IRC channel logs
2023-01-05.log
back to list of logs
<panosalevro>how do i know which use-modules form is needed for a specific variable? when trying to do a system reconfigure, it says that 'swap-space' is unbound <podiki[m]1>rekado: but sending to a debbugs number only goes to...? where exactly? even as author of an issue I don't get replies unless it includes my mail (and I'm not subscribed to the bugs or patches list) <tremon>hmm, the Kea test suite is failing because of DHCP6_INIT_FAIL: unable to find suitable interface for generating a DUID-LLT. is there a simple tweak I can use to provide a mock network interface to the builder, or is it quicker to dive in the code and disable those tests? <podiki[m]1>jman: okay, confirmed it is a bug, I have a fix but not sure if it is strictly correct (I think it is). let me know if you filed a bug otherwise I'll submit a patch later anyway <jman>podiki[m]1: amazing, thank you! yes I've sent an email to bug-guix@ but - as before - I still didn't receive the bug acknowledgement reply :) <nckx>The system can be very slow (hours; exceptionally days). These delays are far more likely than loss of messages. <nckx>We don't run that system and have little insight or recourse when things go south. <oriansj>does anyone know why guix build doesn't have an option to download and build a package by name? <oriansj>so that one could do something like this: guix package -A | awk '{print $1 "@" $2}' | xargs guix build to download all guix sources and build all guix packages <apteryx>the_tubular: it was forgotten on the patch tracker <the_tubular>Most distro updated it in 2022, I'm not sure I understand <the_tubular>ACTION slaps singpolyma around a bit with a large trout <oriansj>the_tubular: walleye would be much more effective <David[m]1>is it possible to install gnu guix system on a raspberry pi? <oriansj>David[m]1: yes but it is a little painful if you turn off substitutes <acrow>There seems to be a problem with qa.guix.gnu.org... certificate error... <David[m]1>with an official image can I expect to be able to use all of the pi's hardware? <oriansj>David[m]1: not the hardware which lacks free drivers <acrow>podiki[m]: hmmm... I'm still having the problem. Let me try something else. <podiki[m]1>the page loads for me but on the patches page it doesn't show anything but mentions a certificate error <acrow>podiki: I've got the same error on both icecat and flatpak/firefox... still I did just do a guix pull, update, etc and I haven't logged out and re-logged in. <podiki[m]1>yeah, page loads (not a certificate problem on the qa server) but get "#<x509-certificate 7f2240f506b0>patches.guix-patches.cbaines.net#<gnutls-certificate-status-enum expired>#<gnutls-certificate-status-enum invalid>" on that page instead of content <podiki[m]1>cbaines: seeing a "tls-certificate-error" on QA pages <podiki[m]1>yeah, that looks like an internal error, not one of outside reaching QA server <podiki[m]1>not sure what time zone cbaines is in, maybe a note to guix-devel is better <CoinCoinME[m]>should i install a extra package to add additional grub menu like system windows for boot other system? <KarlJoad>Is there any documentation on the bag representation that Guix uses? I ask because of the new blog post about derivations. <lechner>CoinCoinME[m] / Hi, you probably have grub already, but dual boots are hard to manage. If you have UEFI, it may be better to use two Grubs and rely rEFInd or on EFI variables to switch between them <lechner>acrow / Hi, is the qa issues link the same content as MUMI? <apteryx>shouldn't a package having the '(supported-systems (list "armhf-linux"))' field prevent me from building it on my x86_64 host? <apteryx>it just prints a warning: warning: package raspberrypi-userland@0.0.0-0.54fd97a does not support x86_64-linux <podiki[m]1>KarlJoad: probably coming in a forthcoming blog post in that series? I forget what was planned but you can see more either on mailing list or bug tracker for the posts <KarlJoad>podiki[m]1: I will have to check the mailing list then. The bag format is one that I never heard of on Nix and I think is unique to Guix? <podiki[m]1>I don't know, I look forward to hopefully reading more about it soon though :) (I assumed most of the low level stuff like that came from Nix, but I don't know) <ConvolutedSquare>I have a simple shell script that uses sed, cp, and grep. How do I go about running it in a GEXP? End goal is to have it run as an activation script. <lfam>the_tubular: Regarding that CVE from 2018, I did forget about it since March. But, although it was discovered and fixed in 2018, the public didn't notice it until 2022. <ConvolutedSquare>lechner: what does the match part do? Is it required or can I keep it lean like the python example? <ConvolutedSquare>Oh looks like it's just to parse command-line arguments for when the script gets executed, right? <acrow>lechner: you caught me away from keyboard -- that link fails with a tls-certificate-error. <acrow>It does not appear to be a failure specific to that one issue. It also seems to have the side effect of causing all the QA tags to flip to 'unknown'. https://issues.guix.gnu.org/60537 for instance is now tagged with QA status 'unknown' but I happen to know that it was succeeding earlier this afternoon and there have been no new changes. <acrow>led-lightbulb: That is the subject of that issue (starting at [PATCH]) but I don't understand the earlier things in that line. <lechner>my question was more about the helpful bot <acrow>lechner: oh, I think I see, so an issue, )Xng(, has been opened. I think that issue demonstrates this tls-certificate-error but, probably many, other issues are also impacted, e.g. https://qa.guix.gnu.org/issue/60535, also produces the same tls-certificate-error. Are we on the same page? <acrow>lechner: I think the bot idea is good. People here on IRC can sorta observe the hands of progress. :) <lechner>acrow / not sure about the certificate error. the bot only links to the issues <acrow>lechner: so if I mention an issue here it will provide a status, no? <lechner>or just #60537, as well as a bunch of URLs <apteryx>lechner: do I understand that the led-lightbulb bot is under your control? <lechner>apteryx / only in the sense that i am at your, and the project's, disposal <lechner>i.e. i would be honored to donate it <apteryx>OK! I'd have a suggestions, if I may: would it be possible to streamline the output, removing any extraneous info (the leading )???( code and the short t.ly URL <apteryx>the rationale being that the full issues.guix.gnu.org URL (which contains the issue #) is already printed <apteryx>otherwise, nice initiative, I like it! <podiki[m]1>maybe a few characters for the QA status even? (though that's on the issues page already) <apteryx>full name could be abbreviated too, for a bit more privacy (e.g., from Mike F.) <podiki[m]1>I wonder if this could become a way to follow bug changes (though might get to be too much traffic here?), i.e. botname follow bug #xxxxx and you'll be pinged with QA changes or something else <apteryx>perhaps if the bot sends you private notifications <podiki[m]1>I'm still confused by when/if debbugs sends responses to anyone, is it only if their email address is added? as the bug number address doesn't seem to go to the submitting author <podiki[m]1>I guess this could be part of mumi/qa as well, via email <apteryx>I think debbugs is basic: sends to you if your are registered to the mailing list, or if your are in the X-Debbugs-CC: mail field <lechner>apteryx / well, i am not sure i can effectively advocate for shorter bug identifiers without that information. as maintainer of Lintian and lintian.debian.org I have potentially responded to thousands of bug reports. i'd like to keep Guix from making a base 10 mistake <apteryx>ah, so the )xxx( thing is an alternative unique ID for the Guix bugs? <lechner>if you squint, the bug is being horned. or perhaps squeezed by Lispy parentheses. i like the symbolism of both <apteryx>OK, I see the Huh? short URL is to address that. I had actually already