IRC channel logs
2023-05-29.log
back to list of logs
<rekado>sammm: thanks for the report about the logging <rekado>no idea what happened on bayfront; the service was still running <efraim>perhaps it's related to the shepherd bug? the one where restarted services don't actually stop and start again but hold file descriptors? <PaulePanter>Hi. How can I run the precompiled guix binary as an unprivileged user? <PaulePanter>It has /gnu/store/… hardcoded in the shebang line and an unprivileged user is not allowed to create these directories. <janneke>you cannot install on a foreign distro (in this way) without root privileges <PaulePanter>janneke: That’s what I feared. Thank you for the confirmation. <civodul>janneke: congrats on the native GNU/Hurd boot!! <janneke>civodul: thanks! phew, that initial "guix system init" support and "second boot" problem took me two days of hacking! <janneke>ah, that would be a nice fix to have <civodul>actually, i saw you had two GRUB menu entries <civodul>did you managed to get one Guix not step on the other's toes? <janneke>no, guix system init /hurd steps on linux's toes <janneke>but i follow that by a guix system reconfigure, that adds menuentries for everything in /hurd/boot/grub/grub.cfg ;-) <janneke>(there is still quite some room to smooth-out the experience ;) <guixfren>Is there any work being done on updating to Gnome 44? <guixfren>I wonder if this could be a good opportunity to start contributing? It's work to do but I don't believe it should be that difficult (hopefully anyway) <janneke>guixfren: i guess you could look into the gnome-team branch <janneke>i'm sure the gnome team would appreciate some help <lambdanil>janneke: I'll see if I can figure something out, I'm not very familiar with the GNU way of development which is currently the biggest barrier of entry for me <ieugen[m]>hi, I would like to learn more guix and my target for this is to try to update lokke package that seems to be old. <ieugen[m]>Is there a guide on how to update the package and build it locally? <lambdanil>You could go ahead and make a separate definition - copy the definition found with "guix edit lokke", put it in a new file, import the necessary modules, then you can finally go ahead and try switching the source tree and messing with the build. <ieugen[m]>I guess my question can be summarised as: After cloning guix repo, updateing lokke commit id, how can I build the package ? is cd guix ; guix build -L . -K lokke enough? <efraim>then afterward it'd be `cd guix; ./pre-inst-env guix build lokke' <ieugen[m]>I ran guix build -f gnu/packages/guile-xyz.scm lokke and it said that lokke is not a package. And it's true: (define-public lokke <jlicht>ieugen[m]: The -f flag should refer to a file that evaluates to the package expression. If you just want something that works _now_, put 'lokke' (sans ') on the last line of the file <janneke>ACTION is trying another approach to resurrect native builds on the Hurd <janneke>instead of adding gcc-boot0/hurd to avoid the gratuitous coreutils-boot0 dependency, let's try to build coreutils-boot0... <janneke>it seemed such an easy fix to add gcc-boot0/hurd, but it's used in more places than %boot1-inputs, as jpoiret pointed out... <Parnikkapore_m>ieugen: the -f means you want to install the package that the file *evaluates to*. It doesn't work with the stuff in gnu/packages/ <Parnikkapore_m>"-L ." sounds like it might work, although I don't know for sure. <ieugen[m]>thanks, I guess I a trying to find how people rebuild packages from guix when they change versions? <Parnikkapore_m>Apparently the official workflow is to build Guix with the source tree, then use the newly-built Guix. The chapter in the Guix manual should guide you through it. <ieugen[m]>running `guix build -L . -e '(@ (gnu packages guile-xyz) lokke)'` gives me the same error as I got with -L . lokke . <ieugen[m]>I think I have to do things ina a gux shell, from guix source <ieugen[m]>Parnikkapore 😁: I did follow the steps, but during make I get an error when building the .ru info . <ieugen[m]>make[2]: *** [Makefile:5146: doc/guix.ru.info] Error 1 <ieugen[m]>Also, I got more context from your method pjals `;;; note: source file ./gnu/packages/pkg-config.scm` `Unbound variable: $PKG_CONFIG_PATH` <ieugen[m]>(running on Debian, not inside guix shell - I can't make inside shell because of info/texi compilarion ) <efraim>have you run bootstrap and configure yet? <ieugen[m]>it's an env path for pkg-config - used when building software (during ./configure phase I think) <ieugen[m]>thanks, I think I will need to get more familiar with guix and pay the price of reading the contribute docs. <janneke>terribly slow progress building coreutils-boot0 <janneke>first, need -Wl,-rpath -Wl," (assoc-ref %build-inputs "libc") "/lib" <janneke>now IGNORE_UNUSED_LIBRARIES_CFLAGS= to