IRC channel logs
2022-12-04.log
back to list of logs
<nckx>sneek: later tell civodul: Thanks for the ‘make cuirass-jobs’ hack. <sneek>nckx, podiki[m] says: can you check the irc logging? seems it stopped making new dates and may have stopped completely? <nckx>That's what the above goggles-bot peekabooery was about, which has resumed the logging but I don't know the root cause. <lxsameer>hey folks, I ran guix repl via `guix shell -D guix -- guix repl --listen=tcp:123123` and connected to it via geiser, and then evaluated a buffer containing an `operating-system` expression and I got this error message, what could be the issue https://dpaste.com/B5Z8TFT6K <jlicht>lxsameer: Not sure if relevant to the issue you're having, but why the `guix shell -D guix --'? <lxsameer>jlicht: well i installed dependencies using `guix install` first but when I tried to use geiser, for some reason it couldn't find guile-git <lxsameer>unmatched-paren: ok let me try without guix shell <jlicht>but that might just be my geiser setup; it freezes consistently when trying to send any expression to the geiser-connect REPL <lxsameer>unmatched-paren: I've installed guix using the install script on my gentoo box, and I checked out guix repo and build it as well (did not install it) and I use that checked out path with geiser, could it be related? <PotentialUser-30>Hi,me again :-) I have guix in foreign distribution and after guix pull the checkouts are in /tmp/user/.cache/guix/checkouts/ Is it possible to configure guix that way that it uses another path ?? <zamfofex>Hello, Guix! Would it be sensible to have home services for desktops? I currently run ‘startx’ “manually” (through a script), and I feel like it would be nice to have a home service for that. <PotentialUser-30>But then all other "consumers" of /tmp/user/.cache would move also? If i only want to move guix ?? No way i afraid ?? <efraim>oriansj: do you have a pool user already setup, and with permissions to mount that nfs share? <neox>Hi there ! I'm kind of new to GNU Guix, and I just installed GuixSD on my two computers. I have encountered no problem so far with my laptop, but for my fixed workstation that's a bit more complicated it seems, for the file-system point of view. My problem is that I have an encrypted (LUKS) LVM volume on top of a RAID array. I modified my /etc/config.scm to address that, but there is a warning during the reconf <neox>Well I don't know what relation there is between my problem and sqlite3 and compiler headers <jorge[m]1>Hola, estoy revisando el manual 12.9.9 Servicios de escritorio pero no estoy claro como actualizar enlightenment pueden ayudarme por favor ? <zamfofex>neox: I think that’s probably a message sent to you by antipode regarding a problem when you were last here. <neox>Well, I think there is a problem with device-mapping service. I see in my log a "erreur : system*/tty : variable non liée" (which I don't know how to translate, perhaps "not linked variable" <nckx>(neox: Don't repaste on my account, because I'm not able to help, but please use a friendlier paste service like paste.debian.net in future.) <neox>nckx, why a "friendlier" ? <neox>This paste service is free software <neox>And hosted by a french free software organization <nckx>It returns HTML when I tried to curl it, and seems to want a heavyweight browser to run code on my machine… <nckx>Yeah: ‘ Error decompressing paste, due to missing WebAssembly support.’ <nckx>It wants to run code on my machine. <neox>nckx, oh.. Yes this is a privatebin instance, meaning a zero-knowledge pastebin with e2ee <neox>So yes, definitely free js code <nckx>neox: JS code would be less bad, because it can be read by humans. <reyman>another thing i could try is rust 1.65 because deno_crypto say build success with 1.65 nightly <nckx>Anyway, sorry for the tangent. Carry on. <nckx>Just be aware that some folks won't see you paste or bother to decrypt it. <neox>nckx, this is actually code (I mean it). Not obfuscated and you can read it... <nckx>‘variable non liée’ would be ‘unbound variable’ (so, undefined). ‘Shouldn't’ happen… <neox>Thank you for the tangent, this is something I easily forget <neox>> ‘variable non liée’ would be ‘unbound variable’ (so, undefined). ‘Shouldn't’ happen… <neox>Okay... Any idea how I can debug that ? <nckx>neox: It's WebAssembly though. It can be disassembled, but that's not the same as easily read. <nckx>neox: If you paste the logs + config elsewhere I can take a look. <nckx>I've compiled wasm out of my Firefox, it's not going to happen 😉 <nckx>I don't see an obvious unbound system*/tty. Which version of Guix is this, neox? (‘guix describe’) <reyman>i try to call rust-1.65 into native-input: but variable is not recognized, strange, because rust-1.65 exist in llvm <nckx>reyman: There's no exported variable called ‘rust-1.65’. <nckx>You can use (@@ (gnu packages rust) rust-1.65) as a quick hack, then later discuss exporting rust-1.65. <reyman>into native-inputs : list `(,rust-latest "rust") ?don't work <nckx>The old style of inputs would be `(("rust" ,rust-latest)) — but just use ‘(list rust-latest)’ now. Note that Guix does not provide a ‘rust-latest’ variable though. <reyman>i define (define-public rust-latest (@@ (gnu packages rust) rust-1.65)) <nckx>That is also what the only firefox package I'm aware of does, but for an older version: (define-public rust-firefox (@@ (gnu packages rust) rust-1.61)) <nckx>reyman: cargo is an output of rust. <nckx>You'd use (native-inputs (list … `(,rust-latest "cargo") …)). <nckx>One does not imply the other, so you'd need both: (native-inputs (list rust-latest `(,rust-latest "cargo") …)). <reyman>i try that (native-inputs (list (rust-latest `(,rust-latest "cargo") clang-toolchain-14 git-minimal ninja gn ccache ))) <reyman>but that return this same error : Wrong type to apply: #<package rust@1.65.0 gnu/packages/rust.scm:649 7f7e68a56b00> <nckx>→ (rust-latest …) means ‘call rust-latest as a procedure’. <nckx>(list a b c), not (list (a b c)). <nckx>‘Call as a procedure’ == ‘apply’ in Scheme/lisp, once you know that, the error makes sense. <neox>nckx, I'm using guix a4d52f0 <neox>I commented the file system part of my configuration and changed the mountpoint and I'm able to mount the lvm volume from shell <lxsameer>where do you store your GUIX_PROFILE env var? in the profile file? <apteryx>is anyone else sometimes hitting "org-src--edit-element: Marker does not point anywhere" when C-c ' in org-mode, on a src_block? <lechner>indieterminacy[m / nckx / Hi, i am surprised fosshost stuck it out until now. they generously hosted part of lintian.debian.org for a few months, but i had to move away because based on private messaging they became wobbly in April 2021 (which sadly wasn't that long after they started) <nckx>lechner: Yeah, they have been making questionable decisions for a while. <nckx>lxsameer: Depends on why you want to set it, I guess. <nckx>neox: I forgot my kernel doesn't have dm_crypt support :o) so of course it failed here. Building one now. <nckx>lxsameer: Why are you setting GUIX_PROFILE? <reyman>hum nckx, i relaunch the compil but this is rustc and cargo v1.60 that are used in definitive <nckx>lxsameer: Thanks. But why? <reyman>export PATH="/gnu/store/6l1pbfzfd5wgyazx4gbsgqgp0p43kn8m-rust-1.60.0-cargo/bin <nckx>lxsameer: So that the variables adapt to subsequent profile updates without re-sourcing? <lxsameer>nckx: I'm still trying to understand how guix works, but without those values I won't have access to the stuff I've installed <nckx>lxsameer: Is this on a foreign distribution? <nckx>Not my area of expertise, but Guix should add an /etc/profile.d file for those… <nckx>Is there one? Is it not being sourced? <nckx>I guess you could just copy lines 511-538 manually. <lxsameer>nckx: should I source that file in my profile? <nckx>As to the ‘where’: that's partly a choice, partly depends on how your distro Does Things. I'd use /etc/profile.d/guix.sh for cleanliness, and add code to your /etc/profile to source it. <nckx>There are nicer ways and uglier ways to source it, but as long as it works (and you don't add it to .bashrc), there's no real ‘wrong’ answer. <nckx>(Adding it to .bashrc will break guix shells — don't.) <nckx>ACTION → dinner party; good evenings all. <neox>Have a nice day nckx, thank you <mbakke>proot-static hangs during the check phase on all my machines ... what am I doing wrong? <lxsameer>I source `/etc/profile` in my ~/.zprofile and I have `zzz-guix.sh` in my `/etc/profile.d/` dir, but still guix is not available until i source the default profile and the current one, is that normal? <civodul>mbakke: we seem to have substitutes though, weird <sneek>civodul, you have 1 message! <sneek>civodul, nckx says: Thanks for the ‘make cuirass-jobs’ hack. <podiki[m]>is there a reason why a package wouldn't get a share/doc/pkgname/ copying or license file? I thought they all do <lechner>lilyp / thanks for the pointer to leaf! now i use it, too (while working through its rich options). it's very clean, and totally superior <civodul>podiki[m]: it could happen if no relevant file is found, or if the build system lacks the phase that copies that file <civodul>the 'install-license-files' phase, that is <lilyp>kinda late to start using it now with use-package slated for upstreaming, but you're welcome <lechner>lilyp / should i stay with use-package? <lilyp>if it works for you, why not? <lechner>i am not sure i ever liked it that much <lechner>lilyp / are you thinking about switching? <lilyp>perhaps I will, I'll have to look at emacs 29 then <podiki[m]>civodul: hmm I guess meson-build-system doesn't...? let me check <zamfofex>civodul: Regarding patch 58654: You seem to have added wld and swc, but not velox. Was that intentional? <lxsameer>I don't have anything to set the PATH and INFOPATH in my zshrc and those values are set by gentoo <johnabs[m]>Hi all, I just wanted to reach out about an issue I posted on the trackers website. I've tried replying to the thread twice, but it has yet to appear either time and it's been weeks. I tried by responding via email and via the little chat box on the website itself; is this a known problem? <permcu>Hi, guix pull errors "In procedure fport_write: Connection reset by peer". Is this a connection problem on my side? Or a more general problem? <podiki[m]>johnabs: I don't know that the box on the website works (it was disabled before, but is back?), email to the bug number should work though, where did you send it exactly? <johnabs[m]>podiki[m]: The bug number is 58789, the box is there for me still, and I found why my email failed, I'm pretty sure I clicked reply on the wrong email address 😅🤦♂️ <johnabs[m]>Maybe the box was disabled when I tried to use it? <podiki[m]>it may be back but still not work, which deserves a bug report :) <rrs>Hey all! I'm trying to build MLIR infrastructure, but clang-toolchain can't find libz.so.1 during build despite cmake finding it at config time. It's a fresh Guix install on top of Ubuntu 22. Here's a paste with package details if useful: https://pastebin.com/8QxAPANU ... Tried a few things at the suggestion of good people over at SystemCrafters <rrs>#guix ... specifically making a shell but same result. Anyone saw this problem before? <zamfofex>I don’t know if that’s the issue, but when you say “a fresh Guix install”, do you mean you haven’t run ‘guix pull’ yet? <civodul>zamfofex: hmm no, not intentional, sorry; i'll have a look after dinner <rrs>I have done `guix pull` and even rebooted in the meanwhile <zamfofex>civodul: I hope you can enjoy your dinner! 🙂 <viivien>You know how you are supposed to clean your house fully every year or so? Well, let’s guix gc my system that I unattendedly-upgraded every day for about a year :) <viivien>You know how you stare at an open fire during the cold winter days? Well, you know how I’ll be watching the guix gc output <johnabs[m]><podiki[m]> "it may be back but still not..." <- Oh, should I report it normally, or is there a different location to report website bugs to ? <viivien>(it’s going to take a loooong time, it’s only at [0 MiB] deleting … the fifth item) <johnabs[m]>unmatched-paren: You got it, thanks for the clarification :) <Obikawa>Hello. I am trying to stop GDM from autorunning. I have (services (cons* list of services (modify-services %desktop-services (delete gdm-service-type)))), but GDM stubbornly keeps showing after boot. Do you have any ideas, please? <panosalevro>are there any tips on how to get a proper resolution with nomodeset? <Obikawa>unmatched-paren: Thank you. It makes sense - I have seen a tip to remove /usr/var/gdm on Reddit, but it seems the path had changed. /var/lib/gdm is untouchable - "Device or resource busy". I am now figuring