IRC channel logs


back to list of logs

<caffe``>indeed, the openfwwf firmare works with my b43 card on trisquel
<buenouanq>fresh guixsd install
<buenouanq>attempting to run guile gives:
<buenouanq>ERROR: In procedure scm-error:
<buenouanq>no code for module (ice-9 readline)
<random_auroras>Is Guix essentially the de-facto Guile package manager? I can't seem to find any and Guix seems to have the packages...
<OriansJ>sneek: later tell janneke that we need his perspective on a possible new debug tool for you to massively cut your debug/devel time
<sneek>Got it.
<OriansJ>sneek: botsnack!
<caffe``>okay, i officially hate Trisquel.
<random_auroras>caffe``: Ah?
<caffe``>i'm not sure i can get on board with the whole idea that removing a user's freedom to read and choose what licenses they wish to accept, is somehow advancing the cause of software freedom. it seems more like a freedom being taken away, and choices being shoved down your throat. I can understand not wanting to include non-free software, or not wanting to package it... but to remove the user's choice is no better than what proprietary
<caffe``>vendors do.
<caffe``>i give up on this.
<OriansJ>I guess caffe failed to realize one can still install non-free software on Trisquel by just downloading and installing ubuntu debs
<OriansJ>I wish, I had the chance to discuss and educate them further, such that they understand that just because one chooses not to host non-free software does not mean non-free software isn't available for the system in question.
<yrk>OriansJ: it's somewhat akin to accusing a vegetarian restaurant of taking away people's freedoms
<OriansJ>yrk: actually it is more like accusing a vegan food delivery service of preventing you from going to McDonalds
<yrk>OriansJ: pretty much
<OriansJ>yrk: perhaps it indicates we might need to be more careful in our education of new members of our community about the responsibilities of the user and the responsibilities we as maintainers have assumed. Perhaps more verbose documentation about writing custom packages and binary blobs might be in order.
<random_auroras>If ubuntu debs are compatible, couldn't such a user simply use ubuntu repos?
<random_auroras>Perhaps pinning it so that it's only used on-request (like debian backports).
<buenouanq>Are all the services going to eventually be changed from `(...-service' to the `(service ...-service-type' form?
<buenouanq>I have no preference between them, I just like things no be consistent ( ._.)
<marusich>I want to know the output path of a fixed output derivation, given just the name, the hash algorithm, the hash, and whether or not it's recursive.
<marusich>I know that I can create a derivation (e.g., via derivation or raw-derivation) with the appropriate values for name, hash-algo, hash, and recursive?, and then apply derivation->output-path to it.
<marusich>However, I'm curious: is there any other easy way to get the output path? I noticed that we have a procedure called fixed-output-path in (guix store), but it doesn't seem to do what I want.
<platoxia>My system date is borked (GuixSD): 'date -d@0' returns Wed Dec 31 18:00:00 CST 1969.
<marusich>Hmmm, it seems I may have been mistaken. It looks like fixed-output-path may be exactly what I want. Perhaps I was just confused earlier.
<platoxia>Bah...nevermind. I'll just have to use the -u switch.
<marusich>platoxia, The command 'date -d@0' asks date to tell you the time at epoch timestmap 0.
<marusich>So, that looks correct to me, and it doesn't indicate anything is wrong.
<platoxia>I know, I'm in CST...which is utc -6...1800
<marusich>Well, when you pass @0 to the -d option, you're asking it to print the time at 0 seconds after the epoch.
<marusich>Sorry if you already understand; I just wanted to clarify in case maybe you didn't know.
<marusich>Details can be found in the manual at: (coreutils)Seconds since the Epoch
<platoxia>I guess I thought my system time would be set to UTC and just converted to my local looks like my system time is set to local time though, which is why it is giving me Wed Dec 31 18:00:00 CST 1969...that time is Epoch -6.
<marusich>Oh, I see what you mean
<platoxia>My configuration is right from the book so I guess the guile script that sets the system date is bad.
<marusich>I don't think the manual addresses how to set the hardware clock.