get rid of -Wl,--as-needed <janneke>jpoiret: doe we have a bug open for broken hurd native compilation? <janneke>jpoiret: i've just built coreutils-boot0, i'm opening a bug with patch so that "people" may discuss <ytc>could someone please explain what is purpose of declaring the packages in the home profile? i don't see those packages in my PATH variable if i didn't install them with "guix install" but wrote it in home-environment's 'packages' field. <ieugen[m]>I got that version of lokke but it's very old <ieugen[m]>right now I am trying to run guix from source in container and when I make it make[2]: *** [Makefile:5146: doc/guix.ru.info] Error 1 <dcunit3d>if i reply to a help-guix mailing list with nnimap in guix with help-guix@gnu.org in the CC field, will that appear in the mailing list? i've already posted to the list, but i was having trouble with gnus last night. the mails are showing as sent in my webmail <dcunit3d>oh, nvm. sorry to bother y'all. i'm just not sure what is supposed to happen or whether i've configured things correctly. <ieugen[m]>I copyied guix.es.texi over guix.ru.texi and guix.zh_CN.texi . Also cp guix-cookbook.de.texi guix-cookbook.ko.texi <ieugen[m]>I have managed to build updated lokke locally <janneke>ACTION builds gcc-boot0 for the hurd, `guix build hello' still hangs <janneke>moving forward manually through commencement... <janneke>is there some trick to find a dependency cycle? <f9>i have important thing to tell you all <efraim>janneke: are there other spots where gcc-boot0 is brought in explicitly? or inherited from? <efraim>I saw the typo in (let-system (syttem target) but then you reverted it <efraim>ACTION is looking at janneke/wip-hurd <efraim>libstdc++ maybe, before zlib-final <janneke>efraim: thanks, i haven't tried those <janneke>there seems to be a problem with %boot2-inputs <janneke>ah, sure both glibc-final and gcc-final hang <janneke>crap, this worked 3y ago, why don't we have this under test? <efraim>what if you unrevert using gcc-boot0/hurd for gcc-boot0-intermediate-wrapped and gcc-boot0-wrapped, but without the typo? <efraim>861c4de08990ca4c54b6bcca54aec3f63b6ed175, but with system instead of syttem <janneke>In procedure string-contains: Wrong type argument in position 1 (expecting string): #<procedure system (#:optional _)> <janneke>so it seems that `letssystem' doesn't work here <janneke>ACTION is consistent in their typos... <davidl>nckx: thanks for the link. A possibility is to keep a TODO.org file in the git repo (TODO-headers like PROPOSED, IN PROGRESS etc.), I wonder what you would think of that? Although I don't really wanna suggest more work to current maintainers, and it might not be worth it in the end. Though it might also be a base for GSOC projects, maybe funding, etc. <janneke>these new gnulib stackoverflow tests are terrible <janneke>i found they fail for riscv, and it seems they fail for the hurd too <janneke>okay...so mig-boot0 also hangs, while bison-boot0 and flex-boot0 build... <efraim>I'd check the with-boot0 %boot0-inputs <apteryx>hm, do we have a service that can be extended to set environment variables globally (e.g. in /etc/profile) ? <janneke>aahhh, mig has changed to using git-fetch! <janneke>it's our (jpoiret and mine) own fault after all <efraim>wait, it's the same %boot0-inputs for flex-boot0 and with-boot0! <efraim>oh, I completely skipped over that commit <efraim>I wonder if you could pre-seed the source. not that it'd really help <janneke>ah wait, we already did this loop, we can use a savannah tarball and add autoconf-boot0 <mwette>This is crazy (fun) using guile to configure an OS. <mwette>I'm getting error in config.scm. I want to remove network-manager-service-type, but getting error that it's not defined. Do I need a user-service-module? If so, what? <davidl>nckx: on a related note with the libreplanet page. It has things like vscodium as a suggestion, and it would probably be useful to keep a note somewhere in like a TODO.org file of the javascript situation for that and similar issues with some other software, like for example licensing issues with ZFS or veracrypt. <janneke>jpoiret: headsup, i've reverted your mig patch and upgraded it to the latest savannah tarball <janneke>and added autoconf-boot0, automake-boot0, ...but still struggling there... <mirai>mwette: you need to use (modify-services …) <mirai>and perhaps (use-module (gnu services networking)) <mirai>or (use-service-module networking), whichever you prefer <cbaines>this morning I think we had 80%+ substitute availability for 5 architectures <cbaines>although now it's only just over 80% for x86_64-linux <mwette>mirai: Thanks! -- I did manage to try those, and have a modify-services, copied from the web somewhere. Chasing other issues now. <jlicht>it's not even a functional change. Perhaps we can still have weekly/ <jlicht>or even monthly 'disruptive' days, were stuff like this can just be pushed. <mwette>What's the syntax to remove a service within modify-services? (network-manager-service-type => ???) <jlicht>mwette: (modify-services %base-services (delete mingetty-service-type)) <janneke>jpoiret: ah, just your package upgrade <mirai>iirc it would be (modify-services %... (delete the-service-type)) <mwette>jlicht: thanks - just found that in the manual <jlicht>(adapt for %base-services/%desktop-services and service-type, of course) <jlicht>janneke: I'm imagining you going "Eureka!" about every 20 minutes today :-) <mirai>apteryx: turns out there's already a ipv6-address? predicate in gnu/services/base.scm <mirai>(it's insufficient since it uses inet-pton but nonetheless) <apteryx>I think that's where my version started from <mirai>ah ok, I happened to see it while I was looking for syslog service <janneke>cpu.symc:83:43: error: 'mach_port_name_t' undeclared (first use in this function) <janneke>omg, with-boot0 does some "magic" that breaks my inputs <janneke>*wo0T* guix build hello no longer hangs -- jpoiret ! <janneke>guess we'll find out where it breaks in a while <elevenkb>Has anyone had trouble building network-manager in the last couple of days <elevenkb>sorry it isn't network-manager but modem-manager. <elevenkb>There is a problem getting substitutes for network-manager from bordeaux. <dcunit3d>elevenkb: i can't stay for long, but i'm getting a 404 on network-manager <elevenkb>that's probably why network-manager is 404'ing. it depends on modem-manager which doesn't build. <janneke>hmm, our hurd bootstrap binaries might be too old... <janneke>error: unknown type name 'long_natural_t' <janneke>error: unknown type name 'rpc_long_natural_t' <ekaitz>efraim you have mail: I need to solve this asap <ekaitz>it's supposed to be just a minute and then we can get to work <ekaitz>efraim: when is your project ending? june? <ekaitz>just for everyone to know what are we talking about: I got more funding from NlNet + NGI assure to work on the RISC-V bootstrapping <ekaitz>this time we are going to work as a team <ekaitz>good night efraim thanks for being that fast <jonsger>lechner lilyp btw, updating eudev resulted in 4229 dependent packages to be rebuilt <janneke>why oh why do people keep inventing new names for "int" "long" "unsigned" etc? <mirai>they should have picked something like Ada if they were looking for strong typing <janneke>well, ada doesn't even bootstrap, so i cannot take that seriously <mirai>ACTION is unaware of the bootstrap section of guix <janneke>the ada compiler is built using--only exists because of--some binaries magically produced in the past <mwette>janneke: What are you trying to do? nyacc can replace random C typerefs. <janneke>mwette: i'm working to resurrect the native guix build on the hurd <janneke>nyacc is awesome, no problems there ;-) <mwette>ty; nyacc/lang/munge-base.scm(expand-typerefs) <janneke>new types got introduced somewhere (glibc, gnumach, ...?) and now new sources do not build anymore with our bootstrap binaries <janneke>jpoiret: made some progress, but it looks like we need new bootstrap binaries for the hurd <janneke> 19239 Floating point exception| /gnu/store/99pfm62kgkl0z5d7aswaixdbxj9bffi0-mig-1.8+git20230520/bin/../libexec/migcom -subrprefix __ -prefix __ -header /tmp/guix-build-glibc-intermediate-2.37.drv-0/build/hurd/hurd/tioctl.__h -server /dev/null -user /dev/null <janneke>resurrecting the native build is going to take some effort still, but for now <oriansj>elevenkb: well what about it do you like best? <elevenkb>gee, mostly that when packages break I have an alternative to waiting for them to be fixed. <elevenkb>oriansj: I'm a linux n00b, so it's not like I can just debug a failed package build quickly. <oriansj>well, not everyone should be experienced in every package <elevenkb>oriansj: the thing I don't like about them right now though is that they don't let me download substitutes. <oriansj>and yes, I agree guix should do just a little more testing prior to pushing packages into main <elevenkb>like I already have the package in the store but... I have to rebuild it so it seems. <oriansj>well I am one of the #bootstrappable crazies, who disable substitutes and always build from source <elevenkb>oriansj: good on you! what is your build machine, I'm curious? <elevenkb>I can't imagine building on my laptop. A 10yr old i5-4308U with 8GB of RAM. <oriansj>an IBM x200 that has been libreboot'd running Guix <elevenkb>maybe i should actually try that one day. :p. <elevenkb>i use sway and emacs pretty much. i'm still clinging to firefox for youtube mainly. <dcunit3d>oh wow nyxt 3 is available. i've been looking forward to C-x C-b in the browser <oriansj>elevenkb: you might find yt-dlp quite helpful for reducing your web browser needs