IRC channel logs
2023-02-18.log
back to list of logs
<dncrash[m]>Hello! Quick question: If I install openssh with guix install how could I get it started with herd? I don't see it in the output of sudo herd status and can't start it with herd start ssh <mirai>dncrash[m]: you need to add openssh-service-type to the services field in operating-system first <madage>I've just sent some emails on #61246, most importantly: python-pygit2 requires libgit2-1.4 <bjc>i do not know how i feel about the bot always changing its name. part of me thinks its adorable and cute, and another part thinks it shouldn't be countenanced <madage>also gitless just complained on test phase... <madage>hmm.. this is probably because pygit2 is one of its inputs.. <ehmry>I installed guix using the graphicall installer, and on reboot the new system it does go anywhere, the last message on the console is "Error: Driver 'pcspkr' is already registered, aborting..." <ehmry>the installer usb could boot all the way <lechner>ehmry / you can ignore the pcspkr message, but it should boot. what do you see? <lechner>dani-g5x[m]1 / which compiler, please? <ehmry>lechner: some messages about iwl wifi blobs, nothing that the installer doesn't also show, as far as I can tell <lechner>dani-g5x[m]1 / everything is in one file (which you can split up). transferring services from one computer to another takes two minutes and work every time <lechner>ehmry / can you switch to another virtual console with Ctrl-Alt-F2 or so? <ehmry>lechner: i didn't think so, I'm going antoher install and then I'll check it <mirai>*sigh* why does it look like a test for GDM wants me to rebuild things like webkit and an endless list of deps <ehmry>lechner: no, can't switch consoles <ehmry>lechner: no, the console messages are there, but then nothing happens <lechner>i think your GDM takes over and then gets stuck <ehmry>lechner: I had to boot with `modprobe.blacklist=radeon,amdgpu` same as the installer <nckx>apteryx: I've pushed the NSS graft. I feel a bit… hm. I'm going to keep an extra eye on problem reports tomorrow… <mirai>but on second thought I think the GDM tests should be kept (sans the tmpfs part) <lechner>ehmry / sounds good! you are an expert <ehmry>no, I have no idea how to configure the command line now <ehmry>lechner: yes, I'm doing an update? <lechner>you want to do a 'guix pull' at your earliest opportunity. are you logged in as an unprivileged user? <lechner>if so, just remember to use that executable from thereout out. root does not have access to is, so you have to use 'sudo' <lechner>anyway, it's a 'guix pull' followed by a 'guix system reconfigure /path/to/config.scm' <lechner>when a new "generation" does not boot, just select a previous one from the Grub menu <lechner>as insurance, you may also want to do a 'guix install git' and then source the profile per the instructions that follow <lechner>otherwise, you can get the config of a previous "generation" with 'guix system describe' <lechner>unfortunately, the sun is setting where i live and i have to go, but i will be back here to help you in about twenty-five hours, if needed. meanwhile, this channel has many lurkers who are far more capable than i. good luck! <apteryx>anyone else struggling making sense of audacity audio configuration? <apteryx>it doesn't seem to mesh well with pulseaudio, although it says it supports it <apteryx>it doesn't show up as a recording application in pavucontrol for example <mirai>but does it "get/put out the sound"? <apteryx>it shows some signal on the vumeter, but the recording signal appears null <apteryx>nevermind, I could make it record my default input <apteryx>but what I'd like to record is my system monitor (what I "hear") <apteryx>perhaps my question is more about why is the soundblaster exposed in such an odd way via pulseaudio. I can't find any means to record from AUX IN 2 for example. <apteryx>the best I've managed to do so far is crank up the Aux2 volume in the "playback" section in alsamixer, and I can hear it <apteryx>alsamixer also has a Aux2 input in the "capture" section, which I'd like to record directly from ideally. <mirai>tbh, I can't make heads or tails out of audacity input handling (this with pipewire+fedora) <mirai>I see some "nondescriptive" names <mirai>though I'm sure under either pavucontrol or any other tool the soundcard(s) show a lot more i/o options <mirai>if all you