<lfam>What features does the Qt version have bdju?
<bdju>It can connect to a remote daemon a bit like transmission-remote-cli and transmission-remote-gtk, but these two remote clients can't rename files in a torrent, an incredibly important feature for me. Supposedly the Qt version has the remote support built in and also supports renaming files in the torrent file list.
<bdju>I've also found that "stig" can rename remote torrent files, but it's cli/tui only and not packaged on guix yet.
<bdju>Unsure if tremc can rename files. It's possible.
<bdju>I've been testing from NixOS where reverse of Guix, stig was packaged and tremc wasn't. Although the stig build fails right now so I actually ended up trying stig from an arch machine. It's all a bit of a mess.
<lfam>Tremc is a really trivial program, it should be possible to use on any OS that has Python
<lfam>It doesn't even get "built". It's just a Python script
<bdju>Ah, interesting. I usually stick to what's in the package manager, so I must confess I didn't try very hard.
<pkill9_>i have aweird issue, if i run a container and give access to /dev/dri, mesa loads the GPU driver, but when i also add the --network flag to llow network access, it can't
<mmcar>finally got guixSD installed on my x60 thinkpad. When it boots though it freezes at a point and there's just random noise on the screen. Is this a common problem? Not really sure what else to do.
<pkill9_>i think when you give network access to a guix container, all X applications try to access the X server over a network protocol rather than a unix socket, and i think X11 over network doesn't support GPU acceleration
<Blackbeard>lfam: guix system vm-image is really interesting
<mmcar`>Got my install working for the first time! Turned out a certain kernel module had to be blacklisted for it to boot properly. How long is 'guix pull' supposed to take on first run? Mine has only displayed 'Updating channel 'guix' from git repository at [..]' for about 10 minutes now.
<sulr70>After re-configuring the system with `guix system reconfigure /etc/config.scm`, I can not enter the gdm. It shows 'Removed session c1.'. There is terminal login prompt and I can login to the command line. I fixed the problem by changing the ownership of files under /var/lib/grm/ according to https://lists.gnu.org/archive/html/bug-guix/2019-09/msg00211.html While, I still have no idea why it happened.
<apteryx>which means the patch is no longer necessary
<raghavgururajan>apteryx I did try strict=no before using patch, but it disables all strict flags and not just -Werror.
<raghavgururajan>apteryx STRP is superseded by ZRTP, which is used when SRTP is set to NO. ZRTP for belle-sip is provided by BZRTP package.
<apteryx>SRTP and ZRTP are not the same thing? SRTP is like some TLS encryption of the RTP streams, while ZRTP is some new way to negotiate secrets and encrypt. If you have a VoIP provider that uses SRTP you can't use ZRTP as a drop-in replacement.
<apteryx>superseded as in the sense they are obsoleting SRTP support entirely?
<raghavgururajan>sneek, later tell apteryx: 1) The last gsm patch broke the package. xD. Nothing gets installed in output folders bin, include, lib and share. I fixed the issue by sending new patch (#40445).
<raghavgururajan>sneek, later tell apteryx: 2) Also, mediastreamer did not technically build with SRTP. I saw "MS will be built without SRTP support" during the build. I fixed the issue by using another patch, which I included in the new revision patch (11-add-mediastreamer-v3).
<sneek>brendyyn, rekado_ says: I’m sorry you find issues.guix.gnu.org “really bad”. Could you please send constructive criticism as a reply to the thread “Brainstorming features for issues.guix.gnu.org” on guix-devel?
<janneke>xavierm02: however, this weekend we got our first hello world guix hurd vm booted
<xavierm02>I saw. That's what made the thing too believable
<xavierm02>Somewhat unrelated question: How far off is Hurd from being a good idea for VMs in QubesOS?
<xavierm02>And if one were to switch from Linux to Hurd, is it really just about changing two lines in the config.scm or there's a lot of stuff to do?
