IRC channel logs

2019-11-01.log

back to list of logs

<bdju>can the topic have control codes? you could put the part about the substitute server being down in red
<bdju>guix.gnu.org (including substitute server) is currently down
<leoprikler>joke's on you, that part is already red in Emacs :)
<bdju>haha nice
*nckx looks up IRC control codes.
<wdkrnls>Hi Guix
***ChanServ sets mode: +o nckx
<sneek>wdkrnls, you have 1 message.
<sneek>wdkrnls, nckx says: You're asking Guile to load a module (<channel>/packages/cran.scm) that doesn't exist. Is there such a file, and does it start with (define-module (packages cran) …)?
***nckx changes topic to 'GNU Guix | guix.gnu.org (including substitute server) is currently down | 1.0.1 is out! get it at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs and patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel is logged: http://logs.guix.gnu.org/'
<wdkrnls>How can I recover from dependency hell?
<nckx>leoprikler, bdju: That look like not-crap to anyone?
<wdkrnls>For example: guix upgrade: error: profile contains conflicting entries for r-mass
<wdkrnls>
<nckx>Either HexChat hides teh codez or they got stripped completely.
<wdkrnls>This happens for me with many R packages.
<wdkrnls>This even happens when I try to upgrade all R packages with 'guix upgrade '^r-[a-z0-9]+$.
<nckx>wdkrnls: While still mentioning only R packages as conflicting?
<wdkrnls>nckx: Thanks for your help with the my personal channel problem. I left off the top level directory in the repository.
<wdkrnls>nckx: yes, R packages are the only conflicts.
<wdkrnls>(and R itself)
<nckx>You're very welcome.
<nckx>wdkrnls: Hm. That's not what I expected, but I fear I am a manifest man, and never use ‘guix upgrade’.
<nckx>(Manifests and upgrading all packages at once don't have this problem.)
***nckx changes topic to 'GNU Guix | ⚠️ guix.gnu.org (including substitute server) is currently down ⚠️ | 1.0.1 is out! get it at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs and patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel is logged: http://logs.guix.gnu.org/'
<nckx>Horrible. Perfect.
***ChanServ sets mode: -o nckx
<wdkrnls>Ah, that must explain my: "guix substitute: error: connect: Connection timed out" problem :)
<nckx>Indeed. The whole Web site is down, so Guix is basically uninstallable ☹
<wdkrnls>nckx: when you say "Manifests and upgrading all packages at once" are you using "and" as a logical "or"?
<nckx>Yes. Yes I am.
<nckx>‘guix package -u .’ gives you the same advantages without a manifest.
<bdju>nckx: sorry I didn't respond. the colors didn't come through in irssi either. I see the emoji, although they're aligned rather sloppily. partial overlap with their ajdacent chars. prob on my end.
<nckx>bdju: I think the colour codes never made it out of my client. I agree the emojos are ugly as ⚠, but let's consider that a twisted feature when it comes to grabbing attention. 🙂
<nckx>Oh, it works now.
<janas>On irssi as well, no emoji overlap with the other characters
<nckx>Welp, they had their 5 minutes to shine, guix.gnu.org is back.
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | 1.0.1 is out! get it at https://guix.gnu.org | videos: https://guix.gnu.org/blog/tags/talks/ | bugs and patches: https://issues.guix.gnu.org | paste: https://paste.debian.net | Guix in high-performance computing: https://hpc.guix.info | This channel is logged: http://logs.guix.gnu.org/'
***ChanServ sets mode: -o nckx
<sturm>I'm getting an error running "guix pull" as root on my Guix System: "uix pull: error: Git error: object not found - no match for id (229818064a154a38dff1c0a9a6bc3ba66a2d4925)"
<sturm>Works ok as regular user
<str1ngs>sturm: does cleaning roots ~/.cache/guix/pull/ help?
<sturm>No error if run locally built guix from current master.
<sturm>str1ngs: I'll just try
<str1ngs>also is root maybe using a channel?
