# IRC channel logs

## 2020-08-16.log

back to list of logs

<apteryx>nckx: I could successfully build webkitgtk using with the added 6 GiB of ZRAM (zstd), it took 108 minutes on that machine.
<nckx>apteryx: Nice!
<nckx>What a spectacular difference.
<apteryx>(yes; but note that it's not 108 minutes vs 15+ hours -- that was for ungoogled-chromium)
<apteryx>I'll reattempt ungoogled-chromium from now
<mroh>good luck!
<drakonis>is there anything to create regular containers without the store?
<apteryx>mbakke: ah, that's a shame, the nouveau driver mostly requires binary nvidia firmware to provide accelerated video decoding: https://nouveau.freedesktop.org/wiki/VideoAcceleration/#firmware
<drakonis>and when did ld-wrapper get added?
<apteryx>hmm, evolution-data-server test failed: 26 - test-book-client-custom-summary (SIGTRAP)
*apteryx rebases on the last couple hundreds of commits
<apteryx>arf... triggered a webkitgtk rebuild.
<brettgilio>Anybody here using virt-manager and libvirt successfully on guix system?
<brettgilio>I'm having a difficult time
<drakonis>what kind of issue are you having?
<drakonis>and do you have the libvirt daemon running?
<brettgilio>I'm starting libvirt with services field in my config.
<brettgilio>Let me paste the error
<brettgilio>This error starts on launch
<brettgilio>It's some dbus error
<brettgilio>But idk much of what to make of it
<drakonis>its not finding the socket lol
<brettgilio>It says the path doesn't exist. But it's there
<drakonis>does libvirt have access to the socket?
<drakonis>looks like it is a known bug?
<drakonis>do you have dbus running?
<drakonis>brettgilio: run as root
<brettgilio>drakonis: thing is, running it as root is not how it's done on any other distro.
<drakonis>the other distros use polkit to avoid it
<brettgilio>I see
<drakonis>it calls to polkit when it needs to access the socket
<drakonis>the socket is owned by root
<drakonis>is there anything i can use to run system containers a la lxc?
<drakonis>without mounting the store, that is?
<drakonis>hmm, perhaps i should write a shepherd service for that
<drakonis>as i would really like to run a few services inside containers running systemd
<pkill9>brettgilio: I have this in my system config
<pkill9>(service libvirt-service-type ;; For virtualisation using Qemu.
<pkill9> (libvirt-configuration
<pkill9> (unix-sock-group "libvirt")))
<brettgilio>Same
<brendyyn>Is there any node/npm importer for javascript?
<brettgilio>brendyyn: not yet
<brendyyn>They new mathjax 3 seems to require a lot of packages, and if tests are enabled on all the dependencies of dependencies it explodes out into even more packages
<brendyyn>so im trying to learn about javascript packaging
<brendyyn>currently when i make node-build-system package it makes essentially the whole git repo the package, and there is no minification. is javascript meant to be compiled or minified ?
<brendyyn>having a brief look at an arch linux package it seems it does get minified
<brendyyn>one useful thing is that packages sometimes have a package with the true source and another for the minified "source" that is distributed
<brendyyn>it is similar to rust in that there is a package.json with all the dependencies listed. it would be nice if those could be automatically parsed for the guix package
*apteryx just found a way to automatically load debug files for any profile, through .gdbinit
<brettgilio>pkill9: are you getting host to guest networking?
<brettgilio>If so, ho
<brettgilio>How
<pkill9>i don't know, i haven't used it in a while
<brettgilio>pkill9: I found your gitlab. I'll compare
<raghavgururajan>Test!
<raghavgururajan>nckx: You around?
<bqv>drakonis: thought you were full time nixos nowí
<bqv>*?
***terpri_ is now known as terpri
***apteryx_ is now known as apteryx
<Kimapr[m]>am i correct that if i somehow get my store corrupted i should not guix init from that guix system and expect store of new system to not be corrupted too?
