<rekado>pkill9: oh, then just look at “glibc-dynamic-linker” in gnu/packages/bootstrap.scm <PotentialUser-21>Hi, I was wondering, is there a way to create an environment out of a specific set of channels/guix revision? <pkill9>also you can use --profile for guix pull, and then run `source /path/to/guix/etc/profile` to put it in your path <matt`>i'm having trouble with the btrfs module when compiling a custom kernel <matt`>i keep getting "kernel module not found "btrfs"" <matt`>i think my kernel config settings are fine, although there's always a change I've messed that up. i tried specifying "btrfs" directly in (initrd-modules ...) at the end in the operating system declaration <xavierm02_>has anyone built the new ungoogled-chromium? How long does it take? <PotentialUser-21>pkill9: inferiors don't see to really to what I want. Maybe a manifest with inferiors in it, <emacsomancer>rubic88: I was compiling on an X230 i7 w/ 16Gb RAM. Tried three times. <emacsomancer>xavierm02_: according to my current calculations, aleph null seconds <matt`>is there any way to see the default .config for the linux libre kernel? <matt`>i.e. file that contains OPTION=y,m or not set <matt`>i think i see it. it's in the aux-files dir <leungbk>when i have this and use the emacs package module in the haskell.scm file <leungbk>trying to install ghc-hindent gives -=: <leungbk>but without the extra arguments for #:modules, the recipe builds properly <leungbk>i feel like i've imported all the right modules in the arguments, so i'm not sure what's going on. i'd appreciate any help <anon321anon123>Quick search through reference manual and mailing lists didn't help <Gamayun_>anon321anon123: If the encrypted filesystems are to be mounted at boot, you can just add them in the same way as an encrypted root. <Gamayun_>Not sure if you want them defined but not mounted. <anon321anon123>Gamayun: so I can unlock them with only one password once for all? <anon321anon123>I installed for reference encrypted system via graphic installer, but I have to type password for 3 times <Gamayun_>Is it an encrypted root which has a keyfile for other filesystems? I seem to remember there was a way to do that. <anon321anon123>Gamayun: yep, that is what I am doing right now on my debian, I am looking for something like this in Guix <xavierm02_>anon321anon123: It's easy to reduce that to 2: put a keyfile only readable to root somewhere in the first drive that's decrypted, and then add this keyfile to the second drive with cryptsetup <xavierm02_>anon321anon123: But getting to one time is hard because the first time, it's grub asking for the password, and then to have a keyfile ready right after that, you have to add the keyfile to the initramfs, which is a bit more complicated <anon321anon123>Yeah, I've never seen grub asks for password, in my current setup I enter password only after grub finish its job <Gamayun_>anon321anon123: That is so you can have /boot encrypted too ;-) <anon321anon123>I tried figure out how to do this, because I want to use debootstrap, but I couldn't <Gamayun_>anon321anon123: The file-system declaration has a 'dependecies' field, so you can specify if another filesystem (holding the keyfile) needs to be mounted first. <anon321anon123>So, right now my Guix system have encrypted boot? Will it help against evil maid? <anon321anon123>Ahh, via config I can specify keyfile to decrypt partition, am I right? <Gamayun_>Hmm, I haven't tried with a keyfile myself actually. <anon321anon123>By the way, how can I find which options can I use in config.scm? This is main concern for my right now <Gamayun_>anon321anon123: In the manual under System Configuration. :) <Gamayun_>Maybe not. I think the most up-to-date manual will be the one on your system rather than the website one (might get refreshed more often now). <anon321anon123>Ok, will have to find how to invoke manual. Also, is only ext4, and I think, fat and btrfs only supported by now? <tune>should I guix pull as both user and root or does one cover both? <Gamayun_>tune: They are separate. Some people might use just one of them though. <anon321anon123>Yep, seems like service for wireguard is missing, at least in manual on actual system. Well, I guess I have to learn how to write my own, hope they are not complicated <anon321anon123>Or configuration interface, don't know how to name this properly right now <brendyyn>anon321anon123: they look complicated but everything basically comes down to putting the right config