<katco>hi, i think the common-lisp build system might be broken for "package-inferred-system" style common-lisp packages. the build system appears to find the dependencies, but then replace[1] "invalid" (within the context of the store) characters with `-`. i think this breaks asdf as it then cannot find the package. "package-inferred-system" packages almost always contain a `/` [2] which is flagged as an invalid character. does anyone have <katco>hm. this may be better as an email ***Guest2237 is now known as sturm
<sturm>On GuixSD, is it a problem that the "user-homes" and "term-auto" services don't start? I can't seem to find any info about these in the manual. The only output I can see is "Service term-auto could not be started." in /var/log/messages. Same for user-homes <lfam>sturm: In practice it should not be a problem <lfam>I see the same things on my system and it still works for me <lfam>It would be good to figure out what's up and fix the services <sturm>Thanks lfam, I've never noticed any actual problem, it's just that little voice in my head that doesn't like ignoring errors :) <lfam>I agree, someone should figure out what's up :) <dannym>If you have a system with a head, term-auto will not start and that's by design. gnu/services/base.scm default-serial-port tries to determine the kernel console (by parsing /proc/cmdline at boot). However, guix services are declarative - so the decision whether to include an agetty (and thus call default-serial-port) has to be made earlier (on reconfigure). But it doesn't have information there, and the only thing to do is to always <dannym>start the service and then fail if there's no serial console after all. <dannym>If you have a headless system it will probably boot with a "console=ttyS0" argument and then the agetty will also start up and enable you to login. <dannym>But the decision of whether to boot with serial console is done by u-boot and/or the user, so it's not really known in advance. <dannym>user-homes doesn't start up for me either - huh I've never noticed that before. Might be worth a bug report to bug-guix@gnu.org ... <AK_>mood: wow I'm there after some months :-) <q6AA4FD>this is only sort of guile unrelated, but there's a build check that sed appears to be failing of it's own accord, would it be bad practice to make a package to replace sed that skips that test? <civodul>janneke: %gash-bootstrap-guile and all, oh ooooh! <civodul>rekado_: thanks for the texlive update! *civodul finds sources of happiness on guix-commits <janneke>civodul: yeah -- i found it finally time to share; however, it's not yet done i'm kinda stuck @gcc-4.7.4 right now <civodul>janneke: oh, gcc causing troubles again? <g_bor>civodul: I am finally making some progress... It seems that I could overcome the jmod files in the openjdk, there are still differences, but it looks much better. <g_bor>The video documentation project is going fine, the framework is almost complete, we would like to onboard roptat if possible to have a look at the video translation workflow, mainly to check if we should mark the strings in the svg files not meant for translation as such. <janneke>civodul: the configure/make system has dependencies that are fulfilled by %bootstrap-coreutils&co, but breaks when using all gash+*-mesboot substitutes <janneke>my strategy now is to replace tool by tool to see which one[s] break the build <g_bor>do you have a build log to share? <civodul>janneke: and i guess it breaks in non-obvious ways deep down in the build process? <civodul>jas4711: the issue was raised before, but i'm not sure what the outcome was :-) <civodul>in some cases i find it useful to distinguish between root and user by showing a prompt <civodul>in other cases it's nice to be able to copy/paste directly <jas4711>civodul: right, i understand. maybe there should be a texinfo macro that generate the # vs $ in a way that doesn't interact badly with cut'n'paste <jas4711>ng0_: never mind. a bit more googling found the answer: boot kernel with iomem=relaxed <civodul>don't miss the very nice video by rekado_ ↑ ***rekado_ is now known as rekado
<roptat>What's the workflow you're thinking of ? <jonsger>rekado: is this ICG conference something bigger? I'm not into the bio scene... <roptat>Maybe there's something I can try? <jonsger>civodul: the berlin.guixsd.org.pub key is used for ci.guix.info.pub, am I right? <g_bor>we will have a discussion on the project at 12:00 UTC, and I will ask lsl88 to invite you. <g_bor>Do you have an account on gitlab.com? <roptat>I do, but I need to check when 12:00 UTC is <roptat>ah, I might not be available at 12:00, but I'll try to participate as soon as possible (12:15 maybe) <Swedneck>i have added ~/.guix-profile/share to $XDG_DATA_DIRS to get menu entries from guix-installed packages, is there any way to make guix automatically modify .desktop files to append "(guix)" to the entry name? ***ng0_ is now known as ng0
***dddddd_ is now known as dddddd
<rekado>jonsger: yes, ICG is pretty big. It’s organised by BGI, which is one of the world’s biggest genetics research centres. (They also design and build sequencers.) ***janneke_ is now known as janneke`