<efraim>Kimapr[m]: I'd go with 'sudo guix gc --verify=repair,contents' to see if that can fix it
<Kimapr[m]>hmm thanks didn't know about that
***rEnr3n8 is now known as rEnr3n
<Kimapr[m]>is there any reason why the check freezes at "checking hashes..." and gives no feedback?
<efraim>it's checking, it can take a while
<Kimapr[m]>okay it found some errors
<fnstudio>hi, when adding my personal channel https://example.org/personal-packages.git, is it possible to add some form of authentication so that the git repo is not publicly accessible?
<nckx>fnstudio: Yes! It requires ssh-agent to be installed and doesn't seem to be documented, but fetching channels over SSH is supported.
<fnstudio>nckx: that
<fnstudio>nckx: that's great, thanks so much!
<bluekeys>I've installed guix onto a debian vm (hosted by xen on qubesos) using the script on the site. The hello package is installed successfully. Step 8 states: "The binary installation tarball can be (re)produced and verified simply by running the following command in the Guix source tree". Where is the source tree?
<sneek>Welcome back bluekeys, you have 2 messages!
<sneek>bluekeys, nckx says: You're passing a nested list (nss stump (%base) as packages. %base-packages is already a list, so move it one line down outside of the (list …) call.
<sneek>bluekeys, nckx says: Typo, forgot a ) in my first sentence.
***jonsger1 is now known as jonsger
<nckx>That smells ooold.
<nckx>This is also what guix pull pulls.
<nckx>Faithful sneek: botsnack.
***jonsger1 is now known as jonsger
<bluekeys>nckx what smells old?
<nckx>My message.
<bluekeys>oh, you can see that... I thought only I could. It's been a while since I logged on
<bluekeys>Is there a way I can update the documentation to include a hyperlink to the git repo?
<nckx>You can edit doc/guix.texi and send as many patches as you want, but the manual's full of that URL already. A ‘Texinfo link’ (whatever those are called) to ‘Building from Git’ would be better.
*nckx AFK.
<johnmd>Blender 2.79b does not have Filmic Blender color management (https://www.blender.org/download/releases/2-79/#filmic-color-management) installed (wrong build option?).
<leoprikler>apteryx: "Webkitgtk, still smaller than an embedded OS" would make for a nice slogan, don't you think?
<nckx>sneek: later tell johnmd: Filmic Blender isn't part of Blender. A package for it would be very welcome!
<sneek>Got it.
<nckx>leoprikler: [citation needed]
<nckx>* for selected embedded OSes, possibly by the same folks who brought you WebKit, restrictions apply.
<johnmd>Filmic Blender is actually part of Blender since version 2.79.
<sneek>Welcome back johnmd, you have 1 message!
<sneek>johnmd, nckx says: Filmic Blender isn't part of Blender. A package for it would be very welcome!
<leoprikler>I goofed anyway. I think apteryx noted an embedded OS taking 1h30m (i.e. 90m), not 130m, tehepero
***jonsger1 is now known as jonsger
<johnmd>The version of Blender compiled by Guix doesn't have a colormanagement folder. I get this error: "Color management: using fallback mode for management" and I don't see Filmic Blender in the settings (unlike Blender on Ubuntu, Arch, Windows, Mac, etc.).
<johnmd>Also, Filmic Blender is included in the settings of Blender 2.8x on Guix.
<johnmd>but not in 2.79b.
<Kimapr[m]>Why does any download guix does freeze after sleep-wakeup? Shouldn't it fail or continue instead?
<nckx>I'm downloading the latest blender but it's currently at 1.5% :-/
<nckx>Kimapr[m]: Guix handles flaky network connections very badly in general. Yes it should.
<nckx>johnmd: Thanks for the report. I'll look into this later if nobody beats me to it. Really have to go now.
<terpri>johnmd, i see the same thing, but what is it supposed to be using for color management? (e.g. what would typical output look like)
<johnmd>it's using files in the colormanagement folder (which is currently missing in Guix's blender@2.79b). Usually (other GNU+Linux distros), this folder is located at /usr/share/blender/2.79/datafiles/colormanagement.