files on the hard drive and then running the program <brendyyn>although one would not put private keys in the config <tune>anyone know how I can change what time icecat thinks it is? it differs from my system time and it makes things like chat logs annoying <tune>ah it's the resist fingerprinting thing... <str1ngs>it can't track you if you time travel :P <quiliro>Ĉu kio oni devos uzi antaŭ 'sudo guix system reconfigure config.scm': 'guix pull' aŭ 'sudo guix pull'? <swedebugia>There is an error in guix scripts pack that breaks compilation. <swedebugia>no binding `zip' to hide in module (gnu packages compression) <rekado_>but there is a “zip” binding in that module <nckx>tune: Unless you have a good reason to pull as root, don't. sudo [-E] is a beautiful thing. <tune>I used to do sudo -E but it creates issues <tune>ownership of files changes and then next time I update as the user it doesn't work until I delete something, chown/chrgrp something else <tune>doesn't seem to happen if I just change to root <tune>I realize this is due to an issue but I've never managed to figure it out <nckx>tune: Can you be more specific? On ‘average’ systems, the only time you need sudo [-E] is when running ‘guix system reconfigure’. <tune>perhaps the issue is from the pull <tune>I'm still not really clear on how to update user/system despite being on guix system for 1.5 years now. I keep hearing conflicting stuff <tune>I mean I *am* updating the system, just clearly not the "right" way <tune>so just to clarify, I only need to guix pull as my normal user without any sudo, and then I can do a guix package -u or a guix system reconfigure afterwards? <nckx>Yes (with the usual ‘for the average use-case’ caveat). guix pull && guix install someproggy && guix package -u && sudo [-E] guix system reconfigure. <tune>cool, thanks. I'll update my aliases then <nckx>Marlin[m]: Just some ‘should I guix pull as root’ ‘probably not’ discussion as is mandated by law to occur regularly. <nckx>tune: Let me say for completeness that ‘su -; guix pull’ etc. isn't *wrong*, but it does imply that root is being used as a ‘full’ user that has its own packages different from tune's and the system-packages. Valid, but not how root is used on most installations. And as you've noticed, you then have to be very careful in keeping track of who's pulling & installing what. <nckx>Marlin[m]: guix pull *is* per user. <nckx>Hence the confusion, because people are still (and quite understandably) used to ‘Imma manage my packages, time to be root’ on other systems. <nckx>The only command for which *I* use sudo is ‘guix system reconfigure’. <tune>I do have 90% of my packages on my user, I'm just unclear on how to update the system part I guess. I think I get it now <tune>> hint: After setting `PATH', run `hash guix' to make sure your shell refers to `/home/brad/.config/guix/current/bin/guix'. <tune>does this mean I've broken something? <nckx>tune: No. It just means ‘run this command now to make sure the newest guix is used’. <nckx>We (guix) cannot affect the shell it was invoked from. <tune>oh, alright. I was worried it meant my path was screwed up, and I was wondering if that command should give an output or not <nckx>All new shells will automatically use the new environment variables. <nckx>tune: command -v guix (like ‘which’ but more reliable) should print /home/brad/.config/current/bin/guix now, if you do want to double-check. <tune>hmmm... I get something slightly different <tune>/home/brad/.config/guix/current/bin/guix <tune>I don't have a ~/.config/current <nckx>Sorry, stupid typo from me. <Marlin[m]><anon321anon123 "But how do I update system packa"> you from the thread? <nckx>Why I didn't just copy & paste we'll never know 😒 <tune>I considered the possibility but thought I'd stay on the cautious side <tune>yeah the quote cuts off short <nckx>It looks like this: <Marlin[m]> <anon321anon123 "But how do I update system packa"> you from the thread? <tune>I'm applying a manifest for the first time in a few months, I'd been doing imperative updates so I could roll back when something was broken <nckx>Which isn't horrible, but it gets annoying when there are 10 such lines on your screen and it's very not-IRC. <tune>I don't think I got that hint updating the other way though <nckx>tune: You can still roll back a manifest, I think, although I've never rolled back my user profile. <tune>nckx: oh I'm probably thinking of partial upgrades <tune>like holding back one package <tune>I