<cnmne>ngz: I think you were logged out, by the time I figured it out, but I ended up finding my error. Turns out I didn't export anything with define-public. feeling pretty silly. The funny thing is I definitely knew I needed it, but got turned around after using `guix import gem ... >> gnu/packages/ruby.scm' and tricked myself into not needing it. woops ! <shtwzrd>This simple service declaration causes xf86-video-intel to be pulled in (which I can't build on this machine -- perhaps an issue itself?), but I don't understand why -- I'm explicitly excluding x86-video-intel module from xorg <shtwzrd>it's a little long. but deleting that service causes it to no longer pull in xf86-video-intel <shtwzrd>drakonis1: hm, you're right. I thought xorg would still get required due to xfce, xlock, etc but actually it's not in the list of things to download or build. <shtwzrd>but, how does xorg end up pulling in xf86-video-intel? No packages take that as an input, whatsoever. The only reference it has in all of guix is %default-xorg-modules, and I'm explicitly overriding that list to exclude it. :/ <drakonis1>but right now, do you have to reduce the amount of modules? <drakonis1>its probably not the right solution as the pinebook doesn't have a lot of storage space <shtwzrd>no, I'm just trying to get rid of xf86-video-intel because I can't get a substitute for it, and I can't build it either :( <shtwzrd>maybe I should try to fix the issue with it not building. I just figured skip it altogether since it's technically not needed but <shtwzrd>xf86-video-intel dies with: checking which acceleration method to use by default... configure: error: UXA requested as default, but is not enabled <shtwzrd>but a few lines before that I can see: checking whether to include UXA support... no, but the command invoked to build the thing in the first place has the "--with-default-accel=uxa" flag <drakonis1>the folks with more experience might be able to <shtwzrd>package def for xf86-video-intel actually has a supported-systems field in it, which does not include any non i686 or x64 systems. So perhaps it's actually a bug that xorg even pulls this in when arch=aarch64 <shtwzrd>and no problem of course :) Thank you for your help, I hadn't even realized that my config didn't pull xorg at all so you've already helped me out :) <drakonis1>it wouldn't if you were to remove set-xorg-configuration <shtwzrd>yeah exactly, I had noticed that removing that service meant no more building xf86-video-intel, but hadn't realized it also meant no more xorg, until you told me haha <drakonis1>then get back to learning how to do advanced things <pkill9>can a service add a kernel argument? <shtwzrd>pkill9: I would think not. Services don't run until the kernel has already booted. <pkill9>shtwzrd: i mean in the config file <pkill9>as in, can you add a service that add a kernel argument to the OS config <shtwzrd>pkill9: the operating-system declaration takes a kernel-arguments field though <pkill9>like how some services add packages <shtwzrd>oh, you want to dynamically add kernel arguments to the config based on the included services? I don't know if that can be done, sorry <raingloom>hmm, gnome.scm is pretty huge. should i add Geary to it or should i split it off into something else for compile time's sake? <pkill9>i would just put it in gnome.scm, and let it be split up later <bandali>see which one has similar apps i guess <raingloom>evolution is in gnome.scm, so i figured it should also go there ***cualquiercosa is now known as Telior
<cnmne>hello guix! I get an error when running `guix import gem unicode_utils' for an empty vector. I'm not where to even look for the error there. I thought it might be the underscore, but that seems silly and anyway it works with `guix import gem unicode-display_width'. I've tried quoting and that's about all I know to try. <efraim>Actually I checked my config and fold delete doesn't work, I assume I tried alist-delete too <apteryx>cnmne: I've never looked at the gem importer, but potentially it's an issue where it's looking for some field in a file and doesn't find it, and doesn't guard against this case. Perhaps it can't be imported for some reason; if so, it should fail gracefully. <shtwzrd>efraim: thanks :) I actually am specifying it explicitly though (no alist-delete or similar), I just supply the whole list of modules I want, without xf86-video-intel in it. But it still tries to build <shtwzrd>My guess is that the modules form only impacts what modules get loaded in the xorg configuration, not which ones get built. I've written a bug report for it <valignatev>Hey guix, is there a livestream of fosdem2020 anywhere? <sneek>civodul, you