***janneke` is now known as janneke
<janneke>civodul, g_bor: no i don't really have a build log; the errors are usually symptoms not causes and very silly; an awk script that is invalid because sed or grep or gash made an error while producing it <janneke>my strategy now is to first find the tool that is problematic, resurrect the build and then try to find the real cause <janneke>i'm really tempted right now to first repackage gcc-4.7.4 into a gcc-4.7.4-core (upstream stopped doing that after 4.6) <janneke>hmm, would gcc-4.6.4 be good enough to build gcc-4.9.4? i could try that! <civodul>janneke: heh, so many possible paths! :-) <roptat>arg, I've got a step while building sbt that requires network access <roptat>the tool uses a tool that uses protoc (protobuf from google), but instead of the system-wide tool, it tries to download the latest version from maven <roptat>I have no idea how to disable that behaviour... <janneke>civodul: yes -- i'm using gcc-core-4.6.4 now (at least to bisect), that's a lot faster/friendlier <civodul>roptat: newish build systems tend to do that kind of thing frequently... <roptat>hm... actually not the latest, but the one it was compiled with <roptat>it's not even a build system, it's just a tool to generate some sources <apteryx>civodul: yes, I have some things to try regarding the network-manager-openvpn issues. Yesterday didn't have much energy to attack this at night, but today is Friday... :-) <civodul>i'm confident you'll get to the bottom of it very soon :-) <roptat>it works! I could embed protoc so it would be choosen instead of a remote one, and specify the version to prevent it from picking the latest instead of ours <roptat>I had to fix an issue where a script was generated at runtime with /usr/bin/env too <roptat>now the rest of the build continues and fails the *right* way :p <ng0>jas4711: what flashrom problem? <ng0>if I made no notes, I do not remember <rekado>booting the kernel with iomem=relaxed is the fix <ng0>doesn't guix have this in the manual these days? <rekado>fellow humans! I’d like to push the GNOME 3.30 stuff to core-updates (because the glib update is required), but I can’t test it as core-updates seems to be in disarray. Python 2 doesn’t build, for example. <rekado>is anyone feeling responsible for core-updates at this point? <rekado>who would like to be the core-updates “meister” for this round? <Swedneck>heads up: the ssl cert for the irc logs is expired since dec 30 <rekado>bavier: You added EXTRATESTOPTS=-j~d with parallel-job-count to Python 2 (on core-updates), but the check phase fails because it says that “-l” and “-j” cannot be used together. <rekado>Swedneck: thanks. It probably isn’t actually expired; the server just hasn’t been reloaded. <rekado>it’s fine now. Thanks again, Swedneck <Swedneck>ah, it seems the matrix logs for this channel aren't viewable for guest users <bavier>rekado: yes, it worked at the time, but some update in python@2 broke later <bavier>rekado: iirc someone has a fix, maybe efraim *rekado feels an itch to write a tiny Guile frontend for the logs <janneke>samplet: i have just reached until gcc-4.6.4 with the Scheme-only/Gash bootstrap; just updated Gash again and rebuilding world in the hope we are feature complete for the bootstrap <samplet>janneke: Wow! I saw your work on Awk. Very impressive. <janneke>samplet: thanks! both sed and awk are not yet good enough to bootstrap until gcc-4.x; but grep is! we can improve on those lateron anyway <janneke>i tried a fully functional approach for awk, i guess we'd better rewrite that using set! and a global %variables alist or so... <janneke>...and it would be so nice if we could tree-il SED and AWK execution (... and SH?) <samplet>janneke: I have been thinking about that a bit, too. <samplet>janneke: Don’t get too excited. Just idle thoughts. :) <samplet>I want to split things up a bit, still, though. <janneke>samplet: a little bit excited, is that ok? ;-) <janneke>when my wip-bootstrap branch builds somewhere this weekend i intend to mail gash/bootstrap status to guix-devel; i think we may need to fully support guile 2.0 and want to plan for a real 0.1 release (update doc, compile warning cleanup, dead code removal, ...) <samplet>Do you think we could pull out the utilities to a separate project before then? <samplet>We could start with the wip-merge branch, so all the Autotools stuff is in place. <ibanja_>Hi... I just discovered guix while searching for a way to distribute my admin scripts to some of my computers which run on two different distros--void and devuan. Would this be a good solution for what I am trying to do? <samplet>Then, break of gash/commands.scm from the shell. <bavier>ibanja_: guix could work for that, esp if there are supporting packages that need to be installed too <janneke>samplet: oh, i forgot about that; yes i have no real hurry -- what split do you have in mind exactly (esp: utility/library wise)? <ibanja_>bavier: so does it handle dependencies from the underlying distro, or from guix? <janneke>i exposed a couple of "built-ins/*" as shell commands too; that's necessary for the bootstrap <bavier>ibanja_: on a foreign distro, the only thing guix uses from the host system is the kernel <janneke>false, true and pwd come to mind -- it would be nice if the interface of both built-ins and command were the same (now they are just very alike) <samplet>janneke: I’m currently thinking about three projects: gash (the shell), file-utilities (e.g., rm and chmod), and text-utilities (e.g., grep and sed). <ibanja_>bavier: ok... thanks for the feedback <samplet>janneke: You mean that non-shell programs are exec'ing