IRC channel logs


back to list of logs

<Noisytoot>lle-bout, I've sent it, although it hasn't appeared on yet
<lle-bout>raghavgururajan: in "d249e0a0eb * gnu: gst-plugins-bad: Fix indentation." qtbase etc, is that expected to indent so much to the right..?
<lle-bout>Noisytoot: great!!
<lle-bout>Noisytoot: once you have a link and all, add it there also: - similar to GNOME 40 upgrade
<lfam>It should appear now Noisytoot
<lle-bout>lfam: posted also
<Noisytoot>lfam, it doesn't appear
<Noisytoot>It says "Archives are refreshed every 15 minutes"
<lfam>Try waiting about 15 minutes
<lle-bout>lfam: this worked quite well: guix package --search=. | recsel -e 'dependencies ~ "patch@2.7.6"' -p name,synopsis,dependencies
<lle-bout>yours had false positives out of the box
<lfam>I did think it had too many results
<lle-bout>lfam: so I am thinking the only legitimate reason to include patch as inputs is to reference it with assoc-ref somehow
<lle-bout>lfam: but in these cases, couldnt we use (which "patch") instead?
<lfam>I've noticed that there can be a namespace collision when referring to 'patch' in packages
<lfam>If there is a patch file being used, it gets complicated
<lle-bout>lfam: I propose we first do such changes and push to master once verified they don't cause world rebuilds, then move on to hidding the patch package and creating a new one? WDYT?
<lfam>I'm not sure what that means for these packages
<lfam>I'm not sure I understand what you are proposing
<lle-bout>lfam: removing all the dependencies on the patch package besides the build system
<lfam>Oh, right
<lle-bout>then hide the patch package using (hidden-package ..) while still referencing it in the build system, then create a new patch package
<lfam>Is that how the foo/stable thing works? Or is that not useful here?
<lle-bout>lfam: I have no idea (yet)
<lfam>We should try to re-use that, if it's relevant
<lle-bout>lfam: I don't think so, I think it had to do with things only being required as native-inputs
<lle-bout>But then the naming /stable doesnt make much sense to me
<lfam>The name isn't important. It's just an implementation detail
<lfam>Just like foo/fixed
<lle-bout>lfam: well I like to make descriptive names, for /fixed I think it makes sense.. but stable hmm
<lfam>"Fixed" has another meaning in functional programming and we confused some people with that
<lfam>I just don't want to solve this problem in a new way, if foo/stable would work
<lfam>We should try to avoid having multiple ways of solving the same problem
<lle-bout>lfam: sure, but I don't think it applies, let's check with Mark
<lfam>To me, it's a similar task: how to deliver security updates for interactive use, but not affect build-time uses
<lle-bout>lfam: hmm I see
<lfam>But, I didn't look into the details of Mark's work yet
<lfam>His work is usually very good, though. He specified and wrote the grafting code
<lle-bout>lfam: cool! I'm reading the commits, it does look like it could help
<Noisytoot>It's appeared:
<Noisytoot>What section should it be on
<lle-bout>Noisytoot: create a new one
<lle-bout>Noisytoot: name it for example: 'Web services'
<raghavgururajan>lle-bout: I used the indent.el script. I didn't do it manually.
<lle-bout>raghavgururajan: I see.
<lle-bout>Noisytoot: thanks for filling up the wiki!
<lle-bout>raghavgururajan: if you have any difficulty with git-send-email please email me at lle-bout AT zaclys DOT net for questions, asynchronicity is easier for me these days
<lle-bout>After the 17th April it will be easier for me
<lle-bout>raghavgururajan: I think all your patches are fine, I reviewed all until gst-editing-services update, then gpupnp-igd etc. these are new I didnt review yet
<lle-bout>raghavgururajan: also some rebase needs to happen on core-updates (no conflict) some new commits appeared
<lle-bout>raghavgururajan: I have difficulty reading into the plan you laid out in the disroot calc page, it would help if you outlined that plan in the ML
<raghavgururajan>> lle-bout‎: raghavgururajan: if you have any difficulty with git-send-email please email me at lle-bout AT zaclys DOT net for questions, asynchronicity is easier for me these days
<raghavgururajan>> lle-bout‎: raghavgururajan: I think all your patches are fine, I reviewed all until gst-editing-services update, then gpupnp-igd etc. these are new I didnt review yet
<raghavgururajan>Thanks! Yeah, anything after gst-editing-services are not ready.
<raghavgururajan>> ‎lle-bout‎: raghavgururajan: also some rebase needs to happen on core-updates (no conflict) some new commits appeared
<raghavgururajan>Sure and thanks!
<lle-bout>raghavgururajan: the only thing I have to say now is add more comments or more verbose commit messages so we understand what you're trying to do
<lle-bout>In every commit, and in the future too
<lle-bout>raghavgururajan: telling the *what* is not very useful in commit messages or comments, I am not sure whether the *why* should be located in commit messages or comments, I tend to do a bit of both, comments for when that's relevant to someone reading the code, commit message when it's more relevant to someone reading the history, sometimes a bit of both when it's a bit of the two
<lle-bout>raghavgururajan: One question I have is about the xorg servers used in tests, it never quits, does it get killed automatically?
<raghavgururajan>Makes sense.
<raghavgururajan>I was under the impression that it gets killed at the end of build process. Did you notice otherwise?
<lle-bout>raghavgururajan: in many cases the spirit of your current commit messages works fine, e.g. "small" changes with not much impact besides the thing it's changing itself, but I am thinking for such core packages it might be important to explain more there
<lle-bout>raghavgururajan: I didnt notice otherwise but it puzzled me!
<raghavgururajan>I started to do more comments, may be I should do it more. Thanks for the feedback. Will do.
<lle-bout>raghavgururajan: I hope it will not be too troublesome to do!
