IRC channel logs

2021-06-12.log

back to list of logs

<civodul>:-)
<roptat>I still don't understand
<vagrantc>yay, the build-depends for guix on riscv64 are all there ... let's see if i can patch it up enough to build guix.deb
<roptat>it's refered to only in guix-configuration (which I replace) and in guix-publish-configuration (which I don't use)
<civodul>weird
<civodul>vagrantc: yay!
<iskarian>yoctocell, how exactly are you invoking etc/committer.scm in your workflow?
<drakonis>oh, guix is in the hn frontpage
<drakonis> https://news.ycombinator.com/item?id=27477230
<drakonis>orange senpai has noticed us
<drakonis>anyways, this was weird
<drakonis>orange site has noticed us, prepare for impact
<roptat>drakonis, I don't think it's the first time
<iskarian>You know, I've been wondering... why is downloading substitutes so slow? It's not the actual speed that's slow, it's like there's a delay of 2-5 seconds between each package
<drakonis>roptat: i dont think the poster is already an user
<pkill9>iskarian: this may have been changed but, I think guix opens a new connection to the server for each download instead of reusing one
<iskarian>ah yep that'll do it
<pkill9>i don't have any gap actually, how old is the guix you are using?
<pkill9>I think it got updated to reuse a connection
<pkill9>maybe you're on an older guix
<dstolfa>i wonder when icecat will get updated...
<dstolfa>it's been a while
<dstolfa>oh, nevermind
<dstolfa>it's been 33 days since the last big update
<dstolfa>interesting
<ixmpp>> iskarian wrote:
<ixmpp>> You know, I've been wondering... why is downloading substitutes so slow? It's not the actual speed that's slow, it's like there's a delay of 2-5 seconds between each package
<ixmpp>I feel doing them in series is a mistake
<ixmpp>Parallel is life
<iskarian>"<pkill9>: maybe you're on an older guix" I just pulled & reconfigured, maybe it's a bit faster than before, but still plenty of delay
<iskarian>And watching the network activity, there's definitely pauses
<dstolfa>i've noticed this too, but it has never bothered me much
<dstolfa>however, it is something to improve
<iskarian>I also wonder if the downloading is done in series with unpacking/installing them?
<dstolfa>there's no huge delay, but it's noticably slower than apt
<pkill9>yea could be faster
<marusich>Hello, beautiful Guix
<drakonis>greetings friend
<marusich>What's up?
<drakonis>not much, what about you?
<marusich>ixmpp, the downloads happen in parallel if you specify --max-jobs=N where N is greater than 1.
<ixmpp>Problem solved
<marusich>Hopefully :)
<marusich>I'm getting ready for dinner, but also trying to sneak in debugging time on https://debbugs.gnu.org/cgi/bugreport.cgi?bug=48941
<marusich>I suppose I'll ask upstream to see if they have any advice.
<vagrantc>though --max-jobs can overwhelm a system if it actually ends up building N jobs with X cores where X == nubmer of cores in your system and N is biggish
<marusich>It can't build on powerpc64le-linux due to the test hanging, so ~2k packages can't be built.
<marusich>Interestingly, I heard Nix uses --max-jobs=N and --cores=1, where N equals the number of available threads (e.g., the output of nproc). The rationale is that some packages fail, or their tests fail, when built/tested in parallel.
<marusich>I've often wondered if that is better than what Guix does, which is --max-jobs=1 and --cores=N, but I never have taken the time to test. I want to believe that our packages can build in parallel.
<marusich>It would be neat if somebody did some benchmarks.
<vagrantc>it's probably pretty package-specific
<vagrantc>but would likely improve reproducibility
<marusich>It would also be neat if our daemon had some kind of work-stealing algorithm for scheduling the jobs, rather than the coarse-grained flags --max-jobs and --cores.
<vagrantc>e.g. "guix" is unreproducible due to parallelism ... guile-* in general
<marusich>Right, the tests for guix fail when run in parallel also, which is lame
<marusich>But very common :)
<marusich>I am no better; most of my projects do not use parallel tests. The shame!
<vagrantc>building guix on debian riscv64: [ 93%] GUILEC gnu/services/dns.go ... let's see :)
<marusich>vagrantc, it might work sometimes, but it definitely did not in the past.
<vagrantc>yeah, i disable parallelism in the guix package
<marusich>guix edit guix shows that bug https://debbugs.gnu.org/cgi/bugreport.cgi?bug=21097 prevents it
<vagrantc>originally just for tests, but now for the whole package
<marusich>Oh, I still build guix in parallel. Is there an issue with that?
<vagrantc>reproducibility
<marusich>Oh, I see.
*vagrantc filed a bug about it...
<vagrantc>oh, i didn't file a bug, just commented on: https://issues.guix.gnu.org/20272
<marusich>OK, bug reported upstream, time to eat dinner
<vagrantc>dpkg-deb: building package 'guix' in '../guix_1.3.0-2~20210611~0_riscv64.deb'.
<vagrantc>in a moment, will figure out if it actually works :)
<ixmpp>Shit :D just set PID1 to my user with setuid
<ixmpp>Guess live-hackable code has downsides too
<vagrantc>did this patch fail to apply because it was an attachment rather than in the body of the message? https://laminar.cbaines.net/jobs/patchwork-test-series/6449
<vagrantc>hrm.