don't think manifests support that <nckx>ItsMarlin: I'm surprised (s/&/disappointed) that your Matrix client doesn't give an option to disable that at least on IRC channels, given how it doesn't fit the culture at all. <nckx>The culture is to manually hack around the lack of context by manually quoting things and otherwise do manual work 😛 <nckx>But it's usally not needed and starting to read a line only to realise it's an old repeat gets old quickly. <nckx>tune: I've never needed to do that (if something breaks I just fix it, but that will turn on me one day when I urgently need an up/downgrade, I know). There's probably a way, and it probably involves inferiors, which is Guix's powerful but hand-wavy answer to everything. <anon321anon123>I couldn't find how well Guix containers provide security. Is it like docker/LXC, or it just enhanced chroot? Can I run possible malware at container? <nckx>anon321anon123: Containers don't provide security, everyone involved in their creation has always been very adamant about that. That includes LXC, although Docker is purely marketing-driven so who knows what they promise. <nckx>That said, guix containers are newer and have seen less testing so assume less security than that 😛 <anon321anon123>Hm, right, if I want security I would run VM or even malware on another machine, but still, LXC relying on some linux features like cgroups, namespaces, etc. Does Guix do the same? <nckx>Yes. Don't ask me exactly which ones (cgroups no; namespaces yes is all I know). <nckx>anon321anon123: There's a lot of room for improvement in how and how much Guix uses them, if you'd be interested in contributing to that part. <anon321anon123>Well, I am not a real programmer, but hope Guix would be kick start for me :) <nckx>anon321anon123: Nor am I? ☺ I hope it does! <nckx>dutchie: Which command could I use to reproduce that? <dutchie>guix environment guix --pure -- make <dutchie>hang on, maybe i broke it myself actually <nckx>dutchie: Yeah that error means you're trying to ‘call’ expat as a procedure… <nckx>I don't think we support that yet :o) <dutchie>thought the license field quoted a list for me *nckx just discovered they'd been using a completely incorrect syntax for the SSH authorised-keys form for years and it only just started failing today… <g_bor[m]>I was wondering about two things, as I was doing some packaging. <g_bor[m]>1. Is there any feed/ml I can get info when the master/staging/core-updates branches get merged. My usecase is that I would like to get notified when master is merged onto staging, so the I won't need to add the same commits to both, simplifying the merge, but can start working on the staging feature as soon as possible. <g_bor[m]>2. Is there a way to get guix download working for any origin? I was thinking something along the lines of guix download (package-source packagename), so that I don't have to repeat the download after a failing hash. <mbakke>g_bor: The 'guix-commits' mailing list may be of use to you. <g_bor[m]>mbakke: Yes, I am subscribed, and also watching it, but it is a quite high volume list, and I am always afraid that I miss something :) <g_bor[m]>Maybe I could add a filter to have merge somewhere collected, but it seems like suboptimal. <rekado_>g_bor[m]: “guix download” does not have an “-e” option, sadly, but I think it would be cool to have. <g_bor[m]>rekado_: I might have some time to have a look at that in the coming weeks. I feel this would be a really useful feature. <cap>g_bor[m]: Hey, how is the status about your config file? <g_bor[m]>here it is, I just pushed it so that you can have a look. <g_bor[m]>but as I said, there's nothing fancy going on there. <g_bor[m]>Once I setup my dotfiles, that will be more interesting. <wellton>Hi.. I use an T400 with libreboot and Guix System 1.0.1 =) But, two questions: 1) my bluetooth not works =( my lsusb returns this: Broadcom Corp. BCM2045B (BDC-2.1) ... I installed bluez and b43.. anybody can help me? 2) The libreboot not find the guix`s grub. anybody can help me?? <wellton>In other computer (other T400 with libreboot with PureOS and Linux-libre) the bluetooth works. <g_bor[m]>cap: one thing that is missing is the swap setup, as it currently does not seem to work with a uuid. I would like to fix that first. <g_bor[m]>It seems that upon swap activation the dev mapper links by-uuid are not up yet... <g_bor[m]>maybe putting swap activation after udev could work, but I don't know if that is acceptable. <wellton>Hi.. I