<sturm>hmm, I don't think I've configured a channel for root - has worked similarly up until now
<sturm>removing ~/.cache/guix/pull for root doesn't seem to help
<sturm>but "sudo -E ./pre-inst-env guix pull" works
<str1ngs>and $ stat ~root/check with stat
<str1ngs>err sorry bad paste. stat ~root/.config/guix/channels.scm
<sturm>str1ngs: nope, doesn't exist, thanks
<str1ngs>I didn't think it did but that error might indicate bad channel config
<str1ngs>sturm: try guix describe as user. then guix pull --commit=<hash> and reuse the user hash
<str1ngs>pull as root of coursae
<str1ngs>course*
<sturm>str1ngs: that's what I normally do, and was seeing the error there too. I'm getting a slightly different error now though "guix pull: error: Git error: the reference 'refs/heads/master' cannot be peeled - Cannot retrieve reference target"
<sturm>(possibly because I started the "sudo -E ./pre-inst-env guix pull" and cancelled it?
<str1ngs>might be a remote issue. what hash are you pulling?
<str1ngs>though refs/heads/master can exist on disk. so could be a permission issue
<sturm>8b438f85cf85c1a22273eb6164c7f67f41f29bf4
<sturm>wouldn't think it's a remote issue if it works with guix pull from source though, and also as regular user
<str1ngs>8b438f85cf85c1a22273eb6164c7f67f41f29bf4 pulls fine for me. I didnt complete the build though
<sturm>I'll let the guix pull from source run through and see if it still occurs with installed guix
<str1ngs>have you tried pulling without sudo. eg. sudo su - . ~# guix pull ?
<sturm>yes, both as sudo and via "sudo su" followed by guix pull
<str1ngs>the thing is sudo -E ./pre-inst-env guix pull . really does not have much effect
<str1ngs>is it's calling guix from $PATH and it's pulling the source anyways . so it's kinda redundant to do that
<sturm>hmm interesting
<str1ngs>./pre-inst-env is good if you are working on package modules
<str1ngs>get ./pre-inst-env guix build bash . say you added something to bash
<str1ngs>s/get/like
<sturm>unrelated question - is the "." required there? I haven't used it before
<sturm>as in "bash ."
<str1ngs>./pre-inst-env you mean?
<sturm>"bash ."
<sturm>or was that just punctuation?
<str1ngs>no sorry that was end of a sentence for me
<sturm>:)
<str1ngs>why I put the space. it's confusing sorry haha
<str1ngs>sturm: this is an odd issue. I'm wondering if it's filesystem related
<str1ngs>not hard dirks but permissions maybe
<str1ngs>disk*
<sturm>str1ngs: I *am* mildly suspicious of whether this SSD is on the way out. It has done a o*t of swapping while building guix packages on machines without enough memory in the past.
<sturm>*a lot of swapping
<sturm>no issues so far though
<sturm>still the same issue with standard root "guix pull" after successful run of root "./pre-install-env guix pull"
<sturm>reconfiguring now to see if that help
<sturm>nope that hasn't helped
<davexunit>dongcarl: what's the significance of the --rpath-link stuff in that code you showed me?
<dongcarl>davexunit: one sec!
<dongcarl>davexunit: This might not be entirely needed, but in our gcc package, we modify the default specfile which introduces some rpaths: http://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/gcc.scm#n251
<dongcarl>davexunit: I don't understand it all fully so it might not be needed
<davexunit>dongcarl: I mean, I think it is significant because I do want to avoid runpaths.
*nckx logs into build server, sees fftw test suite has been running for 648h (a month). Sound familiar to anyone?
<davexunit>I was able to build a cross toolchain using your code, which is great. but since it's a cross toolchain the configure script fails to find gcc. I guess I need to specify --target
<dongcarl>Right. One more tip for you...
<dongcarl>davexunit: I set the CROSS_*_PATH variables manually cuz I was running into trouble: https://github.com/bitcoin/bitcoin/blob/master/contrib/guix/libexec/build.sh#L30-L55
<dongcarl>permalink: https://github.com/bitcoin/bitcoin/blob/222b7d0ca795c7306cd2103473c1c54c60e701f9/contrib/guix/libexec/build.sh#L30-L55
<dongcarl>davexunit: might not be needed, but just in case you run into trouble like that
<davexunit>dongcarl: thanks. good to know.
<davexunit>I might try to put together a native toolchain rather than a cross compiler
<dongcarl>davexunit: sounds good!
<davexunit>but then again I'm kind of lost and confused so who knows where I'll end up ;)
<davexunit>but what you've shared is really helpful. gives me a good direction to head in.
<dongcarl>Glad!