out how to break through this one. <Obikawa>Could it be, that Guix executes GDM when it evaluates the architecture test in desktop-services-for-system? And the missing directory merely prevents GDM from running? <Obikawa>Never mind , yes. It would be executed on guix rebuild. <mbakke>does 'guix build --no-grafts --check proot-static' work for anyone? if so, what is your kernel version? <mbakke>note: you can cancel the build if it prints no output for a minute or so <unmatched-paren>could someone who has worked on the website please tell me how you tested your changes? <mbakke>unmatched-paren: guix shell -m manifest.scm, then 'haunt build && haunt serve' <unmatched-paren>mbakke: OH I didn't know about serve, i've never used haunt before :) <mbakke>the last successful tests depends on the number of cores on your machine ... I'm guessing you have 12? :) <mbakke>ACTION has run that test on too many machines <mbakke>verified that it fails on ext4, f2fs, xfs, and btrfs <unmatched-paren>though that's virtual cores, not real cores, right? (that's a real distinction, isn't it...?) <mbakke>I'm fairly certain the hanging (actually looping) proot test is a regression with newer kernels, but not sure how to bisect the issue <mbakke>thanks for checking unmatched-paren :) <nckx>I have disabled the box of lies. <johnabs[m]>nckx: Oh okay, thanks xD I guess I don't need to submit that bug report then? <nckx>But thanks for bringing it up. <unmatched-paren>nckx: it doesn't seem like it'll be fixed anytime soon, so maybe we should just remove it entirely? <nckx>I think the current state is OK. <nckx>But yes, it has proven rather unreliable. <podiki[m]>it was disabled for a while (after I had reported this issue) but I guess came back on some update but not fixed <neox>Hi again people. How could I make a mapped-device depend on another mapped-device ? <neox>I have a LUKS on LVM on RAID, and it seems a race condition occurs <nckx>podiki[m]: I disabled it in August 2021, it was re-enabled in March 2022. Do you mean more recently? <nckx>neox: Not possible yet, I think. <neox>Okay thanks nckx. I'm not taking more time on this, then x) <nckx>Understandable. A bug report might help, though. <podiki[m]>nckx: I wasn't paying attention to when it was re-enabled so that must be it <nckx>I apologise for any mail lost (where ‘any’ is at least ‘some’, and probably more). <podiki[m]>hopefully it wasn't much more than the one person above and no amazing patches were left in the ether <nckx>I can't imagine submitting patches through a Web form. Still, I suggest we log (effectively journal) entire messages somewhere before it's re-enabled. <podiki[m]>I can't either, but it is a copy/paste and a mouse click away :) <nckx>I'm sure there's some pithy Law saying, effectively, ‘if you allow something, it will be done’. <nckx>Right. That was the one referenced (through a merge) in the comment. <podiki[m]>which suggests a fix was deployed but I guess it didn't work or was it not pushed to the server itself? <nckx>Why it has since broken, I do not know. <nckx>That part's for tomorrow :) <podiki[m]>did it used to ask for an email in submitting? otherwise one wouldn't get a response (tricky enough already as often cc's are missed in bug reports I noticed) <civodul>zamfofex: for some reason the velox patch isn't showing up in debbugs.el, so i grabbed it from issues.guix.gnu.org <nckx>podiki[m]: It was always a required field. <Obikawa>unmatched-paren: I solved my problem -> since the system installation I had a (set-xorg-configuration) service, which enables GDM on its own. I don't think I need it with Sway, so I cut it out. Thank you for your assistance. <lebowski>Hi guix! i have a question about lukesmithxyz-st package. Is it really okay to call "git submodule init" and "... update" during package building? I can imagine cases where diffrent kinds of proprietory code or malware may appear on user's computer this way. <podiki[m]>you can do a recursive git checkout though, but in the source field, after fetching sources there is no network <lebowski>well yeah, but when I as a user will install this package and for example wiil try to use built binaries, they will have network access <nckx>What binaries do after being built is their