IRC channel logs

2021-02-19.log

back to list of logs

<g_bor[m]>civodul: is this renaming because other types of hashes will be supported?
<OriansJ>probably because no hash will be the perfect solution forever (see git and SHA1)
<milkey>that's what I figured
<civodul>g_bor[m]: yes, it's part of a patch series that made it in 1.2.0
<civodul>i think the release notes mention that
*rekado sees activity on wip-build-systems-gexp, holds breath
<jackhill>hmmm "guix pull: error: while creating symlink '/home/jackhill/.config/guix/current': No such file or directory"
<jackhill>fresh install, trying to make my home directory a btrfs subvolume. The skeleton files didn't get copied in, or chowned correctly. I suppose that's becasue /home/jackhill already existed, which is fine, and I did it manually, but guess I messed up something along the way…
<jackhill>guix pull -v3 doesn't give any more information…
<rekado>does .config exist?
<jackhill>rekado: no
<rekado>can you create it and try again?
<lfam>jackhill: Sorry about that. It's been fixed recently
<OriansJ>jackhill: you can manually link to /var/guix/profiles/per-user/$username//guix-profile
<lfam> <https://bugs.gnu.org/46269>
<raghavgururajan>jonsger: Shall we discuss about gnome tomorrow?
<shcv>how do I customize the system /etc/profile?
<jgart[m]>What is the recommended practice for packaging a go package that does not have a go.sum and/or go.mod like this one: https://0xacab.org/leap/shapeshifter
<jgart[m]>Ah, I think I need to set the following to: (#:import-path "https://0xacab.org/leap/shapeshifter"
***ece4 is now known as ece
<jackhill>rekado, OriansJ: thanks, manually creating .config/guix first as stated in #46269 got pull working for me.
<jackhill>sneek: later tell lfam thanks! $46269 was exactly it.
<sneek>Got it.
<lispmacs>hi, I'd really like to try out Qucs, but it doesn't currently build because of the python2 problem. I'm trying to build it with time machine but am having trouble finding a guix commit where Qucs or a dependency doesn't have a build problem. Does somebody know of a month or a commit where it definitely was building successfully
<pocketroid>Greetings programs
***lukedashjr is now known as luke-jr
***lukedashjr is now known as luke-jr
<Shmiggles>HELP: When using guix pack -R to create relocatable binaries and unpacking (tar xf) to local directory, I am unable to delete the gnu folder (permission denied with rm -rf). Not sure how that is even possible
<avalenn>jgart: import-path should be the Go import path, without http://
<civodul>Hello Guix!
<mothacehe>hey civodul!
<kmicu>( ^_^)/
<jonsger>mothacehe: what does us block from merging simple-cuirass
<mothacehe>jonsger: nothing, I did want to write a system test but completely forgot about that one
<bhoefling>My company asks if I could move the store from "/gnu/store" to somewhere $HOME. Is that possible? AFAIK, there are problems with substitutes, right? I found some older hacks from Pjotr to relocate, but nothing new. What's the state about this?
<Rovanion>bhoefling: Possible hack: Symlink it.
<bhoefling>Rovanion: Is it that easy? Just move "/gnu/store" to $HOME/my-store and ln -s $HOME/my-store /gnu/store ?!
<bhoefling>I thought there were some troubles with substitutes and libary references hard-coded in the binaries?
<Rovanion>Honestly I have no idea, but it wont hurt to try.
<Rovanion>The paths in the binaries will still be correct, there's just a symlink in the path instead of a directory at one part of the path. But again, just guessing.
<wonko_the_sane>bhoefling: you may be interested in using 'guix pack ..' and '--relocatable'
<bhoefling>wonko_the_sane: To produce the 'guix pack', I would need a store somewhere :-) I don't want to use another machine to hack this together. I want to use a real Guix(TM) :-)
<wonko_the_sane>bhoefling: out of curiousity what's the rationale ?
<NieDzejkob>IIRC there's a check somewhere to bail out when /gnu/store is a symlink, dunno what the rationale for that is though
<NieDzejkob>you'll probably have better luck with bind mounts
<brendyyn>the guix chat log search seems to be down
<NieDzejkob>not for me, I can even see your message there
<brendyyn>the search functionality
<civodul>hmm something's wrong with the goggles service (chat web interface)
*civodul reports it
<boombim>Hey guys. I've installed `flameshot` with guix but unfortunately `flameshot gui` command doesn't work. Can you guys help me with it?
<civodul>hi boombim!
<civodul>s/guys/people/ :-)
<civodul>i never used flameshot so can't help
<civodul>if you don't get a timely reply here, you can report the issue by email to bug-guix@gnu.org
<civodul>with the command that you typed and its output
<civodul>as well as the output of "guix describe"
<nckx>‘flameshot gui’ just exits without printing anything, exit status 0.
<nckx>That's on Wayland BTW.
<boombim>nckx yep. But it should open screen to select screen region
<boombim>I'm on x11
<brendyyn>same for me on x11
<boombim>it works with flameshot from debian repo.
<boombim>but with flameshot from guix it doesn't
<boombim>civodul send bug report to bug-guix as well
<nckx> https://github.com/flameshot-org/flameshot/issues/1195#issuecomment-749960588
<nckx>Purrhaps.
<nckx>I can't test that because ‘$ flameshot’ → qt.qpa.plugin: Could not find the Qt platform plugin "wayland" in ""
<boombim>It will not work on wayland without gnome
<boombim>it works only on x11 or on wayland with gnome
<brendyyn>nckx: i think i reported the bug for that ages ago
<boombim>it's not bug
<boombim>flameshot doesn't work on wayland without gnome
<nckx>lol
<boombim> https://github.com/flameshot-org/flameshot/issues/446
<boombim>I am going to close this since we decided not to use the XDG portal for KDE and Gnome. Additionally the clipboard was fixed in #1282 .
