IRC channel logs

2019-05-12.log

back to list of logs

<apteryx>I had to pass the --build-users-group option to guix-daemon
<apteryx>although I have an issue where it seems the substitute should be available but it's always getting built locally (openssl)
<PotentialUser-22>Des français dans le coin?
<PotentialUser-22>Guix ISO installé uefi proce?
***Glider_IRC__ is now known as Glider_IRC
<lafrenierejm>atw: The full path to my guix clone is '/home/lafrenierejm/source/org.gnu.savannah.git/guix'.
<lafrenierejm>Oops. Previous message was in response to nckx.
<lafrenierejm>atw: Yes, `make check` gives the "names too long" error even after stashing my changes. I'm now just working on the latest master (2ce144eebb3e86e6f52509499d25a134309f24d0).
<nckx>lafrenierejm: I'm getting ready for bed; did you find anything relevant in the mentioned config.log?
<lafrenierejm>nckx: I have looked at the config.log that's created. grepping for "length" doesn't seem to give any hints as to how to fix the problem, though. It just lists the various lengths.
<lafrenierejm>The values listed in config.log for `guix_sbindir` and `guix_sysconfdir` both start with `/usr/local/` after having run ./configure with `-localstatedir=/var`. /usr/ does not exist on my system.
<nckx>lafrenierejm: Same here (except for sysconfdir but I set that to /etc manually), and I never had any trouble.
<nckx>(Also, prefix=/usr/local, but in practice you never ‘make install’ guix.)
<nckx>ac_cv_guix_test_hash_bang_length=93
*nckx → 😴
<nckx>Good night, and good luck.
<OriansJ>well this is weird: https://paste.debian.net/1083419/
<lafrenierejm>nckx: Good night to you, too. Thank you for your help.
<OriansJ>grep -R 1xs8fnhh48gs41wg004r7m4r2azh9khmyjjlnvyzy9c6zrd212x2 ~/.cache/guix/ returns nothing and neither does the other hash.
<jackhill>Hi Guix: I ran: `guix archive --export libreoffice > /dev/null` but received the error: "guix archive: error: hash of path `/gnu/store/xva6rqxkimlkhjzxm41bcbfvyxv725b2-libreoffice-6.1.5.2' has changed from `2dc8cbb29020fcd9f5760fe1d78c2fd781feceb816a2b873a497bbab9a098220' to `4fe455dd730a5fca3618f3186ea30b3990f84fffe5136bb97198fc784333e6aa'!"
<jackhill>is that because of grafting?
<reepca>jackhill: nope, grafting still follows the functional paradigm (instead of re-evaluating the function with different inputs to get the new inputs, it just applies a patching function to the outputs to make them refer to different store items).
<reepca>or rather, a more accurate description would be "produces new outputs that are like the old outputs but that refer to different store items" instead of "make them refer to different store items".
<apteryx>hmm, has anyone seen guix size reporting multiple times the same package with different hashes?
<apteryx>I have four variants of Python 3.7.0... which have the same size... eh?
<reepca>apteryx: 'guix size python' only reports one python of size 74.6 MiB for me
<reepca>my guix is from may 5, though
<apteryx>I'm checking the guix size of a particular OS config
<apteryx>you can generate it too: https://paste.debian.net/1083554/, if you'd like to test
<apteryx>I use: guix system docker-image robot-framework-system.scm
<apteryx>this will spit a tar.gz, that you should open to see what is the /gnu/store/xxx-system generated, and guix size that
<apteryx>the guix size output looks like this: https://paste.debian.net/1083555/
<apteryx>I need to get some sleep, but I'll be curious to see if someone else can reproduce it!
