<katco>Petter: i figured out what was different from my old setup... i didn't have gcc-toolchain installed through guix so it was an odd mismash of guix files and system files <katco>Petter: by guix package -i gcc-toolchain, i get the expected C_INCLUDE_PATHS in ${GUIX_HOME}/etc/profile which solves the issue <Guest72938>civodul: i installed with a usb stick as described in the link. this was a working install that was running fine for about a week until i rebooted today and got the isolinux screen. <civodul>Guest72938: GuixSD doesn't use/install isolinux; maybe something else interfered? <Petter>katco: Aha. Thanks for sharing your findings, I'm interested in Go related issues. ***Digitteknohippie is now known as Digit
<lfam>I missed earlier conversation. Is everything okay with the Go package? <Petter>Yes, katco had an issue compiling a program that required C headers. <Petter>They solved it by installing gcc-toolchain. <Petter>Well, isn't it the flasher of boot firmware that shows up. <Petter>Saw you had a successful operation updating Libreboot. <Petter>I'm inspired to do it as well now. <paroneayea>Guest72938 probably needed to ln -s /boot/grub/grub.cfg /boot/grub/libreboot_grub.cfg <lfam>I love seeing "oops goops" <lfam>It seems like the Hydra interface is less useful when there is only one evaluation of a jobset. Once there are two evaluations, some useful filters appear <rekado>Icecat crashed three times today :( <rekado>ABORT: X_CopyArea: BadDrawable (invalid Pixmap or Window parameter); <rekado>file /tmp/nix-build-icecat-45.3.0-gnu1-beta.drv-0/icecat-45.3.0/toolkit/xre/nsX11ErrorHandler.cpp, line 157 <sneek>civodul, you have 1 message. <sneek>civodul, davexunit says: yeah, if the URLs aren't complete then that's a silly oversight on my part. I'd say "file a bug" but I don't have my shit together and there is no bug tracker. <rekado>in my arm VM I need to build automake from source, even though it looks like we have a substitute on hydra. <civodul>rekado: "guix build automake -s armhf-linux -n" suggests there's no substitute <civodul>actually /gnu/store/r8jjapxmm2ysmzp1097wya2xnzslk66l-automake-1.15 was unduly marked as failed <civodul>now you should be able to get a substitute, at least from hydra.gnu.org <civodul>(the 404 may be cached on mirror.* and in /var/guix/substitute/) <thomasd>I'm switching from my distro-provided emacs 24.5 to guix' emacs 25.1, and have the impression that it runs much faster ?! <thomasd>Does anyone know of an emacs benchmark that I could run to see if it's really true? <civodul>thomasd: everything runs faster with Guix, of course <thomasd>quigonjinn: thanks, but it's not so much about startup performance. It just seems more "snappy" overall <thomasd>civodul: yes, actually it's mostly OK. Some fonts were displayed badly in emacs, but I could fix that with "fc-cache -f", which I learned about from an older thread here <thomasd>(and maybe by adding a ~/.config/fontconfig/fonts.conf, I didn't debug it in a scientific manner) <thomasd>Now I'm at a point where I'm not sure if the fonts in Guix' emacs look exactly like those in Ubuntu's emacs. <thomasd>But the performance difference feels very real (when using magit, CEDET, mail, ...) , . I'd like to find out how/why. <thomasd>Hmm, running a pure elisp benchmark (bubble-sorting 1000 random ints), the Ubuntu emacs is faster than the Guix one (~1.5 seconds vs 1.7 seconds) <thomasd>I'll try to do some more thorough tests, and see if anything comes up. <rekado>icecat crashed for the 6th time today. Using epiphany now :( <civodul>i wonder how it can be this unstable <civodul>it doesn't change much code compared to Firefox <ng0>do you know why it breaks? <rekado>ng0: ABORT: X_CopyArea: BadDrawable (invalid Pixmap or Window parameter) <ng0>this is using the latest firefox esr or some other version as base for icecat? <thomasd>is there a way to specify the contents of a local directory as a package source? <civodul>(package (source (local-file ".")) ...) <thomasd>aha... I think I'm approaching enlightenment w.r.t gexps :) <thomasd>is it possible to define “local packages”, say, within another package definition. <thomasd>like (let ((my-pkg (package ...))) (package (inputs `(("mypkg" ,my-pkg))) ...)