IRC channel logs


back to list of logs

<nckx> /etc/letsencrypt/renewal/*.conf files all have timestamps 10 minutes old. There's no reason for them to have been modified at all, and they probably weren't, but it leaves a bad taste.
<civodul>katco: good that it worked :-)
<civodul>nckx: i think it's webroot, no?
<civodul>note that there's an mcron job that runs "certbot renew"
<katco>civodul: tyvm!
<civodul>so the only thing that might need to be done manually is "herd restart nginx"
<nckx>civodul: Right, it's not very stateful.
<civodul>normally there's a deploy hook that allows certbot to restart nginx
<civodul>but it is/used to be broken
<civodul>so the certs had expired?
<nckx>civodul: That would be a separate (and less severe) bug, though.
<nckx>civodul: Yep, but andi- caught it and I caught andi-'s message and all was well.
<civodul>so yeah, it must be that deploy hook
<civodul>i thought we had fixed it
<nckx>civodul: Deploy hooks run after renewal to clean up and reload services, they can't shut down servers beforehand (and that would be a hack anyway).
<nckx>They're not even invoked until everything's over.
<nckx>There used to be other hooks but I know for a fact all that changed after I last used certbot myself.
<civodul>wait, we're not using the certbot service on berlin, i think it didn't exist back then
<civodul>but bayfront.scm is using it
<civodul>so we should just do the same
<civodul>well, for later
*civodul -> zZz
<nckx>civodul: O… K. Good idea, but… how did this work before? 😛
<civodul>maybe it didn't work before on berlin
<civodul>i must have been confused
<nckx>So I was just the (hopefully) last in a long row of human cron jobs. OK.
*nckx quietly closes the door to the hot-dog factory & goes to bed as well.
<nojr>hello, does someone know why qgis launches with lots of python related errors?
<NieDzejkob>nojr: Do you happen to know whether this was always the case?
<nojr>NieDzejkob: yes, everytime I launch the app
<nojr>I'll launch right now and write down all the errors, it's like three
<nojr>Update of view in private qgis.db failed.
<nojr>table vw_srs already exists
<nojr>Private qgis.db
<nojr>first error
<nojr>couldn't load 'processing' plugin, second error
<nojr>Couldn't load plugin 'MetaSearch' due to an error when calling its classFactory() method
<nojr>ModuleNotFoundError: No module named '_struct'
<nojr>this is the PythonError
<nojr>An error occurred during execution of following code:
<nojr>import pyplugin_installer
<nojr>An error occurred during execution of following code:pyplugin_installer.initPluginInstaller()
<nojr>An error occurred during execution of following code:
<nojr>those are all the errors
***jonsger1 is now known as jonsger
*NieDzejkob is waiting for genv --ad-hoc qgis to do its thing
<ryanprior>Is there a more straightforward way to set an environment variable for a Guix package build than to add a phase that sets it?
<ryanprior>Grep suggests #:environment-variables
<ryanprior>But I don't see that in any package definitions, only in services.
<ryanprior>I see a lot of `(add-before 'build 'set-environment-variables ...)` which is too bad
<NieDzejkob>Where should a data structure library go? Someone suggested to move python-intervaltree out of bioinformatics.scm since I am going to use it from cybersecurity.scm. python-xyz.scm sounded like a good choice until I realized that it depends on intervaltree (a C library), and leaving that one in bioinformatics.scm creates a cycle
<NieDzejkob>ryanprior: in some cases, like CC=gcc for example, #:make-flags is enough
<ryanprior>In this case the build script is specifically checking an env variable.
<NieDzejkob>In that case, I don't see any alternative to a setenv phase
<jonsger>nojr: I have this private db "issue" as well with qgis. Python errors only occur when installing plugins like "Geocoding"
<nojr>jonsger: the thing is that this is a fresh qgis install, BTW i forgot to mention that I run guix on a foreign distro
<nojr>ubuntu 20, to be exact, the thing is that I realized GIMP had some issues with Python as well, since it's using python2, but I was able to fix by
<nojr>specifying it's path on GIMP's settings
<nojr>but, as far as I know, qgis uses python3? not sure why it's not finding all the modules it needs
<butterypancake>how come my install includes things like gnome-shell and gnome-settings-daemon? Where are these things coming from? My services are set to %desktop-services but I don't see anything gnome related under there
<bdju>butterypancake: %desktop-services comes with GDM, which is a Gnome thing.
<bdju>(or at least it used to, I'm not currently using %desktop-services to verify)
<bdju>emacs-general is still failing to build :(
<butterypancake>bdju: oh yes, I'm currently using GDM. It seems a little heavy though. Any suggestions for alternatives? I'd use startx but apparently that's non-trivial to get working
<bdju>I think you need elogind to get startx working
<bdju>SDDM is okay, not great at all, though, honestly.
<bdju>I think someone was packaging "ly", not sure if that's ready yet, but it looks pretty lightweight and nice.
<bdju>I'm using wayland, so I do "exec sway" instead of "startx", but it's fairly similar probably
<marusich>bdju, are you using Guix System? How do you tell it to run Wayland? (And sway?)
<bdju>marusich: Yes, I am on Guix System. I have "sway" in my system profile and I boot to a TTY (just running base-packages rather than %desktop-services), then from the TTY I login and then run "exec sway".
<marusich>I see. Running "exec sway" launches Wayland, too?
<bdju>You can just run "sway" also, but I heard it was better to use exec. The difference seems to be that it then logs me out if Sway crashes/exits, so then there's not a shell logged in as me.
<marusich>I don't know much about how Wayland (or X, for that matter) do their start-up, so I'm a little curious
<bdju>Uh, yeah, it all launches together. I'm guessing if you were to run some other Wayland compositor like wayfire you would run its name instead.
<bdju>Similarly I don't put "wayland" in any package lists, just "sway".
<lle-bout>marusich, hey, I'm trying to install docker with GNU Guix so I can then use a GNU Guix service to run Cuirass, do you think I could still use a service without a GNU Guix system?
<marusich>Huh, I see. I guess I expected to invoke something like "start-wayland" since there was a "startx" I thought you had to invoke from the CLI to get X working, but I suppose it's to be expected since they're prertty different.
<marusich>Is Docker a dependency of Cuirass?
<lle-bout>marusich, you just run `sway` for example, since it has a compositor built in
<lle-bout>marusich, no but I wanted to run Cuirass cleany as a service
<marusich>Oh you want to run Guix System it inside a Docker container?
<marusich>I suggest not using Docker if we can avoid it.
<marusich>Can we use a VM instead?
<lle-bout>Probably yes, it's just heavier and resource management is more complicated
<lle-bout>It's probably simpler to use Docker or any other container thing to run a GNU Guix System and then run the service as usual
<lle-bout>Bonus is that Docker doesnt require a kernel, so we don't have to sort the kernel for powerpc64-linux for it
<marusich>Maybe it's worth a try, then
<lle-bout>There's podman but last time I checked it wasnt in GNU Guix
<lle-bout>Docker is building so I'm waiting
<lle-bout>Gentoo for ppc64 doesnt have working docker
<lle-bout>runc fails to build because of some build option unsupported on ppc64
<lle-bout>I don't know Gentoo enough to fiddle with that
<marusich>lle-bout, does "guix system docker-image" work?
