IRC channel logs
2018-02-15.log
back to list of logs
<pkill9>is there a meta package that contains basic tools like sed and grep? I thought coreutils has it but it doesn't. It's for use with `guix environment` <civodul>pkill9: there's no such meta-package <civodul>there's 'sed' and 'grep' and 'findutils' though <civodul>i understand it's not as convenient as one would like :-) <pkill9>hmm i'm compiling something (veracrypt) which looks for 'cc' but can't find it in the `guix environment` even thoguh i gave it gcc-toolchain. on my distro 'cc' is just a symlink to /usr/bin/gcc <pkill9>is there a package i need to add to the environment? <adfeno>pkill9: I think it might aswell accept the CC variable to be set to "gcc". <pkill9>the variable where adfeno? In the veracrypt source? <pkill9>I changed the MakeFile to refer to 'gcc' where it was referring to 'cc' <adfeno>pass something like this to `make': CC=gcc <adfeno>I think some package recipes use this <Digit>ACTION wishes guix well, so well, oh so well, and wishes he had more to offer other than well wishing, but is spread too thin, so instead, every few weeks/months, he mentions/actions his love for guix, and thinks that encouragement will have to be enough as his contribution to guix <civodul>there aren't any other themes coming with GuixSD proper <civodul>if you write one, you could share it :-) <lorslara>I installed GuixSD a few days ago and I'm just trying to learn how to use it <rekado_>M-x guix-hydra-* with berlin.guixsd.org fail with gateway timeouts. <civodul>i guess that varies over time, given that it's still slow <civodul>rekado_: looks like it's working currently <pkill9>when using `guix import json`, imoprting a json file with 'license: false' or 'license: null' doesn't work <civodul>pkill9: could you send that to bug-guix@gnu.org so we keep track of it? <civodul>i suppose it should gracefully handle it instead of crashing <pkill9>ok, i just send it to that email? <pkill9>civodul: you can actually define the license as #f (aka 'false') <pkill9>so I think it's just not handling it <Apteryx>civodul: Morning! I'll try replacing invoke with a open-pipe* based approach which will both collect exit-status *and* output of program when it failed. I had done something like this in the (fiasco finder) script. <civodul>Apteryx: i'm not convinced we need something fancier than the current 'invoke' <civodul>usually program will write to stdout/stderr anyway, so you can just look at the build log to find out what went wrong <civodul>i'm in favor of keeping things simple here <davidl>taylan: I don't know if it matters, but I have a line just prior to "(dependencies mapped-devices))" that says "(needed-for-boot? #t)" <civodul>taylan: what error do you get at boot? LUKS partition not found? <civodul>you should double-check the UUID, and remember that it's a LUKS UUID, not an ext2 UUID nor a GPT UUID <civodul>'guix system init' now checks for possible mistakes in LUKS UUIDs, but 0.14.0 didn't <rekado_>ugh, I have accidentally embarked on a java updating spree on core updates :( <rekado_>I’m really just trying to build that bioinfo pipeline with a python 3.6 dependency on core-updates <rekado_>but there are so many dependencies, and some of those are broken on core-updates <rekado_>I still hope we can merge core-updates soon. <taylan>civodul: I commented out the uuid and used /dev/sda1, still no luck. maybe that just broke it further though, as I'm getting a different error now: "pre-mount actions failed". I think it previously was LUKS partition not found, followed by a byvevector. <civodul>rekado_: oh, looks like you're less optimistic than i am <civodul>taylan: "LUKS partition not found" means you provided an incorrect UUID <civodul>you should check with cryptsetup what the right UUID is <taylan>btw two questions: 1. can I reconfigure the system without needing to re-run all of "guix system init ..." from the install media? 2. can I access a shell from the REPL I'm put into when boot fails? <civodul>rekado_: is there anything non-java i can do to help you on core-updates? <civodul>taylan: 1. not really; 2. try ",bournish" at the REPL to get a Bourne shell-like interface <taylan>civodul: do I need to strip the dashes in the UUID? <rekado_>civodul: less optimistic? No, I think we should go ahead and merge it already. The little problems with broken packages will be sorted out when it becomes a problem for people. <rekado_>civodul: it looks like I’m now past the java problems (I’ve disabled a few non-essential failing tests) <rekado_>now it’s building a few haskell things, but I expect those to pass. <civodul>rekado_: whenever you feel ready, please go ahead and merge <rekado_>well, I wanted to have it built on hydra, actually <civodul>taylan: no, 'uuid' is a macro that checks the stringly-type UUIDs at expansion time :-) <rekado_>but my attempts to schedule a new evaluation just timed out. <civodul>there are still 3.7K scheduled builds, but i'm sure they'll succeed ;-) <pkill9>is it possible to pipe into the `guix import json` command? <mbakke>rekado: I can help with the Python packages, do you know some important ones that are currently broken? <pkill9>i tried `guix import json -` but it treats '-' as a literal file <rekado_>mbakke: actually, I don’t. I can’t get the list of failing builds from hydra at this point :-/ <mbakke>pkill9: It would be nice with a patch for the import command that treats '-' as stdin :-) <pkill9>yeah, though to be consistent it would probably make sense to do the same for all other places it takes a file e.g. `guix build -f <file>` <pkill9>dunno if that would complicate things <mbakke>Each of the guix subcommands are really self-contained scripts, so I guess we'd have to add it on a case-by-case basis. <mbakke>Perhaps it is possible to separate