business. <nckx>Guix's wget package will download proprietary code and/or malware with the same efficiency & ease as any other wget package. <lebowski>still, this is bas (at least i see it so) - if there is a package which has a git submodule with malware and there a lot of packages thich have this package as a dependancy - they all will be infected <nckx>s/a git submodule with // <nckx>The git submodule doesn't matter here. <lebowski>okay, thanks everyone for the explanation <nckx>ACTION → sleepy times. o/ <unmatched-paren>mirai: you should use the modules field of shepherd-service instead of that (begin ...) <civodul>unmatched-paren is blazingly fast :-) <unmatched-paren>also, maybe use make-systemd-constructor for services involving sockets :) <podiki[m]>unmatched-paren has a paren to find and no time to waste! <lebowski>okay what if i use (recursive? #true) in git-reference but guix does not download submodule? <unmatched-paren>mirai: i think you also need to add user-processes as a requirement? not certain though, i'm mainly familiar with writing home services <mirai>unmatched-paren: it doesn;t change much <unmatched-paren>mirai: try adding a #:log-file argument pointing to "/var/log/radicale.log" to the make-forkexec-constructor <unmatched-paren>then, when you reboot into the previous generation after the hang, the stderr (and possibly stdout? not sure) should be in that file <mirai>I am using system vm for this <podiki[m]>for the recursive issue, change something in the hash field <podiki[m]>that will force a redownload; I forget how to phrase this exactly but guix doesn't see a change in the package definition and sees the already downloaded source <mirai>unless there's a way to put a service start in "disabled" mode <nckx>ACTION back, never mind. <unmatched-paren>lebowski, podiki[m]: i believe it goes something like "guix looks to see if there's a store item with that hash, and if there isn't downloads it" <nckx>lebowski: The correct hash is 1b0ligrswvz307bbx5jp8wnnqz52v5s4gcgakxy4i3jvccalm2if. <unmatched-paren>i vaguely remember seeing a mailing list (?) post saying "btw stopped by default services are now supported" <nckx>lebowski: Guix sources are mostly content-addressed. Since you're pointing to the wrong source (=hash), Guix will use the wrong source. <mirai>civodul: don't know if you got my message yesterday but how does the 'this-record' and the serializer field work within the define-configuration macro? I don't think I am doing anything wrong here? (https://paste.centos.org/view/96b48bf6) <lebowski>nckx: how u calculated new hash so fast? 0_o <podiki[m]>ah right, the hash of the source, not the package hash (e.g. you change the package definition then there is a new derivation and a new build, but will use the same source if the hash was found) <nckx>lebowski: I ran ‘guix build’ on the snippet you shared. <nckx>podiki[m]: There's an asterisk there, as changing the file-name will also change the behaviour, but yes. <nckx>podiki[m]: Well, not anything: people change the url and expect it to make a difference. <podiki[m]>funny cause I forgot about just this with the change to a recursive checkout last night, and then re-remembered <podiki[m]>right, change in the resulting file name/hash <nckx>Put the (munged) URL in the store file name, or hash it. <viivien>Is guix gc output accurate? It tells me it has deleted 740 GiB so far, but there’s no way that there was that much space dedicated to guix on my hard drive <nckx>It's a bad idea but both would work. <unmatched-paren>i see. i suppose that would cause a rebuild of *literally* everything though? <nckx>unmatched-paren: Yes, for no good reason. <nckx>The same file is the same file, no matter where you look for it first, that's the point. <podiki[m]>also with mirrors, I guess you'd use the "mirror://" form not the final url it lands on? seems tricky <viivien>nckx, are we back to the times when free anti-virus software would claim to clean thousands of vulnerabilities (or rather, would be able to clean these vulnerabilities if you purchased the pro version)? <nckx>We could, however, enforce a file-name. Having any default at all (foo-checkout) might have been a mistake in hindsight. <nckx>The real problem here was that lebowski forgot/was unaware of the file-name field. This package as-is would not be accepted in Guix for that reason, but there's no error message/hint. <mirai>unmatched-paren: ah, I see the option, it's auto-start? <nckx>viivien: Yes, correct. It's a ruse to sell Guix Pro. <mirai>though it doesn't seem like it helps as I suspect its crapping out on illegal scheme code <unmatched-paren>nckx: Only 100 GuixCoins. 