IRC channel logs


back to list of logs

<jlicht>nckx: I don't see how that would work with the php-fpm service as it is right now
<nckx>OK, that's what I feared. No biggums.
<andreas-e>nckx: I am just reading, not writing. narinfos are not cached by nginx, supposedly because they are also cached by "guix publish".
<nckx>A gig of RAM for everybody! 's Not like I'm selling shared hosting. Thanks jlicht.
<andreas-e>nars are cached by nginx, although they are also cached by "guix publish". But we reserve only 10GB for the nginx cache.
<nckx>andreas-e: Why re-cache like that? Does nginx cache extra gooder? I'm a big nginx fan, but that seems a bit much.
<nckx>It's slightly faster, I get that, but only slightly.
<andreas-e>I do not know. The gurus should answer that question. Maybe in the past we did not activate the "guix publish" cache? It comes at a price - when first asked about a nar, with caching "guix publish" will reply it does not have it.
<andreas-e>Then build it, and reply positively at the next request.
<civodul>andreas-e, nckx: on berlin we disabled nar{,info} caching some time ago
<civodul>we should do the same here
<nckx>Yeah, that's why I don't use it. guix publish → on-the-fly-lzip → nginx caching proxy.
<nckx>The first download is slow, but it actually happens, and since the first download is me, that's a big deal.
<nckx>But this sounds like it's doing both just so it can say sendfile on the box or something?
<andreas-e>In berlin.scm, it looks as if nar files are still cached.
<nckx>Or a forgotten remnant from before baking?
<andreas-e>civodul: Which caching? guix publish or nginx?
<jackhill>I just had a daydream about an incremental approach to getting a guile deamon by using guile as an extension language for C++ daemon and slowly moving things over rather than needing a fully functional guile daemon before switching.
<nckx>nginx I presume.
<andreas-e>I think nar files are still cached by nginx on berlin.
<andreas-e>I am getting confused. Did you not just establish that we should disable guix publish caching, and enable nginx caching? It sounds plausible to me.
<jlicht>nckx: you can still try your luck with the `file' field of php-fpm-configuration :-)
<nckx>jackhill: 😳 *slow blink* That is... poetic. Guile, the C extension language, preparing all this time to liberate us from C++.
<nckx>andreas-e: Me? I didn't mean to suggest anything.
<nckx>My preferences don't dictate squat.
<nckx>jlicht: Hehm. You caught me... I was...
<andreas-e>Since I used to be the only person in the world using bayfront as a substitute server, I find your suggestion compelling ;-)
<jlicht>you could even generate parts of by looking at default-php-fpm-config for inspiratation; I _think_ you might even get with generating two configs as g-exps using default-php-fpm-config: and then creating a new 'mixed-file' that simply does "include #$gexp1\n" "include #$gexp2" as the `file'
<jlicht>parts of it**
<jlicht>ah never mind; you'd still be stuck with the "www" pool name. Enjoy all the php goodness either way nckx!
<zimoun>rekado: every when indexes new bugs?
<nckx>I thought it was 30s.
<darkpsi>icu4c has changed there download url will do a PR when i get my system up and running.
<andreas-e>Enough for yesterday, see you again!
<roptat>what's the variable to make guix print backtraces without trunkating them again?
<zimoun>nckx: I search with ’author:me’ the last bug report I sent yesterday evening is not on the list. But it is available at Maybe the issue comes from the number of entries returned, maybe there are limited.
<nckx>The worker is (very; too?) busy working but I don't know enough about Mumi internals to say.
<civodul>nckx: yeah i was referring to nginx caching
<nckx>zimoun: Search being broken is a separate thing though. rekado knows.
<nckx>It's not related to fetching new mails.
<nckx>civodul: And disabling or keeping?
<zimoun>nckx: thanks
*nckx .oO Why am I excited by a DNS server update? But Knot 3 looks cool.
*zimoun says zZz!
***ChanServ sets mode: +o nckx
<tribals>hello again!
<tribals>Does `gnu/packages/firmware.scm` export `make-arm-trusted-firmware`? I'm getting error "error: make-arm-trusted-firmware: unbound variable" "hint: Did you forget `(use-modules (gnu packages firmware))'?" when trying to use it. And I also did `(use-modules (gnu packages firmware))` What I'm doing wrong?
<nckx>tribals: No (you can tell by the use of define*, not define-public, and that it's not explicitly exported in the (define-module ...) header.)
<nckx>You're doing nothing wrong, the variable's not exported & the hint is misleading.
<nckx>Use (@@ (gnu packages firmware) make-arm-trusted-firmware), I guess.
<tribals>nckx: thanks. Noob questiong, huh: what does it mean extra "star" to `define`?
<nckx>tribals: Are you familiar with lambda*? It's to define what lambda* is to lambda: syntactic sugar to easily define procedures that take #:keyword #:arguments.
<nckx>I think it's equivalent to (define (lambda* ...)).
<tribals>nckx: thanks
<nckx>Ah, it's in the Guile manual: (guile)lambda* and define*.
<nckx>make-arm-trusted-firmware takes an optional argument, which is also provided throug lambda*/define*.