<boombim>Wayland support is not perfect but I would like to have more specific issues to track new problems.
<nckx> https://github.com/flameshot-org/flameshot/issues/446#issuecomment-717380307 ... headpalm facedesk.
<avalenn> https://www.tweag.io/blog/2021-02-17-derivation-outputs-and-output-paths/
<avalenn>Having no idea of the challenges posed by content-addressing I found this post interesting
<avalenn>I wonder if ERIS integration with Guix faces the same challenges or if they approach it totally differently.
<civodul>avalenn: the more i think about it, the more i think these "content-addressed paths" are a mistake
<civodul>it's addressing a problem (space/bandwidth usage, primarily) at the wrong place
<civodul>that breaks many things, which in turn calls for rather complex workarounds
<avalenn>content-addressed have the problem that it lives in purely static world
<avalenn>really good for caching and sharing but as soon as you want to refer to a yet-uknown content it does not fit
<civodul>yeah
<civodul>the store is just memoization and it has to be keyed by the function arguments
<boombim>nckx but it's so useful screenshoter
<nckx>And Wayland/GNOME/XDG folk broke useful screenshooters, it seems.
<nckx>But it should work on X11.
<nckx>Does it not work if you start a copy in the background as explained in issue #1195?
<nckx>Not saying we can't polish it further (you seem to say Debian's works out of the box?), but it would help narrow down the problem immensely.
<nckx><civodul> avalenn: the more i think about it, the more i think these "content-addressed paths" are a mistake
<nckx>Part of me is glad to hear that.
<nckx>Just because the conversation's been one-sided ‘of course this is the goal because IPFS whee’.
<boombim>nckx could you send the link to issue that you mentioned?
<nckx> https://github.com/flameshot-org/flameshot/issues/1195#issuecomment-750245645
<avalenn>past is static, archivists long for a tool managing the past and IPFS could just give it
<boombim>nckx got it but i just don't understand what should I do.
<avalenn>I like what IPFS try to do. But I start to see the big limitations of the approach.
<nckx>boombim: Start ‘flameshot’ in the background (&) for the GUI to talk to over D-BUS, then invoke a ‘flameshot gui’ instance.
<nckx>( flameshot &; ) && ( sleep 2 && flameshot gui )
<boombim>yep looks like it works.
<boombim>when i start flameshot in background it works correctly
<nckx>OK!
<nckx>That's very good to know.
<boombim>I started flameshot from terminal and then invike "flameshot gui" from another and it works
<nckx>Now, you say Debian's own flameshot ‘just works’ when you type ‘flameshot gui’?
<boombim>yep.
<boombim>I bind it to my screen key
<boombim>I press it and it starts in background and then launch
<boombim>no issues
<nckx>We should be able to replicate Debian's behaviour. Could you send a bug report to bug-guix@ ?
*nckx has to → AFK.
<boombim>nckx I did but looks like it rejected
<boombim>should I register there?
<nckx>No, it's an open list. I don't see any messages in the queue. Define ‘rejected’?
<nckx>If you received mail, what did it say?
<boombim>Yep. Here is it
<boombim>You tried to send a mail to one of the GNU bug mailing lists that uses
<boombim> http://debbugs.gnu.org to track its bugs. Your request was:
<boombim>    Posting of your message titled "Flameshot but"
<boombim>Your mail has been rejected by the debbugs moderator. The moderator
<boombim>gave the following reason for rejecting your request:
<boombim>"Non-members are not allowed to post messages to this list."
<boombim>If you have questions about this, you can try emailing the relevant
<boombim>GNU project's general discussion mailing list.
<boombim>If you cannot find that list, you can mail help-debbugs@gnu.org.
<nckx>Which address did you use?
<nckx>Exactly?
<nckx>"Non-members are not allowed to post messages to this list." is certainly not true of bug-guix at gnu.org.
<boombim>bug-guix@gnu.org
<boombim>I've sended to this address
<nckx>o_O
<nckx>That is bizarre and untrue; I'm looking at the settings right now, they don't say that.
<nckx>boombim: Could you send it again (describing everything we've discussed here, not just ‘flameshot gui doesn't work’)?
<nckx>I'll keep an eye on the queue.
<boombim>ok
<nckx>Really have to go. Be back in 7 minutes.
<boombim>I've sent message.
<nckx>Nothing yet...
<boombim>I didn't receive any reject message
<nckx>Nor should you: https://www.tobias.gr/gna.png
<rekado>civodul: what’s wrong with goggles?
<rekado>(or rather: which *new* part is broken?) :)
<nckx>I think the indexer has once more died.
<rekado>I’d like to fix the redirect error today.
<rekado>ah,
*rekado logs on to bayfront
<nckx>Unless we haven't discussed something called ‘guix’ since 4 January.
<rekado>it’s a rare topic these days
<rekado>the indexer is running, but I see a bunch of Xapian::DatabaseLockError
<rekado>looks like more than one process trying to write to it
<nckx>boombim: Have you received another rejection notice? Your mail never made it to the moderation queue.
*nckx AFK longer this time.
<boombim>nckx no. I haven't receive any message
<wuui2>Why is rav1e 0.3.5 package called "rav1e", but version 0.4.0 is "rust-rav1e". Also why keep two versions, there's literally no reason to use 0.3.5 instead of 0.4.0.