20% off if you donate your machine to become a --hivemind-- *build farm* node. <mirai>whatever's happening inside that string-join is upsetting guile <podiki[m]>the change in source but not hash is perhaps obvious once you realize, but is a little tricky when you haven't seen it before. maybe a clear note in the source reference in the manual if it isn't clear already? <Andronikos>My package fails to build on configure phase since cmake does not find "CMakeLists.txt" since it is in a subdirectory called src. How can I fix this? <Andronikos>Other question. This package needs a long time to compile. Is there some cache or do I need to recompile it everytime? <unmatched-paren>Andronikos: obviously, it doesn't rebuild if there's no change in the derivation :) <podiki[m]>you can also use the -K flag to keep failed output, and you can go into the /tmp/guix-.... directory to mess around (it has environment variables used in the build environment) <nckx>podiki[m]: The more random ‘oh, and also…’ things we add to the manual the fewer people will read all of them. I wonder if my mandatory file-name idea has flaws (probably!), i.e., if we can make *Guix* do the work, not offload it onto each user. <unmatched-paren>ACTION away, bye \o :) also, there is no hivemind. the Glorious Master-Brain "Deep GNU" promises. <nckx>unmatched-paren: Damn right. <lebowski>nckx: oh i thougt about sending patch. is this package definition so bad? <nckx>ACTION turns off unmatched-paren's shock collar… for now. <lebowski>okay i need to add dependencies and will probably send a patch 0_o <podiki[m]>nckx: well at least the documentation should be clear what the source hash really means (maybe it is already, I haven't looked recently); but yes I agree overall there is only so much you can do <podiki[m]>packaging examples and tutorials where you hit common problems are useful for that <podiki[m]>reading all the examples (all 20,000 or however many packages we have) is great, but the process of problem solving packaging comes with experience...some tutorials of the real life trial and error and learning could be good starting points <nckx>I agree that ‘origin Reference’ spends a lot of wind on explaining how exactly content-hash is implemented, and none on what it means. <podiki[m]>we can distinguish between reference and a guide/tutorial and what one wants to know reading either <nckx>podiki[m]: <there is only so much you can do> Well, my point was that Guix-the-programme could help here. By refusing to accept origins without an explicit file-name. Could start with a warning. I think all hash-based confusions I've ever seen here were due to (1) having an origin without explicit file-name (2) building it (3) realising they wanted a recursive checkout, adding #t (4) boom. <nckx>So something like -checkout-recursive, I dunno? <nckx>Sorry if that doesn't make sense, I'm exhausted. Sleep attempt 2: commencing. Bye! <podiki[m]>no I hear you and agree better feedback to the user will help <Andronikos>Unknown CMake command "qt5_add_translation". Does someone know how I can fix this? <Andronikos>I think it is because qt-linguist is missing but will this teach cmake the command? <apteryx>does cross-compilation works for linux-libre? <apteryx>ACTION tries guix build linux-libre --target=arm-linux-gnueabihf <oriansj>I have isolated the sshfs mount point problem <Andronikos>podiki[m]: Tried that already. It misses some packages that probably has a cmake configuration it requires. <civodul>roptat: hi! should we cherry-pick the NLS update on 'version-1.4.0'? <oriansj>efraim: the reason was the mount isn't checked if set to #f <Andronikos>apteryx: Yep. After some research it is probably because of missing qt5-linguist. Guess I need to package that as well to confirm. <mirai>what is the ungexp of a (package-propagated-inputs