IRC channel logs
2023-01-20.log
back to list of logs
<ieugen[m]>the tar.gz is unpacked ok and there is a single file which should be put on the path <nckx>ieugen[m]: Just install it into #$output/bin (and make it executable if it's not) <ieugen[m]>something like this: `(#:install-plan '(("bb" #$output/bin/bb))) <nckx>Guix doesn't know which binaries need to be in $PATH, or more correctly, its PATH just refrers to the union of all /bin and /sbin directories for each installed package. <nckx>Just "bin/bb" for the second element. <nckx>Actually I think "bin/" even suffices, to avoid repeating "bb". <nckx>ieugen[m]: It will install to /gnu/store/hash-foo-1.2.3/usr/bin just fine, but usr/bin is not in PATH so it's pointless. <nckx>Note that this package will run only on foreign distributions, but if that covers your use case for now, enjoy. <Fare>if I have guix running on top of debian on one disk (say sda), what's the right way to system install on another disk (say sdb) without disrupting debian? <nckx>sudo mount /dev/sdbN /mnt && sudo guix system init /etc/guix/system.scm /mnt — the destination is never implied, so as long as you don't explicitly set it to / your Debian root will be safe. <nckx>(Side tip: if you use a boot partition and/or ESP, mount it under /mnt in the expected spot.) <nckx>(So /mnt/foot, not /foot.) <nckx>And do make sure that you're not targetting /dev/sda with the boot loader, but I assume you're that clever :) <ieugen[m]>nckx: the bb binary is statically compiled so it might work <nckx>ieugen[m]: You are very right, I was looking at the wrong tarball. <ieugen[m]>there are a lot of devops tooling pacakged like this (java compiled with graalvm, go pacakges, etc.) which I plan to use <ieugen[m]>is there a cli command to add/list/remove channels to guix? similar to what nix has? <ieugen[m]>this makes using extra channels much more easy <singpolyma>You don't even need a channel to install something in guix, just a file and you can install with install -f or -L. <ieugen[m]>so if I push the package definition to git and get the file reference as an https url I should be able to use it? <singpolyma>(I know, it's a gnu channel, I shoulda wget -o -) <jgart[m]>See the README for mailing list and issue tracker in case you'd like to contribute or discuss more <ieugen[m]>sure, I am still exploring but if things go my way, I'll send a PR for an extra channel <lechner>ACTION lowered the creative threshold for a new good bot name <rekado>mirai: I have a vague memory of submitting a patch for sxml documentation… <rekado>there’s a lot of documentation and a bunch of examples in the original source files that Guile wraps <rekado>hmm, looks like I only sent a patch for sxpath docs, back in 2014. <lechner>nckx / Hi, which font do i need in order to see that UTF-8 codepoint in Emacs, please? <nckx>The bot's IP (gnu:baca) is more familiar to me at this point than the names. <nckx>lechner: font-google-noto has it, even Guix's slightly older version. <jgart[m]>android element client over matrix bridge must be buggy... <nckx>lechner: It was just a person bowing. <lechner>i know, but i am tired of typing M-x describe-char <nckx>Seeing it rendered wouldn't necessarily make that clear; it looks more like someone being caught napping at their desk. <jgart[m]>Is there a service that currently provides substitute availability across Guix channels or lets you filter by Guix channel or substitute provider? <jgart[m]>I guess --substitute-url from guix weather... But does it differentiate package foo in channel bar from package foo in channel baz? <akirakyle>Is mbakke around? I'm wanting to try to get guix system reconfigure working again on my aarch64 machine and currently https://issues.guix.gnu.org/60719 is blocking me. It looks like mbakke authored the commits that unbundled the openbios-qemu-ppc rom and introduced this issue. The question is, why should aarch64 even be building that for <akirakyle>qemu-minimal which is only supposed to support the host's architecture so building ppc firmware would seem to be unecessary? <apteryx>thanks for developing it, by the way! it's growing into a useful, friendly bot. <bienjensu`>I’m trying to build a cmake package, dwarfs (mhx/dwarfs on github), but part of its configure script expects for the source dir to have vcs info. Both `git-fetch` and `url-fetch` of the tarball appear to strip the vcs info (specifically the ‘.git’ in the source directory, which the configure step is looking for). Is there any way to override this behaviour of the fetch step? <apteryx>no, because