<fps>i still think setting up a torrent tracker just for guix and run a client on every guix install to share substitutes would be a quicker way to easing the load of distribution and building substitutes <fps>[quicker than waiting for a solution like gnunet emerge as being useful] <davexunit>civodul: thanks! not bad for a quick hack ;) <civodul>i have +/- converted the news entries we had to use Haunt <zilti>Could it be that there's no Guix support for the "Intel Wireless 7260" chip? <davexunit>zilti: intel wireless chips require nonfree firmware which linux-libre, the kernel GuixSD uses, removes. <zilti>davexunit: ...oh. I didn't know that. Hmm... <thomasd>inside a build script, is there a variable that gives us the current build directory? I'd need it in a (subsitute* ...) expression <mbuf>is there a deployment tool used with guix? <mbuf>for provisioning and configuring guix systems? <thomasd>mbuf: not yet, I think. People are currently working on that, though. <lfam>thomasd: You can use (getcwd) <thomasd>lfam: thanks! Maybe I need another cup of coffee :) <thomasd>mbuf: you can search the irc and mailing list archives for "guix ops" to find the discussions <janneke>thomas...you're sure you need it, if it's cwd? <thomasd>janneke: it's a makefile which adds build-dir/lib to the RPATH <thomasd>so I want to remove the -Wl,-rpath,... using substitute* <thomasd>(I don't understand why this is done in the first place, probably a mistake/sloppyness?) <thomasd>Though it's the second package where this happens. <efraim>sneek: later tell ng0: I tried the psyclpc source again and it magically worked, first try <rekado>woo, my changes to the xwidget feature are now in Emacs master :) <rekado>ACTION just pushed 49 commits to update bioconductor packages <marusich_>Somehow, I seem to have joined the channel twice? <thomasd>are any more updates to the kde frameworks packages planned? <civodul>thomasd: package updates, surely :-) <civodul>or did you mean changes to the organization of that file? <efraim>did we get all of the kde-framework patches? <efraim>or were there more patchsets/levels <civodul>efraim: surely you were the one who kept track of it, weren't you? ;-) <thomasd>I'm trying to package kdevelop, but the rabbit hole runs deep <thomasd>for example,there's something called ksysguard, which is missing <efraim>I think I was the one that kept on insisting on modular qt in the patches <thomasd>I was hoping for a resident Qt/KDE expert :) <efraim>i'm good at banging my head against the wall at least :) <thomasd>the modular thing sometimes makes it difficult <thomasd>for example “libksysguard” wants to install something in the “kauth” store directory, which is not allowed <mbuf>virt-manager is not yet available in guix? <mbuf>is anyone using guix on bare metal? <civodul>mbuf: several of us use GuixSD (the complete OS) on the bare metal, sure <mbuf>and does linux-libre include grsecurity/pax patches? <civodul>struggling with Qt and getting crazy? :-) <thomasd>civodul: seeing how everybody here uses erc (?), I'm surprised this doesn't happen more often. <civodul>thomasd: you think ERC is more represented here than on an average channel? :-) <thomasd>for reproducibility, can I keep references to the build dir in the package output? <civodul>thomasd: that's OK because the build directory name is always the same <thomasd>ACTION needs to learn to use paredit-mode <Petter>civodul: Do you have the commands to find out which KMS module and Xorg driver is in use at your fingertips? <Petter>If not, I'll just keep searching. <civodul>Petter: for Xorg, i just check /var/log/Xorg.0.log <Petter>Does this mean KMS module is i915? <rekado>Petter: this means that the module “drm_kms_helper” has one user, namely the “i915” module. <newdan>Interesting... "We just don't know how to offer all the features distros have in a functional matter." Bit of a Guix/Nix noob, any idea what specifically that might be referring to? <thomasd>I was just going to ask the same thing <rekado>Petter: I’m sorry, I also don’t know. “lsmod” lists kernel modules. All I know is that both “i915” and “drm_kms_helper” are kernel modules. “i915” is probably the DRI module for your Intel graphics chip. <jmi2k>Anyone here knows how to have root encrypted with LUKS, and make GRUB prompt for the pass at boot (/boot is in another partition)? <civodul>newdan: some things are more difficult in a functional setting, such as fast security updates; other