<vagrantc>missed the patch defining the bootstrap binaries. hah.
*vagrantc tries, tries again
<brendyn>What chance is there that build side code can escape and infect the system?
<solene>hello
<lispmacs>did Guix have at one time a lincity-ng package...? I have a vague memory of playing it on a Guix system, but I can't seem to find it now
<lispmacs>i guess that must have been from my Debian days
<ixmpp>sneek: later tell abcdw would be nice to be able to `Include <file>` in ssh/config... im not able to use that module right now cause i need that
<sneek>Okay.
<tissevert>hey guix
<asrar>hello folks, i'm interested in guix, i 'm using nix right now but want to try out guix, it seems simpler with better docs. does anyone have resources on migrating from nix to guix
<emestee>asrar: there's a comparison between nix and guix in the manual but I think your best bet would be to watch an introduction to guix on youtube
<asrar>emestee i only see nix in the acknowledgements section in the manual, can someone link to it
<emestee>asrar: uh I appear to be wrong, must have been something I saw on youtube
<emestee>probably DistroTube channel
<yoctocell>asrar: are you on nixos or are you using nix on some other distro?
<asrar>i'm on ubuntu, i am hoping to try guix without installing it globally/root, i saw a video which highlighted using guix using the relocatable option
<yoctocell>ah, i don't have any exprience with that, is there any reason for not installing it globally?
<asrar>so my idea is i download a tar ball and try out guix in a folder within my home dir
<asrar>there's no particular reason, but it's just that this option seems to be the most appealing, since i won't be using profiles or managing software globally
<yoctocell>asrar: i found this: https://github.com/pjotrp/guix-notes/blob/master/GUIX-NO-ROOT.org
<leoprikler>asrar: technically, you'll always be using at least two profiles (one for guix, one for your packages), but I can understand your concerns about global management
<asrar>Thanks, i watched the video here https://archive.fosdem.org/2020/schedule/event/guix/, in this video the speaker builds a set of packages and packages them into a tar ball, and then they can be used anywhere .
<asrar>i thought there must be a way to package guix itself similarly, so that i can install other software using it.
<leoprikler>If you're just trying to `guix install guix', you'll hit a chicken-egg problem, but rootless Guix does work (with some caveats).
<leoprikler>They are mostly described on the rootless Guix page, so I hope there'll be no surprises.
<asrar>Ah, i see, i actually thought guix would be able to pack itself using this command https://guix.gnu.org/manual/en/html_node/Invoking-guix-pack.html, and then i could simply use it.
<asrar>The term used there is `relocatable binaries`, so i mean can guix not itself be used as a relocatable binary
<emestee>asrar: you can definitely install guix on top of an existing machine without any damage
<emestee>in fact this is my preferred way to run things these days as I am sorely disappointed with the direction debian and ubuntu are taking
<emestee>so I find myself more and more using packages from guix on ubuntu machines
<projectmoon>I had issues using guix on void. Jami specifically
<projectmoon>Never really further tried it
<projectmoon>But hopefully can install on new laptop, which has been pushed back yet another week 🙃
<emestee>in the future I would prefer guix SD but currently it has issues with UEFI boot
<asrar>emesteej Yes, it seems to be it will work fine, but i have nix installed globally already. what if i manually create a /gnu/store directory, can i then guix from there, without it modifying the profile and path
<nckx>Good morning, Guix.
<nckx>Just FYI, the name ‘GuixSD’ has been buried for years.
*nckx goes to get another stake…
<yoctocell>Will guix automatically try to fetch the source from software heritage if the upstream source is missing?
<nckx>Yes.
<nckx>But SWH doesn't archive everything.
<yoctocell>nckx: the source I specified is a hg repo on bitbucket and obviously doesn't exist, but I was able to find it archived on SWH, but when trying to build the package guix doesn't try to fetch anything from SWH.
<yoctocell>It only says "Trying content-addressed mirror at berlin.guix.gnu.org..."
<nckx>‘obviously doesn't exist’?
<yoctocell>bitbucket shutdown hg support a while ago
<nckx>Ohkay. So how did you find the hash?
<yoctocell>oh, no
<nckx>That said, I'd expect it to still print a failed attempt to query SWH…
<yoctocell>i will try to clone the SWH archive and get the hash
<nckx>OK. Yeah, so neither Guix nor SWH use the URI for much, hashes are what identify sources.
<nckx>The URI is just a hint as to where it can be found if missing.
<nckx>yoctocell: Is this something you can share here?
<yoctocell>nckx: sure, gimmie a sec
<yoctocell>nckx: https://paste.debian.net/1200946/
<yoctocell>i downloaded the tarball from SWH, ran `tar', and guix hash -r on the directory, but guix still doesn't seem to try to fetch from SWH
<kitzman>guix pull --no-substitutes fails on a freshly downloaded qemu image (1.3)... any advice? it can't find strverscmp.
<kitzman>i already installed glibc
*nckx isn't well-acquainted with Guix's SWH code…
<nckx>But it looks like (guix swh) only supports git for now, yoctocell.
<yoctocell>nckx: ah, okay
<nckx>kitzman: You can remove glibc; Guix doesn't work that way (packages don't sniff around your system using that happen to be installed or not). Can you copy the error message and paste it to paste.debian.net? If not, a screenshot would be all right. Which package is actually failing? Which QEMU image (URL)?
