IRC channel logs

2017-04-21.log

back to list of logs

<OriansJ>This is gonna be a dumb question but, How does one encode in a guix build script the functional equivalent of "./bin/foo --bar1 blah1 --bar2 blah2 --bar3 blah3 "?
<lfam>OriansJ: You want to know how to invoke a shell command in a package recipe?
<OriansJ>lfam: effectively
<lfam>OriansJ: Here's an example: https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/packages/backup.scm#n517
<OriansJ>lfam: the problem is that I am trying to convert this http://git.savannah.nongnu.org/cgit/stage0.git/tree/bootstrapping%20Steps.org into a guix build script
<ivanbakel>Hey guys, I know you must get a lot of questions, but I was wondering if I could some help with an install issue
<ivanbakel>Every time I try guix system init, guix complains about the m4 package being an invalid gzip
<ivanbakel>Is there a way to easily fail it gracefully, or am I looking at some other problem?
<buenouanq>fresh install on actual hardware?
<ivanbakel>Yes
<ivanbakel>It's weird, because I also tried guix package -i m4@1.4.18, and it couldn't find that version
<ivanbakel>Though that's listed in the gnu package list
<OriansJ>ivanbakel: have you tried adding --fallback ?
<ivanbakel>I have not - is that an option for system init or package --install?
<ivanbakel>It's not finished yet, but that looks to have done it - it's certainly past the package download point
<ivanbakel>Guess I should read error messages more carefully, I assumed --fallback was an install flag
<ivanbakel>Thanks for the help
<buenouanq>there's been a lot of talk about --fallback this week
<buenouanq>there should maybe be a blurb about it in the install guide
<ivanbakel>I think that's definitely worthwhile, but I'm also apparently illiterate
<ivanbakel>Is the guide hosted on a git repo?
<apteryx[m]>ivanbakel: if your installing you should have the info manual opened in one of the tty
<apteryx[m]>The install guide is a section of that info document. It's also available online: https://www.gnu.org/software/guix/manual/guix.html. The chapter 7 should guide you through the install.
<ivanbakel>No, as in, I am reading the install guide, thank god
<ivanbakel>I meant to make changes to it
<apteryx[m]>The reference manual is authored in the texinfo format. You'll find it in the tell under doc/guix.texi!
<apteryx[m]>s/tell/repo
<CharlieBrown>Texinfo is hard.
<apteryx[m]>But the result is worth it :)
<CharlieBrown>True. I just wish the literate programming facilities for it were better.
<Gareth422>So, I managed to install GuixSD, finally, by my grub installation is wonky. How do I chroot into my installation and reinstall grub?
<Gareth422>Chroot complains that it can't find bash.
<Gareth422>*but my grub installation
<Gareth422>Well, I just changed my volume label so it boots, now. No need to fix it.
<buenouanq>that doesn't sound like something you should do by hand
<buenouanq>why not let guix system reconfigure take care of it?
<Gareth422>I have no idea what I'm doing.
<buenouanq>life secret: no one really does regarding anything
<bavier1>sneek: later tell lfam I see 29 patches currently not used by packages
<sneek>Okay.
<bavier1>sneek: later tell lfam and 4 patches in use but not in dist_patch_DATA
<sneek>Okay.
<efraim>The later is probably worse
<sneek>Welcome back efraim, you have 1 message.
<sneek>efraim, catonano says: thanks for explaining the old macbooks sense to me !
<Apteryx>Is there a smart word for a file which is a "non-link" (not a hard/symbolic link) ?
<Apteryx>Just a refresher: return #f when everything is OK, #t otherwise?
<civodul>Hello Guix!
<wingo>morning civodul :)
<wingo>civodul: you want a guile release today to fix the syntax thing?
<civodul>wingo: and a pony! :-)
<civodul>that would be nice of course
<wingo>:)
<civodul>but that's up to you!
<wingo>well i can just make a quick release, it would fix the gc 7.2 thing too
<civodul>ok, sounds good
<wingo>which is probably good, avoids problems later
<civodul>yes
<civodul>so you're restoring equal? on syntax objects, right?