<lle-bout>I am thinking there's some rebase-fu I don't know about that could make this surprisingly easy but yeah
<raghavgururajan>Nah! I just have to include comments via rebase -i
<raghavgururajan>Hope it doesn't cause conflicts while rebase --continue
<lle-bout>raghavgururajan: I am always a bit scared you give up because I give you review comments that are troublesome to follow
<lle-bout>Since it's a bit the reason why the GNOME upgrade has stalled
<lle-bout>Part of it, probably
<raghavgururajan>lle-bout: I don't give up on Guix, never. :D
<lle-bout>raghavgururajan: If not Guix, maybe this GNOME upgrade, but glad to know about it! I am happy that's the case! Because we can't afford giving up because else core-updates will remain broken in ways that nobody else will be able to fix.
<lle-bout>raghavgururajan: We have time though, so no rush!
<lle-bout>After the 1.2.1 release I am hopeful we'll get help!
<lle-bout>raghavgururajan: thanks a lot for your work, I need to go sleep now!
<link2xt>looks like cargo-build-system can only vendor dependencies from crates-io, but not from git
<link2xt>if there are git dependencies, .cargo/config should also have an entry for each of them
<jgomo3>I thought the package emacs-recutils would add the rec-mode. It is not happening for me. Is there another package which add rec-mode to emacs?
<jackhill>jgomo3: it used to, but I guess it that packages slipped through the cracks with a chnage to where we install site-lisp files (or doesn't set the search path or something). Would you mind opening a bug?
<jgomo3>jackhill: Sure, I will, but before: are you sure is not my system which is wrong?
<jackhill>jgomo3: doesn't work for me either. I already had it installed because I used it in the past, but I guess not for some time.
<jackhill>and I appreciate you opening the bug, that way we probably won't forget about it :)
<jgomo3>jackkill: Cool. Done!
<awb999>any idea, ho I would be able to run vscode (visual studio code) via guix ? It seems not to be available as a package. And if I switch to guix sd, then I would need to have some kind of idea how to run it.
<jgomo3>awb999: Maybe it's Flatpack version?
<jgomo3>jackkill: I had to (require 'rec-mode). After that, the rec-mode works fine.
<jackhill>jgomo3: ah I see, I guess I made the same mistake. Well, at least there is pleasure in company. If you want to close the bug, you can send that information to . Otherwise, I'm happy to do it since I lead you astray
<jgomo3>jackhill: Thank you. I'll close it.
<jackhill>sounds good :)
***iyzsong-- is now known as iyzsong-w
***lukedashjr is now known as luke-jr
<genr8_>can I make a suggestion regarding the internal bootstrap process
<genr8_>there exists such a thing called "xz-embedded" from the same
<genr8_>and a new version has come out 2021 - 02 - 01. instead of jumping through so many hoops to keep falling back to gzip or needing to build an entire xz so early, cant we just stick to that
<genr8_>i assume theres something im missing, but i want to bring up that new release.
<jackhill>genr8_: maybe send mail to guix-devel. I'm not sure the people most familiar with the boostrap are around currently
<mothacehe>hey guix!
<efraim>genr8_: agreed, send a mail to guix-devel. I certainly like the idea for unpacking xz tarballs. Doesn't help that IIRC we still default to xz compressed tarballs after applying patches or source snippets
<sss2>hi all
<sss2>how can i skip configure phase for gnu build system ?
<civodul>Hello Guix!
<leoprikler>sss2: delete 'configure
<sss2>leoprikler, thx
<meo>is there a substantial difference between deploying a guix system or guix over an existing (freshly installed) OS?
<sss2>looks like it will be hard way
<sss2> - i need this program to build packages
<sss2>as i can see it have custom build process
<sss2>is here some generic procedure to write "manual install" packages ?
<sss2>as i can see whole preparing, configuring and building is done by runing ./bootstrap , also install method is not provided
<sss2>i found large patch including rebar here but looks like it is not what i need
<efraim>I can say the go importer is working nicely. I have almost everything unbundled from keybase
***iyzsong-- is now known as iyzsong-w
<lle-bout>hello! :-D
<Frosku>Has anyone got aws cli (v2) working on guix?
<lle-bout`>rekado: hello! :-) what's the URL for downloading patchsets on Since search doesnt seem to work well in I can't find your older message and I don't keep IRC logs so.. :-S - Thank you!
***lle-bout` is now known as lle-bout
<raingloom>hmmm, does --keep-failed not work with offloading?
<efraim>raingloom: it's a known limitation
<efraim>partly because it's not clear if it should be kept remotely or brought back from the offloading machine
<roptat>hi guix!
<roptat>just pushed new translations for the website: Esperanto (100%), Russian (40%) and Korean (70%) \o/
<roptat>we have wonderful translators, don't we :D
<civodul>roptat: yay! thank you!
<civodul>it's amazing
<civodul>roptat: i think we have to update the nginx conf for auto redirects, right?
<raingloom>efraim: could just provide an explicit choice then. in my case i woldn't mind ssh-ing into my big desktop machine that i offload larger builds to.
<roptat>civodul, ah you're right
<contrapunctus>I'd like to install Guix, but I have an LVM setup and I'm not familiar with Guix or its DSL...the news file shows an example of defining an LV, and the docs have an example of defining a mount point, but how do I define an LV as a mount point? 🤔
<roptat>civodul, pushed, the redirection should be active after the next berlin reconfiguration
<civodul>roptat: woohoo, thanks!
<sss2>hi all again
<sss2>i need "build system" which do nothing but allow define all stages manually
<sss2>i want to create package for
<efraim>I wonder if I create a patch set with --threaded if I still need to wait for a bug id
<roptat>sss2, you can start from the gnu-build-system, and remove/replace/add phases, that's how we define all other build systems
<roptat>there's also the trivial-build-system, but it's... complex because it doesn't have this notion of build stages
<sss2>looks like i got compilcated task from the very begining (
<sss2>but ok
<sss2>still trying to implement ejabberd
<roptat>yeah, packaging a language that doesn't have a build system in guix yet is complex
<roptat>basically, you'll have to create the build system yourself ^^'
<sss2>i meant not this
<sss2>ejabberd have a lot of deps
<roptat>ah, ok
<sss2>and this rebar is dep of other deps
<sss2>so, for now i need to create rebar package
<sss2>and use it for all other deps
<sss2>doable of course
<sss2>but unexpected (
<civodul>on a Coreutils rewrite in Rust (i.e., under a permissive license):
<civodul>i say: Gash FTW!