<wuui2>And the package definition for "rust-rav1e" 0.4.0 is in the wrong file (crates-graphics.scm, it should be in video.scm).
<leoprikler>Probably because the latter was generated through the crate importer and the committers/reviewers didn't check for that.
***mrmerso[m] is now known as gDmZbE[m]
<rekado>re goggles: the xapian database is fine (“quest” shows recent records), but the records don’t appear when searching with guile-xapian; my guess is that sorting before filtering isn’t quite working.
<ngz>wuui2: rust-rav1e is an intermediate package, which is not built (#:skip-build? #t)
<ngz>rav1e is an end-user package, build, and with depedencies, so it needs to be carefully checked.
<wuui2>ngz: Ok. I see.
<ngz>leoprikler: the committer checked for that, actually...
***gDmZbE[m] is now known as Fd9a[m]
<daphnis>i'm a bit confused about the application setup in the installation guide. it says to install glibc-(utf8-)locales and export GUIX_LOCPATH.... -- shouldn't this go into .bashrc or something? anyway, after doing this, i still get the same `hint' about installing ...-locales when i do `guix package --list-installed`
<civodul>rekado: sorry for the delay; re bayfront, the upgrade is still ongoing
<civodul>i'll send a message to guix-sysadmin
<rekado>evidence: http://logs.guix.gnu.org/guix/search?query=nick%3Arekado+%22process+trying%22
<rekado>must be a filtering problem
<civodul>daphnis: hi! perhaps you're using a locale that's not part of glibc-utf8-locales?
<civodul>what does the "locale" command print?
<daphnis>civodul: i'm using the british utf-8 one
<daphnis>(hi!)
<civodul>daphnis: "ls $(guix build glibc-utf8-locales)/lib/locale/2.31" shows that en_GB.utf8 isn't there
<civodul>so you probably need to install glibc-locales rather than glibc-utf8-locales
<daphnis>civodul: aha; thanks!
<daphnis>that was a strangely small selection
<nckx>daphnis: Never in .bashrc. The installation script (etc/guix-install.sh) creates an /etc/profile.d drop-in that sets it correctly if the required locale package is installed.
<nckx>daphnis: Yes, there's a constant grumbling to remove or hide or rename it (for use by test suites only), as it's so tiny and confuses people.
<daphnis>nckx: i see, so the export line just there so one doesn't have log in and out
<nckx>The counter-argument is that all locales is a lot of locales and nobody ever uses *all* locales.
<daphnis>or out and back in rather
<nckx>Correct!
<nckx>Running . /etc/profile should do the same.
<daphnis>and would be a lot shorter
<nckx>But it's probably better to be explicit in the manual than to assume that all just works ¯\_(ツ)_/¯
<nckx>(Although you'll run into trouble later if it doesn't, anyway.)
<ngz>nckx: Could we provide multiple glibc-small-subset-locales, with a combination on en_US and country locales
<nckx>We could, but it would be better not to curate subjectively at all.
<nckx>(Whether we group them by continent, ‘culture’, whatever, is still very subjective and folks like me would end up installing ~3 of those big packs just to get their 5 locales.)
<ngz>hmmm, and generate a glibc package on the fly during installation depending on selected regions?
<nckx>But that's a lot of individual packages.
<ngz>Ah well.
<nckx>ngz: Selected where?
<civodul>"guix size glibc-locales" gives a good idea of what the tradeoff is :-)
<ngz>I thought the install process offered to select regions
<nckx>There's already a make-locale-something procedure.
<nckx>ngz: Doesn't locale-definitions already do that?
<ngz>Anyway, if there is no solution, there's no problem. :)
<civodul>package parameters would provide a way out of this
<civodul>or we could ignore the problem like everyone else and not even emit a warning, after all
<nckx>ngz: I think the ‘problem’ we're discussing only affects foreign distroes, not Guix System, because it already does that.
<nckx>Correct me if I'm mistaken.
<civodul>yup!
<ngz>OK. Luckily, I am one of the happy few who can use glibc-utf8-locales on their foreign distribution. But I was worried about the others.
<ngz>Maybe the install script (on a foreign distro) could do the same thing as the installer? It could accept arguments about required locales, and proceed as usual otherwise.
<nckx>ngz: The problem is how that choice is ‘represented’ in the manifest.
<nckx>Without parameterised packages.
<ngz>OK. So let's implement parameterized packages first! Easy ;)
<nckx>Or 645 auto-generated locale packages.
<ngz>I'll leave it as an exercise to the reader, of course.
<nckx>Generating 645 packages is trivial, it's just a procedure, but it rather goes against how we currently Do Things in Guix.
<ngz>nckx: Bah, we have almost 2000 auto-generated packages in crates-io.scm
<dftxbs3e>hello! is it possible currently to alter all packages in GNU Guix with package transformations during the invocation of one GNU Guix command?
<dftxbs3e>e.g. --without-tests=*
<nckx>ngz: T_T
<ngz>I know that feeling…
<ngz>But, hey, it works©!
<nckx>I think I like that emojo because it's up to the reader whether those are tears
<nckx>or blood.
<ngz>Fortunately, in the environment I live in, eyes pouring tears are more common than those pouring blood.
<ngz>BTW, was there a discussion about Rust packaging in the last Guix Day?
<nckx>sneek: Later ask boombim: could you /msg me the e-mail address you used to mail bug-guix? Unless you're Pastor Patrick Agugoesi, your mail never made it through and I'd like to ask the GNU sysadmins if they know why.
<sneek>Okay.
<nckx>sneek: eat a bag of botsnacks.