<marusich>Have you tried setting it wtih "date"?
<platoxia>I should check bios time
<marusich>Also, the "hwclock" may be useful
<marusich>"date" can set the system clock, but I don't think it can set the hardware clock.
<marusich>"hwclock" can set the hardware clock.
<platoxia>thanks, that looks like just what I need
<marusich>You may also be able to do what "hwclock" does from the BIOS.
<platoxia>Yup, that must be it. hwclock --show is giving me local time
<platoxia>Man, I haven't even run windows on this system since last year, lol
<Sleep_Walker>I have to do something about that docker - that looks like last blocker preventing me to use GuixSD for work :/
<snape>Sleep_Walker: you can package it ;)
<snape>that would be a great contribution, I believe lots of people need it
***knst` is now known as nordstrom
<snape>(it's probably not so easy though)
<nordstrom>+1 for docker
<rekado>past experience has shown that big complicated packages can be packaged by describing the dependencies publicly and asking for collaboration.
<efraim>is there a reason we haven't updated gnupg to 2.2.6? it built fine on my aarch64 board
<snape>efraim: it's in core-updates
<badlock>Hi. Is this the support channel for GuixSD?
<snape>oh no, it's pius, sorry
<snape>badlock: yes
<badlock>snape: Ok, thanks. :)
<badlock>I can't boot the system with the partitions in label or uuid mode (config.scm). It always returns a message "Waiting for partition root to appear". I only got it through set root=(hd0, msdos1); linux /gnu/store/w38../bzImage noapic root=/dev/sda1; initrd /gnu/store/87r../initrd; boot but... now the boot stops on the USB keyboard line.
<badlock>*Booting from grub cmd.
<efraim>snape: are you sure? I don't see it there
<badlock>I have a generic Dell keyboard. Any help is welcome.
<snape>efraim: sorry, I said just afterwards: "it's pius"
<efraim>snape: ah, ok
***knst is now known as nordstrom
<snape>badlock: it would help to have your config.scm (you can paste it to or similar)
<badlock>The installation did not report errors.
<badlock>I tried label,uuid and... device (seen in mailing list).
<badlock>I boot from grub commands but it doesn't pass from the keyboard.
<snape>badlock: is your partition table MBR or GPT?
<snape>oh yes
<badlock>It's an old machine that works with free drivers in Linux.
<badlock>No previous problems except noapic (bios time).
***knst` is now known as nordstrom
<snape>badlock: is /gnu/store on /dev/sda1?
<badlock>search --file /gnu/store... => (hd0,msdos1) and manually mounted to check
<snape>so is it that search --file --set command that doesn't work?
<badlock>It worked from cmd.
<badlock>Bye. See you soon!
<Apteryx>hmm? /tmp/guix-build-webkitgtk-2.20.1.drv-0/webkitgtk-2.20.1/Source/JavaScriptCore/API/JSHeapFinalizerPrivate.h:29:41: fatal error: JavaScriptCore/JSContextRef.h: No such file or directory
<Apteryx>hmm? /tmp/guix-build-webkitgtk-2.20.1.drv-0/webkitgtk-2.20.1/Source/JavaScriptCore/API/JSHeapFinalizerPrivate.h:29:41: fatal error: JavaScriptCore/JSContextRef.h: No such file or directory
<rekado>Apteryx: is that file in a directory on the CPATH or C_INCLUDE_PATH?
<Apteryx>it's at ./Source/JavaScriptCore/API/JSContextRef.h
<Apteryx>will need to dig a bit to learn about if this is in CPATH or C_INCLUDE_PATH
<Apteryx>in CMakeLists.txt there is: add_subdirectory(JavaScriptCore)
<Apteryx>that CMakeLists.txt file is at ./Source/CMakeLists.txt
<Apteryx>the error happened while attempting to build: make[2]: *** [Source/JavaScriptCore/CMakeFiles/LLIntOffsetsExtractor.dir/build.make:103: Source/JavaScriptCore/CMakeFiles/LLIntOffsetsExtractor.dir/llint/LLIntOffsetsExtractor.cpp.o] Error 1
<Apteryx>I can reproduce it with: ./pre-inst-env guix build --cores=1 -K webkitgtk
<Apteryx>(I've updated the version to 2.20.1 though)
<Apteryx>yep! I was starting to look at GnuCash since the old webkitgtk 2.4 failed to build with recent ICU. There is a fix here: but the real fix would be to move to GnuCash 3.0 which can use the latest webkitgtk, which I'm attempting to do.