<usney>If my internet connection is not secure, will installing guix from script can be exploited?
<lle-bout>marusich, I'm building Docker for now, will try that soon
<usney>My internet has been acting very strangely for the past week and a half
<lle-bout>usney, TLS should keep you relatively safe even if it's not perfect
<usney>yes I know but...
<usney>that's the problem
<lle-bout>So if you download GNU Guix's installer script through HTTPS (HTTP over TLS) then it should be OK
<marusich>OK. Assuming you can use Guix to build a Docker image, you should be able to get away with booting the docker image via "docker run --privileged $theimage". It'll run the same start-up scripts that would normally run, so the Shepherd will launch all the processes, including guix-daemon.
<usney>it is in browser I am having issues with https
<lle-bout>marusich, that would be awesome!
<marusich>The reason I'm hesitant about running Guix System in a Docker container is not because it doesn't work, but because if there is a problem building something, it could perhaps be difficult to differentiate between "a problem with running Guix System in Docker" vs. "a genuine problem with the package that failed to build." I know of no blockers other than that.
<lle-bout>usney, are you in a country with severe internet censorship?
<usney>is this the right link?
<usney>no I am not but it could be now with the crazy president running now
<lle-bout>usney, yes that's the link!
<usney>I keep getting websites I use often to switch from https to http randomly.
<usney>like gmail, facebook and ebay
<lle-bout>usney, what system are you on right now?
<usney>I use https everywhere
<marusich>Running Guix System in a Docker container works, but it can be finnicky. That's why the --privileged option is necessary. You can't do all the things you might expect to be able to do (e.g., set up the builder sandbox) in a Docker container, but you can do them in a VM.
<usney>I am using a debian based system
<lle-bout>You can turn on HTTPS Everywhere with enforced HTTPS to block all HTTP, that's what I use
<lle-bout>There's an option
<usney>yes I did that already
<usney>still doesn't work
<marusich>Also the thing I was rabbit-holing on earlier was actually a problem with Guix in Docker. See:
<lle-bout>"Encrypt All Sites Eligible is ON"
<lle-bout>usney, if it doesnt work, then are you sure you don't have malware?
<marusich>That particular problem won't be a blocker for us though.
<usney>yes that option is enabled lle-bout
<usney>I don't think I have malware on linux.
<lle-bout>usney, and so when it goes to HTTPS you get a warning?
<lle-bout>to HTTP *
<usney>no not all the time
<marusich>In short, if you try to "guix gc" dead directories in the store that came from a prior layer, Guix will fail to GC them even though it reports no error.
<usney>only one time on ebay I got a warning from https everywhere
<lle-bout>usney, what kind of Internet access do you have? Is it a home ISP you pay for?
<lle-bout>usney, Ebay may not have good practices everywhere on the site when it comes to HTTPS
<usney>facebook and gmail says it is unencrypted because it was exempt
<lle-bout>That's unusual
<usney>I reloaded the page that https everywhere blocked and it loaded as https
<lle-bout>marusich, hmm I see
<marusich>The implication would be that your image might grow larger and larger without bound as you build more things - unless you do something about it, like mount the store on a separate volume outside the container.
<lle-bout>marusich, we could do a regular chroot :P
<lle-bout>but I'm not sure GNU Guix manages that smoothly
<lle-bout>Especially when it comes to services
<marusich>I think, we can just try using Docker if it's easy. Once we can build a VM image using guix, use a VM.
<marusich>Although at that point, we could probably boot bare metal?
<lle-bout>usney, it's kind of worrying, can you describe more about your network access conditions and what your machine actually is?
<marusich>I guess we don't need grub working to boot a VM, so a VM may be a fair intermediate target.
<usney>yes I know it is very worrying
<lle-bout>usney, do you have any reason to think you are being targeted?
<lle-bout>OK, I see.
<usney>I am doing legal stuff
<usney>with a lawyer
<usney>so the connection going in and out has nothing to do with that?
<lle-bout>usney, what do you mean, nothing to do with what?
<usney>ever since the connection has not been stable it started doing that
<lle-bout>Before running into worrying conclusions; it's possible that you've got a crappy router
<usney>I am not sure
<usney>can I pm you?
<lle-bout>Yes, go ahead
<apteryx>eh, finding the 'ag' package was near impossible. I had to do: guix search ag | recsel -e 'description ~ "grep" && description ~ "search" && description ~ "fast"' to find it
<lle-bout>marusich, I have a working config, it should work out of the box with Linux-libre so
<lle-bout>I'll try that next
<lle-bout>GRUB works and is ported to ppc64
<marusich>apteryx, you can use regexes in package search, so in the case of a small word like "ag", it helps to add word boundaries: guix package --search='\<ag\>'
<apteryx>marusich: good tip! it works great. thanks :-)
<apteryx>there isn't any bash completion for shepherd, is there?
<marusich>lle-bout, the bootstrap binaries built successfully. I'll give you their info in a sec.
<lle-bout>marusich, cool, are they identical?
<lle-bout>marusich, on my end: "go-1.4-bootstrap-20171003" failed to build because of a GNU Guile code error
<lle-bout>Looking into it now
<marusich>To the ones I built before? They are identical to a set I built from a differnt Guix commit (before we merged it into master), yes.
<lle-bout>marusich, you said you'd run 2 rounds
<lle-bout>If they're identical then great!
<marusich>Ah, I was hoping someone else would run it once
<marusich>Knowing it builds identically on my machine is great, but knowing that it builds identically when another person does it would be super nice.
<lle-bout>Oh well, I could. Some time. But if you ask explicitly some people might
<lle-bout>efraim, or mbakke
<marusich>Yeah, makes sense. I'll send an email out with some details.
<lle-bout>marusich, - go failing
<lle-bout>The GNU Guile trace isnt very helpful
<lle-bout>I think it fails on: "(loader (car (find-files ld "^ld-linux.+")))"
<lle-bout> /gnu/store/b03hmpmgfhz078y8gbv2p9w4rmzsm29a-glibc-2.31/lib - doesnt contain ld-linux so that's probably why
<lle-bout>but on x86_64 it does. Interesting.
<enderby>hi, i've been having problems with self-built ardour6 on debian with guix-installed lv2 plugins, wondering if there are plans to upgrade ardour to 6.00? (pretty please? :)
<lle-bout>well it looks like powerpc64-linux doesnt have ld-linux at all
<lle-bout>even on Fedora it doesnt
<lle-bout>(well fedora is little endian but that should be similar)
<lle-bout>enderby, hello! I'm sure it'll get updated eventually, not certain it is any hard either. Are you sure you can't write a GNU Guix package for it? If you could build it by hand, you should be able to modify the existing package at least.
<lle-bout>I think everyone who's working on GNU Guix is quite busy already, so definitely can use some help. I personnally don't have commit access so I can't push updates.