want to do is "record" some audio playback in the system, I'd use ffmpeg (or parecord?) <apteryx>I found that in record device, I can select something called "SB Audigy 1 [SB0090]: Multichannel Capture/PT Playback (hw:0,2)", and then this gives me 16 distinct recording channels <apteryx>now to find which one corresponds to Aux2 <apteryx>they all give me the same default mic input... <apteryx>ACTION puts this into the figure it out later bucket <old>How can I have gcc-toolchain for ARM <old>I'm trying guix build --target=arm-linux-gnueabihf gcc-toolchain <old>got: build of /gnu/store/9aj1pq9csbz5j3fwl18cxjhcksqkzvn4-gcc-12.2.0.drv failed <old>lots of compilation error <apteryx>I'd define it declaratively in a manifest or such <old>there's a way to specify target in specification? <apteryx>see cross-gcc in (gnu packages cross-base) <old>guix build -f my-gcc.scm <old>my-gcc.scm: ((@ (gnu packages cross-base) cross-gcc) "aarch64-linux-gnu") <old>weird that it does not work directly <old>e.g. I get /gnu/store/97a5d4n310ghx4c5wgiyy4cg6xdlxkf9-gcc-cross-sans-libc-aarch64-linux-gnu-10.3.0 with the cross-gcc <old>but guix build --target=aarch64-linux-gnu gcc-toolchain@10 does not work <old>I must be doing something wrong here, but the manifest trick will do <yasht>Hi guix, is there a reason that Mesa is kept on 21.3.8 and not updated to 22? <irfus>Hi yasht. No specific reason, AFAIK, other than that updating mesa would trigger rebuilds for too many other packages. <mala>are the guix folks looking for more machines for their substitutes/build environment? if so, what specs would they need to have? <AwesomeAdam54321>mala: Yes. Regarding the specs, it doesn't matter but having a CPU about as powerful as a Talos II is very appreciated <jpoiret>i think the consensus currently though is that it is more of an organizational problem, rather than CI being underpowered. Big updates like mesa are currently queued in a branch called core-updates which is merged into master periodically but not often enough <jpoiret>at the last guix days and on the MLs there's been discussions about moving to a feature branch system instead, which means that updates like mesa would be merged faster <mirai>there's tunable builds in guix <mirai>but can that be coordinated with this script? <nckx>mala, jpoiret: Some more (both powerful & reliable) aarch64 nodes would be nice, though. I'd dare say that CI is underpowered there, even if it's not the only problem. <jpoiret>ah, you're right. I always seem to forget about the "other architectures" <jpoiret>maybe i should buy myself an arm board, just to remedy this (or risc5) <nckx>RISC-V would also be nice, although I can't say if it's better to wait until the porting effort is further along. <nckx>(Assuming RISC-V prices follow the curve of major arches; maybe they don't.) <jpoiret>maybe the fact that very few people have the hw has been slowing the port down <jpoiret>efraim was talking about a ~150$ board at the Guix Days <nckx>jpoiret: I stopped short of suggesting a ‘shared’ RISC-V machine because maybe there already is one, I didn't actually check 😛 Maybe one of the porters is handing out SSH access. Maybe not. <jpoiret>I'd find it more motivating to have actual access to the board though <nckx>ACTION goes to poke the suspiciousl idle POWER9 box. <nckx>Oh, you mean actually sending/reimbursing it? Sure. <nckx>There are one or two, just not official. <nckx>Or another way, this is the official Matrix room for Guix, since it's bridged. <jpoiret>nckx: I meant just buying one for myself to explore it! and also be able to test other archs than x86 locally <jpoiret>in other news, do we know why core-updates evaluations on CI keep failing? <nckx>jpoiret: Oh, this is news to me. <nckx>jpoiret: You mean grep? Yes, that's weird… <nckx>‘building of `/gnu/store/cnx6nd7lh4kv4n4sazyw5iyjilpflbmc-grep-3.8.drv' timed out after 3600 seconds of silence’ <nckx>Where ‘silence’ is an empty log file. <nckx>Yeah, I don't think so :( <nckx>There's nothing suspicious about the evaluation log itself, it ‘legitimately’ fails due to grep failing. <nckx>‘guix build grep.drv’ on berlin is ‘waiting for locks or build slots’. I'll let it, but I suspect it will fail again within the hour. I hope to at least capture the output of this manual build then. <bdju>I ran `guix refresh` thinking it would show me stuff I've installed myself that was out of