read it once already. <lechner>least significant digit comes first, base 60, and no zeroes or O's <lechner>someone on #guixcn even offered a Guix REPL <lechner>sorry about the code style. i am a Guile (and Scheme) noobie <lechner>as odd as the "horned" bug IDs are, we would never need more than four digits (12MM) <lechner>flatwhatson / Hi, '=' stands for the zero digit, which I abolished together with the lowercase and the uppercase Os in order to avoid any confusion among them. You can try )=1( or )==1( <apteryx>podiki[m]1: re debbugs, the best practice is to always wide reply and include everyone in the discussion <apteryx>that way people filtering on Return-Path gets the reply in their inbox <podiki[m]1>apteryx: I agree, but I sometimes see replies (via our frontend) as someone might just send to the bug number (easier to reply to all via emacs debbugs, otherwise it is not as obvious I think) <ArneBab>I’m seeing plasma-desktop in the package list, but the texinfo still says that kde is missing. What’s the state of KDE in Guix? <nckx>oriansj: Heh, that's basically how guix.tobias.gr builds the world. None of that 'features' CI bloat! <sneek>Welcome back nckx, you have 1 message! <nckx>I've heard of it but not tried it. <rekado>podiki[m]1: apteryx: I’m on an unhealthy diet of dog food by only interacting to issues through mumi, so I send mail almost exclusively to <number>@debbugs.gnu.org <rekado>you’re saying that I *have* to explicitly include the submitter’s email address in my replies so they get sent a message? <rekado>lechner: my lisp-addled brain refused to parse “)…(” as delimiters. Can’t even type them! <nckx>lechner: I know I just bikeshed it yesterday, but... Other contextbots don't ping the OP. I think making it less intrusive might help its reception. The OP by definition knows the subject, so the ping could be interpreted as 'hey, *you*, use my new hornycode notation'. <nckx>Also they are called hornycodes now sorry. <ConvolutedSquare>Hey nckx, any chance you know how I can trigger a gexp right after a system reconfigure? Activation scripts get called before bootloader installation :< <nckx>ConvolutedSquare: I have to go, but I think Shepherd restarts services after GRUB is installed, no? Not a recommendation per se (I don't know what you're doing, but do appreciate the updates & will catch up later). Relying on stateful triggers ('hooks') is quite unguixy. <rekado>lechner: do you have a git repo for the bot? Or does it live in paste.debian.org posts? <ConvolutedSquare>nckx: I am doing something quite disgusting, frankly. Trying to sed grub.cfg to fix up the paths, and then copy all of the kernel/initrd files over to the root /gnu/store. I am not exactly sober either, and it's 2:20AM. Can't quite get it to work but I think this is my limit for tonight :P <jman>podiki[m]1: w.r.t. guix container + FHS emulation -> ticket #60566 now created (if you want to link it to your patch) <lechner>rekado / Yes, as far as I know Debbugs does not copy anyone on bug comments (although the Debian fork copies the package maintainer there) <cbaines>bayfront is having problems as the Guix Build Coordinator is using excessive amounts of memory (64GiB heap). This'll affect things like the website. I'll kill it, once I can SSH in... <Guest54>it just makes me feel like im actually in control of my system <civodul>cbaines: ouch, thanks for the heads-up! <cbaines>I actually didn't manage to SSH in before it seems to have died on it's own, maybe the OOM killer <cbaines>I'm going to add earlyoom to bayfront to try and make this less disruptive when it happens <elevenkb>say I have a `guix shell` that makes some info manuals, what is the best way to make them available to my home profile's INFOPATH? <lechner>rekado / thanks for the suggestion to guile-irc. it is lovely <elevenkb>adding `texinfo` as a dependency does work. <elevenkb>lechner: yep, now just need to set up `emacs-envrc` to reset `info-directory-alist`. <civodul>maybe we could run lightbulb on project infra? <mirai>does debbugs control allow tagging of issues by someone who is not the original submitter? <sneek>Welcome back mirai, you have 1 message! <sneek>mirai, nckx says: It did! I'm subscribed to the list and not (currently) using a MUA that's clever enough to highlight ‘personal’ mail, but it worked. <civodul>lechner: there's a lot of stuff in that maintenance repo! <lechner>civodul / yes i have been reading, and learning lots. thanks so much for Guix! i am a total convert <sneek>unmatched-paren, you have 1 message! <sneek>unmatched-paren, podiki[m]1 says: looking forward to reading your blog post in full detail with a nice cup of tea today :) great work! <civodul>we'll shorten the publication delays next time, i promise ;-) <ekaitz>unmatched-paren: now I can find you here: good job with the blogpost <ekaitz>this series is something I've been waiting for for a really long time <oriansj>nckx: is the setup of guix.tobias.gr available so that I learn from how it? <nckx>oriansj: Nope, it's proprietary shame-based technology. <nckx>I just wanted to point out that your example wasn't hypothetical. <lechner>Hi, is the certbot service supposed to renew my certificates automatically? <nckx>Leave the Guix installer running long enough and you'll start receiving rottlog dead.letters. <nckx>lechner: That is its purpose. <lechner>nckx / what if nginx is also running on the same host?] <nckx>That's how it renews them, I believe. By using (extending) the already running nginx service. <nckx>Well, that is what's supposed to happen. I gather it's not happening. <nckx>There's also an eggchicken if you don't have certs yet, because then nginx will fail to start. <lechner>that part i solved, but i only have six days left! <nckx>(E.g. /var/log/letsencrypt/letsencrypt.log) <nckx>I don't use Certbot for my own LE certs but it's what berlin uses, and most renewals often work a lot. <unmatched-paren>writing this blog post will be hard because i don't actually understand the store monad myself :) <lechner>nckx / looks like my webroots are not set correctly. how do you manage your LE certificates, please? <nckx>I call dehydrated from mcron, but there is no Guix service for that. <nckx>And I use dns-01 challenges, not HTTP. <nckx>So nginx is blissfully uninvolved. <nckx>Problem with ‘my webroots are not set correctly’ is that I can't help you with only that info, and once you type enough details to be helpful you'll automatically know what to fix. <nckx>(berlin's working configuration is in maintenance.git, by the way, as is bayfront's I believe. They're complex though.) <lechner>nckx / thanks! as a certbot contributor i'll get those webroots set up right before long, but i am curious about dehydrated. which DNS server do you use, please, and how do challenges propagate from the host that actually needs the certificates? <mirai>nckx: did you write custom snippets to use with dns-01 as well? <mirai>lechner: did you run the /var/lib/????