IRC channel logs

2017-07-04.log

back to list of logs

<quiliro>;;; Failed to autoload read-pid-file in (shepherd service):
<quiliro>;;; ERROR: missing interface for module (shepherd service)
<quiliro>ERROR: In procedure open-process:
<quiliro>ERROR: In procedure open-process: Cannot allocate memory
<wigust>ng0: Hi! Did you solve this https://lists.gnu.org/archive/html/guix-devel/2017-02/msg01082.html ?
<ng0>yes
<ng0>see a much later thread, this month
<ng0>(or bug thread?)
<ng0>I'm thinking about the best way to improve the documentation on this
<ng0>one moment
<ng0>wigust: I think this one: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=27386
<ng0>I lag behind at the moment, I will finish this
<ng0>I don't understand how the control mailserver of debbugs works
<ng0>reopen… but how?
<ng0>there are no provided examples for how I should send these commands
<ng0>simply put it at the start of the body of the message?
<ng0>yep.
<wigust>ng0: Thanks! I also need to mention that need to add `export` before $GUILE_*.
<wigust>.bashrc: export GUILE_LOAD_PATH="${GUILE_LOAD_PATH}:/run/current-system/profile/share/guile/site/2.2"
<wigust>GUILE_LOAD_COMPILED_PATH="${GUILE_LOAD_COMPILED_PATH}:/run/current-system/profile/lib/guile/2.2/site-ccache:/run/current-system/profile/share/guile/site/2.2"
<reepca>When the manual says that build processes need to be launched under separate UIDs for determinism... any examples what specifically could cause them to interfere with each other? I ask because I notice that the daemon doesn't use user namespaces, and wonder if that could eliminate the need for a pool of users.
***Guest74941 is now known as sturm
<efraim>reepca: maybe protecting against a malicious package trying to break out of its build environment AND into another one? Where the file permissions would act as defense
<efraim>I don't remember if we require a kernel with user namespaces
<efraim>Or when it was added
<efraim> http://man7.org/linux/man-pages/man7/user_namespaces.7.html says namespaces are a linux specific feature, might interact poorly with the hurd
<Muto>I installed Guix on my laptop computer, when I boot into my Guix partition, it starts with the Guile listener. I couldn't find anything in the reference manual about this. Is it normal?
<catonano>sneek later Muto most of the people here are European. Your last message (about the listener on your laptop) was very early in the morning. Maybe you could try the mailing list
<sneek>Got it.
<rekado_>efraim: on the hurd virtualisation is easier and we would not need user namespaces.
<rekado_>sneek: later tell quiliro When “make” fails just try again. Could be that you ran out of memory.
<sneek>Got it.
<civodul>Hello Guix!
<sneek>civodul, you have 1 message.
<sneek>civodul, wingo says: take a look at this backlog around 22h30
<civodul>wingo: on #guile and #guix?
<wingo>guix
<wingo>moin :)
<civodul>i'm looking at https://gnunet.org/bot/log/guix/2017-07-03#T1437793 but i'm not sure what i should be paying attention to :-)
<civodul>the Shepherd failure?
<efraim>reepca: rekado_ says not to worry about user namespaces on the hurd
<rekado_>civodul: https://gnunet.org/bot/log/guix/2017-07-03#T1437677
<rekado_>civodul: starts with “well, it is bedtime for me”
<wingo>civodul: yeah what rekado says. i think we have an 'accidentally quadratic' thing to fix :P
<wingo>my hope is that the compiler algorithms themselves are fine, it's just this runtime issue of marking large stacks
<wingo>obviously there are improvements to make in the compiler but i am hoping they don't have to be algorithmic
<rekado_>ACTION notes that boostrappable.org (and thus bayfront) is down again
<civodul>wingo: oh, glad that you found about this
<civodul>what happens if we disable the stack mark function and let libgc do its thing instead?
<civodul>it won't be precise
<civodul>but maybe good enough?
<wingo>you could give it a go :)
<civodul>indeed, probably not today though
<wingo>i think it would be a useful experiment but i prefer to fix our mark function
<civodul>yes of course, that would be even better
<jsierles>any tips for dealing with this situation? guix package: error: unsupported manifest format
<rekado>jsierles: can you give us more context?
<rekado>jsierles: what I do in situations like this is grep for the error message
<jsierles>rekado: i'm just running 'guix package -i julia'
<jlicht>'sup guix
<jsierles>i found some irc logs that suggest rolling back generations, but not sure how to do that.
<jsierles>i set my .guix-profile manually to point to the binary tarball profile
<jsierles>ok, the daemon seems to be in trouble: unexpected Nix daemon error: unexpected end-of-file
<jlicht>jsierles: You can find older profiles in /var/guix/profiles/per-user/<your-user>/
<jsierles>jlicht: yeah. i linked to them, but it looks like the daemon is unhappy.
<rekado>ACTION goes afk for a few hours
<jsierles>is it possible that if i did a 'guix pull' and ran the daemon, then rolled back, that the db could be corrupted?
<jsierles>any way to know when upgrading the daemon itself is not backwards compatible?
