <buffergn0me>Is anyone running Xorg on 1.1.0 using the xinit package (without a session manager?) <buffergn0me>I don't think that package has been working for a while <buffergn0me>As-is, xinit in that package has the wrong path to X hard-coded <buffergn0me>Even tweaking a startx script, I run into the problem of X having a hard-coded module path, unless it is running as root <buffergn0me>Which is a problem, because the video card drivers live in modules in packages like xf86-video-ati <buffergn0me>I looked into how Debian works around this, and they have a suid wrapper executable (/usr/lib/xorg/Xorg.wrap) that will run the server as root if KMS is not available <buffergn0me>My Xorg does not even get that far because /dev/dri/card0 or /dev/dri/anything is not there <buffergn0me>Which is to say, it sounds like a lot of work to get the xinit package working <buffergn0me>What is the way to nominate a package to be deleted? <apteryx>buffergn0me: what do you mean by "nominate a package to be deleted" ? <buffergn0me>I think it is ok to make everyone run a session manager. OpenBSD did that last year <buffergn0me>It does not work, it will take a lot of work to make it work, and when it works it is just a big setuid security hole <apteryx>I think it should be fixed, unless it's unmaintained and buggy (which is not the case AFAIK). <apteryx>I meant buggy upstream. That's its broken in Guix is a Guix specific problem <pkill9>all the xinit package needs is a working path to X to execute <buffergn0me>...and then all the other problems I described happen <pkill9>yesss I unloaded the i915 kernel module <apteryx>buffergn0me: don't display manager need the same setuid to work? what do they do differently? <srandom> is why OpenBSD decided to force everyone to use display managers <srandom>If give up xinit , I want to use a login manager like xdm <buffergn0me>apteryx: They start X as a daemon basically. That way there is no opportunity to pass command line arguments to it, which is how that exploit worked <jackhill>leoprikler: I just read #41174. Thanks for investigating. <buffergn0me>srandom: I am looking at the manual now, and it looks like for 1.1.0 you get a choice of GDM, SLiM, and SDDM <srandom>There are many issue in gdm. I only have one computer successfully started it (using guixsd). If I can, I still hope to use xdm. After all, it is more traditional, and I am familiar with it <apteryx>slim works well, although it hasn't seen a commit in years. I'm not sure how safer that can be assumed to be compared to xinit <xelxebar>pushcx: Did you do a guix pull between the install and uninstall? Counterintuitively, uninstalls can actually install stuff. <xelxebar>The reason being that uninstallation ends up creating a brand new profile, one generation after your current one. This is different than rolling back to the generation before installation. <apteryx>pushcx: I think grafts might be at cause <apteryx>try 'guix remove hello --no-grafts' and to check <srandom>Is there any way to calculate the size of the system built by config.scm? <srandom>I am building a chromebook system on my guixsd server and I am worried that 16G emmc cannot store it. <buffergn0me>Anyone else getting 'localhost' for hostname -s in 1.1.0? <calher>cbaines: I already had handbrake installed, but it did not behave as if libdvdcss was already installed, when I installed it. <Starfish>guile: warning: failed to install locale <Starfish>hint: Consider installing the `glibc-utf8-locales' or `glibc-locales' package and defining `GUIX_LOCPATH', along these lines: <Starfish> export GUIX_LOCPATH="$HOME/.guix-profile/lib/locale" <Starfish>See the "Application Setup" section in the manual, for more info. <srandom>you need 'guix system reconfigure /etc/config.scm' <srandom>I guess this error is due to the global glibc-locales not being upgraded <Starfish>So I installed Guix in a foreign distro and read the manual. According to the manual I installed glibc-utf8-locales and exported the GUIX_LOCPATH, but I still get the errors. <jackhill>Starfish: on a foreign distro there are two environments to be concerned with. It sound like you took care of the environment of your user where you run the guix command, but guix-daemon may not have access to the locales. You should also install a locale package in root's profile (since on a foreign distro that's the environment that the daemon gets run in). <jackhill>you're welcome. Hopefully I am right/it helps :) <jackhill>I'm afraid it's bedtime for me though. Goodnight! <Starfish>I installed the locale package, added the path for root user, and restarted the daemon. Unfortunate it still doesn't work. I get the same error or hint that says I need to install locale and set the path and it's for both me and root user. Consequently (I guess) when I launch the guix version of icecat, I can't see any fonts. ***dingenskirchen1 is now known as dingenskirchen