<janneke>xavierm02: interesting questions; as for the config.scm: we have a custom way of building the vm; not through "operating-system" yet, no shepherd/services yet
<civodul>we're not there yet, but it could well be a couple of lines in config.scm to choose the OS
<janneke>so, that's probably a long path still with some open questions for me: do we really need qemu for a system reconfigure; etc
<janneke>otoh, `guix' runs on debian hurd, builds everything up to "hello", so lots of what needs to be done is "just work"
<xavierm02>@janneke "A lot of things are already packaged in Debian, so packaging this stuff in Guix is 'just work'" :D
<civodul>janneke: i've pushed two commits on Savannah, you can rebase and push your on top if you want
<xavierm02>When wandering in the internet, trying to figure whether it was an april fool or not, I found an IRC log where someone said that Guix was going to start using some typed language (Shen I think) and I was really happy for one second, until I realized it was a joke with the remainder of the sentence "to detect run-time errors" (or something like that) ...
<janneke>civodul: great; testing here without the remount,rw hack first
<janneke>xavierm02: guile is a strictly typed language ;)
<janneke>i could imagine that some static typing could be feasible and might actually be helpful
<brendyyn>when only some is used you have gradual typing, which incures a high performance penalty.
<xelxebar>running guix pull and then guix system reconfigure, the latter is giving me a warning to run guix pull and guix system reconfigure...
<xelxebar>Unfortunately, I seem to have broken the user's setup somewhere; guix pull is failing
<xelxebar>Probably due to the broken profile symlink?
<xelxebar>.config/guix/current points to a symlink that references a nonexisting path
<brendyyn>running sudo guix pull may have caused some damage im not sure
<xelxebar>Is there a standard way to fix these things? Or is manual intervention reasonable?
<xavierm02>@brendyyn I don't think you necessarily get a performance penalty. You can think of dynamically typed as statically typed with a unique type that's a big sum type of every possible type, and so each habitant comes with a tag that tells you which component of the sum it came from, i.e. which type it has. When adding static typing, one often wants to get rid of the tags (because you already know which tag it's going to be
<xavierm02>thanks to the static type), and so in this case, yes, you loose time adding and removing tags every time you switch from a statically typed context to a dynamically typed one. But you could also leave the tags, and then the program would run exactly as if no static typing had been done, and the static typing would just be used to detect errors.
<brendyyn>bricewge: sorry for wasting your time. it was really a bad coincidence that the first time it failed it showed errors that looked related, and it failed for your commit, but not the previous one
<ennoausberlin>Hello. Is there a way to install chicken scheme eggs on guixsd? The store is read only and the CHICKEN_REPOSITORY environment variable is ignored
<sneek>Welcome back ennoausberlin, you have 1 message!
<sneek>ennoausberlin, nckx says: I forgot to tell you to remove ‘insmod hfsplus’ since that can only confuse GRUB.
<leoprikler>Does chicken have a load path variable similar to guile, e.g. CHICKEN_LOAD_PATH?
*guix-vits "`guix build -L . gcc-0-toolchain ;+objc`" probably oolite is a great game, and they've used something similar to Flatpak before it become a thing... But c'mon, aren't games should be trivial to package?
<rekado>what makes you think games should be trivial to package? They are often among the more difficult applications.
<anadon>Reading of 0install docs, they have mailing list entries from 2018 referencing guix, but their official documentation only references a dated nix. Reading that, I don't see a significant justification of using 0install over nix. Since their development has been pitted in a nessicarily loosing battle and they haven't migrated, I'd bet that they just have an attachment or outstanding contracts to support 0install. Still, maybe someone sho
<mbakke>octave on core-updates fails to build with missing references to libstdc++.so symbols. I suspect it is because it picks gfortran's libstdc++.la, which references a libstdc++.so that does not exist. I don't understand why this issue shows up now, though.
<mbakke>wait, gfortran does have libstdc++.so, hmm
<efraim>Didn't we recently get a gfortran-toolchain package or something?
<mbakke>efraim: the -toolchain packages are generally not useful in package definitions, as a full toolchain is already available
<mbakke>changing the order could work, though; thanks for the tip
<lfam>efraim: About the rust crates, what stage of work would you be able to help with? I've imported a fresh batch on master but haven't started removing duplicates and handling upgrades, or cleaning up the UI bits
<efraim>I was thinking of grabbing some of the rand and other math packages for their own module
<str1ngs>rekado: where you thinking of reviving emacs guile?