things have to be approached differently, such as plugins, environment variable settings, etc. <civodul>but overall i think we're doing ok on these fronts <Petter>rekado: Thanks! I'm inexperienced in this area of kernels and graphics, so any help is valuable. <rekado>Petter: the kernel trace also shows that the error is triggered in the i915 module (i.e. the graphics driver) <davexunit>we should look at how firejail sandboxes x11 and copy it <davexunit>we can easily offer all of the features firejail does, with the added benefit of being able to compose it with all of the other guix features <Petter>rekado: Ah, right. I haven't heard of i915 before, so I didn't even notice it in the kernel trace. It's just a bunch of jibberish to me :/ <Petter>Hm, looks like one can basically get the KMS module from the kernel trace then. If one knows what to look for. <Petter>(I'll be back later, have to go.) <rekado>“what happened to guix?” asks a commenter on the linked Reddit thread. <rekado>ACTION builds openssl 1.0.2j for the third time <rekado>had to run “guix gc” after building all of our CRAN and Bioconductor packages. <dvc>what's the status of guix 0.12.0? I keep thinking it's due any time now... <civodul>ACTION manually starts an openssl 1.0.2j build on hydra.gnu.org... <civodul>we need to finalize & merge core-updates first <davexunit>will we ever be able to pre-build stuff for grafts on hydra? <davexunit>ACTION is also sick of building openssl and other things <civodul>davexunit: we do, but i don't know why so many things were gc'd recently <rekado>I thought this happens when .links gets too big and then the GC runs. <rekado>maybe the out-of-space errors lead to a failure in registering GC roots? <civodul>rekado: i don't think that's what happened, not at that scale <rekado>can't wait to replace hydra with cuirass <dvc>the berlin thing is on a monday right? <dvc>if it were on a weekend maybe, but monday :/ <davexunit>"What happens is that when something doesn't fit into the metaphor, it's just stripped from the software." <dvc>davexunit: is there anymore context to this? that's just a noob IMO <rekado>davexunit: I don't know of any case in Guix where that's done. <OrangeShark>"everything has to be specified" Someone has to specify it? <dvc>that doesn't apply to nixos - he was referring to functional package managers in general <davexunit>bavier: but that's not an example of something not fitting into the functional metaphor <rekado>bavier: oh, that didn't even occur to me :) <bavier>some examples from the commenter would be nice; nothing comes to my mind <rekado>hey, I thought our pypi importer was updated to produce "pypi-uri" instead of these really long URIs. Was this not merged? <civodul>davexunit: ↑ it's about the issue i have with the pivot-root test <civodul>i'd like to know which kernel versions trigger this problem <davexunit>civodul: does the test get skipped when user namespaces are unavailable? <rekado>error on 3.10.0-327.28.3.el7.x86_64 <rekado>a.out: test.c:46: main: Unexpected error: Invalid argument. <civodul>yeah, because user namespaces aren't supported i guess <bavier>I get the same "operation not permitted" <bavier>SUSE 11 with linux 3.0.101-0.47.71 <lxo>civodul, no idea, I've never had much luck either way. reporting to a more specific list has helped sometimes, especially when I had a patch <bavier>efraim: thanks for the x265 package. that makes my handbrake work a bit easier <rekado>paroneayea: I just realised that I should have done this the other way around: first upgrade R, then CRAN packages, then Bioconductor. Oh, well… :) <rekado>next up: update R to 3.3.1 (I’m surprised the release is so long ago already.) <paroneayea>ACTION wonders why signing with magit never works... <rekado>I really don’t like to sign up to these bugzilla instances everywhere <paroneayea>rekado: if only we had working federated identity! <rekado>but I’ve been having some odd magit-related problems recently <rekado>paroneayea: isn’t that what zot/red does? <paroneayea>rekado: the main direction I'm interested in is the Linked Data Signatures + HTTP Signatures direction <rekado>some magit problems I see regularly: rebase fails because index.lock exists; sometimes signing fails because pinentry doesn’t pop up at all; very slow diff (e.g. after all these R updates the diff for bioconductor.scm took many seconds to display) <rekado>paroneayea: thanks, will take a look when I’m home <rekado>I feel a little