use an T400 with libreboot and Guix System 1.0.1 =) But, two questions: 1) my bluetooth not works =( my lsusb returns this: Broadcom Corp. BCM2045B (BDC-2.1) ... I installed bluez and b43.. anybody can help me? 2) The libreboot not find the guix`s grub. anybody can help me?? In other computer (other T400 with libreboot with PureOS and Linux-libre) the bluetooth works. <rekado_>wellton: what do you mean by “not works”? What did you try? What error did you get? <swedebugia>My git is broken. Https error. I set all SSL env. variable that I know of but they seem not to be well documented in the git documentation. :( <swedebugia>fatal: unable to find remote helper for 'https' <- is the error. <rekado_>swedebugia: is GIT_EXEC_PATH set to $HOME/.guix-profile/libexec/git-core ? <rekado_>the only SSL variable that matters for git is GIT_SSL_CAINFO <swedebugia>rekado_, it seems this variable is not correct. Hope that helps :) <swedebugia>I finally got the recipe of grin to compile only to realize that none of the rust-1.xx toolchain substitutes are available :p <swedebugia>Anyone who knows how long time it would take to compile the whole guix rust toolchain on a 2-core 2.2Ghz celeron? <nly>(assq-ref %guile-build-info 'top_srcdir) returns "/tmp/guix-build-guile-2.2.4.drv-0/guile-2.2.4" <nly>building guile embeds the build-path <nly>this might be an issue for reprducibility and such *rekado_ restarted the build <mbakke>nly: Luckily the build container always uses the same path, even if the system is configured with a different TMPDIR, or has multiple failed builds laying around. <rekado_>swedebugia: you can check the build duration for past builds on ci.guix.gnu.org <rekado_>most of the build nodes are pretty old and weak, though probably still faster than a 2-core celeron. <rekado_>(we’re currently finalising the order for a replacement of all but two build nodes) <swedebugia>Maybe we could have an additional time measurement: Ghzminutes. E.g. if the build took 60 minutes on a 3Ghz machine it took 20 Ghzminutes. <swedebugia>The build node would have to communicate the total Ghz used for the build. E.g. if 2-core 3Ghz= 2*3Ghz. <xavierm02>Starting to build ungoogled-chromium. Wish me luck <cap>g_bor[m]: Thank you really mutch. <minall>how can i find how a module is constructed in guix....more specificaly, the gnu module that is loaded in every config.scm config file <minall>what is the most basic definition? <minall>i can see that the gnu module is made up of other modules <minall>but i would like to get to the "God" module <minall>For example, the gnu module is made up of= gnu build system and others, but how can i see what those modules are made up of? <nckx>minall: What do you mean by ‘made up of other modules’? <nckx>The ‘gnu’ module is entirely contained in gnu.scm. <nckx>If you mean (gnu packages foo), that's a different module gnu/packages/foo.scm. It doesn't inherit anything from gnu.scm unless explicitly imported. <saslibre>What we mean is that, gnu module is composed or uses other modules, what are those modules composed of? <matt`>I'm setting up a system with two luks-encrypted devices and btrfs on top of that. However, when booting I'm only prompted for the password to the first device. Grub then goes directly to trying to setup the btrfs filesystem and fails because it's missing the 2nd device. <matt`>Anyone have experience with this? <rubic88>xavierm02:ungoogled-chromium took me > 4 hours to build on my laptop (X220 w/ 16GB). Won't build with 8GB. <matt`>I define the encrypted devices with (define cryptnvme (mapped-device (source (uuid "uuid")) (target "target") (type luks-device-mapping))) <matt`>then declare them in the os with (mapped-devices (list dev1 dev2)), where dev1 and 2 are target names from definition <matt`>filesystem is defined with (define fs-root (file-system (mount-point "/") (type "btrfs") (device (file-system-label "btrfs")))) <matt`>whoops entered too soon on that last one... <xavierm02>rubic88: Gah. I only have 8G. It's at 47%. When does it stop if you only have 8G? <matt`>(define fs-root (file-system (mount-point "/") (type "btrfs") (device (file-system-label "btrfs")) (options (btrfs-mount-options "btrfs/guixsd")) (needed-for-boot? #t) (dependencies `(,dev1)))) <rubic88>xavierm02: Somewhere north of 70-80%, I recall. It might work for you, but it locked up my system after 2 attempts. <xavierm02>rubic88: By locked up, you mean you had to reboot or more trouble than that? <rubic88>It was the