***christianbundy is now known as Guest63901
***catonano_ is now known as catonano
***terpri_ is now known as terpri
<apteryx>Formbi: C-z to stop (interrupt) the job, then 'bg' to tell it to continue running in the background; it's a feature of the OS/Bash called job control, see "info (bash) Job Control Basics" <guixy>I need a firewall on my guix system. Can anyone point me in the right direction? <guixy>I'm surprised that guix is designed for security, but there is nothing in the manual (last I checked) about firewalls. <apteryx>there is a service to declare iptables rules <guixy>Thanks! I'll look into that. <guixy>It says that nftables is being developed to replace iptables. What is it supposed to fix? <guixy>I'll add something about firewalls in the manual's description of iptables-service-type and nftables-service-type and send a patch sometime. Unless anyone beats me to it... <apteryx>does someone run Guix on actual ARM hardware? <apteryx>I'm trying to build a package using QEMU emulation and it crashes or segfaults. I'm not sure it's due to QEMU or the application. <guixy>apteryx, I'm trying to install it on my bananapi-m2u. The package manager is working with armbian, but I've had problems getting the system to work. <guixy>I've also looked into putting it on a raspberry pi, but there's no libre bootloader. <guixy>If you want to put the system on a raspberry pi, you're on your own getting the bootloader to work. <guixy>I think there are guix users working on making the system for librem 5. <apteryx>yeah, I've heard about the Rasperry Pi having freedom issues on that front (the GPU is used to bring up the system, but there's no free driver for it or something like this). I thought there were some breakthrough but I haven't followed closely. <guixy>I thought the gpu instructions were documented, but if that was true there probably would be a boot loader already... <guixy>linux-libre takes longer to cross-build for armhf on x86-64 than native ungoogled-chromium, which I didn't think was possible. <guixy>Guix hangs running the linux-libre script. <guixy>There's some python, but most of it is bash. <guixy>There's also some sed, perl, gawk. <guixy>But it looks like that's only the deblob-check script. <guixy>No python found in deblob script <apteryx>But most of the time used by the deblobbing process occurs in that Python deblob-check script <guixy>It takes a finite amount of time to build linux-libre natively on ARM. So until I set up my bananapi as an ARM guix substitute server, I'm stuck using it to build system images. <guixy>ooh... linux-libre tarball for arm has a substitute available... <guixy>But the kernel itself is absent... <lfam>DMA-BUF Userland Memory Heaps (DMABUF_HEAPS) [N/y/?] (NEW) <jackhill>Are you seeing this is `make oldconfig`? I've found the documentation that is browsable in `make menuconfig` somewhat informative. I don't know where it lives in the tree. <lfam>jackhill: Yes, I'm using oldconfig. It's my first time configuring a kernel build <lfam>I want to use 5.7 but Guix doesn't have a config for it yet <lfam>Is "not yet" for any particular reason? Or from a generally conservative approach to new kernel features? <guixy>What package provides the 'host' command? <lfam>guixy: The "utils" output of bind. It can be installed with `guix install bind:utils` <jackhill>lfam: It seemed unlikely that the feature would see use right away. I'm not expert, but I've configured a fair number of kernels (no idea how well) in my Gentoo days ~5 years ago. <lfam>You're more of an expert than me <lfam>jackhill: So, the reasoning is that increased compilation time and disk usage would not be worth it yet, because it would not be used? <lfam>If the feature became popular, could we expect people asking us to enable it later? <lfam>I don't have a sense of how to approach this task, so I'm open to advice and stories <jackhill>the other argument, is to provide it so that if people want to develop against it. Not enabling it now there is a danger that we'll overlook enabling it later, and we'll need to so that people usecases won't break. <lfam>I've been involved in guix development for years, but I can think of only one case of "can we enable this kernel feature" <lfam>So I do wonder about that <jackhill>lfam: I think the mindset is different from distro kernels than when I was doing them only for me. I know that there has been some discussion on this on the mailing list, but in some ways I wish all FSDG distros could collaborate on kernel configs <jackhill>for this particular thing, it would be nice to know the types of hardware that would use it in their drivers. If they all require non-free firmware… However, the slides mentioned DRM (direct rendering manager) which is used with free graphics too. <lfam>Regarding "the rules", I disabled HDCP support on AMD platforms (it's a new option) <lfam>Bit of a shame IMO