<nckx>s/using/using things/
<nckx>‘Which package is actually failing’ → because ‘guix pull’ still builds packages.
<kitzman>nckx: the first thing it tries to do is to build module-import-compiled, and that's where it fails. the build log is at https://0x0.st/-Le3.log (easier to use curl)
<nckx>past.debian.net has a ‘raw’ link but 0x0 is perfectly fine too.
<kitzman>wait it might actually work without "--bootstrap" flag
<nckx>‘This option is only useful to distribution developers.’ Which isn't an excuse for it to be broken, of course, but at least I get the exact same error as you when I try it.
<kitzman>okay that derivation worked. ^^ . I wonder why that might be the case, that it fails.
<nckx>I don't understand the error either.
<nckx>I don't think --bootstrap gives you any meaningful ‘purity’ or reduces trust in the VM image though. I'm assuming that's why you tried it.
<kitzman>maybe a dependency is not specified? should I open a bug?
<kitzman>yes
<nckx>Sure, go ahead.
<nckx>But do note that it's a developer option, not a ‘make Guix pure’ option.
*nckx AFK.
<dstolfa>hello guix
<munksgaard>hi!
<dstolfa>nckx: why is GuixSD still on the website? maybe it would be good to just rename GuixSD to Guix System on the website?
<leoprikler>which website?
<yoctocell>nckx: I managed to patch (guix swh) and (guix hg-download) to fetch from SWH \o/
<nckx>dstolfa: Did so years ago?
<nckx>I remember being told to leave historical blog posts alone, are you looking at them?
<nckx>yoctocell: Sweeet.
<nckx>& hi dstolfa & munksgaard!
<munksgaard>dstolfa: https://guix.gnu.org/en/download/ says Guix System, if that's what you mean?
<nckx>dstolfa: I've grepped the Web site and can't find any GuixSD usage that isn't historical.
<dstolfa>nckx: you're right, it was a blog post
<dstolfa>namely this one https://guix.gnu.org/en/blog/2017/porting-guixsd-to-armv7/
<dstolfa>:)
<nckx>I'm not thrilled about them saying GuixSD either, but I'm sure I said everything I needed to back then.
***soheil_ is now known as soheil
<asterope>Is there a way to get package's downloaded origin files store path?
***soheil_ is now known as soheil
***soheil_ is now known as soheil
<nckx>asterope: https://paste.debian.net/plain/1200953
<nckx>But then use (package-source …), I think.
<nckx>Lemme see…
<nckx>…yep, that works: https://paste.debian.net/plain/1200954
<nckx>asterope: That's assuming some kind of development/scripting. If you're ‘just’ writing a package, Guix does all the magic for you and you don't really need to do anything clever. What's your goal?
<leoprikler>On a somewhat related note, is there a way to conveniently form a tarball out of a (local-file ...) origin?
<dstolfa>sneek: later tell nckx: i've been planning a few more changes to strongswan, some of which depend on the service patch landing, others don't. should i wait and focus on getting this sorted first, or should i submit a few that don't (e.g. updating the version)
<sneek>Will do.
<leoprikler>Hmm, perhaps using the store monad…
<dstolfa>sneek: botsnack
<sneek>:)
***smartin1 is now known as smartin
***jonsger1 is now known as jonsger
***MidAutumnMoon7 is now known as MidAutumnMoon
<Noisytoot>chrome://settings in ungoogled-chromium links to the chrome web store, which contains proprietary software
<dstolfa>Noisytoot: you can't install it though
<dstolfa>at least not through the web store, you'd have to manually find the addon file and install it after tweaking a few options
<Noisytoot>Is there something like LibreJS for (ungoogled-)chromium?
<dstolfa>Noisytoot: best i found is noscript and similar
<dstolfa>it just straight up disables JS and you can selectively enable it
<solene>note to myself, when you deny everything inbound on the firewall, think about skipping lo0 to avoid madness
<ecbrown>Noclip[m]: any chance you picked up the mantle on snapper/timeshift? i found your discussion searching the logs
<ecbrown>from february
<ecbrown>i am in a similar place now and want to replace my zfs toolz :-)
<ixmpp>> In procedure variable-ref: Unbound variable: "#<variable 7fe4bd2e7470 value: #<undefined>>"
<ixmpp>Getting real sick of this
<ixmpp>A+ error reporting
<ixmpp>Much detail, many data
<ecbrown>solene: lol yes, this is the problem with the iptables in the current iptables recommendation
<dstolfa>ixmpp: this isn't even the worst of error reporting you'll find in FP
<dstolfa>guile's error reporting is not too bad by FP standards
<ecbrown>solene: can't telnet to localhost ports :-/
<dstolfa>i think ocaml takes the cake for me
<ixmpp>dstolfa: i guess im spoilt by haskell
<dstolfa>"error: line 106", where line 106 is `let f ...` and it's a function that continues on for about 500 lines
<ixmpp>Either way, that particular error pisses me off cause i have NO information on where or what the issue is
<dstolfa>ixmpp: i see you haven't used monad transformers and lenses extensively
<dstolfa>in combination
<dstolfa>error messages created by that are really fun
<dstolfa>but in haskell's defense, its error reporting is decent
<dstolfa>ixmpp: the sooner you accept that all programming languages suck in some way, the closer you will be to happiness in programming
<dstolfa>(the good thing with FP is that they usually suck at a fundamental level and the only thing you can be angry at is maths for being undecidable)
*ecbrown is looking for simple way to "prime" a local substitute server
<ecbrown>i have thought about cuirass just running and guix build --keep-going `guix .... ` manual example
<bricewge>ecbrown: Why would you need btrfs snapshot in Guix? I don't get the use case since Guix includes rollback
<ecbrown>data
<bricewge>Makes sense
<dstolfa>bricewge: guix solves the system provisioning problem, and with guix home it solves dotfiles and all that, but it doesn't solve the "my cat pictures are important" problem ;)
<bricewge>dstolfa: Yes but for that you have backups
<dstolfa>yeah, and they can be done with btrfs :P
<ecbrown>dstolfa: in my great re-orginization, i am thinking about shoving all my documents into git/git-lfs
<ecbrown>just push everything out to a monster git repo on my home server
<dstolfa>ecbrown: you might find it more efficient to do with svn for binary files
<dstolfa>as bad as svn is, it handles binary files much better
<ecbrown>then i can move between computers and contexts and just get what i need. i have zero need for stuff from e.g. 20 years ago
<ecbrown>dstolfa: yes, for sure. i would do this in perforce at work
<ecbrown>but tooling and mindshare and git-lfs make for a nice libre solution to big binaries
<dstolfa>fair
<ecbrown>it's my biggest problem with git: why must it pertain to "source code"
<ecbrown>delta algorithm, etc.