<xelxebar>Starfish: I've seen that happen when your user glibc-locales are out of sync with the one in the daemon's profile. *xelxebar rains a hellfire of M&Ms down upon calher ***apteryx is now known as Guest80643
***apteryx_ is now known as apteryx
<sneek>lprndn, you have 2 messages! <sneek>lprndn, ryanprior says: we should put together a system definition that contains all the pantheon stuff so we can start testing it in a VM and work out the configuration. <civodul>efraim: thanks for the Shepherd blog post BTW! <efraim>civodul: :) I think I'm going to have to have a follow-up with some other cases, like using mcron <dissoc3>i installed package hackrf and it has user space library. where are those located in guix? <dissoc3>i was looking for hackrf and not libhackrf. sorry very tired. deep in rabbit hole <PurpleSym>Is it just me, or is the package checksum of boost-signals2 wrong? I get 13i5j43nggb46i6qpaf7gk53i7zp7pimphl7sydyfqz2m9yx5cdy <nixo_>PurpleSym: I get 13i5j43nggb46i6qpaf7gk53i7zp7pimphl7sydyfqz2m9yx5cdy too <janneke>efraim: about the user shepherd shepherd services, how does the bootstrap work, i.e., how/when/by whom is your user shepherd started? <rekado>efraim: would it be enough to wrap the (guix-environment …) call in (zero? …) to satisfy shepherd? <janneke>efraim: i love the post (just like your earlier mail -- but better)...but how do i "start" ;-) <nixo_>PurpleSym: They changed the tag <nixo_>It was added in february 2019, and the tag referred to fe476cb0a1a57d570f305ec959affedef48caac4 probably <efraim>janneke: I start my 'shepherd' by hand, after login I run 'shepherd' in a terminal <efraim>rekado: I haven't tried wrapping it in zero, i'll try it out <PurpleSym>nixo_: Urgh. Not sure I even need it though. Previously my rstudio-server package worked fine with just boost and now with 1.72 it complains it can’t find boost_signals :/ <efraim>rekado: herd: exception caught while executing 'start' on service 'fccache': <efraim>Throw to key `quit' with args `(0)'. <efraim>I'm not sure it ran. it probably did and the zero? ate the output and then raised an exception <nckx>Power failure @ my house, Overdrives are down for the day. <efraim>ludo and i came up with a more complex version with forking but it was bit much for the blog post <janneke>efraim: right; i'll start playing with that, thanks <janneke>i wondered if i missed an obvious way to have users services started upon boot <efraim>I don't know a lot about xsessions, I think some others are using some sort of 'unless [ -S ~/.config/shepherd/sheperd.sock ] shepherd -s ~/.config/shepherd/shepherd.sock;' <kmicu>[Jokin’] Thanks to Guix mailing lists I know How to get off of Ruben Safir's spam list. *kmicu lols at Ludo being in CC too. <civodul>nckx: thanks for letting me know, i've commented them out in /etc/guix/machines.scm on berlin <civodul>i hope you can still see the light nonetheless :-) <kmicu>Probably the message is from yesterday (on gmane). <kmicu>Oh, finally removing hydra.gnu.org. Sweet. <nckx>civodul: Thanks. I'll add them back tonight. <nckx>True enlightenment requires wi-fi, we'll see. *efraim knows enlightenment has wifi support ;p <nckx>Spake too soon. Now it is. 502. <zimoun>nckx:are you using the anonymous clone or the member clone? <nckx>zimoun: The exact link above. <zimoun>nckx: now you have 502 too, right? <nckx>zimoun: SSH doesn't seem to work either. ‘Cloning into 'guix'…’ but no network traffic. <zimoun>nckx: thank you for mastodon info. :-) <nckx>zimoun: Yeah… I submitted a (very) simple git mirror patch for berlin before The Event. Need to pick up that ball. <nckx>It just called ‘git’ every n minutes and that wasn't good enough, had to use Scheme 😛 I think the problem was that the Scheme bindings don't support mirroring. <nckx>zimoun: After a minute or two of doing nothing the SSH clone starts transferring data. So if you have an account, give it a chance. <nckx>That doesn't fix guix pull of course. Where'd I save my shouty template... <zimoun>nckx: I do not have an account on Savannah if it is what you mean. <nckx>zimoun: …now HTTP works for me, so this isn't an off-line VM issue. *nckx puts the ⚠️s away for now. <civodul>kmicu: on guix-devel? i think i don't see it <zimoun>nckx: Well, back what we discussed on December, and because these kind of situations are annoying for the end-user as me, why not implement an "official" mirror and fallback to it when savannah is down? <kmicu>civodul: info-guix iirc, let me pull fresh Guix lists from