<lle-bout>enderby, you can try $ guix refresh ardour - it may work
<lle-bout>well it seems there's no updater for ardour. Well either way, try updating the version from 5.12 to 6.x and see if it works
<lle-bout>marusich, oh no.. "go tool dist: unknown architecture: ppc64" - go 1.4, first element of the bootstrap chain, doesnt know about ppc64 at all
<lle-bout>marusich, gcc 7.5 which we have at this stage contains gccgo which is Go 1.8 compatible as explained here:
<lle-bout>It should suffice, but that's not how GNU Guix does it
<bdju>anyone using waybar? I see the waybar wiki says to install otf-font-awesome. guix has font-awesome, just not sure if the otf- part is significant
<lle-bout>marusich, hmm: - maybe it's not ported at all to big endian? that would surprise me, it worked in gentoo
<lle-bout>bdju, are you having an issue? try fc-cache -rfv and then fc-list - otf should stand for OpenType?
<lle-bout>fc-list should tell you if the OpenType variant is there
<bdju>no, have not actually ran waybar yet
<bdju>bounching between tasks at the moment
<bdju>thanks for the tip, though
<marusich>lle-bout, I see. That is mysterious if it was working on Gentoo. By the way, I wonder if the little-endian bootstrap will work out better now that time has passed...might be worth exploring too, if the last time you looked into it was months ago.
<lle-bout>marusich, Gentoo uses a 1.8 bootstrap, it's maybe cross compiled
<lle-bout>marusich, little endian will not change a thing, experimental support came in 1.5
<lle-bout>marusich, - CTRL-F ppc64
<bdju>could the guix font-awesome package be out of date? I'm missing a couple icon in the default waybar config after installing it, but some appeared that were missing before. so, I suspect it's using some too new for the guix version.
***apteryx is now known as Guest90849
***apteryx_ is now known as apteryx
<mroh>oh, I just realized that emacs-debbugs has a shortcut 'C' for sending control msgs. nice!
***apteryx is now known as Guest4964
***apteryx_ is now known as apteryx
*elais sent a long message: < >
***apteryx is now known as Guest55299
***apteryx_ is now known as apteryx
***wxie1 is now known as wxie
<civodul>Hello Guix!
<mroh>hello civodul!
<reepca>guix package --search= > /dev/null
<reepca>Throw to key `parser-error' with args `(#f "Unknown command" abbr)'.
<efraim>lle-bout: I'll try building the powerpc64-linux-gnu bootstrap-binaries from aarch64-linux, hopefully it doesn't shutdown that board :)
<efraim>lle-bout: go-bootstrap "should" know about ppc64, I'm pretty sure it supported POWER5 before it upgraded the minimum to POWER7 or 8
<efraim>ly is still segfaulting on me, I'll probably post what i have to the mailing list
<efraim>i kinda hacked entrance to work, it doesn't want to read its config file but I can plop files into /usr/bin where it expects them
<reepca>aha, looks like it's a texinfo issue
<reepca>seems the source of the issue is @abbr{} in libuemf, when those are removed the error disappears
<reepca>what's the proper way to comment on a certain commit?
<reepca>can I just M-x gnus-summary-followup-with-original the email sent out by guix-commits?
<efraim>This is a follow-up to commit FOO
<efraim>oh, there's a macro? I always do it by hand
<efraim>I guess you could change it to @dfn{The Longer Text} (TLT)
<bricewge>Hello Guix!
<civodul>reepca: you can do that, but maybe set To: guix-devel instead
<civodul>in the past i mistakenly started a discussion on guix-commits, which seemed odd :-)
<reepca>guix-commits actually sends them out with Mail-Followup-To:, which causes the gnus followup to automatically replace to 'To' field with guix-devel
<civodul>ah good
<PurpleSym>Is there any way to speed up `guix environment`? I have a manifest with ~50 packages and it takes a full minute before it jumps into a shell.
<civodul>a minute, uh
<civodul>the first time or even on subsequent runs?
<PurpleSym>On subsequent runs. It does not actually build/download anything.
<civodul>"time guix environment guix -- true" runs in 1.4s here
<civodul>PurpleSym: could you open a bug with the manifest, and send the the output of "time guix environment -m thething -- true"?
<civodul>and the output of "guix describe"
<PurpleSym>Takes about 10s here.
<civodul>pretty bad
<civodul>spinning disk or SSD?
<reepca>can confirm ~10s for guix environment guix on spinning disk here
<PurpleSym>SSD, but it’s a Laptop and I’m running guix inside a virtual machine (KVM). That probably adds some overhead.
<civodul>ah yes
<civodul>reepca: that's terrible
<reepca>although I'm also building bootstrap tarballs at the same time elsewhere
<civodul>ah, that surely doesn't help
<PurpleSym>On my main PC it takes 1.7s, so it’s either the hardware or the manifest.
<civodul>or the VM i/o
<reepca>so I ran it twice, both actually from within a 'guix environment guix', so the profile already exists in both cases. First time it took 10.8s, second time it took 4.4s
<civodul>PurpleSym: if it's in a VM, we'd also need to check what kind of file system and block device you use
<PurpleSym>Ah yeah, I should add /var/guix is on an NFS share :/
<civodul>NFS is slow anyway
<PurpleSym>Do you have a good recommendation? I think sharing /var/guix is the only option in multi-machine setups, right?
<civodul>you didn't explain it's a multi-machine setup :-)
<civodul>but yeah, if it is, you need to share
<PurpleSym>Oh, I meant, a faster/better alternative for NFS?
<reepca>doesn't our use of WAL mode for the database mean that it won't work over NFS?
<reepca>(see disadvantage number 2)
<PurpleSym>With --debug=4 I see a lot of locking operations. Is that the reason it’s so slow via NFS, civodul?
<civodul>did you check the blog post above regarding multi-machine setups?
<civodul>normally, everything is delegated to guix-daemon
<civodul>so clients don't access the store and the database directly
<civodul>that way, slowness is mostly due to the client/daemon communication latency
<civodul>(aka. "it's our fault")
<PurpleSym>Yeah, I’ve used it as a starting point for my setup.
<civodul>but then, if your store is on NFS, there's slowness all around, for instance when loading all these .scm/.go files
<PurpleSym>Ah, ok, I see.
<civodul>it's ok on high-speed networks on HPC clusters IME
<civodul>but if you're doing NFS in a VM... maybe it's not so great
<civodul>those layers come at a cost :-)
<reepca>speaking of expensive locking operations, I've posted a patch series that addresses the issues I've found thus far with (guix store database), civodul
<PurpleSym>Our entire environment is virtualized, no access to physical machines. Alright, I’ll have to think about that again.
<efraim>tangentially related, is there a good way to test btrfs compression w/o daemon deduplication speeds vs ext4 w/ daemon deduplication?
<efraim>If I want something to start when my window manager starts I should put a .desktop file in ~/.config/autostart/ or would a shell script be enough?
<civodul>reepca: awesome, thanks reepca, i'll take a look
<civodul>efraim: i use good ol' ~/.xsession
<efraim>I guess that works
<efraim>[[ $(herd status) ]] || shepherd
<efraim>I tested it with true and false
<efraim>should it be executable?
<lprndn>Hello guix!
<civodul>well, it's mostly static binaries, so it's "easy"
<civodul>but still
<civodul>hey lprndn!
