<ng_>"Downloading 3wb66d...-texlive-texmf-minimal-2015 (428.4MiB installed).. " this fails.. what is this? texlive-minimal? <cehteh>but using google translate might be worth it <cehteh>Another "package" tells you whether a number is positive. This "package" is available in version 3.1. No really! <anthk_>well in the end I installed guix over trisquel <Jookia>paroneayea: FWIW I think eventually all functions like that should be put in a p2p filesystem that we can draw from to make a literate web of programming <str1ngs>hello, guix seems to want to build alot even though. I've authorized to use substitutes. Is there someone to avoid building so much? I'm really not into build a kernel with modules today :( <Jookia>str1ngs: What's it building? Have you run 'guix pull' lately? <str1ngs>but even if I say package -i emacs. it wants to build most of the world <str1ngs>guix (GNU Guix) 0.9.0 do I need to pull? <efraim>The binary substitutes from the 0.9.0 release were already garbage collected <str1ngs>also hydra.gnu.org is pretty slow for me :( <str1ngs>well that doesnt make sense to GC those. since anyone that is doing a guixsd is going to have to build most of the world? <Jookia>It's intended that everyone runs 'guix pull' <str1ngs>ok then I missed something in the documentation. my bad :( <Jookia>Not sure if it's in the live USB documentation (I haven't checked) so it's possibly not your fault. Oh well <str1ngs>off topic question can guix-daemon run as an unprivledge user? <Jookia>That's an interesting question, though I don't think so since it'd have to write to /gnu/store <Jookia>I think you could change it but it'd require rebuilding all the packages <str1ngs>aye since things like the dynamic linker are hard coded :( <Jookia>Even if you got past dynamic linking, applications would still want to read from /gnu/store for documentation and other data <str1ngs>going to be hard to convince people create a store. just so I can use /gnu <Jookia>You can't get a sysadmin to install Guix for you? <str1ngs>there public machines. luckly there are part of the GNU cfarm so it might be doable. <str1ngs>updating list of substitutes takes forever is that normal? <jmd>So it seems that for the purposes of "build --keep-failed" failures in the "check" phase are not considered failures. <jmd>Is that intentional? <civodul>jmd: "guix build" does not know about phases etc. <civodul>so it considers a failure when the derivation fails to build <jmd>That is not what I experience. <civodul>normally 'make check' exists with non-zero on failure, in which case the 'check' phase exits with #f, leading to a derivation failure <jmd>I started a build with --keep-failed, one of the tests fails, yet the build directory is gone. <taylan>it's currently on version 3.1.0! <taylan>(and (integer? x) (positive? x)) so difficult in JS to warrant a package it seems <str1ngs>I was reading about that on HN. kinda sum's up the node culture don't you think? <jmd>On further tests it seems non-deterministic. Sometimes --keep-failed is honoured. Other times it is silently ignored. <jmd>How can I find out what is going on? <str1ngs>updating list of substitutes takes forever. <civodul>jmd: it's unlikely to be non-deterministic; could you send more info on how to reproduce the problem? <rekado>str1ngs: unfortunately, it currently is like that. <rekado>we're painfully aware of the problem <jmd>Then run ./pre-inst-env guix build -K pspp <jmd>It'll build; then one test fails; and the build directory is gone. <jmd>(at least that's what happens for me) <civodul>and the thing eventually run 'make install' and succeeds? <civodul>after each build phase there's a message saying whether it succeeded or failed <str1ngs>rekado: ok does it not do any caching at all? like can I expect that to happen everytime I install something? <jmd>civodul: "make check" happens before "make install" doesn't it? <jmd>So it's not running "make install" then. <jmd>Let me run it again, and see. <jmd>civodul: Where is the build log saved? <jmd>The build ends with : <jmd>phase `check' failed after 522.9 seconds <jmd>builder for `/gnu/store/kwrz1bhj0w7r4fm0a68xsvbsb196dqzd-pspp-0.9.0.drv' failed with exit code 1 <jmd>guix build: error: build failed: build of `/gnu/store/kwrz1bhj0w7r4fm0a68xsvbsb196dqzd-pspp-0.9.0.drv' failed <jmd>But the /tmp/guix-build-.... file is not preserved. I expected it to be since I passed the -K flag. <civodul>jmd: normally there's a message like "keeping failed build in /tmp/guix-build..." <jmd>No. That is what I don't understand. <civodul>didn't you use --keep-going instead of --keep-failed? (just in case ;-)) <jmd>No. I used -K (upper case) <jmd> ./pre-inst-env guix build -K --no-substitutes pspp <civodul>nothing special about your system, like TMPDIR value of guix-daemon, etc.? <jmd>I don't think so. It's a bog standard guix on debian setup <jmd>I havn't specifically set anything up like that. <civodul>i'll try locally, but i don't think i'll be able to reproduce it :-/ <civodul>phase `configure' failed after 35.2 seconds <civodul>note: keeping build directory `/tmp/guix-build-pspp-0.9.0-g48deb6.drv-0' <jmd>Yeah. I found that if it fails during configure or make, then the build is saved ok. But not with failurs during test. <jmd>(which, as you can imagine, makes it difficult for me to find out why this test is failing) <civodul>jmd: with the recipe you send, it works for me: <civodul>phase `check' failed after 272.0 seconds <civodul>note: keeping build directory `/tmp/guix-build-pspp-0.9.0.drv-0' <jmd>Maybe if I kill the daemon before it finishes it'll leave the directory there. <jmd>Got any better ideas? <rekado>jmd: you could add a phase after the tests and make it return #f <civodul>yeah, that shouldn't make any difference, but why not <jmd>That might be a plan. <civodul>surprise surprise: hydra.gnu.org is completely overloaded! <ng_>could've told you that yesterday, took me ~18 hours for a normal guix package -u <ng_>update; fail; update; fallback; repeat <jmd>rekado: No. Of course that doesn't work, because if the check phase fails (as it does) my post-check phase never happens. <ng_>so when I want to run an executable in a build phase, is this done in the (system) I see often? what I need to do is execute settings/psyced; cd back to src/; "make all" in src; cd to util/; make subs <civodul>ng_: i have the same problem right now, due to the server's unavailability :-( <civodul>that