***lukedashjr is now known as luke-jr
<dftxbs3e>would it be acceptable to merge fwupd upstream without any default repository (that might have nonfree firmware) configured? ***lukedashjr is now known as luke-jr
<dftxbs3e>sneek later tell marusich I ran "$ git cherry-pick dcaf49b8f5 c5498d5925 a8fd25c51a 52f0b3db6e a1cdd9de3c" on top of master (commits from wip-ppc64le branch), basically I ignored all previous wip-ppc64le commits and started working on top of master with your bootstrap binaries commits ***lukedashjr is now known as luke-jr
***txgvnn is now known as Guest95054
***iyzsong-- is now known as iyzsong-w
<xelxebar>So, with a luks-mapped-device, both grub and initramfs ask for password. On other systems, I've suppressed the latter by storing a keyfile directly in the initramfs. <xelxebar>However, this seems non-ideal on a Guix System since that would effectively store the key in the store. <xelxebar>Is there a ready-made solution in Guix to this double-password prompt issue? <apteryx>dftxbs3e: there are already free firmware packaged in Guix, see (gnu packages firmware). Wouldn't it be simpler to add this firmware as a Guix package? <apteryx>I'd trust a Guix package more to install firmware than a 3rd party service. <apteryx>then firmwares can be added directly at the operating system level via the operating-system-firmware field. It default with ath9k-htc-firmware and openfwwf-firmware. <apteryx>a comments for why there are two licenses would be in order, I think (are they for different files? dual license?) <apteryx>dftxbs3e: do you want a copyright line added to the file? <remby>do you guys care that the output doesn't line up in `guix package -l` ? ***txgvnn is now known as Guest15828
***iyzsong-- is now known as iyzsong-w
<terpri>xelxebar, i don't think there's a ready-made solution right now unfortunately, for exactly the reason you mention <xelxebar>terpri: Thanks. Not what I wanted hear but sort of expected. <terpri>xelxebar, but...grub does apparently have "multiple initrd support" (per NEWS, but seems undocumented) which should allow the kernel to combine the public initrd in the store with a private initrd containing a keyfile <xelxebar>Very interesting... Thanks for the pointer. I'll have to take a look. <terpri>no idea what that looks like from "within" the early boot process, but it annoys me too so i might play with it later. (gnu system linux-initrd) probably has the relevant code <terpri>random googling suggests the grub side of things might just require passing multiple arguments to the initrd grub command (vs. the single arg documented in the manual) <xelxebar>Interesting, how does this work from the initrd's perspective? Like filesystem overlays? <dftxbs3e>apteryx, hello! thanks for looking, I do not care much for copyright line, licenses are for different files yes.. <dftxbs3e>apteryx, about firmware, fwupd specially implements upgrade mechanisms which can be arbitrarily complicated depending on the hardware. It's also possible to update offline from specific cab file which could be built by GNU Guix from firmware source (like that broadcom free firmware reimplementation). <dftxbs3e>Hardcoding cab files in GNU Guix looks fine by me, and more coherent with the functional model <dftxbs3e>However, I am not sure all firmware supports to be upgraded directly to latest from any version <dftxbs3e>fwupd may make sure to do version-to-version updates.. which would be more complicated to do with GNU Guix <terpri>xelxebar, apparently cpio files can basically be concatenated? i don't know what that looks like "inside" the booting system but the --repl kernel argument drops you into the initrd environment... <terpri>(and would be very easy to follow other distros in allowing keyfiles, assuming it does look something like an overlay fs) <terpri>luks-device-mapping is probably the thing to modify if someone figures out the multiple-initrds stuff, as that's where cryptsetup is called (in the open-lunk-device procedure) <xelxebar>Oh, I vaguely recall cpio being essentially a concatenation of files (with appropriate headers), but never gave it much thought. The catenative proprety is really cool! <xelxebar>I was basically wondering what happens in case of conflicts in directory hierarchies. <xelxebar>Given that the format is really easy to stream, I would assume that it boils down to last is winner. <kndr>I'm trying to create an inferior in a manifest and draw several packages from there, but the trouble is, doing guix package -m manifest.scm asks for authentication. <kndr>The channel I'm trying to download packages from is obviously a custom channel <kndr>and, as far as I know, `package` subcommand does not have a --disable-authentication keyword <kndr>Any way to go around this? <kndr>btw, looking forward to FOSDEM :) ***txgvnn is now known as Guest51549