but the FSDG prohibits that type of DRM <jackhill>I don't really know either. Despite what I initially said, I'm leaning towards enabling it. <lfam>True but it would be nice if Guix was an option for people that watch tv online. It's a lot of people <jackhill>Trisquel's kernel is probably too old to help us. Maybe we can take a peek at parabola and PureOS? <lfam>It would be useful to look at theirs but I am going to try doing it on my own at first so I can learn <jackhill>lfam: hope I helped a little and didn't confuse you more ;) <guixy>Anyone know if nftables supports blocking specific domain names? <guixy>If not I'll block those domain names just like the manual suggests blocking facebook. <lfam>Not at all jackhill, thank you! <jackhill>lfam: cool. Thank you for giving it a go. The care and feeding of Linux-libre is imporant work. <jackhill>I'm going to bed. Good luck and have fun! <mroh>can someone confirm '(exception unbound-variable (value #f) (value "Unbound variable: ~S") (value (qtbase)) (value #f))' in guix-package-cache.drv after `guix pull` <mroh>oh. nvm, its in a channel somewhere, sorry for the noise. <greenrd>I am trying to install guix SD with a different kernel version, but I am getting os.scm:13:14: error: linux-libre-arm64-generic-5.7: unbound variable <greenrd>I have (use-modules (gnu packages linux)) <greenrd>Even when I change it to say (kernel (@@ (gnu packages linux) linux-libre-arm64-generic-5.7)), I still get the same error <greenrd>ah, I was using an old version of guix, that's why ***jonsger1 is now known as jonsger
***sputny1 is now known as sputny
<peanutbutterandc>Okay... this is strange... I ran `guix environment guile` and then tried running ./autogen.sh (git clone) and it complains about automake not being installed. And it isn't. I thought guix enrironment would take care of it all... <davidl>so I sent a patch following the contributing instructions in the manual as closely as I could - how long does it usually take for a patch to be reviewed and applied to master? <peanutbutterandc>davidl, I think mine took long enough for me to feel pleasantly surprised by it being accepted. :) 15-20 days or so? But I have only sent in two patches. And the next one was accepted within a few days (it was a follow-up to the first). <davidl>peanutbutterandc: oh, thats a while. I think Ill not wait then for that one to be accepted before starting to work on the next one. <peanutbutterandc>davidl, But mine wasn't all that urgent. So I don't know for sure. You'll probably have to ask more regular contributors. But yeah, I think starting to work on the next one might be a good idea. :) <euandreh>Is there a place where I can download logs of the #guix channel? logs.guix.gnu.org only gives me the option of visualizing them online. <euandreh>But I'd get the HTML files, but I wanted text files <euandreh>so when searching through them it's easier to read locally <stikonas>yeah, that's true... Maybe some tool can do a reasonably good job of converting html to text <smileyface>Hi. Is there a way to install Guix on a live-USB to demo the OS? I would like to install it on my laptop but I'm not sure about the compatibility with my hardware so would like to try it first. <smileyface>euandreh: I read this page, but from my understanding this is just for the purposes of installing Guix onto the computer permanently. Would I be able to e.g. get into a desktop environment within the installer? <aadcg>hi. I spent a patch to update an emacs package. Eventually the update is already in place and was already commited by another user. Should I close the thread at issues.guix.gnu.org? Or I don't have permissions do to so? <smileyface>euandreh: will try the installer and see if I can install a WM etc. there <fnstudio>hi is there an obvious guix equivalent for `dpkg -L <package-name>`, i.e. a command that displays all the files installed as part of a given package? <fnstudio>on a different note, i think my guix-installed apps (I'm using Guix on a host OS) can't be found by my i3 window manager <fnstudio>the same apps can be reached (and launched) from a terminal <fnstudio>but they don't get launched if bound to a key-combo in my i3 config file <fnstudio>i suspect that has to do with the fact that i need to inform i3 of the Guix path <fnstudio>not completely sure how i can do that though... <fnstudio>maybe i need to tweak my xinitrc which is the place where my i3 gets launched? <peanutbutterandc>has anybody here compiled guile from source? I want to know where the build binary of guile resides once `make` is done (before make install) <janneke>peanutbutterandc: libguile/.libs/guile <fnstudio>(yes it seems it worked... adding the path in my xinitrc makes guix apps visible to i3) <peanutbutterandc>janneke, (Hey there!) Thank you! That is one secret place. It throws "error while loading shared libraries libguile-3.0-so.1. Is there