<str1ngs>sturm: anything in dmesg?
<sturm>str1ngs: I have to admit, I don't really understand dmesg - is it a log where the newer things are at the bottom, or do I need to look at the whose output?
<nckx>sturm: It's just a log. dmesg -T prettifies the timestamps.
<sturm>thanks nckx - nothing new in dmesg that I can see str1ngs
<peanutbutterandc>Hey there, I am trying to package a cmake project and I am getting a warning that says it couldn't find python interpreter or perl interpreter in PATH and yet, I have tried putting them in (inputs) (propagated-inputs) and (native-inputs). Any ideas please?
<peanutbutterandc>This is how the package definition looks as of now: https://termbin.com/q0en
<str1ngs>strum: does ./pre-install-env guix pull as root work?
<sturm>str1ngs: yes it does
<str1ngs>sturm: maybe purge all of ~root/.cache/guix then.
<str1ngs>note *cache* not config
<sturm>str1ngs: that seems to have fixed it!
***pbac is now known as peanutbutterandc
<str1ngs>sturm: maybe I should have suggested that first :P
<peanutbutterandc>I have this strange thing going on: `guix describe` from root complains about the command not being obtained with `guix pull` when in fact it is. Other user's guix describe works just fine. Any explanations please?
<bdju>I'm getting a "The page isn't redirecting properly" error in icecat trying to log in to cisco netacad. I've tweaked addons, deleted cookies, restarted my browser. nothing works
<bdju>I managed to log in on my phone (also using icecat, but I'm sure they're somewhat different)
<raghavgururajan>Hello Guix!
<raghavgururajan>What is RTM (Release to Manufacturing)? The GNU Guix Wikipedia page has this line "Working state: RTM (Release to manufacturing)".
<fps>i installed guix on another system yesterday (no ssh nor X access atm) and for some reason guix pull complained about not being able to migrate generations because the profile path already existed
<fps>hmm, and on another machine xorg doesn't start correctly anymore (and the console has quite a few messages about dbus not being able to find its socket)
<fps>oh, i was out of video memory in the vm :)
<fps>let's see
<fps>yeah, increasing it makes xorg work
<fps>oh well, not really. it seems slim is borked now,too
<fps>herd does weird things to xorg-server, too
<fps>herd restart xorg-server
<fps>does not restart it
<fps>but rather disables it
<xelxebar>Hrm. Installing on a foreign distro, I'm hitting a chicken-and-egg problem.
<xelxebar>Want to install glibc-locales, but would rather use substitutes instead of building everything locally
<xelxebar>however guix archive --authorize reports "guile: warning: failed to install locale" and then appears to sit doing nothing
<gnu_srs2>Hello, is it possible to extract Shepherd from Guix and create a Debian package running as init, i.e. PID 1?
<xelxebar>Actually, I believe that warning is just throwing me off.
<xelxebar>ugh. Nevermind. Just a RTFM problem. I was supplying the key path as an argument instead of putting it on stdin :/
<aerique>What's the difference between `(use-modules (gnu packages wm))` and `(use-package-modules wm)`?
<aerique>I'm using the former since that was suggested when running `sudo guix system reconfigure /etc/config.scm` but I'm seeing the latter here for example: https://notabug.org/thomassgn/guixsd-configuration/src/master/config.scm
<aerique>Also, recompiling the kernel when I just add `(screen-locker-service …)` is not ideal ;-)
<leoprikler>What's the difference between `(use-modules (gnu packages wm))` and `(use-package-modules wm)`? <- only a syntactical one, it's sugar
<leoprikler>the kernel should not be recompiled if you just edited the file, did you perhaps guix pull before?
<aerique>I have done some pulls in the past yes. So that's it, thanks.
<aerique>I'm also using a custom nonguix kernel.
<leoprikler>That should hopefully not make a difference (other than the non-freeness of course)
<leoprikler>if the kernel is already built, there should not be a rebuild
<joshuaBPMan>Hello guix!
<PotentialUser-79>hi, "guix pull" doesn't see channel from channels.scm. And update channel I commented =)
<rekado>using the cuirass database we can restart an evaluation by deleting the corresponding checkout. Can we do the same for a single build?
*rekado wants to add restarting of builds to the admin interface
<fps>hmm, what service starts dbus?