<roelj>Is there an example of an ‘nginx-configuration’ that serves two domain names with and without SSL?
<jonsger>roelj: you just need to overwrite the (listen '()) field as it defaults to 80+443 ports
<roelj>jonsger: Okay, so I have two ‘nginx-server-configuration’ blocks, one for port 80 and one for port 443. Now I want to put multiple domain names under both nginx-server-configuration. How do I do that?
<jonsger>don't know.
<jonsger>I'm not even sure what you want to achieve with this configuration
<roelj>well thanks anyway :)
<lcatcat>How can I customize the kernel in Guix System so that it guarantees that a certain feature can be disabled?
<lcatcat>Guix's default kernel seems to enable some features that are very bad for the server(on kvm).
<lcatcat>I tried to use other kernel configuration before, but it will fail in the initrd when booting. I have no way to debug why the initrd failed.
<mroh>lcatcat: the cookbook ( has several examples
<davidl>hi, I tried to package a pip package fastscript and Im receiving this error in the build log: "AssertionError: missing expected setting: version" - any ideas how to fix that?
<davidl>I can see in the file that it checks for some things like that but still I dont know how I can fix it.
<lcatcat>mroh, I read this, but it only supports enabling features, not disabling features.
<guix-vits>davidl: an wild guess: maybe it want you to specify if this a DEBUG or RELEASE build?
<davidl>guix-vits: I actually figured it out just now - for some reason the settings.ini file is not in the tarball in the pip package repo.
<davidl>( parses settings.ini and verifies the parsing which fails)
<mroh>lcatcat: you can provide a kernel .config file. there you can disable/enable whatever you want. I build my kernels that way. you can also disable features the same way, as you would enable them, i think.
<guix-vits>davidl: while i'm stuck thinking "what DEBUG and RELEASE have in common with PYTHON?"...
<davidl>guix-vits: ^^
<tinga>Hi. How do I retrieve the signing key for the guix distribution (key 3CE464558A84FDC69DB40CFB090B11993D9AEBB5)?
<tinga>gpg --recv-key 3CE464558A84FDC69DB40CFB090B11993D9AEBB5 => key 090B11993D9AEBB5: new key but contains no user ID - skipped
<tinga>Ah, this worked (my * are people right saying gpg is too complicated): gpg --keyserver hkps:// --recv-key 3CE464558A84FDC69DB40CFB090B11993D9AEBB5
<janneke>hmm, how is gexp->file different from program-file?
<janneke>i need the output produced by program-file, but my code won't run
<janneke>it "runs" with gexp->file, but the result is...well, not a program-file
*janneke needs to create a patch with mail
<leoprikler>janneke: program-file is the declarative variant of gexp->script
<apteryx>I'd like to understand something about Shepherd services; when a required service is started, it's just forked (depending of the procedure in the start slot) and then the service requiring it is started, right? What canbe done in a situation where B requires A but A starts too slowly for B to notice that A is started?
<janneke>leoprikler: thanks...that helps!
<janneke>so, i'm in the wrong domain and where gexp->file works, there gexp->script will also work and produce a "program-file"
<leoprikler>Probably, though I'm not quite sure whether "program-file" is an unambiguous term here.
<leoprikler>apteryx: I don't think B is started before (start A) returns, so you can try delaying that until e.g. a PID file is available or something
<civodul>tinga: try with: --keyserver
<civodul>it should be there
<civodul>janneke: gexp->script is a monadic procedure, whereas program-file is a regular procedure that returns a "lowerable" object
<civodul>"file-like" object as the manual says
<civodul>when possible, you'll prolly wanna use program-file
<apteryx>leoprikler: good idea, thanks
<janneke>civodul: thanks, yes -- i've created some (prolly warped) code that needs both, but lower-object on program-file seems to do the trick
<boeg>is it just me or is there no gui for handbrake in the official channels? The handbrake package only seems to have the default output and its the cli version?
<pinoaffe>boeg: it seems like the gui is available under the command `ghb`
<pinoaffe>not to be confused with the party drug of the same name
<boeg>pinoaffe: interesting - alright, thanks
<boeg>bit weird instead of something like handbrake-gui
<pinoaffe>yeah, but it seems to be a decision of the handbrake devs, as I cannot seem to find anything in the guix package definition that would change the output path
*janneke hates bash
*janneke is fighting exec foo "$@" magic
<leoprikler>what would be your preferred way of writing exec foo "$@"?
<janneke>well, i want to iterate $@, but still be able to do exec foo "$@"
<janneke>so, args="$@" ... [shift] ... exec foo "$args" would be nice
<apteryx>leoprikler: the "upstream" systemd service definition for containerd uses the 'notify' systemd mechanism:
<janneke>but i cannot find a way for that to work
<reepca>(apply execl "foo" (command-line))
***sputny1 is now known as sputny
<janneke>i can probably do an element-wise copy while shifting, but eh...
<janneke>and also $@[$i] does not seem to work, avoiding shift and doing exec foo "$@" later
*janneke goes to search the interwebs
<leoprikler>apteryx: can we patch containerd, so that it instead writes to a PID file?
<leoprikler>It would appear, that there is a call to sd_notify in this case, so either we implement that, or we patch that away
<apteryx>leoprikler: perhaps! it'd be nice to have a way in Shepherd to allow reusing that notify thing
<leoprikler>Hmm, it appears, that notify is written to a normal Unix socket (?), so perhaps we can just listen on that
<apteryx>leoprikler: I'm not sure, but I thought the application to use systemd-notify needed to be linked with the systemd-notify library. Perhaps that's not true.
<janneke>yay, /me found a solution; for i in "$@"; do ..; no shift, keep $@ intact
***sneek_ is now known as sneek
<apteryx>leoprikler: to use sd_notify requires #include <systemd/sd-daemon.h> in C/C++ languages.
<apteryx>I'm not sure about Go (which is the language containerd is written with)
<leoprikler>on the sender side, nothing stops you from writing your own receiver
<apteryx>I was just wondering because we don't have any systemd package, so even on the sender side it could be a problem
<leoprikler>do we have containerd?
<leoprikler>or anything else that uses sd_notify?
<apteryx>we have containerd
<apteryx>I guess the sd_notify support in many applications is disabled since it's currently useless to us.
<civodul>leoprikler: no sd_notify here
<leoprikler>I see, so IOW we can't rely on applications to call sd_notify.
<civodul>that'll be in Shepherd 1.0
<civodul>(joke or not? you decide!)
<apteryx>that gives us nearly a year to implement such support (deadline April 1st 2021).
<civodul>seriously, we could implement it
<civodul>but so far, it didn't seem to be a showstopper
<leoprikler>IIUC, we already have the shepherd socket to communicate with, right?
<tinga>civodul, nope, that appears to be the default, so was the *reason* for the problem: "You are probably using the keyserver, which has an owner approval system – it will strip all user IDs unless the owner of the corresponding email address has allowed them to be published." (
<tinga>civodul, note that yes the key is there, but, is ignored by gpg
<tinga>I first thought it was a problem with the key itself ("why would the guix key not have a user id?") that's why I came here. Till I found that pasted statement.
