<quiliro>how come 'guix package -I grub' shows nothing even if i am booting with grub-efi? <reepca>quiliro: it's because you don't have it installed in your profile. Just because it's on your system, in use, doesn't mean it's "installed". <quiliro>and guix package -I tor shows nothing even if my /var/log/messages shows Tor is running? <quiliro>and 'herd status tor' shows tor is running? <reepca>both of those are part of the system profile, rather than the user profile. <quiliro>reepca: ok, i, kind of, get it. what part of the manual says that? <quiliro>reepca: i want to read more about it <reepca>I'm not sure exactly where in the manual would best summarize it - just looking at "invoking guix system" I see that a lot of the options (build, roll-back, switch-generation) are similar to "guix package". I notice there isn't a command there that lists installed system packages or anything, but if you look at /var/guix/profiles/system/profile you'll see that it's very much like a user profile. <reepca>well okay, on second thought "build" isn't similar to guix package, it's similar to guix build, but you get the point <reepca>ah, I see now that there's a "manifest" file in that directory that lists all of the system's packages. <nckx>Indeed: the manual doesn't seem to explicitly explain the ‘system profile’ concept, at least not using that term. <nckx>(reepca: sorry, I was out, but I see you got your answer anyway) <nckx>Maybe I can write something readable to briefly touch on them in section 3.1, but don't hold your breath. <nckx>Because I'd write ‘a system profile is exactly like a user profile, but for the system instead of for a user’ in far more words, start to condense my prose, and realise that. Then I'd go do something else, annoyed. <reepca>heh, gnunet_bot, faithful as ever, still breaks on non-ascii input :D <quiliro>perhaps having a bad document is better than having no document <quiliro>the bad document is correctible by someone with little knowledge <quiliro>reepca: please go ahead and put the info...if am told how and what,i can do the carpentry <nckx>quiliro: OK, I promise to give it a try this week. <quiliro>reepca: where is the system profile defined? in the file that 'guix system reconfigure' and 'guix system init' use? <reepca>quiliro: grep tor /var/guix/profiles/system/profile/manifest <reepca>the list of packages the profile uses is sort of implicitly created from the same file that those commands use (typically "/etc/config.scm") <nckx>Or if you want to get all fancy: guix package -p /run/current-system/profile -I <nckx>(That's an I as in Igor Stravinsky). <reepca>the services, kernel, bootloader, and explicit system-wide packages all can add packages to the system profile <quiliro>guix package -p /run/current-system/profile -I tor <nckx>*sudo guix, by the way, or errors wil occur. <quiliro>and grep tor /var/guix/profiles/system/profile/manifest <nckx>You're right. Same on my system (which is also a tor node). <nckx>Oh, it worked without sudo? Here I got an ‘unsupported manifest format’ error in that case. <joshuaBPMan_>quiliro: The system profile is defined by both the file that guix system init and guix system reconfigure use. <joshuaBPMan_>BUT since you are more likely to use guix system reconfigure, then that is the file that defines the system profile. <quiliro>grep tor command shows a bunch of /gnu/store <Xe>quiliro: | grep -v store <nckx>quiliro: Are you saying you never ‘asked’ for tor but it's on your system anyway? <reepca>nckx: out of curiosity, which font are you using right now? I just see a box where M-x describe-char tells me a winking face should be <nckx>In that case: what does ‘guix gc --referrers /gnu/store/*-tor-*’ say? <quiliro>nckx: i installed tor with guix package -i on both root and user <nckx>quiliro: Removing a package won't delete it from the store (that's by design). Is that what you were expecting? <quiliro>nckx: i was expecting the service not to run <quiliro>/gnu/store/9k7xgy5hr1x4ia7rhdnqhr9q9sxyq2yy-shepherd-tor.scm <quiliro>/gnu/store/jdya6zh91bv0scyh10vbp77av7baiha7-profile <quiliro>/gnu/store/jh3wnm7ifa6jrwj3as933bvg2p9v4966-tor-0.3.0.9.drv <quiliro>/gnu/store/zcxm1q9ypjv8nlix34123369k3i6d8j7-tor-0.3.0.8.drv <quiliro>/gnu/store/cj8n8n635hly0b4yz9gvfafvy3sk8vz9-tor-0.3.0.9.drv <reepca>quiliro: have you restarted since removing it? I suspect it doesn't stop the service when it removes it from your profile <nckx>reepca: Cantarell. I'm running hexchat from Guix on a very busted (as in, won't even update itself since March) Ubuntu system. I just can't be bothered to fix it. Soon all will be GuixSD and happiness. <quiliro>i did not remove it today and have started the machine several times <quiliro>and guix package -I tor shows nothing on bot users <quiliro>reepca: that is exactly the problem i have...tor is started even though it has been removed <nckx>quiliro: If you never had tor in your config.scm, it's not being started by the system. <quiliro>and it is not on the system profile that i used for 'guix system reconfigure' last time <quiliro>nckx: i installed tor on users: quiliro and root...but then removed it from both <nckx>quiliro: Just installing it wouldn't make it start at boot. Who or what is starting it? <quiliro>but tor starts and i must herd stop tor <quiliro>i dont know what user...but i suppose the user is root <joshuaBPMan_>Hello, in the hurd roadmap, there is a develop a DHCP client in scheme...why is that