IRC channel logs


back to list of logs

<civodul>bauermann: you could try passing MALLOC_CHECK_=2 in the build environment to get a backtrace a core dump + backtrace hopefully as early as possible
<civodul>passing it through valgrind could also hint at the problem
<bauermann>yes, I was setting things up to make texlive-build-system call luatex using valgrind, but the MALLOCK_CHECK_ option is probably easier. thanks for the suggestion
<ix>anyone awake still?
<ix>i'm wondering how 'inherit' works with (operating-system)
<vagrantc>#guix spans many timezones :)
<ix>i want to split my definition into many many modules
<ix>i'm not sure where inherit is documented
<ix>is it a deep merge?
<ix>can i do it multiple times?
<ix>what's the merge strategy
<vagrantc>in packages you can have packages that inherit from other packages which inherit from other packages ...
<ix>so it's linear inheritance
<vagrantc>not sure how it applies to operating-system, but my guess is it is just guile and should work ...
<ix>i mean i can test things out i guess
<ix>but i figured there'd be docs about
<bauermann>I don't know if it applies in this case (guix in particular and GNU projects in general have a much stronger emphasis in documentation than most), but often:
<civodul>ix: has an intro on 'inherit' in the context of packages
<civodul>but it works the same on operating systems
<ix>right, okay
<roptat>civodul, how does inherit interact with this-package?
<civodul>pretty well :-)
<civodul>this-package is a "self reference", so it refers to the package at hand
<ix>(sidenote this is so much cleaner than nix)
<roptat>so if I use this-package in a package, and inherit that package, it would refer to the new package, right?
<roptat>great! thanks :)
<civodul>that's its raison d'être, even :-)
<roptat>I see
<roptat>still rebuilding core-updates... I haven't reached my new python yet
<roptat>I hope this is my last try, then I'll need to adjust one or two dependents
<civodul>oh and i've just pushed Maxime's world-rebuild changes
<roptat>and I should be able to send 4-5 patches this week-end
<roptat>pass --optimize, remove windows binaries and test modules, prevent building the static library and move idle to a separate input
<roptat>I expect up to 2x speed up, and around 25MB reduction in size
<civodul>roptat: these are much welcome improvements
***danderson` is now known as danderson
<apteryx>uh, (deprecated-package "rust-rav1e" rav1e) seems to trigger an error about rav1e being unbound, but its module is imported
***sjb is now known as sjb-clone
<raghavgururajan>> civodul‎: i was thrilled when i watched your DebConf talk back then!
***Kimapr6 is now known as Kimapr
***sneek_ is now known as sneek
<raghavgururajan>> ‎ixmpp‎:
<raghavgururajan>Fellow XMPPian. ☺️
<clacke>i'm on here =)
<raghavgururajan>clacke: Nice!
*raghavgururajan is on
***Kimapr4 is now known as Kimapr
<clacke>I've been on XMPP before but the bridges I was relying on went down, so I was on Matrix again for a year. When the bridge didn't come up immediately in the liberapolypse, I found a new bridge, uninstalled Element and went right back to XMPP =)
<clacke>Native Matrix rooms work great on XMPP too
<clacke>the virtual MUC for any Matrix room is at on XMPP
<projectmoon>Is there a stable XMPP bridge now?
<raghavgururajan>I see.
<raghavgururajan>Hello Guix!
*raghavgururajan woke up from nap but thought it was another new day xD
<tissevert>hi guix
<jonsger>how do you folks get the `core-updates` to the subject of a patch via mail. --subject-prefix="core-updates" didn't worked for me
<jonsger>okay, editing the patch itself does work :)
<raghavgururajan>jonsger: I usually do `subject-prefix="PATCH core-updates"`.
<nckx>Wazzurp Gweks.
<sneek>Welcome back nckx, you have 1 message!
<sneek>nckx, raghavgururajan says: Finally was able to get rid of the errors/issues by doing [1] `guix pull -d` (as root and user) [2] `guix package -d` (as root and user) [3] `guix system -d` (as root) [4] `guix gc -d` (as root) [5] `guix pull` (as root and user) [6] `guix upgrade` (as user) [7] `guix system reconfigure` (as root) [8] Repeating steps 1 to 4.
<nckx>Sounds like a fun night.