<leoprikler>probably dbus-service-type
<fps>i'm just asking because login into i3 started to fail (screen just gets black) and i see some errors about dbus in the Xorg log
<leoprikler>what errors exactly?
<fps>since i don't have an x server it's hard to get them to a paste site..
<joshuaBPMan>fps: haha. I've had days like that before!
<fps>hmm, there once was a mouse console tool in the olden days
<fps>hmm, systemd definitely has the way better logging support ;)
<leoprikler>fps: there's emacs-debpaste :)
<fps>actually, i was just confused. i had an .xinitrc and .xsessionrc and it went into xmonad instead of i3. i use different bindings in xmonad..
<fps>slim, y u no tell me what you log me into?
<aerique>fps: gpm for console mouse
<joshuaBPMan>morning guix. I'm trying to create a simple guix service to run a simple web site. I want it to run a the command "guile web.scm", but I'm having trouble getting it to work on reconfigure...
<joshuaBPMan> https://paste.debian.net/1112646/ would someone mind taking a look at my service thus far?
<joshuaBPMan>the error I'm getting is "In procedure struct_vtable: Wrong type argument in position 1 (expecting struct): guile"
<leoprikler>do you want to run this service as your user when the system starts?
<leoprikler>or would it suffice to run it in your user session?
<joshuaBPMan>leoprikler: It would suffice to run it in my user session, but I would rather learn how to run it as a service when the system starts...that way i can learn how to add services to guix proper.
<leoprikler>I see
<leoprikler>I personally don't have custom services in my config.scm, but some things do strike me as odd
<joshuaBPMan>I can't really seem to find much documentation about adding your own custom services.
<joshuaBPMan>the documentation for the shepherd seems different than the services I've found in the guix source code.
<leoprikler>s/custom services/custom Shepherd services/
<leoprikler>indeed, they are different
<joshuaBPMan>hmmm. I wonder why.
<leoprikler>you can't write guix code inside your shepherd init.scm
<leoprikler>because Guix needs to know more than just the shepherd part
<joshuaBPMan>leoprikler: are you saying if I want to add a shepherd service...then I had to add it into the guix repo? and recompile it?
<leoprikler>not quite
<leoprikler>you can add it to your /etc/config.scm as well
<joshuaBPMan>I've been running guix system for a while now. I use sway.
<joshuaBPMan>Right now, I've got it as a module.
<leoprikler>but the code you write for /etc/config.scm is different from shepherd/init.scm
<leoprikler>anyway, back to topic
<joshuaBPMan>ahhh. gotcha.
<leoprikler>guile-web-service-type should extend shepherd-service, passing guile-web-shepherd service
<joshuaBPMan>leoprikler: I believe that you are talking about the end of the file?
<leoprikler>yep
<joshuaBPMan>where I define guile-web-service-type?
<joshuaBPMan>ok.
<leoprikler>also, the profile-service-type is probably not needed
<leoprikler>you're very likely going to have `info guile` in guix
<joshuaBPMan>leoprikler: gotcha. I just copied the code from mariadb...I think.
<joshuaBPMan>leoprikler: I just find it annoying that the guix documentation doesn't have much information about how to make your own services.
<joshuaBPMan>at least I haven't been able to find it.
<leoprikler>the shepherd service looks good to me so far
<joshuaBPMan>nevermind.
<joshuaBPMan>I just found a writing your own services...
<joshuaBPMan>section
<leoprikler>The Guix manual has been described as terse before, so I can somewhat understand you.
***dekenevs is now known as mitescugd
<leoprikler>user and group in the exports is probably wrong
<leoprikler>also, there is a bigger problem: using a module does not really help you find the service from your config.scm (unless you hve a custom channel)
<leoprikler>s/hve/have
<joshuaBPMan>leoprikler: I think I have it working as the module.
<joshuaBPMan>my config.scm file has a (add-to-load-path (dirname (current-filename)))
<leoprikler>that's pretty hacky, but okay
<leoprikler>one problem less
<joshuaBPMan>hmmm...well I had it the use module working before...
<joshuaBPMan>now I'm getting the error that 'did you forget (use-modules (guile-web))'
<joshuaBPMan>grrr.
<leoprikler>did you forget (use-modules (guile-web))?
<leoprikler>after adding the directory to the load path?
<joshuaBPMan>no. I have that there.