/renew_certificate.sh manually before? <mirai>and did you rsync the letsencrypt directory from another host or is this a pristine setup <lechner>mirai / from what i remember, it was a pristine setup. i pulled new certificates manually while nginx was stopped. i never liked the HTTP challenge method though <mirai>by pulled you mean running the renew_certificates.sh script right <lechner>no, i ran certbot certonly --standalone <nckx>lechner: My DNS set-up is unsophisticated. I run Knot on the same machine, so I can just nsupdate localhost and there is no need to exchange secrets. <mirai>lechner: there's your problem <mirai>the service sets up some symlinks, permissions and idk what else and if you copy the directory or mess with the directories it won't work as intended <mirai>does it look like there have been renewals lately? <mirai>nckx: your dns setup with knot sounds interesting, are you hosting the nameservers yourself or do you have a split-horizon setup? <nckx>You're still thinking too sophisticated :) <nckx>I host one myself (a.ns), the rest is an army of free secondaries I've cobbled together. <nckx>The puck.nethers of this world. <lechner>did you read that jared won a county contract to provide rural service? <nckx>Already a precious in my collection. <nckx>(Yes, I brand them because I'm a twit, and also I was providing services for a while [hence that hook] and went full pre-make fake.) <nckx>lechner: No, but I'm happy for him. <mirai>admittedly it's incredibly rudimentary <lechner>mirai / thanks for that certificate website. there have been no renewals <mirai>and that I never liked the sleep workaround <nckx>I used to poll, but couldn't justify the complexity vs a single if inelegant ‘sleep’ line. <PotentialUser-10>guix package -l " I can see the vim is installed, however I can't launch it. <lechner>what's the output of env | grep -e ^PATH <nckx>You must forgive me for biting, but why not ‘echo $PATH’? <lechner>i like the equals sign up front, i suppose <PotentialUser-10>PATH = run/setuid-progrmas:/home/username/.config/guix/current/bin:/run/current-system/profile/bin:/run/current system profile/bin:run/current system profile <nckx>We all have strange desires. (And I guess ‘declare -p’ is too noisy.) <nckx>$HOME/.guix-profile/bin should be in there. <nckx>This is after sourcing ~/.guix-profile/etc/profile? If so, what does it contain? <PotentialUser-10>my guix_profile includes /bin /etc /games /include /lib /libexe manifest sbin and share <nckx>I meant etc/profile specifically. You'll see that it sets PATH. It is, to say the least, odd that it would fail to do so if you sourced it. <nckx>But from ‘progrmas’ I gather that you can't paste(?). <PotentialUser-10> yes I can't copy paste because I didn't installed any X server on my virtualbox so I have only have vt/ <nckx>No problem. There's wgetpaste in Guix, but that implies installing :) <nckx>You could always install it, invoke it as ~/.guix-profile/bin/wgetpaste. <PotentialUser-10>I understand, so I entered the command cat profile.However I think it is empty <nckx>I assume that's short for ‘cat ~/.guix-profile/etc/profile’? <nckx>So its size (ls -l or whatever) is zero? <nckx>Could you install any random new package (such as wgetpaste) and see if it remains empty? Guix will generate a new …etc/profile each time you create a new, unique new set of packages. <lechner>mirai / which symbolic links does the certbot service create that i may be missing, please? <mirai>lol, raw.githubusercontent.com seems to be blocking wget <PotentialUser-10>I have realized a very strange thing if I type ls -alF the time for every file is 1 jan 1970.also profile is 0 byte <PotentialUser-10>if I type guix install it gives error guix install: error:unspoorted manifest <lechner>PotentialUser-10 / the dates in the store are all fake. your issue is that the contents are missing <mirai>lechner: I have no idea, I deleted my directories but that was only when I noticed that I was getting issued certificates left and right but they weren't being saved <mirai>in your case I don't know, maybe the service does work but it renews a bit closer towards the expiry date <mirai>\/var/log/letsencrypt I think <mirai>what's in /var/log/letsencrypt/letsencrypt.log <lechner>PotentialUser-10 / how about which guix ? <lechner>PotentialUser-10 / that looks right. i am not sure what to do and would probably reboot <lechner>what happens when you run guix describe please? <lechner>you can try sudo herd restart guix-daemon <PotentialUser-10>it prints the package installed, but at the end it will print unspoorted manifest form <PotentialUser-10>I installed guix system on my virtualbox with automatic systm instalation I didn't configured anything yet. <lechner>are you doing this as root or as yigit? <lechner>did you try sourcing the profile again? <lechner>i think your environment may not be right <nckx>lechner: No, the problem is apparently that the file really is empty. <nckx>PotentialUser-10: Can you run ‘guix gc --verify=contents,repair’? That's a bit extreme, but I'm not sure what's going on. This file (really: a symlink) is generated by Guix, it's not something that depends on your environment and it should never be empty. <nckx>For comfort, you can run PATH="$HOME/.guix-profile/bin:$HOME/.guix-profile/sbin:$PATH" in your current shell(s). It will temporarily add everything you install with ‘guix install’, so you can at least use vim etc. <PotentialUser-10>nckx I have run the command guix gc --verify=contents,repair. however my ~/.guix-profile is still empty and I can't run anything related to guix pacakge manger without getting the error: unsupported manifest format <nckx>PotentialUser-10: Did it output anything about damaged files? <nckx>Which versions do you get for: grep -m1 version ~/.guix-profile/manifest ~/.config/guix/current/manifest <nckx>Should be 4 & 3. If you don't get 2 results, check whether one of those files is empty too :-/ <nckx>I have to go again, for now. <apteryx>rekado: or just use a spoof email address different from the list email <apteryx>(when sending emails from the frontend of mumi) <apteryx>it's not that I don't receive any emails when the list email itself is the Return-Path, it's just that they get filtered as all the others to a guix-patches folders which contains thousands of emails. When someone wide reply to me, the Return-Path is instead their own email, and it reaches my INBOX instead, which makes it much more likely that I'll see it and reply promptly <lechner>Hi, is there another accepted method for patch submissions aside from git send-email please? <apteryx>lechner: you could try patman, which automates git send-email under the hood <apteryx>you can also attach patches produced with 'git format-patch', but that's a bit more work to apply and review <apteryx>so git send-email or patman are preferred <lechner>i have been dinged for attachments before <lechner>it seems odd that one is not allowed to bring their own email client <PotentialUser-10>nckx I am going to write about this error to devs. because I think it is a bug rather then a error. <rekado>lechner: you can use your own email client, of course <nckx>lechner: Why is one not? <nckx>PotentialUser-10: That's what I feared, if not expected. <lechner>please send your ways connecting git send-email to gnus <nckx>You can ‘rm ~/.guix-profile’, run ‘guix gc -d’, then try to ‘guix install’ something again. Then check those 2 empty files. <mirai>lechner: I resigned to using git send-email <nckx>You do realise that people moaning about your MUA don't speak for the project, only for themselves. <nckx>If there's someone who will only apply your patches if you use Yandex Mail, you can use Yandex Mail to appease them, or you can ignore that person. <mirai>I think it's simpler to use it for sending patches rather than any kind of contraption you spin with $YOURMUA <nckx>Our lack of reviewers makes the latter not as easy as it should be, but it's still an option. <efraim>from mutt, attachments or git-send-email are equally easy to apply, but git-send-email is easier to make inline comments in the patch <apteryx>jlicht: hm, I'm getting 'WARNING: Unknown setting get_maintainer_script' with patman after updating my profile <nckx>PotentialUser-10: Hm. OK, let's be a bit more meticulous: <nckx>…or let's quit, we can do that. <apteryx>hm, the patch is applied: u-boot-patman-get-maintainer.patch <mirai>you can always reply with your conventional MUA <apteryx>ah, seems I dropped a hunk, only the test gets included... <lechner>nckx / sorry, i think i misunderstood your MUA comment earlier, but i am done whining for now <apteryx>rekado: if you could ack and have a last look at #60224 v3, that'd be great! Perhaps I could merge it along a fix to patman patches, as this will rebuild all u-boot packages <nckx>Uhm. ‘guix pull’ → ‘fatal: unknown index entry format 0x64720000’ <nckx>Did the curse transfer to me. <bjc>well i was just about to pull, but i think i'll wait a bit~ <lechner>wow, i had a day like that yesterday, although it did not involve guix pull (or Guix) <nckx>False alarm (for you, not for me), it was not from Guix. <nckx>My linux repository is dommage'd. <nckx>apteryx: Took me a sec to grok, but no. <jlicht>apteryx: good to know it's not just me :P <apteryx>jlicht: thanks for reporting it! I'm reintegrating the patch, and if it tests I'll send a fix <ekaitz>any i3 user here? did you ever find Can't locate AnyEvent/I3.pm in @INC (you may need to install the AnyEvent::I3 module) (@INC contains: /gnu/store/hy6abswwv4d89zp464fw52z65fkzr7h5-perl-5.34.0/lib/perl5/site_perl/5.34.0/x86_64-linux-thread-multi /gnu/store/hy6abswwv4d89zp464fw52z65fkzr7h5-perl-5.34.0/lib/perl5/site_perl/5.34.0 <ekaitz>/gnu/store/hy6abswwv4d89zp464fw52z65fkzr7h5-perl-5.34.0/lib/perl5/5.34.0/x86_64-linux-thread-multi /gnu/store/hy6abswwv4d89zp464fw52z65fkzr7h5-perl-5.34.0/lib/perl5/5.34.0) at /run/current-system/profile/bin/i3-save-tree line 19.?? <ekaitz>oh it looks like there's a thread in the ML <acrow>lechner: Does the bot report new bug reports here? New=novel,recent,just-posted. <nckx>telmo[m]: Your best starting bet is fonts-google-noto. <nckx>It should have an N'Ko font, but I don't know if our package does. <mirai>is "guile: warning: failed to install locale" supposed to occur during a "guix package build" ? <lechner>acrow / it does not, but i may be able to modify MUMI to alert the channel of new filings <lechner>mirai / i have seen the locale error on my own software when i used make-locale carelessly <telmo[m]>nckx: but in gnome settings not is available <telmo[m]>the reason for question its why i want create my first semitic conlang with this script <acrow>lechner: Can issues be accessed using the new concise identifiers yet? <lechner>acrow / i read myself into MUMI but am reluctant to propose such a patch until rekado develops positive feelings for my goofy identifiers <lechner>i have been suggesting that people direct negative feelings toward the bugs instead <apteryx>jlicht: fix for patman sent for review! <apteryx>ACTION patches their profile with 'guix install /gnu/store/6jm39v0f0i8nn2s8q1jbj0ybf7fdfvh1-patman-2022.10' <podiki[m]1>rekado: re: debbugs replying, yes that's my understanding, the bug number address goes to...? just the mailing list I guess? <podiki[m]1>jman: thanks I see it now, will write a short note to go with my patch and send it today <apteryx>I'm out of ideas for #59700 (best location of %facebook-host-aliases in our doc); should it go to the cookbook instead? <apteryx>OK; would you like to reword your patch this way? Then I guess if it's nicely organized in the cookbook we could merge it <mirai>tbh, the entry feels completely arbitrary given that there's more bad apples than them but it "works well" as an example of hosts-file usage <apteryx>right. it's more of an example than anything else <apteryx>guix-patches is slow again, it seems <mirai>better yet would actually be turning hosts-file into a service as if you were to use that snippet, it overwrites your "localhost" entries <civodul>re #59700, i'm in favor of the status quo, even if strictly speaking it's not a service <acrow>lechner: Your codes are more succinct but knowing one does not give you a sense of how recently the issue occurred or allow you to compare the recency of issues without doing a lookup. <civodul>but yeah, we should turn 'hosts-file' into a service <bjc>what's the right way to go about complaining about led-lightbulb? <mirai>bjc: knock at its maintainer <bjc>ok: it's overy verbose, has ugly horns, the identifiers are less memorable (to me) and only act as a pointer to the real deal, it uses a url shortener, and the site it points to requires javascript and a login <bjc>there's a place for #number -> url, but led-lightbulb ain't it (for me) <apteryx>to explain the horned short bug identifier <bjc>i don't think the short ids work as well as numbers, personally. they look like gibberish and have the same memorability <apteryx>I'm not convinced it's useful *yet*, given the URL includes the full identifier and we're not in the millions yet <bjc>there's better ways to shorten when we get there than the current encoding <bjc>and, like, i grew up with phone numbers. relatively long strings aren't that hard to remember, especially compared to arbitrary letters <mirai>imo the thing with arbitrary letters is that they share the same flaw of passwords <mirai>they're not that human memorable <bjc>and this encoding is *way* worse than that. capitalization matters, it has other non-letter symbols. it seems like something that made sense on paper, but wasn't actually vetted with real people <nckx>These )( are in no way shorter or easier to remember, they just pack the bits. Humans are pretty good at doing that internally already. <nckx>I can remember phone numbers. I cannot remember base60 strings of the same length. I'm not special. <unmatched-paren>i think something like ``[#NUMBER] TITLE <URL>'' would be nicer than just the number and url <apteryx>agreed; don't really like to see real names plastered all over either, it's not useful information <acrow>Maybe the bot could return a tiny_url reference to minimize bandwidth and screen real-estate? <bjc>url shorteners shouldn't be used. they break the web and track you at the same time <acrow>lechner's desire for conciseness is good. <nckx>Oh, I hadn't even noticed those. <nckx>Shows where I'd stop reading :) <nckx>bjc: Third-party ones add another party to the tracking, but nothing more. And we could use g.gnu.org/N or whatever (note: I don't think we should, just pointing it out). <nckx>unmatched-paren: Mmm, shouty! <bjc>yeah, one hosted by gnu would at least be trustable not to track you <nckx>No, that part I can't agree with. <nckx>I'm just saying the tracking would be limited to 1 party (us/GNU) instead of 2. <bjc>gnu's already got you, though, since they host the issues website <apteryx>we already have 2 URLs for bugs, let's not add a 3rd one <bjc>stop agreeing with me! <acrow>Both the full description and link should not be necessary bc the description ought be searchable from issues... <bjc>the description adds context for people on the sidelines, and the link is a convenience <nckx>So just a bot that echoes the number back at you. <apteryx>I liked the templated proposed by unmatched-paren <nckx>That is the most logical. <nckx>If you leave out the description, better leave out the bot. <nckx>acrow: On the contrary, there seems to be a rough consensus building. <apteryx>oh yes, I like dropping the leading #ID, since in the URL anyway <acrow>Actually I think the description is what you want. <nckx>acrow: Then I misread ‘Both the full description […] should not be necessary’ (but how?) and there's even more consensus :) <bjc>i really like the url <unmatched-paren>so it would make it easier for them to jump to the issue page if the URL is included <acrow>Well what are we trying to achieve? I think it depends on that. <bjc>the title gives you a general idea, the url gives you a quick way to dive in if you're interested <apteryx>I thought the MUMI bug URL is helpful to most people <nckx>IMO: (1) save the waste of time of everyone (including lurkers) looking up each posted bug# by hand to get context (2) save even that click in the common case where you just want to know ‘oh it's the bug about x’ not the entire discussion/patch. <nckx>And by looking up, I mean having to copy/remember the number, and open it in your favourite issue viewing solution. <acrow>If someone types the issue number here, well, you've got it. A bot might help make that otherwise obtuse number more understandable with the issue title. <acrow>I also think the IRC channel, as the center of what's happening for guix, would be better if a bot announced incoming, new issues for general interest. <mirai>but that's the purpose of mailing lists no? <mirai>it'd just increase the snr here <acrow>don't mailing-lists provide summaries as regular intervals? <unmatched-paren>apteryx: i mean something like ``<commitbot> New commit "..." and 21 more'' <acrow>Yes, but I'd say more signal and less noise. <unmatched-paren>the last commit is generally the most relevant, so it could display only that <mirai>you could have a guix-commits and a guix-issues channel