? <ng0>I'd be curious if a local package can reference another local package as dependency, they are not aware of each other and come bundled in the source of each package <davexunit>if you have a reference to that object, you can do all sorts of things with it <thomasd>but `guix package' will not know this local package <thomasd>so could it be gc'd if (and only if) the “owning” top-level package is removed? <thomasd>yes, I'm trying to understand how it would all come together. but probably I should just play around with it <thomasd>I'm wondering what would happen if I tried to install a package which refers to a “local” package defined in the way above. <davexunit>thomasd: the packages that 'guix package' can see are dependent upon what packages are in guix itself in the (gnu packages ...) namespace, and the additional modules you add to GUIX_LOAD_PATH or with the --load-path flag <thomasd>davexunit: ok, that was also my assumption <davexunit>thomasd: you are still confusing things. a "local package" has no meaning. <davexunit>if you want to install a package that you wrote that is outside of the guix source tree, you need to use the -L flag or set GUIX_PACKAGE_PATH appropriately <davexunit>package objects completely capture their dependency graph, so your example of a package expression inside of a let expression is no different than any other package expression <davexunit>there is literally nothing special about packages. I can't stress that enough. <ng0>my question was more related to, I checkout 2 branches of an subversion repository. one currently has a (simple) guix package included, the other would need this. I think this is what the pre-inst-env etc is about, setting the environment variables <davexunit>ng0: you would need to arrange to create a guile environment that was able to evaluate both of them <davexunit>and be able to reference one package from the other's module <davexunit>the particular example you provided I would highly discourage <ng0>it's for development versions of software <thomasd>davexunit: thanks. there's still something I wonder about, but I'll just have to experiment and see how it all works together. <davexunit>but it's doable, if each project provides a guile library. <ng0>but I think it'd be easier to outsource this to a branch/path where all packages are in modules <ng0>i'm still damaged from dentist operation, just thinking out loud <davexunit>I keep a separate repository for my own special packages. <thomasd>davexunit: how does `guix package -l` or `guix package -r` know about packages which are not defined in (gnu packages ...). Probably this information is kept in the store database (at least I think I recall there's a database there somewhere?) <davexunit>thomasd: the store database is not involved. <thomasd>let me take another guess: it's the current profile, then? <davexunit>the guix tools query all modules on GUIX_LOAD_PATH and accumulate the package objects exported from them. <thomasd>ok, I'm going to have to login as an anonymous user before we continue this :) <ng0>davexunit: I knew of some methods before, I'm just starting to wonder if my initial move on gnunet.org was indeed not so good and the separate package collection (for development versions) would be better <davexunit>ng0: I don't know anything about gnunet, but if it's work that you want upstreamed then it's best to keep it in guix itself <ng0>it is upstream work i do <davexunit>thomasd: modules are first-class things in Guile, so you can programmatically traverse the symbols it exports, which is how Guix comes up with a list of all packages. <thomasd>so if you install a package which is not in any (currently visible) module, (i.e. by evaluating a (package ...) expression, or by removing something from GUIX_LOAD_PATH), it effectively becomes invisible to tools <davexunit>but Guix is a library, so you can write code that does whatever you'd like with any package object <thomasd>so I just did a test, and installed a package from my GUIX_LOAD_PATH in my profile... <thomasd>in a session where GUIX_PACKAGE_PATH is not set, it doesn't show up with, e.g. `guix package -s <pkgname>` <thomasd>but it does show up (luckily) in the list of installed packages `guix package -l` <davexunit>the name and version information were saved in the profile <thomasd>well, ok, that actually is what you'd expect, isn't it? :) <bavier>civodul: thanks. I think this was the article that was being discussed recently at my workplace <bavier>oh, OpenHPC. that's another one I don't know how to feel about <bavier>it seems that the bundling-and-sharing idea is important <thomasd>so actually HPC needs persistent images, we could use smalltalk or lisp machines ;) <mbuf>For USB installation, I need to create a partition /dev/sdc1, but, dd should be invoked on /dev/sdc? <Apteryx>mbuf: I don't believe you need to create any partition there. <Apteryx>Just copy the raw image using dd, it should contains its own filesystem. <mbuf>Apteryx, no need to use mkfs either? <civodul>bavier: Singularity's 'sexec-suid' is a big chunk of C code, that tells a lot on how little the technical side of things really matters... <civodul>so it tries user namespaces and otherwise falls back to calling the suid thing <civodul>the suid thing is 99% the same code as the non-suid program <civodul>no, it's really the same code, with just a few lines in #ifdef SUID <civodul>there's 1 line of debugging message every 3 lines of code <civodul>that makes it look more secure i guess <rekado>I get this error trying to build a new python package: Aborting implicit building of eggs. Use `pip install .