IRC channel logs
2019-05-18.log
back to list of logs
<Marlin[m]>But i need amd's gpu firmware to have proper video <vagrantc>guix doesn't support proprietary firmware <samplet>Marlin[m]: Search the Web for examples. The policy on this channel is to not advocate non-free software, so you can’t get help here. <samplet>Well, you can get all sorts of help here, just not that kind. :) <vagrantc>i get the impression modern intel CPUs actually require microcode in order to even boot ... but i guess that's outside the scope of guix, since it is loaded very early in the boot process <samplet>vagrantc: I think the CPUs have the microcode built-in, so if you are willing to treat it just as hardware that you can’t update, it works with free software. At least, that is the rational why old ThinkPads “Respect Your Freedom” even though they require microcode. <samplet>AIUI, lots of devices have all kinds of non-free firmware. Right now, the FSF only takes issue with the kind that has to be explicitly loaded. (The rest is considered unfortunate but tolerable.) <vagrantc>i *thought* some of the newer ones *required* getting regular updates ... but i guess i don't know how that could even be technically enforced <samplet>It could be. I don’t pay much attention to new processors these days. I suppose you could have a two-stage thing where a small processor loads the microcode for the “real” processor. (I hope no one from Intel is listening!) <Marlin[m]>Aren't there any workarounds or recommendations for newer amd cards? I'd love to run guix on bare metal, but it linux libre always breaks on my pc <Marlin[m]>i'm mostly stuck with debian and the firmware package atm <Marlin[m]>and libre compatible hardware is way too expensive where i live, at least for desktops <fosterite>Hello, I did `guix package -i stumpwm` but it's not showing up in the desktop session list at the login screen. I see that the stumpwm package makes an xsession, but I don't know how to check if it's in the right place <Blackbeard[m]>and do guix system reconfigure according to the instructions there <fosterite>I will note that I have i3 installed and it's showing up <Blackbeard[m]>fosterite: if I remember correctly it will fail and ask you to add a few SBCL things <fosterite>OK, it's running. Why should I run it regularly if my config.scm hasn't changed? to use latest versions of packages? <fosterite>Blackbeard[m]: guix system: error: exception caught while executing 'start' on service 'user-homes', any idea what this could be? <fosterite>type ^l until you figure it out I still don't know <fosterite>Blackbeard[m]: don't worry about it didn't hurt anything and didn't happen next time I reconfigured. thanks for helping me with stumpwm <rekado_>Marlin[m]: instead of “clear” you can use Ctrl-L; “clear” is provided by the “ncurses” package, I think. <calher>Wait, Ungoogled Chromium is OK according to the FSDG? <brendyyn>calher: its ok according to guix developers <calher>brendyyn, are they violating the FSDG? <Marlin[m]>any way i can compile a different kernel on guix? <brendyyn>calher: thats what the debate is about. there is no proof that there is a violation, but also no proof that there isnt <calher>I'll use it to run proprietary web applications. <rekado_>Marlin[m]: yes, you can compile a different kernel and specify it in the operating system configuration. <rekado_>calher: if you’re interested (and have lots of time) you can read up all about this on the various mailing lists. It was a bit of a meandering discussion, and in the end it’s all about Russel’s teapot. <calher>rekado_, is that some sort of term related to Bertrand Russell? <rekado_>calher: yes, it’s about proving a negative, which I’m afraid this project won’t be able to do. <str1ngs>on GuixSD what provides the poweroff command ? herd? <rekado_>I don’t have a poweroff command on my system. <str1ngs>poweroff, shutdown generally are provided by the init system. so herd in this case <str1ngs>I think historically you use to have to pass arguements to shutdown. and poweroff just was easier to use if that's what you indended to do. <vagrantc>funnily enough, halt existed before poweroff and poweroff just seemed like more to type <vagrantc>but for some reason it gained popularity <str1ngs>maybe, but I rather be explict with reboot, poweroff then with halt and flags <str1ngs>herd's halt does not handle reboot either. maybe that's what you mean by no flags? <calher>Has anyone written a recipe for Telegram Desktop? <xelxebar>What'are the default login details on the prebuilt qcow2 images? <xelxebar>I'd assumed root is setup with an empty password, but I'm unable to log in. <calher>Looking through the Guix manual. What is the command to bring me to the package definition of a given package? <str1ngs>calher: if you use emacs-guix you can do M-x: guix <RET> p e <calher>What is a simple widow manager that uses