date, instead it's been going over an hour, exceeded a github rate limit, and half the lines are for things I'm not even familiar with. questionable defaults, maybe... <mirai>bdju: that computes for every guix package I think <bdju>oooh it can take a manifest file. neat. thanks. <nckx>You could pip—oh, or that :) <nckx>ACTION never got used to the manifest life; still pipes ‘guix package -I’ into stuff; is happy. <nckx>‘successfully built /gnu/store/cnx6nd7lh4kv4n4sazyw5iyjilpflbmc-grep-3.8.drv’ ☺/☹ <nckx>No errors to capture. Went through a full GCC build too, so a lot never made it back from any build node. <nckx>New evaluation forced elsewise. <jpoiret>i've been wanting to help with core-updates but only have my laptop with me, building all the bootstrap is not something I want to spend time on <nckx>Of course, I wouldn't wish that on anyone. <nckx>In better news, POWER9 is finally building things other than old ‘guix’ revisions. <nckx>ACTION wonders how many people still run Guix on such machines… <XADE[m]>How can I add a grub 'Menu Entry' that loads menu entries from another .cfg ?? <nckx>I think that would require adding ‘configfile’ support to ‘menu-entry’ first, XADE[m]. <mirai>turns out the GDM test suite currently in guix is not deterministic <nckx>gpg: signature packet: hashed data too long | gpg: read_block: read error: Invalid packet <nckx>ACTION wonders what that's about. <mirai>how can I mount this guix image <mirai>-drive file=/gnu/store/k73rzs1d4grhzl4qq00ifc966vrqncxm-disk-image,format=raw,if=virtio,cache=writeback,werror=report,readonly=on <mirai>it was one of qemu's parameters <nckx>mirai: Look at the -P option of losetup, since you can mount only file systems (partitions), not whole partitioned devices. <nckx>So losetup -Pf, I'd say, but untested. <mirai>eh, ok but that just seems to configure /dev/loop0 ? <mirai>what next? it won't exactly mount using 'mount'? <bjc>with ‘-P’, you should have loop0p[0-9] as well, for the individual partitions <nckx>I expected it to create /dev/loop0p1. <nckx>Does fdisk -l …disk-image show partitions. <mirai>looks like I completely missed that loop0p1 <XADE[m]><nckx> "I think that would require..." <- np <XADE[m]>Can I have both 'grub-bootloader' and 'grub-efi-removable-bootloader' ? <XADE[m]>So that i can boot from both legacy and uefi <nckx>Sorry, I don't know (I doubt it, not out of the box, without writing custom code), but the past days here have suggested it might be a good idea to support multiple bootloaders. <bjc>i'm pretty sure ‘operating-system’ only has a slot for one bootloader <bjc>but past that, won't writing to the mbr mess up gpt? <nckx>Yes. ‘The operating system supports multiple bootloaders’ means you can choose exactly 1 from n. <mirai>XADE[m]: I'm curious about this. Why do you want both legacy and uefi? <platoxia>The Guile manual it references '/usr/local/share/guile/site/math/bessel.scm', where would this be in Guix system? <XADE[m]>mirai: I'll be installing it on external drive <XADE[m]>And will be booting on different boards <nckx>bjc: How would that happen? <bjc>i'm looking it up, rather than asking. i'm not very familiar with how gpt actually works beyond a high level <nckx>XADE[m]: If you're familiar with installing GRUB, you can always install the ‘BIOS’ one manually. As long as it loads the grub.cfg maintained by Guix it will see new generations. The manually installed GRUB not being updated shouldn't really be a problem IMO. <nckx>bjc: To me, this would also imply that you can't use GPT on non-UEFI systems, but maybe I missed your point. <platoxia>You can us GPT on BIOS-based systems...I'm doing it now. <nckx>XADE[m]: If you're not familiar, it's a bit involved (you need to create a tiny <= 1MiB ‘BIOS boot partition’ for the GRUB to hide in). <platoxia>The Guix installation instructions are pretty clear on how to do it. <bjc>so gpt reserves the first 512 bytes for mbr, allowing for some amount of backward compatibility <nckx>It's in the inimitably-named ‘Keyboard Layout and Networking and Partitioning’ section of the manual. <mirai>XADE[m]: and these different boards don't support EFI? <nckx>bjc: Enough to look for the rest of GRUB elsewhere, yep. (I didn't know it was 512 TBH; would have guessed 440.) <bjc>it's because of traditional sector sizes <bjc>i'm not sure how this works with 4k-only disks, but if you're using those, you probablo don't care