<xelxebar>Could someone sanity check me? Does this command fail for you too: guix time-machine --commit=c072887bb0b37aea36d6ba6e0dc636858b4b440a -- build nyxt <xelxebar>Trying to build an older version of nyxt, but its toppling over when trying to build guix-package-cache: (exception unbound-variable (value #f) (value "Unbound variable: ~S") (value (rust-1.47)) (value #f)) <g_bor[m]>I am perparing the Outrechy particpation of the 2021 summer round. <g_bor[m]>I would like to create a libreplanet page, but I can't see how to do that. <g_bor[m]>I would like to bring up a gsoc-2021 and an ourteachy-2021-summer page. Is this because I need some additional access rights? <cbaines>I think I can see how to create pages <cbaines>g_bor[m], I just tried going to a page that didn't exist, and the text of the page says this: <cbaines>There is currently no text in this page. You can search for this page title in other pages, search the related logs, or create this page. <cbaines>"create this page" is a link, I haven't tried clicking it yet <zjgkkn[m]>Hi! `guix install xmonad ghc-xmonad-contrib` is not enought to run xmonad. GHC is also required, but do not installed as dependency. `guix install ghc` would install ghc 8.8.3, which cause an error while `xmonad --recompile`. Problem was solved by installing ghc@8.6.5. <zjgkkn[m]>How can i prevent upgrading GHC by `guix package -u` in futere? <htgoebel>Do I need to register somewhere or is it okay if I'm listed as a participant? <rekado>xelxebar: I was unable to reproduce this. <rekado>xelxebar: the command produced /gnu/store/z2bwj3dcpma02n9036gkf49qcgns1bwm-nyxt-2-pre-release-4 without errors. <rekado>perhaps the problem is with your current Guix <rekado>I used Guix at commit 0981f872cb1fb94fcdf3f4d00bd08c6a6b61ed8d <rekado>do you have a channel enabled that refers to rust-1.47 before it existed? <xelxebar>Yeah, very possible. Think my machine has some acpi issues. Has been randomly falling back to bios like a hard boot when I leave the machine alone for very long. <xelxebar>I do have a private channel, but nothing with rust in it <xelxebar>rekado: That said, trying again with your commit, it looks to be getting much farther O.o <xelxebar>How did you go about choosing said commit? <cbaines>arrgh, I hate core-updates conflicts :( <rekado>xelxebar: I didn’t choose it. It’s just my current Guix. <xelxebar>Hrm... I guess that means builds can fail due to far away changes on unrelated commits :/ <xelxebar>Incidentally, nyxt 2-pre-release-5 that was crashing and giving all sorts of problems seems to work flawlessly when installing without grafts! <pinoaffe>can I safely remove the ~/.cache/guix directory? <efraim>xelxebar: how about now-ish? with staging merged in there should be few/no current grafts <leoprikler>pinoaffe: ~/.cache should imply "safely removable" <xelxebar>I'm trying to manually produce a hello-world guile script in the store but only able to get a *derivation* that supposedly produces this. <xelxebar>Just running this line in the repl: (run-with-store (open-connection) (gexp->script "my-hello.scm" #~(display "my hello"))) <xelxebar>That returns a derivation, so how do I compile/lower (?) this into a file in the store? <PurpleSym>Can I somehow invoke launch-environment/container from (guix scripts environment) manually in `guix repl`? I tried (@@ (guix scripts environment) launch-environment/container), which is not giving me a reference :/ <liltechdude>Hello. What about to add with non-english translate? <rekado>PurpleSym: it’s unfortunate that so much of (guix scripts environment) is private <rekado>it would have been nice for the GWL