<jlicht>Sorry, no clue :O
<reepca>I don't think "guix pull" updates the daemon
<efraim1>I still don't understand what gemma does, but I've written a patch to make it build on all our architectures
<reepca>jsierles: rolling back one generation can be accomplished with "guix package --roll-back". First, let's set your .guix-profile back to what it was before (symlink ~/.guix-profile to /var/guix/profiles/per-user/<your-user>/guix-profile). Then run "guix package --roll-back", then we'll see what happens.
<jsierles>reepca: rollback also throws that error
<jsierles>unexpected end of file, from the daemon. so guessing something is corrupted
<jsierles>i suppose if i want to start from scratch, whatever is in my store will be useless without the database being in sync
<jsierles>ok, just wiped the store and db. starting from scratch worked.
<jsierles>now to rebuild the world again...
<boskovits>hello
<boskovits>I need some help
<boskovits>I guess I miss some input for a build, but cannot find out which
<boskovits>The problem seem to be related to PATH.
<boskovits>Before I source path running ant gives me a message, that build.xml not found, which is ok.
<efraim>sneek: later tell lfam https://www.pine64.org/?page_id=7147 4GB RAM and 64 GB emmc for $75, should be supported by 4.12
<sneek>Got it.
<efraim>sneek: botsnack
<sneek>:)
<boskovits>After sourcing the PATH from the build however, I get JAVA_HOME is not defined correctly.
<jsierles>on 0.13.0, i'm seeing failures building Python 3.5.3: https://gist.github.com/jsierles/8c26e4e7c171606344f1589a27232511
<jsierles>i guess this means that a release does not guarantee a complete set of working packages for a given architecture?
<civodul>jsierles: unfortunately sometimes package build processes are non-deterministic, and sometimes they depend on details not under Guix's control (kernel, hardware, etc.)
<civodul>jsierles: are you on Btrfs?
<jsierles>civodul: no
<civodul>anything particular about this file system?
<jsierles>i'm building inside a docker container which is on overlayfs
<civodul>here you go :-)
<jsierles>how did you guess it was related to the filesystem?
<jsierles>OK, so maybe I need to set TMPDIR to a normal fs?
<civodul>without looking in details, this Python test is about inode numbers and the likes
<civodul>it's the kind of detail that can change on "unusual" file systems
<jsierles>ok. good catch
<civodul>that said, the right thing for us would be to fix the test to tolerate these changes
<civodul>and to report it upstream
<civodul>so if you feel like emailing bug-guix and/or the CPython folks, that would help improve this
<civodul>setting TMPDIR to a "normal" fs should help, yes
<jsierles>ok, thanks. i saw tar tests even failed too
<civodul>may be fs-related as well
<jsierles>maybe this is related to the previous problem
<jsierles>trying the TMPDIR fix. at least at the end of all this, i will have a good guide to using guix under docker.
<civodul>yep :-)
<jlicht`>hi guix! How can I have a service which does not 'autostart' in my operating system declaration?
<jsierles>civodul: ok, working much better now!
<jsierles>seeing another issue now. gettext build is stuck on 'test-lock' process
<jsierles>output is stuck at: Starting test_rwlock ..
<jsierles>i see it's related to multicore machines, and fixed in core updates
<jsierles>if i want to make a 'guix pull' reproducible across containers/servers, can i just jot down which profile is being used after running it, and ensure that profile is used as the default?
<jsierles>to avoid having to run 'guix pull' all the time.
<jsierles>assuimg i'm using guix pull with a git hash.
<pmikkelsen>hello guix
<oriansj>hello pmikkelsen
<civodul>jsierles: re test-lock, yeah, it's a nasty problem fixed in core-updates
<civodul>only shows up on machines with "many" cores (more than 4 or 8)
<civodul>for 'guix pull', yes i guess you could pin to a specific commit
<civodul>it's not very convenient ATM
<jsierles>yeah. so if i wanted another machine to use the profile generated by the pinned guix, would it be downloadable as a substitute?
<jsierles>or do they have to run guix pull and wait for builds?
<jsierles>i'm on an 8 core machine. maybe i should use a bigger one as builds seem to use all cores
<sirgazil>Howdy! Verifying the authenticity of guix 0.13.0 binary installer with GPG, I get messages like: "gpg: no ultimately trusted keys found", "gpg: WARNING: This key is not certified with a trusted signature! There is no indication that the signature belongs to the owner". What does that mean?
<elreydelaswasas1>how to decrypt a partition in guixsd
<elreydelaswasas1>?
<civodul>jsierles: you can fiddle both with --max-jobs and --cores
<civodul>often --cores=4 is the best you can get, so then you can increase --max-jobs
<ng0>my xfce4 update series now fails differently after the xfce package changes you made. but I'm no gtk expert yet. Is anyone interested in reviewing the series (actually one patch which simply includes many updates which can only be updated all together)
<elreydelaswasas1>how to decrypt a Luks encrypted partition in the guixsd operative system?
<elreydelaswasas1>in the part ofthe guixsd installation where we edit the config.scm template how must we configure the template if we live in Colombia and we have a Luks encrypted root partition?