on the roadmap? <quiliro> Requires (user-processes loopback syslogd). <quiliro>i will reboot to test if i could disable it this time <quiliro>i would like to remove it completely <nckx>quiliro: I just can't understand how herd would start (or even know about) a service that was *never* in your config.scm. <ng0>so remove it from the system config file and reconfigure? <ng0>that's the normal way to do it <nckx><quiliro> i never used tor on my config.scm <lfam>Do we have a package that can take an audio file and associated CUE file and split the audio file into the tracks listed in the CUE file? <reepca>quiliro: how did you get the service to run initially? You installed it and then... did it just automatically start? <ng0>I have this problem with python with the getmail 5.0 update.. it is more of a runtime problem with my specific configuration: NameError: global name 'errno' is not defined <nckx>quiliro: To forcefully remove all tors from the store, you can try ‘sudo guix gc --delete /gnu/store/*-tor-*’. But I suspect it's depended on by *something*. <ng0>I'm not sure if the getmail author did reproduce what I suggested.. <nckx>And then that won't work. <quiliro>nckx: are you sure that command will not break my boot? <nckx>quiliro: Scout's honour. <reepca>quiliro: even if it did, you could just boot to the old configuration from grub :-) <ng0>I used to use getmail-passwordeval, the 5.0 release was to add password_command so that you should be able to use password-store etc <nckx>joshuaBPMan_: phantomas (or was it phant0mas?) is the main Hurdperson. I'd ask sneek to ask them if was sure about the name. <quiliro>reepca: guix gc: error: build failed: cannot delete path `/gnu/store/43i5hicd3p5l10804wds7dxsclk45bxs-tor-0.3.0.9.tar.gz.drv' since it is still alive <nckx>quiliro: As long as you don't force it to remove live paths. <nckx>quiliro: ...which is apparently the case. OK. <nckx>quiliro: Yes and no. It's started by ‘the system’, and run as root, but won't have anything to do with root's profile. <ng0>the thing which depends on tor is the system configuration. remove (tor-service) and/or hidden-service from the system config and run system reconfigure. <ng0>there is nothing which just starts tor, at the moment <ng0>unless you wrote a service yourself. <nckx>quiliro: Basically, if config.scm doesn't (and, according to you) never has, herd won't start tor. Period. <nckx>If you did write a service yourself and didn't tell us that, that would be not very nice. So I assume that's not the case. <quiliro>you were right.... tor is in another config.scm....i thought i used the one without tor <ng0>hm.. I think the way password-store acts is not compatible with plain getmail 5.0 <reepca>quiliro: happens. I once spent a ton of people's time chasing down issues with some open firmware for a wifi adapter only to discover that I left the -a flag out of ifconfig <nckx>quiliro: :-D. I should really get rid of this cruft. It was fun in 2008. <nckx>quiliro: You're welcome. I'm curious how you ended up editing one file and reconfiguring from another, since both commands would require an explicit path. <nckx>quiliro: Maybe set up your system/workflow so that can't happen again :-) <nckx>There. No more emojibake. <quiliro>i have one lightweight/desktop.scm on /etc/ and the other on $HOME/ <quiliro>anyone used (console-keymap-service) ? <cehteh>is it me or is rxvt-unicode broken recently? <quiliro>nckx: it does not work for me.. i have it right below (locale) on (operating-system) <quiliro>/home/quiliro/lightweight-desktop.scm:9:0: error: extraneous field initializers (console-keymap-service) <lfam>cehteh: What goes wrong? <quiliro>(console-keymap-service "dvorak-es") <nckx>It's a service, so it needs to be under (services … <cehteh>all characters are "s p a c e d l i k e t h i s" <lfam>cehteh: Sounds like a font or locale issue <reepca>HDMI->VGA adapter finally arrived 17 days after ordering it with 4-7 day shipping. <nckx>ACTION goes to look up what a Spanish Dvorak looks like. <lfam>cehteh: It's working for me on Debian <reepca>Time to reboot and maybe, after several months of ordering graphics cards and adapters, I can finally use all 3 monitors! <ng0>can someone who's using getmail and password-store test getmail 5.0? I wonder if it's just me or a reproducible error behavior. The package works fine, it's just this new option <ng0>cehteh: that's normal <ng0>cehteh: for many many years (10+) I have two options which fix this behavior <quiliro>nckx: i have learned a lot of guixsd through this error <ng0>cehteh: normal as in you are supposed to fix it by yourself <quiliro>i underarstand now that the system profile is not the root profile <cehteh>well thanks .. but i really think thats annoying <ng0>cehteh: long config is long, I'm searching for my fix <ng0>at the very least it involves this: URxvt*letterSpace: 0 <ng0>rxvt-unicode is an old piece of ... <quiliro>so root does not need to guix pull and guix package -u <cehteh>what terminal emulator would you recommend then? <lfam>ng0: Where do you put that option? <quiliro>because none of the programs are installed in root profile <nckx>quiliro: I never even use the root profile. I can't think of a use for it in a world of sudoers. <cehteh>i am willing to switch to a better alternative <ng0>cehteh: I don't expect to see a fix for this upstream ever <ng0>lfam: in the normal place? .Xresources <nckx>quiliro: It's mainly a place for packages to get