have 1 message. <sneek>civodul, vagrantc says: Thanks for 2032d8473d11711b88fd3e48644c569dee32fa42 ci: Cross-build for riscv64-linux-gnu. Very exciting! <civodul>comes from a discussion with dongcarl <efraim>No promises, I might try to write up my rust conversations while on the plane <leoprikler>sneek later tell raingloom gnome.scm is fine, but raghav-gururajan is looking to split that up into multiple parts later on <g_bor[m]>I remember there was some dicussion that it be not needed to return #t from the phases. <civodul>we should check whether there's still code that relies on it <civodul>actually Mark H Weaver was initially taking care of that transition, i'm not sure what the exact plan was <efraim>Last night I hacked together a %vim-install-phase but I wasn't happy with it <efraim>Too many differently named directories for me to actually like it <efraim>I did like the implementation overall though. New build module misc-build-phases to hold a collection of single phases to be added in *efraim boards flight in about a half hour <efraim>I guess I could look at the fonts and see if it's worth creating just an install phase that can be dropped into most of them <leoprikler>Was it not, that it used to be you could return anything, but now it's required to be #t? <ng0>with sleep and people, fosdem is actually enjoyable. not like what I did in 2016 :D <ng0>the mes talk and where it's heading is fascinating <divansantana>leoprikler: my timezone is set to as per (timezone "Africa/Johannesburg"). But qutebrowser and chrome still show GMT. epiphany is correct. I think it's guix specific packaging but I'm not sure. <bdju>I know icecat's resist fingerprinting setting messes up the timezone, but I'm not sure about the other browsers. <efraim>I missed the mes talk, I'll have to grab the recording ***ng0_ is now known as ng0
<jonsger>a "rm -rf po/guix" and then "git restore -- po" helped <str1ngs>divansantana: do you have a simple website I can test that on? <str1ngs>divansantana: nvm I see the JS code now <str1ngs>divansantana: this returns the right timezone for me <str1ngs>I tested using qutebrowser and alert(new Date().toString()) <divansantana>str1ngs:wow. That's interesting. Latest qutebrowser from guix I guess? on guix system? My qutebrowser is stored in a profile under ~/.guix-extra-profiles/ <str1ngs>divansantana: yes, I'm using guix. I packages qtwebengine and updated qutebrowser. so I want to make sure there is not an issue with that. <divansantana>str1ngs: thanks hugely for packaging it. I really appreciate that. <str1ngs>divansantana: what is the output of $ date <divansantana>let me clear .config/qutebrowser and .cache/qutebrowser and try again. <apteryx>divansantana: At least in Icecat, the timezone is set to GMT as a privacy measure, caused by the ResistFingerprinting setting. <apteryx>perhaps the same is true for other browsers, I only use Icecat. <str1ngs>09:19:02 INFO: Mon Feb 03 2020 09:19:02 GMT-0800 (Pacific Standard Time) <str1ngs>divansantana: if you start qutebrowser from a terminal and use new Date().toString() with jseval it will output to console <str1ngs>maybe if there is a localization issue you will see a warning or error aswell <divansantana>str1ngs: tried jseval alert(new Date().toString()) which I only get 19:24:29 INFO: No output or error on cli <divansantana>or new Date().toString() which I get 19:24:47 INFO: Mon Feb 03 2020 17:24:47 GMT+0000 (GMT) <divansantana>doing the same with qutebrowser -d -l vdebug I doesn't help. <str1ngs>divansantana: it's odd but not apparent why that does not work for you. <divansantana>perhaps I have to run some command within it's python environment. <divansantana>str1ngs:I thought it was a packaging issue, but perhaps not if it works for you. Let me ask on #qutebrowser <str1ngs>divansantana: it's potentially a packaging issue with qtwebengine. <divansantana>str1ngs:I do get the same behaviour on chrome via a guix package. And that is not default on chrome. <NieDzejkob>divansantana: is this guix system or a foreign distro? <str1ngs>divansantana: can you clone this repo https://github.com/mrosset/testqt . and then in that directory do. guix environment --ad-hoc qtbase qtwebengine qtwebchannel qtdeclarative gcc-toolchain then qmake && make && ./main <divansantana>NieDzejkob:guix system. echo $TZDIR says /gnu/store/9mmsilz9avdl49i6a6nj5mzfyim8ihv2-tzdata-2019c/share/zoneinfo <NieDzejkob>are you starting your browsers from this shell? If not, can you try doing that? <divansantana>str1ngs:done. Launches guessing qtengine. what do I do then. Not sure how to navigate. <divansantana>str1ngs:it returns QDateTime(2020-02-03 20:01:13.463 SAST