if there was a (guix environment) module with more public procedures. <rekado>xelxebar: right, build-derivations does this <rekado>note that instead of gexp->script you can use program-file, which can simply be lowered <PurpleSym>rekado: I guess I’ll have to copy&paste the relevant code then… <xelxebar>rekado: I think I'm confused: (run-with-store (open-connection) (lower-object (program-file "my-hello.scm" #~(display "my hello")))) <xelxebar>What am I supposed to do with the procedure (mval?) returned from lower-object? <kndr>no, it's pop-os + guix-as-a-package-manager <kndr>i'm hopeful purism delivers one of their librems to me in the next ten years <PotentialUser-47>how do I list all the environments I've created with `guix environment`, so that I may clean them up once I'm done with them? <xelxebar>PotentialUser-47: Unless you explicitly created gc roots for them, then you can just let general guix gc clean them up. <xelxebar>If you created roots, then guix gc --list-roots should give you some hints. <kndr>btw, how to use a local directory in `channel` definition (instead of `url`) <kndr>i _think_ i read on this somewhere in the docs, but can't find it anymore <rekado>xelxebar: give me a few minutes; I’ll show you an example then <rekado>xelxebar: there might be a simpler way to do this, but this is the way I understand well enough. <xelxebar>rekado: Thanks. So essentially the data flow is like this: gexp -> program-file -> lower-object -> build-derivations, using mlet to thread the store connection through as needed. <paulj>I am having some fun with dwm, and would be interested in any comments or experience from others. dwm needs to be recompiled to change configuration. To this end, I have created a custom channel, and then made package definitions to inherit the relevant package definition from the standard code, but to use code in my own git repos. This is all working as expected. Since this configuration is for me, it doesn't seem correct to install <paulj>the packages in the system configuration, but instead I should install them in the user environment. When I do this, of course slim doesn't see it, and I can't select dwm to log in. I then thought that logging into the console and running startx (or equivalent) should work (I don't really _need_ a login manager!). I removed the relevant section adding slim as a service, and the system started in the console. Running xorg is then not <rekado>the mlet (and all the other m* things in (guix monads)) lets you define a monadic computation <rekado>that’s the only thing that needs the store connection <Rovanion>Is it possible to compile a program in a `guix environment` using an old version of glibc? I've tried `guix environment --pure --ad-hoc glibc@2.29 gcc` but after running `rm a.out && gcc printer.c && ldd a.out` the result is that glibc-2.31 is still in use. <mhj[m]>Hiyo all, how does the Guix system fare on a Pinebook Pro currently? I was trying to build an image for it based on a template, then I found something else that worked initially, but I just can't seem to get it right. Any Guix wizards here? ***chrislck_ is now known as chrislck
<civodul>mhj[m]: hi! i think some tried running it on a Pinebook Pro, but i don't know what the outcome was <NieDzejkob>paulj: in my case it's ratpoison but not dwm, but I didn't have any troubles with using slim-service-type in my system config + ratpoison installed in ~/.guix-profile/bin <NieDzejkob>ah, right, you probably want to create an ~/.xinitrc <roptat>another English question. Since 0 is a plural, should I say "no features change"? or "no feature changes"? <NieDzejkob>"In a bugfix release, no features change" / "This release contains no feature changes" <roptat>so, depends if change is a verb or a noun? <roptat>in my case, it's "as long as no features change, we can do that and that" <roptat>so change is a verb, and we use the plural on features <leoprikler>in this case it would be "no feature changes/none of the features change" <efraim>do we have fuse support for file-system declarations in the os-config? <roptat>so, you almost got me with the simple explanation :p <roptat>not like you'd find any that's not context sensitive <roptat>should we add them to the website