<lle-bout>hellooo! (again)
<roptat>it's not like it's actually going to be used, but it looks like a nice project, if it weren't for rust
<roptat>I mean, I understand why some people want to rewrite everything in rust, but the bootstrap chain is prohibitive
<yoctocell>abcdw: Should we have a way to create an XDG MIME .desktop file using the Scheme API?
<lle-bout>roptat: once mrustc or the gcc rust frontend can compile the reference rust compiler then it'll be OK, but what I am more worried about is that this new coreutils is under a permissive license.
<civodul>roptat: yeah, that vision is somewhat interesting technically but the implications (getting rid of copyleft) are worrisome
<lle-bout>civodul: +1
<yoctocell>abcdw: If say someone wanted to use Emacs dired as their file manager, we could have an option that would automatically create the .desktop file and configure XDG MIME to use it
<lle-bout>The main reason I stopped contributing to Rust is that the culture is fundementally anti-copyleft
<lle-bout>civodul: I also like Rust technically that's why I am so happy there's a Rust GCC frontend being developped then maybe some copyleft Rust libraries etc. can grow from there.
<lle-bout>Copyleft or else, but at least not automatically permissive.
<roptat>ok, thanks for making it clear :)
<rekado>lle-bout: it’s now deployed.
<abcdw>yoctocell: It's not urgent, but will be useful in some cases for sure. If you want to cover this feature, please do)
<yoctocell>abcdw: Ok, I will include it to make service more complete. :)
<abcdw>yoctocell: cool, will be happy to start using it (:
<civodul>there's a bug that OpenJDK currently refers to previous versions of itself, making the closure huge
<civodul>but i can't find that bug, does anyone remember?
<roptat> \o/
<roptat>as well as and :)
<Noisytoot>lle-bout, You could fork the Rust coreutils under the GPL
<lle-bout>Noisytoot: what do you mean? If it's under a permissive license it's under a permissive license there's nothing you can do about it
<lle-bout>rekado: hey!! thanks so much!!
<Noisytoot>lle-bout, If you modify it, you can release your modifications under the GPL
<lle-bout>rekado: the link you sent for me is a blank page, is that expected?
<lle-bout>Noisytoot: I'm not sure I have modifications, it is likely there's more momentum on the permissive version socially now
<lle-bout>rekado, civodul: What's troublesome with the "Rewrite It In Rust" culture is that it also carries the "Rewrite It Under a Permissive License" side of things most (if not all) of the times
<lle-bout>Basically most people who spend the effort with rewriting in rust arent much aware about Free Software political ideas or vision in any way, they just like Rust and go with the flow
<lle-bout>I have experienced multiple projects where they were under AGPL even, and the original author has got a contract with a company and asked everyone to sign a CLA then re-licensed under Apache2 and everyone signed it without questionning anything, that was quite troubling to me. When asked people said they just did not care they just liked Rust.
<Noisytoot>Rust also promotes Discord on its website
<lle-bout>Noisytoot: There's that but also most development teams use Zulip right now
<lle-bout>Zulip is Free Software
<roptat>about the wiki: "and notes about WIPs (among other things)." WIPs (and its definition) is not translatable :/
<roptat>I know how to fix WIPs, but not the attribute that gives the definition
<roptat>(I mean in most languages the translation would still be WIPs I guess, but in French at least, it's weird because we don't use the plural mark on acronyms)
<civodul>roptat: where do you see that string?
<civodul>oh i see
<civodul>more generally: let's avoid jargon and acronyms
<civodul>IMO this should be understandble by everyone, not just insiders and techies
<civodul>roptat: maybe you can just change that to "notes about work in progress, among other things."
<roptat>sounds good
<roptat>civodul, fixed
<roptat>I still don't understand (from the manual): "To add the binary path @code{_jll.jl} packages, you need to ..."
<roptat>what do we want to add to what?
<Noisytoot>What's the difference between login USER and ql USER at the GNU Hurd login screen?
<roptat>no idea
<Noisytoot>How can I change the keyboard layout to UK English on GNU Hurd?
<roptat>set-keyboard-layout doesn't work?
<roptat>or the keyboard-layout field in the os declaration?
<roptat>oh, I think I remember it was not possible to use the extension (like I could set fr, but not fr-bepo)
<Noisytoot>roptat, -bash: set-keyboard-layout: command not found
<roptat>I meant from the os declaration
<roptat>if it's for xorg, then setxkbmap should work
<Noisytoot>I don't have xorg installed
<roptat>then I don't know, I only used the debian hurd at the time (~5 or 10 years ago)
<Noisytoot>I'm using bare-hurd.tmpl
<roptat>maybe "loadkeys" from a tty then
<Noisytoot>-bash: loadkeys: command not found
<roptat>well, you'll have to install it I guess, not sure which package it comes from
<sss2>is here function like install-file but for directory ?
<roptat>Noisytoot, apparently it comes from the "kbd" package
<roptat>sss2, instead of install, you could use copy-files-recursively
<roptat>I'm not sure what the difference is exactly
<sss2>i do not see it anywhere in packages
<roptat>mh... did I confuse it for something else?
<sss2>this does exists
<roptat>ah correct!
<roptat>I confused it with delete-files-recursively, but copy-recursively would do what you want
<sss2>i should probably create public repo to annoy you devs with my stupid questions more efficiently
<sss2>one package from this chain already building and even working
<roptat>does that mean you managed to build the dependency? :D
<sss2>first one
<paulj>Good afternoon all! I appear to have introduced an error into my system config, but as yet I haven't been able to put my finger on it. If I load guix repl, then ,use the definition, I get an error: "Wrong number of arguments to #<procedure cons (_ _)" but I don't see which line number has triggered this error. Is there a way to get the repl to give me more information?