<joshuaBPMan>hmmm. I should probably go do some actual work now. Thanks for the help leoprikler!
***ng0_ is now known as ng0
<kirisime>So, it seems that offloading builds was failing last night because ci.guix.gnu.org was down and the --no-substitutes switch doesn't affect the machine doing the building.
<kirisime>Is there a simple workaround, or should I claim it's an issue deserving attention?
<leoprikler>The workaround would be to run the offload remote with --no-substitutes, I guess.
<nckx>raghavgururajan: ‘RTM’ (‘we sent the files to the factory to be printed on boxes and pressed onto CDs’) is complete nonsense. It should be changed to ‘Current’ as in the Trisquel and Hyperbola articles.
<nckx>kirisime: I think ‘--no-substitutes’ should just be passed on to any offload machines, yes.
<nckx>Keep the common case sane.
<kirisime>nckx: Yesterday, the build processes would just freeze like any other guix build with substitutes enabled.
<kirisime>And today I tested the setup again by running `guix build --no-substitutes linux-libre' and the build machine just did the substitution on its own and didn't build a thing.
<nckx>kirisime: Yes. There are in fact 2 (related) bugs: what you note about --no-substitute above, and the fact that substitution is ridiculously fragile (it still regularly crashes here when the HTTP response isn't to Guix's liking).
<nckx>And, as you noticed, the default time-outs aren't great 🙂
<nckx>(By ‘bugs’ I mean ‘annoyances’, please do open actual bug reports if you want.)
<kirisime>There's a timeout?
<nckx>😃
<nckx>I dunno, I'm an impatient boy.
***seswu_ is now known as ixtlan
<kirisime>nckx: Recommend me a font that'll let me see those unicodes though.
<nckx>You missed :) and :D, i.e. nothing, but: font-google-noto.
<nckx>Less a recommendation than a ‘it's the only Guix font I know that has 'em’, tho.
<leoprikler>I have the noto font and some other installed, but Emacs still won't show the Emoji. What do?
<kirisime>Is guix missing fonts? I wrote a package for the small bitmap tewi font.
<nckx>leoprikler: I think that's an emacs issue.
<nckx>kirisime: What do you mean by missing? We don't have all the fonts, more free ones always welcome.
<Franciman>Hi
<Franciman>I successfully installed guix, it's awesome!
<nckx>Hullo Franciman! Thank you.
<Franciman>I mean guixSD
<Franciman>One thing I didn't quite get is how to make RAID volumes
<Franciman>I have two disks
<Franciman>I tried to add a mapped-device, but I get this error, device /dev/md0 not found
<Franciman>should I first make the raid device?
<Franciman>and then write the config?
<Franciman>or what?
<leoprikler>Franciman: s/guixSD/Guix System/ :)
<nckx>Franciman: Yes, mapped-devices only handles already-exisiting arrays.
<nckx>You have to create them with mdadm & mkfs &c.
<Franciman><leoprikler> Franciman: s/guixSD/Guix System/ :) <- thanks!
<nckx>This isn't (yet) automated.
<Franciman>oh I see
<Franciman>thank you nckx !
<nckx>Good luck 🙂
<Franciman>do you think that software raid is good enough?
<Franciman>I don't care about the speedup, I have two ssds, it's already super fast
<Franciman>but I am afraid the software managed RAID could be slow
<nckx>Linux software RAID isn't just good enough, it's usually the best choice.
<Franciman>oh, perfect
<Franciman>final question, which font do you use? I have the default font, and I am missing a lot of characters
<nckx>No, it's as fast as it can be. And faster than most consumer
<nckx>*-level RAID cards.
<nckx>IMO safer too.
<nckx>Franciman: I actually install all of them…
<nckx>If you're looking for emojos, mine are displayed by font-google-noto. Except in Emacs, but that's OK for me.
<Franciman>thanks
<Franciman>I am mostly annoyed on iceca
<Franciman>cat*
<Franciman>I can't see numbers
<nckx>Franciman: I think guix install font-gnu-freefont-ttf fixes that.
<nckx>(Or putting it in your manifest or wherever you've put icecat.)
*nckx → AFK
<Franciman>ty
<nckx>Does ‘guix build vhba-module’ work for anyone here with an up-to-date Guix?