` to install from source. <rekado>what does this mean? Does our python-build-system support this? <rekado>I'm having problems building R interfaces. The compiler would always complain about conflicting declarations of math functions. <rekado>e.g. cmath:84:17: error: declaration of C function ?double std::abs(double)? conflicts with <rekado>In file included from /gnu/store/pbz6h59l4c4y77j5ik45krlvp4xq79qq-r-3.3.0/lib/R/include/R.h:40:0, <rekado>I see this in two separate programs: spams and shogun. <rekado>another error that's repeated a lot is "error: template with C linkage" <mbuf>my USB data card dongle is being detected as mass storage. Is there a way to switch it to use cdc_ether for mobile Internet? <rekado>I wonder if there's something wrong with our R package. <civodul>we've seen issues with math functions with the new libc, IIRC <rekado>it's also the reason why I cannot upgrade shogun to the latest release <rekado>because the R interface cannot be built. <civodul>apparently R.h provides conflicting declarations <civodul>so you need to find out why it provides those <civodul>it could be its configure machinery that incorrectly determines that it needed to provide them <Apteryx>Would someone know how I can make installed GTK+ themes "visible" to a theme switcher such as lxappearance? <rekado>civodul: okay, thanks for the hints <rekado>so, looks like spams is wrong. The R NEWS file mentions that R headers should not be included inside an extern "C" block. <mbuf>The e1000 driver is not a free software driver? <Apteryx>Strange enough, if I just change the theme name under the .gtkrc-2.0 file, it applies the theme, but lxappearance still don't know where to look for the themes it seems. Maybe there should be an enviroment variable set which I'm unaware of. <mbuf>igb-5.3.5.4 Intel drivers are non-free? <Apteryx>Anyone having icecat crash a lot? ABORT: X_CopyArea: BadDrawable (invalid Pixmap or Window parameter). <bavier>Apteryx: I've had crashes recently too <bavier>ooo, firefox is going to require a rust compiler in the 2017 timeframe <davexunit>so we'll get an icecat that needs rust in the 2018 timeframe ;) <katco>is anyone else having trouble with calibre? i get a stack-dump upon launching <katco>"ImportError: No module named QtWebKitWidgets" <Apteryx>What would be the best way to find the source which the libgtk-x11-2.0.so library was made from? <Apteryx>I've browsed the web but I can only find the string I'd like to locate in a language ".po" file, with a comment pointing to gtk/gtkthemes.c, which I can't find anywhere. <Apteryx>The string which I'm trying to track down is : "unable to locate theme engine" as I want to understand where GTK is looking for the themes. <enderby>katco: yeah i've had that problem with calibre too :( <katco>enderby: there was a thread on it back in july i think. i thought the author said they were going to roll back a dep and push <katco>enderby: but no such luck =/ <ng0>we do not package alsa-plugins? ***htgoebel1 is now known as htgoebel
<htgoebel>I try to get the next level of guixperience and try running a container with nginx and ssh in. Based on `gnu/system/examples/bare-bone.tmpl`. I fail to access the ssh port, though. <htgoebel>Are containers the wrong approach? I'd prefer containers since there is no need to copy a vm out of the store :-) <bavier>htgoebel: running the vm directly out of the store will not work for you? <htgoebel>bavier: The docs say: "To run the image in QEMU, copy it out of the store …" <bavier>I'm not sure that one can ssh into a container <bavier>but I haven't played with such things much <htgoebel>bavier: I've done this in docker. The trick is to start init within the container. <davexunit>htgoebel: our containers don't do networking right now <davexunit>but you also don't need ssh to access a container <davexunit>you can create a container process and just join the container with 'guix container exec', I believe. <davexunit>keep in mind that 'guix system container' and 'guix container' are proof-of-concept, experimental tools and not something I would rely on right now. <htgoebel>davexunit: Thanks for pointing this out. Otherwise would have wasted a lot of time :-) <fr33domlover>Hello, newbie question: I use rxvt-unicode from apt-get install and it reads my configs from ~/.Xdefaults . Now I did `guix package -i rxvt-unicode` and it doesn't see that file - where do I need to place it so that Guix's rxvt sees it? <davexunit>so for that we need a virtual networking implementation and a way to forward ports <davexunit>it's a feature that I couldn't get to when I did the initial container implementation <davexunit>fr33domlover: guix does nothing special in this regard <davexunit>you can try running it with 'strace -e open' to see what files it reads <fr33domlover>davexunit, but there's the guix profile and files under it