<civodul>ah, ok
<civodul>but that still shouldn't prevent gpg from verifying signatures
<civodul>weird that it's ignoring it
<tinga>It's not importing it.
<tinga>The linked page also says "Future GnuPG versions will accept keys without an UID, although it won't be terribly useful except for direct fingerprint-based comparison.)"
<civodul>what version do you have?
<civodul>i think i imported stuff from with 2.2.20 or so
*janneke cleans-up the Hurd startup even more, cutting hurd-pkg->rc-script dependency
<tinga>civodul, gpg (GnuPG) 2.2.20
<civodul>hmm, ok
<terpri_>nice work janneke
<tinga>civodul, I haven't had this before so I guess it's a recent change. The linked page " Asked 8 months ago "
<terpri_>does hurd support amd64 yet?
<civodul>anyway, you can also get the key from
<civodul>tinga: ↑
<civodul>terpri_: not quite
<civodul>janneke: yay
<terpri_>is it being worked on?
<civodul>more or less
<civodul>there's been progress over the last few years
<civodul>but the Hurd has very few volunteers
<civodul>so everyone can make a change
<tinga>civodul, libgcrypt 1.8.5 fwiw
<andi->I've once more tried to send an email to and my mailserver logs successful delivery. Those mails just never appear on the thread: Is there anything I should be doing instead?
<tinga>civodul, note that now that my gpg has the key, "gpg --recv-key 3CE464558A84FDC69DB40CFB090B11993D9AEBB5" doesn't complain anymore (Total number processed: 1 unchanged: 1). Seems gpg is simply refusing to add a key to its keyring without a uid.
<tinga>civodul, so if you have run --recv-key recently just to update signatures or so, and already had a uid for that key on the keyring, then that's why you didn't see the issue.
<tinga>(I guess, given the information I have.)
<terpri_>andi-, i see some extra messages on compared to
<terpri_>so just out-of-sync websites, apparently
<andi->I wasn't even aware there was another interface to this
<andi-> doesn't have any reference to besides in the mail address :/
<apteryx>lle-bout: I've built the ppc64 bootstrap binaries and had a different checksum only for gcc
<civodul>andi-: probably has a slight delay
<civodul>but your message should show up after a while
<civodul>like an hour at most
<andi->civodul: first mail is from ~16h ago or so
<apteryx>andi-: is it your first mail to the list?
<apteryx>I think it goes into a moderation queue the first time
<andi->but it is displayed here:
<apteryx>ah right
<andi->I was also able to instruct control@ to unarchive the issue
<tinga>guix-1.1.0 build from source docs say it requires "GNU Guile 2.2.x", but then checks guile 3 as well; which version is preferred for Guix?
<bhartrihari>Hi, has any mailing list software been packaged in guix?
<civodul>tinga: 3.0 is preferred
<civodul>hi bhartrihari
<civodul>bhartrihari: not sure, but you can check at or using "guix search"
<tinga>(civodul, might be worth changing the README to reflect that.)
<civodul>will do
<bhartrihari>I'm a bit amazed that GNU mailman hasn't been packaged yet (or so `guix search Mailman` tells me).
<apteryx>am I the only one to not have icons in Evince?
<apteryx>I don't use GNOME
<leoprikler_>I think there are some icon packages propagated by the gnome meta package, that need manual installation in that case
<leoprikler_>particularly adwaita and hicolor
<apteryx>I have hicolor-icon-theme, gnome-icon-theme and arc-icon-theme installed.
<apteryx>I'm surprised it can't gracefully fall back to hicolor (that's the point of hicolor)
<leoprikler_>do the icons you're trying to load exist in hicolor?
<apteryx>strangely I have Adwaita in my user profile, but it's not listed as installed. Probably propagated by one of the icon packages I have installed.
<apteryx>leoprikler_: seems I'd have to look at the source to know
<NieDzejkob>how can I make `guix size $pkgname' include its propagated inputs?
<reepca>I think there's a race condition in (guix nar), in with-temporary-store-file. If after adding the temp-root the file doesn't exist, it's clear that it was gc'ed, but if it does exist, it's possible that it was gc'ed and then created by some other process's with-temporary-store-file. In that case the current process would clobber that other process's temporary file.
<NieDzejkob>reepca: But the content would be the same anyway, no?
<reepca>I don't see why it would be, the only criterion used for the temp file name is "doesn't exist yet".
<NieDzejkob>ah, I misremembered what that does. Hmm.
<reepca>actually, for that matter, couldn't another process's with-temporary-store-file clobber the current process's temp file after it deletes it prior to evaluating the body?
<civodul>NieDzejkob: re "guix size", you can't
<ryanprior>A common thing I do while packaging is to clone a git repo, check out a certain tag, and then `guix hash -rx`. Is there a tool for that? Something like `guix download` but for git basically?
<NieDzejkob>ryanprior: Not yet ;)
<guix-vits>NieDzejkob: citing one fellow, "yes, but not yet".
<Hugal31>Hi, I am trying to play with a Qemu image of Guix System, but I don't find the system configuration file at /run/current-system/configuration.scm
<sneek>Hugal31, you have 1 message!
<sneek>Hugal31, nckx says: I'm not aware of a way to set the substitute cache TTL or disable it altogether, but it would probably be accepted as a patch.
<Hugal31>Oh, it may be because I am running a read-only fs
<Hugal31>No, it wasn't that
<ryanprior>If I rename a profile directory, will Guix still know about it? How does it know where profiles are stored?
<apteryx>ryanprior: I think it registers them under /var/guix/gcroots/auto/
<apteryx>by their hash, so I guess moving the profile links that you have somewhere wouldn't matter
<apteryx>Note that I assumed the question was about user managed profiles, such as can be generated with guix package -m manifest.scm -p your-profile
<marusich>hey there
<User28>hello I am using linux mint 19.3 I installed guix package manager a while ago from the shell installer script ( gotten from
<User28>I installed and set the locales according to 2.6.1 and set substitue server ac
<User28>cording to
<User28>I have two problems-
<User28>whenever I install/remove/update anything the following error pops up repeatedly
<User28>substitute: /gnu/store/q19l04vd2za80mk1845pz7r8cz29qk43-bash-minimal-4.4.23/bin/bash: warning: setlocale: LC_ALL: cannot change locale (en_US.utf8)
<User28>the version of mpv listed at is 0.32.0
<User28> the latest which I want however guix pull && guix package -u does not update my
<User28> mpv.
<marusich>For mpv, what version is shown by "guix package --show=mpv"?
<User28>also the one which I currently have
<alextee[m]>is there a way to rm -rf a directory?
<alextee[m]>(rmdir) doesn't seem to work on non-empty dirs
<efraim>User28: what does 'which guix' show?
<alextee[m]>efraim: thanks!