Qt::LocalTime) <str1ngs>divansantana: try now with git pull && make && ./main . make sure you are still in the environment <str1ngs>divansantana: I added a javascript call to date <divansantana>str1ngs:first row QDateTime(2020-02-03 20:30:39.450 SAST Qt::LocalTime) second "Mon Feb 03 2020 18:30:39 GMT+0000 (GMT)" <str1ngs>divansantana: second row seems wrong. so not a qutebrowser issue <str1ngs>also for me I get "Mon Feb 03 2020 10:44:14 GMT-0800 (Pacific Standard Time)" which is right <str1ngs>divansantana: I don't know why mine works and yours does not. <str1ngs>divansantana and QDateTime works as well <str1ngs>it's something qtwebenine or javascript related <jonsger>is it intended that guix.gnu.org doesn't redirect to https by default? <bandali>but afaik gnu.org uses a header of some sort such that once you visit over https, subsequent requests over http will be redirected to https <kkebreau>Is anyone here able to build the x86_64 version of LLVM 3.9.1 on the core-updates branch? <kkebreau>Never mind, there's a commit that looks like it addresses this issue. *civodul just ran: guix package -r $(guix package -I ^guile |cut -f1) -i guile-next{,:debug} $(guix package -I ^guile-| cut -f1| sed '-es/^guile-/guile3.0-/g') <civodul>ArneBab: yeah i had already been using Guile 3 when adapting Guix, but it's like i'm officially switching all my stuff to 3.0 now :-) <ArneBab>that’s really cool! I already rejoiced when I saw guile3.0-wisp in the tree :-) (thank you!) *jackhill sees that there was a mgmt config talk in the minimalisted languages at FOSDEM, and wonders how its ideas could influence/combine with guix deploy <civodul>jackhill: the talk there is mostly about their DSL, which is interesting in many ways, but Guix is not looking for a DSL ;-) <civodul>the tool itself is an imperative deployment tool actually, AIUI <vagrantc>how would one go about getting a list of all guix packages sorted in the order they need to be built ? <vagrantc>don't need it to be exact, but the packages with the fewest dependencies first would be ideal <civodul>internally the daemon does the right thing <civodul>so it's not needed if you go through it <jackhill>civodul: thanks. I guess I should watch the video :) <civodul>yeah, so i guess i'm not really answering your question? :-) <jfred>is there something I'm missing? <vagrantc>basically, i want to rebuild the world without substitutes and run guix challenge on the results ... i could just iterate alphabetically and then some builds with already be done by the time you get to them <civodul>"guix challenge" without any arguments will traverse the store, if that's of any help <jfred>aha, so it's not just me then! <vagrantc>civodul: is there a way to "guix challenge --build" if no local build... ideally only if the remote has a build? <vagrantc>then it could just be one command iterating through all the packages <vagrantc>it'd be a waste of build time if there's nothing remote to challenge, for example <civodul>vagrantc: there's no such option but it'd be worth considering, i agree <civodul>well, "guix challenge" was initially thought as a way to challenge servers based on what's available <civodul>i had never thought of also having it trigger builds <vagrantc>are there any stats on how long substitutes which aren't "current" stick around for? <vagrantc>i know there was some process to try and keep substitutes for released versions indefinitely? <vagrantc>are the patches in gnu/packages/patches needed for a build from git? <jfred>civodul: ah, figured it out - "herd doc udev list-actions" <vagrantc>becuase other than that issue around unclear licensing, i think my debian packages are ready to upload <vagrantc>civodul: now that i think about it ... guix build --no-substitutes && guix challenge is probably just fine <ArneBab>I cannot install spatialite-gui, because wxwidgets@2.8.12 fails to build. <vagrantc>or rather: guix build --no-substitutes PACKAGE && guix challenge PACKAGE <vagrantc>is there any conception of what a "base system" is? <vagrantc>apparently it only operates on packages ... maybe that's a more interesting bug report *vagrantc chews on these feature requests a little more before submitting bugs <pkill9>are there any VPS providers that will deploy a guix system based on a guix config? <vagrantc>ok, for a "base system" i can build a system image and then parse the installed packages from guix challenge. close enough. <vagrantc>this can be more meaningful sets of packages to target. :) <vagrantc>bare-bones.tmpl desktop.tmpl lightweight-desktop.tmpl will be interesting package sets to start with <vagrantc>does guix challenge issue a different return code for inconclusive vs. differed?