<wingo>ACTION waits for an updated guix to get my environment straight
<wingo>yeah
<civodul>(haven't checked mail yet)
<civodul>awesome
<wingo>fyi hydra is serving http://mirror.hydra.gnu.org/guix/nar/y3nhc5n2misr91ydp907s50irjcssdfv-texlive-20160523b-texmf.tar.xz
<wingo>which is like 2 GB
<wingo>i thought that thing was not substitutable
<civodul>indeed
<civodul>well, i think the source is substitutable
<civodul>but nginx is supposed to filter it
<civodul>bah
<civodul>ok hydra.gnu.org returns 410 for that one now
<civodul>thanks for the heads-up!
<wingo>tx
<wingo>jeez i am getting segfaults now related to syntax objects, yarrrr
<wingo>ah it was because i added code to hash.c to hash syntax objects, it had compiler warnings because it didn't know about scm_syntax_expression, assumed it was int->int, when i fixed unrelated warnings and rebuilt it didn't rebuild hash.c
<wingo>so i thought everything was fine when in reality nope :)
<wingo>thanks c
<civodul>heh
<wingo>civodul: downloading http://mirror.hydra.gnu.org/guix/nar/y3nhc5n2misr91ydp907s50irjcssdfv-texlive-20160523b-texmf.tar.xz still proceeds
<wingo>for me anyway
***jonsger1 is now known as jonsger
<ng0>was it snape who took on the work with 0ad? if so, do you still need the branches on my side? I'll delete them if you no longer need them
<wingo>civodul: so when do you expect syntax objects to be equal? i am getting non-equalities sometimes due to the ribcage in the wraps
<wingo>which i find to be pretty weird
<wingo>like when comparing #'foo with #'foo, their wrap components are different
<wingo>maybe that is an artifact of the test though, humm
<wingo>yes i think that's what it is.
<civodul>ok
<civodul>the 'gexp' macro collects syntax objects in a alist in a first pass
<civodul>then it traverses the tree again and looks up objects in that alist
<civodul>so it's similar to (equal? #'foo #'foo) i think
<snape>ng0: you can delete them, I have downloaded the git repo.
<ng0>okay. otoh, i think I'll wait to push the delete until the patch is o nthe list
<snape>ok, I should send it this week-end.
<ng0>cool
<wingo>i was using pass-if-equal which expands #'foo in two different environments
<wingo>that was my problem
<wingo>civodul: would you mind fixing the texinfo thing? that way i can roll a tarball
<wingo>currently it wants to download the 2G file
<wingo>er, texlive, rather
<wingo>with guile 2.2 + guix i get lots of things like
<wingo>Found valid signature for /gnu/store/i92s6sqzlps5wasx09wv3lckn2lj3dm7-perl-5.24.0
<wingo>From http://mirror.hydra.gnu.org/guix/nar/gzip/i92s6sqzlps5wasx09wv3lckn2lj3dm7-perl-5.24.0
<wingo>Downloading i92s6s…-perl-5.24.0 (50.8MiB installed)...
<wingo> perl-5.24.0 492KiB/s 00:30 | 14.2MiB transferred
<wingo>gzip: stdin: invalid compressed data--format violated
<wingo>guix substitute: error: corrupt input while restoring '/gnu/store/i92s6sqzlps5wasx09wv3lckn2lj3dm7-perl-5.24.0/share/man/man1/perlrun.1.gz' from #{read pipe}#
<wingo> perl-5.24.0 492KiB/s 00:30 | 14.2MiB transferredkilling process 7938
<wingo>i wonder if it's a force-output thing, something related to ports perhaps?
<wingo>like, guix download on that file works but then when i go to build it it fails to extract the tarball or something
<Petter>wingo, have you seen this thread? https://lists.gnu.org/archive/html/guix-devel/2017-01/msg00139.html
<wingo>i think zero-length nars are another issue
<Petter>Ok.
<wingo>it could certainly be though that the cached NAR is not valid or something
<wingo>since i am running a guix with guile 2.2 i think more things might break, dunno
<wingo>come to think of it, it's a terrible plan to use guix on guile 2.2 to cut guile 2.2 release tarballs :P
<wingo>too many things to break
<wingo>Petter: you are right though, that file appears to be corrupted
<wingo>rekado_: do you have the ability to clear corrupted cache entries?