impetus I needed to upgrade ram, among other things like containers. <rubic88>xavierm02: Sorry, just meant I had to power cycle. *xavierm02 is thinking of bying more ram <rubic88>Just running mail on localhost for cron. Any changes I need to make to my config.scm to get it working? <nckx>xavierm02: What're we building today? <xavierm02>according to rubic88, my laptop should freeze in the next few minutes but I still have hope <bandali>rvgn, i saw your post on help-guix. i never used gnu gnash, anyone know if it even works? <bandali>the last release seems to be from 8 years ago, and the project page says “SWF v10 is not supported by GNU Gnash.” <bandali>i personally wouldn’t hold my breath :/ but if anyone’s able to help and package that soon, great, hope it works out for you <bandali>cheers rvgn, sorry i can’t be of much help <bandali>i too had to deal with a similar thing (our school uses some bullshit flash application for TA trainings) and i had to resort to using one of their lab computers that ran windows to do it <nckx>Oh god, I just read your mail. My (very sincere) condolences, rvgn :-/ <g_bor[m]>rvgn: I've also read it, and gave you a quick answer. <nckx>As much as I hate steering people away from Guix: if this really is time-sensitive I'd test in a Trisquel (or whatever) VM whether Gnash is even a solution. Then you still have time to come up with a plan B if it isn't. <g_bor[m]>yes, my answer also boils down to that... <nckx>Rather than waiting for it to be packaged in Guix; maybe it doesn't work; is this a packaging bug or a Gnash bug; no idea; let's test with another distro; and you're at the same point anyway with precious time lost. <g_bor[m]>I've been trying with gnash years ago, and sometime even then it did not work. The effort is much appreciated, but it is very hard to shoot at this moving target... <nckx>g_bor[m]: From moving to (almost?) dead with little in between, no? I thought Flash was X-( <rubic88>xavierm02: Hoping my prediction fails, for your sake. <rvgn>g_bor[m] I have replied to that thread. <nckx>Even worse that it's mandatory. <g_bor[m]>nckx: in happier parts of the world maybe, but here I still see it sometimes... <rvgn>nckx I thought of USB booting trisquel, but do not have any access to usb drives atm. :( <nckx>The last time I needed Flash was for my riding licence several years ago. I think I was just lucky. <rvgn>Not sure how to use VM in Guix. In trisquel, I use to download virtualbox, create a vimage/vdisk, boot from ISO and that's it. In guix, I don't know where to start. <nckx>Gnash played those animations just fine. <nckx>rvgn: Yeah, qemu is great but its command-line interface is very intimidating. <rvgn>nckx yeah. Is there any other VM managers? <nckx>It's partly deprecated syntax but ‘qemu -accel kvm -hda img.foo -net user’ should work (although I'm not 100% sure about that ‘user’). *nckx installs virt-manager to see if it's usable on Guix. <dongcarl>mikegerwitz: Did you end up revisiting that patch? <nckx>rvgn: Is it even in Guix? <nckx>It was the first thing I searched for but maybe I typo'd that bad boy. <nckx>rvgn: virt-manager needs the libvirt service or it won't even give you the option to create a VM. Still, might be worth a try if command-line qemu isn't your thing. <quiliro>how can i know in what file a module is defined? <nckx>g_bor[m]: I assure you that Belgium isn't a happier part of the world than Hungary, at least as far as software freedom (or even ‘not mandating obsolete shit’) is concerned ☹ <quiliro>for example where is (gnu packages) defined and how is it found it is in that file? <nckx>quiliro: (gnu packages foo) → gnu/packages/foo.scm. <nckx>So (gnu packages) → gnu/packages.scm <nckx>It's literally that simple ☺ <nckx>A module is defined with define-module. <nckx>For once, things are simple and perfectly logical in computing land. <Marlin[m]>quiliro you may also do guix search package | grep scm <rvgn>nckx So I have to enable libvirt service in system config, do system reconfigure, install virt-manager and create VM. Correct? <quiliro>but I cannot find module (guix utils) <nckx>Marlin[m]: That maps packages to files, not modules to files. <nckx>quiliro: ? guix/utils.scm <rvgn>nckx Cool! "service libvirt-service-type" should do right? <nckx>rvgn: I'd say that if it doesn't work out of the box with the default values that's a bug, so yes. <nckx>I don't see any parameter that couldn't have a sane default pre-set. <quiliro>nckx: i am using emacs-guix's