a wrapper somewhere? Or some env var that I should fix? <janneke>peanutbutterandc: the pre-inst guile can be run with: meta/guile <peanutbutterandc>janneke, I did `env LD_LOAD_PATH=`pwd` ./guile` from inside libguile/.libs/guile but now it throws Pre-boot error; key: misc-error, args: ("primitive-load-path" "Unable to find file ~S in load path" ("ice-9/boot-9") #f)Aborted <peanutbutterandc>janneke, I skimmed through the README but it appears to only have a "using guile without installing it" section and not a "testing your changes" section.... <janneke>good; yeah it's all "easy" or "obvious" once you've seen it ;-) <janneke>you could also add readline to your environment: --ad-hoc readline (and reconfigure, make) <peanutbutterandc>janneke, You are a good teacher. :) Thank you for not pointing it out yourself. lol That leads me to another question: I tried to build guile from source using `guix environment guile` but the autogen.sh failed. The environment didn't contain automake and friends. Should guix environment have taken care of that for me? <janneke>"guix environment guile" provides all you need when buildig from a released tarball <janneke>there is, however, a guile package in guix that's built from git, and thus includes the autotools <janneke>that's guile-for-guile-emacs; it's a bit of a trick/hack; but you can do: guix environment guile-for-guile-emacs --ad-hoc readline <peanutbutterandc>janneke, I see. So for those of us trying to compile from git clone have to take care of our own... :( [You always help, sir. :)] <OriansJ`>peanutbutterandc: a silent channel does not mean no one is listening but rather they haven't something to say yet. <Formbi>would it be possible to use the current Guile in Guile Emacs? <janneke>Formbi: Guile Emacs is not generally usable <smileyface>Hi. The guix installer cannot detect my NVME drive. How can I solve this issue? <smileyface>Nvm. It is there in the manual installer section. Strange that the guided installer didn't detect it. <roptat>are you trying the latest iso, or 1.1.0? <smileyface>Thanks I will try it. Should the updated installer be able to remove an existing LVM partition on the drive or should I remove it beforehand? <roptat>it should be able to do it, but I'm not sure, i never tried myself <smileyface>Any way to check signature/checksum of latest image? <roptat>it's built by the build farm, not by a maintainer, so I don't think it's signed <smileyface>OK. I guess it confirms packages during install anyways? <roptat>actually, the build farm signs the image, but I don't know of an easy way to check it <smileyface>It is still not there when I select either "Guided - using the entire disk", "Guided - using the entire disk with encryption" or "Manual" in the guided installer. It shows up if I change TTY and do lsblk. ***terpri_ is now known as terpri
<guixy>I'm looking in the source for how guix defines the gexp shorthands <guixy>Where are these shorthands defined? <Formbi>but I'm curious how would the speed look like on a newer Guile *pkill9 wants a replacement for firejail that uses guix environments <lfam>I made a working kernel config for 5.7 on x86_64. But I wonder how to create viable configs for other architectures? <lfam>It's not going to be feasible for me to virtualize them all <jackhill>lfam: cool! Yeah, that's an interesting question. Call for testers I suppose? I also only have x86_64 hardwre currently. <lfam>I wonder what the person who used to do this did. They would handle 32 and 64-bit Intel, as well as 32-bit ARM <lfam>We will test the update on the build farm but I first need to make the config file <roptat>smileyface, can you send a bug report then? to bug-guix@gnu.org, describing your problem <lfam>I guess I can try `make ARCH=i386 oldconfig` ***familia_ is now known as familia
<noob31>Hello, it seems that "guix system vm" is ignoring my file-systems list in the operating-system definition. <noob31>(How) can I get additional mount points to put data on without having to <noob31>mess around with --share, 9p and rewriting the run-vm.sh every time to get security_model=mapped-xattr? <noob31>I'd like to keep using "guix system vm" (i.s.o. vm-image) to be able to use the Host's /gnu/store. <lfam>noob31: If you want to use a vm-image you need to use the command `guix system vm-image` <lfam>The `guix system vm` command creates lightweight immutable virtual machines that don't have any storage of their own, so the file-systems configuration is obviated <lfam>But, you won't be able to share the host's store with `guix system vm-image` <lfam>So, you'll have to choose between "share the host's store" or "configure file-systems" <smileyface>roptat: Hi sorry for the late reply, was AFK. I am currently formatting the drive to see if it works after that. Either way I will send a bug report tomorrow describing the issue (and whether formatting solved it) <noob31>lfam: ah, that's a pitty, I was so hoping to just get away with "run-vm.sh ... -drive file=data.qcow2..." :S <noob31>lfam: but thanks for the confirmation, I was starting to doubt what little knowledge I have about guile and worrying about my guix setup :) <lfam>In general, using a vm-image is pretty space-efficient because the qcow2 format uses a "balloon" mechanism <lfam>So you can tell it to make a 50 GB image, but it will only be as large as it actually needs to be <emacsen>I'm running the installer and when it gets to the partitioning, it throws me back to the beginning of the installer <noob31>yes, it's dynamic but from my previous experience with guix on a foreign distro, <noob31>I guess I'll look into --share some more <roptat>smileyface, thanks, I'm also on and off, so no problem as long as there's a reply eventually :) <noob31>hmm, it ate my "with my usage I tend to get some dozens of GB in /gnu/store that I was trying to avoid multiplying" <lfam>I guess it depends on what kind of system you are trying to virtualize <noob31>yes, some development boxes (over multiple weeks, for sure; but also with multiple versions of some dependencies...) <noob31>and guix gc is still something I haven't wrapped my head around enough to not having to re-download so much with the next guix pull again... <emacsen>okay, the installer appears just plain boinked :( <lfam>emacsen: I recommend using the manual process <emacsen>lfam, selecting the manual process throws me back to the beginning (language). Unless you mean using the shell as the installer <lfam>It's not difficult and it's more flexible <lfam>I'm not taling about the installer GUI <emacsen>okay... I mean I can do it. I'm a little surpised it's necessary but sure <lfam>Well, it's necessary since the GUI isn't working for you <lfam>In general, I recommend installing a very simple system at first, and then reconfiguring into the more complicated system that you aim to use. <lfam>Guix is very forgiving after it is installed, but the installation process is more fragile <emacsen>yeah... and the fact I have to keep the system plugging in due to wifi isn't making it easier :) <lfam>Right. That's a real pain point for Guix <emacsen>lfam, yeah I know it's not Guix's fault. It's just a bit of a high bar for a new machine. <lfam>It's really a shame. And as far as I know, there is nobody working on free drivers for recent wifi chips. So the problem will only get worse <lfam>Sooner or later wifi will be replaced by LTE and then the free software community will really have a problem <emacsen>lfam, I was talking to dustyweb about this yesterday. The OpenBSD song about it is right :) <lfam>I hope that some people can get motivated to work on it <emacsen>lfam, the sad thing is that this laptop runs coreboot but no free wifi drivers :) <lfam>Oh my! That is the kind of non-solution I find so frustrating <lfam>"Just don't use a computer" :) <lfam>At least wifi is backwards compatible so old chips will work with new access points <euandreh`>smileyface: came back just now. Now I see that you want a live USB. Is 'guix system disk-image' closer to what you're looking for? <emacsen>just out of curiosity... is brtfs actually usable? <butterypancake>Is there an accepted way to add kernel modules to a blacklist? Ideally I'd like to blacklist pcspkr from a system configuration file <butterypancake>In the guix cookbook it mentions some convoluted way of adding kernel options during the compilation of the kernel but that seems a little overkill and I'll admit it went a little over my head <guix-vits>butterypancake: kernel-arguments in Manual? (IDK, as usually :P) <guix-vits>There was some arguments to blacklist modules(?). <kmicu>emacsen: not only usable but even rock solid if you read its manual. <apteryx>emacsen: I'd say more than usable. I can recommend it. If you go Btrfs, makes sure to enable zstd compression for the performance boost and disk space gain. <emacsen>I also seem to be able to download the nightly installer now <emacsen>sadly the whole Linux disk system has changed and the whole nvme thing is new to me <butterypancake>guix-vits: It looks like the default kernel arguments are already a kernel module blacklist that I likely don't want to overwrite. The default list is set from the variable default-modprobe-blacklist so maybe I can do something with that variable? <rekahsoft`>Hi all, I had a question about console fonts. Namely, I do not have the fonts I use installed in my system profile, but wanted to have them only installed in my user profile, but it seems like this may not be possible. I say this because I can't seem to get certain characters to show up on a tty, but they do when I load X. To use a font on a tty, I assume I will need to install it in my system profile and configu <roptat>I'm trying to package the Android SDK, and one of the packages fails with "JavaLibrary.mk:127: external/icu/icu4j/adjust_icudt_path.mk: No such file or directory" when its Makefile tries to include another one that exists in another repository (a fork or copy of icu4j with a Makefile that lists source code, etc) <roptat>I think it wants the external sources to build them and bundle parts of icu4j in the result <GNUtoo>roptat: There is the option of cloning both repos in one package, but I've no idea if it's apropriate or not in Guix <GNUtoo>(I'm new to Guix, guiles and cheme) <guix-vits>rekahsoft`: You can try to simply configure the service. Services, afaik, aren't need their dependences to be installed to any profiles at all. <GNUtoo>It's (ab)used in other distributions like Arch Linux up to the point where they build adb in 1 package and don't have fine grained dependencies <GNUtoo>So in some cases it doesn't make sense to do that as the dependencies could be reused by other packages, but here I don't see another way, though someone else might have a better idea <guix-vits>rekahsoft`: i think You need to "modify" the service, as it's already present in %base-services (as console-font-service-type). <butterypancake>if you repeat a linux kernel parameter twice, is the value overwritten? Like if I specify "modprobe.blacklist=something" twice in kernel-parameters, what happens? <roptat>maybe I could change it a bit to not build the sources, and use our existing icu4j, but I don't feel like modifying the way android is built too much, as that will lead to issues... I'll probably just add the icu4j sources to the inputs and change the path to its Makefile <guix-vits>butterypancake: what about the module_blacklist=? *GNUtoo will send patches for android-libcutils and android-liblog soon-ish *GNUtoo has a slow pre-inst-env with things like that: WARNING: loading compiled file /home/gnutoo/work/projects/replicant/guix/guix/gnu/packages/zwave.go failed: In procedure load-thunk-from-memory: ELF file does not have native word size <roptat>you've compiled guix with a different version of guile than the one you use to run it <butterypancake>guix-vits: sorry, I'm not sure why you mean by module_blacklist= *GNUtoo uses guix official binaries in my system while I compiled with ./configure --localstatedir=/var (so it uses the system stuff) <butterypancake>I suppose I could use that. I'm not sure what the difference is. Since Guix uses modprobe.blacklist I'm inclined to think that it's more proper but I'm not sure <GNUtoo>The question would be where that modprobe\. stuff is implemented <GNUtoo>(in systemd? udev? gnu sheepherd? etc) <butterypancake>well the guix config is very simplistic. It just gives the kernel a parameter that looks like "modprobe.blacklist=usbmouse,usbkbd" <GNUtoo>so if modprobe.blacklist is documented in Guix, it's probably going to work <apteryx>I'd think it's just passed as a kernel boot parameter <GNUtoo>yes, the question is what recovers that afterward <GNUtoo>many userspace programs have access to /proc/cmdline <GNUtoo>So you can do write a custom daemon that gets stuff from there if you want <GNUtoo>like an init daemon that looks at "quiet" <GNUtoo>or whatever argument you put there (like nowayland, or whatever) <guix-vits>i thought that module_blacklist= should just make some module won't load, not matter what(?). <GNUtoo>systemd already recovers some arguments like quiet or systemd specific stuff <GNUtoo>so my advise is just to follow what's documented to work, else you probably need to go deep into the implementation details, though it would be interesting to understand it <butterypancake>well I don't see any documention on what to use is kinda the problem :P <GNUtoo>On Parabola I can put module.blacklist=radeon and still load the radeon module with 2 things: <GNUtoo>- systemctl start lightdm (or any other dm) <GNUtoo>So if I disable the desktop manager, it works fine <GNUtoo>This means that in Parabola at least, that is not done in the kenrel and it's done somewhere in userspace during boot <GNUtoo>So anything after that (like lightdm or a shell) can still load the blacklisted modules <GNUtoo>You could try it to see if it does the same in your case <guix-vits>GNUtoo: module.blacklist or module_blacklist? <GNUtoo>I never heard of module_blacklist, to maybe it does something different <GNUtoo>admin-guide/kernel-parameters.txt: module_blacklist= [KNL] Do not load a comma-separated list of <GNUtoo>I don't see any module.blacklist in Linux Documentation/ though <GNUtoo>So maybe they do something different <GNUtoo>static int load_module(struct load_info *info, const char __user *uargs, int flags) <GNUtoo>So module_blacklist probably rejects the module at the kernel level <butterypancake>well, it looks like you're correct that things look at /proc/cmdline after the fact. the modprobe.stuff isn't handeled by the kernel at all. modprobe queries it later. not sure what would query module.sutff <GNUtoo>So prevent any module named "radeon" from loading => module_blacklist <GNUtoo>Prevent the module to be loaded automatically at early userspace boot => module.blacklist <GNUtoo>I've patches in Linux so I hope I can <GNUtoo>(though the patches are usually only for small stuff) <GNUtoo>And reading C++ for instance is harder for me than Linux C style <butterypancake>I still wish I knew what would happen if I just specified modprobe.blacklist twice though... <GNUtoo>so it's a matter of time spent on things <GNUtoo>(If I was into C++ I'd have an easier time reading C++) <butterypancake>I'm very hesitant to experiment because this could break my system <butterypancake>C++ seems pretty gross but I sometimes I do wish I could do object oriented in C. <GNUtoo>butterypancake: you could try manually if it's not risky <GNUtoo>like edit grub menu manually eith 'e' at boot <GNUtoo>and press CTRL+X or F10 to boot after that as explained by the text on the screen <GNUtoo>Note that lightdm and similar programs have the habit of loading modules for you <GNUtoo>So the radeon case described was not a made up example, it's real <GNUtoo>they tend to load gpu kernel modules <butterypancake>I'd love to just use startx but apparently I'm not allowed to do that in guix :P <GNUtoo>What module are you trying to blacklist? <GNUtoo>oh ok, that's not a gpu, so lightdm shound't be an issue <butterypancake>ok, I'll be back in a bit once I've done some testing. Wish me luck! *GNUtoo whish you good luck <Formbi>butterypancake: I have that: (kernel-arguments (list "quiet" "modprobe.blacklist=pcspkr,snd_pcsp")) <roptat>now I'm having some issues with includes. Where is JniHelp.h? ***lle-bout_ is now known as lle-bout
<butterypancake>ya you can specify modprobe.blacklist multiple times and it just works. I did not need to go down this rabbit hole :P <roptat>ah, libnativehelper, parts of the SDK <GNUtoo>21:46 < Formbi> butterypancake: I have that: (kernel-arguments (list "quiet" "modprobe.blacklist=pcspkr,snd_pcsp")) <butterypancake>Formbi: you should probably add back in the %default-kernel-arguments variable as it seems to have a bug fix in it <butterypancake>it blacklists usbkbd and usbmouse as there is some race condition apparently <GNUtoo>Is it considered rude to add people in CC for patches? <GNUtoo>like the people that touched the part I'm patching? <GNUtoo>(I don't know Guix's community practices yet) <butterypancake>I would personally appreciate it but I'm really new to debbugs and I don't know the debbugs etiquette either <butterypancake>Once I forgot to CC someone that previously reviewed the patch and I think it took longer because of that :P <GNUtoo>(It's in common use in Linux, but in my experience community practices differs in different communities) <GNUtoo>(Though too long --cc list are considered bad in Linux) <butterypancake>I suppose I shouldn't have assumed everyone would use the tool the same :P my bad *GNUtoo stopped keeping track of which community use what rules and justs asks <GNUtoo>In some cases people asked me to reply to all while in other I was supposed to reply to the mailing list only for instance <GNUtoo>And the issue is that people being wronged often assume that the way they do thing is "the" way to do it <GNUtoo>whereas in reality it's "the" way to do it in that community <butterypancake>I think it partially depends on the volume of emails. In emacs-devel you reply to all parties involved since they might miss it in the torrent of emails being sent to the list if they gotta check the list <GNUtoo>indeed, it depends on how people are used to handle that huge volume <GNUtoo>If they want to ignore the reply or not <GNUtoo>which isn't always obvious to me <GNUtoo>It's not a big issue but I prefer things to go smoothly if possible <butterypancake>I usually just remind myself that I can only control myself, not how other people react, and just go for it without too much thought. If it was important then someone would have written it the contibuting section of the manual <butterypancake>Did a quick read of the contributing section again just to refresh my memory. No email etiquete mentioned <GNUtoo>I guess it's mostly the standard one with the usual small variations I mentioned *GNUtoo will be back later <butterypancake>GNUtoo: ok there was a little email etiquette mentioned. When doing a patch set like it looks like you did you should try to only use 1 debbugs number by sending the second email to the number of the first. I'm sure no one will be too upset but I thought I'd mention it <GNUtoo>tissevert: some build time dependencies are not needed at runtime <GNUtoo>For instance you might need a special tool to compile a package which isn't needed to run it <tissevert>yeah, which is why you expect to have both sets of packages separated in two variable, wouldn't you ? <GNUtoo>so if you just want to install an already