seriously annoys me, to say the least <ng_>yes.. do you have any eta on the new server? <civodul>so far we've been waiting for Libreboot to work on server motherboards <civodul>that was supposed to already work, but then we learnt it did not <civodul>and then it was supposed to happen "soon", but it's not there yet <civodul>we'll see what can be done, and maybe consider plan B while waiting for Libreboot to be usable on those motherboards <rekado>apparently it's a compatibility problem when using the RAM we ordered. <rekado>I wonder if we could take some other RAM when it's still unknown if this can be fixed and how long it's going to take <civodul>nully: do you have more insight on Libreboot machines with those K16 motherboards? :-) <civodul>we need to find a way to move forward, it's just ridiculous <civodul>the more we try to remove load from hydra.gnu.org, the slower it is <civodul>currently it's idle (no CPU, no io), but "ps" takes 10 seconds to complete <ng_>i have this substitute line and I think this might be solved in a shorter way, but i know too little atm about substute: (lambda _ (substitute* "settings/psyced" (("exec ./configure-do --prefix=/opt/psyced --bindir=/opt/psyced/bin --libdir=/opt/psyced/world --libexec=/opt/psyced/run --with-setting=psyced-current $*") ("exec ${CONFIG_SHELL} ./configure-do --prefix=/opt/psyced --bindir=/opt/psyced/bin <ng_>--libdir=/opt/psyced/world --libexec=/opt/psyced/run --with-setting=psyced-current $*")))) <ng_>i don't want like 300 chars on one line <rain1>if you use system* you can have each arg as a separate string, which lets you put line breakn in <ng_>I just need to put the ${CONFIG_SHELL} after the exec <ng_>there's nothing like "insert X between A and B" ? <jmd>civodul: Did you try that on a GuixSD system or a Guix under something-else system? <civodul>GuixSD, but it's the same daemon anyway <jmd>civodul: I have a suspicion it might be a permissions thing. <benchmark123>It might be a good Idea to deploy a Benchmark Server in the future in addition to build-server to test for performace degration in hand with functionality. <ng_>not sure if that was spam or just impatient to wait for a reply... <ng_>well the software is real <ng_>but from wikipedia reading it's a big project. <ng_>and it supports hurd <ng_>a software which endorses hurd <ng_>I'll add it to my org to keep track of it <jmd>Does guix run on Hurd? <ng_>there's a hurd port going on <ng_>documentation sucks. it'll be a guessing requirements game <ng_>my guess is parts can be packaged, but the entire thing would have to sit outside of guix master <ng_>but I just looked at it briefly, no time to study it yet <rekado>I wouldn't bother looking at it when it's just some "please package this" posted by a drive-by account with a link. <ng_>yeah.. i was just curious <ng_>it wsa not very polite, but i'd like to check out the project later <ng_>i have xonotic on my list for this year, reading the debian packaging on it I have expectations that it will be challenging :) <ng_>hopefully it will all be easier once I can do guix packages as fluent as ebuild <ng_>can I do something like (lambda _ (begin (system* "boats") (system* "motorboats") (chdir "util"))) for multiple system* 's? <rekado>sure, but I'd check the return value <rekado>(and (zero? (system* "this")) (zero? (system* "that)) ...) <ng_>with (zero? in the beginning? <ng_>the fun part will be to package psyced after this, which has the function to reconfigure itself and update itself, and how to do this for systems which are not guixsd as I will write a system integration for this for guixsd <jmd>Where does mktemp do its business when under guix build ? <ng_>where in the guile documentation can I find info about the (and)? <ng_>is it similar to + ? <jmd>ng_: and is simply the logical AND of its arguments I think. <rekado>ng_: when in Emacs info mode, open the guile manual, then hit "i" for the index, then type "and" and hit enter. <df_>guix-daemon is running under shepherd <df_>which actually leads me to the question, what/where is the configuration that shepherd uses to run guix-daemon? <jmd>Aagghhh! postgres wont work unless /bin/sh exists. <ng_>rain1: could you explain what you meant with splitting up the lines when I want to substitue something (or, if possible just insert something between two existing words in a line)? <rain1>something like this might work out if you use quasiquote <ng_>i think I can work this into substitute.. i hope <ng_>stripping comments in nodejs can give you a performance boost of +50% ? wtaf. <rain1>Do guix package definitions have a way to check a gpg .asc? <rekado>I don't know if it does. I guess it would be nice to have signature checking built-in. <jmd>Is it possible to write to /bin during the check phase? <ng_>say I download the package, check the signature, calculate the hash. then the same thing happening on the build systems again would make sense, to recheck their download <rekado>you can write to $prefix/bin though. <jmd>rekado: Won't help me. pg wants '/bin/sh' <rekado>do you have a half-working recipe you would like to share? <jmd>rekado: What would I patch it to? <rekado>e.g. the result of (which "bash") <jmd>But (which "bash") changes depending on what has been installed. <rekado>it depends on what "bash" is at build time. <rekado>bah, that's annoying: "jar" will gladly keep the timestamps of classes as they are but will set the timestamp of a manifest (even if it already exists) to the current time. <rain1>does anyone elses terminal get really screwy if you resize it? <ng_>urxvt does that in some case. <rekado>it seems that the "internal size" remains the same <rekado>whenever I have a long command it wraps around and overwrites the prompt. <rekado>I thought it's something to do with my prompt settings <rain1>im just using default settings <rekado>I don't know why this happens or how to fix it. <rain1>If I find a fix i'll mention it <ng_>use tmux, it fixes what you can't see. <rekado>I don't see this problem in Emacs shell-mode. <rain1>how can I dig through my list of gc roots and find things to remove? <rain1>I figured out the resize issue, there is no fix though <rain1>can a package 'cherry pick' from git repo <rain1>I might just wait for it to get into master <davexunit>rain1: you can add patches using the 'patches' field <rain1>oh of course! why didn't I think of that <rain1>absolutely bizarre patronizing writing :P <rain1>"Okay developers, time to have a serious talk" - hindsight is 2020 <rain1>it's crazy people are pushing back against abstraction after this, not the real problem: the trust heirearchy that allowed npm to forcibly remove/replace one package with another <davexunit>well, you don't honestly think that 1 function per module is a good ratio, do you? <davexunit>but then there's the "bundle everything" crowd that has become very vocal after all this, and that's equally insane in the opposite direction. <rain1>let me pull up a relevant link <davexunit>if I saw a function like left-pad for guile and I wanted in my guile project, I'd just snarf it. <rekado>one function per module seems a little excessive, but I agree with rain1 that the problem here is more about npm (and the insane dependency graphs) <davexunit>the insane dependency graphs are partly caused by the 1:1 function to module ratio <davexunit>it's like having SRFI-1 broken into one SRFI per procedure <davexunit>but this is *exactly* what the popular lodash library does <rain1>it would be annoying to important every single function you wanted from srfi-1 <rain1>but you could just make a metapackage that did that and re-exported them <rain1>there's no good reason to do it <rain1>i don't see any problems with it though <rekado>it makes dependency management difficult <davexunit>we need to write *hundreds* of packages to get even the most basic node software to work <rekado>there's a threshold at which the complexity of dependency management outweigh the benefits of code reuse. <ng_>reading about javascript package management 2016 makes me less willing to learn and work with javascript <rekado>I really don't like to work with JavaScript. <davexunit>a lot of people are defending the use of left-pad because "it's another place where bugs can happen" if you were to just copy it into your source tree. <davexunit>but there are other packages like "is-positive-integer" ***Kyle is now known as Guest28935
<rain1>those are obviously silly :P <davexunit>I think the node folks have the rekado's threshold set very low <davexunit>to the point where a single line of javascript is worth abstracting into a module <davexunit>imagine if linux had a separate repo for each of its drivers <rekado>I'd like to have some more modules for Guile, though. <civodul>what a waste if everyone kept rewriting isPositiveInteger! <rekado>I feel bad about ripping stuff out of Guix (like some of the monad stuff) <ng_>let#s fork linux and split up each function into a guile module <civodul>rekado: i agree that it sucks big deal <civodul>currently my strategy is to move everything into Guix, but that's very selfish ;-) <davexunit>I think the monads module could be incorporated into Guile itself <ng_>rekado: well i'm not excited to learn it, but there's jobs in javascript so i might just for that reason <davexunit>rekado: but the general monad framework would be the same <rekado>I once started a monads library for Guile but then stopped as I didn't really have any use for it. <civodul>rekado: i think Alex S. extracted (guix monads) into something more generic, adding monads on the way <rekado>ng_: I was once really desperate to find a job (back in China) that I would have even accepted Java work. <civodul>that's definitely the right thing to do long-term <rekado>ng_: but now I think it's better not to work in IT when the only options are unethical or frustrating. <ng_>rekado: but maybe I will get employed by my partner if she goes into academic work after finishing master, and I'll just do everything around the house.. so I can halftime work on free software <ng_>another one of my many options.. all other options are years away in planing <civodul>rekado: i'm on the "frustrating" side right now, i dream of better days <rekado>I'm glad I'm not paid to be a software developer. <ng_>bootstrapping a business out of nothing but ideas is another option, but this will take some years^^ <jmd>rekado: What I you paid to do? <rekado>I had most fun when I wasn't working for somebody else. And I was delightfully unproductive. (In your face, capitalism!) Too bad paying bills gets increasingly harder that way. <rekado>jmd: I'm not really sure, actually. <ng_>I like what I do now. the only sad part is living at the lower edge of income through the situation I am currently in. <ng_>ACTION afk for some hours <davexunit>I would like to just work on guix full-time ;) <iyzsong>I get 504 from our hydra. when will 0.9.1 out? <davexunit>being a "professional" web developer for some number of years now has been helpful in identifying the things I want to fix in the software world. <civodul>iyzsong: this will a long week-end here, so that'd work well for me, maybe Monday or so <fuzzyhorns>i constantly get 504 from hydra, i gave up and used mirror <civodul>yes, it's terrible, but nully just explained me what happened <civodul>basically, we can hope for a return to normal sometime, but it's unclear how long it'll take <mark_weaver>there was also a proxy misconfiguration that made things a lot worse. <rekado>"report faked system time to programs without having to change the system-wide time" <davexunit>"In order to prevent such a terrible tragedy from occurring ever again during <davexunit>our lifetimes, `left-pad.io` has been created to provide all the functionality <davexunit>of `left-pad` AND the overhead of a TLS handshake and an HTTP request." <iyzsong>I also run a nginx hydra proxy for myself, but failed to serve (substitutes will fail) at subdir like 'my.host/hydra'. have to use location '/' with another port... <civodul>iyzsong: could you report a bug regarding the inability to use a subdirectory in the URL? <iyzsong>civodul: ok, I will remake the bug. but maybe my nginx configuration is wrong. <civodul>ACTION considers setting up string-append.io <davexunit>lol new "Docker for Mac and Windows" beta requires signing an NDA, will likely be proprietary software. <rain1>isn't it just a lxc install script <davexunit>it's a common misconception that docker uses lxc <davexunit>they have their own container implementation <davexunit>with optional support for using a different backend like lxc <df_>I believe they originally used lxc <davexunit>much like we have call-with-container, which doesn't use lxc <davexunit>originally, maybe, but I've read their Go code for their own implementation <rain1>I see, is there any reason why they