installed by accident when first using guix, and then never used again. <ng0>lfam: or do you mean the getmail thing? <lfam>ng0: I was wondering about the urxvt thing. It works for me from Guix and I don't set that option, unless Debian is hiding it somewhere <cehteh>i mean on debian rxvt-unicode my my favorite terminal emulator .. just works, and is *fast* .. the later is really important <lfam>But I just purged the Debian package and my Guix urxvt still looks right <lfam>I'll restart X to double-check <nckx>ACTION goes AFK to continue exciting sewing machine repair. <ng0>lfam depends on your WM I guess. <ng0>I have it in there since at least before 2007 <cehteh>looks like Xorg freezes in this vm now too <quiliro>nckx: i was explained before, but i still dont understand the difference between 'guix package -u' and 'guix package -u .' <lfam>I wonder if the root problem is something else. I just restarted X after purging Debian's rxvt-unicode package, and my urxvt still looks good <ng0>you just need to patch it, which is easy in guix <ng0>I have an example in one of my repositories <quiliro>nckx: i did not mean to send the question explicitly to you...sorry <lfam>quiliro: I can explain it <lfam>quiliro: `guix package -u` is the same as `guix package --upgrade`. I'll use the long form for clarity. <lfam>`guix package --upgrade` takes an optional argument. For example, `guix package --upgrade icecat` will upgrade IceCat. <lfam>But, you can omit the argument, and Guix will upgrade everything in your profile. However, if you try `guix package --upgrade --fallback`, Guix will interpret '--fallback' as the argument to '--upgrade'. <lfam>So, you should pass '.' as the argument. That is a regular expression that sort of means "everything". <lfam>In conclusion, whether or not the '.' is necessary depends on whether or not you pass any arguments after '--upgrade'. <quiliro>lfam: excellent explanation...it would be nice to have it in the manual <cehteh>grr .. that computer needs more ram :D <lfam>I agree, this documentation could be improved. Or we could improve `guix package --upgrade` to not create tricky and surprising results <quiliro>please consider me to make any work for guix <quiliro>perhaps it would be useful to have tasks to be done published somewhere <lfam>The fact that '--upgrade' accepts regular expressions is very powerful, but also tricky for those who don't know regular expressions. <lfam>Sometimes there are surprising results. I tried to upgrade syncthing once and it also upgraded qsyncthingtray :/ <cehteh>some day ago i noticed that --fallback --upgrade and --upgrade --fallback are not the same <quiliro>lfam: that i did not know...it is useful but tricky <lfam>cehteh: Yes, that's the case I just used in my example. You have to do '--upgrade . --fallback' so that '--fallback' is not interpreted as the argument to '--upgrade' <cehteh>would one consider that a bug does it try to parse '--fallback' as argument? <lfam>It's definitely working as designed but I'd argue that the design should be changed. <lfam>However, it's not a priority for me to work on this :) <lfam>In the meantime I *always* do pass the '.' to `guix package -u` <cehteh>--upgrade for all and --upgrade-regex <mandatory argument> maybe --upgrade-package <argument> for non regex <cehteh>or the gnu style --upgrade=<argument> <cehteh>that would fix this 'optional' argument semantic too <ng0>ha! I solved my getmail issue <ng0>it was all about the syntax :) <ng0>more details in my follow-up post on the getmail mailinglist <lfam>You'd still have to watch out for '--upgrade-regex --fallback' :) <ng0>so, getmail 5.0 patch is all good from my side <lfam>But I do think that splitting the regex option into something else would be helpful because then you can blame the user more easily ;) <lfam>However, my favorite approach would be a graphical package management front-end for the people who will never ever choose to learn about regexes, which is most people <cehteh>na in the back there should ne aleays some simple command line interface <lfam>It's quite simple once you know it ;) <cehteh>well one with unambigous syntax :D <lfam>Actually, I think it's a great CLI overall. This is one of the only stumbling blocks in my opinion <quiliro>cehteh: i like your solution to --upgrade <lfam>The commands are all discoverable from `guix --help` and you can nest and compose them powerfully <cehteh>some user defined aliases would be nice. like git does <lfam>What's wrong with making aliases the normal way? <cehteh>guix upgrade -> guix package --fallback --upgrade <cehteh>that wont work as nicely, you cant make a 'guix upgrade' alias from the shell <lfam>Ah, because you can't pass arguments? I use shell functions for that purpose <cehteh>guix-upgrade may work, but then you have to remember whats with - and whats without <lfam>I have lots of Git shell function "aliases" like: gci() { git commit "$@" ; } <cehteh>and aliases only work on interactive shell, whereas buildint aliases (from the tool) work from scripts as well <lfam>I see. I never would have thought to use aliases in scripts, since I like to keep them as portable as possible. <lfam>But it's good to learn why people use Git aliases, because I never got it <cehteh>actually i dont use them anymore <cehteh>but for guix it looks like it makes more sense to me <lfam>I think Guix aliases were discussed on the list before, but I don't remember the outcome <cehteh>well its not