<nckx> http://paste.debian.net/plain/1112676
<str1ngs>I was able to get guix-daemon to run in a rootless container so no need to root access at all \o/
<puoxond>nckx: Works for me (commit: 79f0bd7).
<bavier>str1ngs: cool
<str1ngs>sneek: later tell dattashantih3 . I was able to get guix-daemon to work with a rootless cgroup. Though that might not get around your user mapping issue. But it's progress atleast
<sneek>Will do.
<davexunit>dongcarl: hi, me again. I was curious where/how you actually use packages like the xtoolchains that you showed me. I've been getting into some weird situations where I think guix is trying to build a bunch of stuff from source that it doesn't need to which indicates to me that I've made some mistakes. :)
<nckx>puoxond: Thanks. Time to clear random caches.
<nckx>str1ngs: …blog post…? 😉
<dongcarl>davexunit: I use it to create a reproducible build environment for Bitcoin Core releases
<str1ngs>nckx: I'm not sure it's time for that. since I'm using another package manager to achieve this. I'm hoping to replicate this in a more guix way.
<nckx>Fair enough. Still impressive.
<dongcarl>We previously had a system that had reproducibility only because we built inside an Ubuntu VM
<str1ngs>nckx: I'm either going to try to use nsenter with some bash scripts. or try to get the feature in build.cc
<dongcarl>With Guix we have bootstrappability, which is attractive for security-critical applications
<dongcarl>davexunit: most of it is here: https://github.com/bitcoin/bitcoin/tree/master/contrib/guix
<str1ngs>nckx: currently I'm using https://github.com/mrosset/via. which Iam the author of. and it supports rootless cgroups. so it's easier for me to reason what's going on.
<nckx>str1ngs: That's vaguely Guixy.
<davexunit>dongcarl: thanks!
<raghavgururajan>nckx Thanks!
<gnu_srs2>Hello, is it possible to extract Shepherd from Guix and create a Debian package running as init, i.e. PID 1?
<davexunit>gnu_srs2: it would totally be possible. shepherd is actually maintained as a separate project.
<gnu_srs2>davexunit: tks :)
<davexunit>np
<janas>hmm the docker build tests seem to be failing after my last upgrade, and the substitute server doesn't have it built either
<nckx>raghavgururajan: Would you be willing to do so? Wikipedants don't take kindly to being corrected by project maintainers.
<raghavgururajan>nckx Sure :-)
<nckx>‘Guix System Distribution (abbreviated GuixSD)’ is also woefully out of date.
<nckx>raghavgururajan: Thanks!
<nckx>Actually, the whole infobox still uses GuixSD. Bah.
<nckx>And the home page has changed 😛
<raghavgururajan>nckx Yeah, lot to change. I will do my best :-)
<raghavgururajan>nckx Btw, I wanted to ask you this. Do you know how to use JACK or sndio, instead of pulseaudio, for alsa?
<nckx>raghavgururajan: Not on Guix [System].
<raghavgururajan>nckx Hmm. I have ben told that pulseaudio is bloated and has issues (https://www.hyperbola.info/todo/sndio-migration/). So thought I could give jack or sndio a try.
<raghavgururajan>nckx Is it also same for networkmanager? Would I be able to use an alternative in Guix system?
<nckx>raghavgururajan: Guix has a Connman service, a manual networking configuration service (neither of which I've tested). The only thing stopping you from using another tool is writing a service yourself 🙂
<raghavgururajan>nckx Gotcha! Thanks.
<nckx>s/a manual/and a manual/
<nckx>Hyperbola unironically uses the ‘word’ KISS, so I don't trust their opinion on anything technical. Their claims about PA may or may not hold water.
*nckx → AFK
<leoprikler>I'm pretty sure, those problems with Pulseaudio are due to configuration most of the times.
<raghavgururajan>nckx I see. What about things against systemd? Did guix adopted shepherd because of guile or because of things against systemd? Just trying to learn. ;)
<raghavgururajan>* :)
<leoprikler>Probably because of Guile
<leoprikler>Guix is pretty Scheme all the way
<raghavgururajan>leoprikler I see.
<raghavgururajan>So in general, does guix takes design flaws of programs into consideration before adopting porting to guix?
<leoprikler>you mean before packaging?
<raghavgururajan>I mean, something beyond scope of FSDG.
<raghavgururajan>leoprkler Yes
<raghavgururajan>> ‎leoprikler‎: you mean before packaging?