did it <df_>is that libcontainer? <davexunit>because lxc has a different focus than they do? <davexunit>I deliberately chose to not use lxc either, because it's better to have everything written in Guile, down to the syscalls. <rain1>can we make/use a container for arch linux in guix? <rekado>rain1: it's a matter of finding the right package versions and binding them to expected locations in the container. <rekado>it's possible but the command would be rather long. <rekado>would be nice to have some simple declarative expression we could use to specify the properties of a container. <rekado>(not just packages but bind mounts as well) <rain1>I packaged lxc but I coudln't get debootstrap or pacman done <davexunit>rain1: the containers can be used for whatever you'd like <str1ngs>I would just use pacman to create the container <davexunit>you specify the mounts as the first argument <davexunit>alright we're all talking about a bunch of different things here <davexunit>I suspect y'all are talking about "disk images" <iyzsong>civodul: it work now, was my nginx configuration problem :-) <rekado>davexunit: I just mean a container :) <davexunit>I think both str1ngs and rain1 are confusing the two <davexunit>makes it difficult to talk about when we're not on the same page <rekado>what I meant is creating a fake debian by binding packages to /bin, /usr and /lib inside the container <davexunit>if we packaged debootstrap you could make a debian system for real <rain1>that's difficult, i had problems when i tried <rain1>str1ngs, how do you go without it? <str1ngs>just use debootstrap but create the rootfs of the container <rekado>thanks to roelj I found a couple of problems with the ant-build-system and fixed them locally. Will send a patch soon. <str1ngs>that would be hashed as a drv correct? <rain1>but you could just put somewhere in home <str1ngs>no I would tarball rootfs, created by pacman -r and put it in the store <str1ngs>but I know nothing about how guix does containers <str1ngs>now you have whole rootfs that is hashed and deterministic <str1ngs>you could start with a busybox rootfs <rekado>so you want the whole rootfs as a big blob in the store? <rekado>I thought we could do this on the package-level and add individual packages to the container instead, bind-mounting them from the store. <str1ngs>would that be a huge performance issue? I assume blobs use blocks? <rain1>both are good we just need to try it <str1ngs>I dont understand the store enough yet, either. <davexunit>rekado: should be easy to just mount a profile at / <str1ngs>davexunit how is that contained though? <str1ngs>I was taking an hour just to update substitutes :( <rekado>davexunit: yeah. Just need to be sure that the target locations are as expected. <str1ngs>ahh you mean mount the profile in the containers / <civodul>str1ngs: yeah we've all been suffering :-/ <bavier>I noticed a strange thing earlier, doing `guix environment --substitute-urls=http://mirror.hydra.gnu.org guix` downloads/builds packages, then a later `guix environment guix` will try to download/build the same packages. Is that expected? <str1ngs>mirror.hydra.gnu.org is this an example or an actual mirror? <str1ngs>if it is an mirror is it a fast one? <rekado>civodul: your efforts to keep it all together are much appreciated! <davexunit>depending upon if the cache has the substitute in question <str1ngs>you know /gnu/store is alot like ipfs. <str1ngs>add the DHT features of ipfs. and you dont need mirrors <davexunit>ipfs is orthogonal to the trust we're talking about <civodul>str1ngs: that sentence does not make much sense to me :-) <davexunit>reproducible builds are a necessary feature in order for distributed binary distribution to work <str1ngs>davexunit: yes orthogonal is the right word <davexunit>because peers must have consensus on what a build product is <davexunit>it might be an answer, but there are other systems that could be used, but it doesn't solve all of our problems inherently. <davexunit>distribution is probably the least of our worries <anthk_>any way to setup DPI on Guix? not Guixsd <str1ngs>guix system init ~/config.scm /mnt borks . does --no-grub imply no kernel as well? <str1ngs>or the grub setup handles the kernel in the store? <str1ngs>linux /gnu/store/x90f821cs9qq821lizvym2ib6wj6ddjm-linux-libre-4.5/bzImage --root=root --system=/gnu/store/7qrr36yz7c1swml0xq57b68bhpa8y7qm-system --load=/gnu/store/7qrr36yz7c1swml0xq57b68bhpa8y7qm-system/boot <str1ngs>ok, so I can use my system grub to boot the usb drive. see how this goes :) <str1ngs>I just thought it weird /boot was empty <mark_weaver>str1ngs: and initrd. they are symlinked from the created 'system', so even 'guix system build' makes them. <str1ngs>what happens with system where boot is on another partition? <mark_weaver>the kernels and initrds are in /gnu/store, referenced from grub.cfg <str1ngs>I'm doing a from another system install because of wifi issues <str1ngs>sorry I'm not installing from the usb installer <str1ngs>I'm using another host system. but I think I have it sorted out. <str1ngs>I can install grub by and and use the generated grub.cfg <mark_weaver>sure, iirc, grub.cfg is installed even with --no-grub <civodul>ACTION successfully upgraded his profile minus texlive/qt things \\o/ <mark_weaver>the only difference is that GRUB itself is not installed <mark_weaver>civodul: I had started to, but then I realized that I should hold back until I've finished testing my improved wip-faster-grafting code <civodul>mark_weaver: there's currently no grafts though <mark_weaver>that's why I want to hold back. I'm currently running with grafts, before the security-updates merge. ***nckx|offline is now known as nckx