<bricewge>dstolfa: Of course
<ecbrown>my storage unit has 200T and if i wanna check in a 10T file then do it dammit
<roptat>I managed to build guix from my x86 system, with qemu, and pushed the nar to the armhf system, looks like I'll be able to update it
<ixmpp>Except by deleting code and seeing if it goes away
<ixmpp>> dstolfa wrote:
<ixmpp>> ixmpp: i see you haven't used monad transformers and lenses extensively
<ixmpp>> in combination
<ixmpp>> error messages created by that are really fun
<ixmpp>No i have, but even then i still get *some* info
<ixmpp>My point is this is literally impossible to debug now
<dstolfa>ixmpp: you can always use the repl and debug it that way
<dstolfa>impossible is a very strong word
<ixmpp>Nope
<ixmpp>Its a sheperd service gexp
<ixmpp>Unless i can? But i dont know how
<dstolfa> https://guix.gnu.org/manual/en/html_node/Invoking-guix-repl.html
<ixmpp>I know how to use guix repl ._.
<dstolfa>so what's the problem then? i'm pretty sure you can use gexps in there and try to build up the thing you need?
<ixmpp>How do i debug a shepherd script with it
<ixmpp>Those dont get evaluated, theyre gexps
<dstolfa>ixmpp: you may have to explain your problem a bit better, where is the error happening?
<ixmpp>Ok but, how do i do that?
<ixmpp>I have a shepherd service, im trying to do wacky stuff in it's "start", when i run the service i get that error
<dstolfa>so, once you do a `herd start foo` you get that backtrace?
<ixmpp>yes
<ixmpp>So how do i simulate that in repl?
<dstolfa>it would depend on your specific code, what is said "wacky stuff"?
<dstolfa>are you trying to carry around variables from other gexps, maybe you are forgetting to ungexp something?
<ixmpp>Nope
<ixmpp>im just trying to import new stuff
<soheilkhanalipur>Hello!
<ixmpp>The vast majority works
<ixmpp>Then i change one tiny thing and get this blackbox error
<ixmpp>I have no idea whats wrong and no real way to find out
<ixmpp>Unless you can help me get this in a repl
<soheilkhanalipur>How do I bridge ADSL connection to LAN in Guix system?
<dstolfa>ixmpp: this is still pretty vague to me, i think the only way i can try to help you is if you share the diff that causes the problem
<bricewge>ixmpp: Debugging shepherd is clunky
<ixmpp>Exactly
<bricewge>Do you have the derivation for your service?
<ixmpp>The service.scm?
<ixmpp>I do, yeah
<bricewge>No shepherd-FOO.scm.drv
<bricewge>Or something like that
<ixmpp>Yeah
<ixmpp>Thats what i meant, heh
<bricewge>Ok, ther should not be any gexp in the result of that derivation
<soheilkhanalipur><soheilkhanalipur "How do I bridge ADSL connection "> ☝️
<ixmpp>bricewge: Yeah no that has no gexps in it, its clean
<soheilkhanalipur>Help!
<bricewge>You can load that into shepherd with "shepherd load root shepherd-FOO.scm"
<ixmpp>Same error
<ixmpp>Well, when i run it (cause its not autostart)
<roptat>ixmpp, can you run that .scm code in the repl'
<ixmpp>roptat: it's a shepherd service, not a script
<ixmpp>So it builds but *running* it causes the error
<ixmpp>Ive shared the code in #nonguix
<ixmpp>Gotta brb though
<bdax>soheilkhanalipur: I don't have an answer for you, but you might get more answers if you gave more details, perhaps say how you would do what you want to do in another distro
<abcdw>ixmpp says: would be nice to be able to `Include <file>` in ssh/config... im not able to use that module right now cause i need that
<abcdw>What prevents you from doing so?
<roptat>ixmpp, I mean, it's some code that gets executed by the shepherd at some point, can't you execute it from a repl?