<ngz>I believe I remember something about it was informally planned.
<nckx>...or don't.
<nckx>ngz: Not really my jam but I thought so too.
<apteryx>rekado: I think pigx is broken again on master
<sneek>Welcome back apteryx, you have 1 message!
<sneek>apteryx, jlicht says: the pip included with python on core-updates works fine!
<apteryx>jlicht: because it's newer than 20.X I guess
<apteryx>thanks for the report!
<rekado>apteryx: thanks, I’ll look into it
<ngz>I wish we could do something like this: (define-package rust-whatever ... (arguments `(#:cargo-inputs '(("rust-input-A" "rust-input-A-version" "rust-input-A-hash") ...)) ...))
<ngz>without "rust-input-A" being the name of a package in Guix, of course
<rekado>can they just be origins?
<ngz>Good question
<ngz>I don't think so, because "rust-input-A" still have its own Cargo inputs.
<rekado>apteryx: it’s fine on commit a03a3ab0434fc39c19d1cfda7d7ff95ad97c5af3
<avalenn>ngz: I wondered about the duplication with Go too.
<rekado>that was yesterday evening
<apteryx>rerekado: oh nevermind, it just built fine; I think I was mislead by its substitute unavailibility
<apteryx>sorry for the bother :-)
<rekado>phew
<apteryx>the amount of r packages required to build it is mind blowing :-). I must have shuffled 4 GiB across the network to the offload machine.
<raghavgururajan>Hello Guix!
<nckx>o/
<apteryx>hi raghavgururajan!
<raghavgururajan>o/
<apteryx>rekado: it's strange, weather had reported that: 0.0% (0 out of 16) of the missing items are queued at the time it found pigx to be missing
<apteryx>but it was built and made available by the CI shortly after
<rekado>apteryx: it’s annoying that there are so many GHC packages needed to build it
<rekado>only because of pandoc
<rekado>and rmarkdown uses pandoc
<apteryx>I see! GHC packages are huge
<PotentialUser-84>I tried to package rav1e 0.4.0 several weeks ago, but failed and gave up. The problem is that when I try to compile rav1e, I get several error messages like this: "crate x required to be available in rlib format, but was not found in this form". Does anyone know how to compile the required crates in rlib format (I don't know rust...)?
<PotentialUser-59>i'm trying to create my first package but can't use subtitute* which i found while looking at some other packages: Unbound variable: substitute*. any idea what could be wrong?
<ngz>you need to use (guix build utils) module
<ngz>This is where it is defined.
***db48x2 is now known as db48x
<nckx>Matrix is doing well.
<boombim>nckx looks like I've recieved successr response tham my report was appreved half of our ago.
<sneek>Welcome back boombim, you have 2 messages!
<sneek>boombim, nckx says: could you /msg me the e-mail address you used to mail bug-guix? Unless you're Pastor Patrick Agugoesi, your mail never made it through and I'd like to ask the GNU sysadmins if they know why.
<sneek>boombim, nckx says: cancel that ☝, your mail arrived! ...3 hours after you sent it.
<boombim>oh, you already said it
<boombim>sorry, I was absent
<nckx>boombim: Yep, I got back and your mail landed in my inbox for approval. Weird.
<nckx>No problem, that's what sneek is for!
<kondor>I'm getting a bunch of store-protocol errots on opening pipes and system calls in a gexps. I learned I shouldn't involve #$output in a call to `system`, as well as not opening pipes on #$ unquotations
<kondor>Does this make sense to anyone?
<raghavgururajan>sneek, seen guix-vits?
<sneek>I think I remember guix-vits in #guix 26 days ago, saying: sneek: botsnack.
<mdevos>How can I run a vm where data persists after reboot? I tried "guix system vm", and adjusted the "-drive file=..." argument to refer to a qcow2 image that uses /gnu/store/...-qemu-image as backing file and removed "-no-reboot", to no effect.
<mdevos>I'll use 'vm-image' in combination with --volatile for now. Seems somewhat wasteful, though.
<mdevos>(Disk space and time spent building bootloader stuff)
***jpoiret9 is now known as jpoiret
***yjftsjthsd9 is now known as yjftsjthsd
***penguwin9 is now known as penguwin
***leoprikler_ is now known as leoprikler
***db48x2 is now known as db48x
<boombim> https://paste.debian.net/1186126/
<boombim> guys I tried to install vim-full because i need system clipboard feature but it failed...
<boombim>can you hekp me please?
<nckx>boombim: I'm not your guy, dude. See "View build log at '/var/log/guix/drvs/ns/zargr9s510lfr0680mkg4m4lfimwcd-vim-full-8.2.2067.drv.bz2'": please paste the (decompressed) log to paste.debian.net, with bzcat/bzless/etc.
<nckx>I suspect a test failure, vim is notorious for those.
*nckx flies away.
<ngz>Thankfully, we have Emacs ;)
<ngz>which builds perfectly fine, BTW. Just sayin'
<ngz>Error is: "ld: /gnu/store/imh5xxqw10dql4crlngbbjh4r24raf4j-expat-2.2.9/lib/libexpat.so.1: error adding symbols: DSO missing from command line"
<ngz>This kind of error is above my pay grade.
<ngz>hmm and the line above: "ld: /gnu/store/…-python-3.8.2/lib/python3.8/config-3.8-x86_64-linux-gnu/libpython3.8.a(pyexpat.o): undefined reference to symbol 'XML_FreeContentModel'"
<ngz>Not sure if both relate.
<ngz>Probably.
<boombim>nckx it's 10 thouthands strings in file. looks like debian pastezone is down....