<raghavgururajan>Hey Toby o/
***jonsger1 is now known as jonsger
<dstolfa>nckx: do you have some time later today or tomorrow to take a look at the strongswan patches? i have some time to fix things during the weekend if necessary :)
<dstolfa>i've been using it for a few days now without issue, and swanctl is still TODO + docs
<nckx>Yes, that's on the menu for today.
<dstolfa>thanks! i'll be around on and off, have to get some groceries, nice day and all, but will be around :)
***pjotpr is now known as pjotrp
<muradm>hi, can someone point to an example of rust package with Cargo.toml as virtual manifest?
<muradm>i.e how to properly build rust cargo project with subprojects
<muradm>this what I get for now: error: manifest path `/tmp/guix-build-greetd-0.7.0.drv-0/source/Cargo.toml` is a virtual manifest, but this command requires running against an actual package in this workspace
<muradm>should I build every subproject and then use them as inputs to main one? O_o
<ixmpp>raghavgururajan, clacke: heh, hi
<ixmpp>clacke, projectmoon: that bridge has one massive bug: you can never leave a room
<projectmoon>That sounds like a problem
<ixmpp>It is, i had to block 5 of them
<ixmpp>Theres a 4(?) year issue on github but they dont care cause it doesnt affect matrix
<daviwil>Hey folks! Hopefully this is an easy question to answer. I'm creating a custom Guix installer image and I'd like to add a service to the default list %installation-services (which isn't exported). I'm trying to use `(operating-system-services this-operating-system)` to access that list like so:
<daviwil>When I do that, I get this error: "guix system: error: this-operating-system: cannot be used outside of a record instantiation"
<daviwil>Seems strange beacuse I'm using `this-operating-system` inside of an `operating-system` instantiation
<raghavgururajan>Hmm. `guix archive --import` just stays active and does nothing. No error or abort.
<roptat>daviwil, mh, I'm not sure how this-operating-system is supposed to work, but oviously you can use (operating-system-services installation-os) instead
<muradm>daviwil, i had similar problems, so ended up using %installation-services as base
<daviwil>roptat: *headslap* Thanks, that totally makes sense
<roptat>raghavgururajan, it's waiting for standard input
<muradm>I suppose one should not attempt to apply general "inherritance" approaches with guile/guix
<roptat>it's not the same as inheritance in C++ for instance, but I'm not sure about the details
<raghavgururajan>roptat: Oh I pass the path to nar file in the command. `guix archive --import /path/to/foo.nar`
<muradm>daviwil, try taking (operating-system ...) outside of define, I suppose it should work
<raghavgururajan>roptat: Am I doing it wrong?
<dstolfa> -- this might be relevant to guix?
<dstolfa>(the whole thread)
<daviwil>muradm: Could be that. Using `(operating-system-services installation-os)` got me past this particular issue at least
<roptat>muradm, daviwil ah: "This macro may only be used in the context of the definition of a thunked field.", which systems is not
<bricewge>guix archive --authorize write to /usr/local/etc/guix/acl on guix system o_O
<daviwil>roptat: Ah! OK, I didn't see that in the docs
<bricewge>What is happening?
<roptat>bricewge, oh, someone forgot --localstatedir=/var
<roptat>is it the guix package from guix, or a local checkout?
<nckx>daviwil: You could (not saying this is good style of future-proof) force access to the variable you actually want with (@@ (gnu system install) %installation-services)))
<daviwil>nckx: Cool! I haven't seen that pattern before
<roptat>raghavgururajan, guix archive --import < /path/to/foo.nar instead
<bricewge>roptat: -_-a local checkout
<roptat>bricewge, yeah, then reconfigure with --localstatedir=/var
<bricewge>roptat: I did not expect it to fail that way. Thanks!
<daviwil>Interestingly, now I get this error after basically just reassigning the same services list: "guix system: error: more than one target service of type 'account'"
<daviwil>I would expect that it'd replace the service list with the new one I give it
<roptat>bricewge, it's even worse, if you forget --localstatedir, it might get the store and database out of sync, and that can lead to troubles if you run the daemon or gc from the checkout
<roptat>daviwil, ah I was wrong, you can't use operating-system-services, because it appends operating-system-essential-services and operating-system-user-services
<roptat>so probably, use (operating-system-user-services installation-os) instead
<roptat>that will prevent the append to happen
<daviwil>roptat: Thanks! I should have paid closer attention to the install.scm file
<roptat>daviwil, I'm looking at (gnu system) actually ^^
<bricewge>ropat: Ok i'll have a look at that to
<daviwil>roptat: Yep, seems that image creation is proceeding now, thanks for the help!