<bricewge>سُـھِـیـل: You could use NetworkManager with network-manager-service-type, and configure the rest non declaratively
<user__>Hello, did anyone here have trouble launching Evolution? Apparently: "The schema default value for key 'primary-tasks' in schema 'org.gnome.evolution.calendar' was rejected by the binding mapping function."
<solene>in scheme how do you check that a list is not empty? I tried (list?) but I'm surprised to see (list? '()) returns #t
<ecbrown>i thinkn '() is a list, so #t. it happens to be empty
<ecbrown>the predicate list? would test for whether something is a list
<solene>I come from common lisp, I was confused because '() is nil in common lisp ^^'
<bdax>apparently it's `null?`
<bdax> https://www.gnu.org/software/guile/manual/html_node/List-Predicates.html
<solene>thanks
<jonsger>can I do the `herd start cow-store /mnt` when I only have a debian rescue system while install?
<solene>I'm hacking the networking services file and I see many occurences of #$ or even #$@ but I don't understand what it is exactly, I assume it's related to macros. Could someone point me where to look for information about that?
<bdax>solene, you may already know it, but seeing as you referred to it as 'scheme' I'll mention that guix uses 'guile' specifically, so when searching for answers to problems, search for things related to guile not scheme, as the other schemes can be quite different
<roptat>solene, those are related to gexps
<solene>bdax: ok, I need to think guile and not scheme
<roptat>#~ is gexp, #$ is ungexp and #$@ is ungexp-splicing (similar to how quote, unquote and unquote-splicing work)
<roptat>it's code staging
<roptat>user__, I don't use evolution, sorry I can't help you. If nobody can answer here, you might have more luck asking on help-guix@gnu.org (the first message is moderated by a human, so it can take up to a day to get through, subsequent messages go through immediately)
<solene>i really don't understand all it's about when reading the documentation ^^'
<user__>roptat: It's OK, I'll manage
<solene>english is hard
<bricewge>solene: That's the hardest part I game across in Guix
<bricewge>Recently abcdw did 2 live stream about it, maybe it can help you
<bricewge> https://www.youtube.com/watch?v=6JPHQJTlUIM&list=PLZmotIJq3yOI0cPPQ07urjm6VMnb8GDSQ&index=3
<roptat>yeah, it's not easy, there's even a paper on it: https://arxiv.org/pdf/1709.00833.pdf
<solene>thanks!
<solene>yeahhh! after 2 hours I found my issue :D :D
<solene>I had a service variable that is expected to be a list and I picked #f as a default
<solene>i couldn't compare without an error becaues either iwas testing a boolean or a list
<solene>and $#@ on a boolean explodes
<bdax>I recently saw https://guix.gnu.org/blog/2019/guix-reduces-bootstrap-seed-by-50/ and I was very impressed to see that guix is thinking about this problem, and is making some remarkable headway in that direction. It then occurred to me that it is strange that I can't think of any other distros that have publicly considered this problem, or sought to
<bdax>do anything about it. Perhaps that's because it's a hard problem, but does anyone know of any other practical attempts to solve the trust problem with regard to OS distribution? Perhaps in the worlds of nix, gentoo, crux, netbsd, openbsd, etc?
<solene>bdax: at OpenBSD we encourage distributing the signify public keys on various media for cross check. The key for release n+1 are in release n. But at some point you have to trust something (or multiples sources here)
<solene>you can easily find the keys on mastodon, reddit, twitter, cvs and github repos
<bdax>ah okay, that seems like a good idea, but would rely on key exchange parties (https is insecure, ask me for links if you want)
<bdax>and is that signing the binary release, or the source release?
<bdax>as binary release signing still requires trusting whoever is producing it
<solene>only the binary release is signed. If you got the signify key from various sources to check it's the one you can tust, it's the one from the OpenBSD team so if an iso downloaded is fine with the key you can expect it to be trusted
<solene>sources are not signed
<solene>OpenBSD 6.9 was release last month, you can already get the keys for 7.0, 2 medias to cross check https://cvsweb.openbsd.org/src/etc/signify/ and https://github.com/openbsd/src/tree/master/etc/signify
<dstolfa>in practice what solene says tends to be sufficient, but one would always want to push things further :)
<bdax>still, it relies on https for security then; https://blog.cryptographyengineering.com/2012/02/09/trustwave-announces-name-change/
<solene>bdax: you could ask OpenBSD users on #openbsd what the key is for 7.0
<solene>and see if it matches what you found
<bdax>irc is no more secure than https though
<solene>bdax: no, but the more sources you get, the more chances you have to trust the key
<solene>dstolfa: indeed, I'm curious how this could be improved though
<ecbrown>i guess you have to go see theo
<bdax>but it's all coming to you through the same channel, i.e. your isp and al lt's connected to
<bdax>all*
*dstolfa wonders why offlineimap is so slow at initial fetch
<dstolfa>it's just email damn it
<solene>bdax: if someone is able to hijack every requests from your computer to multiples sources and fake people on communities, you are in reallyyyy big troubles :D
<ecbrown>dstolfa: email providers throttle heavily
<ecbrown>
<bdax>solene, it's more possible than you think: https://blog.cryptographyengineering.com/2012/02/09/trustwave-announces-name-change/
<dstolfa>this is why i want to have a custom email setup...