*apteryx zzz
<reepca>for what it's worth, 'guix size'ing my system's profile shows two python-3.7.0's
<brendyyn>Guix simply redefines what dependency hell means
<buenouanq>upward and onward
<brendyyn>my /gnu is 77GiB :/
<reepca>brendyyn: one-upped ya there, mine's 86GiB
<brendyyn>reepca: what command do you use to measure it? I used ncdu
<reepca>du -h -c -s /gnu/store
<brendyyn>mine shows the .links directory as 74GiB
<efraim>du -sch takes too long on my machines to bother running :/
<efraim>I check /home and subtract from df -h
<brendyyn>I still have a .config/guix/latest. should i remove this link
<rekado_>brendyyn: you probably don’t need it.
<g_bor[m]>Hello guix!
<g_bor[m]>As per yesterday's discussion with nckx here, I pushed the fix for icedtea directly to master.
<brendyyn>Noob question. I notice there are many versions of GCC; 4, 5 ,6 ,7 ,8 , 9. But the default one of the older ones. Arch builds with llvm 8 but this doesn't even seem to exist in guix. What goes in to choosing the default versions and what would happen if they were all updated to the latest, would everything break? would the things that didn't break run faster?
<roptat>hi guix!
<nckx>brendyyn: Some things break when updating GCC; these are fixed or built with an older GCC; but all this happens only on core-updates since it rebuilds all the things. Don't know if there's a GCC upgrade in the works there ATM.
<brendyyn>nckx: I notice the default gcc is 5, which puts it major versions behind the latest. would the update change it to 6 or what?
<brendyyn>4\
<nckx>brendyyn: As for LLVM 8, it's probably missing because nobody's done the work to add it yet. Maybe things have changed and it's a better compiler now, but most of the GCC → LLVM switching hype was ideological/political, with some mumbling about compilation speed to sugar the pill. As a GNU project, we can just stick with GCC ;-)
<efraim>IIRC core-updates is on GCC 7
<nckx>brendyyn: Oh, no idea, I think 7, indeed.
*nckx has soured on LLVM since; probably time to give it another chance in my heart.
<rekado_>g_bor[m]: thank you!
<brendyyn>nckx: well there was that mess with rms not wanting gcc to improve in certain ways, so maybe its not all that unreasonable
<brendyyn>nckx: or unreasonable, but anyway i was curious
<nckx>IIRC ‘improve’ here's subjective and was blown out of proportion, but 🤷, I'm not really taking sides anymore.
<sirmacik>hey, how can I get package.scm for a program with guix saved in to directory chosen by me?
<dutchie>sirmacik: you could just `guix edit package` and then use your text editor to save it
<kmicu>brendyyn: packages lags behind compiler updates so being conservative with default version requires less maintenance work.
<brendyyn>fair enough
<kmicu>(In practice new compiler version can make code slower too (I’m looking at you GHC). Wasting days on maintenance to maybe save some miliseconds is a terrible tradeoff.)
<divansantana>I'm trying to reconfigure my guix system on an older generation. (Because newer generations, have large packages that take long to build, then fail). Is this the right way of doing it? sudo -E /var/guix/profiles/system-88-link/profile/bin/guix system reconfigure ~/.config/guix/myconfig.scm
<divansantana>I'm guessing the above is not the best way, since it fails to find modules defined in my channels.scm
<cbaines>divansantana, if you're using a recent version of Guix, I'd check what guix pull --list-generations says
<divansantana>cbaines:that fails, even though have a recent guix version https://paste.debian.net/1083582/
***paroneayea is now known as dustyweb
<sirmacik>dutchie: thanks!
<jackhill>reepca: ah, okay got it. Thanks for the explanation.
<jackhill>I wonder if this means my disk is not returning the right bits.
<jackhill>I guess next time I'll use btrfs :)
<arshin>jackhill: no don't
<arshin>it's still full of caveats every step of the way
***apteryx_ is now known as apteryx
<Aurora_iz_kosmos>I'm wondering how `guix pull` & generations interact. Are previous generations garbage-collected when no user uses them anymore?