<apteryx>raghavgururajan: sorry I think the rename from mediastreamer to mediastreamer2 might be unwanted after all; I just realized their source tarball is named 'mediastreamer'. Their doc speaks of 'mediastreamer2'... But maybe it's OK to stick to 'mediastreamer'.
<sneek>apteryx, raghavgururajan says: 1) The last gsm patch broke the package. xD. Nothing gets installed in output folders bin, include, lib and share. I fixed the issue by sending new patch (#40445).
<sneek>apteryx, raghavgururajan says: 2) Also, mediastreamer did not technically build with SRTP. I saw "MS will be built without SRTP support" during the build. I fixed the issue by using another patch, which I included in the new revision patch (11-add-mediastreamer-v3).
<efraim>we're at about 870 pacakges in crates-io.scm, I was hoping we were nearing full enough :/
<lfam>efraim: Okay... it's a few hours of work to re-sort everything to the point where Guix still builds, so I'd rather not do that work if we are just gonna move stuff into other modules
<rekado>but that’s not the legacy software I was referring to earlier.
<rekado>I’m currently implementing commenting for mumi / issues.guix.gnu.org
<rekado>(new mumi features are new legacy software :))
<str1ngs>rekado: Guile Emacs would be interesting project I think.
<roptat>not sure what the patching strategy should be... I want to move some packages to maven-bootstrap.scm, while changing them so that they use install-from-pom as their install phase
<roptat>how many patches is that, and in what steps?
<roptat>should I have one patch per package, moving them one by one and changing them at the same time, in a separate patch, or should I move everything at once, then patch them in individual files, or move everything at once, and patch them all in one patch?
<roptat>but guix refresh helps when I have enough courage to update the ocaml stack
<vagrantc>"guix refresh" has some features to check for new upstream version with specific importers ... i've often wondered if that couldn't be generalized to at least have a hook in a package description to check for new versions
<roptat>for those on github, you can also "watch" the repository in "release only" mode, so you get notified when there is a new version. There is also a hidden atom feed
<vagrantc>for uri-fetch ... you could usually just use the (source (origin (uri ... and update the version ... a simple checker that looked for similar tarball names with different version parts substituted would be quite doable (in debian, there's uscan for this sort of functionality, which has a regex to match versions)
<vagrantc>sometimes it's more complicated, but often it's just a bunch of tarballs in a directory
<rekado>for R I’m subscribed to the announce mailing list; for R packages I run the updater about once a week.
<rekado>everything else I do when I feel embarrassed.
<guix-vits>atw`: idk. do you know how to do that 'by hand'(y/n)?
<rekado>Parra: you wrote “different properties”, so I wasn’t sure if you mean the “properties” field of the “package” record. (It’s rarely used, e.g. for recording the upstream package name or to hide the package from the command line.)
<rekado>for different source origins I think it’s complicated because the “source” field is not thunked
<Parra>I see, I didn't refer to that, I just want to know how to handle that case
<Parra>and what about this, I can make a package per architecture, then a final package which adds a dependency dynamically depending on the current architecture
<rekado>Parra: I think what we did in some cases was to add these different source archives to native inputs and then to unpack the right one in a build phase.
<efraim>lfam: I got them sorted in alphabetically and removed duplicates. once 'make' passes I'll push it
<guix-vits>mfg: in the worst case (with obvious disadvantages) you can just copy the definition of the %desktop-services in your sys.config and edit them directly (but keep an eye on updates, he-he).
<mfg>guix-vits: that's what i currently use, but i wanted to improve a bit :D
<NieDzejkob>How do I reproduce "xmodmap -e 'keycode 62='" in the keyboard-layout record I pass to operating-system?
<noobly>guix-vits: i rebooted to get a clean slate, and after about ten minutes the wifi stopped cold turkey. ping 18.104.22.168 and ping gnu.org both fail. ps e | grep dhc only returns one process
<noobly>apologies in advance if i'm overlooking something basic. FWIW, i start my wifi with sudo ./wifi, where wifi is `sudo ip link set wls3 up`, `sudo su -c "wpa_supplicant -B -i wls3 -c<(wpa_passphrase "network-name" "password")"`, `sudo dhclient`
<rekado>noobly: on poor connections nscd may be too eager to cache failures.