<efraim>sneek: later tell lfam i had no issues building mesa or gtk+ from staging on aarch64
<sneek>Got it.
<wingo> http://mirror.hydra.gnu.org/guix/nar/gzip/i92s6sqzlps5wasx09wv3lckn2lj3dm7-perl-5.24.0 is bad
<efraim>sneek: botsnack
<sneek>:)
<htgoebel>Hi, is somebody knowledgeable in Java development around? I'd need a bit of help on sisu.inject to give maven the last kick to run.
<wingo>also bad: http://mirror.hydra.gnu.org/guix/nar/gzip/i92s6sqzlps5wasx09wv3lckn2lj3dm7-perl-5.24.0
<wingo>er
<wingo> http://mirror.hydra.gnu.org/guix/nar/gzip/6lcsaddzj6q3facaidajc6ws0yp0m3dq-mpfr-3.1.5
<snape>are dependencies introduced by gexps in snippets counted as inputs or build-inputs?
<snape>(by snippet I mean origin-snippet)
<wingo>good question
<htgoebel>snape: What do you mean with "dependencies introduced … in snippets"? Example?
<snape>example: un unrtf package, are autoconf, automake, m4, grep and coreutils build-inputs or inputs?
<catonano_>htgoebel: what kind of help do you need with sisu.inject ?
<snape>s/un/in
<catonano_>ACTION going to lunch, will read later
<htgoebel>catonano_: It's a long error message ,please see my message on the mailinglist.
<htgoebel>snape: IC. AFAIK these are no dependencies at all, since they are working on the source - which is done prior to building any package. They are already required when running "guix build --source unrtf", while the inputs are not Some guix core-develeoper should know in detail.
<snape>makes sense. Thanks!
<ng0>efraim: what was this kind of device you are porting GuixSD to arm?
<Petter>I'm trying to build xfce4-notifyd. In the store I find a .desktop file, and a dbus .service file. I don't understand how I should tell the system about these; or maybe it just works automagically on reboot?
<wingo>Petter: generally things installed by packages don't have much in the way of magical service integration
<wingo>guix is a bit more explicit for better or for worse
<wingo>for dbus specifically i think what happens is that the package has a corresponding service
<wingo>e.g. the colord package has a service, and that service extends the dbus service, adding in dbus service extensions that come from the colord package's dbus files
<wingo>but notifyd is a user thing i think right?
<wingo>so it's not a system question afaiu
<wingo>there i think the user's dbus (different from system dbus) will look in ~/.guix-profile/share/dbus or whatever for available services
<wingo>iirc anyway, this stuff keeps getting paged out :P
<Petter>I'm not sure what it is. I can run xfce4-notifyd-config, and when I click "Show Preview" it shows an error box with: "GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name org.freedesktop.Notifications was not provided by any .service files"
<Petter>But this service file is in /gnu/store/.../share/dbus-1/services/org.xfce4-notifyd.Notifications.service
<Petter>Looks like I should make a symlink from .guix-profile/share/dbus-1
<Petter>I don't see how any of the services I have in .guix-profile/share/dbus-1 make their symlink though.
<wingo>Petter: did you install xfce4-notifyd yo your user?
<wingo>i.e. guix package -i ...
<wingo>that is usually what it takes to fix it
<Petter>Not yet, I've just been running it directly from store.
<Petter>Oooh, let me try.
<Petter>Hey, preview works!! :D
<wingo>:)
<Petter>Thanks!
<Petter>And more importantly, the command `notify-send` now works. I think I'm done here. *celebrates* :)
<wingo>:)
<efraim>ng0: I have an odroid-c2 I've been using mostly and I've unboxed the firefly 3399 with 4 GB of RAM but haven't started with it yet
<ng0>okay, thanks. noted, so that I can either get started myself or tell others who are interested
<civodul> https://gnu.org/s/guix/packages/reproducibility.html is slowly back in business!
