<apteryx>kmicu: the kernel's btrfs wiki suggests that there is a global reserve designed to handle deleting files after the file system got filled up. Perhaps file a bug if you can easily trigger bugs with that by using Guix on a full btrfs file system :-)
<chkno>Hi. I'm new to guix. What's the polite but non-intrusive way to say "Ow, this bug hurts." in the issue tracker? (Like other bug trackers have star counts, thumbs-buttons, +affectsmetoo links, cc lists, subscriber counts, etc.)
<PotentialUser-86>I'm seeing that my messages didn't get logged... Is it safe to assume no one else saw them?
<str1ngs>leoprikler: the benefit is users can pick what platform they want. ie ncurses or maybe the prefer QtWebengine
<leoprikler>okay, so I can get that you may want something ncurses-based if you really like terminals, but you're still essentially writing three (admittedly quite similar) browsers for three different platforms, where users may have certain requirements w.r.t. look and feel, how data is stored, etc.
<samplet>str1ngs: No messages beyond “STARTUP” and “ACTIVATE”.
<brettgilio>Just wanted to quickly remark how much I love cherry picking packages from inferiors, and using the time-machine for development environments. I keep my system guix back a few weeks, I keep my kernel pinned back several dozen revisions, all in all my system is staler than an Italian crouton except in places where I want it fresher, and thats exactly how I like it
<str1ngs>samplet: anything in the terminal where you started nomad?
<str1ngs>brettgilio: time machine sound intresting.
<brettgilio>str1ngs: I haven't ran guix pull in weeks. It's wonderful
<GNUtoo>LibreCat: I meant about compiling the .scm file, loading it is probably about the same amount of time on my system due to some technical issues in my setup
<GNUtoo>(I'm using a guix daemon which is built with another guile version that what I used to build a local guix to hack on)
*GNUtoo assumed that you were talked about guix source code with desktop.scm like you have linux.scm and whatnot
<GNUtoo>For the rest you can simply remove xfce4 as explained before, guix gc can also help keeping the size of /gnu smaller
<GNUtoo>I also heard that somebody was working on replacing Yocto with Guix, and since Yocto rootfs sizes were in the hundreads of Megs years ago (I didn't touch it since), we might have more improvements in this area
<kmicu>apteryx: I assume that’s one of those experimental features (like quotas or rait56) that do not work in practice, otherwise all those folks on btrfs ml/irc attaching additional drives to make room have no sense. 🤷
*GNUtoo would like to be able to produce tiny rootfs + kernel that fits into 4M or 8M with Guix but that's probably still far away
<GNUtoo>(In the meantime we have LibreCMC for that)
<kmicu>apteryx: ok, after checking it makes sense, that GlobalReserve is exaclty for situations when we need to attach a bigger ‘boat’ to fix a full disk issue, it’s not to let us unallocate space in a usual manner.
<LibreCat>removing xfce does give a performance boost
<LibreCat>i am starting to think that my pc is dying
<rekado>mothacehe: no objections to reconfiguring ci.guix.gnu.org; just try not to reboot it without confirming that /store contains the latest kernel and initrd
<rekado>(otherwise we won’t be able to easily boot it)
<GNUtoo>hmmm, I've now managed to compile a bigger Android component
<GNUtoo>It's called libsamsung-ril and it's part of Replicant
<GNUtoo>however the Android.mk has several targets (LOCAL_MODULE), and it built only one of them (not the most interesting one)
<GNUtoo>I've got the same issue with my Android version of libsamsung-ipc where this time it built only the most interesting target (the library itself)
<GNUtoo>The good news though is that I got that far
<GNUtoo>(and in not that much time, most of it was spent waiting for the tests as I had issues between guix and guile versions)
<LibreCat>are you trying to port guix to a samsung phone
<GNUtoo>Not really, I'm looking for a saner build system for some low level Replicant components
<LibreCat>whats the diffrence between replicant and lineageos without google play services
<GNUtoo>Replacing the whole Android build system by Guix is probably not doable due to the lack of time and the fact that it needs to be maintained, so right now I'm just trying that for 1 package I need which lowers the maintenance
<GNUtoo>To really do that (Replace the Android by Guix) we'd need more people onboard
<GNUtoo>LibreCat: LineageOS makes all the hardware work, but to do that they use nonfree libraries
<GNUtoo>So for instance you had a backdoor in the library implementing the modem protocol on many Samsung devices, and it may still be there,
<GNUtoo>we don't know if that backdoor was accidental or on purpose though.
<GNUtoo>It would be neat to have the official tor-browser use Guix and remove nonfree software recommendation, but until then the tor-browser is not available for GNU/Linux ARM for instance
<str1ngs>LibreCat: nomad is a extensible web browser using guile scheme.
<GNUtoo>Some distributions like Trisquel are also lacking
<GNUtoo>So it's a matter of if your use case is covered or not (as usual I guess)
<LibreCat>how long would it take to compile trisquel for arm64
<str1ngs>your probably better of using Guix for arm64. I regularly build packages for my pinebook on my 3900k
*GNUtoo would prefer the about:addon removed from the official tor-browser and to trick their developers to use guix for better reproducible builds as I don't know if a fork could be distinguished from the stock(s) tor-browsers
<GNUtoo>(if forks can be, then you loose most of the anonimity)
<str1ngs>GNUtoo: my solution has been to write my own browser. not fork some other browser :)
<LibreCat>i am sceptical about building packages on a modern pc because of ME/PSP
<str1ngs>I would like to add GNUNet support to nomad. and possible tor
<str1ngs>LibreCat: I think those are well mitigated now. but yes hardware is always a concern not matter what free OS you are using.