<dstolfa>i hate how slow this is lol
<bdax>easily automatable by someone with enough resources
<ecbrown>well, i use fastmail, they are reasonable and support cyrus
<dstolfa>ecbrown: well this is work email
<ecbrown>yeah, outlook is dreadful
<dstolfa>how did you know it's outlook (it is)
<dstolfa>is it that common? :(
<bdax>and you still require trust in the openbsd team, and no offense but perfect people are rare https://blog.cryptographyengineering.com/2012/02/09/trustwave-announces-name-change/
<bdax>that last link I intended to be: https://www.theregister.com/2010/12/15/openbsd_backdoor_claim/
<ecbrown>because i have tried every solution of getting office imap into gnus that one could imagine
<ecbrown>there is nothing else ;-)
<ecbrown>davmail, local imap
<solene>bdax: old news
<dstolfa>hah
<dstolfa>ecbrown: we used to use hermes, that worked really well
<dstolfa>but then someone higher up decided: hey, let's move to exchange!
<bdax>solene, still relevant
<dstolfa>everyone got pretty annoyed at this, and a solution might yet come up with some custom-hosted thing
<ecbrown>we had lotus notes until a couple years ago
<dstolfa>but yeah...
<bdax>though I will say the irc channel is better than https
<solene>bdax: this is not better from linux in that regards
<solene>you need to trust the people who write and manage the code anyway
<bdax>I'm not saying it is better in linux, I started this applauding guix for their efforts in this direction
<dstolfa>ecbrown: i don't really understand why so many things are moving to exchange, it's awful
<bdax>though I continue to be surprised that only guix has really offered something substantial here
<solene>weren't linux reverting hundreds of commits from an university recently? ^^ the code is produced by human and is too complex to be reviewed correctly IMO
<solene>bdax: any effort in that regards is worth
<bdax>all other communities shrug it off, like 'mossad gonna mossad'
<bdax>with an honourable mention to openbsd
<ecbrown>dstolfa: calendaring, q.e.d.
<solene>back to guix, I reworked my firewall service to make it cleaner :D
<solene> https://pastebin.com/JqdGtZnK
<solene>I got made for the for-each line 32
<solene>mad
<ecbrown>dstolfa: plus mail is a service no one wants to run anymore. too many problems
<Noisytoot>ecbrown, someone has to run a mail server
<dstolfa>ecbrown: paying a sysadmin to run a mail server or having everyone complain to you daily for not being able to get their work done because exchange is hot garbage
<dstolfa>i wonder which one is more efficient :D
<solene>and then everyone will complain they can't reach people using gmail or microsoft because they are in spam
<ecbrown>every sysadmin i know in charge of email is miserable
<ecbrown>both of them
<irfus>I've run a mail server for personal use for around 7 years now. Only once had an issue with undelivered email, and that was to some german university that didn't allow encrypted connections
<irfus>granted I'm the only user, but this is still an in-production setup
<solene>irfus: you are lucky, I'm running mine since 2010 and I still have issues with gmail and Microsoft cloud users, I get into spam. My IP is totally clean, never blacklisted, I'm compliant to everything
<solene>:/
<ixmpp>abcdw: theres no option to?
<ixmpp>Its exclusively host and match blocks
<irfus>tbh, running znc on my server has required more actual maintenance than the email parts. I hope the luck lasts, :D
<abcdw>ixmpp: Ok, I see. Can we put Include in default-options, it will work for all hosts because of "Host *"? Or it must be top-level?
<ixmpp>Top-level abcdw
<ixmpp>Either that or a text trapdoor would be nice
<abcdw>ixmpp: ok, can you create a topic for that in rde-devel please?
<ixmpp>Sure
<ixmpp>Just a sec
<ixmpp>Oh nvm, html email banned :|
<ixmpp>This is real ergonomic i love email workflows
<ixmpp>abcdw: https://lists.sr.ht/~abcdw/rde-devel/%3CXNh-7CLvmQPB70oQ7r5c1spO7t8NotWPYKyeKKBqnTF44zmB31BDYAUhUNrUNo9Miftg2MaQxOrMcY33ZuC4kg%3D%3D%40fron.io%3E
<soheilkhanalipur> https://unix.stackexchange.com/questions/654000/bridging-adsl-to-lan-in-guix-system
<abcdw>ixmpp: Thank you!)
<ecbrown>soheilkhanalipur: is this your question? are you trying to set up nat-dhcp?
<ecbrown>my advice is to look at nftables
<ecbrown>in other words, it is the same for guix as other gnu/linux systems
<ecbrown>probably need dhcpd-service-type as wel
<iskarian>Does anyone happen to have experience using the vmware video driver for more than one display?
<spot272253>the kiwi-irc on guix.gnu.org to get here requires SSL to be set in order to connect, otherwise it errors
<spot272253>I am trying to get guix running on a hosted VM. I have the live image up, and I can shell in. When I do `guix system init my-config.scm /mnt` where /mnt is the file system of the provisioned SSD, I fail with 'no space left on device'. I'm not too familiar with the overlay filesystems in `mount -v`.. How do I figure out what device root is on?
<spot272253>That's the filesystem filling up.
<spot272253>Or could I change the bind mount of /gnu/store mentioned in the manual to use swap?
<spot272253>ie a tmpfs
<gnarlf>Hi all, simple end user here having a hard time with Guix System but really interested in it. I want to adopt it as my main and only OS in my librebooted thinkpad.
