IRC channel logs
2023-06-26.log
back to list of logs
<pmf[m]>Hey, I submitted an patch updating a package almost six weeks ago, and there's been radio silence since then. It's for the package praat, which I need for work; the version currently in the main guix channel doesn't even build, so I'd consider it rather critical to get this fixed. I've been using the version in this patch since I submitted it, without issue. What is the process for escalating patches? Should I forward it to the person who last mod <juliana[m]>You can CC members of the teams most affiliated with the code of a patch using the various functionality of the etc/teams.scm script in the Guix source repository. I'm not sure what to do if there's no clearly-affiliated team - perhaps the general mentors team? <juliana[m]>posting a polite request here has also proven effective in the past, i have noticed ;) <juliana[m]>that said while i can't merge a patch i'd be happy to offer a peer review <podiki[m]>anyone setting a pam limit? it has stopped working not sure when: i see the correct setting in /etc/static/security/limits.conf yet it hasn't taken effect <podiki[m]>e.g. set a nofile limit higher and yet ulimit -Hn still shows the original lower <podiki[m]>been using this for year(s) at this point, not sure what changed, anyone know? <pmf[m]>Thank you! I don't believe that there is a team assigned to this particular file, but a second pair of eyes would be greatly appreciated :D <juliana[m]>i have gotten absorbed in doing some admin work for my community's guix pubnix and am just now wrapping that up; one moment while i obtain sustenance and i shall get on that review <lilyp>juliana[m]: mentors should be CC'd if it's your first time contributing or you have questions that really need to be answered before pushing <sozuba>can i run -> guix system init /mnt/etc/config.scm /mnt again? I ran it already and the systemc completed installation,but it boots to grub rescue.Now if i made somechnages to config.scm, can i re run this command? guix system reconfigure your-config.scm is only when chrooted right? <sozuba>Should (use-modules (gnu) (additional modules)) have additional modules to have luks lvm work? <podiki[m]>sneek: later tell apteryx my pam-limits config didn't seem to do anything in lightdm but does with gdm, maybe that's the cause <juliana[m]>lilyp woops accidentally just reviewed a patch you did hopefully that's not a faux pas - i forgot to check if someone else had reviewd it yet <sozuba>HiltonChain[m]: boot should be unencrypted for luks2, but not for one, i guess. <HiltonChain[m]>I assume relevant GRUB modules are installed to /boot instead of built-in, can you check it? <sozuba>finaly booted. I guess it had to do with the install 'guix system init /mnt/etc/config.scm /mnt' pulling in outdated packages, especailly grub. Once I ran, 'guix pull' and again 'guix system init ..... ' again and rebooting, it worked. <xelxebar>So, this is fun. Shepherd is pegging at 100% CPU usage right out of boot on my laptop. It'll settle down after a while, but I have no idea what it's working so hard on. <xelxebar>This is PID 1 shepherd, just in case that wasn't clear. <sozuba>HiltonChain[m]: i guess the relevant modules are inbuilt, at least in my case,considering that i was able t boot now. But if you want tme to check,i would be happy to. Just let me know how to check. And checking that is a good thing to do. <xelxebar>Is there a good way to get a flame graph of procedures or something, to figure out what exactly it's grinding on so hard? <sozuba>Just to clarify this -> 'Once I ran, 'guix pull' and again 'guix system init ..... ' again and rebooting, it worked.', I did this, updated the packages in the live image and then -> ran guix system init.... again <sozuba>so i guess its best to update the packages in the live image before starting installation,if installing manually at least. <ennoausberlin>for this package? I tries to compiles c-code during the build process, but fails for now. <rekado>ennoausberlin: yes, this is free software <rekado>you can add a different gcc package to the native inputs <rekado>this npm install business is really confusing. I haven’t yet found any good reason for the behavior to differ between the two