<terpri>hm, yeah, with 2.83.3 there's no colormanagement folder under datafiles, odd
<Kimapr[m]>does guix use curl, wget or something else for downloading?
<fnstudio>now, this may very well be OT, but i'm familiarising with the packaging process and was pondering what the options are in terms of uploading a signed tarball to gitlab
<terpri>CMakeLists.txt mentions "OpenColorIO" which could be required for that to work properly, maybe
<fnstudio>wait, forget that, i might be mixing things up
***jonsger1 is now known as jonsger
<terpri>lol opencolorio requires git to build...not to download a repo but to apply a diff
<atw>when attempting to sudo guix pull I'm encountering an odd error: "guix pull: error: Git error: could not open '/root/.cache/guix/checkouts/pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq/nix/.gitignore' for writing: Structure needs cleaning". What can I do about that?
<leoprikler>why do you want to "sudo guix pull" in the first place?
<terpri>atw, that seems to be a filesystem error rather that a git-specific error...fsck?
<terpri>failing that, nuke /root/.cache/guix maybe (*hopefully* it's only a cache...)
<terpri>it seems to happen with multiple FSes (xfs, ext4 at least)
<fnstudio>is guix download supposed to be used against the tarball of the final binary? i suppose so?
<atw>leoprikler: to, uh, update the root user's guix? should I be doing that?
<leoprikler>depends
<leoprikler>are you on a foreign distro?
<atw>no
<leoprikler>then you don't need "sudo guix pull"
<fnstudio>in particular guix download doesn't seem to support ssh-agent based connections?
<terpri>fnstudio, i don't understand what you want to use "guix download" for. typically, it's used to get the hash of a *source* tarball, to help with writing package definitions
<fnstudio>terpri: yep, but suppose the final source tarball is not publicly available, eg when it is kept as part of a private repository
<terpri>fnstudio, then the simplest method would be to get a local copy via scp or something, and use file:/// uris for package definitions
<fnstudio>i understand guix provide some support for downloading from "private" channels but that mechanism doesn't seem to be available for guix download (i'm experimenting with this for the first time, i may very well be wrong)
<fnstudio>terpri: ok, cool, i'll try that - thanks!
<atw>huh, thanks leoprikler, didn't know that
<terpri>well, guix download takes a plain URI as an argument, so the availability of extra channels doesn't really change anything
<terpri>there is also "guix hash", to get the base32 hash of a local file/directory
<leoprikler>for clarity: the guix binary, that you use (even in guix system) will not be root's, but your own
<fnstudio>terpri: guix hash might actually be what i was looking for! thanks
<terpri>np
***jonsger1 is now known as jonsger
<raghavgururajan>nckx: Is it possible to dependency between service-types? Like, foo-service-type and bar-service-type are two different service, but adding the bar-service-type will also add/includes foo-service-type.
<kkebreau>mbakke: In commit a749caa74e2, you removed git as an input to the elixir package. Did it just not show up in the requisites of elixir? Because one of the phases looks for git... :-/
<kkebreau>Perhaps git-minimal is enough? I'll find out once I'm done building these wxwidgets dependents.
*terpri compiles blender with OpenColorIO as an input
<terpri>maybe this will fix the colorspace problem, maybe not
<terpri>...as a native-input rather
***jonsger1 is now known as jonsger
<terpri>"WITH_OPENIMAGEIO ON" seems like a good sign
<raghavgururajan>kkebreau‎: Thanks for your patch for MozJS-68 on wip-gnome. I cherry-picked it for my work. :-)
<kkebreau>raghavgururajan: You're welcome! Your work on GNOME is looking really good; please keep it up! :-)
<leoprikler>raghavgururajan: sounds like you want to extend that other service type?
<raghavgururajan>kkebreau, Thanks!
<leoprikler>there's no dependency resolution, though
<raghavgururajan>leoprikler, Wouldn't extending a service require further arguments though?
<efraim>Is there a way to run mysqld and put it in the backgroud? Normally I'd skip the test but it's easier to debug the test than to debug the service which uses the package
<raghavgururajan>leoprikler, For example, network-manager-service-type will not work without wpa-supplicant-service-type. In that case, I would like to make the former to include the latter.