it'd break reproducibility (the .git metadata is not immutable) <bienjensu`>apteryx: thanks, I thought that might be the case. Going to patch the CMakeLists.txt or versions.cmake file instead to conform with guix packaging requirements. <Guest53>Ay y'all! I'm doing the whole Erase Your Darlings thing, put together a system config that rolls BTRFS subvol ('/') back to a blank snapshot (as part of a custom `mapped-device-kind-open` proc), but Guix doesn't actually handle an empty root very well :p <Guest53>My thought right now are that I could a.) roll back to a snapshot of a `guix system init` directory (which would never be updated...), b.) try to figure out what directories need to exist and roll back to an empty file tree (if that's enough...) <Guest53>c.) modify system activation to do whatever needs done (I've been working to keep my config in one file, this would mean patching Guix...) <Guest53>d.) Roll back to blank and then actually run `guix system init`, on-boot (I imagine it would take as long as it does under any other circumstance? and how would I know the system store path?) <Guest53>e.) Pull off something clever with `#:volatile-root?` and a few `mkdir-p` calls <Guest53>Option E intensely preferred, but I'm going to need to sleep on it and browse the sources for a bit to figure out what the best option'll be <Guest53>Reducing unnecessary disk writes would be great, and I don't data in the "blank" snapshot, so I'm going to be looking for a way to use overlay mounts and will probably have to give in and patch guix (short of forking the whole initfs into my config) <Guest53>I guess I might be able to pull what I need out of `guix system init`, everything short of store population, but it occurs to me that I might get past the new few roadblocks just by making sure that I make the directory for the store mount, oof, I'll check that <Guest53>Ah, well, that was easy; creating /run seems to have done it. Would ya'll be interested in adding a single `mkdir-p` call in `linux-initrd.scm`, or where-ever suites your fancy? Disclaimer, my config sure looks like that was all I needed, but there could be breakage that I haven't found yet <Guest53>hmm, this isn't gonna play so well with hibernation <Guest53>> rolls BTRFS subvol ('/') back to a blank snapshot (as part of a custom `mapped-device-kind-open` proc) <Guest53>This part right here, which is how I'm getting away with not patching the initrd itself <Guest53>Mapped devices can't depend on each other, can they? <Guest53>I'm using volatile root, will try turning it off to see what happens tomorrow, but maybe that's enough to make hibernate "just work"; gotta try it! <ArneBab>It would be great to get this merged quickly! <cgenie[m]>Can I build a system definition as a docker image? It seems guix pack didn't work for me <cgenie[m]>ah ok, it's guix system image --image-type=docker my-system.scm <rekado>danialbehzadi[m]: looks like you forgot to add a commit message <rekado>danialbehzadi[m]: can you tell us why the other changes were needed to update the package? <rekado>ArneBab: merged with commit 0d169401cbc9e1a418290d4d6cb03dac8babcde3 <ArneBab>rekado: besides, a really nice thing is that I could use my fix instantly by running with /path/to/pre-inst-env guix shell maven :-) <danialbehzadi[m]><rekado> "دانیال بهزادی: can you tell us..." <- The build system was changed from setup.py to pyproject.toml and bash-completion was added. <cgenie[m]>i added --shared=/tmp/psql=/tmp/psql to provide a socket interface to connect to the db, i can only connect as postgres user, not as "test" <cgenie[m]>i also added (extension-packages (list postgis)) and it seems it's not installed as well (select * from pg_extension returns only plpgsql) <cgenie[m]>performing post-bootstrap initialization ... sh: line 1: locale: command not found <cgenie[m]>also, i can specify (postgresql postgresql-15) however, create extension postgis throws an error that postgis was built against psql-14 <ekaitz>civodul: I need to buy some hardware for my project (if they give me the funds). Is there any reasonable way to donate it to Guix when we are finished with it? <civodul>yes, i guess that would be the best option <jakfrost>Hello GUIX community, I am new to Guix, but not GNU. I am looking forward to getting more proficient with it. <ekaitz>jonsger: we don't have it yet, but it will be risc-v <numerobis>Hi #guix! A few monhtts ago I reported an with gitolite: specifically, the creation of the 'hooks' directory fails on repository creation. The issue (I think) came from