<str1ngs>I think I fixed the grub issue. (device "/dev/sdb1") should point to the root fs <mark_weaver>str1ngs: in the 'grub-configuration' form, 'device' should normally be something like "/dev/sdb", (without the "1") <mark_weaver>but if you pass --no-grub, I guess that field might not be used for anything <str1ngs>mark_weaver: the root label needs to be set as well <str1ngs>source_dir doesn't exist. Please specify --target or --directory <mark_weaver>if the 'file-system' form has (title 'label), then the string in the 'device' should be a label, and the label needs to be set on that partition <mark_weaver>but the 'device' in 'grub-configuration' is a different story <str1ngs>I'll see if I can install it by hand <str1ngs>Installation finished. No error reported. <str1ngs>yay it works. now I have to setup wifi :) <rain1>wondering if this segfaults: mkdir -p A/B ; mpv A/* <lfam>It will trigger ~200 rebuilds <lfam>Many are trivial but some are not, such as libreoffice <rekado>lfam: I don't expect much to break with that update (as it only affects tests). But it would take hydra some time to rebuild 200 packages on each supported platform. <rekado>hydra is already a little behind <rekado>due to the recent I/O problems with the machine <lfam>Exactly, that's why I am thinking it should go to a staging branch and be merged once it's built. But I'm a little impatient because I am biased re: letsencrypt ;) <davexunit>ACTION really wants a shepherd interface for emacs <civodul>ACTION hopes our Emacs Hero has seen this suggestion <davexunit>and now I wish I had the same for managing the daemons <davexunit>like just having a list of all registered services and their state would be nice <davexunit>and then maybe you could mark them and apply an operation or something <davexunit>mark with 'm', press 'x' and be prompted with the action to send <civodul>yeah pretty much like in guix-installed-packages <civodul>i think our Emacs Hero has already most of the code, in fact <rain1>i have to do autogen.sh after patch shebangs or I get an error about bad interpreter /bin/sh <rain1>but then it happens on ./configure <rain1>should i insert patch-source-shebangs a second time? <rekado>maybe you don't need to run autogen at all. <rain1>there is no configure file without it <rekado>we're passing CONFIG_SHELL to the configure script by default <rekado>you should see that in the build logs <str1ngs>I guessed as much. but there are still going to be some scripts that are to autotools based <str1ngs>in which case there is a common way to add those to be patched <str1ngs>and things like python. mind you /usr/bin/env is all that is getting patched? <str1ngs>is there a way to add a package to a root directory. ie /mnt <rekado>I don't understand the question. <rekado>the shebang patching depends on what packages are among the inputs <rekado>if there's a python shebang and the package has "python-wrapper" or "python2" among its inputs it will patch it out. <str1ngs>can guix install a package to /mnt after or during init? <str1ngs>how do I add a package either in the config.scm or using guix <davexunit>or boot into the system and reconfigure from there <str1ngs>I need to add wifi. so I can pull packages <davexunit>but if you have no network access, then you're a bit stuck. <davexunit>you'll need to boot a system that you can get on the network with <str1ngs>I cheated sudo chroot /mnt/ /gnu/store/7cdd8s466qyjh64m0byq0rz9gk1jid40-bash-4.3.42/bin/bash --login <str1ngs>going to have issues with the daemon this way :( <rekado>I don't understand what you are trying to do. <rekado>I don't know what you mean by "add a package to /mnt" <str1ngs>I need to install packages into /mnt <str1ngs>I've just chrooted and started a daemon its working <rekado>do you mean that the store is in /mnt/gnu/store? <davexunit>you can only install packages into the store, nowhere else. <rekado>or do you want the profile to be located on /mnt? <str1ngs>I'm pretty sure all of FOSS knows what package means <str1ngs>btw the guix terminology not very user friendly <rain1>to use guix edit do I just symlink something to my git checkout? <rain1>is there instructions on how to do this? <davexunit>this way your development code will be on the load path <davexunit>we use the source information stored in guile objects to find the file and line <rain1>./pre-inst-env: line 71: /gnu/store/2qx3kn7l07xlfcfdgw0whp1by5b9z3cq-profile/bin/guile: No such file or directory <rain1>trying in: guix environment guix --ad-hoc guile <rain1>just for doing 'make' before trying that pre-inst command <rain1>oh i probably have to reconfigure <str1ngs>I'm going to do this install again. but I'm pretty confident these instructions wont work without guix pull <str1ngs>so that should be updated or bug report <lfam>str1ngs: I recommend you read the latest version of the manual (from our Git repo). It includes that change as well as many others. <lfam>It's unfortunate and confusing that the online manual does not get updated quickly <lfam>I think we should host both the "release" version and the current version on the web somewhere <lfam>The latest version of the installer also recommends the user run `guix pull` in the MOTD <str1ngs>since the substitutions are GC'd already <lfam>So, I'm sorry you are having the annoying issue, but hopefully new users won't have the issue once we issue our next releaser <str1ngs>I'm not bothered by it just letting you know :) <lfam>We also plan to keep substitutes around for longer in the future, but currently we simply don't have the storage capacity <lfam>We recognize this is a problem for new users (and all users, really) <rekado>actually, I think that we still keep 0.9.0 substitutes on hydra <rekado>just not sure if all of them are there. <rekado>in any case it's best to use the latest. <lfam>rekado: Right, the "dead" ones are gone I think <rain1>is git diff the correct way to make a patch? <rain1>do i need to pass in extra flags? <lfam>rain1: git-format-patch is preferred <str1ngs>so maybe not an issue if these change is in master <rekado>formats a patch for the last commit you made. <str1ngs>lfam: also the terminology is really kinda hard to follow. <str1ngs>can there not be a better name the substitutions and derivatives? <lfam>str1ngs: I don't think end users need to know about derivations. As for 'substitutes', I think it is acceptable once a user understands that our system transparently provides substitutes for building from source <lfam>You are "substituting" a binary for a source build <lfam>Well, maybe you can find an abbreviate if you hunt. But if you read our code you will find complete words and sentences. <str1ngs>that's fine I know there is a distinction. I'm just thinking of the masses here <lfam>I believe this decision was made early on, and I really like it. It makes it easier for English speakers as well as speakers of other languages <rain1>can I make that into one patch <lfam>rain1: Do it on another branch, and then do `git format-patch master` <lfam>That's my advice anyways <davexunit>where we prefer descriptive, albeit longer, identifiers as opposed to abbreviations that may be