<anadon>Is there a policy for removing packages? The very old verision of django has a python2 variant that is no longer maintained upstream.
<noobly>rekado: thanks, I've since run that command. hopefully this conenction stays up now
<noobly>rekado: the connection is indeed very poor, btw
<noobly>unforunately the connection was lost again. i rebooted, so should i be running `invalidate nscd` before `dhclient wls3`?, or after?
<FennecCode>If I wanted to run Zoom securely in some sort of box, should I do that in a container or a VM?
<leoprikler>VM or pure container, but you'll never be 100% secure on x86.
<FennecCode>Why is that specifically? And what ultimately is the difference between a VM and a pure container? Also, what makes a container "pure"?
<FennecCode>Sorry, I don't mean to barrage questions. 😅 I'm just excited to learn about this and not really sure what to ask.
<leoprikler>x86 has hardware flaws, that enable side channel exfiltration of information. E.g. if you have some process running in VM1 and some AES in VM2, VM1 can (through some magic) get to the key used in VM2.
<mfg>Hm so the error was to use '(service-types...) instead of (list service-types...) what is the difference between those two?
<leoprikler>A VM emulates hardware, etc. a container is a bit more lightweight.
<leoprikler>Forgive me about the pure thing, that was my confusion with pure environments, which also work similar to containers.
<FennecCode>Ohhh~ Right. Through Spectre/Meltdown and such? In that sense, there wouldn't be any possible way to have true isolation so long as it's the same physical machine.
<atw`>ooh somebody packaged emacs-form-feed. I was gonna highlight that as a "good for beginners" contibution
<leoprikler>You can containerize to protect yourself from inadvertent changes, e.g. service A and B both want to store information in the same file – you make one file for each and don't tell them about each other.
<leoprikler>containers are also a crude way of getting reproducible builds on distros without functional package managers, so that's that
<FennecCode>I keep reading about more and more security vulnerabilities that pop up for Zoom, and I'm thinking today is the day I finally install GuixSD on my desktop. Having a little more easy access to insulation is sounding really nice right now.
<ngz>hmmm I cannot compile a recent checkout of Guix : error: failed to load 'gnu/tests/docker.scm'. Am I alone?
<FennecCode>That, too. 🙂 It sounds nice for essentially a more comprehensive, language agnostic form of pyenv
<civodul>bavier: i think we should shorten the file names of patches for xplanet ('guix lint' should complain)
<civodul>noobly: try "herd restart nscd", but that seems weird
<noobly>civodul: ok, thanks, I'll try that after it fails again in a few minutes
<mbakke>civodul: git-fetch and similar runs on the remote node I think?
<civodul>i forgot, depends on whether #:local-build? is true
<dongcarl>So there's a reproducibility bug for gcc, which is present in all released versions right now... It is because the `ltmain.sh` was bootstrapped with libtool 2.2.7a, which didn't sort `find` output...
<dongcarl>I'm wondering what's the best approach here... Making a separate patch for each gcc version doesn't seem right to me...
<noobly>civodul: restarting nscd has also failed. I run dhclient with -v -d flags, and it looks like it's only ever renewing the lease once. it says 'bound to $my-IP -- renewal in 42515 seconds, but long before that time comes it says 'recieve_packet failed on wls3: network is down' (it's not, just a poor connection)
<civodul>noobly: i don't know, you can also just turn off nscd
<raghavgururajan>>apteryx: raghavgururajan: sorry I think the rename from mediastreamer to mediastreamer2 might be unwanted after all; I just realized their source tarball is named 'mediastreamer'. Their doc speaks of 'mediastreamer2'... But maybe it's OK to stick to 'mediastreamer'.
<raghavgururajan>I think MS2 is fine as they will be changing the changing the name of release to MS2 soon.
<mfg>because this has no distinct service-type or what is it?
<atw`>rekado: I have some ideas for contributions that should be easy. Focus would be familiarity with basic packaging, but also familiarity with git send-email, debbugs, and other stuff that I struggled/struggle with :)