<emyles>sneek: later tell ArneBab: looks like the location to look in for a CMakeLists.txt is hardcoded in guix/build/cmake-build-system.scm:39: (abs-srcdir (getcwd)) <adfeno>And now sneek enters the repeated messages dance :) <adfeno>Or will it only talk to ArneBab if this one says something? <rekado>adfeno: “cmake ../source” is defined in the cmake-build-system ***atw` is now known as atw
<emyles>adfeno: oh whoops, I see now I replied to the wrong person, sorry <mekeor>is there a simple, straight-forward way to use %desktop-services but add some custom parameters to (slim-service) ? <buenouanq>there a guide anywhere about installing guixsd on something with uefi? <Apteryx>buenouanq: Is this not covered in the user manual yet? <ZombieChicken>Can someone remove ArneBab from the room? He's spamming connect/disconnect messages at a rather alarming rate <Apteryx>Is there a way to detect if I'm in a build session? <Apteryx>I need to make Python aware of 0) build mode 1) environment mode ($GUIX_ENVIRONMENT is defined) or 2) User profile mode ($HOME is defined... although this is a weak test). <efraim>is it wierd that I know I'm building tar because I recognize the test suite? <janneke>efraim: i think people are amazing pattern matchers, but i have the idea that you actually have a slightly different question <efraim>after removing the ld-wrapper-boot0 workaround, gcc-final compiled on aarch64 <efraim>had a non-reproducable error in the test-suite on gettext-boot0, and now i'm hanging on one of the tests in tar ***snape` is now known as snape
***jonsger1 is now known as jonsger
<efraim>looks like gcc-5.4.0 doesn't take -mcpu=armv8-a or either of the -mfix-cortex-a53-xxxxxx flags, so looks like i'll be diging into why that test hangs ***ShalokShalom_ is now known as ShalokShalom
<myglc2>Anyone out there using IPMI Serial over LAN (SOL) <myglc2>... or is there another scheme to capture console messages that make it into /var/log/... ? <snape>civodul: I just found the same import bug in openssh-service too. I'll send a patch tonight, and do a check for all services as well <snape>((guix build utils) missing) <snape>usually people don't have this bug because nscd provides the module <kyamashita>Other packagers: How does one deal with the "ld" binary not finding libraries like "-lpthread", "-ldl", and "-lc"? I'm close to packaging a Free Pascal bootstrap compiler, but I'm stuck here. <snape>civodul: and then it depends on the services order in config.scm <kyamashita>Also, using the same procedure on a filesystem that compiles with the FHS works fine. <snape>the module is loaded once and for all by the first service that needs it <snape>maybe that would be a bug, because such missing modules are hard to catch <civodul>we should run each snippet in a separate module <kyamashita>P.S. The glibc package present in that definition was part of my attempt to investigate the errors. <civodul>kyamashita: if you look at gcc.scm, it has "hacks" to allow gcc to find libc and all <civodul>specifically the thing that modifies the "lib" spec <davexunit>aww, iyzsong packaged guile-sdl2. how nice. :) <davexunit>now to get another haunt release out the door... <davexunit>I also have another project I'm trying to get a first release done for. <catonano>davexunit: I tested Haunt a few hours ago. It can't find the code for the comonmark reader. Keep this in mind for the next release ;-) <catonano>anyway, I can't wait for a new release ! <catonano>davexunit: I installed it ! I even tried in an environment with guile-commonmark in it ! <davexunit>you need to build haunt in an environment that has it <catonano>davexunit: I installed the guix package ! <OrangeShark>catonano: might need to have it included as the inputs <catonano>OrangeShark: I'll take a look at the package definition. Thanks <civodul>cantstanya: i think Haunt 0.2 didn't have that feature <OrangeShark>catonano: the one included in haunt's git should have it <catonano>OrangeShark: yes, the one in the repository has guile-commonmark in its propagated inputs <ng0>Only 4 more applications to package and I can configure my system with LXDE, see if I got it all covered then. <ng0>those commits depend on the updates I've just sent, shall I just resend the whole series then with the new packages added? <ng0>well, depend as in they are in the same branch and same file <ng0>I doubt we need lxsession.. but I don't know the desktop well enough <ng0>but the content of its bin/ says otherwise. <civodul>anyone making stickers for FOSDEM? :-) <davexunit>(or give some to paroneayea who will give some to me at libreplanet) <civodul>yeah it'd be nice, especially that quite a few of us will be there! <civodul>i did that a while back with the old logo but forgot the details of the company i used etc. <paroneayea>yeah, Guix and Guile both have such nice logos now... <ng0>I wonder how I will test lxde.. it's just a bunch of applications, so