XCB? <calher>str1ngs, thank you! That works very well! <str1ngs>calher:I think you can do this with guix repl. but it's more tedious <sirmacik>I can ssh into machine, signing-key is imported <str1ngs>sirmacik: did you setup /etc/guix/machines.scm with private key? <str1ngs> thoug, this seems to suggest ssh is working, since get gets guile version from offload server <str1ngs>seems to disconnect when sending files <str1ngs>can you check that guix describe match on both machines? <str1ngs>match the remote server with local machine with guix pull --commit=<hash> <str1ngs>though you can use channels too I think. <sirmacik>I'll try to pull and reconfigure guix on the second machine <str1ngs>this, might not fix that error. but it won't hurt either <sirmacik>why is that both machines running guix with pull from today are on diffrent commit? <sirmacik>str1ngs: ok, I've pulled commit but it didn't fixed the error <str1ngs>I don't why but the connect closes when it's sending files the remote server <sirmacik>if I have package.scm with multiple outputs can I automate somehow generating hashes for downloads? <cbaines>sirmacik, do you mean the package.scm contains multiple package definitions? <sirmacik>I'm trying to update it and see if it fixes bug I'm encountering but I'd rather generate hashes automatically then looking for urls by hand and so on <cbaines>guix refresh -u can update the package definitions for some packages <sirmacik>there isn't any guix command that will download files described in package.scm and output their hashes to put in file? <cbaines>there is guix download, but that requires you to give it the URL <cbaines>I usually just build the package, and then copy the correct hash in to the definition <sirmacik>hm, so how can I just bump the versions? <sirmacik>I did guix edi an saved the file in my ~/ <sirmacik>guix build doesn't let me build docker.scm file <cbaines>sirmacik, what command are you using? <cbaines>I think guix build does have a -f argument? That was what I was going to suggest? <cbaines>guix package also has --install-from-file=FILE (or -f) <cbaines>also, unless you're really intending to keep these changes separate from Guix, it's probably easier to work on this change within the Guix repository <sirmacik>butit gives me not something we can build <cbaines>first thing, have you got a checkout of the Guix git repository? <cbaines>it's up to you, both approaches will work, but if you make the change inside the Git repository, you'll end up with a patch you can submit <Marlin[m]><sirmacik "if I have package.scm with multi"> Where do you get these packages from? <cbaines>I thought it was going to maybe take another week, but I surprisingly managed to get all the lint tests passing this morning \o/ <janas>I'm trying to run guix from git using the ./pre-inst-env command, but the daemon fails because it can't find the zlib shared library <janas>I'm using the guix system, and I've double checked that I had zlib package installed, any ideas? <janas>nevermind, I think I found the problem <str1ngs>in emacs, how do I get geiser to use Guix REPL instead of Guile REPL ? <samplet>str1ngs: I’ve never tried, but it looks like you can do “guix repl --listen=/tmp/socket”, and then use “M-x geiser-connect-local” to connect to “/tmp/socket”. <samplet>There may be an easier way, but I’ve never looked into it. <str1ngs>sample: thanks, that should do the trick <str1ngs>not sure why I did'nt think to try that myself lol <str1ngs>samplet: hmmm guix repl does not have a ---listen flag <samplet>str1ngs: Huh. It’s in the manual. :/ <samplet>It says “--listen=unix:/tmp/socket”. <Marlin[m]>My laptop should be able to run guix, but the installer keeps failing <str1ngs>samplet: ahh and I've seen that before. I was missing the protocol so working now. thanks <samplet>str1ngs: That’s my fault (I forgot it in my first message). Sorry and you’re welcome! :) <str1ngs>samplet:the funny thing is I use this, when interacting with a web browser I created using guile. and I didn't think to do it for guix . lol <samplet>Ha! The extra funny thing is that I never bother using it! I only know that it exists. <samplet>I remember your browser. How’s it coming along? <str1ngs>my thinking was I'd reuse the Guix REPL that emacs-guix created. <str1ngs>it's coming along well. slow but surely. I know have a working guix package which is nice. <samplet>str1ngs: Do you have a public repo? I would be curious to take a look sometime. <str1ngs>samplet: feature-qt is the developer branch <str1ngs>samplet:I also have a substitue server for qtwebene. should be covered in README.org <samplet>str1ngs: Thanks! I will enjoy taking a look at this later. <str1ngs>samplet: it is fun to hack on though. <samplet>That’s okay. I am mostly interested in the source code. I find it interesting to see how people structure projects in Guile. <str1ngs>samplet: I'm using