package-here) <mirai>(map car '#$(package-propagated-inputs package)) "works" and shows package names as strings <apteryx>Andronikos: at least on Qt6 linguist appears to be part of qttools as podiki[m] suggested <podiki[m]>i've seen translation stuff in QT need lrelease (from qttools-5), but different error message if I remember <Andronikos>I fixed it. Now I have "util-linux package too old. lsblk does not support --json (used for disk <Andronikos> enumeration)" as well as "lsblk: failed to access sysfs directory: /sys/dev/block: No such file or directory" <apteryx>I'm trying ot build an old kernel version for an embedded board (4.9.11), and it errors at build time due to having headers from linux-libre-5.10.35 offering conflicting definitions <apteryx>where do the headers get added to the build? <Andronikos>podiki[m]: You were right. I needed qttools but forgot to use version 5. <apteryx>I guess my kernel would prefer an older version of glibc <apteryx>Andronikos: that's just because lsblk wants to see /dev/block or something, which doesn't exist in the Guix build container <podiki[m]>Andronikos: great! no more extra packaging needed :) (well for this part at least) <Andronikos>Oh wow. That is exactle the package I am trieng to package. <lxsameer>hey folks, I'm using `guix system` to build a customize image, is it possible to copy a file to a certain location while using `operating-system` ? <podiki[m]>see base services, special-files or extra-special-file <lxsameer>podiki[m]: thank you, I want a file to be copied to the user home dir <podiki[m]>along with local-file if it isn't from a package <podiki[m]>oh...maybe guix home then, but that's a different configuration <Andronikos>apteryx: Why do you not have a problem with outdated lsblk? I recently updated Guix. <FlaminWalrus>Anyone ever run into phantom packages? I just librebooted a T60 and am getting Guix set up, but somehow I've got SDDM running on tty7 without installing it either in my user profile or through my system config (paste.debian.net/1262886). The config is closely adapted from one on my desktop that doesn't pull in SDDM... <podiki[m]>that's the config you have reconfigured to, or are you still in the version from the installer (that may have had a desktop set up)? <apteryx>FlaminWalrus: that's not phantom package, that's guix substituting GDM for SDDM on systems where building rust the Guix way is not yet possible <apteryx>FlaminWalrus: but... T60 being x86_64 that shouldn't happen there, hmm <FlaminWalrus>Well, where's it coming from? It's not in base-packages, nor base-services, from what I could tell in Guile. I prefer to startx from my bashrc on tty2, rather than use a login manager <apteryx>via mrustc (GCC exhausts the 4 GiB of memory limit attempting to build it) <apteryx>FlaminWalrus: if you use %desktop-services, that pulls in gdm-login-service <apteryx>which in your case will get swapped to sddm-login-service <FlaminWalrus>I don't use %desktop-services though; I just append the stuff I use to %base-services... <podiki[m]>did you see my question earlier, did you reconfigure to this config and reboot? <podiki[m]>or is it fresh from the installer which had a generated config based on what you answer in the questions <podiki[m]>(or did you use that config in a manual installation?) <lxsameer>podiki[m]: I've decided to use skeletons <lxsameer>podiki[m]: but does this looks ok ` (skeletons `((".ssh/authorized_keys" ,(local-file "./authorized_keys"))))` <podiki[m]>I haven't used skeletons before, but the format looks correct <oriansj>ok since guix currently has a bug with sshfs mounts that don't allow them to be created with (mount? #t) but only #f what is the most guix way to mount them on boot? <podiki[m]>note that local-files get added to the store if I understand <podiki[m]>lxsameer: so if you are trying to do anything with secrets, you wouldn't want to do that. these are public keys I take it <lxsameer>podiki[m]: yeah, just want to be able to ssh to the machine later <podiki[m]>what's the build log it references say? and the path will be relative to the config file I believe <lxsameer>oh maybe .ssh does not exist for the user <podiki[m]>wouldn't by default, not sure how to handle that if skeletons doesn't already