IRC channel logs

2019-04-07.log

back to list of logs

<wednesday>who do i go to to add some python packages for me, i wrote what is needed for mypy
<apteryx>wednesday: have you considered adding them yourself? Or hire someone? Most people around here contribute on their free time, and unless your needs match their needs, it is unlikely that they will commit to "add some packages for you" :-).
<atw`>apteryx: my reading is that wednesday meant that they have written some package definitions and are looking to contribute them (wednesday, correct me if I'm wrong there ☺). The help for contributors is https://www.gnu.org/software/guix/manual/en/html_node/Submitting-Patches.html#Submitting-Patches and boils down to git format-patch and emailing guix-patches@gnu.org
<wednesday>atw`: yea you're right, and I,ll look into doing that maybe tomorrow morning
<quiliro>hello
<apteryx>wednesday: oh, sorry for my misunderstanding! I second the answer from atw`.
<wednesday>apteryx: am I right in saying you use license:expat for MIT licenses, and what do I use for Apache 2.0 licences?
<apteryx>about expat, yes, and fro apache 2.0, I believe in Guix it is called asl2.0
<apteryx>(license:asl2.0)
<apteryx>in the doubt, you can check the known licenses in the module (guix licenses)
<wednesday>awesome, thanks
<reepca-laptop>how do thunked record fields interact with match?
<quiliro>i want to use emacs shell mode to ./configure and make on readline package...
<quiliro>atw` suggest running guix environment readline and then ./configure and then make
<quiliro>i ran guix environment readline but then ./configure says there is no such file (inside [env])
<reepca-laptop>quiliro: you mean that the file ./configure doesn't exist?
<reepca-laptop>quiliro: which version of the source are you working with? Checkouts straight from version control usually have an extra first step that generates configure scripts and makefiles.
<quiliro>reepca-laptop: guix environment readline will not download the source on the current directory?
<reepca-laptop>quiliro: nope, it just sets up the environment so it has everything needed for building it
<reepca-laptop>quiliro: To get the source used in the package definition: tar xf $(guix build --source readline)
<quiliro>oh cool, so i just need the source now...before i needed the environment... i will just cd to the source environment i downloaded with git...
<quiliro>oh!
<quiliro>nice work guix devels!
<quiliro>so all i need to do is cd to the newly dowloaded readline source....
<quiliro>i am in love with my guix!
<quiliro>and with my emacs shell mode!
<reepca-laptop>👍
<atw`>I'm glad reepca-laptop beat me to that, I was struggling to remember ☺. quiliro, some other things I do: M-x compile, then guix environment readline -- ./configure, or I start emacs from within a guix environment
<quiliro>atw`: what? please describe
<quiliro>C-h haha
<atw`>well, M-x compile can be handy for running a command over and over again, as often happens when trying to compile something
<atw`>and the command in this case is "guix environment readline -- ./configure", which means "in an environment suitable for building readline, run ./configure"
<atw`>I mention it as an option, but do whatever works best for you!
<quiliro>C-h k compile
<quiliro>atw`: thank you for the tip
<quiliro>but first i must have run the command to download the source
<quiliro>and cd to that dir
<atw`>yup
<quiliro> atw`: but how can i tell compile to download the source and cd to that directory before running ./compile and then run make?
<quiliro>atw`: is it not better to open a buffer with shell mode
<quiliro>M-shell
<quiliro>and then run "guix environment readline", and then run ./configure and then make
<quiliro>on that emacs shell
<quiliro>?
<quiliro>atw`: otr is there a way wich is easier on M-compile?
<quiliro>s/otr/or/
<quiliro>s/with/which/
<apteryx>M-x compile is really just a way to present the output of a command (any command) so that is is convenient to jump to errors, etc.
<apteryx>it's not magical
<apteryx>the flow you outlined at the shell is correct
<quiliro>apteryx: ok...thank you both
<quiliro>nice info to jugle
<reepca-laptop>quiliro: It's a rather common use case that you'll be editing some source, saving, and trying to recompile after your changes. M-x compile makes M-x recompile available, which remembers the command for you
<reepca-laptop>and you don't have to switch buffers :-)
<atw`>for downloading and cd-ing, I would run those in a shell, as they're just done once. I find myself frequently re-running ./configure, make, and make clean, so M-x compile (for which I have a convenient binding) is nice for those
<atw`>exactly, reepca-laptop ☺
<apteryx>Can Guix deal with authentication (say, SSH) for channels?
<reepca-laptop>sneek: botsnack?
<sneek>:)
<reepca-laptop>ah, good
<reepca-laptop>sneek: later tell civodul: I find myself in (guix store database) wanting to use derivation-input-output-paths from (guix derivations), but if I try moving it to (guix store derivations) then it ends up indirectly needing (guix store) through &derivation-missing-output-error as it is currently. At no point is any "talking to a separate daemon process" done, though.
<sneek>Got it.
<reepca-laptop>sneek: later tell civodul: Any ideas on how to split it up?
<sneek>Will do.