autotools, QML for the UI components. C and some C++ for QT. and then scheme were possible <str1ngs>samplet: it's highly inspired by emacs <g_bor[m]>I am trying to update ovmf. The proble is that the new version depends on openssl, bundles it as a git submodule, and insists on building in as part of the build. What would be the proper process here. I also have a similar problem with nominatim, which bundles osm2pgsl, and the build heavily depends on it. It seems that in these cases I can't get by without pathcing the build system if I want to unbundle. Wdyt? <nckx>g_bor[m]: The latest Borg does the same thing. I'm patching the build system, but it's not enjoyable. <g_bor[m]>nckx: Ok, I did dig into the source a bit, and ovmf needs a very specially built version of openssl. So it would be needed to path the build system in nontrivial ways, and create a special openssl package, with all the magic already implemented in the EDK II build, and that version is not useful for general use, so could not replace the system openssl. What is your recommendation? <g_bor[m]>Some of the main differences are some architecture specific flags, optimizing for image size, disabling syslog. <g_bor[m]>I believe it would be not so hard to define a version openssl with that... <g_bor[m]>EDK II is a very special build system, I am not into it very much... I guess for now I will send a working patch, and mark it as help needed. <nckx>g_bor[m]: I'm surprised (well, that's not the right word…) that patching the EDK build system is the most work here, not creating the openssl variant itself :-/ <nckx>In the end, you're the one doing the work so you get to decide. I'd be less weary of using the bundled library ‘for now’ (examples of ‘XXX TODO unbundle some day’ abound it Guix) if the library in question weren't openssl. *nckx is all for Doing Things Properly, until it amounts to telling volunteers to do a lot more work… <devilishtype> nckx hey get back to work while you're chatting you can't be contributing to the project >:| <nckx>devilishtype: I'm not contributing as it is; turns out that typing Scheme with one hand is much harder than telling people what to do on IRC. :'-( <devilishtype>oh man, this is going to turn into an interesting discussion if whether that's a paraphilia or not. *nckx knew the moment they hit Return that they had made a Mistake. <devilishtype>"I love guix maybe a little too much, I contribute to it one handed" <nckx>‘While clad entirely in leather.’ <nckx>(It was a motorcycle accident.) <nckx>What an enigmatic Web site. <amz3>I take that as compliment :) <nckx>(You might want to move that last div into the body, although it does add to the overal enigma.) <nckx>devilishtype: You don't? <Marlin1113>maybe i'd be able to run it baremetal with a custom kernel <str1ngs>Marlin1113: you can build a custom kernel yes. what specific hardware support do you need? <Marlin1113>hopefully they'll release the firmware source someday *nckx vascillates between between welcoming all new Guix users and pointing out that we won't provide support for that. <nckx>Marlin1113: It's that bad? It won't even work? <Marlin1113>It only will display stuff with low res, and won't show the gui installer <nckx>My previous laptop had AMD graphix — good riddance — but it was only 3D stuff that needed the blobz. <Marlin1113>i think amd is closer to being a fully libre modern gpu <str1ngs>yes, it's looking that way. though as you know some cards I guess don't work <nckx>Marlin1113: Oh, that's actually being discussed? I thought you were making a ‘like that will ever happen’ joke. Cool. *nckx does not play games or mine cryptoes or simulate genomes so doesn't really know the first thing about GPUs. <Marlin1113>their ray tracing 3d render engine is free software i think <Marlin1113>it just needs a repo, but the source is available when you download it <Marlin1113>although opencl on blender still requires proprietary drivers <Marlin1113>umm, does anybody know the default login on the qemu image for guixsd? <nckx>My personal history with AMD is such that there is a lot of lost trust to rebuild, but if it's true that they're slowly opening up that's great news. <Marlin1113>the amdgpu driver is libre, is just the firmware that isn't still <Marlin1113>i think it's just the time it takes to get people to maintain repos and change the licensing <nckx>Marlin1113: Sure, amdgpu is even in Guix, but IIRC it had to be blacklisted to even release 1.0 because even using KMS for the (text-mode!) installer required the blobs. <Marlin1113>the Vega and newer cards also use Rocm (radeon open computing) and the opencl standard, instead of icd <nckx>That might just be a bug, but still, it's not a great sign that the open part ends as soon as you want to modeset. <Marlin1113>which basically means the open drivers will have newer versions of opencl or something <nckx>Marlin1113: Interesting. Again, opencl, not my world, but from what I've heard about it any amount of openness is an improvement already. <Marlin1113>best part is how they put Linux OS on the graphics <nckx>Marlin1113: Have you tried ‘root’ with an empty password? <Marlin1113>it says that somebody got 3d working with the ATI drivers on a card like mine <nckx>I wonder if they even use the card at all or something like LLVMpipe (if that's still a thing). <Marlin1113>hey nckx, how can i start a wm from the cli on guixsd? the vm isn't booting to xfce anymore, maybe it's because i installed xmonad <nckx>It might not run Blender (or it might), but it might run GNOME. <nckx>Marlin1113: Errr, ‘startx’, is what I've always read, but I've never tried myself. I'm old and grumpy and just use SLiM + an .xsession. <Marlin1113>i'll try it, stuff is weird when you aren't used to the init system <nckx>Gotta rewrite all the things in Scheme. <Marlin1113>hmm, seems like startx fails because how guix isolate packages <nckx>Of course a search returns the tantalising snippet ‘Running startx complains about a missing X, since xinit looks for X in its /gnu/store directory instead of my profile one. Anyone knows how to fix…’ <nckx>…which points to a 404'd IRC log X( <nckx> Return a startx script in which the modules, fonts, etc. specified in config, are available. The result should be used in place of startx.’ <nckx>but it's not as ready-to-go as it probably should be. <nckx>bt`: ‘Probably’, and thanks for reporting that. *nckx guix pulls && guix builds thermald <nckx>(Your system will/should throttle itself or even shut down even without thermald, but it can help enforce lower limits [=longer chip life and a happier lap] or more user control.) <nckx>(Also, ‘should’ means your system is well-designed which is always a daring assumption.) <devilishtype>computer on your lap? Why aren't you sitting at a desk with a 32 core machine! <bt`>At current when using Guix my system just bakes until it crashes if I close the lid. It's not a issue on other distributions, just some strange interaction between my librebooted x60s and Guix. Along with my graphics issues which I still haven't figured out. <nckx>bt`: I've never had that issue with Intel systems but, yikes, my sympathy. <bt`>thanks nckx, I'm sure I'll figure it out eventually. <devilishtype>did you check to see if it's seeing the lid close event? *nckx assumed this was a strange if deliberate choice, but good point. <nckx>The number of ‘clam shell’ laptop users always astounds me. <bt`>I can check dmesg I know the screen goes black at the minimum. <devilishtype>that's the other thing on hte list for first things I'd try <devilishtype>if the screen goes off... then it's probably getting the event. <devilishtype>or does the backlight go off from some hardware switch in newer laptops? <bt`>Nothing showed up in dmesg but I've got thinkpad_acpi because I was testing something. I didn't really want it to suspend/sleep just because I use a cron job as a alarm from time to time. <bt`>it's libreboot so very likely the hardware handles it. <bt`>it also handles screen brightness. <bt`>I've got thinkpad_acpi off* <bt`>Perhaps, I wouldn't doubt if the clockspeed wasn't lowering at all. <bt`>even with it that is. <bt`>I don't even remember, I think a dual core 1.6Ghz core duo though. <devilishtype>theis is where someone who knows more about guix is going to need to help cause I know there's stuff in systemd for it too <nckx>devilishtype: What stuff is that? Curious. <str1ngs>hmmm why won't docker images start a ssh daemon? it generates keys. this is really odd <nckx>devilishtype: the ‘stuff in systemd’. <devilishtype>you can change pstate with some scripts out of the box in systemd distros <devilishtype>Anyway it would be a dirty hack, but you could just turn the frequency down as long as the button event is getting triggered <bt`>I seem to remember something in my logs about not having a valid pstate or something like that, one moment. <devilishtype>I know one of my old laptops didn't work without ACPI for turning down the frequency <devilishtype>even though, you should be able to change frequency without it <nckx>bt`: Have you ‘guix pull’ed lately? I just built thermald, twice, on master. <devilishtype>nckx: it could be he doesn't have the modules for his CPU <bt`>that's interesting nckx, yes I have. <nckx>devilishtype: Sure. But a log was posted above with a build failure (format string mismatch) that I'd like to fix regardless. <nckx>Or, as it turns out, find out why it's non-deterministic. <nckx>I thought it was a cyber. <bt`>lsmod | grep intel_pstate returns nothing, do I just need to add it as a kernel parameter <bt`>I'll try pulling again nckx <devilishtype>I'm not 100% sure you need it, but I think.. you maybe do? <nckx>bt`: Thanks. You were building /gnu/store/dq6y934sls06g1pq23bxm826sv15fn7k-thermald-1.8, I