the fact that gitolite uses a git executable from the "git-minimal" package, which lacks support for hooks. I just updated my system, and it seems that the issue is still there, and <numerobis>so I was wondering whether this issue was being worked on? (or did I mess up when updating — I'm still a guix noob?) <numerobis>I should mention that I am talking about the gitolite service of guix system, here. <nckx>jakfrost: Guix containers aren't related to the OCI (Docker) implementation of containers, but ‘guix system image’ can create Docker-compatible images if needed. <nckx>Many features are aimed at ‘developers’, but not at ‘Developers’, if that makes sense. Like, don't expect Kubernetes Ansible Integration for Devops® to be considered the coolest thing ever in Guix circles. <nckx>numerobis: What's the bug number? All gitolite/git-minimal/hook-related bugs I can find (in a very quick search) are closed. <numerobis>nckx: I just wrote about the bug here; I did not submit an issue or anything to the official repo (but I'm happy to do so if you point me in the right direction!) <nckx>Reporting bugs here is (1) welcome (2) risky, as IRC is quite ephemeral. It's logged, but nobody peruses the log for new bug reports. <nckx>You can simply send an e-mail to bug-guix@gnu.org to file a new bug. <ekaitz>when talking about build-systems what are implicit-inputs? <ekaitz>also, i'm dealing with zig's build system and they do some crazy stuff with the libc, they embed it in the compiler... should we avoid using that one? <ekaitz>i don't know how does that map to our style of making things <jakfrost>nckx: so the standard container (not system containers) are Linux containers I guess and the system ones are too but "fatter" and you can create system containers to be OCI compliant. <gnucode>morning guix! Guess who managed to set up a nftables service and then lock themselves out of their linode server ? Me! <gnucode>what's annoying was I tested the configuration file beforehand... <gnucode>lechner: are you talking to mirai or me? <gnucode>oh, well, basically I can't ssh into the server anymore. I can put the nftables.conf into a pastebin if you like. <gnucode>I think linode somehow will let you see the boot up process and boot up a previous generation, but I'm not sure how to do that. <gnucode>nope. Before I reconfigured, I did a sudo nft -f nftables-lamora.conf <gnucode>and the firewall seemed to work really well. <gnucode>I could ssh in, my website still worked. But after I reconfigured, I can't ssh in and my website is down. <gnucode>I am pretty sure that link I put up is the nftables configuration that I reconfigured with, but I can't ssh into the box to verify.... <gnucode>also mirai did you creat the function version-major+minor ? <gnucode>and what's the issue with policy drop? <mirai>oh, wait, I might be wrong. Thought there was a reject policy <gnucode>lechner: option 1 I give up computing, all my worldly possesions, and become a monk. <mirai>dropping packets instead of rejecting them is a source of headaches <mirai>makes troubleshooting harder and doesn't really "hide" anything <mirai>idk, does nftables have reject policy? their wiki says "Currently there are 2 policies: accept (default) or drop. " (could be outdated though) <lechner>reject is probably better than drop for most situations <mirai>in general I prefer reject over drop <mirai>at least gives us a clue to where things got wrong <mirai>gnucode: version-major+minor is from guix packages <lechner>gnucode / what's weird? you cannot boot into an earlier generation? <mirai>lechner: when you're being DoS then you'll consider taking other actions <gnucode>when I launch the "weblish" console I don't see anything. when I launch the "glish" console, I see the "error & messages" tty. nginx: configuration file /gnu/store/sanehusanehulr,chusanehuh-nginx.conf is ok. <lechner>mirai / it's so you don't participate in a DoS, i think <gnucode>I'm not sure how to change tty's in "glish". Also I may need to download ungoogled-chromium. Maybe firefox is just a bit outdated for some of this stuff. <lechner>mirai / i think participation would be inadvertent, i.e. just by replying with REJECT <lechner>gnucode / for me, only one of the options worked. i'll try locally <mirai>idk, if DoS is a concern maybe you could write some rule that rejects under some threshold of "unexpected traffic" and starts dropping afterwards <gnucode>oh, by the way, my nftables configuration came straight from guix-src. It's very similiar to the default firewall. <gnucode>mirai: what is chvt