<raghavgururajan>roptat: I added a key to `authorized-keys' to the `guix-service-type'. But system reconfigure fails with buidling acl.drv.
<muradm>daviwil: did you solve it?
<raghavgururajan>I have 3 other keys though, which doesn't cause this error.
<daviwil>muradm: Followed roptat's suggestion to use `(services (operating-system-user-services installation-os))`
<roptat>raghavgururajan, what's the error message?
<roptat>(in the drv.bz2, probably)
<muradm>daviwil: id you will want to change which services should be started or not, look at installation-os it self, it uses it like (services %installation-services), so you can use (services (modify-services %installation-services ....))
<muradm>using it from installation-os some how evaluates everything and you get that error with already defined
<daviwil>muradm: Unfortunately %installation-services can't be used directly, it isn't exported from `gnu system install`. I'm just trying to add a new service to the same list. Got it working now!
<muradm>just like modifying %base-service
<muradm>meh.. then
<raghavgururajan>roptat: My bad. Missing parentheses.
***jonsger1 is now known as jonsger
<muradm>in cargo-build-system, in modified phase, how to access (ice-9 regex)? i tried #:modules and #:imported-modules
<muradm>with no luck
***lukedashjr is now known as luke-jr
<pkill9>i wish guix would show everything it's going to download
<yoctocell>pkill9: guix package --dry-run ?
<pkill9>no, the problem is it doesn't know what it will download
<pkill9>it shows some things
<pkill9>but not all
<pkill9>something to do with it not knowing everything it will need
<pkill9>i think it's to do with grafts
<yoctocell>ah, ok
<civodul>if it's any consolation, it's explained at
<civodul>i'm guessing it's not much of a consolation though :-)
<minikN>Hello. Could someone explain to me if/how it is possible to set env vars inside my /etc/config.scm? I'd like to change $ZDOTDIR, either setting it directly in my scm file or defining it inside /etc/zshenv. For both approaches I find little to no information online.
<iung>Hello! Asking for help with guix installation on btrfs raid0. Failing to mount root subvolume at startup. Example error message from guile-rescue: "kernel: BTRFS error (device sda): devid 2 uuid 65c21020-6607-425b-a36b-5455d80ab0d1 is missing". I can mount it from live usb. I'm confident config is standard and correct. Internet suggests that
<iung>mounting is made too early, that 64-btrfs.udev if placed in iniitsomething should delay it properly. Looking for your help,
<bricewge>minikN: I'm configuring zsh by extending etc-service-type
<minikN>bricewge: Reading and trying to understand :-)
<iung>Iow. I have btrfs raid0 flat layout. I installed guix with config that says 'root' subvolume to be mounted as "/". But it doesn't -- error: subvolume is missing.
<bricewge>minikN: It just populate /etc/zshrc and /etc/zprofile, you should be able to add other /etc configuration file from there
<minikN>bricewge: This actually helps me quite a lot, thanks!
<iung>.. trying to reproduce in a vm.
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | ⚠️ guix pull broken: | 1.3.0 is out! | videos: | bugs & patches: | paste: | Guix in high-performance computing: | This channel's logged: https://logs.'
***nckx changes topic to 'GNU Guix | ⚠️ guix pull broken: | 1.3.0 is out! | videos: | bugs & patches: | paste: | Guix in high-performance computing: | Logs: https://logs.'
***nckx changes topic to 'GNU Guix | ⚠️ guix pull broken: | 1.3.0 is out! | videos: | bugs & patches: | paste: | Guix in high-performance computing: | Logs:'
***nckx changes topic to 'GNU Guix | ⚠️ guix pull broken: | 1.3.0 is out! | videos: | bugs & patches: | paste: | Guix in high-performance computing: | Logs:'
***ChanServ sets mode: -o nckx
<viivien>I’m glad I set up a local clone on my system :)
<mitzman>hm, I might be a bit challenged - i've created a directory X/packages/package.scm. In package.scm, i've defined the module (X packages package), and the package inside is defined publicly. any reason why guix install -L ... package fails? it does read the package file, gives hints about naming
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | ⚠️ guix pull broken: | 1.3.0 is out! | videos: | bugs & patches: | paste: | Guix in high-performance computing: | This channel's logged:'
<nckx>That's clevererer.
***ChanServ sets mode: -o nckx
<pkill9>civodul: why does guix not show everything that will be downloaded again?