false, true, and pwd? <bavier>ibanja_: great, let us know if you have other questions <janneke>samplet: hmm, not sure -- either that or might a bootstrap BASH be doing that? i should have taken notes but will investigate <janneke>true, false and pwd are part of coreutils I think? but it would be good to know...hmm <samplet>janneke: I ask because false, true, and pwd are all “geesh” built-ins from before, so the shell should be able to use them. <janneke>yes, i understand -- a C program exec'ing would be a reason we need them externally; i'll try to look it up where it happened *janneke needs to go afk for a bit <rekado>samplet: I haven’t yet looked at it, but I’ll check it out soon. <davidl>has anyone has anyone been able to install Guix as a package manager in a linux that runs in WSL (windows subsystem for linux)? <lsl88>Hi! i need some help with guile :) <rain1>lsl88: you can ask here or in #guile if you like <lsl88>rain1: it is just something quite silly, I am changing a script a little <lsl88>did not know about that channel <janneke>samplet: read your pastebin, i think i see what you are pointing at -- some line-based unixy template to implement all(/some of) these tools <apteryx>I can 'save' an environment with the --root option, but how do I "reactivate" it? <janneke>apteryx: you source its profile: . ~/the/root/etc/profile <apteryx>janneke: but this leaves me in an impure environment -- what if I want it pure? <rekado>apteryx: you can do “/path/to/env /path/to/bash --init-file /path/to/profile/etc/profile” <apteryx>rekado: thanks. This addresses the pure environment. I feel there should be a -p flag to 'guix environment' the same there is a -p flag for 'guix package'. <samplet>janneke: Yeah. I would like to work towards a more Scheme-focused approach for the text commands, and I think that defining them in terms of SRFI 158 (generators) would be a nice way to do it. *janneke has to look at generators; in python those are something like "broken coroutines" ;-) <samplet>Right. They are not the greatest, but they work really well with ports. *samplet is tired of writing read-line loops. <lfam>Oof, LZMA decompressing a Git checkout of Stellarium is sloooooow <lfam>Oh, I see that it's actually recompressing it to apply a patch <samplet>janneke: I made a new repo for planning Gash 0.1. Maybe we can use it wiki-like to keep track of what we need to do (and what can wait). You should be able to push to it. <janneke>aah, note that I did the geesh->gash transition on wip-bootstrap; it would be nice if we could salvage that somehow <janneke>samplet: 3a0ab42 Move geesh into gash, some doc updates. <samplet>janneke: That’s good! Yes. We will keep it. <apteryx>rekado: yeah (about -p on guix enviromnment) <pkill9>i don't suppose anyone has a way of generating a list of the required module imports in emacs when writing a package definition? <pkill9>i.e. have it read all the package inputs and generate a list of the modules they're from <rekado>apteryx: sounds good to me. Go for it! <bavier>pkill9: I don't know of an existing solution, but you could possibly use (guix discovery) like the unbound variable error procedures do <roptat>it's the second time I see a software that's not able to handle a modification time of 0... <roptat>and of course tests fail because of that :/ <roptat>ah, the test makes sense: java returns 0 in case of error in getting the last modification time <bgardner>Hey all; I finished a clean install EFI install and after rebooting got into a situation where "guix reconfigure" was hanging. I decided to wipe the root partition and reinstall, but now I can't get the installer to write the boot information. It does "copying to /mnt" for about 15 minutes and then gives: "Could not prepare Boot variable: No such file or directory", then "[...]/grub-install: error: efibootmgr failed to register the <samplet>bgardner: Maybe your EFI partition is not mounted or is not mounted where your configuration file expects it. <bgardner>samplet: I did check that, it is mounted at /boot/efi and is the correct type. I also rewrote the partition and reset the esp flag on it to be completely sure it was right. <bgardner>I also added a stanza to the (file-systems ) block per the documentation to refer to the EFI by UUID. Didn't change the error. <samplet>bgardner: I think the important part is the “target” field of your “bootloader-configuration”. It probably already points to “/boot/efi”, though. <bgardner>It does, yes - and like I said this install worked once. I'm sure I've done this to myself in some fashion with leftovers from the prior install but I feel pretty confident I blanked both partitions. I'm puzzled. <bgardner>samplet: I'm going to blank everything and try again, thanks for the response <bavier>ArneBab: is there a vcs repo somewhere that could be used instead? <bavier>that can be used with the git-fetch method <ArneBab>I now copied the git-fetch part from another package and just adjusted the URL :-) <ArneBab>(I want libaom to get ffmpeg with AV1 support) <ArneBab>(because I now have a fast Guix machine at home, so I want to experiment with AV1 encoding) <ArneBab>(sidenote: did you know that most students nowadays don’t know plausible bitrates anymore? Only very few realized that 0.1kbit/s is much too low for music, though 15 years ago I’d have been very intrigued if someone told me that we could stream HD video at 200kbit/s)