<leoprikler>if it's super mega hyper important, then you'd make a list as with %desktop-services
<leoprikler>chances are it's not tho
<raghavgururajan>Technically, wireless won't work.
<raghavgururajan>I see.
<efraim>you can make one depend on the other if that's as much as it's required
<efraim>but for a relatively simple example we sometimes extend the root-profile to add a package used in a service
*efraim leans heavily on "relatively"
<raghavgururajan>> efraim‎: you can make one depend on the other if that's as much as it's required
<raghavgururajan>Yes, that's exactly i am looking for. How do i do that.
<raghavgururajan>Is it how it's done in network-manager and wpa-supplicant services?
<efraim>the keyword to serach for is 'requrement'
<efraim>looks like it
<raghavgururajan>Cool!
<efraim>i never realized how complex the network-manager service is
<efraim>the network-manager shepherd service has a requirement on '(user-processes dbus-system wpa-supplicant loopback), and the network-manager-service-type extends several service types, most importantly the shepherd-root-service-type where it registers itself as a service
*efraim can't help but touch all the files
<apteryx>git send-email question: what am I supposed to put as the value of --in-reply-to= ?
<apteryx>I tried the Message-Id of the emali, but that doesn't work
<NieDzejkob>my bash history seems to indicate that it did work for me when I included the <> in the argument
<apteryx>ah, it does. Not sure why I thought it didn't. Thanks :-)
<terpri>i fixed the blender bug! i don't understand quite what the problem was, but i fixed it nonetheless ;)
<terpri>rip paste.debian.net
<leoprikler>okay, so for version you should probably use 1.1.1 ;)
<terpri>haha, yeah
<alextee[m]>hmm guix upgrade1 hangs
<alextee[m]> * hmm guix upgrade hangs
<terpri>this OCIO thing has...a lot of licenses involved
<leoprikler>wait, since when do we have a firefox package?
<nckx>Hullo Guixagain. o/
*nckx just spent a train ride packaging opencolorio 1.1 (for Filmic Blender); reconnects to the internet; turns out there's a 2.x now.
<leoprikler>it do be like that
<nckx>Or at least it's confusing.
<nckx>‘Core Library v1.1.1’ might be part of 2.x.
<nckx>Just to fuck with us.
<nckx>Hm.
<nckx>leoprikler: Firefox? We... don't?
<mroh>leoprikler: take a look at the location (where its comming from) ;)
<leoprikler>Oops, it snuck in through the channel, that must not be named.
<nckx>terpri: Oh no ☹
<nckx>Dupe work.
<terpri>:D
<terpri>nckx, by all means, feel free to continue
<terpri>i haven't even worked out how many licences apply to this thing
<nckx>terpri: Here's mine: https://paste.gnome.org/pcjzwk517
<nckx>I don't think all the licences you may have found actually apply.
<nckx>But feel free to fact-check that assertion.
<nckx>Is yours better? Mine won't even pass the test suite.
<terpri>yeah yours is better
<terpri>mine doesn't pass the test suite either
<terpri>unbundling also reduces the licensing tangle
<nckx>TBH I'm just ignoring those for now, hoping they can all be cleanly removed once it actually builds.
<terpri>yeah
<nckx>terpri: Mind sharing yours?
<terpri>sure, it's linked somewhere
<nckx>Oh, that's still recent? OK.
<nckx>Speaking of which: is dpaste Tor-friendly? p.deb.net has been down for a while.
<nckx>I think our wgetpaste package uses dpaste.net by default so that would be a nice choice.
<nckx>Why a native-input, terpri?
<terpri>dunno, i didn't try it over tor
<nckx>Maybe a friendly Tor user will help out 🙂
<terpri>good question :p git because it's used...instead of patch for some reason??...at build-time
<nckx>My bad, I meant the blender hunk.
<nckx>Indeed git should almost always be native.