<civodul>(bayfront had not built much since the core-updates merge)
<efraim>I think the odroid will be easier, it boots from a micro SD card without special magic I think
<efraim>I should figure out a custom kernel and the too-slow storage test failures for shephard and wingod/elogind
<wingo>apparently there are new elogind releases in elogind/elogind
<wingo>i have not tested them tho
<wingo>someone from gentoo or something took over mostly, seems to have done a good job
<wingo>ACTION trying to not be a roadblock :)
<efraim>Manjaro also uses them iirc
<efraim>Also once I build an aarch64 install image I'm going to try this guide https://wiki.debian.org/InstallingDebianOn/96Boards/HiKey
<wingo>civodul: neat reproducibility thing :)
<snape>Can a guix package embed a font that is not "free"?
<wingo>ACTION does not know
<ng0>what is "free"?
<ng0>licensing?
<civodul>snape: take a look at https://www.gnu.org/distros/free-system-distribution-guidelines.html
<civodul>which is what Guix follows
<snape>thanks civodul
<snape>ng0: free as in 4 liberties
<civodul>basically the part about "Non-functional Data" would suggest we can add non-free fonts, for example
<civodul>which is more permissive than what Debian does
<civodul>i'd rather not do that, though
<civodul>i don't think we've ever had to answer this question before
<ng0>there are fonts included in 0ad-data where I had not checked the licenses
<snape>the font I'm thinking about is https://www.gnome.org/fonts/#Final_Bitstream_Vera_Fonts
<snape>there is a clause "no copy of one or more of the Font Software typefaces may be sold by itself"
<snape>which, in my opinion, qualifies as unfree
<civodul>oh? indeed
<civodul>/me looks
<civodul>
<civodul>
<civodul>
<civodul>
<snape>(and yes ng0, it is for 0ad-data)
<civodul>oh but see the FAQ: "You can sell the fonts as part of any software package, however"
<snape>yes, I've seen that
<civodul>so if someone were to seel GuixSD CDs, that'd still be fine
<civodul>*sell
<snape>that's why I ask. It is unfree, but it can be part of a free package :)
<civodul>kinda weird
<civodul>we do have font-bitstream-vera 1.10 BWT
<civodul>*BTW
<civodul>is it different?
<civodul>ours is marked as X11-licensed
<snape>to me this is an orror. X11 does not contain the clause I'm talking about
<snape>*error
<snape>and there is a difference between packaging the font and bundling it as part of another package...
<civodul>but perhaps it's a different version or something?
<ng0>though if it is inside a system, is it still sold individual, isn't that again part of a software?
<ng0>font -> 0ad -> Guix/GuixSD
<civodul>our package dates back to 2013, could it be that we were sloppy at the time? ;-)
<snape>civodul: see the link, it's the same
<civodul>uh, that's bad
<civodul>so what did Debian do?
<civodul>i had never thought GNOME's font could be non-free
<civodul>the one for GNOME 3 is definitely free, i packaged it
<civodul>Cantarell that is
<ng0>it's in debian
<civodul>ACTION goes afk for a bit
<ng0>ttf-bitstream-vera - The Bitstream Vera family of free TrueType fonts
<civodul>sounds good
<ng0>we could simply ask GNOME
<snape>ng0, yeah, one can consider the font is part of a larger package, that is Guix
<ng0>why does my icecat still display 45.7 as a version while we have upgraded to 52.x?
<brendyn>Hmm. I use Bidsream Vera fonts, they look great
<brendyn>Bitstream*
<brendyn>They bloody better be "free"
<Petter>ng0, what does `icecat --version` show you?
<ng0>in about i meant
<ng0>this is terminal:
<ng0>[user@abyayala ~]$ icecat --version
<ng0>GNU IceCat 52.0.2
<Petter>You didn't start Icecat from terminal, no?
<ng0>nope
<Petter>Think you need to log out and back in again.
<ng0>yeah.
<ng0>I'll see a new version later, can't log out right now
<Petter>You could start it from a terminal as well.
<wingo>i thought guix fixed that thing where /gnu/store is embedded in copied ltmain.sh
<civodul>wingo: yes, commit 98341757c8
<civodul>is it broken?