built package, you don't need the build time dependencies <GNUtoo>ah sorry I missread your question <tissevert>when I was using arch a long time ago, AUR scripts had a variable for build dependencies, and another for runtime dependencies <GNUtoo>As I understand, Guix has reproducible builds, so it tracks the whole dependency tree <GNUtoo>so for a given pakcage binary you can get that whole dependency tree <GNUtoo>So there may be some abuse of language somewhere which confuse runtime and build time dependencies somehow <tissevert>so it finds itself from the content of the output what dependency is and isn't needed ? that sounds like being able to fix the halting problem… : S <GNUtoo>as I understand inputs -> runtime, native-inputs-> build time <GNUtoo>but keep in mind that I'm ultra new to Guix <roptat>tissevert, inputs are the dependencies. Guix tracks runtime dependencies separately <GNUtoo>So I can make confusions and mistakes <roptat>namely, if the result has a reference to one of its inputs, it's a runtime dependency <pkill9>inputs = runtime, native-inputs = buildtime <tissevert>how does it know there's a reference ? does it parse the binaries to read all the execs too ? <GNUtoo>In git a hash to references the whole history before the commit and the commit, <GNUtoo>so here a the hash of a binary package reference the whole dependency tree (probably build time and runtime) <GNUtoo>This way you can have reproducible computations <roptat>if the result matches the hash of an input, it's a dependency <roptat>if it doesn't match, then the input is not needed at runtime <roptat>I guess this is confusing and I'm not making it very clear... <tissevert>roptat:I'm relieved, it's a «no false negative, maybe false positive» solution <roptat>there could bbe false negative, if the reference is mangled <tissevert>I was just scared of the halting problem, if the solution is just a simple heuristic in works <roptat>like if it's in a compressed file <tissevert>yeah, nothing stopping a mad developer from including the string of a package it wants kept <tissevert>even though it's never called, and fool the dependency tracker <tissevert>(no, I'm not planning to do that, don't worry ^^) <roptat>or, you could trick the dependency tracker into thinking there's no dependency by having something like "/gnu/store/abc" + "def...-my-dep/" <roptat>that's also the reason why we don't compress jar files when building java stuff for instance <roptat>in case it has a reference, we can't have it compressed or we won't recognize it <tissevert>thanks for all the explanation, I really need to make thought experiments beyond the normal scope to understand how a thing works <tissevert>what's the best way to ask a question about a specific package, here ? shall I mail the maintainer, or ask if they're around here on IRC ? <roptat>now the difference between native-inputs and inputs is whether in cross-compilation you would use the native or the target architecture for the package, which usually matches the runtime/buildtime difference <roptat>here is best, if you don't get an answer, ask of help-guix <roptat>ok, libnativehelper works, but I still have missing headers for libcore... <tissevert>I'm wondering if there's a reason the ibus-daemon script doesn't also run dconf-service in package ibus <roptat>no idea, maybe because nobody thought of that? <tissevert>great ! that would mean I could learn and modify a package : ) <tissevert>I don't know, maybe it's my setup still being shaky, but since today I'm able to run ibus-daemon when I start my LXDE session <tissevert>but I can't modify a single parameter in ibus configuration (like, for instance, the checkbox pop back unchecked if I try and check one) <tissevert>I remembered ibus used dconf to store its config and I checked : no running dconf-service on my system <tissevert>I located it in /gnu/store and run it, just to see, it «fixed» my problem temporarily <tissevert>but I don't know if that's the right fix, I've seen something called «gconf» in ibus' derivation, I don't know it's a similar service, maybe people should just pick their favourite config-storing service ? <tissevert>so, I'd love to hear what is the right way to do this, and if it's only my bad setup or if there's something to fix here <tissevert>good night o/ (it's night and way too warm to stay up late here) <butterypancake>anyone here familiar with how udev rules work in guix? I installed libu2f-host which installed some udev rules to it's install location but I'm a little unsure if they are actually being used. I'm trying to get a U2F key to work. Does anyone know if U2F works through IceCat? I know it works through firefox <butterypancake>looking in the directory returned by `herd rules udev` shows me lots of sym links but none of those are the udev file I want to use. Does that mean I need to do something special to get the udev rules?