that important to waste time on it <lfam>I disagree :) It's not a waste if people find it useful <lfam>But of course, someone needs to be motivated to work on it <cehteh>yes .. but i tihnk there are much more important things to fix first <lfam>That's how I feel about this subject too <cehteh>like this rxvt-unicode problem :D <lfam>Do you have any fonts installed and referenced in your ~/.Xresources file? <cehteh>and figure out why xorg hangs in my vm .. looks completely dead now <cehteh>i had no .Xresouces until moments ago <lfam>That might have something to do with the font-spacing problem :) <cehteh>and i'd expect it to start up at least reasonable useable <cehteh>non customized .. but not that crap <lfam>It's surprising how many problems are fixed by the distros, while the fixes never go upstream <ng0>welll... you can have an guix+ script and define aliases as functions in there. <ng0>i have a guix_dev script which saves me some lines of typing <lfam>It's a real problem in some cases, because all the development attention is fragmented between the distros when it should really be concentrated upstream <lfam>Like when mutt development stagnated while distros carried 100k lines of patches against it <ng0>no I mean, distros really try <ng0>but some projects jus treject <cehteh>well the most important problem now is that Xorg is completely locked up :D <lfam>That's true, it's usually a combination of issues <lfam>Yeah, that is a bad problem! <lfam>You should be able to get a console on another TTY and look at logs <ng0>that's how neomutt happened <ng0>by collecting patches <lfam>Yeah, I'm glad that project exists now. Slowly they push them to upstream mutt <cehteh>guix need some knowlege base which is searchable, wiki, or a mailinglist with searchable arcive (guix-trove) or something like that <cehteh>one can post solutions to common problems there <cehteh>ssh session was useable .. Xorg (kvm/splice) was locked up .. mouse was still moving, nothiung in the logs <ng0>afaik neomutt and mutt parted ways in codebase a couple of weeks or versions ago.. there might still be patches going upstream though <ng0>but neomutt is more than just patches now <ng0>codebase is being rewritten <quiliro>how about using synthesis and describe tags to packages for feeding fsf directory? <nckx>About --upgrade: is ‘--upgrade[=foo]’ (only) considered non-GNU? It's my favourite style, and avoids any ambiguity for long options. Of course it wouldn't solve it for short ones. <quiliro>so, in summary: 'guix package -u' does the same thing as 'guix package -u .' <ng0>quiliro: no, I would only want to add or communicate with yet another source (a mediawiki in this case) with the consent (opt-in) of people using the command. <ng0><quiliro> how about using synthesis and describe tags to packages for feeding fsf directory? <cehteh>root 393 0.0 0.0 0 0 tty7 Zsl+ 02:08 0:00 [X] <defunct> <cehteh>switching consoles doesnt work either <nckx>quiliro: could you rephrase that question ng0 quoted? I can't parse it. <quiliro>nckx: what i mean is that directory.fsf.org could use the name, synopsis, description and other tags from 'guix package' <nckx>Oh, OK. Then it's ng0's answer that I don't get. Never mind :-) <nckx>quiliro: I wonder what the copyright situation would be for that. If the descriptions are copyrightable, the individual contributors retain their copyright. On the other hand, the vast majority of descriptions are straight from the home page, Wikipedia, or both. <nckx>I also wish Guix had better and longer descriptions, but no-one seems to like writing them :-( <lfam>Yeah, it's usually the last thing I do <lfam>The worst is when somebody submits dozens of packages without decent descriptions. It feels bad to ask to them to write a bunch of prose <lfam>Usually I do it myself in that situation <lfam>In case anyone is wondering, I used shnsplit (from the shntool package) to split my big FLAC file using the CUE file <lfam>But I had to google to learn that it could do that, so I'll improve the description now <lfam>A modest improvement, but that's how most things happen ;) <nckx>(I'm not saying we should turn 'em into how-to guides or bore people with long and probably outdated lists of supported file formats, but that's a problem I'd be glad to have. It's easier to prune than to write.) <nckx>^ unrelated to your patch which just landed in my in-box :-) <nckx>‘Fun’ ‘fact’: guix pull makes my rusty mail server swap so deep it delays delivery by several minutes. I hope someone™ fixes that soon. <lfam>nckx: Did you read the Guile thread about the excessive memory consumption? <nckx>Yep. But there's no unicode superscript ‘AW’. <lfam>Took me a few seconds ;) <lfam>It makes me especially worried for the armhf port, since few of those devices have >1 GB RAM or fast I/O interfaces <nckx>As long as it's acknowledged as a bug I'm not worried about the long term viability of my herd (hah) of cheap VPSes running GuixSD, but I do hope it doesn't hurt the next release too much. <nckx>People will not like an apt-get update that takes 2 hours on their old netbook. <apteryx[m]>Guile is using 1.571g as we speak (building guix). <lfam>I think it will increase to ~3 GB at the peak usage <lfam>During boot I see that some compiled objects are rejected as if they are malformed, or at least Guile 2.0 objects, and so Guile could be trying to rebuild them during early boot <quiliro>i would be available to edit the descriptions