<gnarlf>So different to all the distros I've used in almost 20 years using Gnu/linux distros. Is like learning again from the beginning
<gnarlf>But also is like the exciting feeling of that beginning
<spot272253>agreed! More fun than I expected packaging things myself.
<gnarlf>Yeah!
<gnarlf>A lot of doubts in a week using the system exclusively but I'm reading the manual yet
<natrys>hi, as guix/SELinux noob, it's not clear to me from manual what `restorecon` command should I be running?
<natrys>right now the daemon starts manually, but not via systemd
<dstolfa>has there been any discussion on home folder encryption in guix? with guix-home in the upstreaming process, it might make sense to bring the per-user home encryption that depends on their password into guix in some way maybe?
<dstolfa>instead of full disk encryption, for laptops, one could opt for just encrypting home
<dstolfa>i'm not sure how it would work in the ecosystem, but just a bit of a wild thought
<spot272253>this looks promising re the 'no space left on device' issue: https://guix.gnu.org/manual/en/html_node/Proceeding-with-the-Installation.html
<ecbrown>spot272253: have you tried making everything on one partition?
<spot272253>I don't think that'll help, the provisioned SSD is 45 GB, it's just the image is not using it. Trying out the cow-store, though I had to pull it out of the os-installation config
<spot272253> https://paste.debian.net/1200984
<ecbrown>have you successfully installed using a) the GUI installer and/or b) the command line option, literally following the directions?
<ecbrown>reason i ask is that i have not had to access cow-store except for `herd start cow-store /mnt' during install
<yoctocell>ixmpp: re ssh toplevel config: i just sent a patch for it https://lists.sr.ht/~abcdw/rde-devel/%3CXNh-7CLvmQPB70oQ7r5c1spO7t8NotWPYKyeKKBqnTF44zmB31BDYAUhUNrUNo9Miftg2MaQxOrMcY33ZuC4kg%3D%3D%40fron.io%3E#%3Cd08bf0e6f395eb0a312b868a6954ab0116e879b1.1623524468.git.public@yoctocell.xyz%3E
<yoctocell>wow, that's a long link
*ecbrown wonders if anyone else has seen a runaway kworker process while using emacs-native-comp
<dstolfa>ecbrown: how do you install emacs-native-comp on guix?
<dstolfa>i'm just using emacs as-is
<ecbrown>flatwhatson/'s channel
<dstolfa>ah
<dstolfa>i guess it'll be upstreamed eventually? :)
<ecbrown>28.1 for sure "ready or not, here i come"
<dstolfa>heh :)
<ecbrown>well, maybe i will test out if i observe this process with slow-macs
<ecbrown>brb
<ixmpp>> yoctocell wrote:
<ixmpp>> wow, that's a long link
<ixmpp>blesséd day, thank you as ever
<dstolfa>ixmpp: did you manage to solve your problem from earlier?
<dstolfa>the shepherd one
<ixmpp>dstolfa: Havent had a second to try things yet
<spot272253>ecbrown: yeah, installed the machine I'm on with the gui. Just going with `guix system init` on a copy pasted config with a hosted VM, and then updating with guix deploy.
<spot272253>if this try doesn't work, I'll probably go from a debian image and use the bootstrap script mentioned in the manual
<ecbrown>+1
<spot272253>if that's what most people use that's probably what I should
<spot272253>'ve done to begin with.
*ecbrown says brief prayer for repair of debian/fsf documentation schism
<efraim>sneek: later tell vagrantc I've pushed a wip-riscv branch to savannah
<sneek>Got it.
<efraim>sneek: botsnack
<sneek>:)
<Noclip[m]>ecbrown: "any chance you picked up the mantle on snapper/timeshift? i found your discussion searching the logs; from february"
<Noclip[m]>Crazy how fast time passes by ... But I'm glad you ask.
<ecbrown>o/
<Noclip[m]>ecbrown: The answer to you question is no, I didn't messed with packaging snapper or timeshift for guix. But I might still have something for you ...
<Noclip[m]>In the last days I installed Linux Mint on a new family computer and I set up btrfs as main filesystem. As you might know Linux Mint comes with timeshift preinstalled so that was of course what I tried first managing snapshots.
<Noclip[m]>ecbrown: Unfortunately timeshift seems to only support snapshoting for the subvolumes @/ and @home/.
<Noclip[m]>This made it mostly useless for me since I wanted to use and auto-snapshot much more subvolumes than just those two.
<Noclip[m]>Next I started reading a bit about snapper on the arch wiki but that tool seemed a bit "over-complicated" or "bloated" to me. I really only need a tool which creates regular snapshots and deletes old ones in the background.
<Noclip[m]>Also I don't like that snapper puts a subvolumes snapshots inside that subvolume itself (at least by default; can this be changed without some hacky bind mount setups?). This makes by-hand rollbacks (with default btrfs tools) more complicated and I don't see a good reason for it.
<Noclip[m]>ecbrown: So in the end I decided to just create my own little snapshot tool ... a short and uncomplicated shell script.
<user__>roptat: I had to install evolution-data-server seperately for evolution to work
*Noclip[m] posted a file: bsnapscript (0KiB) < https://libera.ems.host/_matrix/media/r0/download/matrix.feneas.org/08ce6aeac54fcfc2b6fb1c5f09cf8dd018a5a267 >
<Noclip[m]>e4b2620569a244b0c0841e7dac9db1b237da77448184a6c44e84704f5eb5c4ed bsnapscript
<user__>I'm trying to set up a OpenVPN client, if there's anyone kind enough to share how they set up VPN on their Guix config, I would greatly appreciate it
<user__>I don't have Wireguard
<Noclip[m]>ecbrown: It does exactly what it tells to do, not more and not less:
<Noclip[m]># Create a new read-only btrfs snapshot of $1 inside $2 and keep the $3 most recent versions.