oh... change virtual terminal! I knew you were awesome! <mirai>gnucode: maybe shepherd hung? the configuration could be wrong elsewhere <gnucode>mirai: I'll try it, but I don't that will work. even typing in ls doesn't get me any results. <lechner>gnucode / i'd try this LISH Console via SSH <gnucode>glish does, but weblish shows me just a dot. pressing keys on the keyboard changes nothing. <gnucode>and lish console via ssh does the same thing. :( <lechner>from what i remember, glish will also show the grub screen, but you have to be fast <gnucode>well let me try glish then. If it'll show me the grub screen then I may be ok. <mroh>ups, send it twice. idk why, sorry! <nckx>No problem. Just curious if you encountered problems. <mroh>I send a merge control msg. debbugs still list them seperate. What did I wrong this time? ;) <nckx>If you did anything wrong, Debbugs should reply. <nckx>mroh: I don't see them listed as separate? <torb>Hi! I'm afraid I managed to break the guix gdcm package: I submitted a patch to upgrade to 3.0.20, which built fine in my local checkout, but now fails because of 3 failing tests. I have attached a patch to ignore the tests and make the package build, as a temporary measure, at https://debbugs.gnu.org/cgi/bugreport.cgi?bug=60640#94. Would someone be interested to take a look? <nckx>Is there a trivial way to tell ‘setup.py sdist’ that nobody cares about mtimes inside its .tar.gz? <nckx>Or pass a custom ‘tar --mtime=…’? <mroh>nckx: python-build-system has (ensure-no-mtimes-pre-1980). <nckx>This is not a Python build system package. <nckx>I'm currently writing a phase to nuke mtimes before 'install but (1) this might not work, if setup.py generates files during its operation (2) ew. <nckx>…it did not work for that reason. <nckx>I find this whole behaviour questionable so it's tempting to just disable it. <nckx>Oh, but it used to be reproducible in the old version. Damn it. What changed. <Kabouik>What do you guys use as pdf enfine for Pandoc? I used to use pdflatex, but it does not seem to be available as a Guix package. <mroh>nckx: old version of python-build-system or old version of the pkg? <nckx>This is not a Python build system package. <nckx>Previous package version. <Kabouik>Oh, I see. But that's a huge package to install I'm afraid. I was hoping something more minimal would be available. <mirai>there's various texlive flavours <mirai>besides, how would pdflatex work if it the texlive packages it will want aren't present/ <Kabouik>It would not I guess, I just didn't know pdflatex had such dependencies, and was hoping it would just be a smaller utility <Kabouik>Looking through texlive packages, to see if there's a minimal version <nckx>Self-answer: nothing changed, the previous version already had this bug, but grafting is super-reproducible so that's nice… 😒 <nckx>mroh: So (since you seem to know more of Python than I do you are now my victim) this is what it's doing: https://sourceforge.net/p/lirc/git/ci/master/tree/python-pkg/README.rst — installing a tarball of its own Python sources to /share for… Python reasons, I guess. I don't know what pip3 is or why this would be useful, and ‘If you have installed LIRC either from a package […] you don't need this’ is what tempts me to just completely dis <nckx>I mean, I know pip is a ‘package manager’, but why it needs this or if it's even useful on Guix is beyond my ken. <mroh>nckx: you cant use /lib/python3.9/ in our lirc pkg or want to separate it? or these files have that mtime problem? <nckx>$out/share/lirc/lirc-0.10.2.tar.gz contains mtimes. <nckx>($out/share/lirc/lirc-0.10.1.tar.gz did too, this is not a regression.) <nckx>mroh: Is installing a source tarball like this something that is done/sane/common in Python land? I've never encountered it. <nckx>But I do not much venture into this land. The language confuses me & the people dress funny. <mroh>I dont know much of the python land neither. I think, its uncommon to install python files with gnu-build-system?! (Have to study the makefile...) <nckx>The Makefile is explicitly doing this, there's no mystery in the how, only in the ‘why’ (and ‘is this all just a bit silly and safe to disable’). <nckx>By now I'm leaning towards ‘let's disable it and find out, by listening for complaints’. <mroh>sounds like a good idea to me ;) <nckx>Thank you/sorry for serving as my faithful rubber ducky. <bumble[m]>@unmatched-paren I am novice with guix and scheme, but if you would advise me how to set things up in order to debug greetd-wlgreet-sway-session, I will do that. This