<roptat>I wonder if we could improve the hints a bit, to not suggest a module where the variable is not exported
<roptat>or say "this module contains it, but it's not public" but that might be a bit confusing too
<xelxebar>efraim: Did that build ever finish? Kind of worried that the board caught on fire and you died in a maelstrom of flames or something.
<apteryx>sneek: later tell civodul this pretty much sums up my migration away from helm to ivy:, in case you're interested.
<sneek>Will do.
<apteryx>this link may be of interest as well:
<Helpester>Hello im having some problems installing can someone help me out?
<Helpester>When running the guix system init command it throws an error saying that the ghostscript-9.27 file cant be opened, is there a way to fix that?
***apteryx_ is now known as apteryx
<bdju>I'm in need of the debugging symbols for dino, but it seems there's no debug output for the package
***slyfox_ is now known as slyfox
<mfg>Hi, in which file should 3d printing related software be added?
<mfg>i didn't find anything that sounds like the right place
<cbaines>mfg, the gpx looks 3d printing related, and that's in the engineering module
<mfg>Ah, i will use this module then :) Thx
<xelxebar>mfg: FWIW, I tried searching for '3d print' and that gpx package pops up. guix search has helped me out in the past for finding where to place stuff.
<potato>Has anyone else experienced a bug in the GuixSD graphical installer that once you get to the partitioning section and choose your layour, creating a >1tb partition is going to start the installer over from the beginning
<xelxebar>bdju: Yeah. I've noticed that debug outputs are relatively rare as well. Snooping the defs for packages that do present a debug output don't seem to be doing anything special, so you could potentially try adding debug to dino's outputs and try building locally...
<xelxebar>bdju: And bumbling around in guix/build/gnu-build-system.scm, it looks like debug symbols are automatically stuffed in a debug output if the latter exists.
<xelxebar>Feel like this should be the default.
<mfg>xelxebar: i dind't know i can search words with spacing...
<mfg>is there a good way to find out how the guix package name given the ubuntu package name? i'm trying to findout if the package named 'libwx-perl' exists in guix. Is it part of wxwidgets?
<cbaines>mfg, the Debian/Ubuntu source packages tend to match up better with Guix packages, so I'd check what the source package is for libwx-perl
<cbaines>if it's wxwidgits, then it's probably part of wxwidgets in Guix
<mfg>thx for the tip :)
<andreas-e>cbaines, I just ran a full "guix gc" on bayfront. Only 250GB are freed, so we need to delete gc roots.
<sneek>Welcome back andreas-e, you have 1 message!
<sneek>andreas-e, nckx says: I took the liberty of compressing 55G of /var/log/nginx to save 53G, in case you want to keep them. Deleting them will save the rest.
<cbaines>andreas-e, I'd suggest dropping the configured TTL sooner rather than later, as it'll take a week for the existing gcroots to expire
<andreas-e>Then is one week enough? I suppose for the Guix data service, a small time is enough.
<andreas-e>But if we ever want to use it as a substitute server also...
<cbaines>I don't think there's much harm in dropping the TTL, I'm not quite sure what effect it will have on substitute availability
<cbaines>but if it results in bayfront not running out of space, that might even help with substitute availability
<andreas-e>Since ttl only becomes valid when the item was not downloaded, it should indeed not make much difference if we lower it. It could just happen that bayfront being too popular as a substitute server will result in ttl having no real effect.
<andreas-e>We could also experiment with dropping the "guix publish" cache altogether in favour of a larger nginx cache; when we start offloading again, we will have spare CPU power on bayfront. And it avoids the annoying miss on first request.
<peanutbutterandc>Question: How do I install enigmail add-on in icedove? (Or any other thing to use GPG)?
<nly>sneek: later tell roptat peertube works, thanks.
<sneek>Will do.
<brendyyn>peanutbutterandc: probably same way you do in thunderbird?
<peanutbutterandc>brendyyn, extensions seem to be blocked.....
<andreas-e>sneek: later tell nckx Thanks for the compression, indeed I think we do not really need these logs.
<andreas-e>sneek: botsnack
<brendyyn>peanutbutterandc: I have enigmail in mine
<peanutbutterandc>brendyyn, is it there by default? o.O I'm using icedove 68.12.0. (I am a beginner at this)
<peanutbutterandc>.... non-web-based-email-clients....
<brendyyn>i think i had it installed since years ago
<brendyyn>in what way are they blocked?
<peanutbutterandc>brendyyn, Hmmm.... Add-ons has no search bar, and it only takes to the fsf website.... But I can see that there is a 'add add-on from file' option there. Perhaps I should try that.
<peanutbutterandc>I think I remember icecat also having blocked mozilla addons
<brendyyn>peanutbutterandc: get it from here
<peanutbutterandc>brendyyn, Hmmm.... I'll try that thank you very much.
<brendyyn>peanutbutterandc: basically, thats the way to do it, but we dont link to thunderbird's addons page because it also has proprietary stuff. someone would have to go to the effort of creating a curated list of free addons
<peanutbutterandc>brendyyn, I wonder if the mozilla people could be convinced to create something like It also seems that thunderbird 78 has OpenPGP support built-in: I looked at the package definition, but gnuzilla.scm looks scary
<brendyyn>i think icedove is based of the long term release version
<peanutbutterandc>thunderbird 78 seems to be an extended support release (ESR)
<brendyyn>looks like its quite recent
<mfg>is there a phase which also cleans absolute include paths in c/c++ source files? how do i do it if such a phase doesn't exist?