<raghavgururajan>Yes.
<leoprikler>I think it's more of a community approach
<raghavgururajan>leoprikler I see. Anyway, I saw "udev" service as part of base services. Isn't part of systemd? How come it is in guix?
<leoprikler>Sure, people may be opinionated on this program vs. that one, but the one that you actually manage to build (especially in a reproducible manner) gets included
<leoprikler>and if you can build more than one, then that's fine too :)
<leoprikler>Not sure about udev belonging to systemd, but Guix does have packages, that provide systemd APIs, such as elogind
<raghavgururajan>leoprikler I understand.
<leoprikler>IIRC, elogind was even written for Guix, since GNOME needs the logind API and Guix doesn't run full systemd.
<raghavgururajan>elongind is a fork of logind, done by gentoo? logind is what associated with systemd?
<raghavgururajan>Ah
<raghavgururajan>leoprikler What is IIRC?
<leoprikler>If I recall correctly
<raghavgururajan>Cool
<leoprikler>and yes, udev is part of systemd, but Guix uses eudev, which comes from the Gentoo folks
<str1ngs>i never recall correctly myself :P
<leoprikler>Which just means my Alzheimer's not as advanced as yours :)
<str1ngs>I think eudev is compatible though no?
<leoprikler>eudev should be a udev replacement where systemd is not used
<str1ngs>leoprikler: what were we just talking about? I forget
<leoprikler>(at least that would be their mission statement)
<leoprikler>I'm not so sure either, let's check the logs
<str1ngs>is there a way to automatically start shepherd as a the user daemon?
<leoprikler>I use GNOME autostart
<leoprikler>you could probably do something with .profile or similar dot files
<nckx>I use .xsession.
<kirisime>> `guix pull' now honors `/etc/guix/channels.scm' <- this is nice, thanks
<raghavgururajan>Folks! Check this out: https://harmful.cat-v.org/software/ . It lists half of gnu packages are bloatware. bash, gcc, screen, nano .. how is that even possible?
<raghavgururajan>Even this: https://suckless.org/sucks/ . What is wrong with gcc?
<leoprikler>don't you know everything after 1970 is just bloated?
<janas>raghavgururajan: In general, gcc & friends try to cover lots of use cases, which leds to lots of code. Some people prefer simpler alternatives.
<janas>Also the people behind cat-v have a different design philosophy
<janas>I think they
<janas>'re affiliated with plan9?
<amz3>leoprikler: +1
<leoprikler>According to Epiphany, harmful.cat-v.org's TLS configuration is harmful :)
<janas>I'm a little surprised they support TLS at all
<raghavgururajan>leoprikler :D
<amz3>TLS is bloated.
<raghavgururajan>janas I see.
<kirisime>leoprikler: I had a chuckle as I clicked on `proceed to harmful.cat-v.org' yes
<leoprikler>"Info is bloated, use manpages instead"
<amz3>I like suckless ideas, but st for instance, is far less useable than gnome-terminal. Similarly, the wm of suckless is much less useful that i3.
<leoprikler>Because you can fit the guix manual onto one page ;)
<janas>Yeah suckless has some good ideas for sure, but you do sacrifice some usability
<amz3>that is what I was going to say.
<leoprikler>Do the Plan9 folks seriously use ed as a text editor?
<joshuaBPMan>M-x Butterfly
<janas>leoprikler: I think Rob Pike wrote the sam text editor for plan9?
<leoprikler>sure, but it's listed alongside ed
<janas>oh wow
<amz3>Yes, people still use ed.
<amz3>not me.
<kirisime>ed is the standard text editor. Which is to say, do you have ed installed? I don't.
<leoprikler>yet another +1 for Guix
<raghavgururajan>I think the issue is over-doing KISS principle. Things should be made simple enough, not simpler. The former retains required functionalities with uncomplicated *simple* design. Whereas the latter will lead to removing some required functionalities in the name of making this simpler and causing program to be unusable.
<kirisime>It might have a use case in non-interactive use, like how you can pipe a bunch of characters and escaped newlines into parted.
<leoprikler>kirisime: so like sed?
<amz3>raghavgururajan: true, mind the fact that scheme primitives are simple *powerful* primitives. It does not aim to take away features.
<amz3>raghavgururajan: while remaining KISS.