<Apteryx>anyway, I have to run for now. later!
***otremblay is now known as mister_zombie
<roptat>Apteryx: I think it's already updated in core-updates (not sure)
<rekado>Apteryx: I already have an update for GnuCash 3.0
<rekado>(and for aqbanking and gwenhywfar)
<pkill9>is there any way to know the latest guix commit that has had all the substitutues built for it?
<rekado>pkill9: no.
<rekado>pkill9: but you can go back a couple of commits and run “guix weather” to see if it’s any better.
<snape>that would be a nice addition to Cuirass
<pkill9>when there's a new guix commit, does hydra reuse previously build things?
<pkill9>things built in previous commits i mean
<rekado>pkill9: it does not care about commits much.
<rekado>if nothing changed then no rebuild will be performed.
<rekado>the current state of the git repository only determines the derivations that will be computed.
<rekado>if these derivations have already been built no extra work will have to happen.
<niebie>hi #guix
<niebie>what support is there for swapping profiles on the fly? i've been using guix environment, but i what i would really like is to just swap the active manifest
<zybell_>You mean reboot?
<rekado>niebie: you can “source /path/to/profile/etc/profile” to load the environment variables for that profile.
<rekado>niebie: or you can use “eval `guix package -p /path/to/profile --search-paths`” to make those variables exclusive
<niebie>rekado: so -p swaps the profile active in my shell? i thought that flag was used to manipulate that profile instead, ie to install/uninstall things from that profile
<niebie>rekado: not to swap it in or out
<rekado>it does not swap anything
<rekado>it just specifies what profile “guix package” operates on.
<rekado>“--search-paths” prints environment variables.
<rekado>by evaluating them in the shell they are set.
<niebie>rekado: ok. so you take a manifest and install it with 'guix package -p <some path> -m <some manifest>', then to activate it i just source <some path>
<rekado>niebie: correct
<niebie>rekado: ok i've tried it and it works a treat. not that this is something hard to do, but is there any intention in having a command that manipulates these more directly? i would think it would make sense to be able to globally swap active profiles more cheaply than reinstalling whole manifests
<rekado>oh, you don’t need to reinstall the manifest
<rekado>you can simply run “source /path/to/profile/etc/profile”
<niebie>rekado: say i have a minimal manifest i default to, and want to activate a more detailed manifest globally. i've created profiles for them both as above. i haven't been messing with ~/.guix-profile right now, whould i just link that to the desired active profile?
<rekado>where “/path/to/profile” is the location where the manifest has been instantiated
<rekado>no, ~/.guix-profile is the default profile
<niebie>i'm not reinstalling the manifest for a single shell, but if i want to swap the default i would have to i thought
<rekado>it will be generated automatically when you use “guix package” without “-p”
<rekado>I don’t know what you mean by “swap”
<niebie>right now i have several manifests that i might want to have default
<niebie>not that there is a huge purpose for it
<rekado>to use a manifest by default you would do “guix package -m that-manifest.scm”
<rekado>this will create a new generation of your default profile at ~/.guix-profile with the contents of the manifest.
<rekado>but to just temporarily enable or disable software I’d recommend just to use the “--search-paths” feature.
<niebie>i think i might have muddied the waters here. say i have 3 manifests. i have installed them in separate profiles. i can make the default profile one of these manifests with the -m option. i'm saying that if i already have these profiles available for use with --search-paths, i'd think it would be smart to be able to swap the default between these manifests, rather than rebuilding the default with the -m option.
<niebie>i can work with the current behavior, i just think rebuildling the default profile every time seems a bit heavy handed.
<rekado>you could have ~/.default-profile as a link to the actual profile you want to use.