<Aurora_iz_kosmos>It seems the metadata would be enough to reconstruct them.
<brendyyn>Aurora_iz_kosmos: every user has two profiles. the one in ~/.guix-profile which contains all their packages, and one in ~/.config/guix/current, which contains just the guix package, a generation for every time guix pull is run
<brendyyn>you can see them by running guix package -p ~/.config/guix/current -I
<brendyyn>and delete old ones with --delete-generations that way.
<Aurora_iz_kosmos>brendyyn: Huh, neat, I didn't know that.
<amz3>there is 'gc' subcommand
<Aurora_iz_kosmos>amz3: Aye, but no-referencing generations is necessary before that works.
<Aurora_iz_kosmos>amz3: That's the detail brendyyn mentioned which I had not been aware of. How to do it.
<brendyyn>Hopelly someone will add a more convienient command to guix pull
<apteryx>is there an Scheme API to interact with Shepherd programmatically?
<Aurora_iz_kosmos>brendyyn: It'd be a fairly small PR, all thing said. I'd honestly be satisfied with it simply being documented more visibly.
<Aurora_iz_kosmos>brendyyn: Perhaps in `guix pull --help`
<brendyyn>it already has --list-generations
<brendyyn>ill have a look
<amz3>actually it is easy to create guile packages using guile-build-system
<amz3> https://git.sr.ht/~amz3/guix-amz3-channel/commit/af065f7d1f07f20c2bb9301ed5510e9ec92c6882
<anon12345>hello, i ran into another error that i can't seem to work around today when attempting to configure slim regarding xf86-vesa https://arborescent.tk/config.scm is my config and https://arborescent.tk/output.txt is the error message. It mentions config.log but im unsure how to locate it.
<anon12345>thanks for your time again.
<atw>apteryx: https://www.gnu.org/software/shepherd/manual/html_node/Services.html#Services gets into it, but are you looking for the in-scheme equivalent of e.g. herd start ath the command line?
<numerobis>Hi #guix! Is there a canonical shebang line to include for scripts on guix? Like the '#/usr/bin/env PROGRAM_NAME' on Nix? Something like '#/run/current-system/profile/bin/env PROGRAM_NAME' might work, but is this how other people do it?
<numerobis>*Forgot the '!' after '#'
<brendyyn>numerobis: there was a little discussiionn on the /usr/bin/env thing on https://lists.gnu.org/archive/html/guix-devel/2015-11/msg00414.html
<dutchie>If i'm packaging some non-free software for my own use, what should I put in the license field?
<dutchie>afaict the license constructors aren't exported so i can't make my own and i don't want to just lie
<atw>You can use #f
<brendyyn>Perhaps they should be exported.
***marlon_ is now known as Marlin1113
<numerobis>brendyyn: Thank you for the link! :)
<numerobis>brendyyn: The method 'sudo guix package -p /usr -i coreutils' works well. Would you know whether there is a way of doing this automatically/declaratively within /etc/config.scm?
<archetyp>Hi
<archetyp>GNU Icecat shows no numbers, neither in web pages, nor in the GUI, what is wrong?
<archetyp>OS: GuixSD
<str1ngs>numerobis: is this GuixSD?
<archetyp>Yes 1.0
<numerobis>strings: Well at the moment I'm still in a VM, but yes, I'm planning to move to guix system little by little.
<numerobis>*st1ngs
<str1ngs>numerobis: with GuixSD you can use a special file service
<numerobis>*str1ngs
<str1ngs>numerobis: it won't handle of of coreutils but you can cherrypick
<str1ngs>see my example here http://dpaste.com/0E9XMY2
<numerobis>str1ngs: oh, sounds interesting, I'll have a look. Is this in the documentation?
<str1ngs>it should be documented in services somewhere
<str1ngs>you can then add file-service to services
<numerobis>str1ngs: Amazing, thank you very much. This looks like what I am looking for!