machines. I wonder how many more quirks there are that make our packages fail to build reproducibly. <rekado>one thing I find somewhat concerning is that npm often uses relative locations that go through the root file system <rekado>e.g. '../../../../8a0q583gnkg171mq3pz70k2zjjm22i0l-node-fliplog-0.3.13/lib/node_modules/fliplog' <rekado>instead of just /gnu/store/8a0q… <rekado>I’ve tried stracing the invocation of npm install on the two machines and the output differs wildly <rekado>my next step is to figure out where in npm the files are actually copied or arranged to be copied and inject some logging there <civodul>i'm afraid cuirass-remote-worker is not picking up jobs <rekado>on the cluster Guix arborist resolves locations like this: 'file:/gnu/store/63z2zyh797216kvnyfyfyz83d3ib8mk9-node-fuse-box-3.7.1/gnu/store/gxl42qkmyndrz4dyma65g49lwcbfqnm4-node-universalify-0.1.2/lib/node_modules/universalify' <rekado>uhm, it also looks like that on my laptop (where it works) <rekado>what I do is: cd $(guix build node-fuse-box)/lib/node_modules/fuse-box; $(guix build node)/lib/node_modules/npm/node_modules/@npmcli/arborist/bin/index.js actual <rekado>the output of resolved locations is concerning <ngz>How can I "enlarge" Guile backtraces so they are not truncated anymore in build logs? <cbaines>ngz, COLUMNS=1000 sometimes works I think <cbaines>that is, set that environment variable <ngz>Thanks, this does work. <MatoHota-work>I'm glad that guix locate is available, but as far as I understand it can't be used to search out of what is already installed <efraim>spine-o-saurus: in this case it means you've installed guix in one of the profiles <MatoHota-work>Nobody has an idea where to found *.el files ? Still looking for ada-mode.el :) <Cairn>MatoHota-work: Did you install it from a separate Guix channel? <AwesomeAdam54321>MatoHota-work: I don't have ada-mode in Emacs 28, and the ada-mode package hasn't been added to Guix yet <efraim>it looks like it was included in emacs-26 /usr/share/emacs/26.1/lisp/progmodes/ada-mode.elc <flaminwalrus[m]>MatoHota-work: The help commands generally link to the source files, no? <apteryx>podiki[m]: maybe! Have you tried adding it? <sneek>apteryx, you have 2 messages! <sneek>apteryx, podiki[m] says: my pam-limits config didn't seem to do anything in lightdm but does with gdm, maybe that's the cause <Cairn>MatoHota-work: The way you installed Emacs is fine, I was just curious how you installed ada-mode. Sounds like you did it with package-install or some sort <MatoHota-work>I have installed emacs 28.2 which don't provide ada-mode - as I have discovered <spine-o-saurus>so if the admin runs a pull update for the system, is there a way for users to pull update from local system instead of remote? <MatoHota-work>I was using debian before and emacs comes with ada-mode included as far as I understand <Cairn>MatoHota-work: Well since it isn't packaged in Guix yet, you can just install it using the built-in Emacs package manager <Cairn>M-x package-install ada-mode <flaminwalrus[m]>MatoHota-work: If ada-mode isn't in Emacs or isn't loaded, yeah, it might fuzzy-complete to something unrelated. But when you do have it installed and loaded, this is a good way to figure out where source files are. <Cairn>Yeah, a lot of individual Emacs package maintainer sites grow stale after a while. It's still in the ELPA though. <reed92>Is there a standard way to access graphical programs from within a 'guix home container' environment? I think I need to --expose something but I'm not sure exactly what <Cairn>reed92: Try `--preserve='^DISPLAY$'` <Cairn>reed92: I know that works with `guix shell` <reed92>Cairn: It's saying 'unrecognized option' for preserve <Cairn>MatoHota-work: Yep, `guix import` returned a working package for `ada-mode` as well =) <Cairn>reed92: Alright, then I don't know what I'm talking about. <reed92>Cairn: no worries, I'll keep looking through the docs <Cairn>MatoHota-work: But you already proxy guix? Can you reach elpa.gnu.org at all? <spine-o-saurus>woa, so when i tried installing fluxbox it tried to build but exited with code 1 failure <Cairn>spine-o-saurus: Fails for me too <HiltonChain[m]><spine-o-saurus> "so if the admin runs a pull..." <- It's not "for the system", but "for the root user" <spine-o-saurus>so if there is 20 users on the system they all have to run that pull over the network? <MatoHota-work>It seems that using package-install with the good proxy I got it installed in my .emacs.d repo - thanks for the help Cairn, flaminwalrus[m] <Cairn>MatoHota-work: Glad to hear! <HiltonChain[m]>spine-o-saurus: I assume copying .cache/guix/checkouts would work, I'll have a try. <ekaitz>efraim: did you have some time to take a look to zig-build-system? <3 <efraim>ekaitz: I'm not sure I'll have time today but I can try <ekaitz>efraim: also the 14 of July I start to be 100% focused on our Bootstrapping effort, and I'll need some help, are you free for that? <ekaitz>if you want, you can share your calendar with me (or just an overview) and we can think on how to coordinate <ekaitz>also, anyone who is interested on RISC-V bootstrapping is invited to join us <ekaitz>anyone who wants to join or be informed just email me (grep my name in guix source code and you'll find my email :)) <Cairn>ekaitz: Interested in excitedly reading the blog posts later; not technical enough to actually help, hehe <ekaitz>Cairn: for that you can follow my blog! we'll post everything there and we'll try to post also in guix blog when we merge everything upstream <Cairn>Ooh! Mind sending a link ekaitz? <ekaitz>Cairn: also, don't be overwhelmed by the technical complexity, it's easier than it looks <podiki[m]>apteryx: didn't get a chance to try but will today, thought you might know already :) <Cairn>ekaitz: By the way, what sort of RISC-V hardware would an interested Guix user actually get? What do you use? <ekaitz>Cairn: I didn't need to use anything yet. I just read the binaries that the compilers produce with my bare eyes. <ekaitz>we are looking to buy some Starfives but I don't know if they would actually work, efraim knows more than me on that <Cairn>Your mind is the RISC-V hardware o_O <efraim>The one that I bought most recently that I'm happy with is the starfive visionfive 2. right now I have mine running ubuntu <ekaitz>ACTION only works with easy examples <ekaitz>efraim: so you run guix as a package manager there? <efraim>it's my next target for building an image for <ekaitz>efraim: if you payed for it with your money we can cover the cost (I have funding for it!) <AwesomeAdam54321>efraim: How much effort is it to add RISC-V support to all the compilers that don't have it in the bootstrap chain? <rekado>I keep having problems with ibus on every upgrade <ekaitz>later we can donate the hardware to guix <ekaitz>I may get one, but only if I can properly donate it to guix later <rekado>ibus-libpinyin works in the gnome shell but doesn’t work in the console or in firefox <efraim>the rust compiler I need to fix, it used to work before the core-updates merge <rekado>I keep deleting ~/.config/ibus and ~/.cache/ibus, and recreate the config after every upgrade <efraim>GHC isn't doable ATM, that needs a blessed binary from upstream <rekado>I would be very happy to get some pointers on how to make this work reliably, across upgrades <Cairn>Is the PineTabV of any interest? I know Pine64 stuff tends to blobs for wireless stuff, but... <efraim>rekado: would it be enough to delete ~/.cache/ibus and keep ~/.config/ibus? or are there paths to the store there <efraim>the pinetabv looks interesting, but it's beyond the scope of where I'm interested ATM <rekado>I haven’t yet been able to restore ibus across applications <efraim>I'd like to see about a licheepi 4a to test more riscv cores (and the 16gb of ram) , and I bet the star64 from pine64 is probably good <ekaitz>AwesomeAdam54321: also in C it's a hella lot of fun <ekaitz>that's why nlnet is paying us to focus on it <efraim>ldc is coming along nicely and is probably doable, ocaml I haven't looked at, zig isn't supported yet upstream IIRC <efraim>zig might need some backporting from upstream with their new-ish binary bootstrap <efraim>that's all the compilers I can think of right now <Cairn>spine-o-saurus: By the way, commit 741b534 works for `fluxbox` if you'd like it for now <efraim>julia doesn't build yet, I forget what the problem is there <Cairn>spine-o-saurus: guix time-machine --commit=741b534 -- install fluxbox <podiki[m]>apteryx: there's 2 places for pam with login managers, do you know if both are needed? my use case was with pam-limits, not sure about the other <ekaitz>efraim: node is a problem in general, not specifically with riscv, right? <apteryx>podiki[m]: I'm sorry, I'm very PAM-ignorant :-) <podiki[m]>me too, except for the one setting I need :) <podiki[m]>I'll poke around, maybe the pam files from lightdm will clue me in <efraim>ekaitz: We build node@18 with node@10, and node@10 doesn't support riscv <juliana[m]>Are there any projects to provide Guix packages for every release of software? It's fairly common for package updates to skip point or minor releases, but in cases of eg language-specific dependency management, having precise versions of certain things may be necessary. I know there's guix-past, but there are gaps in its library as well, notably with more recent releases <juliana[m]>Thinking specifically here of the project I was working on when I got into Guix - it was a Python project and I used some environment management tool for the first time, and it inspired me to try Guix. But the specific version of Python the project was using had never been packaged for Guix <juliana[m]>ekaitz where does one sign up to assist with the RISC-V effort? I've done some RISC-V compiler stuff in the past <civodul>juliana[m]: "every release" is a lot :-) <civodul>Guix-Past is the closest you get to that <civodul>but then again, since you can time-machine to any Guix revision, you can get many releases <juliana[m]>every release is indeed a lot, but a girl can dream XD <lilyp>juliana[m]: multiple opinions are always welcome <lilyp>apart from guix past don't forget about guix time-machine; that ought to have you covered on recent, but outdated packages <efraim>I've thought about packaging releases of python but not any libraries, seemed like it'd be useful as an input to other channels that needed other versions of python and were willing to use package-with-explicit-python <johnabs[m]>Hi all, can I get some quick help with an apparent locale issue? I'm not sure what went wrong, but something is breaking libreoffice and throwing errors in julia. <johnabs[m]>To clarify, here's the exact issue with libreoffice, and I think the two are related: (soffice:22905): Gtk-WARNING **: 14:28:07.284: Locale not supported by C library. <johnabs[m]>/gnu/store/y83kzjgbahphv8h7ybf7pqhyrf37mlbf-libreoffice-7.5.1.2/lib/libreoffice/program/soffice.bin: symbol lookup error: /gnu/store/5h2w4qi9hk1qzzgi1w83220ydslinr4s-glibc-2.33/lib/libpthread.so.0: undefined symbol: __libc_pthread_init, version GLIBC_PRIVATE <Cairn><juliana[m]> "Are there any projects to..." <- I've been interested in this as well. It'd be cool to write package definitions which cover a range of versions/commits, and then inherit and modify when the definition needed to change <Cairn>That way, there could be a working definition for every possible version of a software <efraim>johnabs[m]: are you using LD_PRELOAD anywhere? that libreoffice is linked to glibc-2.35, not 2.33 <Cairn>I'm having trouble using --with-input=gcc=*. I can't replace * with gcc@4.8.5, gcc-4.8 or gcc-4.8@4.8.5. What am I getting wrong? <juliana[m]>you need `--with-c-toolchain=<package>=<toolchain>` to switch out your C compiler <juliana[m]>`<package>` is the package whose compiler you're switching out; `<toolchain>` is `gcc-toolchain` or the like <juliana[m]>there is no plain old gcc package because gcc by itself isn't very useful (you still need an assembler, linker, etc.) <juliana[m]>(well, there is a plain old gcc pacakge, just not that is exposed to CLI tools - idk about transformation options actually come to think of it) <Cairn>I looked at this, but I was confused. Do I just specify my package twice? --with-c-toolchain=package=gcc-toolchain@4.8 package <Cairn>I assume the verbosity is for being able to replace the toolchain of a package's inputs as well? <sozuba>Around twelve hours ago, i said that outdated packages (grub) being pulled by the live image in manual install could be the reason my system booted into gryb rescue. Turns out that's not the reason. <johnabs[m]><efraim> "johnabs: are you using LD_PRELOA..." <- I don't believe so, where would I check to verify this? <johnabs[m]>I'm