<andreas-e>mfg: There is no such phase; where did you see such strange software?
<andreas-e>You can always add a phase that uses "(substitute*" on the header files; it can be as involved as you wish, from just replacing one occurrence in one file to replacing several inclusions in all files that you select according to some regular expression.
<mfg>slic3r is the software i'm trying to package
<peanutbutterandc>Hey there, what does it mean when a build fails at 'unpack phase? I can unzip the file from my normal shell just fine. But in the build it exits with error 127
<andreas-e>I would suggest to do a "grep" on existing packages.
<mfg>that substitute* thing is what i did now, i luckily found some examples in the same file :D
<andreas-e>Oh, I see!
<andreas-e>I have also started packaging slicr3.
<mfg>Oh, how does/did it go?
<andreas-e>Extremely complicated, and then I gave up ;-)
<andreas-e>It has a dependency around opengl, which is difficult to build. Nix ships a configuration file with it that has been created on a developer's machine, and it leaks the processor type (such as "skylake") into the file. It did not inspire much confidence.
<andreas-e>I will have a quick look again, give me a second.
<andreas-e>I have this line in a phase just before 'configure:
<andreas-e>(substitute* "xs/src/libslic3r/GCodeSender.cpp" (("\"/usr/include/asm-generic/ioctls.h\"") "<asm-generic/ioctls.h>"))
<mfg>i substituted with the complete store path, does that make a difference? (just out of curiosity)
<andreas-e>The stumbling block was perl-opengl, from mirror://cpan/authors/id/C/CH/CHM/OpenGL-0.70.tar.gz .
<andreas-e>It is easier to not substitute with the complete store path. Actually then it could even be done in a source snippet instead of a phase. I had more things to do in the phase anyway...
<mfg>source snippet means patch?
<andreas-e>No, it is a scheme expression to modify the source, whereas the patch is a "fixed" change taken from a separate file. You can look for "snippet" in the source.
<mfg>okay, i will look into it
<mfg>I have problems even before i need perl-opengl... how does your package definition look like? I _thought_ i had to split the slic3r::xs module into it's own package because the main Build.PL didn't build it automatically, now i think i have some missing referencecs
<andreas-e>mfg: Just as a convenient way of sharing, I pushed the branch wip-ae-slicr, so you can see where I am.
<andreas-e>perl-opengl wants a file utils/glversion.txt, after telling me "In order to test your GPU's capabilities, run this make under an X11 shell". But of course, we do not want to have this depend on the build machine.
<andreas-e>mfg: I think this must be why I am setting the PERL5LIB environment variable.
<andreas-e>And I issue a call to "perl" inside the xs directory.
<mfg>yes that makes sense :D
<andreas-e>I am off to something else, but your interest provides some motivation to get back to it (with the covid lockdown, I lost access to the 3d printer at work, and moved on to something else).
<mfg>bye :)
<andreas-e>Your comments are welcome, of course!
<tribals>Hi, folks!
<andreas-e>Hello, tribals!
<tribals>How to get appropriate toolchain for cross-compilation with `guix environment --ad-hoc`? Currently, I did it by specifying package that depends on right triplet, like so: `guix package u-boot-pinebook --ad-hoc <packages-that-i-need>`. I tried to search for `aarch64-linux-gnu-gcc` or so but found nothing
<andreas-e>cbaines: I have just reconfigured bayfront with a guix publish ttl of 7. Actually maybe we should replace it by 1 (or even 0?) to make sure the cache gets emptied, and then disable this caching altogether.
<tribals>The reason why I want so is that I suppose that guix have needed cross-compiler in place, and I don't want to download it, say, from linaro
<leoprikler>cross-compiling in guix is different from building a package for <triplet>
<leoprikler>IIRC the latter uses qemu-binfmt and builds the entire toolchain as if running on <triplet>
<leoprikler>for cross-compiling, you'd want to make use of (cross-gcc ...) etc. defined in cross-base.scm
<tribals>does it mean that in order to build any package for triplet different to the host, I need qemu-binfmt setup?
<leoprikler>depends on what you mean with "build any package"
<leoprikler>if by "build" you mean "build in the guix sense", then yes
<leoprikler>if by "build" you mean "./configure && make", then no, you can set up cross-gcc to do that
<leoprikler>I think you can also `guix pack` a cross-compiled package, but don't quote me on that
<leoprikler>as to how you'd use that with guix environment: you can load scheme files (pre/post --ad-hoc) with -l
<tribals>I think, I need both. The idea was to enter first `guix environment` to tune up build and ensure it is working, for which I need proper cross-gcc (to invoke it "by hand"), and then create valid guix package to be able to use it in my `system.scm`
<leoprikler>yep, for the latter you'd need qemu-binfmt
<leoprikler>for the former, you just need to invoke (cross-gcc ...) et al. at some point, depending on what you need
<leoprikler>e.g. (cross-gcc "aarch64-linux-gnu") (cross-binutils ...) ...