out the argument handling to a different module. <civodul>note that 'guix import' command-line handling is somewhat factorized <civodul>looks like supporting '-' is feasible <taylan>the /dev/sdxN nodes are supposed to exist in the initrd environment, no? I guess I found my problem... <civodul>if they're not, it could be because the initrd lacks a module <rekado_>I think in a few hours this will be done and I finally have an environment for this pipeline :) <civodul>you could write a testimonial: "thanks to Guix, it took me no more than a couple of weeks to get my pipeline up and running" ;-) <mephi42>I'm trying to use "guix download" to get python sources, but it fails with certificate validation error. wget and firefox download the same URL just fine. How to debug this? <rekado_>mephi42: have you set the usual environment variables to point Guix to the certificates? <rekado_>mephi42: the manual has a section on application setup, including certificate setup, which may be needed. <mephi42>I must have missed that. I'm using Guix on CentOS, so this could very well be the root cause. I will give it a try, thanks! <mephi42>rekado_: using the commands from the manual did not help, unfortunately: https://pastebin.com/HSMVrcQV . I noticed that after installing nss-certs, $HOME/.guix-profile/etc/ did not appear. <rekado_>you installed nss-certs but $HOME/.guix-profile/etc/ did not appear? What is the result of “ls $(guix build nss-certs)/etc”? <rekado_>so the etc directory should also exist in $HOME/.guix-profile <rekado_>you can also set the environment variable to the directory containing the certs on CentOS. <pkill9>had no problems with that so far <rekado_>mephi42: I’d suggest to not set SSL_CERT_DIR to the output of “guix build”, because garbage collection might make that location invalid at some point. <rekado_>but for a one-off command it’s fine. <pkill9>in the package definition, how do i tell `make` what directory the makefile is in? <pkill9>basically the tar file has a 'src' directoyr which contains the source and makefile, etc <bavier`>pkill9: sounds like you want to add a phase that does (chdir "src") <pkill9>i'll see how another package does it <pkill9>just found one that also changes to that particular directory <pkill9>hmm, it fails saying it couldn't find the directory, but it is there <bavier`>pkill9: maybe verify the phase is being executed from the directory you're expecting, maybe by printing the result of (getcwd) <pkill9>how do I print the result of (getcwd)? <pkill9>oh nvm, i'll just change the chdir part <pkill9>ah thansk tha worked, turns out it is in the 'doc' directoyr for some reason <pkill9>i guess the previous phase maybe? <rekado_>pkill9: if this is a tarbomb it would have entered the first directory it could find. <rekado_>in that case you may want to use the url-fetch/tarbomb method. <bavier`>which is nice for people who want to inspect the result of `guix build -S veracrypt` <pkill9>i hate when a tar file explodes all over my home directory lol <lfam>ACTION prepares GRUB bug report about test failures with latest QEMU <lfam>If anyone knows their way around the GRUB test suite, it may be a simple fix, so pipe up :) <pkill9>hmm, i think url-fetch/tarbomb is a little buggy, there's a couple messages in there about it if you do `grep -r url-fetch/tarbomb ~/.config/guix/latest` <pkill9>and it's not working for me, the build directory just has the environment variables file and an empty directoyr named 'source' <pkill9>hmm i may need to adjust some flags or something <pkill9>hmm, might need to set 'out-of-source to false <Myrl-saki>I come from Nix, from what I understand, Guix uses Guile Scheme, how does it handle lazy loading? <davexunit>there isn't a 1:1 correspondence between nix and guix <davexunit>lazy loading is a rather general thing. I guess we'd need to know how nix uses that to know if guix does something similar. <davexunit>Nix (the language) and Scheme are quite a bit different, so we probably implement features very differently <civodul>so what's up with core-updates? is it merged yet? :-) <lfam>Do you know if there is anything left for core-updates that I can help with? <civodul>rekado_ pushed a bunch of fixes earlier today, and said we could go ahead anyway <lfam>I've been running on my home servers for a week now, so I think the basics are working <civodul>it all seems that we're not waiting for anything concrete :-) <lfam>Has anyone tried booting a graphical GuixSD from it yet? <lfam>I'll try booting a GNOME desktop in QEMU <lfam>We should also check that locales besides US English are working properly <lfam>I sent a bug report to bug-grub about the GRUB test failures with the latest QEMU. It's not relevant to core-updates, just FYI <lfam>We don't have the GHC / MADV_FREE patch on core-updates yet <lfam>Building a VM image from desktop.texi now <lfam>civodul: It's taking a while to build what's necessary for the desktop VM. I'll test it whenever it finishes and send an update ot guix-devel. Hopefully we can merge ... tomorrow? <lfam>pkill9: I have to AFK now but you can list the changes with `git log core-updates...master`. In short, all the core packages are updated ;) <civodul>sneek: later tell lfam tomorrow sounds good! <pkill9>anyone know of a good guide to guile in context of package definitions? <pkill9>e.g. introduce guile/scheme concepts using examples of parts of package definitions <pkill9>if not, then i'll write one after I learn guile/scheme and package definition writing <OriansJ>pkill9: well the mescc-tools guix package definition is relatively simple as a basis for learning <OriansJ>The really interesting part is going to be when we replace the gnu-build-system with kaem as it is going to be bootstrappable from M1-macro. <pkill9>i need to make a trivial builder