also getting this: warn:sal.osl:27600:27600:sal/osl/unx/thread.cxx:1068: RTL_TEXTENCODING_DONTKNOW -> _ASCII_US <johnabs[m]>warn:sal.osl:27608:27608:sal/osl/unx/thread.cxx:1068: RTL_TEXTENCODING_DONTKNOW -> _ASCII_US <johnabs[m]> and a similar error in Julia when I first open it :( <Cairn>Alright, thanks for the advice juliana. <sozuba>Initially the issue was frub doesn;t support luks2 at all, even with pkbdf. It only supports luks1, at least in my case. second, everytime yo convert between luks1 and luks2, you have to reinitialise the system, with guix system init /mnt/etc/config.scm /mnt. I guess econfigure also works but i haven;t tried it, cause this us only the initial build. <sozuba>What i don't understand is, if the UUIDs remain the same and config hasn'tchanged , why is there a need re-initialise/re-configure/rebuild the system, when the parameters of the encryption changes. Does it have to deal with what modules are built into grub during this process? i thought it installs the binary and doesn't compile them. <HiltonChain[m]><sozuba> "What i don't understand is..." <- grub-install is invoked, Guix itself doesn't deal with the details. <denys[m]>Did anything change in issues.guix.gnu.org routing today? Down For Everyone Or Just Me says it's up, but traceroute enters a loop at 188.1.239.206 <denys[m]>Must be the network then, but strange, I try connecting from three servers in different areas, and all three can't access <rekado>denys[m]: where are you connecting from? <rekado>denys[m]: did it work for you earlier? <rekado>I only know of the firewall blocking connections that come from unspecified “Russian IPs”, but I have no privileged access to what the firewall does or doesn’t do, so I don’t know if there has been any recent change. <rekado>I’m also not aware of any maintenance going on around this time. <rekado>I’m afraid I don’t have a good explanation for what might be wrong in your case. <denys[m]>Alright, I guess I'll wait until tomorrow and write emails to contacts of nodes in the traceroute chain if the issue persists <sozuba>HiltonChain[m]: thank you for the response. so the grub core image containes the neessary modules, such as luks, lvm, etc? <mirai>I'm surprised at how/why the cups service works <mirai>not only it is strangely structured, the configuration shouldn't even be serializable since somewhere along the process it would do a (string-append "string1" … #f "string_n" …) <johnabs[m]>Hey guys, I just updated after guix pull and gdm is crapping the bed with a white screen, I'm temporarily switching to slim (not working) and now to sddm, but is there a way to get this issue diagnosed? I'm not sure where to look for logs for this service :/ <mekeor[m]>mirai: link to source? which line should invoke that s-expr? <mirai>mekeor[m]: (gnu services cups), look at cups-configuration and note the extensions field <mirai>its type is package-list and the serializer for this type simply returns a #f <mekeor[m]>johnabs: maybe in /var/log/...? i'm guessing :) <mirai>at cups-shepherd-service, look at cupsd.conf-str <mirai>serialize-configuration is called <mirai>actually this whole thing is… strange <mirai>serialize-configuration in this module returns me a string <mirai>it should be returning Gexps <johnabs[m]><mekeor[m]> "johnabs: maybe in /var/log..." <- Okay, question, there are a lot of log files, 4 to be precise, would you happen to know which one may be pertinent? <johnabs[m]>Also, I noticed that the other login managers didn't particularly like my xorg-configurations, so I'm in a bit of a bind trying to debug the issues I'm getting with my screen :/ <mekeor[m]>mirai: do you know where #~ etc are defined? <mirai>uuuuh, this whole thing is suspicious <mekeor[m]>mirai: i wonder if #f does not get spliced or so <mirai>I think it's not serializing at all <mirai>rather, its echoing to a port <mirai>$6 = (string-append "" #t #f #f #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #t #<unspecified> #<unspecified> #t <mirai>$4 = (serialize-configuration (cups-configuration) cups-configuration-fields) <mirai>thinking about it, that line above should only give a gexp <johnabs[m]>Ayy, it was my xorg-settings apparently...though I'm now unsure of how to fix them so I can do what