gmane.io. <nckx>civodul: But not info-guix@gnu.org, or it didn't make it through at least. <efraim>ugh I feel like I should wrap the enlightenment start binary with dbus <nckx>zimoun: That's what I meant. <nckx>It's on berlin. It's probably still there. It's just not wired up to fetch updates. *nckx .oO how bad was this power failure. <civodul>wigust: should be back up now, not sure what happened <nckx>I should go, not to overstay my ‘hi mind if I come over to borrow some wi-fi’ welcome. <nckx>If only coffee shops still existed. <kmicu>civodul: I fetched the fresh news from news.gmane.io vir slrn to not break my Gnus setup. Somone added gmane.comp.gnu.guix.announce mailing list (mapped to info-guix iiuc) and the message is still there. <kmicu>(That’s the only message on that news group.) <kmicu>So I assume that’s a fake mapping to a rouge mailing list and you should contact Lars/Gmane.io to remove it. <civodul>i wonder how it can reach Gmane if it doesn't reach the underlying mailing list <mothacehe>talking about Gmane, I requested the addition of guix-patches multiple times, but it still doesn't show up. Anyone knows why? <kmicu>No problem. Gnus shows me automatically new mailing lists added to a news server so I’ve added gmane.comp.gnu.guix.announce without any hesitation. A lesson for me to assign very low trust for emails not signed with pgp and to monitor Guix related infosphere for malicious actors. <mothacehe>yes twice, but I received a 'request discarded: list exists' <janneke>mothacehe: thanks for noticing/chiming in on my vm-image patch, i meant to cc you... <mothacehe>janneke: No worries! About the closure registration, it made me realize than the database I'm creating with "register-closures" in (gnu build image) could not be cross-platform. <mothacehe>kmicu: ok! For now I'm using "(nntp "news.yhetil.org")" to get guix-patches. <janneke>mothacehe: ah...well, civodul pointed out that we should probaly use `store-database' from guix/scripts/pack.scm for [vm-]image <kmicu>mothacehe: thanks for that info! <janneke>as a (what i thought to be easier) step in between, i tried to remove it altogether; now i realise most probably won't do <mothacehe>kmicu: yw! I just switched to a GNUS + NNTP based mail workflow. Still struggling :p <kmicu>Oh, that’s a personal site. There’s no info whether using it is ok. <mothacehe>kmicu: I think I found an email introducing this NNTP server to guix-devel. <kmicu>I’m still struggling after many years so I plan to migrate away from Gnus and NNTP 😺 <mothacehe>janneke: Oh, this "store-database" in (guix scripts pack) looks very much like what I do in (gnu system image). <mothacehe>janneke: but is a sql database cross-platform, or architecture dependant? <mothacehe>kmicu: haha :) I'm also trying to get gnus-cloud to synchronize GNUS state between machines, but it seems to be broken since always! <rekado>wigust: I don’t get a 504 on ci.guix.gnu.org <rekado>and I just reconfigured it to add Madalin to the sysadmins <rekado>we have confirmed that the public network interface has only negotiated 1G even though 10G should be available <rekado>so tomorrow we’ll meet at the data centre and see if there’s a wrong cable connected to the patch panel <rekado>there might be a short (~10 min) network outage <jonsger>so the day after tomorrow will hopefully be 10G :P <janneke>mothacehe: good question...i would surely hope so... <rekado>jonsger: perhaps, but there’s probably another bottleneck that keeps us from enjoying all the bandwidth <janneke>mothacehe: (previous comment) oh, great -- then we can perhaps combine some things <janneke>mothacehe: yes, sqlite is cross-platform <kmicu>mothacehe: the thing keeping me on Gnus is adaptive scoring. Gnus is like ‘oh, I see you are not interested in Guix patches related to that subsystem, ok I will try to hide them’. So useful. <blobfish>why is guix using outdated versions of linux and gnome? <kmicu>Hi blobfish what versions did you ? <rekado>that appears to be the latest stable version <rekado>what version of guix are you using? <mothacehe>janneke: I should maybe directly ask #grub about that. <rekado>efraim: I suggested zero? because the blog post says that guix-environment returns the status code instead of a boolean <efraim>raghavgururajan: just saw your question. Yeah, I start shepherd by hand <raghavgururajan>efraim, In my system, I get error when I manually start shepherd in terminal. <efraim>rekado: I think when I ran it from a guile repl I got back ("31234" . "0") or something after playing with it <efraim>numbers, not strings $3 = (6206 . 