<boombim>oh i can just attach file
<ngz>boombim: I think I just pasted the error, didn't I?
<boombim>oh yep
<boombim>Thanks ngz
<boombim>Is it enough?
<ngz>No idea.
<cage_>sorry for the trivial question but how can i fetch a tagged git commit?
<sneek>cage_, you have 1 message!
<sneek>cage_, raghavgururajan says: Could you share your pack-def via debian paste-bin with no expiration. I'll see what I can do to make it work. :-)
<cage_>raghavgururajan, i succeded finally!
<cage_>thank you very much!
<raghavgururajan>cage_, Great!
<cage_>there are some problem with a library but i can not understand if it is a bug from my code or what
<cage_>raghavgururajan, thank you again!
<boombim>413 Request Entity Too Large
<boombim>cant attach
<boombim>File is to big
<cage_>my program uses ncurses and, apprently the package built with giux laks UTF8 supports, not sure why
<PotentialUser-84>When I try to compile rav1e 0.4.0, I get several error messages like this: "crate x required to be available in rlib format, but was not found in this form". How can I compile cargo inputs in rlib format?
<apteryx>has anyone configured xterm to open URLs in icecat?
<raghavgururajan>cage_, May be you need to build with unicode support? Like with icu4c and/or libunistring.
<ngz>PotentialUser-84: Could you show your package definition?
<cage_>raghavgururajan, can i rebuild ncurses to support unicode using some directives in my package definition?
<apteryx>oh, I've found happiness, I think: https://aduros.com/blog/xterm-its-better-than-you-thought/
<PotentialUser-84>ngz: Yes.
<PotentialUser-84>ngz: https://paste.gnome.org/p508xlt8w
<raghavgururajan>cage_: Wait! You meant that ncurses itself lacks unicode support or that your package lacks it?
<cage_>ncurses apparently
<raghavgururajan>Ah
*raghavgururajan checks
<cage_>i use a common lisp wrapper for the C library (ncurses)
<cage_>the wrapper try load the libncursesw.so
<cage_>where 'w' is for wide character IIRC
<cage_>i get no error loading the library, the software load and works fine except that lacks accented character and such
<raghavgururajan>It seems ncurses is built with unicode support (--enable-widec).
<raghavgururajan>Yeah, libncursesw.so is for wide char
<cage_>ok
<cage_>thanks so this is likely an issu from my code
<cage_>thanks for checking!
<cage_>*issue
<raghavgururajan>Np!
<ngz>PotentialUser-84: hmmm, maybe it needs a more recent rust-cargo-c?
***daviid is now known as Guest95794
***Guest95794 is now known as daviid
<PotentialUser-84>Maybe, I will try to update it.
<ngz>Also, I don't know why "cargo cinstall..." replaces regular build? Ain't you supposed to "cargo build --release" too?
<PotentialUser-84>Idk, it was already like this, I only edited cargo-inputs.
<apteryx>ghc itself weights nearly 1.4 GiB, ahem.
<apteryx>weighs*
<apteryx>wow, the xterm link above really delivered
*apteryx restarts xterm
<spk121>cage_: it may not be obvious, but, ncurses wide expects UTF-32 codepoints, or whatever your OS is using for wchar_t
<cage_>spk121, yes but the fact is that if i recompile the software without guix everything works
<cage_>on the same machine, of course ;-)
<spk121>cage_: well, sorry then. I'm out of ideas.
<cage_>spk121, no problem, thank you anyway! :)
***tepper.freenode.net sets mode: +o ChanServ
<PotentialUser-15>can someone tell me what the heck is the name of the installer so that i can lauch the installaton ?
<mdevos>what do you mean with ‘launch the installation’? Are you trying to install Guix System via the installation image burned / written to a CD or USB storage medium?
<PotentialUser-15>yes
<PotentialUser-15>i chose to install via shell and now I don't know how to start the curses installer
<lfam>The installer image is a bootable operating system image. It should appear when you boot the computer
<mdevos>are you using the installation image (guix-system-install-VERSION-ARCHITECTURE-STUFF.iso)?
<sneek>lfam, you have 1 message!
<sneek>lfam, jackhill says: thanks! $46269 was exactly it.
<PotentialUser-15>mdevos yes i do
<PotentialUser-15>lfam i wanted to install f2fs first to see if it would appear as an option to the installer
<PotentialUser-15>or to select the partition scheme manually in there
<lfam>It's a little hard to understand what you have tried, and what you are asking
<PotentialUser-15>I have booted on the usb ; and in the menu I chose the "graphical" installer. But I noticed f2fs was not an option so I rebooted, installed f2fs-tools. And now I wish I could start the installer again
<mdevos>PotentialUser-15: the actual installation is done with ‘guix system init config.scm’, the ‘guided graphical configuration’ is just for partitioning the disk and assisting the user with writing a configuration file.
<PotentialUser-15>mdevos I would like to be assisted
<mdevos>Looking at gnu/installer/parted.scm, it seems like f2fs is not (currently) supported by the installer
<PotentialUser-15>is that possible to use f2fs anyway ?
<mdevos>idk if guix and GRUB support booting from f2fs currently. If they do, ...
<mdevos>then you can manually partition the disk. See ‘Manual Installation’ in the Guix manual
<PotentialUser-15>afaik grub has f2fs support since 2018
<PotentialUser-15>but i never tried it for real
<PotentialUser-15>so i just can't use the curses tool ?