I want... <mirai>there's no reason why I should see raw output, they're side-effects <mirai>mekeor[m]: open a repl, do ",m (gnu services cups" <mirai>and play around with serialize-configuration + gexp->approximate-sexp <johnabs[m]>And I fixed my libreoffice and julia locale issue, now just to fix my xorg config without borking gdm <johnabs[m]>Okay, new question! I had this in my guix config, but it apparently borked gdm. Specifically, it was in my (services (append (list (.....))), is this no longer the right place for that, or did something change that I need to take into consideration? https://pastebin.mozilla.org/xQS6wzT0 <johnabs[m]>(Pardon the common lisp tag, it didn't have guile there) <mekeor[m]>johnabs: you could bisect the configuration. does it work when you only remove the (extra-config ...) s-expression? <mekeor[m]>johnabs: did this config work fine up to until now? <johnabs[m]>mekeor[m]: The config worked fine until I ran guix pull and updated <mekeor[m]>johnabs: you could also investigate the xorg-config generated from this guix-config. i think it should be located at /run/current-system/profile/share/X11/xorg.conf.d after guix-system-reconfigure. <mekeor[m]>johnabs: what does 'guix system list-generations 1d' tell? <mekeor[m]>johnabs: i.e. did you upgrade packages related to xorg or so? <mekeor[m]>mirai: for me, serialize-configuration returns a gexp <mekeor[m]>mirai: are you sure it returns a string for you? doesn't it simply "display" a string? <mirai>or do a ,use (guix) + ,build (mixed-text-file "foo" the-gexp) <mirai>mekeor[m]: indeed, it displays a string <mirai>I mistakenly said return a string but the issue still remains <mekeor[m]>mirai: and yes, it calls with-output-to-string which reads whats being displayed into a string, i guess :) <mirai>so what's happening is that its capturing the displayed string and using it as serialization <mekeor[m]>it's using the captured output/string as cupsd.conf-str, yes <mekeor[m]>the extensions will just be ignored for the configuration, i guess <mekeor[m]>extensions are only added to DataDir and ServerBin, it seems <johnabs[m]>Okay, I'm back! First of all, mekeor , that file doesn't exist for some reason after rebooting into either config: I only see XErrorDB XCMS.txt and locale <johnabs[m]>Also, running the following command returns nothing: "guix system list-generations 1d | grep xorg", so I assume that's a no to the second question <johnabs[m]>(when I said either config, I meant both the non-working and the working ones) <mekeor[m]>johnabs: not necessarily, no. could you share 'guix system list-generations 1d'? <johnabs[m]>Also, there are a lot of them due to me finagling trying to get it working again, I hope that's alright <johnabs[m]>Also, is there any other place I could find the X11/xorg.conf.d? I can't find it anywhere... <mirai>mekeor[m]: Someone™ should consider doing a refactor of the cups service <mirai>its incredibly obtuse the way it is structured <mekeor[m]>johnabs: i'm sorry, i can't help. my personal setup is too different, so i can't investigate :/ <mirai>I'd say incorrect even, were it not for “it works though?” <johnabs[m]>But I need those parts...this'll be fun to fix :/ Thanks for your help so far mekeor, I appreciate it 😁 <mekeor[m]>mirai: not sure what you find incorrect? do you mean that with-output-to-string is hacky? <mekeor[m]>there are 10 services that use with-output-to-string <mekeor[m]>these files: gnu/services/authentication.scm gnu/services/cups.scm gnu/services/dns.scm gnu/services/kerberos.scm gnu/services/ldap.scm gnu/services/mail.scm gnu/services/messaging.scm gnu/services/pm.scm gnu/services/virtualization.scm gnu/services/vpn.scm guix/scripts/describe.scm <mekeor[m]>it seems to be a common pattern in guix code <mekeor[m]>many of the occurrences are tight to configuration serialization <mirai>I didn't look at the others yet <mirai>but I can say that cups definitely needs some extra-abrasive polish here <ekaitz>efraim: did you manage to take a look in the zig patches? <johnabs[m]>Does anyone happen to know where the xorg extra-config stuff is supposed to go per chance? I can't find any information in the docs on this, though maybe I've missed it. <ekaitz>johnabs[m]: let me check my config