0) <efraim>from the repl I like (match (primitive-fork) (0 (guix-environment "--ad-hoc" "fontconfig" "--" "fc-cache" "-frv")) (pid (waitpid pid))) <efraim>raghavgururajan: perhaps a mismatch of guile versions between system and user libraries and shepherd? <efraim>do you have anything in your init.scm? <hksa94>I have problems with the rendering of symbolic icons and the Gnome Web icon in Gnome 3.34.x <janneke>...apart from the ugly/complicated bits... <raghavgururajan>efraim Shoot! I do not have my system with atm. The error had something like socket <efraim>I had an error another time where sheperd tried to create a socket in a directory which didn't exist yet (since the user hadn't actually logged in yet) so I changed the command to 'shepherd -s ~/.config/shepherd/shepherd.sock' to create the socket in ~/.config ***Xaa is now known as Guest84863
<Guest84863>Hello there ! I'm trying to get started with Guix and i would like to know if LVM on Luks is supported ? <raghavgururajan>efraim, found the error from irc log. It is /run/user/1000/shepherd/socket: No such file or directory <raghavgururajan>efraim, error: connect: /run/user/1000/shepherd/socket: No such file or directory. lfam was telling me something about XDG dir. <boeg>I want to install guix but my device are not fully supported - the wifi is missing and I dont have a dongle and no ethernet port. Can I somehow get a nonfree guix install iso? <wxie>boeg: No, there is no nonfree guix. You can get a dongle easily from RYF hardware site. <jonsger>lurkki__: does tethering work in the installer? <lurkki__>another option is to generate your own installer image with the wifi crap in it <rekado>Guest84863: LVM works but booting from LVM devices does not AFAIK <Guest84863>@rekado alright tanks. The documentation seems to say i do not need to use LVM and i could create a fully encrypted drive with multiple partitions but i'm not sure if that works <kamil_>Anyone, do you have an idea why does calling the bash binary fails with the error message "while setting up the build enviroment: executing '/gnu/store/*-bash-minimal-5.0.16/bin/bash': No such file or directory" when trying to build a Nix derivation? I've configured Nix as per the example in the Guix manual. <rekado>kamil_: does that file reference a loader at a location you don’t have? <kamil_>rekado, the path is valid and the requested file exists there <mbakke>kamil_: try running $(guix build glibc)/bin/ldd on that file, are there any missing references? <rekado>civodul: I don’t know. I only know about the two clusters at the Berlin Institute of Health and the MDC <kamil_>mbakke, "-bash: /gnu/store/*-glibc-2.31-debug: Is a directory" <rekado>that trick won’t work because guix build glibc returns two values <mbakke>kamil_: oh, whoops... use /gnu/store/1y7g7kj3zxg2p90g692wybqh9b6gv7q2-glibc-2.31/bin/ldd <mbakke>would be neat if we could use guix build glibc:out, but that's not currently supported <civodul>i started working on it some time ago but there were annoyances <kamil_>mbakke, linux-vdso.so.1 (0x00007ffd629e1000), whilst 3 other *.so files point to something like this "=>". <mbakke>only three? is there a ld-linux-x86-64.so reference? <mbakke>kamil_: can you execute that bash outside of the nix store? not sure if the Nix service should be using anything from /gnu/store at all <kamil_>I wasn't clear enough, sorry. There are 4 references, only one--linux-vdso.so.1, is missing a reference; whilst the other 3--libdl.so.2, libc.so.6 and /gnu/store/*-glibc2.31/lib/ld-linux-x86-64.so.2--don't have missing references. <kamil_>mbakke, executing this bash file prints our an error: "bash: shopt: progcomp: invalid shell option name" and drops me into a new bash shell <wigust>mbakke: well, when I installed there was no issue like this :b <kamil_>mbakke, it should be noted that pressing arrow keys only prints symbols like this: ^[[C <mbakke>wigust: perhaps the recent Nix update broke it? have you tried using it since 13c18af1d835db11f3f4cf34e65c7da706625a3f ? <mbakke>kamil_: that's expected, bash-minimal does not include readline or many other things that your bashrc might expect <wigust>mbakke: no, I installed it on couple machines before the update and then I switched to nix-daemon installed by Nix itself <mbakke>kamil_: can you file a bug report? <kamil_>mbakke, I would love to, but I still can't. I really wish there was a gitlab bug tracker for the less adventurous <mbakke>kamil_: most bug trackers require an email address too, no? :) <kamil_>mbakke, they do, but I don't have to reveal one to other participants ^^' <mbakke>kamil_: you could sign up for a secondary email account somewhere, or use a mail alias service like @member.fsf.org (requires membership though) :) <kamil_>Oh. I don't think I'll ever deserve a membership at FSF. I play commercial video games, albeit