<sss2>and "build" is too large word for this
<roptat>paulj, if guile is not being cooperative (even with ,bt), then you can always copy partial expressions from your config that contain "cons" and see what it says
<roptat>paulj, maybe replacing cons with cons* would work in your case?
<Noisytoot>Are substitutes available for Hurd?
<paulj>roptat: I'll do that then, as there isn't any backtrace available. I'll also try cons* instead where there is cons, but I am not sure of any change made recently (since the last successful system reconfigure.
<roptat>Noisytoot, I don't see any from the ci website, but there might be some
<roptat>paulj, if you share your config (on for instance), I can have a look too
<paulj>This is the base definition, and this is called from the specific system definition. The error is somewhere in here!
<roptat>the very first cons in the file has 3 arguments
<roptat>%backlight-udev-rule, etc
<sss2>roptat, what wrong here ?
<roptat>cons only takes too: an element and a list to add the element to
<roptat>cons* takes multiple elements and a list to add these elements to, so you want to use cons* here
<paulj>Thanks - I am looking back over what I had here...
<sss2>looks like version variable is not available ?
<sss2>so visibility does not inherited in lisp ?
<roptat>sss2, right, it's not defined in that context, because it's part of the quasiquote
<roptat>you can escape it with a comma (as in ",version"), then it would bind to the value of the version field in your package definition
<paulj>Ok - now I remember what I did. I introduced a rule for the touchpad, and added it here...
<roptat>mh, having /usr in your packages sounds wrong
<sss2>roptat, why ?
<paulj>roptat: Thank you - it's now working as expected. Guess I need to go and read more about cons and cons*!
<roptat>sss2, we don't have any package with that directory
<sss2>so i should use /lib directly ?
<roptat>instead of /usr/bin, we use /gnu/store/.../bin, etc (by setting --prefix)
<roptat>yes, I think so
<roptat>maybe even lib/erlang/site-lib/base64ur-<version> (or whatever erlang uses typically)
<sss2>this is a good question
<sss2>looks like libs dropped to subdirectory in /lib root
<roptat>this is what we do for python (in lib/python-x.xx), ocaml (in lib/ocaml/site-lib), etc
<sss2>unlike python looks like here is not special lib root
<roptat>if that's possible, would you consider putting them under lib/erlang still?
<roptat>also, you'll have to find a way (maybe an environment variable) for erlang to find them
<sss2>yes yes, i completely agree, also, ejabberd does not see it anyway
<roptat>maybe it's already available, I don't know
<roptat>you'd had an environment variable definition to the erlang package
<sss2>roptat, how to drop some options from passing to configure script ?
<ydeb>hello, does Guix use the Kernel libre with no non-free firmware and no easy way of installing them?
<sneek>Welcome back ydeb, you have 1 message!
<sneek>ydeb, apteryx says: do you have an idea of how --check should be changed to make it more natural to use? One idea: if the package isn't in the store yet, fetch it as a substitutes or build it once, then proceed to build it a second time to validate the result.
<ydeb>(that wasn't me)
<kozo[m]><ydeb "hello, does Guix use the Kernel "> Correct
<roptat>sss2, I'd replace the configure phase by my own, I think
<sss2>roptat, where ?
<sss2>i thought here is function or option
<sss2>seen something like this in code i guess
<Noisytoot>Guix can use Linux-libre or GNU Hurd
<roptat>sss2, here's an example:
<sss2>roptat, ok, got it, but with this approach i will loose possible necessary options which may be adde in future
*abcdw recursevly removed $HOME instead of \$HOME *facepalm*
<lle-bout>abcdw: oh wow did you lose any work?
<abcdw>lle-bout: Nope, nothing important, few minor uncommitted changes, and very small patch for pam in guix.
<abcdw>40 minutes before on stream: Now, I don't have to worry about breaking my operating system, loosing all the time I spent configuring it ><
<apteryx>eh, the emacs-php-mode package installs itself to a lisp subdir, so is not autoloaded
<leoprikler>Fixed that on wip-emacs :)
<jackhill>leoprikler: does this seem like the correct behaviour for emacs-recmode? (I have verified that it continues to work the same way with wip-emacs)
<lle-bout>abcdw: :-S
<PotentialUser-30>Hello #guix! I'm trying to download the guix system ISO installer but it is very slow (30 KB/s) are there any alternative locations I can get it from?
<sneek>PotentialUser-30, you have 1 message!
<sneek>PotentialUser-30, nckx says: Good morning. You could try ‘guix gc --verify=contents,repair’ first.
<PotentialUser-30>I don't have guix installed, I'm trying to download the system installer iso (currently booting from a ubuntu live cd)
<lle-bout>PotentialUser-30: I think that message doesnt apply to you, since you were given a temporary username
<lle-bout>PotentialUser-30: we don't have mirrors for now
<PotentialUser-30>oh sorry about that. Ok I'll wait to see if the speed improves otherwise I'll try later
<lle-bout>PotentialUser-30: where are you located?
<lle-bout>PotentialUser-30: I see, server is in Europe so
<lfam>The ISO installer is hosted by the FSF
<lfam>I think it's in the northeast USA
<roptat>depends if it's stable or latest
<roptat>stable is on, latest is on berlin, in Europe
<lle-bout>if there's mirrors then
<PotentialUser-30>it is, please let me know where to find the mirrors
<GNUtoo>Hi, in regular distros when running "pacman-key --populate <keyring-name>", pacman-key looks into /usr/share/pacman/keyrings/ for the name of the keyring. There are several keyring packages (one for Parabola, Archlinux, Archlinux32, Archlinuxarm, Hyperbola, etc). How can I do that in guix? AFAIK pacman-key and each keyrings are supposed to be in separate packages.
<lle-bout>PotentialUser-30: Try
<lle-bout>PotentialUser-30: mirror list is at
<GNUtoo>I assume I can't simply install the keyrings in the same directory
<PotentialUser-30>Thank You lle-bout!