<User28>efraim /home/useradmin/.guix-profile/bin/guix
<efraim>hmm, so guix pull && guix package -u should update everything
<mbakke>User28: you need to make sure that ~/.config/guix/current/bin comes first in PATH
<mbakke>and you can probably uninstall guix from your profile
<efraim>ah, yeah, I saw the /home/useradmin and glossed over the rest :/
<User28>ok I'll add and try and update
<dissoc3>does anyone else have constant problems with the stumpwm package? every time I update it something breaks
<User28>sorry I only now updated path and did guix pull && guix package -u it showed me this warning- guile: warning: failed to install locale
<User28>hint: Consider installing the `glibc-utf8-locales' or `glibc-locales' package and
<User28>defining `GUIX_LOCPATH', along these lines:
<User28> guix package -i glibc-utf8-locales
<User28> export GUIX_LOCPATH="$HOME/.guix-profile/lib/locale"
<User28>See the "Application Setup" section in the manual, for more info.
<User28>but I have installed glibc-locales and set guix_locpath
<User28>it is continuing after the warning though
<User28>it is building from a different commit than before so I guess thats good but the bash locale warnings persist(substitute: /gnu/store/q19l04vd2za80mk1845pz7r8cz29qk43-bash-minimal-4.4.23/bin/bash: warning: setlocale: LC_ALL: cannot change locale (en_US.utf8)
<efraim>4.4.23 is fairly old, current master has 5.0.16
<efraim>it could now be from a glibc version mismatch
<efraim>I'd ignore the warning until you update the daemon and your packages
<rndd>hi everyone! suppose i'm writing my package definition, what i have to place in (source (origin (uri (commit... ?
<nckx>Mornin' Guix.
<nckx>rndd: ‘It depends’. I asume you mean (uri (git-reference (commit …)))? Ideally, the project uses version-derived git tags and you can use something like (commit version) or (commit (string-append "v" version)).
<nckx>‘git describe’-style refs also work.
<nckx>rndd: If the project doesn't use tags, let-bind a COMMIT and REVISION ("0"). Use (commit commit) and (version (git-version "X.Y.Z" revision commit)), where X.Y.Z is the last real release or 0.0.0 if there never was one.
<User28>efraim: ok thanks how should I uninstall the guix in my profile? I don't really understand why there are two versions(?) one in my guix profile and one in .config guix?
<efraim>'guix package -r guix'
<rndd>nckx: what if i want to package master branch?
<nckx>rndd: See libtsm in terminals.scm for a totally arbitrary example.
<nckx>rndd: You can't package master branch since it will move and your hash will change. You'd use the commit that's currently the tip of master.
<nckx>In general, you can't safely use git branches in Guix packages.
<User28>efraim: so the guix in ~/.guix-profile/bin/guix will be removed by guix package -r guix and ~/.config/guix/current/bin/guix will stay and work?
<efraim>User28: exactly
<rndd>nckx: as i understand i cannot put commit (something like "d6f9deb66358a3563d950c6d2f168d9e85a0e564") in (commit ...)?
<nckx>rndd: Yes, you can. But you want to re-use it in your version field too.
<nckx>Hence the (let …)
<nckx>Did you look at libtsm yet? Is something unclear?
<User28>efraim: how were two installed like this though. I installed through the shell script. Will reading through the installation instructions help me understand this stuff?
<rndd>nckx: oh, i see. i will try something similar
<efraim>probably. the install shell script helps a lot with common mistakes when getting Guix set up. Reading the manual about the installation process will certainly help with understanding what the script does :)
<nckx>rndd: What I meant is that you can't write (commit "master"). It will ‘work’, but the next time someone pushes to master it will break. Worse, it might keep working for you (because Guix re-uses the old checkout since it trusts the hash) but you'll silently be stuck at the old commit. And it will completely break for others who don't *have* an old checkout. Badness for everyone.
<nckx>rndd: Great!
<rndd>nckx: how to find out sha256 for repo tag? 0_0
<nckx>cd your-git-checkout && guix hash -rx .
<nckx>or, specify a bogus hash and guix will tell you what it should have been.
<nckx>You can use 0mwn91i5h5d518i1s05y7hzv6bc13vzcvxszpfh77473iwg4wpr1
<nckx>I made that up based on libtsm so it's guaranteed to be valid but not match anything existing.
<rndd>nckx: that's strange
<User28>efraim: Ok I guess I will do that. is mixing up .config/guix/.. and .guix-profile/... guix's a common mistake?
<efraim>people installing guix into their profile is a common mistake
<User28>does the install script just install into .config/?
<efraim>it looks like it installs guix for root as though root ran 'guix pull', starts the daemon and makes 'guix' available from /usr/local/bin/guix
<efraim>I don't think it specifically suggests running 'guix pull' as a user or installing anything
*efraim heads off to bed
<rndd>nckx: what is the meaning of (package (source (origin (file-name ... ?
<nckx>rndd: It specifies the name of the source checkout in the store, /gnu/store/hash-<file-name>-checkout.
<nckx>Otherwise it's just /gnu/store/hash-checkout, with no obvious way to know where it came from.
<nckx>Guix doesn't care; it's for humans.
<nckx>‘guix lint’ will warn about it.
<rndd>hmmmmm, what's difference between "`" and "'" in scheme ?
<rndd>sorry, it's offtopic
<marusich>rndd, ` is quasiquote, and ' is quote.
<rndd>marusich: ohh, thank you =)
<marusich>does implement any sort of bandwidth throttling?
<nckx>marusich: …not on purpose…
<nckx>rndd: Scheme questions in the context of writing (better) Guix packages are never off-topic!
<nckx>marusich: I think the conclusion was that berlin's NIC is effed, and until then so will be the bandwidth.
<marusich>I guess it's AWS; I'm running Guix in an EC2 instance. I made it an m4.xlarge for now, 4 cores, and supposedly "high" bandwidth, but I'm getting 15 KB/s downloading from even though I get much faster via my local wifi
<nckx>*until it's replaced.
<nckx>Hm, yeah, that's something else.
<nckx>Although latency can trigger it.
<marusich>meh. i'll upgrade the instance type
<nckx>There was someone else here with that exact number last week.
<marusich>it's also possible I'm running with a noisy neighbor who's eating up all my bandwidth
<bricewge>marusich: Probably a peering issue
<bricewge>Try “guix environment --ad-hoc mtr -- sudo mtr”
<marusich>could be. i guess the next netowrk tier up is c4.8xlarge which costs more than i want to pay right now, so it's probably something like that. peering, or my guess is is it more likely just poor placement in the EC2 datacenters
<marusich>restarting the instance may help
<marusich>(on new hardware)
<rndd>nckx: i have some errors, can i show source?
<nckx>Ooh, I'm getting unprecedented averages of 7 MiB/s. Maybe things have been fixed.
<nckx>rndd: Sure. paste( please.
<nckx>Include the errors as well.
***familia__ is now known as familia
<marusich>bah, i'm still getting a whopping 26 KB/sec.
<rndd>nckx: debian paste return 503, what else i'm allowed use here?
<nckx>marusich: I second bricewge's curiosity for an mtr (or similar).
<marusich>if i can build it in time i'll do it :)
<nckx>rndd: Oh, weird. It doesn't ‘really’ matter, we just don't like pastebin dot… net? com? Whatever the ‘well known’ one is, it blocks Tor and is generally hostile. Try
<bricewge>marusich: Maybe mtr is available at ?