built /gnu/store/khlgc59kplzgxs3w432gnspvhc3zj73v-thermald-1.8, so there's a difference somewhere. <bt`>actually in my /var/log/messages: intel_pstate: CPU model not supported <nckx>devilishtype: drivers/cpufreq/Kconfig.x86: config X86_INTEL_PSTATE bool "Intel P state control" <nckx>It's really a module there? <bt`>I think it's a core duo actually predates core 2 duo. <nckx>I'd be somewhat suprised. <bt`>do I need to do anything other than undisable thinkpad_acpi for that because it wasn't working with it enabled either. <nckx>devilishtype: A PC won't boot without ACPI… <devilishtype>I think there is like an diffent module acpi-freq or something <nckx>ACPI can't be modularised. <nckx>Are we talking about the same thing? I suspect not. <bt`>modinfo acpi_cpufreq: module is not available <devilishtype>then, worst case you can just lower it manually with a bash script. <bt`>sounds good, any idea on how to install it, do I need to rebuild the kernel with it as a module? <devilishtype>I'm pretty much a complete guix noob so nckx needt to chime in <bt`>looks like the module has freedom <nckx>devilishtype, bt`: Chill… acpi_cpufreq is just a plain boring old Linux(-Libre) module, nothing unfree about it. <nckx>Sorz. I was just sitting here following the conversation in the corner of my eye thinking ‘what… no… no don't recompile any… noo.’ <nckx>devilishtype: It's possible but the UX is not great ATM. <devilishtype>yeah just build the whole system it should only take 2 weeks ;) <nckx>I'm just a bit annoyed because I don't run Guix's kernel so can't say anything with certainty. <devilishtype>isn't it pretty standard lots of people out there I assume with old intel chips <nckx>My kustom kernel has it so I don't know. <nckx>Don't call my ThinkPad ‘old’. <nckx>So I just started the Guix System installer in a VM, and ‘grep CPUFREQ /run/booted-system/kernel/.config’ (that last file name is a free life hack for y'all on Guix) and it's built in, so (as I suspected) acpi-cpufreq isn't relevant here. <nckx>devilishtype: YES! I was literally just distracted by that warning in my dmesg. <nckx>(I run what some would call a heavily patched kernel or ‘a freaking mess’. And that ‘advice’ was the first post I read. Gee, thanks, no.) <dalz>But I also plan to sleep tonight <dalz>So far it's compiled from 1.25 to 1.30 <dalz>If I interrupt it will it start from 1.25 again? <nckx> devilishtype: I'm not denying or confirming that, just noting that X86_ACPI_CPUFREQ and X86_INTEL_PSTATE are both built-in to the Guix kernel so modprobing won't help here. <nckx>dalz: No, if 1.25 is ‘finished’, but it will start 1.30 from scratch. <devilishtype>I'm just saying I don't know why he couldn't change pstates if he has the module <bt`>It seems like most of these intel vulnerabilities don't effect me just because my CPU is so old, zombieload seems to only effect post-2011 cpus. <dalz>nckx: nice to know, thank you <dalz>It's about to finish 1.30 <devilishtype>bt`: run watch or something on cpuinfo and grep for MHz then close the lid and open it a few times <nckx>bt`: Might be interesting to see the output of cpufreq-info. <nckx>There are several factors that can affect the frequency. <devilishtype>I don't know what default settings are, I use sleep on lid close. <nckx>devilishtype: That is the default for me. <bt`>so it seems to have a few options <devilishtype>bt`: do you want it to underclock all the time or just when it gets too hot? <nckx>bt`: I'm 99% sure that ‘performance’ means ‘never scale’, for example. It's more naive than its name might imply. Try changing it to conservative or ondemand or anything else, really. <bt`>just when it gets too hot would be preferable <bt`>well at least I learned something :p <nckx>It's just funny because the difference between ‘ondemand’ and ‘conservative’ is so preciously subtle, and then here's this ‘performance’ thing that one might assume is ‘biased somewhat towards high frequencies’ but nope, it's oven time. <nckx>(bt`: That being ‘solved’, I'm still interested in your new thermald build, and could you run it again with ‘--rounds=10 --check’, please and thank you so, and you should probably set up suspend on lid-close if you want it.) *nckx AFK for a while because OMG does their hand hurt. <nckx>devilishtype: Are you my girlfriend. <nckx>devilishtype: Oh, and Guix ships with Linux-Libre 5.1.2, which is the version with the MDS ‘fix’. So you are ‘secure’. ′Yay’. <nckx>bt`: Thanks! I assume it will build fine ten time but who knows. <bt`>christ, slim is very unhappy that if failed on boot so apparently it decided to just pin one of my CPUs at 100% making throttling not happen... I'll be back in a moment <devilishtype>yeah you better not use that kernel then or your thinkpad will be a brickpad *nckx recompiles their CPU. <pkill9>mbakke: do you have any idea why ungoogled-chromium is failing to build?