maybe like awesome-wm where I just install everything locally and it works <ng0>other distros have a meta package for lxde <Apteryx>Still can't wrap my head around something: How are python libraries made visible when building a package which has python & libraries as inputs? Or do we not support this use case yet? <Apteryx>Where? I only see this being done in a wrap phase (in the python build system), and this wrap phase only wraps binaries. <davexunit>the gnu build system will set the paths that are declared in the native-search-paths field of all inputs <davexunit>the python build system borrows this phase and many others from the gnu build system <Apteryx>Doesn't this only set PYTHONPATH to /gnu/store/...python-2.7.12/lib/python2.7/site-packages? In other words, to its own site-packages, which is useless since site.py already does that? <Apteryx>Or does this site-packages refers to the combined, synthetic "site-packages" which is created by a profile such as "$HOME/.guix-profile/lib/python2.7/site-packages" <ng0>will lxde cry when I build lxpanel with gtk-3? building with gtk-2 incokes a dependency on libwnck-1.0, while gtk-3 uses some version we have. and the way guix works, this should not be a problem <Apteryx>Each python inputs would have set their PYTHONPATH that way since the python build system borrows the native-search-paths phase from the gnu build system; hence the PYTHONPATH is populated with all the site-packages locations of the different python inputs. <Apteryx>davexunit: Thanks! I had foreseen this important detail. <davexunit>search paths are used all over. 'guix package --search-paths', 'guix environment', etc. <ng0>dependes: libfm with gtk2.. hrum. i hope someone had a solution for this before <Apteryx>Does "guix package --search-paths" visits all the package recipes installed to find the search paths? <Apteryx>I read a bug where "MANIFESTS" files were mentionned, but I am not yet familiar with those. <ng0>I will create an lxde meta package.. would be the most convinient way for this desktop <Apteryx>ng0: How are meta packages handled in Guix? With propagated inputs? <ng0>take a look at gnome <davexunit>Apteryx: it doesn't visit the recipes, those are long gone at that point <efraim>I think my tar test failure was me running out of space. Test 142 used 8.5Gb of space <Apteryx>davexunit: Any idea where I can find the "manifest" file of my current profile? I'm curious about what it looks like. <Apteryx>I now have a clear picture of what can be done to handle the order of precedence of Python libraries and enable pip & virtualenvs in a user setting. <gscacco>Installing "hello" package I have the following error: <gscacco>guix package: error: build failed: cloning builder process: Operation not permitted <tg>what's up with guix-sd on arm? anyone working on that? <ng0>hm.. last progress, I think the bootloader is being worked on? not sure who was/is doing tat <ng0>yeah, would be nice to know the current state. february, august, and july 2016 threads but no idea what's left to be done <ng0>Danny and David seem to be involved in the packaging for arm <ng0>dvc is not around, and for Danny i have no idea wether they use irc <buenouanq>pareidolia: please, if you can. Trying to install GuixSD on a Macbook. Installer says it needs a bios boot partition, but that doesn't work because this is uefi awfulness. <efraim>Did you try with the efi grub variant? <rekado_>kyamashita: a Free Pascal bootstrap compiler? <rekado_>kyamashita: I tried to package Free Pascal but saw that it needs a Pascal compiler, so I tried the GNU Pascal compiler next. <efraim>I think you need to leave an efi partition at the front of the disk and mount it to the "correct spot" and go with the efi grub variant <rekado_>kyamashita: does the bootstrap compiler contain binary blobs? <efraim>At least that's my plan with my old macbook collection <rekado_>is there still enough time for t-shirts? <efraim>GuixSD ~\\/~ define* your-system ***snape` is now known as snape
<buenouanq>rekado_: how do I use the efi grub? something in the os config.scm I assume <CompanionCube>ACTION wonders if anyone has ever considered using GNU Shepherd to manage an Xorg session <efraim>There's something about choosing which grub package in the config <efraim>sneek: later tell civodul its in test 142, sparse05, and I think it might be a disk IO thing or an out-of-space thing, when I built tar with -K it was 8.5 Gb <buenouanq>Know of any trick to make it boot to the usb? I've got a blinking questionmark folder now - Only way I know of to get past it is to reinstall OSX from disk, install refind, then start over. Quite annoying, tedious, and time consuming (;-___-) <efraim>There's a magic key to being up a boot device choices menu, not just c for cd <buenouanq>that's it, but it doesn't show my guixsd installer flash drive <jmi2k>I'm getting "ghc: error while loading shared libraries: libncursesw.so.6 ..." with ghc 8.0.1, but ghc 7.10.2 works fine. Anyone knows why? <CompanionCube>ACTION hasn't really used shepherd but likes the idea...even if he doesn't know guile <ng0>which needs to be fixed <ng0>at least I remember this being the reason why I passed on the ghc8 packaging <jmi2k>Oh so downgrade until fixed? <ng0>aren't they separate packages? <ng0>haskell.scm:268 and haskell.scm:91 <ng0>I want to add some more features to my guix fish completions, then I will send in the package definition to include it, I have it on an local branch where I have more fish related packages <ng0>it's usable but fancy stuff can be added. and text shortened <ng0>in case anyone wants to do it. 0.2.9.9 <efraim>debian leaves 200MB for /boot/efi at the front of the disk <buenouanq>and then I just need (grub grub-efi) in the bootloader declaration? <ng0>is the nomodeset thing for directly in grub edit just written "nomodeset" ? <ng0>because if it is, it solved nothing for the netbook I'm trying to restore <ng0>same blank cursor blinking screen after starting from grub of the usb installer <rekado_>ghc must have lost some references, but “guix gc” still knows about everything that’s needed: <rekado_>LD_LIBRARY_PATH=$(printf "%s/lib:" $(guix gc --references $(readlink -f $(guix build ghc)))) ghc <ng0>for diffoscope, how do I define the output of an build? I can't use simply (guix build openssl) because it has more than one output <lfam>ng0: I just manually copy and paste the path to the output that I want to compare <ng0>damn. I CC'd guix-devel with the last message <ng0>but that's just on osx so far <adfeno>Sorry, I'm new to this, but how is "losing dependencies on X11" a good thing? <ng0>it's for making it work on X11 and Wayland <ng0>where for wayland, without X dependencies this single application would not pull in xwayland <lfam>It's nice to see them paying attention to this sort of thing <ng0>but that will be a long way until you can run wayland on its own <adfeno>Ah, so now they are seeking compatibility with both X11 and Wayland? <ng0>first time in years i looked at lxde <ng0>or tried to (see thread) <ng0>maybe graphics require some blob.. idk. TAILS loads just fine, guixsd installer even with nomodeset acpi=off stays at a blinking cursor <adfeno>Does GuixSD have an "os-release" file in "/etc" or some similar "release" file? <adfeno>ng0: About the video issue: PErhaps enable the boot debug. <ng0>wild that it still works, when I got it for free it was already 7 years old <ng0>no, debug changes nothing <adfeno>Not even more transparent messages? <ng0>just a blinking cursor <ng0>blinken lights without lights <ng0>just a small underscore on a black screen <ng0>i know gentoo, archlinux, parabola, tails worked <ng0>oh debug takes arguments <ng0>debug=all pager=1 , at least in the config file <ng0>maybe it just takes a really long time <ng0>but I just waited 10 minutes <adfeno>OK... Does the "linux" line of the related grub entry has "quiet" or "splash"? <adfeno>These two might prevent us from seeing "debug" messages. <ng0>no, guix installer doesn't come with those two <adfeno>ng0: I think you need xserver-xorg-video-intel and xserver-xorg-input-synaptics <ng0>are those included in the image for the 0.11 release? I'm not sure right now <ng0>and even if not, it should throw some error <ng0>and not "_ ad infinity" <adfeno>I wish there was a way to force framebuffer instead. <lfam>I want to fix the /bin/sh -> /bin/bash replacement in ffmpeg's 'configure' script. I wonder, do SHELL and CONFIG_SHELL also need to be changed from (which "bash") to (which "sh")? <adfeno>ng0: Does the bliking cursor happens after install? <adfeno>Or is it in the boot (after GRUB) of the image? <ng0>I would be capable to deal with anything when I would be in the installer <sneek>civodul, you have 1 message. <sneek>civodul, efraim says: its in test 142, sparse05, and I think it might be a disk IO thing or an out-of-space thing, when I built tar with -K it was 8.5 Gb <civodul>lfam: just merged master in core-updates <lfam>civodul: I hope I didn't create too many merge conflicts with my sh / bash changes <civodul>lfam: there was one in mit-krb5 due to the rename, but that was easy <lfam>I'm preparing another handful of them for the master branch now <civodul>efraim: tar's test suite requires quite a bit of disk space and preferably a file system that supports sparse files <lfam>And then I'm going to prepare and send a security update for openjpeg. Finally, I'll do a bunch of ungrafting commits on core-updates. <civodul>we should probably consider starting core-updates not too far from now <lfam>The patch is still not in the official openjpeg repository, but it's written by an openjpeg contributor, and Debian is considering it. <lfam>Tireless? Haha ;) There's a reason I was mostly absent last week <lfam>Yes, I'm telling you these things so you don't start the core-updates evaluation before I get these