<wingo>civodul: i think so yeah
<wingo>i mean, run autoreconf -vif in any package of yours
<wingo>then check your build-aux/ltmain.sh
<wingo>oddly there are two copies of the script in the ltmain.sh
<civodul>ah yes, /gnu/store/h79kczjrh7rnddszqz9wj107p6s4ibfw-libtool-2.4.6/share/libtool/build-aux/ltmain.sh has the wrong shebang :-(
<wingo>but maybe that is to be expected for some reason
<wingo>two shebangs there
<civodul>two shebangs?
<wingo>grep that file for /gnu/store
<civodul>oh, super weird!
<civodul>what's the story?
<wingo>no idea. i just fixed it again locally to be able to release guile :/
<civodul>annoying
<brendyn>wow, this is like the worst ToS ever https://www.vpsnine.com/hosting/tos.php
<brendyn>You can do anything, apart from all the things you want to do
<htgoebel>Re: Bitstream Vera: Wikipedia says: Because of its liberal license, the DejaVu fonts project is expanding it with additional glyphs and styles, which it releases into the public domain.
<htgoebel>So maybe we can drop Bitstream Vara al all (or make it to include only a few symlinks)
<htgoebel> https://en.wikipedia.org/wiki/DejaVu_fonts say: "the Bitstream Vera license, an extended MIT License which restricts naming of modified distributions and prohibits individual sale of the typefaces, although they may be embedded within a larger commercial software package"
<htgoebel>License see /gnu/store/gkm7hkpkkmrskvlwxiqrxw7nzl7hgivp-java-eclipse-aether-for-bootstrap-1.0.2/share/java/eclipse-aether-for-bootstrap.jar
<htgoebel>Argl. License see https://dejavu-fonts.github.io/License.html
<htgoebel>Explanation of copyright is on Gnome page on Bitstream Vera fonts http://gnome.org/fonts/
<snape>sneek: later tell htgoebel Thanks for the references. I already sent emails about it to the list (#26588). If you think it's not correct, please answer them.
<sneek>Will do.
<snape>does anyone know how to avoid being `id` 0 in guix environment -C ?
<snape>ACTION is fixing ctrl-s crash in icecat
<lfam>Does anyone know about the security-updates Hydra jobset? It is based on Git revision f3acbe324671ba960b0a9be76798e87d0b7bee58, but that revision doesn't seem to exist
<sneek>Welcome back lfam, you have 3 messages.
<sneek>lfam, bavier1 says: I see 29 patches currently not used by packages
<sneek>lfam, bavier1 says: and 4 patches in use but not in dist_patch_DATA
<sneek>lfam, efraim says: i had no issues building mesa or gtk+ from staging on aarch64
<lfam> https://hydra.gnu.org/eval/109617#tabs-inputs
<lfam>civodul, mark_weaver: Any ideas?
<wingo>ACTION did that horrible thing of "guix package -u --fallback" and went away thinking it would actually work
<lfam>`guix package -u . --fallback`
<lfam>Tricky ;)
<wingo>friggin irritating
<bavier>I agree it would be nice to have behavior closer to GNU getopt
<wingo>ACTION building bash a few more times for some reason
<wingo>wtf
<wingo>guix pull --bootstrap just failed
<wingo>In unknown file:
<wingo> ?: 0 [dynamic-func "strverscmp" #<dynamic-object #f>]
<wingo>ERROR: In procedure dynamic-func:
<wingo>ERROR: In procedure dynamic-pointer: Symbol not found: strverscmp
<wingo>builder for `/gnu/store/jagbm4f9npc7n7l9y4b1ncc25vigylky-guix-latest.drv' failed with exit code 1
<wingo>guix pull: error: build failed: build of `/gnu/store/jagbm4f9npc7n7l9y4b1ncc25vigylky-guix-latest.drv' failed
<wingo>i guess because bootstrap binaries are statically linked?
<wingo>ACTION having a terrible guix day today :(
<slyfox>civodul: 78dea6f1d4a85dd9571ccbd604239912ba3a18b8 ("gnu: ld-wrapper: Build with Guile 2.2") did fix core-updates bootstrap! thank you!
<slyfox>civodul: but [ to make sure https://lists.gnu.org/archive/html/guix-devel/2017-04/msg00315.html is not lost ] building 'bash-static' in core-updates still fails with arcane macro expansion error: http://dpaste.com/1SH2Y6Z.txt
<binario>hi
<binario>all
<slyfox>hi