<rekado>in your ~/.bash_profile you’d source ~/.default-profile/etc/profile
<rekado>when you want to switch to a different one you could just change the target of ~/.default-profile
<rekado>upon starting a new shell that profile would be used as the default instead.
<niebie>i see. yea, that is a solution to this that is workable. i might write some scripts that do that for me. thanks for the help.
<roptat>there is an issue if your shell was started in a profile that required less environment variables
<roptat>switching profile that way wouldn't load new env vars
<rekado>roptat: even when using “--search-paths”?
<roptat>no, but you need to use --search-paths in every shell
<roptat>if you switch profile by flipping a symlink, new shells are fine, but old shells may not have all env vars correctly defined
<rekado>or in ~/.bash_profile
<rekado>yes, that’s why I wrote “upon starting a new shell…”
<roptat>oh, I missread
<roptat>but old shells will still be partly set to the new profile
<niebie`>rekado: here is the sort of thing i was thinking about
<niebie`>i'm not sure if it would really be useful outside of me.
<roptat>I like the idea
<niebie`>roptat: another thing the command could do is sychronize manifests and their profiles
<catonano>what is this field: (guile #f) ?
<mbakke>woot, I got substitutes from berlin for my `guix pull`.
<pkill9>hi, i'm running guix system reconfigure, and putting int he block to bind-mount /home/tmp to /tmp is giving me this error: In procedure scm_lreadr: /gnu/store/4hs8h95j3bf6b57bvnrsk6i3knn5g0il-shepherd-file-system--tmp.scm-builder:1:1050: Unknown # object: #\\<
<pkill9>this is the block i'm using in 'file-systems':(file-system
<pkill9> (device "/home/tmp")
<pkill9> (mount-point "/tmp")
<pkill9> (flags (list bind-mount))
<pkill9> (type "none"))
<mbakke>pkill9: "flags" is a list of symbols, so try (list 'bind-mount) (note the quote).
<buenouanq>Any thoughts on my guile thing?
<buenouanq>fresh guixsd install, attempting to run guile fails with
<buenouanq>ERROR: In procedure scm-error:
<buenouanq>no code for module (ice-9 readline)
<pkill9>mbakke: thanks, that worked
<mbakke>buenouanq: Did you install `guile-readline`?
<mbakke>pkill9: Excellent :)
<janneke>buenouanq: that sounds familiar, i think i've seen it
<sneek>Welcome back janneke, you have 1 message.
<sneek>janneke, OriansJ says: that we need his perspective on a possible new debug tool for you to massively cut your debug/devel time
<janneke>mbakke: hey, is guile-readline a new thing/requirement that's popping up?
<janneke>sneek: botsnack
<janneke>hey OriansJ, sounds swell
<buenouanq>If guile is supposed to ship with GuixSD, it should pull in everything it requires to actually run.
<mbakke>janneke: Yes, it was moved to a separate package this last "core-updates" round to trim the guile package.
<buenouanq>Then the core guile needs to be able to run without it.
<janneke>mbakke: ah, great -- it's been puzzling me but always was a "reminder" that i didn't yet enter a dedicated profile yet, and then it was gone again -- still weird experience
<mbakke>buenouanq: Did you back up your home directory by any chance?
<mbakke>If you copy a fresh ".guile" from /etc/skel, it should be able to gracefully deal with missing readline.
<janneke>mbakke: great, i think that was my problem -- i had my own .guile that simply enabled readline
<janneke>so much helpful, accurate information here, so fast -- really makes me feel happy and warm-fuzzy
<agaric>hello guix. is anyone else getting a "util-linux-2.31.1: unbound variable" error with the `guix` cmd? (this started happening after a guix pull a few days ago)
<agaric>just wanted to check if its a known thing before filing a report on help-guix
<mbakke>agaric: Which command do you run to get that error?
<agaric>mbakke: i've tried `sudo guix pull` and `sudo guix system reconfigure ...`. `guix pull` works without the error
<buenouanq>mbakke: I'm mentioning it here not because I can't fix it on my end, but because I feel it should be taken care of for future releases.