<terpri>oh right. no good reason
<terpri>i scribbled it in like 15 minutes just to make the colorspace feature work properly (or...at least fail less obviously, as i don't know enough about blender to test it)
<terpri>your patch is better, i'd just submit/commit that one
***jonsger1 is now known as jonsger
<nckx>terpri: But yours actually builds!? Because mine doesn't.
<terpri>yours doesn't build?
<nckx>I've commented out the OpenGL stuff for now.
*terpri checks
<nckx>I suspect that will take care of the build failure and I'll be looking at test failures next.
<nckx>But if you actually verified that all this work actually makes the buttons go work, that's valuable (and motivating...), thanks.
<terpri>"PFNGLGETFRAGMENTMATERIALIVSGIXPROC" that's quite a function name
<terpri>i feel like i'm looking at a roman building inscription or something
<nckx>That's the one.
<nckx>Weird, has my paste above already been memory-holed? I'd set it to stay up for a month. Useless if true.
<leoprikler>meh, it's lacking polish notation
<terpri>needs more interpuncts
<leoprikler>no, that might make it readable
<terpri>nckx, i'm not particularly familiar with mesa but it seems like it might just be duplicating includes? like maybe it's not supposed to #include <GL/gl.h> when alse #include'ing <GL/glew.h> or something
<terpri>also*
<terpri>"As with most other loaders, you should not include gl.h, glext.h, or any other gl related header file before glew.h, otherwise you'll get an error message that you have included gl.h before glew.h. In fact, you shouldn't be including gl.h at all; glew.h replaces it."
<Formbi>leoprikler: Polish notation?
<leoprikler>oops hungarian notation :(
<terpri>so it seems to be misusing glew
<leoprikler>Hmm, I wonder
<leoprikler>is there a language, that includes the return type of a function in its signature?
<leoprikler>C++ and Java don't
<nckx>terpri: That's plausible (also not a GL nerd).
<nckx>terpri: But wouldn't that mean it breaks everywhere?
<nckx>Maybe some FHS coincidence/search order avoids it I guess.
<nckx>Guix is my favourite test suite. OpenColorIO's is a close second because it just passed without issue \o/
<terpri>\o/
<terpri>nckx, do you know how to check that your changes fixed the blender bug?
<terpri>if not i can make a gif to demonstrate the test
<nckx>terpri: That would be very welcome.
*nckx 's ‘plan’ was just poking around the old & new Blender looking for changes.
<NieDzejkob>leoprikler: as in, in mangling?
<leoprikler>well, mangling is a part of it
<NieDzejkob>I don't see what you mean, then. The return type is a part of the signature syntax-wise?
<NieDzejkob>(in C++ and Java)
<leoprikler>It is not in terms of semantics.
<leoprikler>You can't have an int foo(); and a size_t foo();
<leoprikler>[unless int = size_t]
<leoprikler>It might perhaps make a difference if you declare either of them const
<terpri>nckx, http://terpri.org/tmp/blender-fixed.gif shows how to test from the blender startup screen (without the fix, you don't get the 'Filmic' option under color management, just a default one)
<fnstudio>hi, ssh -T git@gitlab.com works fine but then when i try to pull from a channel hosted on a private gitlab repo i get an error
<fnstudio>guix pull: error: Git error: error authenticating: no auth sock variable
<fnstudio>i assume i'm using some auth mechanism that's not based on ssh-agent...
<fnstudio>and that guix can't "see" it
<NieDzejkob>leoprikler: oh, so taking it into account for function overloading. Rust and Haskell kinda do that. The caveat is that they don't have free-form overloading, you need to use traits/typeclasses.
<NieDzejkob>For example, in Haskell, foo <- readLn will infer the type of foo from its usage and then parse the input into that type
<leoprikler>so Haskell has a readLn : IO -> String and IO -> Int?
<leoprikler>Or is there an implicit String -> Int converter, that the compiler always considers?
<PurpleSym>It’s actually a generic, i.e. readLn :: Read a => IO a
<PurpleSym>So it can return anything that implements Read.