<GNUtoo>The way tor-browser works is to make all tor-browsers look the same to browser fingerprinting, in practice you don't have only 1 fingerprint for the tor-browser but probably something like at least 6 for desktop computers
<GNUtoo>(You've got the 3 security profiles * 2 (because tails blocks adds while the stock tor-browser doesn't)
<LibreCat>str1ngs: what can stop intel from injecting backdoors on the binaries we are building
<str1ngs>I wonder what nomad fingerprints looks like hahah
<str1ngs>LibreCat: yep, that's why we need more open hardware
<LibreCat>str1ngs: but its either too expensive or to underpowered
<str1ngs>atleast with ARM though it's underpowered you get really good power consumption. so it's worth using in place like laptops etc.
<GNUtoo>LibreCat: it's probably irrelevant for that issue: we have reproducible builds in Guix, if an attacker used the ME/PSP, they would most likely not try to modify the binary but use other type of attacks (like DMA attacks, get stuff over serial, etc)
<str1ngs>GNUtoo: Guix has really good hardware mitigation. you can ask in #bootstrappable how they hardware mitigation bootstrapping.
<bonz060>In guix when defining a package, how do you specify a specific version? For example, I'd like `("python" ,python-wrapper)` to use `python-wrapper` for python3.6. How do I do that?
*GNUtoo wonder what would be the next thing when we have hardware to guix bootstrap
<GNUtoo>When we start having FPGAs with RISCV as a valid option that adds a huge layer again
<LibreCat>str1ngs: if intel or amd gets *really* wants to destroy the FSF they can use the microprocessor to sabotage public computers and censor social media so it never reaches the majority of the gnulinux comunity
<str1ngs>LibreCat: yep, thats why we need to support open hardware.
<LibreCat>str1ngs: the microproccesors can also be used to make gigantic ddos attack to the fsf servers
<rekado>bonz060: the thing after the comma (“unquote”) is the name of a variable that is bound to a package value.
<rekado>bonz060: python-wrapper is currently bound to the latest version of Python that is known to Guix.
<rekado>bonz060: if you want a different package variant (e.g. one with different configure flags or using different source code) then you need to a) define it and b) reference that variable by name in the inputs.
<rekado>bonz060: if you build one package with Python 3.6 (for example) you will also need to build all of its Python inputs with that same version, recursively
<str1ngs>LibreCat: the ROCKPro64 looks like a really good board. has 4GB ram. you can use nvme as well.
<GNUtoo>ok, passing LOCAL_MODULE= enable me to choose what to compile
<GNUtoo>That whole Android stuff probably needs some better abstraction for stuff bigger than adb, and the good thing is that guix is made for that (you can do all that programatically)
<GNUtoo>The bad news is that I'd need to really learn scheme to do that
<str1ngs>scheme is a feature. just saying. embrace the parentheses
<GNUtoo>str1ngs: what's nvme in that context? AFAIK it's an API
<GNUtoo>so you've got very different things like a SATA HDD + controller that can be seen as NVME to the OS
<GNUtoo>like what transport is it (PCIe? Something else?)
<joshuaBPMan>Hey #guix! I'm trying to configure pulseaudio...essentially I cannot use the command line to change my output device.
<raghavgururajan>LibreCat, yesterday, you were asking about fonts for good unicode and emojis support. I use `font-gnu-unifont` for good unicode support and `font-google-noto` for good i18n and emojis support.
<NieDzejkob>hmm, I have font-google-noto installed, and emoji work in my terminal, but don't work in IceCat
<rekahsoft>Hi all, what is the best way to go about replacing a package in guix from a thirdparty channel? That is, I want to replace `emacs-all-the-icons` with a version I will write myself that inherts from the upstream `emacs-all-the-icons` package but sets the source location to be melpa instead of melpa-stable.
<guix-vits>guixy: was added in use-anthology-style-naming for convenience: bsd- TAB --> choice, in one place.
<rekahsoft>guixy: Seems like you are also unsure of how channel package name conflicts will work. I am still using a fork of guix instead of maintaining a separate channel, but just ran into an issue where guix packages melpa-stable by default but I need the melpa version for a dependency I added to my fork (that I haven't committed upstream yet)
<rndd>msavoritias[m]: try openssh. i personally like bsd licanse
<Formbi>it would make a difference if you were making the things yourself IMO
<msavoritias[m]>well personally i am more into fairness. GPL-3 specifically is more fair than other non-copyleft licenses so i have a pretty hard line whenever i can. All the applications on my phone are GPL-3 for example
<msavoritias[m]><rndd "msavoritias: i think you should "> I see it as like somebody can close source the code whenever they want. That doesn't seem to me freedom of the user but freedom of the developer
<lfam>And since it has had thousands of authors, it will not be feasible to change, even if the main Linux team wanted to change it
<msavoritias[m]><Formbi "I'm pretty anti-corporate as wel"> Its not enough to be only free for me. It needs to be copyleft. So a company can't close source it against the communities wishes. And the license stays the same of course
<LibreCat>its only a matter of time until gnulinux gets invaded by tech corparations and most common distros have a lot of non free software
<lfam>And I'll add that the main Linux team doesn't want to change it
<lfam>Most distros do have a lot of non-free software