sorry for my colleague who has to package things with conda. <rekado>I’m so much faster with Guix and have guarantees that things will work pretty much everywhere. <rekado>I’m considering to write a little Guile script to export all packages as conda recipes to help them out. <rekado>this won’t help with guarantees, of course. <bavier>civodul: I finally got around to replying to your "guix on hpc" message <bavier>I've been out sick the last few days <efraim>bavier: I kept wondering why it wasn't already in guix. first I thought it was the license, and then hard to package, and then I realized just no one got around to it before :) <rekado>bavier: good point about “guix gc”! I just never told our users about this and always advocate the use of separate profiles (not “guix environment”, not “guix build”). <Apteryx>rekado: Can a package be installed in a separate guix profile but still be "visible" from my main profile? <rekado>Apteryx: depends on what you mean by visible. <Apteryx>I'm looking at guile-emacs, which I'd like to experiment with sometimes, but which also clashes with regular Emacs if installed in the same profile. <bavier>rekado: true, 'guix environment' would help in many cases but would still be at the mercy of race conditions <rekado>Apteryx: you can “activate” a separate profile by doing this: bash; source /path/to/profile/etc/profile; do things; exit <davexunit>bavier: stuff in an environment won't be GC'd while 'guix environment' is still running <Apteryx>rekado: Well, I meant available on the PATH, but I guess that's not possible (both have the same name). Maybe a bash script called guile-emacs.sh which would launch (guile)emacs in the right profile? <davexunit>the connection to the daemon is maintained throughout the duration of the session, which protects the relevant store items <bavier>davexunit: right, I meant "race conditions" on the user end <bavier>like, build package, "oh I should have this in an environment", create environment <rekado>Apteryx: sure, you can have both on the PATH, but the directory that’s first will dominate. <Apteryx>rekado: OK :) Is it cheaper to activate a profile than to create a "guix environment"? <bavier>davexunit: are built outputs protected before the profile is fully instantiated? <rekado>Apteryx: it’s pretty much the same, except that a profile is persistent. <davexunit>the profile is never registered as a gc root <davexunit>they are implicitly protected by nature of the daemon working on them <rekado>Apteryx: the result of “guix environment” is a profile, too, but it may be removed by “guix gc” <davexunit>civodul taught me that in some feedback when I first wrote 'guix environment' <Apteryx>rekado: OK! So the 1st time I create the profile it might take as long as issuing a "guix environment" command, but later on when I just activate it it should be swift? <rekado>Apteryx: all this “activation” step does is setting environment variables. <Apteryx>rekado: Nice! Reminds me of virtualenvs in the Python world. <davexunit>yes, it's like a virtualenv for *everything* <rekado>this reminds me that I wanted to provide “guix environment --save” and “--load”. <davexunit>but lately I can never seem to implement anything <rekado>davexunit: I feel the same. I’m barely even successful at keeping my patches from bitrotting :) <davexunit>well my last patch was for a package we already had <paroneayea>which is a package I made mostly correctly nearly about 8 months ago, posted to the list with "mostly working", but thought it was borken incorrectly, and left to bitrot :) <paroneayea>I also would like "guix environment --save" "--load" <paroneayea>I guess you'd just be saving a symlink to the profile somewhere? <davexunit>I'd also like 'guix environment' with no other options to do something convenient <bavier>how would that interact with 'guix package -p'? <bavier>could you do something like 'guix environment --container --load ~/.guix-profile'? <davexunit>I'd like --load and --save to use a default profile of $PWD/.guix-environment, for convenience <davexunit>and --save can create a new generation if the profile already existed <davexunit>so that way environments can get rolled back, too <davexunit>though I think ultimately an environment is more than a profile <davexunit>many of us would like it if an environment could describe things like programs to launch when the environment is created <davexunit>for example, I would like to create ruby web