<reepca-laptop>sneek: botsnack
<sneek>:)
***Blackbeard[m] is now known as Blacebeard
***Blacebeard is now known as Blackbeard[m]1
***Blackbeard[m]1 is now known as Blackbeard[m]
***cbaines_ is now known as cbaines
<efraim>hopefully the sbcl update causes it to build on armhf and aarch64
<efraim>Same TTY error on aarch64
<olivuser>Hello everyone. Is there a chance that the budgie desktop environment will be packaged with guixsd?
<pkill9>olivuser: if anyone makes the effort to package it, absolutely
<pkill9>does anyone else have a bunch of mountpoints on their desktop in xfce when using gdm?
<OriansJ>olivuser: any desktop environment where there is 1 person wanting it bad enough to package it; will be in GuixSD eventually.
<kmicu>In other words: there’re no plans to package it. Don’t expect it this year unless you want to contribute it by yourself olivuser ;)
<ng0>we should have all the dependencies. it's just a matter of whoever decides to work on it
<puoxond>Hello Guix
<puoxond>How do I replace GDM with SDDM?
<puoxond>I tried this https://paste.debian.net/1076564/, but it doesn't seem to work (I get a black screen instead of a login screen).
<wednesday>last I tried sddm it was starting without an x config, take a look at the xorg log
<puoxond>wednesday: What should I be looking for?
***buffet_ is now known as buffet
<wednesday>if it is starting with a config you should see things like [109829.640] (++) Using config file: "/gnu/store/cvbnyfi8sga3jpskvfy7rjz603bgsiwp-xserver.conf" and [109829.641] (++) Using config directory: "/gnu/store/iii7374j648x87hlfl96572y1v9vvaky-xorg.conf.d"
<wednesday>at the top of the log
<puoxond>There aren't any lines like that in the log. I guess it's still the same issue then.
<wednesday>puoxond: posting your log could be handy too, but im going now, i know slim works fine so you could use that in the mean time heh
<Formbi>why isn't guix pull displaying all the new/upgraded packages now?
<civodul>Hello Guix!
<sneek>Welcome back civodul, you have 2 messages.
<sneek>civodul, reepca-laptop says: I find myself in (guix store database) wanting to use derivation-input-output-paths from (guix derivations), but if I try moving it to (guix store derivations) then it ends up indirectly needing (guix store) through &derivation-missing-output-error as it is currently. At no point is any "talking to a separate daemon process" done, though.
<sneek>civodul, reepca-laptop says: Any ideas on how to split it up?
<civodul>reepca-laptop: i think <derivation>, <derivation-input>, and &derivation-missing-output-error all belong in (guix store derivations)
<civodul>in (guix derivations) you'd only keep the RPCs, basically
<civodul>(we might want to use email for asynchronous discussions tho :-))
*civodul fiddles with the installer
<pkill9>how does glibc know where to look for libraries? i.e. without LD_LIBRARY_PATH
<fanta7531>it just does
<pkill9>strace shows that it looks in /gnu/store/...glibc-... by default
<pkill9>hmm, unless it just takes it's own working directory as the root, i guess it ignores symlink, that part i forgot
<pkill9>oh it has a prefix specified
<atw`>I think there's a great blog post on an in-depth dive into how a "hello world" program is compiled, linked, and run (and then compare on Debian/Guix, on glibc/musl, ...)
<atw`>*a great blog post waiting to be written ;)
<efraim>i believe I found a bug in mcron
<civodul>"gnu/packages/curl.scm:55:4: error: libssh2: unbound variable", is it just me?
<civodul>oh, got it
<pkill9>how long does glibc take to compile?
<pkill9>on a 2.5ghz i5
<efraim>30 minutes on my aarch64 board I think
<efraim>so ... 10 minutes?
<pkill9>oh nice, i thought it was something that takes a few hours
<puoxond>GDM crashes when using EXWM as window manager. I believe it is the same issue that made it crash when selecting MATE prior to commit bfac636.
<ng0>pkill9: http://www.netbsd.org/docs/elf.html explains a bit of that story
<ng0>with some abstraction applicable to guix aswell
<atw`>added to my notes in case I try to write that blog post :)
<pkill9>thanks
<laalf>can someone tell me the current situation of lvm and plasma on guix?
<efraim>nevermind, apparently I didn't find an error in mcron, just an error in my .bashrc
<kmicu>laalf: no changes.
<laalf>kmicu: hm. something like home-manager yet?
<kmicu>Nothing like that was contributed yet.
<laalf>alright, thanks
<rendaw>Hello! I'm porting a system and need to mount tmpfs subdirectories (like /tmpfs/etc) over /etc with overlay. tmpfs mounts are created empty and I need to make the subdirectories before remounting them...
<rendaw>I was doing this in systemd with a service that went between the /tmpfs and /etc mounts that created the directories, but AFAICT filesystem definitions in guix can only have filesystem dependencies
<rendaw>Are there any other ways I can do this without copying/pasting/modifying gobs of core guix code into my system definition?