<alextee[m]> https://boards.4channel.org/g/thread/77262620 there's a guix thread on 4chan /g/
<leoprikler>I expected nothing and was disappointed anyway.
<kkebreau>Same. Typical 4chan.
<leoprikler>"there are <slurs> at Guix trying to repackage shit like Cargo/npm packages"
<leoprikler>Like, that's what a fairly competent package manager *would* do.
<alextee[m]>are there similar websites without moderation btw? i keep finding myself browsing 4chan /g/ because every other place is censored and needs personal info/an account
<alextee[m]>if you're mature enough to mentally filter out stupid posts it's actually not too bad
*dustyweb gives guix time-travel a try
<leoprikler>There are a bunch, but they tend to end up like 4chan.
<dustyweb>alextee[m]: best not to be stressed out about that kind of stuff
<dustyweb>just stay positive and keep hacking!
<alextee[m]>leoprikler: can you recommend some? are they just random image boards/clones of 4chan?
<alextee[m]>dustyweb: yeah ofc, it's just fun to see what people *really* think about topics vs censored places or places that give you "good guy points" on your account and likes
<alextee[m]>i saw 2 zrythm threads on there, they came up on my searches lol
<dustyweb>alextee[m]: maybe, but my general experience is that such communities mostly just want to show off how smart they are by tearing things down rather than building things
<dustyweb>it's much easier to look smart by yelling and swearing about how everyone else is wrong than by doing a lot of hard work to build something that works right
<dustyweb>so I mostly try to focus on people who are interested in the latter
<alextee[m]>i agree, zero productive things there. but good for sharing/viewing critical opinions
<dustyweb>alextee[m]: sometimes it's interesting to peer into /dev/null but beware /dev/null peering back into you
<leoprikler>I think you'll find most of the "critical opinions" on there are rather of the "perhaps women should code in the kitchen" variety to put it lightly.
<dustyweb>yeah
<alextee[m]>still fun to look at lol
<leoprikler>For some reason lefties are usually fine with not using slurs to further their ideas. Not sure if correlation or causation tho.
<Formbi>that is why chuds always cry about being «censored» as in not allowed to insult people
<Formbi>but those guys are right that the Guix team jumped on the bandwagon of cancelling RMS
<dustyweb>I think it's best not to go into this here
<dustyweb>now, on a completely different topic, looks like audacity 2.4.2 switched to cmake... wonder how much work it's going to be to port it over...
<kkebreau>dustyweb: Oh dear, I think I have remnants of my attempt at that port.
<alextee[m]>dustyweb: audacity has an insane build procedure according to the archlinux maintainer
<leoprikler>hopefully it's just adjusting the build system and configure flags *fingers crossed*
<alextee[m]>they depend on random unreleased versions of stuff
<alextee[m]>from their own git branches
<kkebreau>I have two patch files on my computer that modify libmad and libid3tag to install pkg-config files, so I guess that was necessary.
<alextee[m]>i see him complaining about audacity all the time lol , let me find the specific issue
<nckx>alextee[m]: That is the most content-free link I've seen all week. Thanks for the laugh 🙂
<alextee[m]>nckx: :D
<dustyweb>alextee[m]: ah really? :( unfortunate
<alextee[m]>looks like they're listening though, maybe worth waiting for the next release where they'll hopefully clean up a bit
<dustyweb>well this version doesn't have the clipfix plugin
<nckx>Is there a way to verbosely trace CMakeList parsing, similar to ‘set -x’ in bash land?
<dustyweb>but here's a good excuse to try out guix time-machine for something practical then :)
<leoprikler>love all the "just use nix :(" posts
<leoprikler>as if guix didn't already fork the nix daemon and build its own guile stuff around it
<nckx>I don't think it's supposed to make sense, leoprikler.
<leoprikler>Don't just shatter my dreams in such a cruel way 😭️
<dustyweb>wow
<dustyweb>time-machine
<dustyweb>works greeeeeeat!