guix-package-locations and guix/utils.scm does not exist...perhaps it is because utils.scm does not hold packages, just modules....but then how can i know where that module is defined? <rvgn>nckx Thanks! Any way, why the manual mentions libvirt daemon as "server-side" daemon? I can be run on any host right?, evein in home computers? <quiliro>dongcarl: everyone here is trying to help....it is not possible to ask a particular developer to do certain task <quiliro>dongcarl: time and knowledge are not readily available <nckx>quiliro: It's always as simple as concatenating the elements (here, ‘guix’ and ‘utils’) with slashes and ending in ‘.scm’. You don't need a function to do this for you. <dongcarl>quiliro: He's the one who posted that email, I'm just asking if he had more info. I'm sorry I don't mean to burden anyone. <quiliro>nckx: i did not find a way to look at definitions without emacs-guix <quiliro>nckx: i am not refering to combining guix and utils but to find the definitions themselves <nckx>quiliro: I think that call-out was unnecessary and inappropriate. <nckx>quiliro: I don't use emacs-guix, so I'm not immediately familiar with what you're doing with it. Do you mean that you don't know the ‘root’ of where these modules are? *dongcarl is okay because he has had enough coffee :-) <nckx>quiliro: I was referring to your response to dongcarl, not your questions. <quiliro>nckx: i don't think it is unnecessary to tell users that tasks are not done becasuse developers are not reminded <quiliro>in fact, i think it is very appropriate <nckx>It's perfectly fine to ping Ludo' about a 3-month old bug if it would unblock whatever Carl's working on. <nckx>Nobody's is saying or implying that. <quiliro>nckx: i think your call-out was unnecessary and inappropiate <dongcarl>Ah, so the situation was: 1. mike gerwitz posted a series of patches 2. One of them broke on Ludo's system so Ludo emailed about it 3. mike gerwitz is still working on that patch 4. I just wanted to see if Ludo had a tentative fix in mind when he emailed about the bug <quiliro>nckx: you are escalating an argument <quiliro>nckx: you are not solving any problem *dongcarl thinks we can agree to disagree :-) <nckx>quiliro: I was addressing the problem of your inappropriate response to a question that did not concern you. Instead of apologising, you chose to escalate. If you truly want to de-escalate, please do so, now, by ending this senseless discussion. <quiliro>dongcarl: i was just thinking of developer time as a precious resource we must administer wisely...do you think that reminding someone whom is busy helps tehm solve the bug? perhaps there are more important bugs...do you think it is a critical bug that may be important to remind a busy person? <nckx>quiliro: I'd rather help you with your module question TBH. <rvgn>nckx Should I need "service virtlog-service-type" as well? Thinking it could help with the logs if I face any issues? *dongcarl is gunna go back to work *rekado_ manually restarted the rust builds <quiliro>nckx: if you want to help you should address problems without using adjectives on people *rekado_ catches up on the backlog <nckx>rvgn: Mmm, from my reading of the description, it's a log aggregator which shouldn't be necessarry, but I'm no expert here. <quiliro>dongcarl: i guess you and nckx did not understand the point <nckx>quiliro: Please stop this. <nckx>rvgn: Let's keep things minimal for now ;-) *rvgn is doing system reconfigure to enable libvirt service. *quiliro feels offended and censored <rekado_>quiliro: it’s fine to ask people about bugs they were involved with. <rekado_>quiliro: your involvement in this comes across as needlessly aggressive, even if it may not have been your intention. <nckx>quiliro: At no point in time were you censored. I'd like to point that out just in case someone stumbles on incomplete logs in the future. I have no interest in feeding this fire. <quiliro>rekado_: time is scarce, bugs have priorities..... <rekado_>quiliro: let’s not repeat what has already been stated. <rekado_>so, moving on: what do y’all think about adding a “restart build” feature to ci.guix.gnu.org? <quiliro>rekado_: ok...so what is the solution? <rekado_>I wonder if restarting a build should automatically cause builds that depend on it to be restarted. *rvgn will be back after system reboot <pkill9>what would you restart a build for? <rekado_>g_bor[m]: your ideas about this are welcome! What do you think? <nckx>rekado_: Just so I'm not