about mbr at all <mirai>are EFI-less boards common? (in recent hardware) <nckx>I don't think sector sizes play a role today beyond the historical reason why the MBR is 512-n bytes long. You can still read 440 bytes from a 4k-native drive 😉 <XADE[m]><nckx> "XADE: If you're familiar with..." <- Got it <bjc>it's not the reading i'm concerned with, it's the writing <bjc>if you write a “sector” with 512 bytes onto a 4k disk without emulation, you've just clobbered a bunch of data <nckx>This is entering hypothetical space far beyond installing GRUB on Linux though. <nckx>Close the door before we all get sucked out. <mirai>nckx: I've stripped the useless GDM test in hopes to salvage something out from the test-suite in v2 (#60756) <oleander>Hello, already asked this a few days ago but I still can't find a solution. My terminal apps are displaying question marks instead of the "€" character in GuixSD. Since this does not happen in other distros, is there something that I'm missing? I'm using the DejaVu Sans Mono font. <futurile>if you cut and paste in the euro symbol does it display? <apteryx>where is RUNPATH documented in the GNU documentation? <apteryx>ah, info ld covers at least the respective arguments <futurile>oleander: you're using a terminal under X Windows right? <futurile>oleander: so something like XTerm, or GNOME Terminal etc <oleander>futurile: I'm using Sway with Alacritty. <futurile>oleander: do you get the same problem using a different terminal apps (e.g. Kitty) <oleander>yes, I already tried with Kitty and I get the same problem. <futurile>oleander: did you try setting to a different font? I'm not a Sway user unfortunately. <oleander>futurile: I've just tried with Fira Mono, same problem. <futurile>oleander: do you have problems with other unicode characters like 💩 - or just the Euro symbol - I was thinking fonts, or sway right. <oleander>futurile: just the € symbol. #$&@! are displayed. <futurile>oleander: and just to check - if you go to an emoji site (like the one mirai linked above), cut and paste in some of those. That will check if you're getting the full fonts capabilities. <futurile>oleander: if it's not that, the only thing I can think is that that your locale is messed up and impacting it somehow. Though beyond me, honestly. <oleander>futurile: The home page of that website is not displaying some characters correctly. <apteryx>is a RUNPATH entry supposed to be equivalent to a LD_LIBRARY_PATH entry? <futurile>oleander: OK, if it's not just the terminal then that would point that you don't have emoji fonts installed. Do a search for 'utf-8 fonts sway' and see if that helps at all. <oleander>futurile: I get a few little rectangles instead of the correct character. <oleander>futurile: I'm using this locale: (locale "en_US.utf8") <futurile>oleander: not your locale then. I've had problems with that, so that's what made me think of it. <futurile>oleander: I reckon trying to get some of the emoji fonts installed and then see if your browser and terminal will be able to display utf-8 symbols generally. <oleander>futurile: I didn't know setting the US locale wouldn't display the € character in terminal. <futurile>oleander: it changes the key map, but it shouldn't prevent you cut-n-pasting in a euro symbol. So that's why I'm confused. If you didn't have UTF8 in some form it would make sense that you wouldn't see the euro symbol (and other utf-8 symbols). Because $@# etc are ascii. <futurile>oleander: so now if you're going to an emoji site and you don't seen emojis. Then it seems like you might be missing the emoji fonts, or some Sway setting that allows utf-8 to display. <XADE[m]>Is there a way to only generate the grub.cfg and skip the boot loader installation completely <snihil>Hi guix! I was trying today to run cgit with a custom .css file, but realized I don't know a good way to pack that file inside the package. I have tried inheriting the original package and adding one more step after the install step, but I think this somehow overwrites the custom phases that cgit already defines. Is there a smarter way to do this? <mroh>XADE[m] try something like this "(bootloader (bootloader-configuration (bootloader (bootloader (inherit grub-bootloader) (installer #~(const #t))))))" <oleander>futurile: I've tried with another locale that is compatible with the € currency and installing the "unicode-emoji" package but the € character