<tribals>leoprikler: It worked, thanks
<raghavgururajan>Hello Guix!
<andreas-e>I am having trouble configuring offload from bayfront to harbourfront.
<andreas-e>It returns: hint: Make sure `guile' can be found in `$PATH' on the remote host.
<andreas-e>But the "ssh ... guile --version" test works (and shows a different guile version, 3.0.2, then what the previous line indicates, 3.0.4).
<andreas-e>The same with offload to other machines, by the way. Everything is freshly reconfigured and should be compatible.
<str1ngs>andreas-e: hello is this when using guix offload test ?
<str1ngs>also make sure you test the guile with . ssh $machine guile --version . ssh does not use interactive prompts so this can be a factor. maybe your ssh example used that, but just to be sure.
<rekado_>andreas-e: in (sysadmin build-machines) you’ll see that I extend the session-environment-service-type to set GUILE_LOAD_PATH and GUILE_LOAD_COMPILED_PATH globally on the build machines that are (directly) connected to berlin.
*wigust uses ~/.bashrc snippet in machines for offload: if [ -n "$SSH_CLIENT" -a "`type -P guile`" ]; then source /etc/profile; fi
<guix-vits>sneek later tell peanutbutterandc: maybe it's a "zip-bomb"? re: "Hey there, what does it mean when a build fails at 'unpack phase? ... error 127".
<sneek>guix-vits, you have 1 message!
<sneek>guix-vits, vagrantc says: no, i don't use a display manager ... prefer to use sway directly. :)
<sneek>Got it.
<guix-vits>sneek: botsnak
<str1ngs>wigust: I need to improve my $- check it's a PITA on foreign distros lol
<guix-vits>sneek: botsnack
<andreas-e>Thanks for your suggestions! Yes, the "ssh" test did work. And now something else fails (maybe I restarted the daemon in between):
<andreas-e>guix offload: sending 1 store item (0 MiB) to ''...
<andreas-e>exporting path `/gnu/store/qj9p7fpc2jl8v1c4k41wxvzds3i66zcx-export-test'
<andreas-e>guix offload: error: program `/gnu/store/64yyvnga4gzyllwwb8m72y2m43f0a2lm-guix-1.0.1-10.41b4b71/bin/guix' failed with exit code 1
*str1ngs stares at guix-vits
<andreas-e>wigust, I also have a similar snippet in .bashrc. I think it is part of the Guix template for .bashrc.
<guix-vits>str1ngs: Hello, what happen?
<str1ngs>guix-vits: Hello, check this out. notice the mode in the modeline.
<andreas-e>Yes, I did restart the daemon on the offloading machine. So this had some effect.
<rekado_>andreas-e: hmm, that’s an odd error. Haven’t seen that before.
<roptat>hi guix!
<sneek>roptat, you have 1 message!
<sneek>roptat, nly says: peertube works, thanks.
<roptat>cool :)
<raghavgururajan>sneek, later tell nckx: Hey, sorry to bother you with this again. I have re-created my ssh key with ed25519. Could you update it on bayfront please?
<andreas-e>rekado_ et al: I have put the complete result of "guix offload test" here:
<andreas-e>The same happens when trying to offload to dover or milano. As if the offloading machine were at fault.
<str1ngs>is there trust setup?
<andreas-e>raghavgururajan: Could you send a pgp signed message to It would be good to have some more authentication than a message on IRC...
<rekado_>andreas-e: I cannot see anything at the pastebin URL you’ve sent.
<rekado_>just a description of what this pastebin does.
<rekado_>(using Icecat here)
<andreas-e>rekado_: Me neither with icecat, but it works with ungoogled-chromium...
<andreas-e>Have you got another pastebin suggestion?
<raghavgururajan>andreas-e: Oh thanks, I didn't know that email-ID existed. Btw, I also lost my GPG private keys while wiping the disk.
<rekado_>andreas-e: no, I only ever upload text files to my “server”… I’ll fire up ungoogled-chromium
<andreas-e>raghavgururajan: Well, that sounds complicated! Maybe authenticate to nckx by sharing secret information only the two of you know ;-)
<rekado_>andreas-e: hmm, this doesn’t say much. I think the error is rather vague. We don’t see the full command line, nor do we know if this is executed locally or on the remote.
<raghavgururajan>andreas-e: I generally use [1] [2] [3]
<raghavgururajan>They all work well on IceCat.
<andreas-e>rekado_ this is a general problem with "guix offload test"; it does not tell where which command is executed.
<raghavgururajan>andreas-e, Ah, yeah. Thanks!
<andreas-e>Maybe civodul knows what to do.
<daemonspudguy>I never got GuixSD working. It hung on the early boot screen.
<daemonspudguy>I do use an AMD A8-7600, if that would matter.
<civodul>hi andreas-e & all :-)
<sneek>Welcome back civodul, you have 1 message!
<sneek>civodul, apteryx says: this pretty much sums up my migration away from helm to ivy:, in case you're interested.
<civodul>apteryx: very interesting, thanks!
<andreas-e>civodul, I cannot offload from bayfront to anywhere else (harbourfront, dover, milano):
<daemonspudguy>Any help?