missing something obvious: this is for transient failures? <nckx>rekado_: You said that dependency failure wasn't tracked properly in Cuirass, this sounds very related. <nckx>(The restarting dependent jobs bit.) <rekado_>nckx: yes. The daemon has all the details, but it can be tricky to tease them out. <rekado_>is it worth implementing this at all? <nckx>It implies authentication & friends. <rekado_>some of us can, of course, log on to berlin and run “guix build the-failed.drv” <rekado_>what I don’t like about doing this manually is: 1) it’s done manually *nckx wonders if ‘helpful’ ddossers restarting failed builds are our friends in disguise. <rekado_>2) one needs to run “guix gc --clear-failures” the output first <rekado_>3) Cuirass doesn’t realise that things are fine now <nckx>rekado_: Yeah, my first thought TBH was ‘can't this be automated’. <rekado_>so the UI won’t display that a build has been restarted and succeeded on the second run. <nckx>Clicking a button is still manual. <rekado_>but we don’t know if something is a transient failure <rekado_>for some things it’s obvious: the log says the build succeeded but for some reason the build is marked as failed <rekado_>(e.g. because of problems transferring the results from the build node to berlin.) <nckx>A ‘build succeeded after n tries’ warning (with sane limits, and maybe a manual escap hatch) actually sounds pretty cool. <nckx>I know, I'm dreaming other's code now. <nckx>If non-transient failures are cached by the daemon it wouldn't be that expensive, I *think*… <rekado_>all failures are cached by the daemon <rekado_>building a derivation for a cached failure (even though the build succeeded on the build node) is going to yield failure. <rekado_>guess we’d have to make the daemon distinguish between different kinds of failure <rekado_>but that’s more than I’d like to chew on; gotta bite off something smaller <nckx>rekado_: OK, I was thinking only of download failures, but not full disc errors &c. *nckx uses --cache-failures *everywhere*, iz nice, and I've never had to clear them at unexpected times. *nckx tries to grok ‘even though the build succeeded on the build node’. <rekado_>berlin offloads all builds; build node builds things successfully, but crashes before sending the results back to berlin; berlin says “where my stuff?”, marks the build as failed. <rekado_>“crashes” could be anything. It used to happen more often than is comfortable on hydra.gnu.org as well. <nckx>Ooh… I admit to kinda ignoring that above. Is that so common? <rekado_>I think there’s a bug report about that sorta thing. <rekado_>if it’s just a bug, though, maybe we can ignore it in the discussion of Cuirass features. <rekado_>there’s no point in enshrining a workaround in the Cuirass UI. <nckx>That's crossing from ‘networks fail sometimes’ into ‘we obviously have a bug somewhere’-definitions of transient. <rekado_>well, we also have failures that depend on the state of the build node <rekado_>if something gets built on a tiny armhf board with another build running, the new build might fail <rekado_>run the build again later and it might get assigned to an Overdrive and pass. <rekado_>then there’s also non-deterministic build failures :-/ *rekado_ -> afk for a few minutes *nckx wonders whether (properties (gib-ram-per-core . 8)) would be worth its ugliness or not. <mbakke>rekado: Optional restart of all dependent packages would be a great Cuirass feature. <nckx>Wondering while wondering: can build nodes non-fatally reject jobs? <mbakke>nckx: I don't think so, short of going offline. <nckx>Hmkay, so that would add another layer of yuchiness to that. Manually specifying farm RAM on the head node. *nckx .oO Who the hell is poking around my substitute server from a Windows 10 box. <rvgn>nckx when I try to create new VM in virt-manager, I am getting an error "No active connection to install on." <rvgn>nckx But the libvirtd is running. I double checked with herd status. ***emyles` is now known as myles