development environments that automatically launch a mysql server with the correct configuration <davexunit>unfortunately, the GuixSD service API cannot be used for this. <paroneayea>davexunit: something I have been thinking about: <davexunit>bavier: for example, the mysql service creates a mysql user. <paroneayea>why can is it that we only have services for the system profile? <davexunit>not sure I understand. services aren't related to the system profile. <paroneayea>davexunit: but we don't have any way to define services for other profiles and use them meaningfully <paroneayea>and yeah I could use hcron + shepherd as my own user <paroneayea>but why not use the same service definition system? <bavier>ok, so services are currently privileged <davexunit>the system profile in /run/current-system/profile is for convenience, the services refer directly to the packages they need <paroneayea>ACTION has just been thinking about "profiles as sets of scopes" <davexunit>could anyone conceive of an API where privileged and unprivileged services can share mostly the same code? <paroneayea>I don't know, I think I haven't spent enough time with services, but aside from the ability to define users <davexunit>I'm not sure how to do it in an elegant way without lots of (if (privileged?) ... ...) <paroneayea>it would be cool if we could also "reconfigure" various profiles in the way we do for the base system <paroneayea>I'd love to play with my user profile being more declarative, and using services. <paroneayea>ok, but profile + system seems similar to profile + environment, no? <paroneayea>and sometimes, profile + environment, with container or VM, might even *be* profile + system? <davexunit>declarative profiles can be achieved with 'guix package --manifest', which is what I do. <davexunit>paroneayea: yes, I think there is an abstraction there that can be generalized. <davexunit>I use an unpriviliged shepherd to manage all sorts of things already. <paroneayea>davexunit: cool, what do you manage, if you don't mind sharing? <davexunit>emacs daemon, gpg agent, mpd, development servers, stuff like that <davexunit>stuff that runs as my user that the system doesn't need to know about. <thomasd>these keyboard slips are getting to be annoying <rekado>should we start something like shepherd-contrib where we can add user services? <rekado>BTW: what’s up with this error: @ substituter-failed /gnu/store/sk9gij7yiqka9lqymh27wzlcp7r5si2b-module-import-compiled 0 hash mismatch in downloaded path `/gnu/store/sk9gij7yiqka9lqymh27wzlcp7r5si2b-module-import-compiled': expected 75462d0f439a19cc0603da507ddb956e4dce114975988ac9f592a7ba14231a8e, got 1e1156fe87dba11be9865c326e5549b2518b079db3887856a113909cdc0881a1 <rekado>I often get this when building a package after switching branches, so I need to use --fallback. <TheSuperGeek>Hello, i'm testing Guix in qemu on fedora 24, and I have an error starting the install vm : <TheSuperGeek>qemu-system-x86_64: -net default: Invalid parameter 'default' <bavier>TheSuperGeek: have you tried 'user' instead? <bavier>TheSuperGeek: cool, have fun with Guix! let us know if you have any other questions <TheSuperGeek>bavier, i write that the guide have a problem, i'll report it for a fix, if i have others problem i'll ask :D <bavier>ok, I wonder if the 'default' value is not supported on some qemu's <bavier>TheSuperGeek: oh, can you ping anything? <roptat>how does (computed-file) work? I see it creates an record of type <computed-file>. How do I get the absolute path of this file? <roptat>TheSuperGeek, sorry, that was not for you <TheSuperGeek>bavier, that is strange : when i ping a website, it solve the dns query (it find the ip adress) but it can't connect <roptat>TheSuperGeek, if you run "ip a" on your host, is there an interface for your vm? and does this interface have an IP address? <roptat>you may need to enable IP forwarding (echo 1 > /sys/proc/net/ipv4/ip_forward), then <roptat>and maybe also masquerading (iptables -t nat -A POSTROUTING -j MASQUERADE) <efraim>i've heard ping doesn't work inside qemu <ng0>efraim / others: I'll see to get email working in the next days and reply to everything, I don't feel up to working on anything now <ng0>for emacs packages: when I install something which requires another package and usually you had it all installed from elpa, but in guix you have one package reference the dependencies, do I need to install the dependencies into my profile as