hard to understand <rekado>rain1: if you want to merge them, then merge the commits. <rain1>oh okay, ill read up on how to do that <rain1>Runtime cpu detection is not implemented on MIPS, so we disable some features. <rain1>these flags no longer exist in new ffmpeg <lfam>Is anybody else noticing slow delivery via the mailing list? <rain1>yes it was very slow a couple days ago <rain1>so slow I thought it must be sommething else <lfam>rain1: You can use `git log gnu/packages/video.scm` and `git blame gnu/packages/video.scm` <rain1>I want to update this but line 501 has to be commented out <lfam>rain1: Can you test the change and send a patch to guix-devel@gnu.org? <rain1>yes i just thought it would break things on MIPS, but with this rename I think it's an acceptable patch <lfam>Well, we use the mailing list for peer review, so I bet that mark_weaver will want to test your patch. <lfam>It's good to include a little information about the changes. For example, that you think it might break things on MIPS <rain1>I don't think it will break anymore but I will add a note <rain1>i'll also test if it fixes a segfault in mpv ***woodrad_ is now known as woodrad
<paroneayea>I need to figure out what's requiring texlive-texmf in this profile <lfam>paroneayea: Can you try with --no-substitutes? <lfam>Well, you'll only want to build texlive-texmf without substitutes <paroneayea>maybe texlive-minimal will still work with numpy...? <lfam>paroneayea: You could try something like this to get texlive-texmf in your store: <lfam>guix build -e '(@@ (gnu packages texlive) texlive-texmf)' --fallback <lfam>Well, I guess it would also work to add --fallback to whatever operation you were doing originally <lfam>I have no idea why this commit has this old email address on it: fae58fe1d11ad33c1465f0f5ab5898f98f92af88 <mik_>to change my shell to zsh. But my terminal is messed up after rebooting (e.g. can't use ctr+L to clear), and certain programs (e.g. su) are missing from my path now <mik_>is there anything more that needs to be done to switch my shell to zsh? <lfam>mik_: Yes, there is more to be done to make the shell work as expected <lfam>Basically, all the shell initialization and configuration files on GuixSD are using the names expected by Bash. For example, .bashrc, .bash_profile, etc <lfam>So, you will need to adapt them to the names expected by Zsh <lfam>I'm not sure exactly what you will need to do though <janneke>paroneayea: how did you "found the culprit" <paroneayea>janneke: looking at "guix graph" output for python-numpy <davexunit>would be cool to use 'guix graph' on entire profiles <paroneayea>I'm now trying to switch texlive -> texlive-minimal on a bunch of python packages <paroneayea>no one should suffer a full texlive-texmf unless they really want it :) <paroneayea>davexunit: my part is easy, andreas did the hard part :) <paroneayea>also while I was waiting for stuff to build I decided to open up another sicp video and watch it since I was having trouble focusing on anything else <davexunit>putting these packages on a diet is a good idea <paroneayea>but it had Hal wearing a giant dunce-shaped wizard hat <paroneayea>I don't think I want someone to ask me any questions today <paroneayea>ACTION got through the first register machine lecture the other night... very illuminating! <paroneayea>right now I have sicp info manual open and sicp videos queued with a copy of sicp on my desk <davexunit>it really is amazing just how jam packed it is with useful information <paroneayea>I'm so excited that Sussman has a new book coming out on symbolic programming <df_>ooh can I join the party? <df_>haven't got as far as register machines, I'm still on chapter 3 <mik_>lfam: thanks! I'll look into it some. Funny thing is I didn't have the problem until I rebooted <df_>(today's progress was somewhat hampered when I managed to make guile segfault on the deque exercise) <df_>had no idea there was an info version either <df_>but wow, that does look a lot nicer <df_>though now I know I can just read it in emacs... <davexunit>I read that nice web version when I'm on my phone <davexunit>it's responsive so it's pleasant to read on a small screen <df_>well reading web pages in conkeror is pretty close to being in emacs <df_>once this minor operating system problem is resolved I wouldn't mind somebody writing a guile-based web browser <davexunit>is POWER8 a processor architecture like x86? <df_>PPC was based on POWER <rain1>POWER8 is absolutely beastly <df_>rain1: in a good way? <df_>that's my understanding too <df_>it seems to be the only arch that can compete with x86 on performance, but far more freedom-friendly <anthk_>yep, cos Intel with ME doesn't look really secure to avoid data leaking <df_>yeah, aiui the entire firmware for that machine will be free software <df_>but I'm certainly eyeing either that or a libreboot d16 when I can next afford a high-end workstation <davexunit>the main question for me is graphics acceleration <davexunit>but probably I shouldn't be using this board for desktop use <df_>can't say for sure but I suspect that's still an unsolved problem <davexunit>I'm still trying to keep my eye out for a board that supports free boot firmware that can be used as a media center computer <df_>I don't have much need for fast 3d graphics personally <df_>ah right, video codecs <jmarcian`>could this system run with GuixSD out of the box? <davexunit>I'm hoping that AMD's GPUOpen project will eventually lead to AMD graphics cards having fully free drivers <davexunit>jmarcian`: we would need to port Guix to the POWER8 platform <davexunit>and we'd need some of these for the build farm <df_>hasn't there been some success with drivers for the decoding hardware on the MALI? <jmarcian`>with first large bunch of money, I would buy that one for project <df_>anyway, there are server-format openpower machines that might be more suitable for a build farm than the talos <jmarcian`>in relation to guix, how can I truly delete package, so that on next install, it is downloaded again? <lfam>jmarcian`: You use `guix gc -d /gnu/store/...