<lle-bout>PotentialUser-30: the one I gave you is Taiwan, should be closer
<roptat>GNUtoo, is there a way to give another directory to pacman-key?
<lfam>GNUtoo: What kind of keyrings are you talking about?
<lfam>Oh, I think I see: these are provided by Guix packages?
<GNUtoo>roptat: yes, I can patch it but I can give it a single directory
<lfam>For PotentialUser-30, they could have used <>. That should automatically choose the fastest mirror
<lle-bout>lfam: then we should use that link on the website probably
<lfam>GNUtoo: You could use (guix build union) to make a union directory of all the different keyrings
<GNUtoo>oh nice, thanks a lot
<lfam>I would check the sdl-union procedure in our packages for an example of how to use it
<lfam>Yeah, lle-bout, it's not a bad idea. I wonder if any other GNU projects use it like that, or if it's considered a bad idea or something
<lfam>Maybe we are the only ones not using it
<GNUtoo>lfam: with SDL how does it handle the fact that people might not install all the libraries? it depends on all of them?
<GNUtoo>(that would probably be good enough anyway in my case)
<leoprikler>GNUtoo sometimes sdl-union, otherwise you might be fine
<lfam>GNUtoo: I'm not sure I understand your question. When using the sdl-union procedure, you can choose exactly which libraries to put in the union
<abcdw>lle-bout: Sent the "guix home: Call for Early Adopters" announcement to rde-discuss and guix-devel MLs. AFAIR, you asked to ping you on that)
<lle-bout>abcdw: I don't remember but yay!!! :-)
<lle-bout>I saw it independently on my mail box, it's in the pipe
<lle-bout>abcdw: will make sure to watch your latest stream as well, I am not expert schemer so I did not understand various things in earlier streams but really like you make those!!
<GNUtoo>lfam: oh ok, that's how it works then, thanks.
<GNUtoo>my case is a bit different but I'll try to make it work in this way nevertheless
<GNUtoo>I'll just depend on all keyrings and it should work
<civodul>roptat: re JLL, i'm not sure what that means! we should ask nixo (Nicolò)
<lle-bout>abcdw: hey! I have a question, I currently use GNOME Evolution and with GNU Guix home probably I'll want to use isync/mbsync but I am concerned with the way the password is provided to isync/mbsync, what do you think about that? I would maybe some pinentry prompt every time that password needs to be used, also user services for regular synchronization? Does isync/mbsync support live sync?
<cbaines>I have mbsync run "pass ..." for the password, which prompts via pinentry
<Bumblehorse>Could someone give a brief ELI5 as to when you would want to write a service as a 'normal' service or a shepherd service?
<lle-bout>cbaines: okay! I use KeePass so :-S
<cbaines>lle-bout, if there's a command which will output the password to stdout, it should be possible to get mbsync to use it
<lfam>isync does IMAP sync. It's not "live"
<tricon>lle-bout: I used isync/mbsync with pass. Works well, so you should have no issue with a KP client that can output to STDOUT.
<lle-bout>lfam: so you need to manually trigger syncs even though IMAP supports notifications on new messages?
<lfam>And, I use a gpg / pinentry thing to handle the password
<lfam>lle-bout: I run it every 10 minutes with cron
<lfam>I guess I don't know if it can handle notifications. I've never noticed it doing that
<lle-bout>tricon: I've seen none that can invoke a pinentry but I can attempt writing some wrapper script for it I guess
<lfam>The Arch wiki has a way to combine isync with another program to deal with notifications
<lfam>Just in case isync wasn't complicated enough ;)
<lfam>For the password, this is the value of PassCmd in mbsyncrc: `gpg --quiet --for-your-eyes-only --no-tty -d path/to/encrypted/file`
<raghavgururajan>Hello Guix!
<lfam>With gpg-agent, it's convenient enough
<lle-bout>lfam: thanks
<lle-bout>lfam: so every 10 mins you enter your password?
<raghavgururajan>sneek, botsnack
<lle-bout>raghavgururajan: hello!!
<lfam>No, gpg-agent caches it for however I long I tell it to
<lfam>Every 10 minutes would be intolerable
<raghavgururajan>sneek didn't deliver my message to civodul
<lle-bout>lfam: how long? can other processes running as your user read gpg-agent's memory to steal it? I suppose yes, using ptrace attach
<lfam>lle-bout: However long you want
<lfam>I don't really consider my Linux system to be something that protects against itself
<lfam>Like, if there is a rogue process on the system, I don't expect anything to be protected against it
<lle-bout>What's the point of using pass or keepassxc if the password is always available in memory
<lfam>Because, like I said, typing in my password every 10 minutes would be intolerable
<lfam>I'm trading off security for convenience
<lle-bout>lfam: I mean compared to plain text files
<raghavgururajan>civodul: In c-u, glib-or-gtk-build-system doesn't trigger bootstrap. For example. try `guix build farstream --without-tests=gupnp-igd`. Any chance its related to your changes?