now? <roptat>I can take care of that if you think it's a good idea <roptat>btw, did you reconfigure berlin? <efraim>does anyone have an example for mounting nfs directories in an os-config? <paulj>NieDzejkob: Sorry - I've been away! How did you get slim to pick up ratpoison? <civodul>roptat: mothacehe reconfigured berlin recently and i confirm the "fr" redirection works <civodul>as for translations, maybe we can wait until a threshold has been reached, like one third? <paulj>Did you do any specific setup? <NieDzejkob>paulj: as I've said, I don't think it's slim picking up ratpoison by itself, but slim runs my ~/.xinitrc which ends with 'exec ratposion' <NieDzejkob>sorry, it's been a long time since I was setting this up <paulj>NieDzejkob: I'll give that a try - I'll make a .xinitrc with the relevant code to start dwm and the support programs and see if slim finds it. I did think I might have to put something in /etc (or the equivalent guix location). <paulj>What does slim offer you as session? <dongcarl>I'm guessing I need nss-certs as a propagated inputs... But that doesn't seem to fix it <dongcarl>environment variable `SSL_CERT_DIR' set to `/gnu/store/0y63pfqi2626nwsjbksrci1jyd62cxx8-nss-certs-3.50/etc/ssl/certs' <dongcarl>environment variable `SSL_CERT_FILE' unset <dongcarl>roptat: I'm guessing if something is using oscrypto it most likely wants its SSL functionality to work? <raghavgururajan>dongcarl: May be you can use `/run/current-system/profile/etc/ssl/certs` <roptat>bad idea, that wouldn't work on a foreign distro <roptat>and it's not available in the build environment <dongcarl>Right, I need it in a build environment... <roptat>can you set SSL_CERT_FILE yourself, then? <dongcarl>roptat: of course, I can add a phase that does that <dongcarl>Feel free to point me to existing code snippets that do the same thing <roptat>but it should already be set, especially if you already have SSL_CERT_FILE <zimoun>civodul: thanks for the hint. You have not answered about the replacement or not of all the args-fold* by parse-command-line. I think it will improve the thing. WDYT? <roptat>they should both be set when you add nss-certs (or openssl, not sure) <roptat>should be set to etc/ssl/certs/ca-certificates.crt <PotentialUser-47>xelxebar is there any way to associate a logical name to a guix environment? If I restart my laptop and later need to go back to an environment I created previously, using guix gc --list-roots, and sifting through the manifest files there is not very practical. I'm looking for the guix equivalent of python mkvirtualenv <name>, rmvirtualenv, lsvirtu <xelxebar>PotentialUser-47: Well, if you use --root <foo>, then <foo> is just a symlink. You can store that wherever and named however you want. <smigold>Hi. Is it possible to use Guix's Shepherd and Shepherd user services on foreign distros in addition to the distro-specific init system? <shcv>is there an emacs daemon service definition? I didn't see one in the gnu/services directory <shcv>on my gentoo system I'm used to having an emacs.<user> service that is started during boot; but that may not be the standard guix way? <dongcarl>roptat: Right, the build logs say that SSL_CERT_DIR is set but SSL_CERT_FILE is unset... <dongcarl>roptat: I _am_ on an older version of guix, could SSL_CERT_FILE be a recent thing? <xelxebar>PotentialUser-47: No worries. That's a normal feeling :) Guix has its own way of doing things which takes some time and mucking about to get used to. <PotentialUser-47>I'm using .envrc a lot, and was hoping to re-enter the guix environment when I cd to the project directory i'm working on <xelxebar>That root is just a script that executes qemu, IIRC. <xelxebar>Rather, that *build product*. You might as well just take a look directly at whatever it is. <civodul>zimoun: sorry, i overlooked the arg-fold* issue; i don't know actually, maybe it's the right thing to do <civodul>if that's the case you can submit a patch <paulj>NieDzejkob: Sorted it! I removed all window managers from the system, then created a .xsession file with 'exec dwm'. I installed the modified packages in a manifest under the user account. Now slim goes straight into the dwm window manager. Thanks for your help! <paulj>Now for