<andreas-e>raghavgururajan: Thanks for the pastebin links, I just put them to good use.
<civodul>andreas-e: you checked ACLs, right?
<raghavgururajan>andreas-e: Glad!
<civodul>what does "guix describe" say on bayfront?
<andreas-e>No. As far as I know, nothing changed. But I will check.
<raghavgururajan>daemonspudguy: Do you see any error message?
<daemonspudguy>No. It just hangs.
<civodul>andreas-e: guix-daemon 1.1.0-26 had a bug related to 'guix authenticate', which could have this effect
<daemonspudguy>I gave up and installed NixOS.
<civodul>andreas-e: so you can check "ps aux| grep guix-daemon" on bayfront ↑
<andreas-e>civodul: I did reconfigure bayfront and harbourfront and restarted the daemon on bayfront: 1.1.0-27.
<raghavgururajan>daemonspudguy: In which step it stucks? Grub or after kernel load?
<civodul>daemonspudguy: i missed the beginning of the discussion, but whether or not you give up, we'd welcome your experience report on, so we can improve on it
<daemonspudguy>After grub, during kernel load as far as I could tell
<raghavgururajan>daemonspudguy: I see. If you still intrested in setting-up Guix System, let us know.
<civodul>andreas-e: could you restart guix-demon on harbourfront then, because it's running guix-1.0.1-10.41b4b71 according to the paste above, which is "old"
<daemonspudguy>I am, but my computer started making screeching sounds. I think it's the fans.
<andreas-e>1.0.1-10 on harbourfront. Okay, I restarted there to 1.1.0-17.
<andreas-e>I supposedly reconfigured harbourfront today, but my ssh session broke down in the middle; nevertheless, "guix system list-generations" still claims a reconfiguration today. Very weird.
<raghavgururajan>> daemonspudguy‎: After grub, during kernel load as far as I could tell
<raghavgururajan>I see. How long it took? Sometime happens with some hardware.
<civodul>andreas-e: the next step is unsurprisingly "sudo strace -p $(pidof sshd) -s 500 -o log" on harbourfront, then "guix offload test" on bayfront
<andreas-e>Same problem.
<andreas-e>Okay. Latest daemon 1.1.0-27 on harbourfront now.
<andreas-e>civodul: This was necessary, there is progress. Now "guix offload test" complains about an unauthenticated guix key, and I can work from there.
***ezzzc671 is now known as ezzzc67
<andreas-e>Good news! Offloading from bayfront to harbourfront is working again. I am having a look at the other machines.
<daemonspudguy>My computer is up and running again!
<andreas-e>daemonspudguy: With the Guix system?
<daemonspudguy>No, in general. The fans were screeching
<daemonspudguy>So I blew the dust out
<daemonspudguy>I'm gonna see if I can get Guix running in Virtualbox.
<andreas-e>civodul: Thanks for the debugging help!
<raghavgururajan>daemonspudguy: Just letting you know that you will not get help in #guix for virtualbox, as it is a non-free software. But you are welcome to use QEMU or GNOME-Boxes, for which you will get help here. :-)
<daemonspudguy>All right, I'll go install Libvirt.
<daemonspudguy>Wait, it's the open source edition
<daemonspudguy>Of vbox
<raghavgururajan>daemonspudguy: FYI, we do provide ready-made QEMU image,
<raghavgururajan>daemonspudguy: I am not sure about that. IIRC, VirtualBox was not a free-software. :-)
<daemonspudguy>raghavgururajan: the base package is GPL v2
<raghavgururajan>daemonspudguy: Is it? Never mind then.
<raghavgururajan>daemonspudguy: Cool! Thanks for the info.
*raghavgururajan wonders why it wasn't listed at
<daemonspudguy>Probably because some distributions bundle it with the proprietary Extension Pack, even though it is 100% optional.
<daemonspudguy>Or someone never got around to it.
<raghavgururajan>That makes sense.
<andreas-e>cbaines, nckx: I am restarting cuirass on bayfront. With offloading enabled only to harbourfront, package building should be sufficiently slow to not run into space problems.
<nckx>Good morning Guix.
<sneek>Welcome back nckx, you have 2 messages!
<sneek>nckx, andreas-e says: Thanks for the compression, indeed I think we do not really need these logs.
<sneek>nckx, raghavgururajan says: Hey, sorry to bother you with this again. I have re-created my ssh key with ed25519. Could you update it on bayfront please?
<nckx>andreas-e: I'm pretty sure Guix has a log rotation (+compression) service. I'm less sure if that deletes old logs too.
<andreas-e>nckx: And it might not be activated on bayfront.
<andreas-e>How can I set the time with guix? I just added the ntp service to one of my machines, but the time is still completely off.
<raghavgururajan>nckx: Morning! o/
<nckx>ntp isn't designed to hard-reset your time, all such features are considered deprecated hacks. Just use date to set it?
<nckx>I see we offer a ‘allow-large-adjustment? #t’ switch for ntp.
<nckx>Upstream doesn't like that but it might work.
<nckx>raghavgururajan: I've reconfigured bayfront. Forgot if I needed to do more last time. Try logging in.