-package-version`, where ... is the hash <lfam>But, you shouldn't need to do that <lfam>If you are using the previously downloaded package, it's because you are asking for the same package <jmarcian`>lfam: my emacs compiled wrongly, and I could not do anything to get graphics version. <lfam>jmarcian`: What do you mean by "wrongly"? Do you mean that the files are corrupt on disk, or that the compilation didn't work as expected? <lfam>Did you get an error message? <jmarcian`>right now all packages are good, working, also emacs, without X. But when it was compiling, there was some error <jmarcian`>error was in gdk-pixbuf if I remember well, so I got emacs without icons, only one icon was visible <lfam>jmarcian`: The compilation of gdk-pixbuf failed? Can you share the error message? <lfam>But, if the compilation of one of Emacs dependencies failed. then the whole thing would have failed, and you wouldn't the the GUI Emacs at all <jmarcian`>yes I guess I used something to override that <lfam>Did you use --keep-going? <jmarcian`>as otherwise gdk-pixbuf would not compile on my side <lfam>Hm, I'm not sure exactly what to do. Maybe somebody else knows the best course of action? <lfam>What I would do (not knowing the right solution) would be to try again with --fallback <jmarcian`>fallback is to compile, it is ALWAY compiling on my side. <jmarcian`>I could never get substitute for emacs, gdk-pixbuf <lfam>jmarcian`: Can you get substitutes for anything? <jmarcian`>basically, when I used hydra.gnu.org, no substitutes, normally compiling. Later I used mirror.guixsd.org, and it was better, "with substitutes". <jmarcian`>Is there a way that I can setup my VPS, to receive substitutes, that I can use it myself? <lfam>You can run your own hydra (I have no idea how to set that up), run a mirror that pulls from hydra, and you can also use `guix publish`. <lfam>I just did `guix build emac` and everything was substituted <lfam>I did `guix pull` in the last day or two, and I also made sure to do it as root so that I could update the daemon <jmarcian`>what do you update with guix pull? I did it so many times <lfam>I did it for my own user and for root. After doing root, I made sure to update to the latest guix package and then I restarted the daemon <lfam>`guix package` is the command line tool that allows Guix users to install, remove, and upgrade packages, and do other package related things <jmarcian`>I have latest version of guix package manager for many days, and it is 0.9.0, I can make guix pull, it will stay 0.9.0 <lfam>jmarcian`: As root, does `guix package -I` include the guix package? <jmarcian`>guix 0.9.0 out /gnu/store/632msbms2yaldfnlrb5lbnlnmn9yjisw-guix-0.9.0 <lfam>Right, and that is a symlink to the guix in root's profile right? <jmarcian`>and as user, I don't have it, and cannot get it <jmarcian`>yes, symlink to: /var/guix/profiles/per-user/root/guix-profile/bin/guix <jmarcian`>is it wrong that I as user, never can get guix ? <lfam>Okay, root needs to update their guix package <jmarcian`>OK doing it. Like, unpacking '/gnu/store/kwkmwaji3grigwd052psbj46sahzp1dr-guix-latest.tar.gz' <lfam>Is that from `guix pull`? <jmarcian`>now I wait for: loading... 14.2% of 487 files <lfam>Well, it's typical, at least. I doubt it's expected <lfam>Yes, it used to be much slower :) <jmarcian`>and as user, I did "guix package -i guix" -- now i get again emacs recompiling... <lfam>After root's `guix pull`, you will need to do as root `guix package -u guix` <lfam>Well, that shouldn't affect what your user can get, but it will make the preferred mirror the default and give you https substitutes <lfam>`guix package -h` shows that it is 'upgrade' <jmarcian`>pull is updating the guix software (package manager) <df_>so this one confuses me, I don't think my root user has the guix package in its profile <df_>but obviously it can run the guix command <df_>what's the difference? <lfam>df_: Do $(realpath $(which guix)) <lfam>It's coming from somewhere, right? ;) <df_>/run/current-system/profile/bin/guix <lfam>jmarcian`: When I do `guix build emacs`, the result is /gnu/store/1mnrkhqi5fah3g6nb9d2m7akdypk7d3k-emacs-24.5 <lfam>What path are you trying to build? The same one? <lfam>df_: Oh, I guess it's different on GuixSD vs Guix on another system <jmarcian`>I cannot know which one, I don't have yours. Am I expected to have same like yours? <jmarcian`>as soon as I get all packages I need, I am switching to GuixSD <lfam>What packages do you need? <jmarcian`>right now, I cannot mess with emails sending, 5000 per day sometimes <jmarcian`>I am using courier-mta and it is not yet in GuixS <jmarcian`>courier-mta is great SMTP server, with webmail and simple web config with /etc config too. <lfam>It look slike it's written in C and C++. Shouldn't be too hard to package. Would be awesome to have a service for it too <lfam>jmarcian`: On x86_64, with current `guix pull`, the Emacs I substitute is the one mentioned above. <jmarcian`>lfam: when it is packaged, do configuration directories remain in /etc/courier for example? Or are they "user based"? <jmarcian`>I will see with emacs when it finished compiling <lfam>If courier-mta can do "user-based" then you can do it on Guix and GuixSD too <jmarcian`>I wish to know: if courier by default, installs config files in /etc/courier, is Guix generally placing that in that location? <jmarcian`>or guix is moving config files somewhere else? <anthk_>but I miss lazarus IDE in guix, but I'm fine, overall I have an updated gcc :) <lfam>You can put files in /etc yourself, or you can use a shepherd service and Guix will do some magic to put them there for you <lfam>But you'd have to write the shepherd service <jmarcian`>lfam: with shepherd, it would be what, system or user based? <jmarcian`>I understood, there is no need to install any package as root <jmarcian`>which is making me confusion in regards to such system services that require root privileges, special permissions, default locations for configurations... <lfam>On GuixSD installing as root is not the same as installing from the system configuration <lfam>My knowledge in this area is pretty weak. You should try reading some of the services that already exist and have those issues, like nginx <anthk_>lfam a question. Having Guix on a distro, 'guix package -u updates all user packages to the latest question? <jmarcian`>updated GNU Guix successfully deployed under `/root/.config/guix/latest' -- now -u? <lfam>If you want to upgrade all your packages, then yes <lfam>anthk_: I don't understand, can you ask again? <anthk_>lfam I mean, how can I update all guix packages to the latest version? <lfam>anthk_: `guix pull && guix package -u .