system now uses a greetd-terminal-configuration, so I believe blinking-cursor issue relates to greetd-wlgreet-sway-session specifically. If you don't reply to this message, I will not bother or message you again. Thank you for all the help you have given me, <minima>(hi everyone! it's been a long time, good to be back) <nckx>Always nice to see a familiar nick return :) <apteryx>nckx: not sure what your problem is, but if a gnu-build-system package contains some python code, it could probably be split into its own package and use pyproject-build-system to build as usual <apteryx>and bypass the make incantations that install it with pip or what not <apteryx>e.g., patman is included in u-boot sources, but packaged separately <Apo11o[m]>Hello! I created a “package” for starship using copy-build-system to install it as a binary. Not sure why but it only gets installed on /gnu/store and I can’t find it on .guix-profile. How can I have it installed also on my .guix-profile? (didn’t really find much on cookbook&manual about this, if anyone can link the section or provide tips would be appreciated )? <Apo11o[m]>nckx: url-fetch from their GitHub repo, build system using copy-build-system. Works just fine but it’s only on /gnu/store <Apo11o[m]>Tried both ‘guix build -f’ and ‘guix package -f’ <bost>Hi. I'm playing with the Xvnc. The docs says try `ssh -L5910:localhost:5910 -- guix shell xclock -- env DISPLAY=:10 xclock`, however I get 'ssh: Could not resolve hostname guix: Name or service not known' <bost>Is it just me? Or the command syntax is wrong? <gnucode>bost I've never actually used Xvnc... no idea... <bost>ACTION Or is it a bug in the ssh? Ugh. <bost>gnucode: my question is about the syntax of the ssh command. SRY for misleading you (Xvnc questions will come later, most probably :-) <gnucode>bost does ssh -L5910:localhost:5910 work ? <gnucode>I don't even know why the -L option... <nckx>apo11o: ‘guix package -f’ should (and probably did) install it into your profile. Can you verify with ‘guix package -I’ that it was? <nckx>‘guix build -f’ installs nothing. <nckx>If the package is installed but nothing is in ~/.guix-profile/bin/, my guess is there's nothing in /gnu/store/…/bin/ either. <bost>gnucode: hmm it doesn't. It really seems like the example from the documentation is buggy. <bost>BTW, `ssh localhost` works <bost>According to Hitchhiker's Guide to the Galaxy the Notion of Progress has been largely overrated. <apo11o>nckx: it's installed as /gnu/store/<hash>-starship-1.12.0 but can't find it on .guix-profile/bin do I have to specify something else in the code to be installed there? <nckx>Things in /gnu/store are not ‘installed’. <nckx>My question is: does /gnu/store/<hash>-starship-1.12.0/bin/starship exist? <nckx>~/.guix-profile is mostly a union. Its bin/ directory will contain the contents of each /gnu/store/…-name-version/bin/ of each package installed into that profile. If your executable is somewhere else (/gnu/store/<hash>-starship-1.12.0/usr/bin/starship, or /gnu/store/<hash>-starship-1.12.0 is a file, or…) it will not show up in that profile either. <apo11o>nckx: not as /gnu/store/<hash>-starship-1.12.0/bin/starship but /gnu/store/<hash>-starship-1.12.0/starship without the bin directory. I guess I should create a bin dir? <apo11o>thanks nckx and apologie for my ignorance <nckx>apo11o: Yep, you should! And no need to apologise for anything. <bost>gnucode: FYI seems like the correct syntax is `ssh -L 5910:localhost:5910 [<username>@]localhost -- guix shell xclock -- env DISPLAY=:10 xclock` <apteryx>bost: hi! was there a problem with the example in the manual? <gnucode>bost nice. What were you trying to do? why do you need to ssh into the local machine? Are you messing with virtual machines? <apteryx>gnucode: xvnc is for remoting a remote session via VNC <apteryx>it's transport is insecure, so it's typically used via an SSH link <bost>gnucode: not really. I'm messing with a physical machine :) <apteryx>bost: ah, I see it, there's no host in the manual for that SSH command <gnucode>hmmmm. If you are messing with a physical machine? Why is there a need to ssh into it? Just practice? <bost>BTW if fact, what I'm trying to solve that the physical machine suspends to RAM if I'm not logged in using GDM. Hmm. <apteryx>gnucode: it can be useful to test multiple sessions concurrently on the same physical session <apteryx>I use ratpoison myself, but I remote to my own desktop to start a GNOME session to test things sometimes <bost>gnucode: I'm not having and extra monitor for the machine. Only an