my next challenge - for some reason guix isn't updating the dwm installation when I change the git commit in the package definition. I am sure I read about that somewhere...! <NieDzejkob>the problem is that the filename and expected hash of the source match the previous tarball <zimoun>civodul: ok, I will submit a patch. Then it will be easier to know if it is a good idea. :-) <paulj>I am updating the git commit and the hash when I commit the new version. Here is the definition I am using: paste.debian.net/1183870 <leoprikler>btw. Telegram Desktop broke with the Qt upgrade (staging merge) <liltechdude>lil@guix ~/guix [env]$ ./pre-inst-env guix build hello <liltechdude>./pre-inst-env: line 55: exec: guix: cannot execute: Is a directory <liltechdude>lil@guix ~/guix [env]$ ./pre-inst-env ./scripts/guix.in build hello <liltechdude>./pre-inst-env: line 55: /home/lil/guix/scripts/guix.in: Permission denied <liltechdude>./pre-inst-env: line 55: exec: /home/lil/guix/scripts/guix.in: cannot execute: Permission denied *raghavgururajan takes a look at Qt's shenanigans <leoprikler>all the way from bootstrap over configure to make *raghavgururajan takes a deep breath <leoprikler>The breaking component is materialdecoration, which from my POV doesn't even work if I check out the "fix build against newer Qt" commit <liltechdude>hmmm, previously, when I run `make check` it run also `make`? Check is so slow that I skip this part :))) <leoprikler>kinda sad, that telega is still the best working for all your efforts ***amiloradovsky1 is now known as amiloradovsky
<joshuaBPMan> raghavgururajan: thanks for packaging linphone. It's been working amazing for me. <raghavgururajan>Is there a guixy way of creating custom environment-variables in guix system? <leoprikler>there is a service type for environment variables, so you can make that a simple-service if you need to <lfam>I keep harping on this, but can we adjust the issues.guix.gnu.org software so it doesn't send messages that are from "guix-patches--- via <guix-patches@gnu.org>"? Like, can we make that include the submitter's name? <lfam>The messages show the sender's name and address in the Resent-From header <lfam>I'm not sure what piece of software to look at. Is it mumi? <lfam>I'm not sure yet what is the common cause for these messages. I'm guessing they are created via the web interface <lfam>Oh, I had thought it was possible to send patches via the web interface, but I guess that's not true <lfam>It's possible to reply, however <xelxebar>Oh my goodness! After spending all day jumping down rabbit holes, I finally figured out how to create a custom wrapper script for a package! <xelxebar>Along the way and with the help of #guix, I also ended up learning how to manually wrange derivations. :P <xelxebar>Anyway. It's 4:30 am. I really need to get to bed. Night, Guix! <raghavgururajan>> leoprikler: there is a service type for environment variables, so you can make that a simple-service if you need to <raghavgururajan>Oh nice! By any chance you know the service-type name on top of your head? <leoprikler>iirc environment-service-type but my memory is a bit hazy <lispmacs[work]>hi, where in the filesystem can I look through udev rules defined by the active system? <leoprikler>lispmacs[work]: /run/current-system/profile/lib/udev <hapster>I am trying to import a gpg key from another disk to guix, and it proves to be a bit difficult <hapster>first off, I needed to sudo every command related to gpg (is that normal?) <hapster>finally, I managed to import the key <leoprikler>Gratuitous use of sudo is the best way of fucking – pardon my language – your system on any distro. <hapster>so what did I do wrong. I copied the contents of my gpg folder from another machine to this one <hapster>I guess thats already the wrong step, right? <hapster>I mean I proceeded in two independent steps: first I simply copied my gpg folder (pass didnt work, surprise surprise). <hapster>Then I did the --import --export with gpg and copied the resulting files afterwards <NieDzejkob>perhaps you copied it in a way that borked the permissions on that directory? <hapster>NieDzejkob: but then, wouldnt chown solve the issue? <hapster>I mean I literally copied the folder using cp <leoprikler>also copying gpg secrets over the network might be an issue in and of itself <hapster>but what you guys are saying is that normally I shouldnt even have to use --export --import, but that a chown after copying the folder from the other machine should be sufficient? <hapster>leoprikler: its the different hard drives in one machine <leoprikler>hapster: for the record, all files in your ~/.gnupg should be your-user-name:users <hapster>leoprikler: I have only one (pubring.kbx) which exhibits such permissions <rekado>I haven’t read the whole discussion, but I generally only copy with ‘cp -ar’ to keep permissions unchanged <Thrilleratplay_>That would only work if the user and group ids were the same from the original system and Guix system <Thrilleratplay_>The permissions for my keys are -rw-r--r-- (644) or -rw-------(600). There are a few files with the stickbit (1700) but I think these are zero byte garbage files from something I was doing three years ago. <hapster>so I got it working, seems like it had to do with me copying stuff over from the other machine instead of relying on --export and --import <hapster>ho well, seems like stow cannot solve every problem there is :D <hapster>thanks guys, see you at fosdem maybe :) <mzbm>hi everyone, i've got some problems that gnome is crashing when doing some cpu intensive stuff. It seems like it crashes when ungoogled-chromium is needing a lot of resources. After some time (when i use graphql playground) the system crashes and i come back to the gdm login. Is there any way i can debug this? <lispmacs[work]>hi, I'm trying to integrate a udev-rule into my system config. I think it needs to be inserted somewhere in this services section, but am unclear on where and exactly how: https://bpa.st/RENA <leoprikler>udev-service-type takes a list of packages, that provide udev rules <Thrilleratplay_>mzbm By crash, does the system do a full reboot or does your session crash and are logged out to gdm? <mzbm>Thrilleratplay_: only my session crashes and i come back to the login <mzbm>but i don't know if its related <jmarsden[m]>lispmacs[work]: There is a section of the Guix Reference Manual with an example, search it for %example-udev-rule <Thrilleratplay_>mzdm That sounds like an X issue. The log may under /var/log/ as Xorg.0.log or Xorg.0.log.old, depending on the severity of the crash. /var/log/gdm may have logs that could help but doubt it. Also try `sudo journalctl -r` (systemd logs in reverse) and scroll down to when the last crash occurred. Sorry I do not use GuixSD, I am here to ask a question myself but have yet to word it correctly to not sounds like a jabbering mental case. <Thrilleratplay_>mzbm Oh...or you are missing the free nvidia firmware. If Chrome is trying to use hardware acceleration, that may throw a fault if it is missing. <mzbm>hmm ok i'll try another browser and see if this fixes it. thanks!! <Thrilleratplay_>or disable hardware acceleration in ungoogled-chromium or install the free firmware for nvidia drivers. <mzbm>i can't install the firmware because i'm using linux-libre kernel <leoprikler>Free nvidia firmware? Where do I apply for that? <Thrilleratplay_>mzbm Oops. I thought there were partial libre firmware, guess not. Sorry. <Thrilleratplay_>I want to modify the docker image output of guix pack that likely would not be accepted into guix upstream. It seems less messy to create a new guix subcommand instead of modifying or overriding pack. I want to be able to hack on this script, is there a path that can be defined for development? <Thrilleratplay_>How do Guix developers set up their environments to extend package manager functionality? <pkill9>is there a linux file manager that is effectively like an interactive `tree` command? <pkill9>i.e. you can select the items output by `tree` and move them etc <pkill9>i was just thinking, unlike ranger you can see more than one level deeper *civodul looks for a trick to change the uid_map of a process living in a separate user namespace <civodul>uid_map is a one-time thing, so i thought you could unshare(CLONE_NEWUSER) and set up a new uid_map, but apparently not <dftxbs3e>apteryx, thank you for merging the patch! <apteryx>dftxbs3e: no problem! I hope it works alright!