tightly sandboxed <rekado>the FSF is not a club of those who never use proprietary software. <rekado>it’s an organisation campaigning for your right to never *have* to use proprietary software <kamil_>I.. I've got the impression that most people at FSF will go out of their ways to not use it at all at all times <rekado>sure, but when you sign up there’s no questionnaire to assess your “purity” <kamil_>I advocate free and open source software, and despise anything that in GNU's terminology doesn't user freedom like some Apache license programs..'t <kamil_>but I still make a very few exceptions to stay relevant in the modern society <rekado>the Apache license is a free software license that protects user freedom. It may not be as effective at that compared to other licenses, but there’s no reason to despise it. <rekado>(playing games allows one to stay relevant in society?) <kamil_>Well. The apache license forced me to throw away my work because I wasn't allowed to package a community-driven Minecraft laucnher. <rekado>alextee[m]: did you sort out the font problems with the fakechroot backend? <rekado>if there’s a bug it would be good to find it before publication of civodul’s blog post <kamil_>One thing I've ever packaged my life, and it has to be thrown out of window. Even though I packaged it so that it respects the license, I was still afraid that the creator of the program would pursue a legal action against me. He's a hatter of everuthing that isn't a tar archive. <civodul>rekado, alextee[m]: we can always fix bugs afterwards too, but i'm definitely interested in bug reports! <civodul>is it me or Savannah is almost dead? <rekado>civodul: people on #guile also noticed Savannah weirdness <rekado>alextee[m]: could you describe your problem again and include strace output? *kamil_ goes AFK and offline <alextee[m]>rekado: I built a pack with `guix pack -RR -S /opt/zrythm/bin=bin -L . zrythm-next` (zrythm-next is my channel's version of zrythm), then i copied the tarball to my archlinux VM to test (i also asked other people to test on debian) and running `opt/zrythm/bin/zrythm_launch` loads the app but all the text is shown as boxes <alextee[m]>will try without -RR now just to see if the fakeroot is the problem or not <civodul>alextee[m]: you have to explicitly set GUIX_EXECUTION_ENGINE=fakechroot if you want it <civodul>if your Arch machine supports unprivileged user namespaces, they get used by default <civodul>the thing about fonts probably has nothing to do with "guix pack" itself <civodul>prolly a fontconfig issue, as always <alextee[m]>civodul: GUIX_EXECUTION_ENGINE you mean when packing or when running from another distro? <alextee[m]>i deleted some irrelevant logs from the app since it wouldnt let me paste the whole thing (also debianpaste thinks im a spammer now \o/) <civodul>alextee[m]: set GUIX_EXECUTION_ENGINE when running, see the manual *civodul looks at the log <civodul>so it's using the "userns" engine here <civodul>so yeah, i suppose you'd have to run "fc-cache -fv" or something :-/ <civodul>or use "strace -f" to see which fonts it's looking for <alextee[m]>i did try fc-cache with the gnu directory but it still didnt work. ok let me try <civodul>alextee[m]: thanks! i don't see any reference to ".ttf" there, weird <boeg>where are channels defined? I can see i can add them to ~/.config/guix/channels.scm, but on a new installation medium, how do I get the installer to pull from a custom channel? I've tried adding a channels.scm to /etc/ and /mnt/etc/ but doesnt seem to work <alextee[m]>fc-cache -fv does list some picked up fonts btw, but fc-list doesn't list them <sirmacik>while being on install image where should I put additional channels to get packages from there into a system I intend to install? <rekado>boeg: the installer doesn’t pull at all <boeg>rekado: no, but if i do a manual guix pull afterwards <boeg>but it doesnt pull from the channel i added <rekado>you can arrange for a global channels file to be installed <rekado>I think recent versions do check for /etc/guix/channels.scm first *raghavgururajan started to try emacs. Concerned more with whyness rather than howness. <johnmod>How can I install GuixSD on a Beaglebone Black? Are there any pre-built installer images for armv7? <mothacehe>johnmod: I'm working a providing pre-built installer images for a whole range of armv7/aarch64 boards. ***familia_ is now known as familia