ssh-access. (Or maybe VNC later, if I succeed in setting it up) <bost>ACTION bloody hell my some package broke my emacs. Again! :( <mirai>did you modify auto-suspend to off? <mirai>in the case you did, are you logged in as a user through gdm? <bost>mirai: I'd like to. But I haven't seen any `(xorg-configuration (inherit config) (auto-suspend? #f))` example so far. Only for gdm-configuration. So I don't know if it's gonna work <mirai>the user autosuspend will take priority over GDM settings <bost>mirai: user autosuspend? what's that? <mirai>(I haven't reconfigured my system yet) <mirai>bost: check the manual for gdm-configuration and modify-services <bost>mirai: yeah but do I have the gdm-service installed on that other machine of mine? I'm not sure. I guess I should be having it, but I'd like to know to find it out. <bost>mirai: and `guix repl` and %desktop-services returns '%desktop-services unbound variable' so far. <bost>mirai: pgrep show something, but since i dunno what pgrep-switch shows the full command line params... f-it :( <apteryx>bost: if your other machine is using %desktop-services (e.g. you used the desktop.tmpl template as a base) it should have gdm-service <apteryx>you can check if gdm is running via ssh (pgrep gdm) <bost>mirai: this is insane! ',use (gnu services desktop)' followed by '%desktop-services' throws a screen full of text at me. Like in the Matrix movie <bost>ACTION for this you need that and for that you need something else before and for something else before you need some more stuff and some more stuff you need... grrrr <unmatched-paren>it might just be that %desktop-services contains quite a lot of services :) <bost>unmatched-paren: it prints just a half screen of Matrix. Progress! <bost>unmatched-paren: and the gdm seems to be there. wooaaa. <unmatched-paren>mirai: for the second one, pretty sure the zip should be unpacked... <unmatched-paren>i would recommend greetd with wlgreet, but it turns out it doesn't work for anyone but me, and i can't figure out why :( <mirai>unmatched-paren: it is, but the zip is "flat" and contains some directories <mirai>so you have foo.txt dir/bar.txt .... <bost>unmatched-paren: yeah I was assuming that, but you know... road to hell is tiled by unproven assumptions <mirai>but copy-recursively "./" will copy the content from dir/* (as if it changed directory to the first subdir it sees) <mirai>extracting the archive will spray your current directory with files <mirai>it won't extract to some subdirectory <unmatched-paren>hmm... i still don't really understand what the issue is, but maybe try using (method url-fetch/zipbomb) <unmatched-paren>mirai: hmm... could you do --keep-failed and then ls the build directory? <mirai>nvm, it's the zipbomb fetcher that I needed <mirai>non-zipbomb fetcher will cd into the first directory it sees <mirai>there's no way to make it "generic"? <unmatched-paren>mirai: btw you can use ``version'' instead of ``(package-version this-package)'' <mirai>version wasn't available as a key sadly <mirai>hmmm.... so I'll have to copy-paste this snippet across the inherited packages? <apteryx>bost: I'm using vncviewer + Xvnc at the moment, so if you still have questions, don't hesitate <bost>apteryx: Thanks, I appreciate it! I need to run `guix system reconfigure ...` now. <bost>And of course I have some syntax errors in my scheme code and until I fix them my remote machine suspends and I need to wake it up... in short I'm having endless fun over here. <apteryx>don't forget the (auto-suspend? #f) for the remote GDM <apteryx>bost: the mistake you spot earlier in the SSH command is now fixed; it now reads: "ssh -L5910:localhost:5910 YOUR-HOST -- guix shell xclock -- env DISPLAY=:10 xclock" <bost>lechner: Why? Is there any particular reason for not having gdm installed, even if I pretty much won't use it? <lechner>bost / i use it on some equipement that is user-facing but on anything with server functionality i don't know. For example, Gdm stops my NFS exports from working after ten minutes, unless i am logged in <apteryx>lechner: you may have been bitten by its auto-suspend default behavior (#t) <apteryx>unless a physical user is connected, it suspends (even while there are active remote connections such as SSH) <bost>lechner: precisely *that* auto-suspend problem I've been having. <lechner>apteryx / well, the machine no longer suspends. could it be a leftover from when i used some other method to block auto-suspension (i cannot recall which)? <bost>apteryx: wouldn't it be better to change the default gdm config to