<OriansJ>nckx: well if one plays with .wgetrc and sets user_agent.... <nckx>rvgn: I'm sorry you're having trouble ☹ I wish I knew more about *virt*, but I don't. <xavierm02>rvgn: Did you add yourself to the <I don't remember the name> group? <rekado_>nckx: what would be the intended effects of (properties (gib-ram-per-core . 8))? Is this supposed to be a hint to the build node (or the head node) to not attempt the build when there’s not enough RAM? <rvgn>xavierm02 Thanks! I was thinking the same thing. But don't know the name of the group. <nckx>OriansJ: It looks pretty human (& non-malicious, just a bit funny). Hey, more power to 'em, and I hope my friendly about page punts them in the right direction. <rvgn>virt or libvirt or qemu? I already added the user to "kvm" group. <nckx>Roughly by definition. I'm not saying its nice. <nckx>Like extending the timeout, it's a just-work-now kluge. *rvgn gonna try the group name "libvirt" as per legendary world wide web. <nckx>rvgn: Any virty-sounding groups in /etc/group? <nckx>TFM: (unix-sock-group "libvirt") <mbakke>nckx: I typically set --cores to (/ gib-of-ram 8) and then --max-jobs to roughly (+ 1 (/ actual-cores (/ gib-of-ram 8))) on build nodes. <nckx>Encoding RAM in cores. I hate it. I love it. <mbakke>I've intended to try tweaking this on the berlin build nodes, but wanted to get the shiny Zabbix setup up to snuff first. <rekado_>bah, the manually restarted /gnu/store/yssc91ymzz5rzryv13jr229dn6j25smx-rust-1.30.1.drv failed :-/ <rekado_>cargo:warning=../rustllvm/PassWrapper.cpp:1004:80: error: �thinLTOResolveWeakForLinkerInIndex� was not declared in this scope <mbakke>rekado: I think there is a patch on the tracker. It broke from the LLVM 8 upgrade IIRC. <bzp>help please, is there a program that is similar to volumeicon? i3wm use <bzp>and as a change of themes in gtk install lxappearance and the themes do not appear <nckx>rubic88 had the exact same question. <nckx>(About a systray volume icon.) <dongcarl>Is there a command where I can print the default list of substitute servers? *rubic88 is hoping volumeicon will someday be added to guix. May consider it as a project if he gets smarter about packaging ... <nckx>guix repl < <(echo -e ',use (guix store)\n(display %default-substitute-urls)') <nckx>oh wait you didn't want the ugliest possible command? <dongcarl>Huh... I thought hydra was also a substitute server? <nckx>dongcarl: Hydra was just retired. <nckx>rubic88: It's probably (just guessing though) a pretty simple package that should be pretty simple to package. <rubic88>nckx: I was thinking it might be a good starter project. *rubic88 Learning guile by working my way through The Little Schemer. <fishinthecalcula>I'm thinking it's probably some lib that should be added to propagated-inputs, since the package builds and runs correctly. <nckx>bzp: Do you really need an interactive widget? ‘bindsym XF86AudioRaiseVolume exec --no-startup-id $vol_up’ where ‘set $vol_up pactl set-sink-volume 0 +5%’ does everything I need. i3bar's volume icon provides immediate feedback. <nckx>rubic88: Ideosyncratic but awesome book. <nckx>rubic88: Phew. SICP is a classic. It's even in Guix (‘guix install sicp’) but you'll want the freely available videos too. <bzp>nckx: Can you help me activate the keys to increase and reduce the volume? <nckx>rubic88: There are also the Reasoned & Seasoned Schemer sequels but I haven't read them myself. <bzp>nckx: What file should I configure? <nckx>bzp: Ehm, the code above is literally it. I'll paste the lines from my .config/i3/config. <rubic88>nckx: I was thinking of the Seasoned Schemer next but was also checking for anything specific to Guile. <nckx>(I have my i3bar set to auto-hide and display when $Mod is held down, and $Mod+<volume key> plays a nice beep sound I got from the 'net so I can judge how loud is loud. <nckx>rubic88: If you find anything Guile-specific that's not the manual, I'd be interested to hear it. <nckx>A hack to make sure the beep is audible since pactl seems to have a tiny async delay that might very well be machine-specific. If I remove it, the beep sometimes gets played before the sink is unmuted, or one only hears the end of it. <rubic88>nckx: I found pairs were not covered in TLS, so I had to check elsewhere. *nckx honestly can't remember how or where they learnt about pairs o_o <rubic88>xavierm02: Did ungoogled-chromium work? <bandali>was super happy to see this episode with ludo :) just finished listening to it, it’s pretty great <bzp>Should I install something else to recognize the volume keys, the new configuration of i3wm? <nckx>bzp: pactl is part of the pulseaudio package, make sure that's in your profile. <nckx>(XF86foo is just the special key's scan code.) <bzp>nckx: how do I do that? <nckx>bzp: Probably ‘guix install pulseaudio’ then ☺ <nckx>Does ungoogled-**** do a Big Link at the end? That might be what uses so much memory… <xavierm02>It's only using 2G. And I don't think it went over 4G while I was watching.