<lfam>Well, file exfiltration, even by accident, is way easier than reading memory
<lfam>This way, I don't have to worry about untrusted backups, for example
<lfam>I'm not claiming this makes sense
<lfam>It's just what I do
<lle-bout>I am so eager for a better system
<lfam>If I wanted really strong security, I would use a different operating system
<lfam>At least I might try using a smartcard or token
<lle-bout>Maybe IMAP supports login challenges with pubkeys
<lfam>I just think it all falls down if there is actually an attacker, so every layer of inconvenience is just a layer of inconvenience
<lfam>Not actually more security
<lfam>If I was starting again, I might use something besides PGP / GPG
<lfam>For password storage
<lfam>The agent is convenient, though
<lfam>And the least convenient thing is to change
<lle-bout>GNU Emacs would need some 'sandboxed/isolation' mode also
<lle-bout>GNU Emacs is a bit like X11 in that aspect
<lle-bout>(if I were to use GNU Emacs as my email client)
<abralek>lfam: I might missed something, but I use isync (mbsync) and bitwarden
<abralek>lfam: basically I am using PassCmd " get password ab3892cb-c337-48f2-b89b-a90a01206f98"
<abralek>lfam: I keep my master password in a gnome-keyring
<jgomo3>I'm using Keybase file storage and git repos for my secrets, including master password. That way I can access them from any of my devices using Keybase :D
<roptat>is there a reason why my I have two directories in $GUIX_LOCPATH? (2.29 and 2.31)
<roptat>oh, (gnu system locale) lists glibc and glibc-2.29
<roptat>"to ease transition", except we never updated that definition when we transitioned to 2.31 apparently
<jgomo3> "libc suffixes each entry of GUIX_LOCPATH with /X.Y, where X.Y is the libc version—e.g., 2.22. This means that, should your Guix profile contain a mixture of programs linked against different libc version, each libc version will only try to load locale data in the right format. " From the manual 2.6.1 Locales
<jgomo3>After creating a Profile, which is done simply by `guix package --profile=<the-path> ...`: What is the proper way to remove a Profile? Just by removing the directory?
<roptat>jgomo3, yes, that I knew, but I didn't understand why there were two libc versions, now I do, but I think it's a mistake :)
<lfam>roptat: You can remove the old one after all the profiles have been updated to the new libc
<lfam>Like, each user's profile, and the system profile
<roptat>jgomo3, yes, the profile is just a symlink to the store, and a gc root. If you remove the path, the gc root becomes invalid and next time you run gc, guix will remove the root automatically
<PotentialUser-32>Friends, I am a newcomer and know very little about computers. I have had Guix on my computer for several months, but without a lot of personal files, the computer memory (500 GB) is almost full! However, I worked with Debian for more than a year and I did not have this problem. This seems to be because of Guix updates, should I delete something af
<PotentialUser-32>ter update, or anything else?
<roptat>lfam, yes, but it's not like I did it myself, it's in (gnu system locale)
<roptat>%default-locale-libcs is set to (list glibc glib-2.29), which I think is wrong, since the transition would be between 2.30 and 2.31, right?
<lfam>Did Guix ever include 2.30?
<lfam>We should probably remove the workaround for the upcoming release
<roptat>ah, maybe not?
<PotentialUser-32>Any help?
<roptat>lfam, I see a commit from november 29: glibc: Update to 2.30
<jgomo3>PotentialUser-32: Maybe you could run the Garbage collector to free space: `guix gc`
<roptat>PotentialUser-32, usually I run the gc with a specified amount, to prevent it from collecting recent stuff that I might still need, like "guix gc -F50G" would limit it to collect just enough to have 50GB free on the disk
<roptat>every time you update guix, you can get a new version of files. Even though there is deduplication, it can take a lot of space on disk for things you don't use anymore (it's kept around as cache or as part of previous generations that are still available)
<jgomo3>PotentialUser-32: Also, I can imagine you having old generations or profiles pointing to software you don't use.
<abralek>PotentialUser-32: you can use --delete-generations where -d [duration]
<jgomo3>ProtentialUser-32: Until you remove those references, the GC will not remove those packages.
<PotentialUser-32>jgomo3: Yes!
<Noisytoot>roptat, I ran out of space on my QEMU image when installing kbd
<abralek>PotentialUser-32: info guix m Invoking guix gc
<Noisytoot>How much space is required to install it (with Hurd)?
<roptat>huh, not too much I believe, unless you have to rebuild the world
<PotentialUser-32>Thanks! jgomo3, roptat, abralek
<roptat>Noisytoot, guix size says 80MB (including glibc, etc), but that's not counting build dependencies, so if you don't have substitutes, that'll be hell
<jgomo3>PotentialUser-32: You'r welcome :D
<Noisytoot>roptat, It downloaded some stuff from, does that mean it has substitutes?
<roptat>sounds like it
<moorg>thanks again for guix!
<roptat>unless it only downloaded source archives :p
<lfam>roptat: I see. Well, we should revisit this locales thing now, before the release
<moorg>PS, the link to the cookbook on is broken
<moorg> seems to work for me though
<lfam>Thanks moorg
<lfam>We'll take a look at fixing it now
<roptat>that's an autogenerated link, right?
<roptat>right, the texi is "@xref{Top,,, guix-cookbook, The GNU Guix Cookbook}"
<lfam>How does this get transformed into a hyperlink?
*abralek Would it be cool to run guix commands at (guix-jupyter)
<abralek>meh, jupyter is not guix =(
<roptat>sneek, later tell lfam I think we have special logic for that going on in doc/build.scm
*roptat afk
<roptat>sneek, botsnack
<roptat>always give the bots snacks, in case the become sentient and remember who was nice to them and who was not :p
<Noisytoot>sneek, botsnack
<abralek>roptat: is that sneek bot from 8sync tutorial?
<roptat>no idea, I really have to go :)
<abralek>roptat: Au revoir =)
<pineapples>abcdw: Hello! Did you receive my private message? My IRC client indicates that no message has been sent, and I'm confused as to whether it's Freenode's or the client's issue..
<jgomo3>Is the following command an error in the manual section of Invoking guix environment?[1]
<jgomo3>`guix environment ...... -share=$HOME/.local/share/eolie/=$HOME/.local/share/eolie/ ....` ?
<jgomo3>shouldn't it be only one equals (=):
<sneek>lfam, you have 1 message!
<sneek>lfam, roptat says: I think we have special logic for that going on in doc/build.scm
<jgomo3>Like this `-share=$HOME/.local/share/eolie/` ?
<lfam>For containers, --expose (resp. --share) exposes the file system source from the host system as the read-only (resp. writable) file system target within the container. If target is not specified, source is used as the target mount point in the container.