auto-suspend #f ?? I guess me and lechner are not the only people tripping over this problem. <mirai>as soon you login with gdm into any user account <mirai>you'll have the same thing happening again <mirai>I meant that you need both of them <bost>mirai: does the issue mean that with (auto-suspend? #f) the machine doesn't suspend until somebody logs-in, and after that this config option starts to be ignored, i.e. the machine starts to suspend again? <mirai>you could change the user settings with gsettings or the settings panel <mirai>but this snippet just disables it for everyone <mirai>or at least sets a default, don't remember too well <bost>ACTION and now my machine suspends in the middle of `guix system reconfigure`... endless fun <bost>mirai: OK. I think I need to give your code a try at some point. Thx. <bost>lechner: there's no mouse to move on that machine. I need ping it from CLI <bost>lechner: Yeah. I just added xvnc to my system config. <bost>lechner: I mean yes you're right I won't need 70% of the functionality coming with gdm but I'm not sure how to remove it and what happens afterwards. If I brick my machine or something. <bost>lechner: maybe som (remove (lambda (srvs) (eq? srvc gdm>)) %desktop-services) would work. Who knows. <lechner>bost / some people remove the Gdm service. i build up from %base <bost>lechner: can you send a link to their configuration please? <jonsger>ACTION thinks that service removal is way to complicated compared to adding services... <bost>lechner: I guess I won't use the gdm either at some point in the future, when my guix-fu grows stronger. <gnucode>jonsger: do you want a configuration example or service removal ? <gnucode>florhizome[m] I think it's been that way for a while. guix moves pretty fast. emacs-guix needs some love to keep up. <jonsger>gnucode: no I have one in my config. But able to write something like (service remove gdm-service-type) would be nice, from an usability view... <lechner>jonsger / isn't that almost possible already? <gnucode>jonsger: that would be kind of cool. On the other hand I would prefer just one way to do it instead of multiple ways to do it. <lechner>ACTION thinks his time building up from %base-services was well-spent and offers more granular configuration options <bost>lechner: BTW I see you're using e.g. '(load "../../packages/gocryptfs.scm")' instead of (define-module (... gocryptfs)) and #:use-module (... gocryptfs). What's your reasons for that? <bost>lechner: I mean you use load instead of defining modules. <bost>lechner: I don't have any. You're the first person I know about who gets code loaded this way. <jonsger>ACTION builds half of gnome and kde while reconfiguring and don't know why... <bost>lechner: The thing is, I don't know enough about guile modules to claim that using them brings some added value over loading code your way. <lechner>bost / i would prefer using modules but getting them loaded from the same repo requires environment variables that i find to cumbersome to set when a path suffices and the "modules" are not used in any other repo <apteryx>sslv3 alert certificate unknown (SSL routines, ssl3_read_bytes) <lechner>apteryx / should you be using SSLv3? <apteryx>lechner: good question, I don't remember changing the SSL settings <mirai>Signature algorithm SHA1withRSA <mirai>they're sending you a root cert with RSA 2048 bits (e 3) / SHA1withRSA <mirai>It shouldn't matter though since there's also another path? <davidl>for the most recent master commit (2491a60404b69da20c2b4fd952e5d5e326374d1d), Im failing to build. I just use guix shell -D guix, then the regular ./bootstrap, configure stuff. Error is: <davidl>Error: failed to load 'gnu/packages/bison.scm': <davidl>ice-9/eval.scm:293:34: In procedure allocate-struct: Wrong type argument in position 1: #<syntax-transformer <origin>> <davidl>make[2]: *** [Makefile:6851: make-core-go] Error 1 <davidl>make[2]: Leaving directory '/home/user1/VirtualHome/src/guix' <davidl>make[1]: *** [Makefile:5936: all-recursive] Error 1 <jpoiret>davidl: can you do `rm gnu/packages/bison.go` first? <jpoiret>if it doesn't work, just `make clean-go` and `make` <lechner>Hi, how many TB of substitutes are there currently at MDC, please? <halting>exist an testing/unstable repo? like arch or nixOS? <nckx>lechner: I'm running a du on it. <nckx>halting: Well, no; there is no stable branch. <lechner>halting / it's all in one. you run whichever version works for you <nckx>There are ‘staging’ and ‘core-updates’ branches but these do not quite fill the role that ‘stable’ would in