<leoprikler>But even Scheme is decidedly not simple where it is a usability improvement
<raghavgururajan>amz3 True! Modularity is a prerequiste for Simplicity.
<leoprikler>eg (define (fn . args) impl ...)
<amz3>leoprikler: what is the problem with define?
<leoprikler>"you could make it simpler by doing (define fn (lambda ...))"
<amz3>leoprikler: I was a bit too talktive when I said KISS. call/cc is not KISS as in "Stupid"
<amz3>leoprikler: yeah. I don't know why define (fn . args) exists.
<amz3>it is sugar syntax...
<leoprikler>which makes Scheme as a language more complicated, but at the same time easier to use
<amz3>I think guix follow the scheme philosophy, more than the UNIX or KISS principle. because immutable store is not "easy" to come up with.
<leoprikler>If you just go for "simple" all the way, you're never going to have functional package management.
<leoprikler>Because functional package management is not as simple as breaking your system with each upgrade.
<leoprikler>(I'm not endorsing the latter here, if that's not obvious.)
<kirisime>Make no mistake, unix today is the most complex system around. It's gotten here because everything that needed to be functional was instead made simple.
<raghavgururajan>> leoprikler‎: Because functional package management is not as simple as breaking your system with each upgrade.
<raghavgururajan>Pardon? Could you explain in other words please?
<leoprikler>I will, but a bit later
<leoprikler>for now I must leave the keyboard
<amz3>raghavgururajan: they say that guix is complex, unlike other distro where you can break you system doing an upgrade.
<amz3>is more complex in its underlying implementation.
<amz3>at the end of the day, debian is just: tar xvf software-1.0.deb && cp softare-1.0/ /
<amz3>or something like that.
<raghavgururajan>amz3 Hmm. Guix did break my system after upgrade :P . Having kernel panic while shutting down. Anyway, it has been repoted and being worked on I guess. :-)
<raghavgururajan>amz3 I could rool-back. That too. :-)
<janas>Doesn't the scheme spec provide default macros for all of the syntax-sugary stuff?
<amz3>raghavgururajan: ubuntu upgrades fails all-the-time. and you can not roll-back.
<janas>I t=wouldn't think it would count as making the underlying language more complex
<kirisime>raghavgururajan: Have you ever built and installed a bigger program manually, without using a package manager or even make install?
<kirisime>Essentially, all you need to do is make sure the right files are found in the right paths in the filesystem and that those paths are listed in an environment variable.
<raghavgururajan>amz3 Tell me about it. I regreted using ubuntu. :D
<raghavgururajan>kirisime Never. But I planned to learn and try it. It can make me understand some deeper stuff.
<kirisime>But with just that your system becomes one defined by possibly years of entropy, actions you are no longer able to revert or reproduce. The core idea is stupidly simple, but the end results in practise are always unmanageably complex.
<raghavgururajan>krisime I think it is same issue here. The idea excluded the essentials of package management. Thus making things simpler (and non-viable) instead of makings simple enough (straight-forward set of features). :)
<raghavgururajan>Also there is trial and error.
<raghavgururajan>"Discovery requires Experimentation".
*raghavgururajan holded their tongue into saying "hail hydra" at the end. xD
<kirisime>Hail cuirass?
<raghavgururajan>krisime Haha! I was referring to Marvel's Hydra.
<raghavgururajan>* kirisime
<raghavgururajan>I wonder what happend to rekado . Did not see them around for a while now.
<nckx>raghavgururajan: They are. Yesterday, even. Just not as often.
<nckx>Better things to do 😉
<leoprikler>And I'm back.
<leoprikler>raghavgururajan: Seems others already did my explaining, but yeah.
<leoprikler>Copying binaries to some locations is "simple", managing a symlink farm somewhat less so.
<leoprikler>And of course, all package managers are bloat. And so are your programs if you can't run make (or mk) directly from the unpacked tarball.
<bdju>does anyone know if NixOS has something like guix's manifest files?
<bdju>I'm planning to use it on a new computer soon because I will not be able to replace the WLAN card right away and it likely doesn't have linux-libre support
<ng0>leoprikler: texi2mdoc. converts texinfo to mdocml, but with some caveats. I tweaked the way gnunet's info documentation is written to the point where the mdocml output hsa none or only minimal information loss.