<lfam>Your example could work, but the manual's example is correct, too
<jgomo3>Yes, there. You see, is like a typo. It is like the target was repeated twice by accident
<jgomo3>`--share=$HOME/.local/share/eolie/=$HOME/.local/share/eolie/` Instead of `--share=$HOME/.local/share/eolie/`
<lfam>It's not a typo
<jgomo3>Oh, I see. Thaks
<lfam>Did you read the manual section I linked to, specifically the part about "--share=source[=target]"?
<Noisytoot>What are the valid image types for guix system image
<jackhill>Noisytoot: does `guix system image --list-image-types` answer your question?
<Noisytoot>maybe it should be in the manual
<Noisytoot>What's the difference between hurd-qcow2 and qcow2?
<lfam>qcow2 is the standard format for QEMU, supported by everything, and the other one is something else
<lfam>You would only use hurd-qcow2 if you were building a Hurd image
<apteryx>Has qemu ARM emulation become slower recently? Just running the configure script is a slide show.
<apteryx>or perhaps it was always a slide show...
<apteryx>lfam: do you have an actual ARM machine? I'm trying to see if inkscape can actually run on ARM with the latest lib2geom. I doubt so, but I'd like to know.
<lfam>apteryx: No, but we can build on the overdrives by offloading from berlin
<Noisytoot>lfam, Will building a hurd image work with qcow2 also?
<lfam>It's also been very slow
<lfam>Noisytoot: Huh?
<apteryx>lfam: do you feel it regressed with 5.2.0 ?
<lfam>No apteryx. It was already terribly slow before that
<apteryx>OK :-/
<lfam>That's why the build farm couldn't keep up, for months
<lfam>apteryx: There is <>
<lfam>We should probably fix that
<lfam>Noisytoot: It's like I said, if you want to use Hurd, use the hurd-qcow2
<Noisytoot>Doesn't the operating system definition already specify the kernel?
<lfam>`guix system image` builds packages, and they need to be built for the Hurd instead of for Linux. So, hurd-qcow2 makes sure that happens
<apteryx>lfam: yeah, but that wouldn't affect user emulation performance though (it's 9p/file sharing related IIUC)
<lfam>There are other limitations of Hurd that hurd-qcow2 takes into account
<lfam>apteryx: Yeah, idk. I think it uses 9p in a lot of cases
<lfam>That's how files from the host store get into the VM, IIUC
<lfam>Well, I wasn't really paying attention when we started doing this ARM emulation, apteryx, so I don't really know how it works. But, it has already been incredibly slow
<lfam>I mean, always been
<lfam>Maybe it got worse for 5.2.0, I don't know. I was hoping it would speed up but it didn't
<apteryx>right! For the transparent (user) emulation though I think the builds happen in the normal Guix build container, without any 9p fancies. The only magic is the binfmt-misc registrations that tell the kernel that detect when, say, ARM binaries are executed, and launch them through QEMU, IIUC.
<lfam>That sounds like ./configure
<lfam>A lot of ARM binaries being launched over and over again
<apteryx>It's surprisingly simple; it even works for things like Docker now. E.g., if you have binfmt entries for armhf, you can do: docker run --rm arm32v7/alpine uname -a --> Linux e5f71906ae1d 5.11.11-gnu #1 SMP 1 armv7l Linux
<lfam>That's really nice
<civodul>raghavgururajan: i saw your message, i even investigated and found something... but i forgot what
<civodul>raghavgururajan: what was the package you mentioned back then?
<apteryx>from Emacs, C-h i m QEMU RET -> info-insert-file-contents: Can’t find or any compressed version of it
<apteryx>works from the command line; anyone else?
<apteryx>(the qemu package is installed to my user profile)
<apteryx>actually, it fails from the command line as well... hmm
<raghavgururajan>civodul: farstream
<raghavgururajan>civodul: Before that were talking about iqa, which was issue with the source url. This one is diff. YOu can try `guix build farstream --without-tests=gupnp-igd` on core-updates
<namtsui>is there a way to set linux kernel parameters in config.scm? I want to set `nosmt'
<roptat>namtsui, there's the kernel-arguments field in the operating-system declaration
<roptat>you probably want to extend existing values, so you'd add (kernel-arguments (cons "nosmt" %default-kernel-arguments)) to your os declaration
<roptat>does it make sense?
<namtsui>yes thank you
<civodul>raghavgururajan: so we'd need a fix similar to 600effef2b0b7a7e094c8009d8ed294e2145ad33, but in this case i think fixing it correctly will entail a full rebuild
<civodul>lemme see
<apteryx>I'm guessing because file such as '', '', etc. aren't considered for installation.
<leoprikler>jackhill: it may be, that wip-emacs inadvertently reproduces wrong behaviour from master
<civodul>apteryx: the dir entry of qemu looks bogus tho: "One line description of project."
<jackhill>leoprikler: the behaviour is indeed the smae as on master, but I don't know if the master behaviour is wrong or not.
<leoprikler>what's the actual package name?
<leoprikler>emacs-recutils got it
<jackhill>cool. My question is, should (require 'rec-mode) be nessisary.
<leoprikler>there seem to be no autoloads provided, so yeah, that's a bug in emacs-recutils, not in the guix packaging of it
<roptat>oh, I wanted to fix the locale issues people are experiencing when testing the website by providing the locales in the manifest, but I can't use computed-files in a manifest?
<efraim>lle-bout: if it's still relevant for mbsync I use Tunnel instead of PassCmd in my .mbsyncrc : Tunnel "ssh -o Compression=yes -q my-mailserver 'MAIL=maildir:~/Maildir exec /usr/lib/dovecot/imap'"
<leoprikler>you might want to report it upstream
*leoprikler → afk
<efraim>don't need to worry about too many IMAP connections and my ssh connection is cached
<efraim>and it seems I sync my mail every 45-60 seconds after it finishes syncing
<jackhill>leoprikler: thanks for checking. I agree it sounds like an upstream issue
***jess is now known as j
<civodul>roptat: you can use anything in a manifest
<civodul>see for ex. etc/release-manifest.scm
<djeis[m]>So I'm trying to write a package for a program that uses cmake and links against a particular version of LLVM and clang, but I keep getting "ld: cannot find -lgcc_s" when I try to build it. Am I missing something obvious? I've been tinkering with the package definition for about a day now but I admit I'm still a beginner at this.