<andreas-e>I had hoped to do it just once. I thought there was a flag for forcing an update by hand.
*alextee[m] is stacking packages to send again
<raghavgururajan>raghavgururajan@bayfront ~$ echo "nckx is a good lad"
<raghavgururajan>nckx is a good lad
<nckx>andreas-e: Try openntpd + allow-large-adjustment?
*nckx nckxsnack.
<raghavgururajan>Is openntpd-service-type better than ntp-service-type?
<apteryx>andreas-e: ntp can take a few minutes to do the initial sync
<andreas-e>I just used date for now with a few seconds off for testing. Maybe the small error will be adjusted now.
<apteryx>it's supposed to be able to make large adjustments out-of-the-box
<apteryx>out of the box*
<nckx>raghavgururajan: It's smaller, and does less, and is probably a better choice in most ‘just set my time’ scenarioes. openntpd is OpenBSD's ntpd and has far fewer features than the ntpd ‘reference implementation’. OpenBSD people always think their software is better.
*nckx stans chrony.
<raghavgururajan>nckx: Cool!
<raghavgururajan>> OpenBSD people always think their software is better.
<raghavgururajan>LoL, is it though? empirically?
<nckx>Define ‘better’ 🤷
<raghavgururajan>With regards to performance, complexity, bugs etc..
<andreas-e>I do not think my ntp service is working. But the machine may be firewalled.
<nckx>It's probably the least accurate (with chrony being by far the most). It's much smaller, therefore less attack service, and explicitly aims to be ‘good enough’ for people who just want the time to look right, not sync machines with microsecond accuracy.
<apteryx>How many log rotations does rottlog keeps? I have guix-publish.log.34; that seems like a lot
<raghavgururajan>nckx: Fair enough
<nckx>All our ntpds have a client that gives extremely detailed info. I don't remember what they're all called.
<apteryx>andreas-e: any logs in /var/log/messages regarding ntp?
<nckx>Eh, why the mass key file rename in maintenance.git?
<nckx>Ohkay, it's been reverted.
<andreas-e>I reverted it. Somehow I thought it would have no effect, but of course it does.
<andreas-e>I do not like the redundant "-export" infix. And some domain names are wrong.
<andreas-e>There is no need in registering the domain name in the file name, since the key is independent of it.
<alextee[m]>anyone familiar with autotools?
<alextee[m]>If this token and others are legitimate, please use m4_pattern_allow.
<alextee[m]>how do i enable this option in a package deifnition?
<andreas-e>I also think there are remnants of hydra which could be deleted, but it requires a bit more care.
<andreas-e>alextee, I think we need more context.
<raghavgururajan>nckx: I am also creating my GPG keys with ed25519. Based on, what options should I choose for 1st and 2nd step? I guess for 1st step, selection 9 is what I want.
<raghavgururajan>Oh wait second box has Curve 25519
<alextee[m]>it should find gtk2, no idea why it fails at the other test
<andreas-e>(9) looks good. Then most people use curve25519.
<alextee[m]>this is what im trying to build
<andreas-e>Apart from that, NIST is the American, Brainpool the German secret service.
<raghavgururajan>andreas-e: Thanks!
<andreas-e>The Brainpool curves have been created in a verifiably pseudorandom way. So if in doubt, I would prefer them to the NIST curves.
<nckx>Pff 🙂 It's been years since I created mine, and I use subkeys of my older RSA4096 key. That master key is kept off-line in a secret location. Under that master key I have three subkeys: one (ed25519) can only authenticate, one (cv25519) can only encrypt, one (ed25519) can only sign.
<andreas-e>alextee: I do not know. Maybe it will be enough to use glib-or-gtk-build-system?
<nckx>Maybe nowadays I'd go with an elliptic master key if that's possible. Doing so now would change my key fingerprint and won't happen any time soon.
<alextee[m]>i'll just give up on this
<andreas-e>apteryx: Thanks, yes, I see messages such as "localhost ntpd[30306]: Soliciting pool server"; but the time remains off by 1s, which is what I managed to achieve by hand.
<andreas-e>alextee: Last commit 10 years ago...
<alextee[m]><andreas-e "alextee: Last commit 10 years ag"> it's a deprecated standard but i wanted to test DSSI support in zrythm
<alextee[m]>the other DSSI synth we have "whysynth" seems to be broken
<raghavgururajan>nckx: Cool!
<andreas-e>alextee[m]: I tried the following:
<andreas-e>In, there is a case distinction with Linux) in one case.
<andreas-e>There I replaced
<andreas-e>ACLOCALARGS='' by
<andreas-e>ACLOCALARGS='-I /gnu/store/sqypvb555xv0nsq3yv2959m3w5f21cal-gtk+-2.24.32/share/aclocal'
<andreas-e>where the hash is the result of "guix build gtk+@2"
<alextee[m]>oof, i can get that through th einputs hopefully
<andreas-e>This allows to find the AM_PATH_GTK_2_0 macro. But then it complains about a missing AM_PATH_GTK. From what I understand, that comes from gtk+@1, which is prehistoric and not available in Guix.
<andreas-e>The case looks rather hopeless.
<alextee[m]>yeah i wont bother
<raghavgururajan>nckx: Is it possible to change creation data in gpg keys?