though I'm pretty sure once the novelty wears off it will be a dead channel <mirai>and new commit messages on IRC basically boils down to: oh, *shiny* thing just got merged <nckx>If you want to use #guix-commits I won't stop you but this is a channel for humans. <acrow>IDK the commit rate here and it varies but typically isn't it around 50 commits per day? Would that be terribly bad? <apteryx>it would, when rekado merges hundreds of R updates <acrow>I mean adding fifty extra messages to the channel each day. Would that be even noticable? <unmatched-paren>apteryx: but that would be contracted to "New commit "gnu: r-blah: Update to blahblah" and 300 more" <mirai>acrow: the thing is, why add a single extra for almost zero gain? <mirai>I don't see this adding any value tbh <acrow>So people can help each other. <nckx>Why announce a random fraction of commits if people are supposedly interested in these commits in the first place? ‘Ooh, random_package + 300 more’. <nckx>guix-commits was mentioned above. <mirai>acrow: that's why everyone's here :) <acrow>It helps the community stay informed for opportunities to help. <mirai>eh, don't get me wrong but that's the job of mumi / issue trackers <unmatched-paren>nckx: my thinking was that when, say, aerc was merged, it would have sent ``New commit "gnu: Add aerc." and $N more'' <mirai>sifting through forgotten issues or reviewing patches is an excellent way to help <nckx>unmatched-paren: That's a legit but very selective example. Most commits aren't like that. <acrow>Then people interested people would have immediately done 'guix pull's and gotten that hot update much sooner. <mirai>acrow: I'd expect such people to be already using unattended upgrade service <acrow>It makes hanging out in the channel better. <unmatched-paren>nckx: really? aren't a lot of groups of commits basically "add some dependencies then add a standalone package requiring those ones" <nckx>The big data's there for the finding out :) but I don't think so, no. <apteryx>jlicht: seems my email with the patman fix hasn't been sent? <nckx>I think those tidy stories that just happen to fit that format are a minority. <nckx>But that's just my subjective experience! <unmatched-paren>oh well :) guix-commits exists, so i agree that an irc bot wouldn't add much value if that was the case <acrow>If something is broken people will be looking for any signs of progress and won't be able to wait for that periodic update. <mirai>unmatched-paren: it's not guaranteed that the interesting commits are always something "standalone" <mirai>for some, a commit adding/modifying a "library" (or dependency) might be the thing of interest itself <mirai>acrow: that's called bug-guix@gnu.org + mumi <nckx>acrow: Why not subscribe to guix-commits then? <mirai>btw, spamming commits on IRC is a bad fit as you'd probably want threaded replies as well <mirai>otherwise I don't see it feasible to discuss about some particular commit which might have caused X <acrow>You've got me there. I don't know about guix-commits; but, and I love things the way they are, yet I don't think these things need be mutually exclusive. <nckx>It wasn't meant as a gotcha, sorry. <acrow>In general I think we want this channel to encompass the widest possible set of use cases so we have the largest possible audience, no? <mirai>acrow: do a "dry-run" of this idea <mirai>grab some logs and graft commit messages to it as a PoC <acrow>It seems to me that lechner has already demonstrated good work in this area and if we haven't too badly damaged his pride maybe he would accept the positive feedback? <bjc>i'd like this channel to be focused on human to human communication. bots should be there to add context, or maybe a bit of fun, but that's about it, imho <mirai>or if you're using hexchat/any extensible irc client, try doing this client-side <nckx><I don't think these things need be mutually exclusive.> I think that's where we diverge. It's not 100% zero-sum, but I think it's a fact that bots announcing issues/commits have an effect on humans in a channel. And I think the net effect would likely be negative. For me, at least, it certainly would be. <mirai>do a IRC client plugin that does this <nckx>ACTION thought a bit too much in that last message. <mirai>and maybe you could also get the bot to interpret commit ids <nckx>mirai: Are there clients that make it easy to ‘merge’ different channels in one buffer? You could merge #guix, #guix-offtopic, #guix-commits… they'd become… oh no… actual hashtags. <mirai>if you see something of interest and write about it, the bot will show a link to the commit (in cgit) <nckx>(The chat would be incomprehensible to those without cross-subscription, so don't; it's just a question.) <mirai>that's infinitely better than indiscriminate commit flooding <ham5urg>Is it correct to say that a system service modules job is to install a package and configure the sheperd accordingly? <mirai>nckx: not that I am aware of, I'm using hexchat <mirai>though it doesn't seem too hard to write a plugin that would do channel merging <nckx>System services installing packages is controversial. Some people like it, some people don't. I'd certainly not say it's their job. <bjc>there used to be clients that did that, but i haven't used one since the 90s. probably irssi <nckx>mirai: Hexchat bestchat. <acrow>lechner: Is there a minimal subset of what was proposed here that you would care to implement for awhile to see how it goes? <mirai>nckx: the proxy handling and SASL TLScert key management is a bit clunky and could be better <mirai>but overall it's a good client <nckx>I use ZNC for the SASL and don't use proxies, so yeah. <mirai>I've been thinking about using ZNC to iron out some of the crinkles of SASL keys and having some sort of unified session <nckx>ham5urg: No, I don't think that's accurate, but I don't have a pithy substitution. Many Guix system services are akin to ‘services’ on other distributions, e.g. ‘systemd services’. They both configure a daemon (foo-configuration) and the Shepherd to run it (foo-shepherd-service). But neither of those are mandatory. Many system services aren't Shepherd services and don't deal with ‘packages’ at all. <mirai>though it's been sitting in the back burner since forever as there was (is?) no ZNC guix service <ham5urg>But maybe this is a special case as I understand you <mirai>the services don't really install anything <nckx>Not special, such services are probably the majority. But Guix services can do basically anything; choosing to be a traditional ‘run foobard’ one is just an arbitrary choice, not what Guix system services ‘are’. <nckx>mirai: Some do, unfortunately. <mirai>when you're using some service module and you do a guix system reconfigure config.scm, what happens is that the required package is fetched into the store <ham5urg>mirai, I did not installed ssh but it works as in config.scm <nckx>I can't think of an example but some have been explicitly modified in the past to install the package that contains their daemon to the system profile. <bjc>i thought installing stuff (when necessary) was a benefit of the guix model? ie, anything the closure references is automatically brought in <nckx>ham5urg: Which service installs ssh? <mirai>nckx: oh ok, I wasn't aware that there are services that actually tinker with your profile <nckx>Are you using ‘install’ to mean ‘is present in /gnu/store’? <mirai>to be fair I almost wanted to type "install" a few lines above <efraim>'Sustainable Use License' isn't free, right? <nckx>efraim: Doesn't sound like it. <ham5urg>(service openssh-service-type ... will install ssh as I understand <efraim>'You may use or modify the software only for your own internal business purposes or for non-commercial or personal use' <nckx>Yeah the whole ‘Limitations’ section is unacceptable. <nckx>Except maybe the last bit. <nckx>Which I won't rehash today 😛 <efraim>looks like if I ever do upstream devilutionx it'll be stuck at 1.4.1 <unmatched-paren>Looks like yet another attempt to prevent commercial parasitism while completely forgetting (A)GPL exists :) <bjc>the agpl has some things it can't cover, unfortunately <nckx>ham5urg: As you understand, yes: it defines a Shepherd service which refers to sshd, so Guix will automatically deploy (either by building or by substition) the corresponding openssh package in /gnu/store, along with the service. <bjc>firmware, for instance, would be nice to protect, but i don't know of a gpl variant that does so <nckx>But that shouldn't make ‘# ssh’ work. <mirai>rather futile as anyone with deep enough pockets can just reimplement your program and put it under a different license <nckx>Which is what ‘install’ means in Guix. <bjc>nckx: i don't install openssh seperately in my config, yet it exists in /run/current-system/profile/bin/ssh <nckx>(Nor should ‘# sshd’ or ‘$ sshd’ work, in case that wasn't clear.) <mirai>the AGPL clause strength is that it also demands that the source be available, even when the program isn't executed on your premises <nckx>bjc: I can't easily grep guix (or use ‘guix system edit’ here). <bjc>that only applies to things which are accessed via some kind of client, though, right? <nckx>That would mean o-s-t extends profile-s-t. Which is possible, but it is *wholly separate* from running a ssh service. It's a ‘oh and also do this’ move. Maybe for a reason; I can't easily check. <nckx>You don't need to install ssh(d) to run sshd, is what I should have soid. <bjc>nckx: ah yes, i didn't mean to imply that's how services work in general, just that it's how openssh-service-type, specifically, works <nckx>And I should've chosen a better/different example. :) <nckx>I *think* cups is an example of the style I prefer: runs cups without littering my profile with it. <mirai>openssh-service-type has a profile-service-type service-extension <mirai>that's what's causing ssh, scp and the likes to show up under your profile <bjc>i don't know why openssh is like this; it doesn't need to be. guix wouldn't be the first distro that allowed you to run sshd without providing ssh <mirai>;; Install OpenSSH in the system profile. That way, <mirai> ;; 'scp' is found when someone tries to copy to or from <nckx>bjc: Did you check the code? <nckx>bjc: Strictly speaking, yes, one could separate openssh into output or otherwise ensure that the service-type provides *only* scp, not the rest. This just wasn't done for simplicity. <bjc>yeah, i get the simplicity argument. otoh, it introduces ambiguity about how services in general work, which isn't ideal <bjc>personally, i wouldn't mind if scp *didn't* work at all unless it was explicitly installed. i don't find that comment convincing <ham5urg>Is it common to first write a sheperd service file and if used often / adapted often to various machines, to write a service module to automate the deployment more? <nckx>I don't find it convincing as much as I totally understand the person who added it after the 3rd ‘why scp no work it works on Dabion’ ‘bug’ report. <bjc>same, and since i'm not handling the bug reports, i probably shouldn't have too much say in the right way to do it ;) <nckx>Your point is valid and you should feel valid. <mirai>pulseaudio service does the opposite <mirai>you actually need to add it to packages <mirai>but if the scp thing bothers you too much I guess you could "subset" that service-type <bjc>was that pulseaudio config bug fixed? where it didn't source .default.pa.d? <mirai>inherit openssh-service-type <bjc>it doesn't bother me that much. life is full of ambiguity <mirai>and remove/filter out the profile service-extension <mirai>do note that you will need to reboot <bjc>it's only an issue on reboot, so that's fine <mirai>somehow reconfiguring isn't enough for it to pick up <bjc>that seems like a bug? i dunno. the herd auto-restart stuff is still magic to me <mirai>I suspect it could be pulseaudio needing to be killed <mirai>but pulseaudio is not handled by shepherd <mirai>so you can't really restart it <bjc>that would make some sense, but i'm using sway, which doesn't do most things automatically <nckx>PA is started by clients, I think through D-Bus. <nckx>I don't start PA anywhere and yet it works fine on Sway, but I do start D-Bus. <nckx>(I guess some login managers might start it to play a log-in sound :) <GNUtoo>hi, I've a quick question: Does Guix require free licenses for non-functional works (like game data) or not? The FSDG allows licenses like cc-by-nd for that (but not the -nc ones). <GNUtoo>but it's unclear if the question was considered or not <nckx>We don't deliberately stray from the FSDG, in either direction, when it comes to licencing. <nckx>(We might reject packages for other reasons out of its scope.) <GNUtoo>ok, so I assume that cc-by-nd is ok for non-funcional data then, thanks a lot <nckx>ACTION has ‘guix’ grouped if anyone ever wants to use it for a channel bot, BTW. <unmatched-paren>i personally prefer ``$TITLE <$URL>'' over ``"$TITLE" $URL'' but this is *way* better than led-lightbulb :) at first glance i confess i thought led-lightbulb was a spammer :) <nckx>unmatched-paren: The ‘Huh?’ made it a very bumpy read. <nckx>I like how this bot's name documents its purpose! 👍 The lightbulb got questions, too. <elb>what is the best way to find out what scheme module a particular package is defined in? I've never found it in `guix show` etc., and wind up grepping through scheme, which I think is probably not the plan <mekeor[m]>ACTION learned that issues.guix.gnu.org offers a graphql api <nckx>elb: Is guix show's ‘location:’ not the right answer? <nckx>Or do you mean the package itself is unlisted? <nckx>(If so, grep is your unfortunate friend.) <elb>nckx: oh ... yeah, actualy, I think it is what I want, I just didn't recognize it <elb>I just turn gnu/packages/whatever.scm into (gnu packages whatever) <nckx>I'm just glad the easy answer was the right one for once. <nckx>elb: Yes, always, it's guaranteed. <elb>I'm embarrassed I didn't see it ;-) <elb>I was looking for the (gnu ...) form, it didn't occur to me to look more closely at the fields <nckx>ACTION wonders what happened to the ‘my profile is empty… oh my manifest is also empty… bye’ person. <f1refly>what's the best way of setting the gtk theme and font options when I'm using sway? <nckx>Help two (probably) related sods for the price of one: how can I get icons in, oh, say, simple-scan, under Sway? I have hicolor-icon-theme and tried adwaita-, but all icons are grey pages. <nckx>Without hicolor-, it actually prints ‘Could not find the icon 'scanner-symbolic-ltr'. The 'hicolor' theme is not…’. When I add it, the message goes away (so it's found!) but the window looks identical. <nckx>I'm guessing I need gwhizbang-svg-gschema-ggg or something. <lechner>ACTION is immune to most visual stimuli, and can wait <nckx>ACTION has memorised the position of the buttons, but would prefer not to. <nckx>I'm a hotshot with things to do. <f1refly>weirdly, it starts in a submenu, but when i press the arrow back i'm able to adjust settings <bjc>i would love to know the answer to the button icon problem. pulseview has a similar issue (althoug some toolbar icons work? maybe they're built-in?) <nckx>f1refly: Did gnome-tweaks have icons (including CSD title bar buttons)? Doesn't here. <nckx>Oh, but it prints ‘Could not load a pixbuf’. A hint! <nckx>Adding gdk-pixbuf, again, makes the *message* go away, but no actual visual difference. <f1refly>I have another problem that bothered me for a long time - when I enter a directory with media files, thunar will fail to show thumbnails and all the temporary files tumblerd generates in /tmp are not deleted, but fill up /tmp and are all 0 bytes in size <f1refly>any idea how to approach that problem? <f1refly>nckx: you mean the x to close the window etc icons? <nckx>OK, adding gdk-pixbuf {adwaita,hicolor}-icon-theme does fix things. <nckx>Yep, and for simple-scan as well. <nckx>Tasty rubber snacks for all you rubber ducks! Thanks :) <nckx>f1refly: I guess you somehow have it already? Or did you add it once? <f1refly>i has to be included as a dependency somewhere <f1refly>oh yeah, i i have the xfce4-desktop-service-type, that