` <lfam>You need to do it for each user <jmarcian`>I see now your hash showing up on my screen... maybe I get it <anthk_>also, as I installed guix not with guixsd , what's the correct way of upgrade the guix binary distribution when a new release (I.E, 1.0) arrives? <lfam>davexunit: What about, as root, `guix package -u guix`? Is that relevant? <lfam>For getting updates to the daemon? <davexunit>lfam: yes, assuming that you run the daemon via the root user's profile <lfam>I found that I had to do that in order to use https substitutes <lfam>Which was expected, per the commit log <anthk_>davexunit, updating guix globally <davexunit>each user can run their own version of guix, or multiple versions <davexunit>anthk_: that depends on how you start the daemon <anthk_>~root/.guix-profile/bin/guix-daemon --build-users-group=guixbuild <davexunit>so then install a newer guix into root's profile <df_>davexunit: this is quite a confusing issue <df_>guix pull is roughly equivalent to apt-get update, yes? <df_>but it does also update your guix and guix-daemon binaries? <davexunit>it updates the guix client for the user that runs 'guix pull' <df_>what updates the daemon then? <lfam>`guix pull` updates your package definitions and the code that make up the guix package manager. `guix package -u` upgrades the software you install with Guix, including the guix-daemon. <davexunit>the daemon pretty much *never* needs updating because it's protocol is very stable <davexunit>df_: on GuixSD this is handled by 'guix system' <davexunit>if you're using a foreign distro, then it's up to you <davexunit>lfam: *only* if you installed the guix package into your profile <lfam>Right, I'm speaking about Guix on a foreign distro <df_>so is there any reason to run guix package -i guix? <davexunit>on a foreign distro it makes sense if you consider the "global" guix installation to be in root's profile <davexunit>jmarcian`: right now I'm at work on an ubuntu machine that has guix installed <davexunit>but I have a laptop and desktop computer that run GuixSD <jmarcian`>builder for `/gnu/store/249dmav1zmidd64qjb2qs700m7h726m1-guix-0.9.0.71e2065.drv' failed with exit code 1 <jmarcian`>Makefile:4747: recipe for target 'check-local' failed <jmarcian`>it says "Look for 'length' in the 'config.log' file for details" .. but building directory is gone. <jmarcian`>and what is this: "Cannot run tests because file name limits would be exceeded" <civodul>jmarcian`: it's saying that $TMPDIR is too long, presumably <jmarcian`>anyway, I don't need to ask more, now I know how tu run emacs with buttons (which I never use) <civodul>jmarcian`: in the build log above, did you see "/tmp-guix/guix-build-.." or "/tmp/guix-build-..."? <jmarcian`>Wrote /tmp-guix/nix-build-guix-0.9.0.71e2065.drv-0/source/emacs/guix-autoloads.el <jmarcian`>don't know what that with "emacs" appears, but now I have log at least. <civodul>ok so you're running an "old" guix-daemon <sapienTech>hi all, I am submitting my application for gsoc 2016 and it is asking me to submit a draft through gdocs or some other sharing service. Does gnu recommend a particular sharing service or should I just submit my final draft? <civodul>with a newish one you'd get "/tmp/guix-build-...", regardless of the actual $TPMDIR <civodul>sapienTech: we recommend not using Google Doc, which is SaaSS, but posting a file somewhere is fine <civodul>where "somewhere" means something that does not require non-free JavaScript <civodul>anyway, post it to the mailing list, it's archived :-) <sapienTech>civodul: nice idea, yeah i was peeved when the website was directing people to using their proprietary service <jmarcian`>civodul: maybe I shall do what? restart daemon? I recently rebooted... (today) <jmarcian`>even though i was doing pull as root several times, my boy rebooted my laptop, and is still "oldish" <civodul>oh it's you? you changed nicknames no? :-) <jmarcian`>maybe it is me, I don't know if it evaluates correctly <ng_>guix package -i guix? <ng_>don't you normaly have guix in your profile when you symlink it on a foreign distro? <jmarcian`>I have restarted daemon, now I see: Wrote /tmp/guix-build-guix-0.9.0.71e2065.drv-0/source/emacs/guix-autoloads.el <jmarcian`>yes I see now it takes longer time, and I see many PASS tests/*.scm <jmarcian`>I wish to use again desktop computer, fast, not "laptops", better be named "lapheats" <ng_>twitter as a package manager. now packages with ads <mik_>I'm having a bit of trouble running the configure script. I'm stuck at this error: ld: cannot find crt1.o: no such file or directory <mik_>I do have LIBRARY_PATH variable specified, and there does seem to be crt1.o in that directory <jmarciano>now after restart of guix daemon, I got guix in a profile... <jmarciano>ooo, emacs integration makes everything much much easier <jmarciano>this package is obsolete, what it means, old? <mik_>interesting... in order to specify the path to that library file, I had to precede the command providing the path, like so LIBRARY_PATH=/root/.guix-profile/lib ./configure <dannym>jmarciano: Yes, usually was replaced by something better and it's basically just there for backward compat. <jmarciano>it was in my emacs buffer of installed packages <jmarciano>through emacs I have quick access to Guix packages now, very nice. But I guess end users may wish different GUI <koz_>jmarciano: You can reskin Emacs. <koz_>I use a theme called Bubbleberry for example. <jmarciano>I got the theme, but forgot how to activate? <koz_>jmarciano: Bubbleberry is a *dark* theme. <ng_>woo. this netbook is so broken and old, the only distro detecting everything is tails. free hardware wifi connected? debian goes like noope. <koz_>ng_: Even a recent Debian? Surprising. <koz_>Although I have had this happen before. <ng_>ethernet? nope. something with the killswitch acting funny. <kmicu>“#guix user peak: 142 (1h 11m ago) average: 121.07 users (30d period)” — it’s growing nicely ヽ(*^▽^)/ <koz_>I feel sorry for that 0.07 of a user. <koz_>Is Thing on here or something? <koz_>Because 0.07 of a user seems like it's one hand or something. <koz_>Nah, that's too big to be a hand. <lfam>paroneayea: I disagree with that article's take on package signing (and the article it links to, as well). Even if signatures are TOFU, it's better than nothing. <lfam>At least when distributing source code <lfam>I packaged flask recently and noticed it wasn't signed. It seems like a major project <anthk_>ng_, cyberpunk? that reminds me of a MUD I played long ago