<str1ngs>numerobis: https://www.gnu.org/software/guix/manual/en/html_node/Base-Services.html documents https://www.gnu.org/software/guix/manual/en/html_node/Base-Services.html
<archetyp>Oops, I was not meant
<str1ngs>numerobis: err documents special-files-service-type
<str1ngs>numerobis: one clarification, are you taking about scripts you might be packaging? or just some user scripts?
<str1ngs>talking*
<numerobis>str1ngs: just user scripts at this point. I am trying this method now and will report back.
<bavier`>icedtea fails to build on my kernel, log says "not supported" for kernel 5.0
<g_bor[m]><bavier` "icedtea fails to build on my ker"> bavier`: is it still so? I pushed a fix this morning, and now it builds fine for me.
<bavier`>g_bor[m]: ah, ok, I had just pulled last night; didn't expect a fix in the meantime already. I'll try again
<numerobis>str1ngs: I tried the 'extra-special-file' service, but I run into the error 'mkdir: Read-only file system' on guix reconfigure. I'll try other things and see if I can get it to work.
<g_bor[m]>bavier`: please report back with the results. Thanks.
<str1ngs>numerobis: is that something you defined using special-files-service-type?
<numerobis>str1ngs: I was going to, but then I saw in the documentation that using 'extra-special-file' (I basically copy-pasted the two lines in the list of services) is the simple way to achieve the symlink.
<kmicu>archetyp: do you see ‘rectangles’ only in place of numbers?
<bavier`>anyone know why our openjdk's 'javac --version' output includes a "-internal" suffix?
***marlon_ is now known as Marlin1113
<bavier`>I can't get Arduino IDE to build and run because of it :rolleyes:
<kmicu>jackhill: (to balance a negative comment) feel free to use btrfs, I had no issues with it on NixOS/Guix System since ~2014. 😺
<archetyp>kmicu: No, i see lines instead of numbers
<kmicu>archetyp: could you make a screenshot?
<apteryx>I'm guessing that 'guix size' is not educated about grafts and hard links.
<kmicu>archetyp: I guess that’s a very bizzare font issue and installing some more popular fonts could fix it (plus running ‘fc-cache -r’ and restarting IceCat).
<numerobis>str1ngs: I rebooted the VM, and now guix reconfigure works and adds the symlink. :)
<archetyp>I just install almost all fonts ... hope that helps ;)
<GNUtoo>hi, I'm trying to build guix for Parabola (and by extension archlinux), I've modified and built gnutls with guile support, and updated the package definition of guix to try to build the 1.0.0 version, and I've the following issue: "guix/gcrypt.scm:33:2: error: %libgcrypt: unbound variable"
<GNUtoo>Is there anything I could do to check if my guile environement is sane? there seem to be a guile interpreter
<GNUtoo>$ guile gives me: scheme@(guile-user)>
<GNUtoo>(2) https://git.elephly.net/ gives 502 bad gateway
<g_bor[m]>bavier`: I will have a look.
<str1ngs>hello, how do I add a input to an output's runpath?
<archetyp>The problem with Icacat is solved! \o/
<archetyp>Thank you kmicu :)
<kmicu>Glad that worked ( ^_^)/
<bavier`>g_bor[m]: thanks, I'm not familiar with java on other distributions, so I don't know if this is standard or not
<bavier`>same problem with arduino ide's master branch too
<g_bor[m]>bavier`: it seems it's both standard, and not at the same time. The situation is quite confusing. Debian seems to be ok with it, but Ubuntu considers it a bug.
<g_bor[m]>I am looking for a way to disable it, could not find it yet. The build system is bit confusing :)
<g_bor[m]>Disabling the internal part would be ok by me.
<g_bor[m]>ok, I belive I have it :)
<g_bor[m]>Will test this locally.
<lafrenierejm>What package provides support for Zip archives? I only just realized I don't have `unzip` when I attempted to run `dired-do-compress`.