in order to include better descriptions to the packages <lfam>But I didn't have time to investigate that bug more yet <Apteryx>lfam: Nope, 1.571g was about the max I saw while building guix. It just finished now. <quiliro>that would be my contribution to guix <quiliro>how can guix package -show output everything that it does but for only one package and not all mentions to that regular expression? <quiliro>for example 'guix package -s emacs' shows a lot of packages and not only the emacs package <lfam>-s is short for --search <lfam>There is no short form for --show <quiliro>what do you think about this statement: the use of a software is its reason to be...but its reason to be is no justification for the user being under control of the program instead of the other way around....hence freedom is essential on software <rain1>thats important to me quiliro <lfam>I agree with that statement. We made computers and we should control them. <quiliro>i like it....but think that it can be improved <quiliro>oh! i am already learning...used M-x guix <quiliro>easy....just need to tinker a little more :-D <quiliro>ACTION seems as a kid with a new toy <joshuaBPMan_>hello, I just created a guix system vm ... I guess all of that information is being stored in the /gnu/store... <lfam>JoshuaBPMan_: How did it fail to boot? <reepca>Well, turns out the old libre-friendly GPU I got, despite having 3 video outputs, only supports 2 at once. Thankfully I accidentally got another one during my quest to get multiple displays working, and at last, although only one's processing capability is actually being used, I have 3 displays running \\o/ <Apteryx>Has anyone tried WebRTC solutions such as Jitsi Meet on GuixSD? <reepca>only downside is that the third monitor, being much newer, makes me realize how dark the others are <lfam>Turn down the new monitor ;) <lfam>I've had clients ask me to do that many many times <reepca>on a related note, it seems like redshift isn't working properly. At least, when I try running it it doesn't seem to change anything. <lfam>Hm, it works for me. How do you invoke it? <lfam>Ah, that's not enough :) <reepca>¯\\_(ツ)_/¯ was enough on ubuntu, what am I missing? <lfam>On Ubuntu they must install with some geolocation service and a default color change value <reepca>redshift -h indicates that ours has a default color change value as well <reepca>and when I run it it does say "using provide geoclue2" <lfam>Well, this works for me: <lfam>redshift -l LAT:LON -t 5700:2700 -m randr <lfam>Where LAT and LON are my coordinates <lfam>I stay at roughly the same longitude, so I don't need to change it often <lfam>Btw, I wonder how it uses geoclue2, since geoclue2 is not an input. Perhaps it comes transitively from another dependency <lfam>Ah, the package definition says that it comes through glib somehow <lfam>Btw, I'm using i3 on X. I don't know how well it work with a desktop environment or other window manager. And I have no idea about wayland <lfam>Okay, so that invocation should "just work" <lfam>I wouldn't mind if this feature was integrated more deeply into the display server somehow. But that's my fault for cobbling the system together <lfam>I wonder why `redshift` isn't enough for us <reepca>I suspect it has to do with geoclue not working quite as intended. It's in %desktop-services, but specifying the coordinates was what made it work for me. <reepca>Perhaps it has something to do with redshift not being in the list of whitelisted applications allowed to access geoclue (gnu/services/desktop.scm) <nckx>reepca: Indeed likely. That was the problem when I tried — long ago — to run redshift on NixOS. <reepca>I should probably try to figure out what that dbus thing is sometime, and how merely listing a program's name can provide any security. <lfam>Right... could one write a script called 'firefox' and use it to request location data? <cehteh>methinks that redshift makes it a bit overengineered about the geolocation ... and i dont like the idea that it wants to connect the internet either <cehteh>some simpler more intuitive thing a la country/nearest_city would be easier, and a database/textfile from few 1000 big cities around the world should be trivial <lfam>It seems like redshift itself doesn't connect to the internet but instead delegates that to geoclue which is part of the desktop service. <lfam>Well, it's not necessary, as I pointed out earlier :) <cehteh>its not that i dont trust redshift, its about that idont want to initiate a "figure my location out" request <cehteh>yes, you can hardcode the coords, but thats a bit unintuitive for most users, or do you know your coordinates? <lfam>I had to look them up. But, I think that the number of people using redshift and not using it as an integrated part of a desktop environment that uses geo-ip is extremely small. <lfam>Not that many people will put an operating system together from spare parts so that they would need to type the coordinates or city name <cehteh>using the timezone unless more exact specified would already work for a lot people <lfam>Yes, I wonder if it can take advantage of the tzdata db <cehteh>thats not enough to determine exact dusk/dawn times .. but close enough <lfam>The exact times are not necessary since the color change is very slow. Plus, weather will change the time the ambient light changes <lfam>There is a graphical program redshift-gtk, maybe it's more accessible to lay users <cehteh>before redshift