<lfam>PotentialUser-15: It's still not clear what you tried to do but, in general, I recommend following the directions in the manual very closely. After you have installed Guix System, then it's a good time to experiment
<PotentialUser-15>I'm just trying to follow to gfx installation but with an extra step in the shell before
<PotentialUser-15>i would have used another tty but qemu doesn't seem to grab it
<mdevos>there are some nice ncurses-based partition programs. I don't recall which one I used, perhaps it was gptfdisk
<lfam>Okay, but I don't understand how there can be "an extra step in the shell before". The installer boots to the installation interface. There is not a "before" step
<mdevos>lookig at gnu/tests/install.scm, it seems like guix supports booting from f2fs
<PotentialUser-15>I want to install f2fs before...
<lfam>f2fs is a filesystem. It's not something that can be "installed" like a package
<PotentialUser-15>afaik a filesystem is either a fuse program or kernel module, both usually packaged
<lfam>Are you trying to install Guix on a partition with a f2fs filesystem? Are you trying to create an f2fs filesystem?
<lfam>It would help if you tried to be more specific
<PotentialUser-15>2d then 1st
<lfam>I really recommend installing with the "manual" method, in that case
<PotentialUser-15>sorry, I don't get what's not clear but I guess I wasn't
<lfam> https://guix.gnu.org/manual/en/html_node/Manual-Installation.html
<lfam>My recommendation assumes that f2fs is not an option in the graphical installer
<nckx>mdevos: Yeah, cfdisk and cgdisk, one of them's provided by gptfdisk, another by util-linux (which now is fully GPT-capable so installing gptfdisk is generally no longer needed).
<PotentialUser-15>allright then
<terpri>and there's gnu parted of course, though iirc it's not ncurses-based. cfdisk and cgdisk don't seem to be packaged...
<lfam>I'm checking now if the installer image includes support for f2fs
<nckx>terpri: gptfdisk and util-linux...
<PotentialUser-15>ifam thanks
<PotentialUser-15>lfam thanks
<nckx>Guix definitely does.
<nckx>To boot, I mean.
<lfam>Okay, I checked, and the 1.2.0 installer doesn't offer f2fs in the graphical interface
<lfam>The manual process is really not hard, assuming one has some experience working on the command-line. I think that desiring f2fs indicates some experience :)
<raghavgururajan>Anyone would like to try social networking using XMPP? https://issues.guix.gnu.org/46628
<lfam>It offers ext4, btrfs, fat16, fat32, jfs, ntfs, and linux-swap
<nckx>root-on-swap support is experimental.
<terpri>there is indeed a fixed list of installer-supported FSes, in gnu/parted/installer.scm
<lfam>Lol nckx :)
<PotentialUser-15>lfam I have experience with linux but not with guix
<PotentialUser-15>but that's okay i will follow the command line installation
<lfam>You have the right kind of experience PotentialUser-15. And we are here to help
<lfam>Thanks for the reference terpri
<PotentialUser-15>nckx root-on-swap ?
<PotentialUser-15>lfam thanks
<nckx>PotentialUser-15: ignore me.
<lfam>I think you meant 'gnu/installer.scm', terpri :)
<lfam>Or, I don't see gnu/parted/installer.scm in guix.git, anyways
<lfam>We could add f2fs support in the next major release
<terpri>er, gnu/installer/parted.scm rather
<PotentialUser-15>a live iso would be great
<lfam>PotentialUser-15: I think that's the "GNU Guix 1.2.0 QEMU Image" download
<PotentialUser-15>I have only one computer so If a boot baremetal I'd better know how to manage a way to some documentation
<lfam>Oh, I see that I misunderstood
<terpri>used by gnu/installer/newt/partition.scm
<lfam>Ah, thanks
<nckx>PotentialUser-15: What would you consider a ‘live ISO’? I ask because the installer is a live ISO in my book, but the meaning of the term seems to have changed over the years.
<nckx>‘Has a GUI’?
<PotentialUser-15>I mean a system that leave you on a functionnal desktop like any other livecd
<nckx>OK, has a GUI.
<PotentialUser-15>yes
<terpri>the guix install disks ought to have the info manual, but i don't remember if they do or not
<lfam>We could make the installer boot a simple GUI on another TTY
<PotentialUser-15>a webbrowser is useful for documentation
<lfam>It might be nice to be able to include an Xfce system in the installer anyways
<terpri>(they also should include emacs for a good info reader + config.scm editor ;))
<lfam>We already include zile, which is much smaller ;)
<nckx>But doesn't include a Web browser.
<PotentialUser-15>I can see that guix is clearly for power users (yet ?)
<nckx>My first action when booting the live ISO is to install emacs so I can browse the Web while it thinks real hard about my derivations for a few hours.
<lfam>PotentialUser-15: There is still some work to do to make Guix super-easy
<lfam>We could use a GUI for package management, and some improvements to how services are made
<PotentialUser-15>lfam I wouldn't have asked for this one but gnome has something already anyway
<PotentialUser-15>if you like gnome of course
<terpri>it's a bit like gnu/linux in the 00s. easy when it works, but you'd better know what you're doing when it doesn't...
<nckx>PotentialUser-15: Somewhere in between, but it's not intended to be. What's standing between us & that is just... work.
<lfam>Yeah, GNOME is great! It's a sister project, being GNU-y :) I don't remember if Guix has integration with the GNOME package management tool yet, though
<lfam>I only suggested Xfce since it is small
<lfam>And easy enough to figure out for people coming from other operating systems
<lfam>Xfce is small and simple to "put together", unlike GNOME
<nckx>PotentialUser-15: It's relatively easy to build an installer with GNOME today, it's just a huge ISO.