<brettgilio>Hey all. Any of you using virt-manager with host to guest networking? If so, I'd love to see how you do it
<brettgilio>Also, polkit setups so you don't have to run virt-manager under sudo
<alextee[m]>brettgilio: not sure how to integrate it into guix, but i use virt-manager with bridged network
<alextee[m]>you just need to set up a bridge network. i wrote a guide (mostly just a reminder for me) here: https://www.alexandrostheodotou.com/orchestrating-virtual-machines-ansible-libvirt.html
<brettgilio>alextee[m]: the bridged network is not available when running virt-manager as sudo for some reason
<brettgilio>Will look
<brettgilio>Thanks
<alextee[m]>brettgilio: you need to run it as sudo -E virt-manager
*janneke pushes a emacs-guix update fix
<apteryx>something caused rust to be rebuilt on master, which means at least 1221 dependencies. Was that a mistake?
<apteryx>ah, that's my local change to gdb! sorry for the noise.
<apteryx>interestingly 'guix refresh -l gdb' says it has no dependents
<apteryx>gdb is used for the tests of rust-1.19, which is inherited by all other rust packages I guess.
<kkebreau>'guix refresh -l gdb@8.2' shows almost 1400 dependents.
<janneke>apteryx: try: guix refresh --list-dependent gdb@9.1
<janneke>that's a special "fixed" version for packages to depend on
<apteryx>ouch, I see :-)
<apteryx>is something more than --in-reply-to=<message-id> require to get the [PATCH v2] produced?
<apteryx>(when using git send-email)
<apteryx>My first attempt failed. Perhaps I simply got the message-id wrong.
<nckx>apteryx: -v 2 😉
<apteryx>ah! good one, thanks.
<nckx>Sorry, no space.
<apteryx>that option is not even documented in 'man git-send-email' :-/
<apteryx>and they use an exemple showing [PATCH v2] without mentioning it. Eh.
<apteryx>show an example*
<pkill9>where do you put obs plugins?
<pkill9>actually nevermind
<pkill9>then again, i do mind
<pkill9>you can put them in~/.config/obs-studio/plugins
<pkill9>but that's not ideal
<pkill9>i don't know if there is a search path you can specify
<apteryx>nckx: I tried -v3 with --compose, but the last line still read as: GIT: [PATCH] system: Modify GDB skeleton to find debug files for any. Does it mean it's not working?
<apteryx>ah, it works if I don't give it a patch file (and let it choose the last commit with -1 for example)
<apteryx>makes sense. Otherwise I'd have to format-patch the patch with the correct Subject line.
<nckx>Possible. I only ever use ‘git send-email -n’
<apteryx>what's the difference between --compose vs --annotate ? pop an editor vs inline?
<nckx>apteryx: More like write a cover letter vs. edit each patch, respectively.
<nckx>But I'm not sure how that differs from --cover-letter.
<nckx>More things I don't use...
*nckx is a simple fellow.
<apteryx>ah, --compose is just a message, --cover-letter pulls in more details such as the branch description, shortlog and overall diffstat.
<nckx>apteryx: I wouldn't call it undocumented but it's not very discoverable: ‘any format accepted by git-format-patch(1) can be passed to git send-email’, and the man page for g-f-m does mention it. But plenty of people (well, me) just ‘/ -v’ in less and won't find that.
<pkill9>is there any work on packaging jitsi desktop client?
<apteryx>nckx: ah I see, yes, that's where I could learn more about --cover-letter
<apteryx>these two commands overlap
<drakonis>y'all oughta start working on the desktop experience
<drakonis>its way too rough
<apteryx>perhaps you've followed the work of raghavgururajan? Some 330 ish patches landed to make the GNOME desktop experience better.
<apteryx>also, please log any issues you encounter by sending a mail to bug-guix@gnu.org (after checking they aren't already known); this will help.
<drakonis>also, containers.
<drakonis>i'd honestly like to have system containers
<apteryx>anyone running Debian?
<apteryx>does their build of GDB include Guile support?
<nckx>(Not Debian, but u durivutive: ‘Scripting in the "Guile" language is not supported in this copy of GDB.’)