i used the laptop webcam to get color and brightness of the environment, the only flaw was that i never figured out how to set the color temperature (i only modified color gamma) <cehteh>i still have a script hooked on acpi buttons which does that for brightness <lfam>We aren't building the graphical program, it seems <cehteh>self learning ... when one readjusts the brightness it adjusts a table, and the think-button oft the thinkpad does an autoadjust <quiliro>rekado_: were you able to boot again? <nckx>quiliro: I'm writing a short reply to your fsf-directory e-mail. Mind if I cross-post it to guix-devel? <quiliro>nckx: i was thinking of doing that myself.....but did not because i would be bashed for cross posting.....if you want to take the bashing, go ahead...i am usually the one that people bash...but it is nice to share ;-) <nckx>Hm, I've never seen anyone bashed for that but I guess most cross-posting I see is of the guile/guix variety, which is more like sliding a note across the table. <quiliro>it would be nice to have a syllabus to follow in order to lear to hack guix <quiliro>d-d? it is my first time on both lists <nckx>directory-discuss, sorry. I'll just send my reply to guix-devel separately. Peeps can cross-post if they so desire. <quiliro>nckx: haha....you were a chicken just like i was <nckx>(So yes, I didn't sleep yet. I tend towards 48-hour days when no-ones asking otherwise.) <quiliro>i dont remember if i am at guix-devel or only at guix-help <Salt>anyone around who's attempted (/succeeded) compiling Guix+toolchain, specifically into a non-standard path? <nckx>Salt: what do you mean by ‘into [a] path’? <quiliro>looper123: but you can learn how to build it...you need to be persistent and patient....then you can so it....i am a neewbie and i am learning how to package <nckx>Salt: The default ‘/gnu/store‘ location should be fully configurable, if that's what you mean. Unless you've run into problems. <Salt>nckx, I do not have root on the HPC, thus having to install to a non-privileged directory. <nckx>Salt: It just means (re)building everything from source all the time, which can be prohibitively expensive. <quiliro>nckx: 48 hours! i will never do that again <Salt>I've been trying to modify nix-no-root to bootstrap guix, just having issues since the available gcc is quite out of date as are many libs <Salt>we don't mind building everything, in-fact it is somewhat preferable due to being able to change flags, the problem is in bootstrapping guix at all <nckx>Salt: Ah. There have been some recent posts to guix-devel about that, specifically about faking /gnu/store in an unprivileged environment (using proot) to allow using regular substitutes. <nckx>Salt: Pjotr and Ricardo in particular, I believe. Probably others. Lots of HPC geeks using Guix. <nckx>Salt: Did you see those messages? <Salt>yes, I've been in touch with Pjotr and have been attempting things from past mailings :) <Salt>though I haven't actually tried proot yet <Salt>as I've never used it before and thought that bootstrapping would be easier (ha!) <looper123>quiliro: I most likely fail in doing so. Well, openembedded files might be readable enough to give a conversion a try but I'm not familiar with the guix packagemanger nor guile/ sheme. I just read hurd is planed to be build with it so I wondered why not this too? <nckx>Oh. I'm not active in that area myself, so can't give much info without risking misinformation. But I have used proot with some success on a shared system. And that was ages ago. Things can only have improved since. <nckx>ACTION is really going to bed now; I need to be able to drive in 3 hours. <Salt>nckx, thanks much anyhoo, I figured I'd ask here as I'm digging into my scripts yet again <rekado>sneek later tell quiliro About tor: check your system configuration file. You probably have tor-service there. You need to remove it and run "guix system reconfigure /etc/config.scm". <rekado>I'm still/again backing up my files before reformatting my disk. I won't be able to review patches today <efraim>qemu.scm got renamed and then the translation stuff got fixed in a later commit, make sure you don't have a qemu.go anymore <ng0>thanks :) I'm waiting for make to finish <ng0>I ran make clean-recursively <efraim>I try to avoid it on my 2GB machines so I check the commits normally <ng0>so probabyl the .go was in the way <rekado>"do I have to blame someone for an untested commit" -- never <ng0>it would be good if the errors could be different when you have an .go file for an .scm file which no longer exists <efraim>the guix build flag '--with-input=' is recursive, isn't it? ***jonsger1 is now known as jonsger
<sneek>quiliro, you have 1 message. <sneek>quiliro, rekado says: About tor: check your system configuration file. You probably have tor-service there. You need to remove it and run "guix system reconfigure /etc/config.scm". <quiliro>rekado_: thank you...i was told that and it was exactly that by nckx ....it was happily solved.... <quiliro>rekado_: were you able to recover your files? ***jonsger1 is now known as jonsger