is still not displayed correctly. I have to do some research on the sway part. <XADE[m]><mroh> "XADE try something like this "(..." <- Will do thanks <apteryx>can tor be used as a "private" domain name service? <apteryx>say, I want to expose my own machine substitutes to a friend, without going through the hassle of setting up a proper DNS entry? <apteryx>does it work well? (have you tested it?) <AwesomeAdam54321>apteryx: I haven't tested it, but there's an experimental guix-gnunet branch <AwesomeAdam54321>By now GNUnet works fine, but I need to update the packages in Guix first <pukkamustard>apteryx: tor can be used for what you want. you might also want to have a look at yggdrasil which provides you with an IPv6 address that can be accessed by other people in yggdrasil. <apteryx>IPv6 is barely used for home internet in my country, it seems <apteryx>I'll study Tor and GDS, thanks for the pointers! <snihil>OK, I've resorted to patching the original css file from cgit and adding that to the `patches` field of the the `origin` object. <apteryx>madage: hi! I just pushed the update to python-pygit2 with f1234c7569088531862c84d19ca6350a44ea5aa5. Thanks for the heads-up! <efraim>I think the datefudge build failure on i686 and armhf on core-updates is due to coreutils-9+ and 64-bit time_t. Rebuilding a whole bunch to find out <nckx>jpoiret: At last, it seems to be evaluating. However, I did not find the root cause ☹ <apteryx>nckx: is julia currently broken because of the libgit2 update? <apteryx>ah, I see your "fix build" 0fbff222c49b1807fae2089b1462fa6e9aefb79f commit. <apteryx>does 'guix copy' copy the transitive references of a package? <apteryx>it'd be pretty useful if it didn't :-) <nckx>I think ‘guix archive’ does not by default, but it's more low-level (and I think it can still be asked to). <nckx>So it can get confusing if you, er, confuse the two. <apteryx>hehe, thanks, perhaps that's the one I had on mind <VesselWave>Hello, how to use shepherd for user sevices without sudo? <VesselWave>error: connect: /run/user/1000/shepherd/socket: Connection refused` <nckx>And is your user shepherd running? <VesselWave>srwxr-xr-x 1 user users 0 Feb 18 14:46 /run/user/1000/shepherd/socket <apteryx>do you know if other distributions post-process their debug files with something like dwz to compact them? <nckx>VesselWave: Eh, bad ownership would've caused EPERM anyway. Is there a shepherd process (it will show up as ‘guile’ unfortunately, so search full command lines) currently running as your user and listening on that socket? That error implies there isn't. <nckx>apteryx: Bit of a tangent (I don't know about other distributions) but would suggest not using the odd and non-standard [AFAIK] DWZ format if something well-supported like --compress-debug-sections suffices for your needs. Maybe it doesn't, though, since you say ‘post-process’ specifically. <nckx>Of course, being fallible flesh-bag, I'm biased towards the person I know. <mroh>interesting thread, indeed! thx! <nckx>I'm resurrecting some more old patches for upstreaming. Building a kernel variant with make-linux-libre* now does something very strange: it builds in CONFIG_AHCI for no discernible (dependency) reason. <bjc>what's the reasoning behind dwz? saving disk space or something more esoteric? <antipode>AwesomeAdam54321: In which repository is this 'guix-gnunet' branch? <antipode>I didn't find it in the Guix or GNUnet repository. <bjc>cuz it seems like a lot of effort for a negligible change in disk use <nckx>Memory as well I'd say. The storage argument could be partially answered with transparent file system compression. <nckx>I don't think it negligible, but I don't have data. <bjc>that was my thinking, too. fs compression seems a much better idea <nckx>Hence why it's probably not the (only) point. <bjc>hence why i'm asking ;) <bjc>in terms of memory, i'm reasonably sure debugging symbols don't get mapped into the process space <nckx>If only mjw were around. <nckx>OK, I figured it out: (kernel-config … "6.0") is #f today because the matching aux-files/6.0….conf no longer exists. <nckx>This feels like it should be fatal? <nckx>Also implies that my original plan would add maintenance burden for lfam, hmm. <nckx>apteryx: I had no intention of blaming you for the (very minor) libgit2 breakage. Should have posted it off-list. Just wanted to point out a possible bug in your scripts.