<Sleep_Walker>ACTION just learnt how to use package-arguments nad substitute-keyword-arguments :)
<mbakke>buenouanq: For new installs, the updated .guile in /etc/skel should be sufficient, no?
<agaric>mbakke: sorry, what i just wrote may have been confusing. the first two commands starting with `sudo guix..` fail with the error. but `guix pull` as a normal user works fine.
<buenouanq>mbakke: interesting - Was this added/changed like, last night?...
<mbakke>agaric: Which distro are you on? Does `sudo -E` make a difference?
<efraim>there's two util-linux packages
<buenouanq>because no, I didn't add any ~/.guile so it should have been the one from skel
<buenouanq>but they are different
<efraim>agaric: what if you use 'util-linux@2.31.1', with the '@'
<mbakke>buenouanq: readline was split from Guile in the start of February.
<mbakke>But no new release was made since then, so for fresh 0.14 installs the .guile will be wrong :/
<agaric>mbakke: oh wow, sudo -E works, thank you. (i'm on guixsd. and just curious: where would i use util-linux? i don't specify it in my configuration nor have i installed it specifically.)
<mbakke>agaric: I suspect there is a bug lurking here, where `root` would use some of your users package definitions (with the updated util-linux), mixed with some of its own package definitions (trying to use the updated util-linux, but cannot find it).
<zybell_>buenouanq:/etc/skel is *copied* by adduser at the time of run and then *forever* left alone
<agaric>mbakke: i see. FYI: this began happening after a pull about 3 days ago (+/- 1 day). also, `sudo guix gc` works fine without the -E. also, i just did the gc, and i see versions 2.30.1, 2.31, and 2.31.1 in the store. (dunno beyond that - im still very new to guix)
<buenouanq>right, which means that if this is a fully fresh install with new empty users, they should mirror skel
<buenouanq>so that they don't is very bizarre
<mbakke>agaric: Interesting, sounds like it came from the recent `guix pull` improvements.
<efraim>do we have an easy way to create an empty file during a build phase?
<efraim>or is call-with-output-file the easiest option
<agaric>mbakke: okay. i wish i knew more guix/guile to analyse this more. but in the meantime, i'll post on guix-devel to put it on record.
<mbakke>efraim: Guile has a (touch ...) procedure, but it does not do what you might expect :P
<mbakke>agaric: Great, thank you!
<agaric>mbakke: no, thank you :-D
<bavier`>efraim: (close (open "/tmp/foo" O_CREAT)) ?
<efraim>worth a try
<efraim> is quite annoying
<bavier`>debugging datamash test failure on aarch64; learned something about the %a format specifier
<atw>I'm excited about the idea of a Guix-based build tool ( I think Julien raises a good point about Maven.
<efraim>built freeimage finally with non-bundled inputs, building ogre now to see if it works
<nyberg>Is it possible to offload almost all work (compiling) when building a derivation to another machine? As Currently my machine overheats and stops half way during reconfigue.
<mbakke>nyberg: Yes, look here:
<catonano>atw: about the Guile build tool, I don't understand what this "guile" field is about
<catonano>atw: if you or anyone else knows, I'm all ears
<catonano>tomorrow I'll try to prod an emacs pacage in the REPL, anyway
<nyberg>mbakke: Thank you
<ngz>(for-each (cut copy-recursively <> "/a") '("b" "c")), where "b" and "c" are both directories in working directory, should create directories "/a/b" and "/a/c" right?
<ngz>In fact, wrong.
<ngz>It copies the contents of the directories, not the directories themselves.
<bavier`>ngz: I think you need to append the directory name to the destination
<ngz>But then, `cut' won't cut it (huhu) because the same argument is needed twice.
<bavier`>otoh, for two directories, you could just as well write out the expansion in full
<ngz>True. In my real use-case there are 6 of them, though.
<bavier`>ah, ok
<ngz>I used (for-each (lambda (d) (copy-recursively d (string-append location "/" d))) '(...))
<ngz>But I may need to create the directories beforehand.