<quiliro>got disconnected because of this bothersome usb wifi...damn apple...why couldnt it use hardware that released its drivers? <quiliro>rekado_: i did not get a chace to read if you answered <reepca>which of the java packages provides jni.h? <quiliro>that is something i would need in guix...the ability to find which package provides a certain file...i think ap-file does that <quiliro>or even there is a package in debian which finds which package provide a certain command <reepca>naively asking hydra to "find /gnu/store -name $FILE" could give weird results though. According to running that on my system, jni.h is provided by ffmpeg. <efraim>Apt-file on my aarch64 box also says gcj (gcc), and openjdk <reepca>ah I think I see now, didn't realize icedtea had a separate jdk output <quiliro>i did 'guix system reconfigure lightweight-desktop.scm' successfully <quiliro>and (console-keymap-service "dvorak-es") worked perfectly! <quiliro>but it is not installed in the user...nevertheless the command loadkeys exists...but the command cannot find dvorak-es <quiliro>Couldn't get a file descriptor referring to the console <quiliro>even after installing the kbd package <nckx>quiliro: You're welcome! That's weird. It works perfectly here (for any user, and so does running ‘loadkeys en-latin9’ manually). <nckx>I recently fixed a bug where kbd wouldn't search for keymaps recursively. Are you running HEAD? <nckx>If so, and dvorak-es.whatever.the.extension exists, it should work. <nckx>Which is an absolutely useless statement to you at the moment, but oh well :-) <nckx>Where does ‘man kbd’ say it wil look for keymaps (/gnu/store/...)? Wat is $LOADKEYS_KEYMAP_PATH set to, and does it exist? Etc. <quiliro>it is so great to have guix irc log! <quiliro>nckx: $LOADKEYS_KEYMAP_PATH is empty <quiliro>nckx: the problem is in the desktop and on the graphical login <reepca>Is there a way to access the inputs from outside of the phases? Similar to %output? <reepca>ah never mind, %build-inputs, got it <ng0>quiliro: you want 'setxkbmap' <ng0>in GUI login ,there is no easy fix <quiliro>i dont understand why the arbitrary choice of en, instead of the one in the console <quiliro>the guixsd installer asks for loadkeys <quiliro>then that keyboard layout should be used throughout <quiliro>is it me that is doing something wrong, or is it me that has the wrong concepts, or is it a bug? <quiliro>i hope i can install openmolar today <quiliro>i am now able to read and send emails in gnus....i still have to learn how to move them to direetories <quiliro>i still have to learn how to package by myself <nckx>quiliro: sorry, unexpected visitors. I have to leave. Good luck! <quiliro>and could not do it with openmolar yet <quiliro>nckx: visitors are always nice to have <Apteryx>Hmm, has anyone seen this: configure: error: found development files for Guile 2.2, but /gnu/store/a88g2d4nf6dw4xbn1xilwhcyzvciaryg-profile/bin/guile has effective version 2.0. I'm trying to build an old tree. <amz3``>I got this issue several time, not sure how I fixed it <lfam>Apteryx: I think you should re-run `guix environment --pure guix -- ./configure --localstatedir=/var`, assuming your localstatedir is /var <lfam>It looks like Guix is configured to build with Guile 2.0 <Apteryx>Same error. Hm. Maybe I should just rebase and build that huge mess another day (it's a change to the gnu-build-system which rebuilds the world). <quiliro>rekado_: what did you do to recover your partition? did you reformat? <lfam>Okay, then you need to use `guix environment --pure` to manually provide Guile 2.0 for this old tree <lfam>I assume you're working with a commit that pre-dates the switch to Guile 2.2? <Apteryx>That branch is on top of 2b95f247215345c9130b5d6623d739f810224313 which is dated May 28th. <lfam>If it pre-dates the switch to Guile 2.2, you'll need to make Guile 2.0 available, and also make sure there are no Guile 2.2 objects in the tree (make clean-go) <Apteryx>OK, so I would create my guix build environment using "guix environment --pure guix make guile@2.0"? <lfam>Wait, are you trying to build Guile or Guix? <lfam>Oh, I misread your command ;) <lfam>I would not include guix in the arguments to `guix environment` unless `guix environment` also pre-dates the switch to Guile 2.2 <lfam>Because the environment for building Guix since then uses Guile 2.2 <lfam>If it were me, I'd try manually listing the dependencies from the Requirements section of the manual <Apteryx>Any recommendation for writing a Word .docx document in Emacs? Or should I bite the bullet and install LibreOffice... <ecraven>Apteryx: libreoffice can convert from the command-line somehow <ecraven>so just use org-mode, export to .doc and convert <Apteryx>is there a .doc exporter from org-mode? <ecraven>libreoffice can convert that to docx <Apteryx>I see! Cool solution. Will try that. <paroneayea>C-u M-x run-python <RET> guix environment -l /home/cwebber/devel/pyld-signatures/guix.scm -- python -i <RET> <paroneayea>and use python within emacs, using the guix environment stuff <Apteryx>I'm happy to see a healthy bayfront server! It downloads at 1.5MiB/s here. <cbaines>paroneayea, you might want to look at direnv and emacs-direnv, as I find that very useful for doing similar things with ruby <cbaines>you have the .envrc file, which is like a bash script to configure the environment <cbaines>in that you can do something like: eval "$(guix environment --ad-hoc python --search-paths)" <cbaines>and there is support for this in the direnv stdlib as use_guix <cbaines>so you can do something like: use guix --ad-hoc ruby <cbaines>and then from emacs, with emacs-direnv installed, it will modify the emacs environment with direnv <paroneayea>cbaines: super cool... I'll have to look into it more <cbaines>direnv and emacs-direnv are packaged for guix, so it