probably included one of the themes <f1refly>the comment i have next to it says "thunar breaks without it" <nckx>I have such comments too. I just thank my past self for being realistic. <f1refly>my thunar is still broken in a way, but at least i can use it :) <nckx>I don't use Thunar, but does it or tumblerd not log anything to std{out,err} when failing to generate new thumbnails (i.e., empty cache)? <nckx>tumblerd wasn't the thing that was ‘recently’ broken because a version update switched the database schema, was it? I remember something breaking Nautilus but not what it was called. <f1refly>thunar forks and logs absolutely nothing unfortunately, i never understood how you're supposed to read its output <f1refly>i don't know if it broke for nautilus, but i think tumblerd never worked for me on this machine with guix <f1refly>i now remember why i have xfce-desktop! i had an issue where thunar would say gvfs wasn't available and wouldn't use the trash + refuse to mount removable drives <f1refly>$ dbus-monitor --session --monitor interface=org.freedesktop.thumbnails.Thumbnailer1 <f1refly>if anything want's to go and burn poetterings house down i'll pay for travel expenses <elb>The documentation for user management says that the root account is handled specially, but has no link to how it is actually handled. Is it possible to set a password hash for the root user in the (operating-system) declaration? <elb>f1refly: I feel like even as a joke, that crosses a line <f1refly>string "No thumbnailer available for "file:///home/xxx/Videos/xxx.mkv"" <nckx>Oh, I missed that, sorry. Yes let's please keep this a pleasant place. <nckx>Not that the inverse would be an excuse, but D-Bus is not even his. <lechner>ACTION thinks some contrition would be in order <f1refly>i'm a bit too angry at computers for contrition right now <f1refly>it'll be time for that later when im in bed <lechner>i would be for the sake of the group, not yours <nckx>Being angry at computers is a natural state, and a logical reaction to their existence. But please don't disparage others here, including other distributions or projects. <nckx>I've done so myself, here, in equal frustration (Flatpak grr) so it's not that I don't sympathise or consider myself immune. <f1refly>it was merely hyperbole to vent my frustration <bjc>i think what's trying to be communicated is that we're trying to foster a community of generosity and respect. and while the comment is clearly not meant seriously, it does contribute, however little, to the othering of lennart, and that's not conducive to the values we're trying to uphold <bjc>and it's something we have definitely all done, and it's also something we should try not to, simply because all people should be treated with as much kindness as we can muster <nckx>I should have added an explicit ‘…and I was wrong to do so’ above. <nckx>I delayed server recovery by literally 20 minutes so far (and counting) by absent-mindedly typing an unnecessary /gnu/store/<Tab>… <elevenkb>might've found a bug that I can fix even as a programming n00b. <vivien>Hello guix! Does your webcam work? <nckx>(Which is any remotely standard USB webcam, including any laptop I've ever seen ever.) <nckx>USB Video Class, I think. <nckx>I wasn't, but I didn't get into camming that early. <mvnx>I booted from the QEMU 1.4.0 qcow2 image then used `guix system reconfigure bare-bones.scm` where `bare-bones.scm` is the source example `bare-bones.tmpl` with bootloader and filesystems modified as a copy from `/run/current-system/configuration.scm`. It runs fine but next time I boot the qcow2 and pass GRUB it is just a black screen. Any idea why <f1refly>turns out installing ffmpegthumbnailer enables tumblerd to generate thumbnails for video <f1refly>I'm able to repent my rage at lennart now <nckx>Since he was blameless, it's the least you can do. <jackhill>I have this bad feeling that new laptops have cameras that have moved image processing into a blob on the main CPU <nckx>Anyway, that definitely qualifies for a ffmpegthumbnailer (phew) reference in tumblerd, f1refly! <nckx>(I can't tell jokes at this point of hardware-hell.) <f1refly>(although the thumbails still don't show for the mkv files in ~/Videos/*.mkv, but they work in ~/Videos/foo/*.mkv, but whatever) <f1refly>nckx: yes, i think ffmpegthumbailer should be at least mentioned <nckx>Agreed. Would you have the time to file a bug? I don't have time to write a patch right now, and don't want it to be forgot. <nckx>mvnx: Can you at least boot the previous generation at the boot menu? Or does the blackness envelop all? <nckx>ACTION managed to corrupt the qcow2 image just by accidentally booting it before it was complete, great start. Time to restart the download. <jackhill>I suspect that's why the cameras on my newer computers don't work. My 15-year-old USB camera still does though, but it just has the ungly bits locked away where I can't see them… <nckx>ACTION could make some very rude comparisons to vendors being obsessed with their ‘IP’ and thinking that anyone else must therefore even remotely care about it too. No. Everyone has one, and yours isn't even that great. <bjc>they're not hiding it to prevent theft, they're hiding it to save face~ <f1refly>nckx: I have written up the bug report, but I found this https://issues.guix.gnu.org/37276 bug and apparently it should fix video thumbnails as well? So maybe both should be listed in the package description as gstreamer didn't seem to be installed for me as well <bjc>wow. did lechner just update the bot? <bjc>if so: you're a real mensch <nckx>mvnx: I'm not sure we'll ever be able to guess at what's wrong, and since this is ‘just’ a VM, I'd simply start over with a fresh download. But this time, copy /run/current-system/configuration.scm, not bare-bones.tmpl. I'd first reconfigure with an unmodified copy, to make sure that that works. If that boots, make small incemental edits. <lechner>bjc / yeah, sorry the thing made you so upset <bjc>i wasn't *that* upset. and i'm very grateful you could take the conversation in stride. it takes a lot of character to do that <nckx>f1refly: By ‘description’, do you mean literally adding a paragraph to the (prose) description explaining to users that they should install these two packages if they want x support? <nckx>I fear that might be way to obscure. You knew what tumblerd was and that it was suspect. I'm not sure many will. <nckx>There are cases — like yt-dlp — where I think it makes sense to keep inputs like ffmpeg optional: yt-dlp clearly prints ‘yo you might want ffmpeg’ at run time. Your debugging experience was… very much not that (which made it so impressive). I think it makes sense to make it unconditional if by default the previewer can't even preview ubiquitous things like MKV. <nckx>But maybe you meant ‘description’ as in the code, and adding both as inputs (and references!) and then I certainly agree. <f1refly>nckx: no, i meant in the description of the package, but you're right that it's probably too obscure. I was only aware that tumblerd might be the problem because years ago it had an issue where it would try to thumbnail for like 15 minutes and hog a cpu core at 100% during that time <f1refly>the thing is that I'm not sure now if gstreamer should be installed by default as a dependency to tumblerd? maybe my installation is broken in some other way so gstreamer doesn't work properly <f1refly>(or is not considered a backend for video files, as tumblerd clearly states that there are no thumbnailers available for the bigger video files) <f1refly>i was mistaken, gstreamer is still commented out with a FIXME in the tumbler package definition <f1refly>I don't understand why it was done, the corresponding note is "gnu: tumbler: Remove unused gstreamer.", but it clearly wasn't unused because now videos don't have thumbails anymore? <nckx>f1refly: I did not test build, but it's possible that gstreamer wasn't actually used during the build, and/or that no reference to it was kept. If so, #37276 merely codified the status quo, but that doesn't mean the