<ryanprior>How can I refer in a manifest to a package that I installed from a file on my system? <ryanprior>It's telling me "unknown package" but I have the package installed on my system, it's just not in my guix master yet. Do I have to add a channel in order to be able to use it in a manifest? <rekado>ryanprior: the module defining it must be on the load path <rekado>you can temporarily augment the load path with -L <ryanprior>I think I got it working, but I won't find out for a while, because me running `guix environment` with a few packages made Guix decide to rebuild TeX live, ruby, emacs, and a bunch of other stuff first. <rekado>it should not just randomly build stuff <ryanprior>no but I added something to my load path, which it maybe interpreted as a rebuild? <ryanprior>I had been running `guix environment` repeatedly (but without the load path added) and it had been humming along, then once I got my `-L path/to/mypackages` in there it decided to go ham. <ryanprior>gonna go make some coffee, thanks for getting me on the right path rekado :) <ryanprior>rekado: now that it's finished I can say yes -L worked! Also I looked through my history to verify what I'd done <ryanprior>guix environment -m test-deps.scm --ad-hoc sed <ryanprior>(this dropped me into a shell within a second) <ryanprior>guix environment -L /home/ryan/dev/guix-packages --pure -m test-deps.scm --ad-hoc sed <ryanprior>test-deps.scm changed slightly- I added two packages from my local repo. they don't rely on texlive or any of that stuff <gisfa1>"guix pull: error: Git error: unexpected http status code: 502" <PotentialUser-14>What is the default directory of guix initial ramdisk and linux image to show the to grub <ryanprior>PotentialUser-14: I don't any idea, not an unwelcome question though! Hope somebody knows :x <ryanprior>I'm working on a package for the v programming language compiler, for which the binary is called v. Is "v" a ridiculous package/too-short package name even if it's correct? Should I call it "vlang" or "v-compiler" or something? <boeg>Where are channels stored in the system? <ryanprior>Can I define a package with multiple sources? The build system is in one repo while the bootstrap source is in another repo. <ryanprior>I could also make a separate package for the bootstrap source and use that as a native-input but it seems more succinct to have multiple sources if that's possible? <roptat>ryanprior, you'd have one main source, and the other source could be put in inputs (you can put an origin record as an input) <roptat>you'll have to add a phase to decompress or copy the source where you want <roptat>for an example, see cobol.scm which has an origin input for "newcob" <keegans>hey, just getting started with guix : i see that i can add desktop services to automatically start sddm and then load into i3, but i prefer to boot right into a tty then run startx manually. how can i achieve this? i installed xinit but when i run startx as my user i am missing Xorg (which is not an installable user package) <reepca>civodul: I've been trying to figure out how to get eval-with-container working, and I actually don't think it's possible without one of a) linux >= 4.3 b) rewriting everything requiring capabilities in a new user namespace (at least all of the container set-up) in async-safe C c) absurd hacks like wrapping guile's main() <vagrantc>keegans: you need to mark various things as setuid ... i never came up with a satisfactory setup and switched to sway which works thata way out of the box <vagrantc>keegans: if you have a machine that's wayland capable ... <reepca>maddeningly enough, execve automatically drops all capabilities unless you're root <keegans>i've got an nvidia gpu, does nouveau accelerate wayland desktops? <bricewge>keegans: xorg-start-command take a xorg-configuration and return a startx script <reepca>and unshare can't be used to create a user namespace with guile, because guile is multi-threaded even without the finalization thread and without any user-created threads <keegans>bricewge: then i've got to run that as a desktop service though, no? <keegans>that's what i was looking for, but if i have to setuid things i'd rather not <bricewge>Somebody asked a similar thing recently, we should probably provide a example for it <efraim>or say it's an unsupported configuration <reepca>I find it hard to believe that for linux from 2.5ish to 4.3 it was impossible to pass on capabilities to regular executables for non-root users. <vagrantc>keegans: fwiw, other distros just have those setuid things by default <boeg>I dont understand how secrets are provided when they are not explicitly set in ones config scm. I see in the manual that the root password is not set, and has to be changed, and then you have to use passwd as root to set a password for user accounts, but what about for a LUKS encrypted drive? When/how do I set that? Do I do that from that installer manually before I reboot? <vagrantc>keegans: i would guess sway