<Noclip[m]> * ecbrown: It does exactly what it tells to do, not more and not less:
<Noclip[m]>"Create a new read-only btrfs snapshot of $1 inside $2 and keep the $3 most recent versions."
<Noclip[m]>(The hashtag made it super big on the matrix side so I edited my last message ...)
<dstolfa>Noclip[m]: on irc, we can't really see the edit, you just kind of send a message every edit :P
<Noclip[m]>dstolfa: I know, that's why I told you the reason for editing / reposting it.
<spot272253>using the cow-store from the installation config _almost_ worked, which really meant it didn't. I got all the way to the end but building the system failed with an undecipherable symlink error.
<spot272253>I'm just going to use the debian image route now.
<drakonis>spot272253: have you tried running it as root?
<drakonis>if you're trying to init guix from another system, you might need to elevate permissions
<Noclip[m]>ecbrown: The script isn't aware of time and doesn't know what a hour, day, week, month or year is. You need to use cron/systemd/shepherd (or whatever) to call the script regularly (with the correct parameters for $1, $2, and $3).
<ecbrown>Noclip[m]: thanks, i will take a look this sounds like zsnapzend or whatever it is called
<ecbrown>rolling backup
<drakonis>mcron is what you need
<drakonis>just port it to shepherd and mcron
<spot272253>drakonis: yep, running as root. Going to stop going against the grain and attempt to get the rest of my weekend back with just following the instructions, lol: https://guix.gnu.org/manual/en/html_node/Binary-Installation.html#Binary-Installation
<drakonis>also is this a running system?
<drakonis>rather
<spot272253>no, from an image
<drakonis>are you trying to bootstrap it into an filesystem with other files?
<drakonis>this road leads to suffering
<spot272253>nope, clean filesystem. Although I do have a penchant for suffering
<drakonis>i see
<Noclip[m]>drakonis: What do you mean with "port it to shepherd and mcron"? The script doesn't care about the programm that calls it.
<drakonis>convert it into a service
<drakonis>you can use guile inside the service definitions, so you can forego some of the coreutils
<Noclip[m]>Of course you can do that. It's just a small script, feel free to modify and integrate it however you wish. The script shouldn't care about any of this, it should just do what it is supposed to do anyways.
<drakonis>aye
<Noclip[m]>drakonis: You could also just rewrite the script in guile and then put it directly into your operating system definition.
<drakonis>yes
<drakonis>that's the whole point
<Noclip[m]>drakonis: Oh okay xD
<vagrantc>hrm. building a guix.deb on debian on riscv64: Build needed 05:16:49, 562808k disk space ... end ended with a test suite failure and no package
<sneek>vagrantc, you have 1 message!
<sneek>vagrantc, efraim says: I've pushed a wip-riscv branch to savannah
<vagrantc>efraim: cool!
<vagrantc>efraim: i applied the two patches you included in your riscv64 email and am trying to build it ...
<vagrantc>very... slowly.
<vagrantc>not terribly impressed with the speed of the hifive unmatched ...
<Noclip[m]>vagrantc: You're porting guix over to RISC-V? Awesome!
<vagrantc>Noclip[m]: i'm just building it
<vagrantc>i think efraim recently and dongcarl a while ago did some harder bits
<bricewge>vagrantc: Do you still suffer from https://issues.guix.gnu.org/31337?
<Noclip[m]>How easy is it to port over guix to other cpu architectures compared to other linux distros?
<drakonis>you need to get guile to work on the architecture first
<vagrantc>efraim: my attempt at packaging the u-boot stuff for debian worked, but ... fails to actually boot :/
<vagrantc>guile, gcc, glibc, binutils, etc... if all that's done it seems like the hardest part is getting bootstrap binaries and shaking out the occasional assumptions
<vagrantc><--- never really ported anything to anything
<Noclip[m]>As far as I know debian and gentoo have been ported to lots of architectures but both distros are super popular compared to guix and also much older.
<drakonis>guile without jit should work fine i think
<drakonis>but then, lower perf
<vagrantc>yeah, even on this supposedly high-end riscv64 board ... building guix taking over 5 hours ... guix pull will be painful
<vagrantc>might get slightly faster with NVMe rather than microSD for rootfs, but my /tmp is tmpfs so the actual build should be about as fast as it can get once everything is in there
<Noclip[m]>According to this wikipedia table (https://en.wikipedia.org/wiki/Comparison_of_Linux_distributions#Instruction_set_architecture_support) gentoo is just dominating everything in terms of architecture support.
<ixmpp>sneek: later tell abcdw im happy with yoctocell's patch, merge as you please. Would reply but sr.ht's anti-html thing is a shitfest im not gonna play part of
<sneek>Got it.
<vagrantc>it seems to be missing some updates for guix ... e.g. i've mostly been running guix on aarch64 machines (which they don't even distinguish from arm)
<vagrantc>and guix ostensibly supports armhf too ... though maybe not well supported due to lack of substitute servers