<apteryx>do you mean expiry? yes
<nckx>Expiry: kind of, in that you add a new signature (with begin & end dates) every time you extend the validity. So don't bump it every week; it's tedious for you but also bloats everyone's copy of your key. Keyservers are append-only.
<nckx>raghavgururajan: It's also no big deal if your key ‘expires’: you can extend it just as well if you forgot.
*nckx AFK for dinner.
<apteryx>wow. QEMU weights 745 MiB in the store.
<raghavgururajan>apteryx, nckx: Not expiry. I set not to expire. I was looking for to change creation date to 2020-01-01.
<raghavgururajan>*purely cosmetic*
<raghavgururajan>nckx: I already lost 5 or more gpg keys which were set to 'no expiry' and were also uploaded to key servers.
*raghavgururajan has plans to preserve/back-up this new key
<mihi>raghavgururajan, probably there is a better way, but I would probably just LD_PRELOAD libfaketime. However, I have no idea if that is packaged in guix and how easy it is to LD_PRELOAD a lib in guix :P
<raghavgururajan>mihi: I see. Thanks!
<andreas-e>Concerning offloading, do I see it correctly that sources (fixed output derivations?) are built locally?
<mfg>What can be the problem if guix fails downloading a tarball with error 500 but directly curling it works? the link guix tries seems to be a redirect to a new location, is that a problem?
<apteryx>mfg: perhaps try 'guix download' on that link
<apteryx>I'm curious to know if that fails also
<mfg>apteryx: interesting, that just works :|
<mfg>Ah, i see the problem, it isn't guix that's trying to download the package but a script inside that package...
<nckx>raghavgururajan: Yeah, it's trivial to ‘forge’ a date before you sign. Exersise for the reader; I'm sure you've searched by now 🙂
<nckx>I've lost 1 GPG key and blushed for a month. 5... wow. Time to hide a USB drive under the floorboards.
<nckx>andreas-e: Yes, I think that's right.
<pineapples>Hello. Does anyone else have knot-3.0.0 fail to build with the ninja subcommand failed error?
<apteryx>andreas-e: all the inputs of an offloaded derivation are first collected on the local machine, then transferred to the remote build machine. Source derivations can be offloaded as well (these are not fixed output derivations), but will be transferred back to the local machine.
<nckx>raghavgururajan: I don't recommend ‘no expiry’: expiry doesn't affect you, the owner, at all: your private key trumps that nonsense. It's only a signal to others that you haven't bothered to renew it for n years, so something might be up. If you lose your revocation key, a key that doesn't expire is valid forever.
<apteryx>andreas-e: so what you said is correct (fixed output derivations are fetched from the local machine)
<nckx>pineapples: I'm running Knot 3. Could you share more info? ‘ninja subcommand failed’ sounds like ‘*** make: error foo’: a sign that something went wrong earlier.
<andreas-e>Okay, thanks! After an aggressive "guix gc", that means that my offloading machine is doing most of the work for a start.
<nckx>I know building Knot's libzscanner can be quite stressful.
<nckx>andreas-e: It means build nodes shouldn't need Internet access, which is quite cool.
<andreas-e>Indeed, good point.
*nckx has about a minute of lag judging by logs.guix. A good night.
<pineapples>nckx: Sorry, I meant knot-resolver. It fails at lib/xxxx@@kres@sha/ with an error "too few arguments to function 'dnssec_sign_verify'".
<nckx>Ooh... does in inherit/take Knot as an input? If so it probably needs updating. I don't use it, only Knot.
<pineapples>nckx: Yes, knot is listed as an input of knot-resolver
<nckx>knot-resolver 5.1.5 changelog entry: ‘support building against Knot DNS 3.0 (!1053)’.
*nckx tries bumping it.
<nckx>Let's hope it doesn't bump back.
<nckx>Sites with a ‘secure.’ subdomain are weird.
<pineapples>nckx: How is that?
<raghavgururajan>> If you lose your revocation key, a key that doesn't expire is valid forever.
<raghavgururajan>Yeah, I made that mistake for my past 5 or more keys, which are still in key-servers :-/
<nckx>Very ‘downloading PHP in 2005’ to me. ‘Ooh, SSL!’
<nckx>Now it just implies the rest of your site is broken.
<pineapples>Lol! That reminds me of antivirus software companies' 'secured connection online payment processing' stickers with padlock icons all over sketchy websites. Anyway, I hope the new release didn't break my configuration. I've been non-stop tweaking my Guix configuration, and I feel burned out hahah
<mfg>How do i specify a specific package version in my input list? on the cmd line package@version is used, but this seems to be wrong?
<nckx>mfg: When writing Scheme you use Scheme variables, so it depends what the (define-public <foo> (package ...)) line is. Usuall foo-1.x, although sometimes foo/fixed or foo-next etc. is used.
<nckx>guix prints the file:line location of packages, so you can check
<mfg>Okay, now i know where to look :)
<nckx>foo@1 is a valid Scheme variable, but it's just not how things happened.