well? <roptat>TheSuperGeek, /proc/sys/net/ipv4/ip_forward <roptat>maybe it's just ping broken since you can resolv hostnames, could you try "curl gnu.org" inside the vm? (don't know if curl is installed though) <ng0>we should really add some basic tools to the image <roptat>"guix package -i wget", and if that works, wget will be useless ^^ <TheSuperGeek>roptat, it seems to be stuck with : accepted connection from pid 1313 <roptat>good, you can stop it and install the system <rekado>wow, qemu-system-arm is sooo much faster on my workstation in the office. It finished compiling Guix in less than an hour. <rekado>on my machine it would take a day and then fail. <katco>hey all, i need a bit of support. i'm compiling a go program using the version of go in guix. it requires some c headers, and it's stating that asm/ioctl.h can't be found, but it's in my ${GUIX_PROFILE}/include/asm... any ideas? <rekado>katco: having something installed in a profile doesn’t mean that the compiler will pick it up. You’d have to tell the compiler where to look. <rekado>katco: I don’t know anything about Go, so I cannot help much. <rekado>katco: In general I recommend building a Guix package expression instead of compiling in the traditional way. <Petter>katco: Looks like you need to pass the -I flag to the C compiler. Are you doing this? <efraim>maybe with `guix environment --ad-hoc asm' <katco>Petter: well, the go toolchain handles the invocation for me... not sure how it works... looking into that <Common_Era>efraim, I think it was you that originally told me this a while back... does guix-devel still want a guide to distribution installation on an Apple computer? <katco>rekado: ty. i need to be able to build this the way upstream is, so i'll have to focus on getting this working as is <efraim>yes, I'd love to replace debian with guixsd on my 2007 macbook{, pro} <Common_Era>I've got it running flawlessly on my iMac. Can't test it anywhere else, but it should transfer. <Common_Era>Is the 2007 Macbook a PowerPC or Intel processor? <Petter>katco: I'd try: CGO_FLAGS="-I${GUIX_PROFILE}/include/asm" go ... <katco>Petter: ta, let me give that a try <Common_Era>It should transfer to any Intel Mac. I'll have it in soon. <Common_Era>I'll send it soon. I've been writing it for a few days. <katco>Petter: thanks, that has gotten me to another error. i should be able to iterate from here! :) <katco>Petter: i think i was just confused because i just rebuilt this machine, and on my other computer the compilation just worked <katco>Petter: maybe i had the library built using ubuntu's toolchain and wasn't building it each time <Petter>Could it be you had exported the CGO_CFLAGS variable? <katco>Petter: possible, but i don't recall setting that up anywhere <katco>Petter: so, from a guix perspective, do you know why guix doesn't resolve <...> to $GUIX_PROFILE/include ? <katco>efraim: tyvm for your suggestion btw, maybe this question is related ^ <bavier>katco: afaik, system headers are resolved against C_INCLUDE_PATH by gcc <efraim>I have added to my .bashrc PATH=$HOME/.guix-profile/bin:$HOME/.guix-profile/sbin:$HOME/.local/bin:$PATH <katco>bavier: i guess i thought that was part of the magic of guix's linking/hard-linking magic <katco>efraim: i think you can do `source $GUIX_PROFILE/etc/profile` and it will do that for you? <rekado>katco: yes, sourcing the etc/profile file is sufficient (and it works for all environment variables) <jmd>civodul: I've just been listening to your cufs talk. <katco>rekado: ty for confirmation. efraim you might be interested if that's simpler for you ^^ <jmd>Where did you say it was held? <jmd>Almost all the people asking questions had Australian accents. <bavier>I just watched civodul's talk too. it seems it went quite well <bavier>civodul: with the "hello version update" question, you could have mentioned 'guix refresh -u' :) ***jje is now known as Guest72938
<civodul>i thought i mentioned updaters at some point <bavier>the answer you gave answered the question quite well I think ***Common_Era is now known as After
<Petter>I like how you disconnect with a xrandr command. Geek level: pretty high! :D ***After is now known as Common_Era
<civodul>Petter: ah, right, well, i'll probably never get used to those GUIs ;-) <Guest72938>at boot i am now getting and isolinux error "no system image or UI configuration found" with a boot prompt, instead of a grub screen. any tips on getting it to boot with grub as it was before? 