<bavier`>lafrenierejm: the 'unzip' package
<lafrenierejm>bavier`: That... makes perfect sense. For some reason the package wasn't being listed by the guix Emacs package. Thank you!
<bavier`>np :)
<rekado_>GNUtoo: thanks for the report. git.elephly.net is back online.
<nckx>jackhill: +1. Been using btrfs since before the disk format had even been nailed down (~2011?). It's been fine for years. Didn't live up to the hype, and for the love of all your gods don't use raid5/6, but it's certainly not full of caveats all of the way.
<efraim>I use it with compression turned on and no other real special features
<ng0>i had one problem with it since it was introduced, that's all. and this one problem was only able to be fixed by manufacturer. I prefer zfs though.
<taylan>wow, many congrats on 1.0 :-)
<taylan>will have to update next week. I only find free time from weekend to weekend T_T
<kmicu>nckx: jackhill: I concur. Using unstable† features is not recommended unless you want complain later that btrfs is unstable 🤭 † https://btrfs.wiki.kernel.org/index.php/Status
<bt>sorry if anyone responded to my (anon12345) question while I was away, the system I was asking on was unstable, but I've now got my emacs config and such migrated over to my machine. If you wouldn't mind copy pasting your response if you made one it would be much appresiated. If no one awnsered I'll go ahead and re-ask. I'm trying to get slim or gdm running on my new guix system but I'm running into a error with xf86-vesa the error I
<bt>get is here: https://arboresent.tk/output.txt thanks for your help.
<amz3>bt: deadlink
<reepca>I remember this, there's a missing 'c' in there
<bt> https://arborescent.tk/output.txt
<bt>every time sorry
<bt>this is a different error, rolling back pull solved my issue yesterday thanks so much!
<reepca>bt: "guix build --check xf86-video-vesa" works for me, perhaps it's fixed in a newer version? When did you last run 'guix pull'?
<bt>a few minutes ago
<bt>maybe 30 actually
<bt>I can try again
<reepca>I doubt it's changed since then
<bt>okay I won't then do you know where the config.log it mentions would be?
<reepca>If I understand correctly, you're running 'guix system reconfigure <config.scm-file>' and during that this error appears? Then it wouldn't make sense for it to be trying to cross-compile or anything like that
<bt>yes that's correct.
<reepca>if you run guix build --keep-failed xf86-video-vesa and the build fails, at the end it'll add a note "keeping build directory <build directory>", which is usually /tmp/guix-build-<package>-<version>
<reepca>It'll be in a subdirectory of that that hopefully should be easy enough to find
<bt>okay I'll try that now.
<bt>I've still got a lot of my config to migrate over, like my keymap, sorry for the speed. Here's the config.log https://arborescent.tk/config.log.txt perhaps I've got to specify a newer version of the compiler somehow?
<reepca>bt: could you ls -l in the directory config.log is in and see what the permissions for conftest are?
<bt>it's not there
<reepca>hm, configure must have automatically deleted it
<reepca>it thinks it can't execute the stuff that gcc compiles because it tried it once and it didn't work. I'm wondering why that might be. gcc should make its outputs executable by default, right? And its owner should be the same user that created it. Could there be some sort of filesystem flags that would cause permission errors on attempting to execute a program?
<bt>my tmp has no-suid no-dev no-exec so that's probably coing it.
<bt>doing*
<reepca>yeah, that'd do it. I wonder why it'd be like that, though. You're on Guix System, right?
<bt>I set it to that because I usually do, yes this on guix system.
<bt>sorry for bothering you, thanks for your help
<reepca>ah, alright. No problem!
<nckx>While building core-updates, or trying to: while setting up the build environment: executing `/gnu/store/420l27y3qrbvc1cdr8jmbw7dnxp62h9c-guile-bootstrap-2.0/bin/guile': Exec format error
<GNUtoo>rekado_: thanks