IRC channel logs
2023-02-20.log
back to list of logs
<apteryx>I saw a strange situation today where 'guix upgrade' said 'nothing to do' or similar, while it clearly should have updated packages to newer versions; have you ever seen this? <gnucode>apteryx: I didn't know we had a 'guix upgrade' command... <thanos_apollon><gnucode> "apteryx: I didn't know we had a..." <- What were you using instead of ‘guix upgrade’ <quidnunc>jlicht: Okay, let me try to rephrase: The packages I install via guix are not showing up in "bin" either in my profile directory. How do I fix that? <gnucode>thanos_apollon: (nice irc nick. ) And I was using guix pull && sudo guix system reconfigure.scm <apteryx>thanos_apollon: before 'guix upgrade', there was 'guix package --upgrade' :-) <quidnunc>Why does compiling from source require sqlite3 >= 3.6.19 which doesn't seem to exist <apteryx>does a locale specifies whether the clock should be as 24 h vs AM/PM ? <apteryx>reading 'info libc', I think so, yes <wdkrnls><ham5urg> If you contribute an installation image of Guix for the pinephone, I would be very grateful. <wdkrnls>I would love to have a Guix powered phone. <wdkrnls>And a Guix powered router as well... <iyzsong>router is do able, i once setup a x86 router with Guix. <apteryx>uh: guix pack: package 'open-adventure' has been superseded by 'open-adventure' <gnucode>wdkrnls: In my humble opinion, I think that the pinephone is barely powerful enough to run guix. <gnucode>I personally would not want guix on my pinephone updating packages with postmarketOS is super fast. <wdkrnls>Sure, but that is what offloading is for. <wdkrnls>I can't keep Guix in my head and PostMarketOS. I just want to figure out one system. <wdkrnls>And I would rather do it all in scheme :) <gnucode>sneek: later tell wdkrnls that "I have never been able to figure out how to do offloading..." <apteryx>hm, I'm not seeing the new 'guix pack -f rpm' news; something's off? <surpador>Whoops, sorry about that, wrong window :) <Gooberpatrol66>there should be a centralized project to create a router config for guix <henzo_midala>Hi, is it possible to reuse /gnu/store among guix instances or distros? Let's say like a partition. <unmatched-paren>henzo_midala: pretty sure it's possible, as i believe the build servers do just that... <nckx>I can't be expected to *read*. Also, 'morning Guix. <nckx>Obviously that will only work if both 'instances' run at the same time. If you want to 'time-share' different Guix System installations on the same metal (interesting use case) you should be good with sharing a partition--the only gotcha is not to forget to share /var/guix, or at least the db/, or bad things will happen. <vivien>Huh… which one of the instances is supposed to run the garbage collector? <nckx>(That applies to any shared store really.) <vivien>How does it know what parts are live for the other instance? <henzo_midala>In fact, i tried it days ago but was not aware of sharing /var/guix. What happened was I started having Guile errors saying "the source file... Is newer than the complied version" and guix command became extremely slow. <semi>zig is totally broken on my machine - trying to do virtually anything (e.g. zig build) results in "Illegal instruction". Is this a known issue/does zig work for anyone here? <rovanion>Why is guix-science a thing? Guix-science-nonfree is obvious but... <AwesomeAdam54321>rovanion: It's because there are very old science tools that aren't maintained or at risk of having security bugs <AwesomeAdam54321>If neither of those things are true for some packages in guix-science, then those packages should be part of upstream guix <rovanion>I can see that. But Jupyter notebook feels like a newish development? <PurpleSym>guix-science provides packages we cannot include in Guix proper, because we can't build them from source entirely. It has nothing to do with old tool or security. <PurpleSym>Jupyter is a JavaScript minefield for example. RStudio too. <PurpleSym>(But RStudio can also simply run in the browser.) <sneek>Welcome back civodul, you have 1 message! <sneek>civodul, jackhill says: I haven't forgotten about the GnuTLS stuff! I hope I'll get a chance to return to it Monday evening UTC-5 <Kabouik>ACTION was pretty sure the Guix days were yesterday and today after mixing up 2022 and 2023 on the Guix website, and was excited yesterday at watching one-year old videos like they were fresh and new. <Kabouik>I even marked those two days on my calendar to not miss the talks. <futurile1>Kabouik: I hope they were still good videos =-) <efraim>It seems semi is offline. I wonder which version of zig and if zig is built with architecture specific extensions by