<rendaw>Like somehow appending (mkdir /tmpfs/etc) onto the tmpfs filesystem service start procedure or something
<puoxond>I just verified that GDM indeed crashes because of EXWM's .desktop file.
<puoxond>Setting the value of TryExec to the same as Exec fixed the problem.
<pkill9>how do you use the 'invoke' function in a trivial-build-system package?
<pkill9>ah nvm i worked it out
<apteryx>how smooth is a guix pull from the current Guix release (0.16) to current master?
<apteryx>any bumps to look for?
<apteryx>pkill9: I believe the RUNPATHS are embedded in the ELF binaries (which glibc must make use of)
<civodul>apteryx: there shouldn't be any bumps
<apteryx>great :-)
<civodul>puoxond: oh indeed, GDM crashes whenever a .desktop file does not refer to the absolute file name of the executable
<civodul>could you provide a patch for EXWM?
<apteryx>any idea how to refer to a path relative to the guile source file being executed? I'm thinking to use some hack where a package gets customized directly in the manifest file, using sources available in a relative directory.
<pkill9>apteryx: i don't think so, as when i placed symlinks to shared libs in /lib and such and tried to run a binary, it didn't find them
<pkill9>unless i tested the wrong path >.<
<apteryx>pkill9: by RUNPATH, I meant some absolute references to the dependencies of the binary, which gets populated at build time of said binary
<civodul>apteryx: if you use 'local-file' that happens automatically
<apteryx>it wouldn't look under /lib but in under the /gnu/store
<apteryx>civodul: oh, neat! thanks a bunch
<civodul>otherwise you could use (current-source-location)
<apteryx>ok!
<pkill9>apteryx: assuming i didn't try putting the shared lib in the wrong directory, what i did was try running a binary downloaded from the web, which was built using the filesystem hierarchy standard, and see if it found the binary when it uses ld-linux-x86-64.so.2 from the glibc package (by putting a symlink to that file in /lib64)
<pkill9>and it doesn't find it, and i think what's really happening is glibc searches for libraries in it's prefix by default, and on most systems that prefix is /, i think
<apteryx>I think our glibc explicitly disable ldconfig
<puoxond>civodul: I've just sent a patch to guix-patches.
<pkill9>yes it does apteryx
<apteryx>pkill9: so for that experiment to work I think you'd need to resort to patching the runpaths of the elf binary you're looking at
<apteryx>(I've heard this before, never attempted myself)
<pkill9>that's one way, although you need to modify each binary, i'm attempting to create a service that adds FHS support, so far all i've done is re-enable ldconfig, and generate an ld.so.cache file which is created using a generated ld.so.conf file that points to a union of a bunch of libraries
<civodul>thanks, puoxond!
<pkill9>apteryx: if you're interested this is what i've done so far https://paste.debian.net/plain/1076589
<apteryx>pkill9: interesting; I'm not sure why this wouldn't work; maybe try using an unpatched glibc to see if some Guix specific patches are causing your experiment to fail.
<pkill9>then i'll just make a service that puts a symlink to ld-linux.* in /lib64 (and /lib for 32bit binaries), and a symlink to the generated ld.so.cache in /etc
<pkill9>apteryx: oh the experiment is succeeding :)
<pkill9>i just had to remove a snippet from the glibc origin object, as it patches a file from 'use_ldconfig=yes' to 'use_ldconfig=no'
<apteryx>hehe, good!
<civodul>pkill9: looks like you're doing terrible things :-)
<pkill9>yes :)
<rekado>I think I’m done with tensorflow now. There’s still one problem with the RUNPATH of one file, but I can’t seem to fix it and it’s enough to delete the file…
<rekado>unbundled a few more libraries
<rekado>will send a new patch.
***Blackbeard[m] is now known as Blackbear
<linarcx>Gyus, why the log is awful?
<linarcx>I mean, you can download them, without any good gui and features like search ,...
<linarcx> https://bayfront.guixsd.org/.well-known/logs/
<linarcx>Maybe you should look at nixos logs page: https://logs.nix.samueldr.com/nixos/
<linarcx>it has better ui, you can search messages, see the realtime chat and, ...
<ng0>search used to be there, will come back including history. it's just that people working on it are the bottleneck with other responsibilities. it could however be done by people interested in it
<ng0>or we could simply (that is, maintainer to maintainer) hand you the logs if you feel like the waiting time is too long and no one really "wants" to work on this
<ng0>there was this series of accidents which lead to the way logs are the way they are now
<linarcx>ng0: In the hope of coming :)
<apteryx>what would be a nice way to archive the full source artifacts that were required to build a manifest?
<apteryx>can I pass the --manifest switch to 'guix archive'? Or maybe
<civodul>you can't!
***Blackbear is now known as Blackbeard[m]
<rekado>linarcx: ugly logs is better than no logs. That’s why we have ugly logs.
<pkill9>my fhs-support service works, i ran the lmms appimage without any modification to it (other than making it executable) \o/
*civodul -> zZz
<user_oreloznog>pkill9: congrats!