<bandali>nckx, fyi debbugs currently seems to be down too
<pkill9>'again' meaning, can you remind me again, because I can't remember
<nckx>bandali: IIUC, that's handled relatively well: well-configured relays will retry until mails are accepted again, and issues.guix will still show all the bugs (and none can be added so it's not even stale :) — correct?
<nckx>It won't really differ from the occasional mysterious 6h debbugs delays that happen occasionally.
<bandali>nckx, yeah i'd hope so. email is generally quite resilient
<nckx>But thanks for the heads-up of course.
<yoctocell>Is there anywhere I can buy Guix stickers?
<bandali>(and i'm not really familiar with inner-workings of issues.guix)
<nckx>bandali: It uses good ol' mu to mirror debbugs, then operates on that.
<bandali>nckx, nice
<nckx>yoctocell: Good question! …I don't really know. I think cbaines_ printed the last batch back when FOSDEM was still legal; *perhaps* they're willing to post them…
*nckx stupidly tried to check
<bandali>the vm specifically should be back again now
<nckx>Different VM I guess.
<bandali>yeah i meant
<yoctocell>nckx: Cool, hopefully I can get my hands on them :)
<nckx>yoctocell: Which continent, if you don't mind me asking? You can be more specific if you like.
<yoctocell>Sweden, Europe
<nckx>Looking at some Web sites it turns out I don't know much about stickers. We should discuss this on the ML when it's back up.
*nckx AFK.
<ss2>Is someone having problems to open ?
<apteryx>civodul: ./pre-inst-env guix build ppsspp --with-latest=ppsspp -> guix build: warning: could not determine latest upstream release of 'ppsspp' -> proceeds to build the current version; shouldn't that be a hard-stop warning? Otherwise i might think it built the latest version (the warning is soon burried in a ton of output).
<apteryx>hard stop error, I meant.
<yoctocell>ss2: yes, see
<apteryx>civodul: my the next attempt ends with a segfault, eh:
<apteryx>non-deterministically, it seems
<apteryx>oh, this time with a backtrace:
<apteryx>In procedure scm_from_stringn: NULL string pointer
<apteryx>that's with guix from the current master branch (./pre-inst-env guix ...)
<ixmpp>Updating channel 'guix' from Git repository at ''...
<ixmpp>guix pull: error: Git error: failed to connect to No route to host
<ixmpp>Savannah down?
<ixmpp>Ah, scrollback
<ixmpp>Inconvenient that i now cant update
<ixmpp>Until thats over
<viivien>I have a hourly updated mirror at
<mbakke>there is also
<ixmpp>I just added `--url=`
<ixmpp>Seems to be doing something
<apteryx>the opencv test suite is indecently long
<katco>bricewge: thanks for that snippet! that automates getting the installer to output onto the serial console. i'm a little confused though because even though the installer displays now, it looks like it still switches to tty1 after booting? i'm unclear on how i'm meant to continue with installation over a serial console =/
<katco>i.e. i get this: "NOTICE: vt: using tty /dev/tty1 (open_tty() in src/uterm_vt.c:285)"
<katco>and then nothing
<bricewge>I didn't remeber if I actually managed to use the graphical install through the console. But I curious to know since I have a a pcengines too, but not running Guix since it's lacking advanced networking configuration
<bricewge>The switch to the TTY is documented
<katco>bricewge: ah! i have another apu2 that is waiting on the newer advanced networking stuff, but this one will live in an environment where the simpler model is acceptable
<katco>bricewge: yeah i've read through that section a few times. i don't know what to do. i'll probably continue to install this installer image to start openssh by default so i can go for there?
<bricewge>I'm wonderinf if kmscon support serial console, as it's what used istead of a normal tty
<bricewge>Oops, line 324 "kmscon-service-type"
<bricewge>A workadound would be to connect the drive/card of your apu2 on an other computer where you would do a "guix system init"
<katco>bricewge: hm, yeah. despite the name `virtual-terminal` i wonder if we could just put `ttys0`
<katco>bricewge: i think i'd sooner do a custom install image where i can ssh in
<katco>re. connecting my apu2 to other computers
<bricewge>Using the default Guix image to install through a serial console would be great
<katco>yeah it, it would really be great if i could have just popped in the usb drive and had it work like normal. i'm going to try setting the `virtual-terminal` field to `ttys0` and see what happens. barring that, i'll just enable openssh to start by default
<civodul>apteryx: re --with-latest warning, yes, maybe this could be an error instead
<civodul>apteryx: could you email bug-guix for the segfault?