<apteryx>I requested the feature ;-) https://bugs.launchpad.net/ubuntu/+source/gdb/+bug/1891818
<nckx>Does nobody there actually use GDB? GDB'ing without Guile is just... weird.
<apteryx>to be honest I only recently discovered how to make use of Guile in GDB, but I'm far from a very knowledgeable GDB user ;-)
<apteryx>is touching webkitgtk on master ok? 'guix refresh -l' says so...
<nckx>This channel is severely biased to see Guile as a friendly, more familiar alternative to scary ad-hoc languagues/CLIs, while for the rest of the world it's just the opposite.
<nckx>apteryx: Yeah.
<fnstudio>hi is it possible to pull from a channel (an additional custom channel that doesn't include guix) while ignoring the main one?
<fnstudio>the reason i'm asking is that i'm experimenting with a custom/personal channel and everytime i edit it and run guix pull again it takes quite long for it to pull from main
<pkill9>fnstudio: you can get the current version's commit with guix --version and run guix pull --commit=<guix commit> and it will use your currently built guix
<apteryx>leoprikler: I just merged your epiphany fonts fix into master, thank you!
<bluekeys>I'm trying to install gwl. Im on a foreign debian vm. I've done the old guix install gwl, but when I try guix workflow I get a guix: workflow: command not found error.
*kmicu ’s listening to ABBA’s Guixme Guixme Guixme 🎶
<fnstudio>pkill9: super, thanks, will do that
***bkv is now known as bqv
<leoprikler>bluekeys: you need guix environment --ad-hoc guix gwl
<simbaclaws>hi, I would like to try guix on my laptop using the same configuration that I've currently setup on my gentoo system but without any of the non-fsf approved software. Is it possible to use suckless software with guix? And start the installation with something like dwm?
<simbaclaws>I've made my own c patches that I would like to port
<malaclyps>does anyone know if guix weather with no arguments calculates the weather for the entire guix distribution, or the default profile? It's telling me it's looking at 14,711 package derivations which seems a lot!
<nckx>malaclyps: All of Guix.
<malaclyps>ah! thank you!
<leoprikler>you can use suckless software, but if you want to have your custom config, you need to modify the package, as required by suckless tools
*leoprikler wonders how modifying the code to suit your needs is an indicator of the tool not sucking, but they digress.
<nckx>You can pass it an optional list of packages or a manifest. I'm not sure if you can pass a profile on the CLI.
<nckx>Well, it's arguably easier to do so with inheritance than with classical distributions. But I've always considered suckless performance art: ‘all software sucks! see? here's an example.’
<leoprikler>Ah, but at least it's not "bloated" 🙃️
<OriansJ>everything not written in Assembly is "bloated" but who cares? You can buy a chromebook with 4GB for under \$50
<malaclyps>okay, i have hacked together something which looks at my profile, converts it into a manifest, feeds that to guix weather to find out what is lacking a substitute, and then feeds those as --do-not-upgrade arguments to guix upgrade
<malaclyps>i assume there's a far far better way of doing this and maybe it is just an option to guix upgrade
<nckx>Not yet.
<malaclyps>what would be a good name for that option? --only-upgrade-if-substitute ?
<leoprikler>--lazy ;)
<OriansJ>malaclyps: or --I-am-too-lazy-to-build
<malaclyps>lol
<nckx>malaclyps: --its-hot. Anything shorter. But we can bikeshed the name when it exists 🙂
<malaclyps>i am currently too lazy to build ungoogled-chromium which is a fairly high bar for "lazy"
<OriansJ>ooh --hot-take
<leoprikler>I think we should allow reading command-line args from ~/.guixrc
<OriansJ>or --fast-food
<leoprikler>then you could put --do-not-upgrade ungoogled-chromium in there
<leoprikler>I like --fast-food
<leoprikler>--bad-weather sounds like it would build everything that doesn't build on CI and then push it to Berlin.
<OriansJ>--rain-coat
<leoprikler>--slack-off
*nckx reboots IRC.
<leoprikler>In all honesty, though, you should just fix your problem via not upgrading chromium ;)