<mfg>andreas-e: have you tried building wxperl? I get the error described in this issue:
<mfg>It seems like something strange is going on
***ChanServ sets mode: +o lfam
<guixquestions>Question: GuixSD Installer: Upon selecting to partition entire hard disk the installer bounces me back to the beginning of the installer. 2TB HDD. This doesn't happen with a 500GB HDD
***lfam sets mode: +b *!*@gateway/shell/
<guixquestions>Is this an issue with GuixSD or the installer intself? Should I simply attempt a manual installation instead?
<guixquestions>Kind of an interesting bug
<lfam>Hm, that sounds annoying guixquestions
<guixquestions>lfam slightly XD
<lfam>Unless somebody else has better advice, I would recommend the manual installation process
<guixquestions>yeah that's what I'm thinking
<lfam>I'm unfortunately not very familiar with the installer
<lfam>We can help you through the process here
<guixquestions>ok I appreciate that, going to attempt full disk encryption as well with libreboot
<guixquestions>looking into the documentation as I have no clue how to install this manually
<andreas-e>mfg: No, that does not ring a bell at all. Is it an input that is still missing, or part of the lic3r package?
<lfam>guixquestions: The guide found in our manual should be complete and sufficient but, like I said, we are here to help :)
<guixquestions>Thanks again I appreciate that
<lfam>It's a bit more involved that installing e.g. Debian
<lfam>But hopefully not too bad
<mfg>andreas-e: it's a necessary dependency of slic3r, but looking on CPAN makes me think it hasn't been updated since 2017, that would explain this error, the last officially supported wxwidgets i found in the docs is 2.8.x
<mfg>this verison exists in guix i' trying it right now
<andreas-e>An option to consider is to look at what other distributions have done; in particular Nix and Debian; and what patches they have added.
***ChanServ sets mode: +o nckx
***nckx sets mode: -b *!*@gateway/shell/
***ChanServ sets mode: -o nckx
<mfg>andreas-e: i got past this error. Now it says this: PM: Error: Unable to initialize GTK+, is DISPLAY set properly? does guix unset this variable in the build environment?
<andreas-e>I do not think we have a DISPLAY during building.
<andreas-e>What is it trying to do?
<mfg>it's trying to run some tests, i'm currently searching how to disable them
<andreas-e>I think we do have a few packages where something like a "headless X server" is used during the build; but I have forgotten the package name.
<nckx>grep -w gnu/packages for that for example school times.
<nckx>Try that before disabling them; suprisingly many graphical test suites run just fine.
<mfg>i will try it out
<andreas-e>Strange names come up with grep!
<andreas-e>Did we not have a separate package xvfb in the past?
<nckx>AFAIK ‘grep Xvfb gnu/packages/*scm’ returns almost exclusively test suites using this technique? Which strange names?
<nckx>andreas-e: Perhaps you mean xorg-server-for-tests. If there was one I'm not aware of that.
<nckx>mfg: By the way, use ‘,xorg-server-for-tests’, not ‘,xorg-server’, if this works. Helps keep xorg-server off of core-updates.
<andreas-e>Yes, I just found it surprising the package was not called Xvfb. Why do we have a separate one, which seems to be just an older release?
<nckx>Because Xvfb is not a separate package, I guess.
<mfg>so adding xorg-server-for-tests to the native-inputs should be sufficient right?
<roptat>after about 2 hours of hard work, I managed to work around the build issue of my antlr4 package, I'll send a patch after I fix the description etc :)
<andreas-e>mfg: Look for "Xvfb" in gnome.scm. Apparently it is common to do more.
<jlicht>hey guix!
<mfg>Nice! It passed those tests even though the screen got spammed with unknown keysym errors ^^
<nckx>Hi jlicht. Nice mfg!
<nckx>mfg: They all do that, dunno why, commit away.
<jlicht>This might be a silly question, but does anybody have a snippet of code that I could use to have a guix-installed emacs-geiser be lazy-loaded? The specific autoloads for geiser are a bit... interesting, as they seem to simply load everything eagerly :/
*nckx -> zzz. Good night!
<mfg>andreas-e: up until now the package builds -- without the gui. I always get the message "prerequisite Wx not found" even though it's added to native-inputs and inputs.
<mfg>i don't know much perl, so this might also be a search path problem?
<PotentialUser-39>Hi, im having some issues booting in my new system, is there a way to boot from the live iso to my system ?(im an UEFI user))
<tribals>I's me again (. .;)
<raghavgururajan>PotentialUser-39: Did you use grub-efi-bootloader in your system config?
<tribals>I finally give up trying to build fully open source bootloader for my ARM board. Now I'm trying just to take vendor provided binary and pack it as package in order to be used in `system.scm`. The problem is that it needs to be cut by one block (512 byte). I'm trying to do it with modify-phases using dd. Here is my take: (sorry for gist, debian paste is down). It doesn't work. I can't figure
<tribals>out what i'm doing wrong. I tried to find `system` reference in docs, in guix sources - no luck. Could anyone help me please write my dumb package?
<PotentialUser-39>raghavgururajan: yes, i used the bootloader and everithing went smoothly, it asked me to reboot, i did so and i was welcomed by a blackscreen
<PotentialUser-39>It happened in artix too but at least they had a "detect uefi and boot" option in the usb, so im wondering if i can do something similar in here