<civodul>i get the "NULL string pointer" exception too
<bandali>fyi savannah seems to be back again (for?) now
<dstolfa>bandali: thanks for letting us know :)
<nckx>bandali: Thanks!
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | 1.3.0 is out! | videos: | bugs & patches: | paste: | Guix in high-performance computing: | This channel's logged:'
***ChanServ sets mode: -o nckx
*nckx crawls back into their hole.
<bauermann>civodul, I noticed that the texlive package is 2 years old. instead of trying to pin down exactly what is wrong with it, do you think it's too crazy to update it to the current version and see if it solves the heisenbug?
*dstolfa has finally configured emacs
<zap>dstolfa: wow, i think ill never do it, its like lifetime journey :)
<dstolfa>zap: the trick is to never use melpa
<zap>I use guix manifest to spawn emacs
<zap>with all packages
<dstolfa>does guix package all the elpa stuff?
<pkill9>it tries to
<zap>but emacs packages are trivial anyways
<pkill9>i wonder how many emacs packages need patching?
<raghavgururajan>So is I create software bundle using `guix pack -S /opt/gnu/bin=bin foo`, what is the procedure that other user should do?
<raghavgururajan>Is there a tar command that put stuff in correct places?
<bandali>dstolfa, nckx, cheers :-)
<zap>raghavgururajan: I think you just need to unpack it in /opt no?
<civodul>bauermann: mbakke prepared a patch set updating it:
<civodul>i'm sure feedback would be welcome!
<civodul>(i haven't taken the time to look at it yet)
<bauermann>civodul, ah, awesome! do you now if it fixes bug 48064? if you don't I'll see if it does
<civodul>i don't know
<bauermann>(though not looking forward to having `guix download` eating up 3GB of my precious nvme swap because of
<bauermann>ok, I'll test and report back
<civodul>raghavgururajan: like zap wrote, unpack to /opt; but do try -RR instead, where the tarball can be unpacked anywhere:
<civodul>bauermann: gasp, i had forgotten about that bug, ouch!
<zap>yep was just about to write about relocatable packs
<bauermann>civodul, I've been wondering if it's feasible/desirable to rebase guix on current nix/libstore. I'm not sure, but I think it's likely that they have fixed this bug (checking that is on my to do list)
<raghavgururajan>zap and civodul: But that is a symlink patch right? extracted file sshould be at /gnu/store ... profile ..??
<ixmpp>guys, i have a shepherd service that depends on a python package
<ixmpp>it's not a python app, but it uses python
<ixmpp>so i need that python package in the environment
<ixmpp>how do i achieve that?
<ixmpp>i can do it imperatively by doing guix install, sourcing etc/profile, and running it as me
<zap>raghavgururajan: sorry, i didn't get what you wrote
<raghavgururajan>zap: User has to first extract in root folder so that /gnu/store/stuff is formed?
<zap>in case of relocatable pack it doesnt matter where you extract all the links are relative
<raghavgururajan>Oh siee.
<pkill9>i have an idea, if space runs out during build, guix pauses the build and collects garbage
<raghavgururajan>I didn;t do relocatable
<raghavgururajan>only --symlink
<zap>raghavgururajan: then you unpack in root
<raghavgururajan>so after extracting, gnu/.. stuff should be put to /gnu and opt/... stuff should be put to /opt right?
<zap>right sorry, i wrote the wrong comment above, when I said unpack in /opt
<raghavgururajan>no worries.
<zap>civodul: I found some weird behaviour when working with inferior-eval...
<apteryx>civodul: will do when Debbugs come back online
<zap>the repl seems to somehow lag behind. It returns me results of previous evaluations
<apteryx>seems it already did (Debbugs), great
<katco>bricewge: it was a long shot, but specifying ttyS0 does not work. it errors with the following: "WARNING: vt: cannot find the currently active VT (25): Inappropriate ioctl for device (real_open() in src/uterm_vt.c:328)". but i created an ISO with openssh enabled and can continue the installation from there. i just need to figure out how to manually invoke the installer. but my daughter just got home, so a task for later :)
<mbakke>bauermann: unfortunately TeX Live 2020 does not solve the indeterministic build failures
<katco>bricewge: i was able to start the graphical installer manually from an ssh session. this was pretty easy, but it is unfortunate that i had to build my own ISO to enable openssh to get this far.