other distributions. <nckx>apteryx, rekado: Eehhhhh. '/var/cache/cuirass/remote/4z3y0p4bkvllcr646kmi8ny7vdvagx04.log.gz': Input/output error <nckx>And there are tonnes of those. <nckx>halting: That said, Guix is not the uppest-to-datest distro evar. Mainly because we don't have the army required to update and test everything at that pace. <nckx>ACTION leaves, leaving du to do its duing. <lechner>sneek / later ask nckx / Hi, can mirrored substitutes be authenticated even when served from potentially insecure locations? <lfam>Any Cuirass / CI experts available and interested to help me debug my go-updates Cuirass spec? <sneek>nckx, lechner says: / Hi, can mirrored substitutes be authenticated even when served from potentially insecure locations? <nckx>lechner: Only if they've not been tampered with. <nckx>This ain't no Mickey Mouse operation. <nckx>Goofy, sometimes, perhaps. <nckx>lechner: I don't understand. They *are* signed, so they can be authenticated. <nckx>If a file has been tampered with by an louche mirror operator, it will no longer authenticate and Guix will refuse to use it. <nckx>With the right HTTPd settings I think it feasible to serve only a subset. <nckx>lzip is not long for this world anyway. <nckx>What they could also do is operate [part] as a caching proxy instead of a mirror. <nckx>apteryx, rekado: I've started a scrub. 2 uncorrectable errors so far. <apteryx>nckx: how do you get that input/output error? <sneek>Welcome back rekado, you have 1 message! <rekado>lechner: not sure I want to keep using github for this. It’s only there because the parent repo was on github. <rekado>nckx apteryx I don’t know what the I/O errors tell me. AFAICT there’s no problem with the SAN; all other production systems that are connected to it work smoothly. <rekado>but also: we’re still not using multipath and we really should <rekado>lechner: is there a problem with guile-irc you’d like to report? <lechner>rekado / Hi, it would be a suggestion. Please take a look at your emails when you have a chance <lechner>Hi, what is proper remedy for this, please? error: gexp: unbound variable <lechner>for some reason i thought we were supposed to use something else for packaging, but i am probably confused <davidl>jpoiret: thx. That got me past that one error. Got stuck on another though. I tried running make clean; make now. Will see how it goes. <apteryx>what's an easy tool we have to record applications in a compact format? <mroh>record to video? peek, maybe. <lechner>Hi, it's my first time using the copy-build-system. Does this look familiar? ice-9/eval.scm:187:27: Wrong type to apply: "./lightbulb" <singpolyma>lechner: looks like you have something containing a string right after an open paren <rekado>apteryx nckx apparently one fibre channel port is down <singpolyma>lechner: make sure you have a ' or ` before the whole args section probably <rekado>apteryx nckx this means one of the following: port on our server’s HBA is down; FC switch port is down; or cable between them is bad/disconnected. <rekado>multipath would have saved us here <rekado>actually… forget I said anything <rekado>it’s the second head that’s down. This must be node 129’s connection to the SAN. <rekado>the other head (ci.guix.gnu.org) is all green <apteryx>OK. so just a single corrupted file so far? or are there others? <rekado>apteryx: the dmesg output doesn’t look good <rekado>apteryx: have you seen nckx’s email? <lechner>Hi, is it possible to install executables with copy-build-system? <lechner>or in other words, which is the simplest build system i may use to install a single (Guile) script, please? <apteryx>rekado: haven't seen it know, will check <apteryx>lechner: Since it's a Guile program, I'd use guile-build-system <nckx>I'm not available until tomorrow, sorry. The scrub is still running and error count hasn't increased from 2 at about 25% completion so far. apteryx: the errors I reported were for one directory, /var/cache/cuirass. First time I got metadata errors on btrfs! These worry me more because more information-dense that data, in a sense, and more likely to cause fall-out. We might be 'lucky' in that they're transient, but those are still dangerous. <apteryx>looking at dmesg I only see SSH brute force attempts ;-) <apteryx>so 2 errors for 25 TiB; I wouldn't call the firefighters yet <nckx>An attitude compatible with our choice of fs :-) <apteryx>but let's get this 2nd node up and shuffle our precious substitutes there while we can <nckx>Yes, let's make multipath a priority. If this task can be split, let's do so. 