<terpri>nckx, just a small matter of programming
<nckx>And our installer is text-based anyway so it wouldn't really boot to something that makes sense to users, yet.
<PotentialUser-15>that would call for a really graphical installer obviously
<terpri>. o O ( write a CLIM-esque toolkit for the installer with both terminal and gui backends )
<nckx>Yeah.
<nckx>It's going to happen, we just don't know when.
<PotentialUser-15>i see
<PotentialUser-15>that's maybe not the most important right now
<PotentialUser-15>i certainly would prefer more packages and services
<terpri>i remember reading something about a distro-agnostic installer that debian et al. might use in the future, but it's doubtful it'd work for guix
<nckx>Yes, Guix System is far from complete, although it is usable (it's been my only OS for 5+ years now).
<lfam>PotentialUser-15: Any packages you are missing, in particular?
<PotentialUser-15>lfam for now I don't know
<PotentialUser-15>zram ?
<nckx>terpri: ‘distro agnostic’ means it supports both .deb and .rpm!
<nckx>‘We have an if clause’®
<nckx>(I looked.)
<terpri>lol
<PotentialUser-15>i meant *zswap*
<nckx>zswap works swimmingly.
<nckx>(zswap ftw > zram).
<terpri>btw, is anyone successfully using gnome's bluetooth control via the control center etc.?
<PotentialUser-15>wait, I though I misremembered the name but I see the two exist
<nckx>PotentialUser-15: zswap is a better zram.
<nckx>Sadly, zram has better publicity (pushed by systemd etc.)
<terpri>the control panel says i have no bluetooth, but bluetoothctl works and gnome-shell shows the icon in the menu when a device is in use (leading to the pane telling me i don't have bluetooth...)
<nckx>And ‘Linux’ tends to fetishise layers, so putting swap on a virtual compressed block device in RAM sounds sane then.
<nckx>terpri: You're not the first to report that IIRC. Nothing on the bug tracker?
<PotentialUser-15>nckx ok i am using zram currently. How is it better ?
<PotentialUser-15>(how is zswap better ?)
<terpri>nckx, no, but i'll take a look at the tracker and put it on my todo list
<jackhill>terpri, nckx: I have the same experience, and I think this was known at least sometime in the past
<mdevos>terpri: GNOME doesn't detect bluetooth, I use blueman, I think.
<mdevos>blueman's graphical interface works (or was it bluez?)
<jackhill>maybe it didn't get added to the tracker though
<lfam>Does anyone know if there is a list of device drivers that are disabled in linux-libre?
<terpri>indeed, taylan reported exactly the same problem almost 3 years ago (#32166). time for some d-bus d-bugging i suppose
<PotentialUser-15>got it https://stackoverflow.com/questions/18437205/difference-between-zram-and-zswap#18598732
<nckx>zram is very simple: it creates a virtual block device in memory, that is transparently compressed behind the scenes. That's flexible -- you can mkfs.btrfs it, or mkswap it, as many people do -- but pretty ‘dumb’: it doesn't actually know what its being used for. It can waste memory because data is kept compressed in memory (taking up less but some space) when it would be better written to storage altogether (no space at all, more RAM for things that matter
<nckx>). zswap is more complicated because it actually does that. Many people don't care about the difference, which is fine, but it is measurable, and most people are kept unaware, which is not fine.
<nckx></nckx rant> ☺
<lfam>It's annoying, you can't do `guix build linux-libre --log-file` and see the log of the deblobbing
<lfam>Freedom by obscurity, I suppose
<nckx>Don't... don't go there, lfam.
<lfam>I've been there
<nckx>I need to get the popcorn first.
<PotentialUser-15>nckx i've notice zram pages can get easily bloated
<lfam>Anyways
<nckx>I've been there and back and it's never fun.
<lfam>Did you forget I maintain those packages ;)
<lfam>I am the unaccountable bureaucrat
<PotentialUser-15>how about the hibernation support ?
<lfam>Okay, one can read the deblobbing script itself
<nckx>PotentialUser-15: What about it?
<nckx>It exists, if that's what you mean ☺
<PotentialUser-15>is that supported ? on ubuntu I had to a bit of tweaking for that
<nckx>Might be rough, since I don't think many people use it.
<PotentialUser-15>probably because it usually doesn't work
<PotentialUser-15>outofthebox at least
<nckx>If hibernation doesn't work, that's usually the kernel's fault, distributions don't matter much.
<nckx>There's no ‘make hibernation good’ bit that some set & some don't.
<PotentialUser-15>on ubuntu it does, but I had to install a package or too and tweak fstab
<nckx>Which package or two?
<nckx>And which tweak(s)?
<nckx>lfam: Good.
<nckx>I trust you more.
<PotentialUser-15>nckx pm-utils worked for me but the ubuntu doc says you can fallback to swsusp if not
<PotentialUser-15>the tweak was to allow booting the swapped system instead of just booting
<PotentialUser-15>don't remember what exactly
<nckx>PotentialUser-15: See, (and this is not your fault) that makes no sense. You can't fall back to swsusp from pm-utils. pm-utils is a collection of bash scripts. swsusp is the part of the kernel that hibernates. They aren't similar things.
<PotentialUser-15>ah ya, i confused things
<lfam>Now that we fixed ALSA / Pulseaudio interaction on foreign distros, I am able to clear out my backlog of packages that use sound :)
<nckx>Perhaps they mean fall back to swsusp from TuxOnIce (a separate kernel patch that they used to ship).
<nckx>Guix does not have that patch because we don't patch our kernels, and because ToI is basically dead.
<nckx>But maybe that's not what they mean at all so I'll stop talking about it.