changes in :) <civodul>we'll send a note before starting anyway <ng0>ponyc has a weird Makefile <civodul>lfam: i hope it's not Guix making you too tired though, we've gotta pay attention to that <lfam>civodul: Do you know if I should change the values of SHELL and CONFIG_SHELL when changing the sh -> bash substitution in ffmpeg? I'm not sure how those variables interact with everything else. <civodul>CONFIG_SHELL is honored by Autoconf-generated configure scripts <lfam>civodul: It was my "day job". Although I had also reduced the number of patches I reviewed as well. The number of contributions seems to have increased A LOT in the past month or so. I'm hoping to grow more reviewers :) <lfam>If there's time, it would be nice for all the Guix hackers at FOSDEM to discuss the work-load :) <ng0>I think I asked this before, but does it matter when I review? the commiter has to review anyway before commiting <adfeno>ng0: Reading some documentations, it seems bliking cursor and no allowed input is indeed video issue. <ng0>adfeno: okay. at least I have a confimration <lfam>ng0: I think that reviews from non-committers are valuable, especially when you describe if and how you tested the patch. For example, saying "I built this package and it seems to work in my limited testing" is very useful information. <ng0>that's enough for me today, adfeno <civodul>ng0: agreed, it can speed things up and i'm sure submitters appreciate it <ng0>ok, I try to review more often. <adfeno>ng0: Now, if we could only replace the faulty thing.... <lfam>I think that if there is a patch from a new contributor, and somebody who is not a committer just says "LGTM", the value is less. <lfam>I typically reserve those "LGTM" messages for committers with lots of experience. For example, when rekado submits some bio-info patches <lfam>It's just a "sanity check" <lfam>Some have said, let's get this merged <lfam>Basically it means, "The patch looks good, and I have no further feedback" <adfeno>About the absense of "release" file in "/etc": How to make sure that user is running GuixSD? <adfeno>I'm making a script to automate hardware-compatibility report creation, for new-commers to free/libre software... <adfeno>... And I need to some way to check if they are running GuixSD; :) <ng0>Nix includes a file somewhere <ng0>but Guix doesn't, so you haver to come up with some check <ng0>I like it that you can't check the system which is running <lfam>I don't see a reason that we couldn't include a file <ng0>was it just not considered before? <ng0>or was there some reason? <adfeno>... How does one check if shepherd is running? <ng0>shepherd can run on other systems too <adfeno>Yep, but I can make a combination of factors, like so: <adfeno>shepherd && guix-daemon == guixSD <lfam>You could look for /gnu/store and /run/current-system. <lfam>Although I haven't checked if /run/current-system exists on other systems, I hadn't heard of it before using GuixSD <adfeno>If shepherd provides a way to list service-files being used, I could lookinside one of these to see if it has "guix" inside. <lfam>shepherd and guix-daemon can both exist on foreign distros <adfeno>Is it possible to tell shepherd to tell me where is a given service file? <civodul>adfeno: there no real "service file" in the sense of systemd service files <buenouanq>hmmm.. weird error I've never got before with the same config: <buenouanq>no target of type 'polkit' for service type gnome-desktop <civodul>buenouanq: somehow your list of systems is lacking polkit <civodul>polkit is part of %desktop-services, but not of %base-services <buenouanq>I never changed %base-services to %desktop-services <civodul>good thing is that you got an error rather than a system that does not boot ;-) <adfeno>I might need to base my detection on a common combination of factors then. <buenouanq>civodul: well, now that it's going we'll have to wait and see. hopefully I did the efi and grub-efi things correctly <civodul>buenouanq: oh you're doing EFI, that's fun too :-) <buenouanq>and I added (grub "grub-efi") to the bootloader declaration. <ng0>ha, nix has a solution for my ponyc problems <adfeno>I might be wrong but: doesn't UEFI require also a user-generated key? <adfeno>Or does grub-efi take care of that? <buenouanq>oh dear - what would that be and how would I go about it if I have to? hoping grub takes care of it (;-___-) <adfeno>buenouanq: I'm also searching for it (for other purposes: somoene (bum/kang00) is requesting my help (on #myboss channel) to make PXE+UEFI setup). <adfeno>I found some information on how to use grub-efi, but hardly a definitive information as to if it needs key or not. <civodul>basically when installing from the USB image, 'grub-install' will not do the last bit of work <buenouanq>I will not be able to get to a grubprompt though. <civodul>IOW, on your first reboot, you'll need to use the browse feature of your EFI BIOS to select grubx64.efi