status quo was correct. Inputs say *only* which packages should be present when building—not which packages are needed at run time! <nckx>Test #37276… #37276… anyone got #37276… #37276 going once… <nckx>lechner already regretting being a botfather. <nckx>Do let us know when it moves from fun learning motivator to chore. <nckx>If you don't respond immediately it'll start PM'ing you, then your friends, then start posting embarrasing ‘facts’ about you in the channel? <nckx>IIRC NixOS had something like that. <nckx>(No, just spammy e-mails that occassionally malfunctioned.) <f1refly>nckx: I don't think it has to be built against gstreamer to support it, but I might be wrong <f1refly>It can use ffmpegthumbnailer as well, despite not being built against it <nckx>Then why is it commented under inputs at all? <nckx>I don't mean ‘build against it’ during the compilation process, or linked as a library, but you'll need to provide a reference to it (and hence have it as an input) if you want tumbler to ‘pull it in’. <nckx>A reference means that wherever it tries to [say] ‘execvp("ffmpegthumbnailer")’ you patch it to exec /gnu/store/…/bin/ffmpegthumbnailer instead. That way, Guix knows that tumbler needs ffmpegthumbnailer. <FlachyJoe>Hi ! Why many package install *-doc and *-debug as dependencies? IMHO they should only be installed at request. <nckx>I think grafts can lead to that happening. Grafts are a hack to allow for timely security updates but have some spooky (if generally harmless) side-effects. <nckx>FlachyJoe: What's an example of such a ‘dependency’? <nckx>ACTION still hates that word. <FlachyJoe>Try to install ungoogled-chromium, it downloads "gtk 3.24.30-bin" + "-doc" + "-debug" <jpoiret>FlachyJoe: it shouldn't download them, however if you need to locally build it then it'll build all of the outputs <jpoiret>on my side `guix size ungoogled-chromium` doesn't list those outputs <jpoiret>ah but grafts can mess this up, i missed that answer as well :p <nckx>FlachyJoe: However, I can't reproduce this, with or without --no-grafts. <FlachyJoe>jpoiret: I'm a newbee with GUIX but 'find /gnu/store/ -name "*-doc" ' lists many items <nckx>-bin is required to build the profile (not ungoogled-chromium itself). However, I built a profile with u-g (‘guix shell ungoogled-chromium’) without a single /gnu/store/*-gtk+-*-{doc,debug}/ at all. <nckx>There are substitutes from bordeaux, though. <jpoiret>when you build a package with multiple outputs, you'll get all of its outputs at once, but if you substitute it you only get the required ones. The reason for the former is that there's only one build recipe that builds all of the things, they're just sorted in separate outputs <FlachyJoe>But why are they bulit as I don't ask them ? <jpoiret>we don't have a recipe for "just -bin or just -debug", everything is built once. So if you can substitute them from somewhere else, you can fetch the part you're interested in, but if you build it then you'll need to build all of the outputs <jpoiret>they aren't separate packages but rather different outputs of a single package <FlachyJoe>Ok got it, so it needs special recipe to have only what I need. <nckx>Yep. You'll have to rebuild pretty much your entire world for that to work, but it's possible. <nckx>No more substitute downloads for you for anything that (transitively) uses gtk. <FlachyJoe>It seems Guix is not reasonable for a reduced ressource system. <nckx>It's certainly not the focus. <SUPERB[m]>SUPERB[m]: An icon will be removed as I install a new package or upgrade <old>anybody has: build of /gnu/store/r8088hz3i9gz5543ajdk2qv5ngblawk3-clang-toolchain-15.0.5.drv failed ? <civodul>podiki[m]1: the FHS post looks perfect! i'll push it tomorrow noon so it can brighten people's Friday :-) <podiki[m]1>civodul: great, thanks! (I had the date from when I last edited, feel free to adjust as needed) <podiki[m]1>thanks to the mailing list for the Tor example, I quite like that one, it is a nice use case for a container too <podiki[m]1>speaking of FHS, about to send a patch for #60556 that you saw mentioned in the website patch thread <mirai>is it me or did mumi just munched a patch at #60582 <nckx>mirai: Give your mail some (more) time to arrive. <mirai>nckx: It ate the mid 1/2 patch though <mirai>its showing now but disordered <nckx>When I checked it was still merely missing. <ConvolutedSquare>Interesting, can't shutdown or reboot my system from the commandline. Shutdown says "Service root is not running." Wonder what I did last night c: <mirai>you have a service that's either invalid scheme code or crashed shepherd <mirai>perhaps you have a procedure that uses exec-command <ConvolutedSquare>I was messing around with a custom service last night that errored out.. but I think it's all good after rebooting <ConvolutedSquare>Oh I'm getting the following exception: "Wrong type to apply: 603" Any idea what that means? <nckx>Something that evaluates to 603 is being applied—called—as a (procedure). <ConvolutedSquare>Interesting... not sure if it's even worth debugging because my little hack appears to work :) <nckx>The simplest form would be typing ‘(603)’ but I assume it's more insidious :) <nckx>Well, it would be computer, but it's not a number that looks familiar. <nckx>I'll pass if it's all right. I have work of my own. <dthompson>does anyone know if there's a way to defer mounting file systems until *after* the user-homes service has run aside from writing a special service for it? <ConvolutedSquare>dthompson: I wrote a service that does it. Afaik no built in way to do it. <nckx>‘Monitor guix.tobias.gr is back up! It was down for 363 hours and 20 minutes’ OK thanks shut up. <dthompson>ConvolutedSquare: that's what I thought, thanks! <nckx>You two just… found each other? Wow. <lechner>Hi, the output of my guix system list-generations is incorrect. How may I sidestep any type of caching, please? <nckx>I'll have it for dessert later but thanks! <ConvolutedSquare>I just hope it doesn't get lost in the void so that it may help some poor soul in the future <nckx>Well you've set it to expire tomorrow. Not that you need to repaste it. <mirai>civodul: well, the issue is relatively minor <mirai>I think it's fair to think that for anyone thinking on working with guix long term-ish will have a manifest of their own describing what packages they want to add <mirai>ConvolutedSquare: it's #~(make-forkexec-constructor <mirai>you have an extra layer of parentheses <mirai>so you're doing a procedure call on (make-forkexec-constructor) <SUPERB[m]>> <@superbs:matrix.org> An icon will be removed as I install a new package or upgrade <apteryx>I've seen this occur on a old Debian 9 when upgrading jami <apteryx>I assumed it was because the shortcut I had added was pegged to the exact store location of the program, and it disappeared, hence GNOME removed it <SUPERB[m]>apteryx: It’s on xfce and haven’t set any shortcut <SUPERB[m]>SUPERB[m]: An icon will be removed as I install a new package or upgrade
What’s going on? <mvnx>nckx: Sure, I guess that's best. I appreciate it. It wasn't so clear how many steps I was skipping until you phrased it like that :) <lechner>Hi, is the "network-manager-service-type" superior to other network configuration options, for example in that the 'networking prerequisite in the Shepherd actually ensures that networking is up? <civodul>the service remains up even if networking goes down, etc. <mirai>lechner: it doesn't even ensure that the network is ready lol <lechner>i am not sure that's where my issue is <civodul>yes, with the current shepherd we could have a service that auto starts/stops as networking comes and goes <lechner>nckx / Hi, again why please does opensmtpd not always start? <mirai>networking target is way too early <mirai>it has to wait for networkmanager to complete using nm-online <lechner>well, right now it works with nm but fails with connman <lechner>also, guix system roll-back can fail when using guix deploy and Guix on the target is even mildly out of date. i think it was something with record-types, but i am not sure 