<roptat>djeis[m], that doesn't look like a stupid mistake on your side at least :)
<roptat>doesn't sound easy to debug either :/
<djeis[m]>Yeah it's driving me up the wall. I'm mostly trying to avoid having to build my own copy of LLVM 8 and clang on this machine, and I figured it'd be a good excuse to learn guix packaging better.
<roptat>so, I'd check what cmake is trying to do, and if it tries to override LIBRARY_PATH or similar
<roptat>I suppose you didn't override the "gcc" input with llvm, right?
<roptat>so you actually have gcc (provided implicitely) in the build environment
<djeis[m]>I have not replaced the gcc input, no.
<djeis[m]>Just added an llvm input and a clang input at the corresponding versions. Also tried explicitly adding the gcc lib output to see if that would help, it has not.
<apteryx>civodul: I've learned something new: makeinfo splits info files in chunks of 300 KiB by default
<djeis[m]> is the most recent version of my tinkering.
<balance>Hello! Please help with next question: how to install gradle on guixSD?
<apteryx>so I needed specify --no-split to the makeinfo command in my custom QEMU patch
<apteryx>(that enables building/installing the QEMU info manual)
<apteryx>otherwise I'd have to install the 10s of .info-N files it generated along the main .info one
<roptat>djeis[m], sorry I don't know :/
<roptat>djeis[m], if you can't get help here, you can always try to send an email to
<djeis[m]>No worries, I'll keep tinkering. Might learn something eventually.
<roptat>balance, as long as it doesn't download anything other than jar files, with no native library, you can use the gradlew script provided with gradle packages. it's not packaged in guix
<roptat>if it downloads native libraries, it won't work and end with "no such file or directory" because it can't find libraries from the guix system
<balance>roptat: ok thank you for answer, main problem is in bulding "bisq" this is in the FSF high priority projects (second in list). Think that it is may be not so hard to build this project, i am in the ded end right now. Not familiar with java at all...
<apteryx>There's a ton of code in the 'roms' top directory of QEMU, containing things such as uboot, e2dk, and bundled libraries they need. We should probably strip it out.
<bdju>can someone remind me how to close a debbugs issue? something in the subject I think...
<bdju>aha, looks like I put -done after the issue number address in a CC maybe
<bdju>when I want to reply to an issue do I reply to a reply or have to make a new mail and cc any relevant people? is there a guide for this stuff somewhere?
<raghavgururajan>civodul: I see.
<efraim>there's the refcard but it assumes a fair bit of knowledge about debbugs
<apteryx>civodul: the QEMU texinfo manual should be fixed with 8515a506ca
<civodul>apteryx: neat, thanks!
<BlackMug>hi there
<BlackMug>(bootloader grub-bootloader) (target "/dev/sdX")))
<BlackMug>i used efi legacy installation method
<BlackMug>now to define the target in guix in this case it should be where?
<BlackMug>roptat yesterday i said /etc/resolv.conf directory wasnt there = wrong it is there
<BlackMug>according to this /boot/efi is what should be for legacy efi
<BlackMug>but i cant seem to success make guix installer passing that
<roptat>BlackMug, for efi you need /boot/efi, for legacy bios you don't need that
<BlackMug>one sec i will show you my config
<roptat>on your VM, you should have a disk, whether /dev/sda, or /dev/vda or similar, that's the target
*roptat is having fun filling his taxes while debugging other people's problems ^^
<roptat>should have done that earlier
<BlackMug>this is the guix config
<BlackMug>this is my VM partition
<roptat>BlackMug, I think the target should be the whole disk, so /dev/xvda1
<roptat>er, /dev/xvda
<roptat>not 1
<BlackMug>done that and gave me error
<BlackMug>anything else im missing in that config ?
<roptat>you declared /boot/efi to be vfat, but the uuid is for ext4
<roptat>I thought you didn't have a luks partition?
<BlackMug>yes i deleted that guix vm
<BlackMug>started freshly
<BlackMug>guix config coming with many fill in the blank places
<BlackMug>sadly they are not blank but you need to figure where to change
<apteryx>sneek: later tell lfam I just sent a patch addressing the QEMU 9p performance warning you reported
<sneek>Got it.
<roptat>BlackMug, yeah... so that /boot/efi partition, does it exist, and is it vfat (as it should be) or ext4?
<roptat>did you manually create the luks partition then?
<roptat>and finally, what's the error you get?
<BlackMug>roptat /boot/efi i created with mkdir
<BlackMug>only /grub was there in /boot
<roptat>so, remove the partition you declared for it
<roptat>you didn't actually create a partition that should be mounted there, right?
<BlackMug>one sec let me check
<roptat>and since you used msdos partitionning instead of gpt, you need to use the grub-bootloader, not the grub-efi-bootloader
<roptat>(make sure the VM boots in legacy mode)
<BlackMug>roptat yeah forgot to say i actually used today according to the wiki so i used gpt
<BlackMug>not msdos
<BlackMug>if you look at the image i sent to you its very clear how my hard is partitioned
<roptat>oh ok, then you need to create a dedicate partition (say /dev/xvda2) for the efi, with a fat filesystem
<roptat>oh sorry, I saw only one link
<roptat>so xvda3
<BlackMug>oh ok
<BlackMug>and that vfat i keep it right?
<roptat>you'll have to replace the device field for the efi partition: it should be (device (uuid "697B-B26C"))
<roptat>keep vfat, it is correct
<BlackMug>oh then its not ext4
<BlackMug>change it to FAT32 ?
<BlackMug>you are referring to device field in file-systems?
<BlackMug>or where?
<roptat>the device field
<roptat>(file-system (device (uuid "697B-
<roptat>(file-system (device (uuid "697B-B26C")) (mount-point "/boot/efi") (type "vfat"))
<roptat>the first one looks correct