<ng0>internet search for my error output brings me this: libxfce4menu is deprecated and not used since Xfce 4.8. It's been replaced by garcon. I just have to figure out why libxfce4menu gets in there, as I updated garcon on the way of updating xfce4-pannel
<rain1>the qemu image on the guix download page does not have SSH correct?
<sirgazil>After installing Guix 0.13.0, and enabling substitutes, I run "guix package -i glibc-locales" and many packages are building, I wonder why. Also, it is downloading packages that seem unrelated. For example, python 2 and 3 where downloaded, is that normal? It's currently building GCC 5...
<sirgazil>ACTION goes for lunch
<rain1>sirgazil: im a little unsure why this would happen without passing the --fallback flag
<rain1>but i have experienced the same thing, it will build the entire dependency tree to build the package you request
<rain1>I think this is expected
<ng0>what's the current size of hydra.gnunet.org mirror? I just want to make sure that my nginx cache doesn't run out of space after I upgrade the disk
<ng0> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=27298 should I send a new version which just does what I asked in comment 3 about? It would be really nice to get this closed.
<ng0>Because, we ship fossil 1.35, and 2.0 closed the sha problem in fossil and then came 2.1 and 2.2
<ng0>so our fossil is pretty old by now
<elreydelaswasas1>i cant format a partition when i had another partition luksformated because when i try to format that partiton there appears the next error Cannot format device /dev/sdX which is still in use
<elreydelaswasas1>how to fix error guix pull error: build failed: while setting up the build environment: unable to make filesystem ´/mnt/home´ private: Invalid argument in guixsd installation?
<caolanm>I'm trying to install guix as a package manager on top of Debian 9, and I keep getting "substitute: guile: warning: failed to install locale" "substitute: warning: failed to install locale: Invalid argument" when I try to guix package -i anything
<boskovits>hello
<boskovits>I need a bit of help.
<caolanm>I've tried following locales setup in 2.6.1 https://www.gnu.org/software/guix/manual/guix.html#Application-Setup, but I still get the warning
<caolanm>any pointers of where to start untangling this would be great
<boskovits>What is the cleanest way to add a directory to the PATH in a package build definition?
<jlicht>boskovits: Do you want to make use of a certain PATH at build-time, or at run-time of the package-to-be-built?
<boskovits>I need it at build time. I see a method in unrtf for example, where getenv string-append and setenv is used.
<boskovits>I guess I will try that.
<boskovits>ok, i think it will work fine
<Buglouse>I'm trying to find the conf for shepherd services, like for example 'host-name'
<elreydelaswasas1>one of my partitions contains a crypto_LUKS file system how can i decrypt that crypto_LUKS file system?
<efraim>Has anyone set up cuirass on not guixsd?
<rekado>sirgazil: the GPG warning tells you that you have not told the GPG programme whether or not you trust the key.
<rekado>sirgazil: that’s okay. You can change this by editing Ludo’s key in your key store and say that you trust it, but that’s not necessary.
<sirgazil>rekado: thanks :) And the certification message too? "This key is not certified with a trusted signature!"
<rekado>caolanm: the warning is harmless but can be annoying. The fix is to tell the systemd service of the *daemon* about where to find locales (i.e. to tell it about GUIX_LOCPATH).
<rekado>sirgazil: that part says that none of the people whose keys you have marked as trusted have signed Ludo’s key.
<rekado>sirgazil: you would not see this message if 1) you had my key and 2) you would have marked my key as trusted.
<rekado>(because Ludo’s key has a signature on it that was made with my key)
<rekado>that’s for the web of trust
<rekado>there’s either direct trust or transitive trust
<sirgazil>rekado: Ah, I see. Those messages are not clear then, imho.
<rekado>yes, they are not
<rekado>very confusing and sometimes needlessly alarming :-/
<sirgazil>Yeah
<rekado>ACTION goes afk again
<caolanm>rekado: that did it, thank you
<Buglouse>I'm un able tp find the definition for 'host-name' herd service.
<sirgazil>Second time this computer turns off trying to build the packages necessary for "guix package -i glibc-locales". Moving on. Let's see if I can do other stuff...
<catonano>hello guix
<elreydelaswasas1>if i have my computer with the screen on energy save mode and i and downloading the packages for the guixsd operative system which key of thekeyboard should i press to see how is the download of packages iwthout aborting it ?
<efraim>I like control
<jsierles>oh, what does it mean when 'updating list of substitutes' just goes on and on
<jsierles>seems when that happens, i have to restart my build.
<jsierles>guix perform-download: error: refusing to run with elevated privileges (UID 0)
<jsierles>does this mean I can't run 'guix' as root?
<sirgazil>Aaand, computer goes down a third time. It can't take the build load :(
<buenouanq>jsierles: this guix on some other distro? are you using sudo or something weird?
<jsierles>buenouanq: yes, on alpine inside a docker container. just running as root
<buenouanq>I know of no reason why you wouldn't be able to use it as root.
<jsierles>ah well. seems to have been a transient failure
<catonano>hi civodul
<civodul>hey catonano!