would work with noveau, but only tried it on intel and rockchip ... <cbaines>boeg, at least when I've installed Guix on an encrypted partition, I provided the password to use when the encrypted partition was created <boeg>cbaines: I mean - isn't the encrypted partition created by guix system? Does it notice that it is creating an encrypted partition and asks for password? <keegans>vagrantc: ah, was not aware that's how it was done. yeah i'll take a crack at sway , are there any example configurations i can take a look at? <cbaines>boeg, I don't believe Guix handles creating partitions. The installer for Guix can, and you set out what partitions to use in a system configuration, but I assume you'll get an error if you've told Guix to mount a partition that doesn't exist. <boeg>cbaines: right okay - think maybe i should see if i can find a guide for how to set up an encrypted drive wit hguix then :) <vagrantc>keegans: you can either start with a base desktop configuration and remove the display manager ... which i found difficult ... or add a small number of service to a minimal system... <vagrantc>keegans: i don't have my config handy where i did the latter <keegans>ah, ok. i'll try to figure it out then <vagrantc>keegans: i think elogind and dbus might hve been the key service i needed to add <keegans>i'll try a few things and see what i can get to work <bricewge>Hum sorry, it doesn't need setuid it's just put in a unusual path for a Guix System <civodul>reepca: re eval-with-container, i'd happily read more details and discuss! <efraim>bricewge: yeah, the propagate the filesystem libraries patch <bricewge>I have sent a v2 for the one patch that didn't made it into master <efraim>yeah, I have it sitting in my mailbox <bricewge>It's just that I'm not really sure about using propagated-inputs in that case <efraim>that's the part that made me hesitate <efraim>there's nothing to wrap so that doesn't really help, but we'd want udisks to just work with libblockdev <efraim>on the other hand it should just work if the user has already installed those programs for their filesystems anyway <efraim>the difference in guix size goes from 349 MB to 369 MB so not a big jump <efraim>I think it's ok to add them in, I'll add a note explaining why they're propagated <bricewge>Or we could patch libblockdev: replacing the binary name with the full path of our utilities? <efraim>;; These are needed by libblockdev at build time and at run time but are not actually linked against so we propagate them to prevent any suprises. <efraim>yeah I was going to say I found some spots with const gchar ... {"xfs_admin" and others <efraim>in plugins/fs/xfs.c i see mkfs.xfs xfs_repair xfs_admin xfs_growfs and xfs_db maybe <lfam>I see that we are missing substitutes for the usual suspects again :( <efraim>right, but for say plugins/fs/xfs.c:161 we can patch xfs_repair to (string-append (assoc-ref inputs "xfsprogs") "/sbin/xfs_repair") <efraim>looking at plugins/btrfs.c it's only the ones that are \"btrfs\"\, <efraim>it does only seem to call them each in their own single file so we could probably patch them all if we wanted <bricewge>keegans: Don't try what I suggested it doesn't seems to work, sorry. <bricewge>It's better for the references to be apparent, no? <efraim>there's also plugins for bcache and lvm, I think it might be best to just add to the description that if you want more filesystem support you need to install the coresponding helper library <efraim>I think v1 where we just add xfs to the inputs and we add a sentance to the description telling people to install helper libraries as needed <bricewge>I don't understand: why should we keep the 4 fs-utilities as inputs if they aren't used at all? IIUC <efraim>I think they're needed to build the filesystem plugins, but I just took btrfs out of the inputs and it still built the btrfs plugin so now I'm not sure <bricewge>Yep they aren't needed at build time, see the configure summary. <efraim>it's getting late here, I'll write an email tomorrow and think about patching vs propagating vs removing from inputs altogether <efraim>np :) It was good to talk about it ***vagrantc_ is now known as vagrantc
<boeg>I was thinking that I want to actually do some programming in my config.scm, and i thought that I dont want my UUIDs hardcoded in my config.scm but instead generate them. Can I just put something like `(system "blkid -s UUID -o value /dev/sdaX")` where there now is a string and it should work? <boeg>ah, yeah, seems to not work like that <boeg>rekado: got it working, thanks for your example! <PotentialUser-72>Hi, sorry to say sitting her on my laptop with the desktop `guix system init ...` broken. Anyone free to help? 