etc. <davexunit>fr33domlover: we don't alter program behavior <fr33domlover>davexunit, .Xdefaults is an X related file, not specific to rxvt at all <davexunit>if rxvt isn't reading it it's easier a problem with the way we've built it, with rxvt itself, or the version of rxvt we have no longer reads it <fr33domlover>davexunit, looks like it does successfully open the file. thanks for the strace hint. probably some technical issue with the file's content <davexunit>fr33domlover: great, one step closer to the root of the problem. :) <htgoebel>ACTION would find containers with network support really cool, though: Need to wast space for a kernel and such <htgoebel>davexunit: when try to build the vm-image, I get a "Could not access KVM kernel module: Permission denied" error within my base system (which is not guix yet). What could be the reason? <davexunit>htgoebel: your user doesn't have permission to write to /dev/kvm <ng0>there could be a qemu or kvm group, depending on your system <htgoebel>ng0: There is none of these yet. I'll check if installing qemu or some kvm package will bring them <ng0>I just imported all guix-devel@gnu.org mail. it's not so high trafficed as others, but the mails/month are rising <efraim>i have ~18k mail for the past year <efraim>the older mails i move to archive.guix-devel <ng0>it's still counting, but so far it's 25000 <htgoebel>I'm now member of group kvm, kvm has +rw on /dev/kvm, but I still can't create the vm-image. <htgoebel>Are there any other requirements on the base-system? <ng0>3 days no email client. "you have 710 new messages" <civodul>htgoebel: what error do you get with 'guix system vm-image'? <civodul>does 'id' in the same terminal shows that you're in group 'kvm'? <civodul>"Permission denied" suggests you're not <civodul>it's the build user that has to be in 'kvm' <civodul>although paroneayea and i made KVM optional <efraim>AttributeError: 'NoneType' object has no attribute 'split' I just want to take the error and throw it at upstream <rekado>Apteryx: icecat crashes for me too. Way too often. And I lost my session twice. <rekado>now I’m backing up the browser profile each time I start it… <hugo_dc>efraim, single quote is OK, in error you show there is a variable that is trying to access the split method, but the variable is not initialized/assigned <htgoebel>civodul: Oh, this is tricky :-) I need to add all guix.buildusers to kvm. <htgoebel>maybe we should emphasis in the manual or even add an explicit check to the code. The code currently test file-exist, but does not test access. <rekado>enderby, katco: the problem with Calibre is that it needs a Qt component that was previously bundled, but which we threw away. It would need to be packaged separately. <hugo_dc>efraim, what is the problem with bs4? <efraim>well, for parse, another one with the same error <efraim>rekado: thanks, I knew it was something obvious/simple <rekado>(oh, weird, doesn’t work for me anymore.) <rekado>(appending “.patch” to the pull request URL) <AlephTwo>So, I have a question: How do I remove guix without toppling my system? (it's got a large and clunky set of crap installed and guix is taking it over the edge) - alternatively, how do I remove packages from guix and re-reference the original packages without some nasty hacks? <efraim>doh, typo in my jasper update commit <Common_Era>I'm here, I just don't know the answer to your question. <AlephTwo>Fair enough. Right now, vim is my main issue - the vim binary is being sourced from the guix packages, and I really want to source my local stuff w/out having to hack through configs like a madman <efraim>it should just be /gnu /var/guix $HOME/.guix-profile and whatever variables you set in .profile or .bashrc <bavier>AlephTwo: you don't want to just uninstall your guix vim? <ng0>civodul: you wrote about "use flag" like feature in 2015, but I can't find the email anymore. do you remember (or have a better search than I do) which thread or month this was by any any chance? <rekado>AlephTwo: note that uninstalling with Guix does not remove the files. It only marks them as unused. To remove all unused stuff run “guix gc”. <AlephTwo>ACTION ponders the time it takes for guix refresh to finish <rekado>AlephTwo: unless you are a developer you don’t need guix refresh. <rekado>AlephTwo: what are you trying to do? <AlephTwo>tbh, I'm trying to remove it, as it has a bunch of stuff which I installed a year or so ago as an experiment and need removing. I'll try again when I have more time to experiment with guix <rekado>AlephTwo: to get rid of everything Guix-related do as efraim said and remove /gnu, /var/guix, /etc/guix, and the .guix-profile link in your home directory. <rekado>AlephTwo: the largest directory is /gnu, which contains all built packages. <AlephTwo>so many ssh connections, so little concentration. <ng0>civodul: found it: "Optional runtime dependencies in Guix", January 2015