<PotentialUser-15>What I remember is that I had to install something, they try systemctl hibernate and if it doesn't, install swsusp and create a file for systemd to use it instead
<vincele>hello, it's been about half an hour that I cannot ping ci.guix.gnu.org (141.80.181.40)
<nckx>swsusp is part the kernel, you can't install it. uswsusp perhaps?
<PotentialUser-15>maybe
<nckx>vincele: That's strange, you should never have been able to ping it.
<nckx>It does not respond to pings.
<vincele>ah ok,
<nckx>Perhaps you were pinging your ISP's $something? (Captive portal, …, who knows.)
<vincele>I tried ping it because the DL rate was getting abysmal
<lfam>I'm curious, roughly where are you vincele?
<vincele>france
<PotentialUser-15>actually how does power managment work on guix ? I mean : what command / service puts the system at sleep / hibernation ?
<nckx>PotentialUser-15: Hibernation in Linux is the absolute suck. But, what there is, should work more-or-less out of the box on Guix Systems on most hardware. If you ever get so far as to test it, please file a bug if it breaks! I care a lot about hibernation in Guix.
<vincele>near germany
<PotentialUser-15>nckx ok i will
<jonsger>raghavgururajan: are you around
<nckx>PotentialUser-15: Same as everywhere else: echo mem > /sys/power/state (S2RAM), echo disk > /sys/power/state (hibernation).
<nckx>We're weird in many ways but that's the standard Linux interface, no Guixy weirdness ☺
<PotentialUser-15>ah ok
<PotentialUser-15>I didn't about this way, I come from systemd land
<PotentialUser-15>i know systemd like to cover everything
<nckx>PotentialUser-15: It would be advantageous to teach the Shepherd (our service manager, vaguely like our systemd) about suspend/hibernation too, so it can start/stop services required in some cases. But that's TODO, the core commands (possibly called from a custom script that, I dunno, unloads a problematic module) above just work.
<nckx>It would be sugar.
<PotentialUser-15>"so it can start/stop services required in some cases"
<PotentialUser-15>that looks quite important
<nckx>vincele: It does that sometimes, and... we don't really know why? :-/ That sounds terrible every time I say it.
<nckx>PotentialUser-15: Not really.
<nckx>Since it usually just implies the service is broken.
<nckx>(There are exceptions.)
<PotentialUser-15>:p
<PotentialUser-15>like the something-is-writing-to-disk kind of thing ?
<nckx>PM on Linux is just a sad mess that makes me sad to talk about.
<nckx>(I was just writing some PM code, and the whole thing is just gnarly -- not entirely Linux's fault to be fair.)
*vagrantc hasn't had working suspend on any laptop for quite some time
<PotentialUser-15>nckx i will trust you on that, but my experience would not go against
<PotentialUser-15>these days it seems way better though
<PotentialUser-15>at least sleep mode didn't fail anywhere for a long time in my experience
<nckx>PotentialUser-15: Both types of suspend should handle writing to disc just fine. Really, the only thing that *should* happen from user processes' view, is that suddenly time jumps forward 12h and all your network connections have timed out.
<nckx>But that's a big should.
<nckx>But nothing as basic as writing to disc should break anything.
<PotentialUser-15>ok, that's good to read
<nckx>Anyway, feel free to report suspend bugs on the Guix bug tracker, there's always a chance it's not entirely upstream Linux's fault ☺
<PotentialUser-15>To change topic, there is something i will need : nordvpn... Which is not free at all unfortunatly
<PotentialUser-15>But i see in the doc there should be a way to install it anyway, although it doesn't seem easy
<PotentialUser-15>but as far as I can read I could even manage to have updates
*nckx takes this opportunity to go back to kernel land, realises they had some pent-up stress that made its way to #guix, apologises.
<nckx>o/
<PotentialUser-15>thanks nckx
<PotentialUser-15>I know I could just use openvpn but then there is nordlynx
<PotentialUser-15>(nordlynx is a patched wireguard and no configuration is given for it, it's only via tha app)
<vagrantc>well, you could instead use wireguard
<vagrantc>which is packaged in guix
<vagrantc>or package whatever you want in your own channel
<vagrantc>supporting installing non-free software is off-topic for guix, though
<PotentialUser-15>so I can really manage to use say the nordvpn deb repo to get updates from it ?
<vagrantc>doubtful
<PotentialUser-15>that's really the only nonfree software I used I think
<PotentialUser-15>oh and gstreamer plugins of course
<PotentialUser-15>the bad and ugly
<vincele>nckx, it has been a long time since I got DL problems with ci.guix.gnu.org, but I think it started to get somewhat slow a few days ago, and then today it looked stuck
<lfam>vincele: The server is hosted in Berlin, so it *should* be fast for you. But we do get reports of slowdowns like yours often
<vincele>lfam, it usually is able to max out my slow DSL line, with DL being ~1MB/s (I sure would love fiber though)
<jonsger>vincele: what is your ISP?
<vincele>jonsger, free.fr
<jonsger>ah
<rekado>python-scikit-learn fails 108 tests
*rekado shrugs and upgrades
<leoprikler>it has achieved enlightenment
***atw` is now known as atw
<lfam>I'm surprised nobody has tried fixing that Python RCE
<lfam> https://bugs.gnu.org/46631
<lfam>Seems rather dire IMO
<lfam>Wow, package/inherit started to be used everywhere at some point
<lfam>All this inheritance is nasty :(
<lfam>Well, our Python 3.8 is a few versions behind so the upstream fix doesn't apply simply
<lfam>That's a problem