<nckx>rndd: The indentation is correct. You're missing two closing brackets after (sha256 …).
<nckx>So everything below it (build-system, description, …) is being passed as a field of your SOURCE record.
<nckx>Don't forget to remove ‘))’ after license to balance the universe.
<marusich>there you go:
<marusich>not sure what to make of it, since lots of networks like to deprioritize ICMP packets, so I wouldn't be surprised if the responses are reflecting that sort of filtering instead of a performance issue
<marusich>but presumably the latency around telia is odd, since if it were icmp filtering/deprioritization on the reply packets, the other hops after would also be high latency responses
<nckx>marusich: mtr has a --tcp option which I've never used or investigated.
<marusich>I used that
<marusich>i guess in this case the responses wouldn't be icmp packets, they would be tcp packets
<marusich>but i dunno, normally routers don't receive a lot of requests to themselves, so just becuase it took a long time for telia's routers to process the request doesn't necessarily mean they're behaving badly, right?
<marusich>I'm also not sure if mtr is using TTL and relying on the icmp response when ttl hits zero, or if it's trying to send a syn directly to the routers. who knows. honestly by the time i find out, my downloads will be done.
<nckx>Oh, oh, oh…
<rndd>nckx: ok, this is new -_-
<nckx>marusich: That ‘godaddy’ tail puzzled me, because berlin is on DFN (German Research Network). And then I noticed the TLD 🙂
<nckx>Are you really fetching from .info?
<nckx>rndd: You're referring to the variable ‘commit’ on line 8. That variable doesn't exist. Why not use ‘commit’ instead of ‘commit_var’? You're free to call it what you want of course, but there's no ambiguity in Scheme. ‘(commit commit)’ won't confuse it, and if this package makes it upstream I'd ask you to change it.
<nckx>Underscores, brrr.
<rndd>nckx: i'm trying to break as much as possible
<rndd>it helps me to learn
<nckx>(Don't forget to re-indent your expression, it looks wrong to humans now.)
<marusich>oh, maybe i messed up the domain name lol
<marusich>i'm surprised that domain name even exists
<nckx>Wait, it still is wrong. …vxhz8l1k5fyhiyz98m4dax4"))
<nckx>marusich: We still own ‘just in case’, but ‘’ was never us.
<marusich>here is the right one:
<rndd>nckx: i didn't get is
<rndd>what's with sha256?
<nckx>…vxhz8l1k5fyhiyz98m4dax4")) → …vxhz8l1k5fyhiyz98m4dax4"))))
<nckx>(license license:gpl3)))))) → (license license:gpl3))))
<rndd>oh, i just didn't changed definition in scratch for paste
<nckx>marusich: sea→nyk does look… off.
<marusich>yeah it's odd
*nckx not a network person, can't help further. 300ms is about double what I'd expect but not horrible. And no packaget loss. I dunno.
<marusich>nyk is ?
<nckx>marusich: My guess is just home-rolled code for New York.
<marusich>oh, more likely.
<marusich>well hopefully somewhere some telia engineer is hard at work
<rndd>nckx: ok, it pulls, but there is another one
<nckx>rndd: I don't mind helping you but think along. The error message mentions screenfetch. That's weird, your package isn't called screenfetch. Oh, look.
<nckx>Copy/pasted line 29? 🙂
<rndd>nckx: yee
<rndd>nckx: damn
<rndd>it works
<rndd>but why it builds database for manual pages if i don't provide any?
<nckx>It builds them for the entire profile (so the ‘thing’ that you're installing your package into). And if your profile doesn't contain any, it should be fast.
<rndd>well, it is
<rndd>thank you
<rndd>it works
<nckx>😃 I vaguely remember that feeling. Enjoy it. Happy to help.
<marusich>problem solved by downloading sources and copying them into the ec2 instance.
<marusich>now i wait hours to build all the things :/
<TZander>how do I install boost 1.69 ?
<marusich>TZander, did you try "guix package -i boost" or equivalent? I see there is a package named "boost", which I found by running "guix package --search=boost".
<NieDzejkob>TZander: guix time-machine --commit=b8e43e61a3aef8236985b7602bce6eb67bc7fa7c -- environment --ad-hoc boost
<NieDzejkob>(I found that commit ID by issuing 'git log' in the Guix repository and searching with / for "boost:")
<TZander>marusich: sure, I just want to try different versions of boost. Not just the latest :)
<marusich>ahh, ok
<TZander>NieDzejkob: got it!
<raghavgururajan>nckx: Are you around?
<nckx>raghavgururajan: Yes. Get my /msg?
<raghavgururajan>nckx: No, I did not. I was not able to contact you as there was wifi outage at my rental place. Things are back to mormal now.
<raghavgururajan>The /msg is not available on my bridge. Could you please resend as regular DM? Thanks!
<nckx>I don't know what a regular DM is. I'm kind of busy now, so I'll just paste it here.
<nckx> I've added you to bayfront <> You should be able to simply log in now: $ ssh If your local user name is not ‘raghavgururajan’ (but I'm guessing it is), you have to add it explicitly: $ ssh
<raghavgururajan>I meant one-to-one IRC chat. That's okay.
<raghavgururajan>nckx: Thanks so much. I appreciate it.
<nckx>Nothing confidential here or in that link 🙂
<nckx>We'll have to think about a workflow here, since • we can't set up offloading since that would let you push arbitrary binaries, I think? • you'll be working on GUI applications that you'll want to run locally. I guess ‘guix copy’ is the way to go. If anyone has any other ideas (ideally from experience), please pipe up.
*nckx rips themselves away from IRC.
<cbaines>fetching substitutes using guix publish, which is already running on bayfront would probably work
<cbaines>although it may take some fighting against the local and remote caching
<nckx>Oh, I didn't realise bayfront's toots were still publically accessible cbaines. Ignore everything I said then. That, even with some caching work-arounds, is the way to go 😉 Thanks.
<cbaines>Yeah, when it doesn't run out of disk space, bayfront can do a reasonably good job of providing substitutes for x86_64-linux
<raghavgururajan>nckx: First time doing ssh. What client is easy to use?
<cbaines>I'm hoping that the recent change to remove staging/core-updates will have a delayed effect on reducing the disk space
<nckx>raghavgururajan: An ssh client doesn't really need to do much (this isn't Windows), it just turns your terminal into a remote one. ‘guix install openssh’ (command: ssh) is the norm.
<civodul>nckx, cbaines: for cases where you don't want to allow offloading, you can use GUIX_DAEMON_SOCKET=ssh://...
<raghavgururajan>nckx: Thanks. So I can create folders under home right?
<civodul>that allows you to build a thing remotely, but without sending any build result from your local machine
<civodul>(then you have to guix copy it back)
<nckx>raghavgururajan: Your bayfront home directory is yours to do with as you please, assuming you don't fill up the drive.
<nckx>raghavgururajan: So you're able to log in without errors? That's great.
<nckx>Now if you'll forgive me, I'll stop responding to pings for a while.
<raghavgururajan>nckx: Okay. I don't want to mess anything up. Would like to double-check before I do something. I am thinking of cloning wip-desktop locally on bayfront and work patches from there? Any downsides?