shouldn't be too difficult <rfmj>what package is "clear" command line? <efraim>Or from ncurses if it isnt in bash <Apteryx>It seems the lapack maintainers cleaned up their tarball in-place instead of making a new release. The hash as changed, and lots of hidden/test files are now gone in the upstream release. <Apteryx>Should I open a bug about this, expressing my concerns that their tagged releases are mutating? <Apteryx>They use branch. So that's bound to happen again. <quiliro>rekado_: were you able to recover openmolar configuration? <efraim>Apteryx: does archive.org have a copy of the old tarball? <jonjitsu>If I have multiple distros on my system and wanted to share one store among them, is it enough to put /gnu on it's own partition or would I have to share /var/guix too? <quiliro>i had this problem before....i do not remember how to solve it <quiliro>neither machine con connect to the other via scp but the can through ssh <nckx>quiliro: re: keyboard layouts, that's just the way things are. The kernel keymap isn't used by X and vice versa. GuixSD doesn't add an abstraction to set both at once like certain other distributions. That would involve translation and lossiness, too. <nckx>quiliro: again, all I can say is that scp should work (with ‘openssh’ in your operating-system packages) and does here :-) <quiliro>how does the graphical login manager (elogind?) know where to get the keyboard layout? <nckx>Bayfront's back? Bayfront's back! <nckx>quiliro: no idea, sorry. I've never tried X on GuixSD. <nckx>quiliro: and ‘(operating-system (packages (list openssh ; for scp...’? It used to be required, at least. <quiliro>nckx: i am connecting via ssh...it should be up, no? <quiliro>plus i have scp command on installed on both ends <quiliro>nckx: what does ‘(operating-system (packages (list openssh ; for scp...’ do? <nckx>Installed in the way I pasted above? Then I don't know what the problem is. Try grepping the guix tree if you're stuck, I really thought it was documented *somwhere*. <nckx>Installs openssh (with scp) into your system profile (remember that? :-) <nckx>quiliro: ^ ‘(operating-system (...’ does that. <nckx>quiliro: I have to leave again. Try reconfiguring with that in your config.scm and scp should work. <reepca>how does scp find the remote-scp binary? would it look in PATH? Wouldn't that then depend on the profile of whichever user you're logging in as over ssh? Is /etc/profile sourced by default when ssh'ing? <jlicht>reepca: There should be an entry in .bashrc on GuixSD systems AFAIK, dealing with literally this <reepca>it seems to only source /etc/profile if "cat" can't be found. I wonder why that is. <jlicht>reepca: I guess that `cat´ would also be installed in a profile. So basically a heuristic that assumes if cat is not in path, this is probably a guix-based machine <jlicht>reepca: but to answer your original question: I guess scp finds remote-scp in PATH, but does not source /etc/profile automatically (only if you have that line in your bashrc) <quiliro>what imean is that i still get that error after source /etc/profile <quiliro>scp -r -P2222 192.168.100.102://mnt/depurados/Descargas/pumps ./ <reepca>quiliro: could you temporarily edit your .bashrc and replace "if [ -n "$SSH_CLIENT" -a -z "`type -P cat`" ]" with "if [ -n "$SSH_CLIENT" ]"? <reepca>then just try using scp as normal <ng0>quiliro: are both GuixSD? <reepca>there are quite a few shell initialization scripts for various uses - among them are /etc/profile and ~/.bashrc <ng0>I haven't read all of this conversation, but "scp" is only exposed systemwide if "openssh" is part of the (packages) in the operating system config file. <reepca>/etc/profile gets run (technically "sourced") whenever an interactive shell is started. In GuixSD /etc/profile delegates setting of environment variables and such to the user-specific stuff in ~/.guix-profile/etc/profile <reepca>when you start a normal ssh session, that's interactive, so /etc/profile gets sourced and you get your PATH set to what the user would normally have. <reepca>but when you use scp, it's non-interactive, so only the initialization script for bash (~/.bashrc) gets run. <reepca>~/.bashrc as it was before you changed it tried to guess whether it needed to source /etc/profile based on whether it could find "cat" (really common). But "which cat" in my case at least indicates that it's installed in the system-wide profile (I'm not sure how the default path gets setup). So /etc/profile wasn't getting sourced when an ssh session was started by scp, so the user-specific profile wasn't getting initialized. <reepca>Of course, the proper solution to this is, as ng0 says, to have scp installed in the system-wide profile. We can check whether this is already the case using: guix package -p /run/current-system/profile -A openssh <Apteryx>Found a GCC bug trying to build vigra@1.11.0. Can someone reproduce this? <quiliro>$ guix package -p /run/current-system/profile -A openssh <quiliro>openssh 7.5p1 out gnu/packages/ssh.scm:125:2 <joshuaBPMan_>Hello, I'm about to try to install guix on my work machine. I figured I might as well try installing guix using the new ncurses option, just to test it. Is that available in the default installer? Or do I need to download something to use that? <ng0>reepca, quiliro that solution is an 'open undocumented fact' and many people run into this.. or maybe it is documented and I haven't read that part of the docs in a while.