<mbakke>so I considered grafting DBus 1.2.18 to get the fix for CVE-2020-12049, but 'abidiff' reports 305 removed symbols such as "_dbus_atomic_dec@@LIBDBUS_PRIVATE_1.12.16"
<mbakke>all of them are LIBDBUS_PRIVATE so I'd be surprised if any packages actually used them, but who knows..
<raghavgururajan> Permission denied (publickey).
<cbaines>raghavgururajan, I can have a look in to the authentication...
<raghavgururajan>cbaines, thanks.
<cbaines>hmm, I'm not sure I see your home directory
<raghavgururajan>FYI, before attempting to login, I started ssh-agent via shepherd.
<raghavgururajan>`env` gives HOME=/home/rg
<raghavgururajan>My public key's file name is under ~/.ssh. Whereas, it is at
<raghavgururajan>Would that matter?
<nckx>raghavgururajan: No, it shouldn't.
<cbaines>I'm guessing there should be /home/raghavgururajan on bayfront, and there isn't yet, so I'm looking in to that first
<nckx>raghavgururajan: This has happened before.
<raghavgururajan>nckx: I see.
<raghavgururajan>cbaines: Thanks!
<nckx>Something goes wrong when configuring bayfront :-/
<cbaines>I'm trying to reconfigure now... this will take a little while as the machine is under load at the moment
<mbakke>hmm, I don't see anything between dbus 1.12.16 and 1.12.18 that changes symbol visibility, I guess abidiff is just being confused by those version numbers in the symbol name ... odd that it does not display the corresponding 1.12.18 symbols though
<mbakke>I've been meaning to update libabigail for a while, let's see how the new version fares :P
<marusich>anyone here do C or C++ programming? I'm curious to know what a good setup (IDE or equivalent) looks like on Guix System. I'm guessing the answer is probably "Emacs", but you never know... :)
<jonsger>icecat, replaces "Open Source" with "Free Software", costs quite some CPU cycles aka time on my laptop
<User28>why doesn't have any search function? Especially since which is part of the same project does
<apteryx>marusich: I'm using Emacs with GNU Global and emacs-ggtags
<cbaines>User28, is deployed as a static site
<apteryx>and I've been toying with Semantic (the Emacs native parser that supposedly understands C++ at a finer level)
<User28>cbaines: oh ok. pretty user unfriendly though
<apteryx>but it's heavy on configuration, and I'm not yet happy about it.
<cbaines>User28, it's not the intention to not have a search feature, it just requires someone to sort one out, and that hasn't happened yet
<civodul>mbakke: great that you though about using abidiff
<civodul>nifty tool!
<mbakke>it is, I always use it when adding a graft -- haven't seen a diff like this before though, so I'm a little skeptical :-)
<civodul>yeah weird
<mbakke>heh, I just realized the soname is different too, so a graft would not work even if the ABI had been stable
<mbakke>sap dftxbs3e
<davidl>running guix pull gives me a segmentation fault after recent guix pull
<davidl>I have restarted my GuixSD VM, and am getting this error still after the restart
<apteryx>davidl: wow! how recent was that guix pull?
<davidl>apteryx: just now
*apteryx tries
<davidl>I had an old 38/39 days Guix, then ran a guix pull.
<davidl>first as root which worked fine, but then as a user which failed.
<cbaines>raghavgururajan, I think I've managed to set things up on bayfront correctly, can you try connecting via ssh again?
<dftxbs3e>mbakke, I am dying with the heat
<cbaines>In other news, I managed to come up with a figure of ~82% matching for Guix packages
<cbaines>I'm not sure I've managed to get a figure that high before with the Guix Data Service
<davidl>apteryx: I have a /var/guix/profiles/per-user/user1/current-guix.lock file
<davidl>..and guix-profile.lock in the same dir
<cbaines>That's comparing substitutes between bayfront, berlin and the build farm I'm trying to run
<civodul>cbaines: could you come up with a figure of ~99%?
<davidl>is it safe to replace current-guix -> current-guix-41-link with current-guix -> current-guix-40-link in the /var/guix-profiles/per-user/myuser directory?
<civodul>cbaines: seriously though, thumbs up! it means that bayfront has been busy lately
<civodul>davidl: it is, but you can do "guix pull --roll-back" for that :-)
<davidl>civodul: well I have a segmentation fault on guix pull or rather guix <anything>.
<cbaines>civodul, haha, it is possible to "calculate" a higher figure by ignoring the unknown potion... that would be about 93%
<civodul>heh, i see you've already thought about it :-)
<davidl>civodul: and not sure about the lock files inside /var/guix-profiles/per-user/myuser
<civodul>davidl: you can safely remove them
<davidl>civodul: ok thx!
<civodul>the segfault is worrying though
<cbaines>according to the data I currently have, berlin has 85% substitute availability for x86_64-linux packages for this particular revision, bayfront has 63%, and the build farm I'm running has 85%
<civodul>davidl: could you send details about the segfault to bug-guix? (last good revision, first broken revision, architecture, etc.)
<civodul>cbaines: oh i didn't know the availability page, it's really nice
<davidl>civodul: I will try and get around to it, maybe tomorrow or later this week.
<cbaines>civodul, it's a resonably recent addition
<civodul>cbaines: making it all green is a nice trick ;-)
<cbaines>I added it along with some more filtering options on the package derivation outputs page
<civodul>i'm sure our armhf users will appreciate
<cbaines>So you can now select outputs which are available or not available from different substitute servers
<civodul>i just realized that the Data Service could serve narinfos with the signature of your choice
<civodul>or several of them
<civodul>dunno, but i feel there's something to be done with that
<civodul>it already serves narinfos, right?
<cbaines>Yeah, and there's a page that provides the public part of the key (like )
<cbaines>I fixed a few bugs even with the narinfo serving, as I've been using that with the Guix Build Coordinator
<cbaines>Although one of the recent downsides I discovered about the serving of substitutes, is that the derivations related to guix pull take quite a bit of space to store.
<cbaines>As the source files for guix become part of the derivation graph, you end up with a ~30MB (I think) lzipped blob for each revision of guix
<cbaines>Anyway, I digress
<dftxbs3e>marusich, I tried building gccgo but it gets stuck at missing ustat in libgo, which was removed in glibc 2.28
<dftxbs3e>marusich, currently only gccgo-4.9 is available, and we would need gccgo-5
<civodul>cbaines: ah yes, these derivations have big "source" dependencies
<civodul>we could do a trick with fixed-output derivations
<civodul>hmm dunno
<dftxbs3e>marusich, at this point it may be easier to get a VM running, but we'll still need to do that work at some point
<dftxbs3e>marusich, in a comment it's written that bootstrap from gccgo-5 is desired anyway
<nckx>cbaines: You recently mentioned using for reproducibility reports. What would need to be done?
<cbaines>nckx, seems like I can fetch narinfos :) So I should just be able to configure the Guix Data Service to crawl it and fetch the narinfos, would that be OK?
<nckx>That's fine.
<cbaines>Great :D
<nckx>Let's see if my home router survives 😊
<cbaines>What systems/targets do you build currently?