accident <efraim>sneek: later tell semi tell me more about `zig build` not working for you. which version? I'd like to try to debug it <Kabouik>I still have the issues I mentioned last week about resolving websites. It seems to be true for every web request I do, even a ping to google.com will take ~10sec before I get the first reply, then ping normally. Then if I interupt the ping and run it again, it's working immediately. I don't know if "resolving" or "dns" are the keywords here, but I'm open to any suggestions. <futurile1>Kabouik: I won't be much help as I use Guix on a foreign distro. But, things (you've problably tried already) to check are that it's definitly 'resolving' not a network problem generally. So if you ping or ssh to an ip address (local on network, and non-locally) it's fine? If it's name resolution, then you need to check Guix's equivalent of /etc/resolv.conf and the equivalent of systemd-resolvd (a <futurile1>Kabouik: not really solving the problem - but you could run a caching nameserver on your network maybe - as a workaround <apteryx>has anyone seen the rpm format news? I must have done something wrong, because it doesn't seem to show upon 'guix pull' <apteryx>ah, perhaps because the first time I pulled the commit it referenced was wrong, and then on a subsequent pull it would think it's already been processed and not show it again? <abrenon>does git fetch work with git-lfs repositories ? <abrenon>ok, so if I simply add git-lfs to the native-inputs, it will be work automagically ? does it perform some kind of git lfs install on its own ? (sorry for the vague question, I've just discovered git-lfs and am still unsure how it works and what could be required to leverage it) <efraim>sneek: later ask vagrant do you know if u-boot provides a UEFI interface for aarch64 and riscv64 by default? I'd like to redo some of our images <efraim>sneek: later ask vagrantc do you know if u-boot provides a UEFI interface for aarch64 and riscv64 by default? I'd like to redo some of our images <cc0a3hr>Hello I want to install this GNU/Linux-libre distro on my system. I currently using Artix Linux. I want to switch to fully free distro. So I wonder if I can. I use nvidia GPU and open-source drivers doesn't work. This can cause problems installing free distro. <abrenon>I've opened my laptop during a conference, someone saw the sticker on it and I've just been asked "do you really use 'guix' ?" : ) <abrenon>so it is probably safe to say the name is starting to get known outside its user base (although it still scares people a bit) <nckx>Sticker with or without the name? <nckx>I had that experience in the hotel lobby during the previous FOSDEM. <efraim>At debconf I got known as the guy who was good at debugging debian packaging issues <civodul>abrenon: what was the tone though? like, "do you *reaaallllly* use Guix?!" :-) <abrenon>civodul: I understood it as genuine curiosity : ) <efraim>I wonder if the Rust Team™ can make librsvg-bootstrap unnecessary <efraim>ok, I got rust-1.64 building as rust <mirai>normalize-extra-args **procedure <civodul>mirai: hi! i wonder if the default sanitizer could be made "good enough" that custom sanitizers are unnecessary <mirai>civodul: it's possible, I was also thinking whether it would be "easier" to simply implement this by "creating a new type" instead <mirai>right now, we only have the type predicate (as a procedure with name type?) <civodul>i think we might want to start by looking at the use case closely <civodul>perhaps we'll end up with what you propose <civodul>sounds like hand waving, sorry about that :-) <civodul>to me the question is more about how we can support field deprecation in define-configuration <mirai>the convo might be missing some context but the idea was that we'd like to have the field getters return only one kind of object, in this case we only want user-account records even though we want to pass either strings-or-records <civodul>can this be done with a custom define-configuration "type"? <civodul>oh maybe that's what you had in mind above? <mirai>greetd-configuration implements this behavior but it uses record-type* which meant that using sanitizers was possible <civodul>right (there are other examples of deprecated fields with record-type*) <mirai>civodul: sure, it was another avenue I thought <lechner>Kabouik / I think you have a DNS issue. Can you run a traceroute instead and see if the hop times are reasonable? We can then look into your name service issue <oleander>Hello, I'm getting this message after a guix pull && sudo guix system reconfigure /etc/config.scm: "shepherd: Evaluating user expression (and (defined? (quote transient?)) (map (# ?) ?)). I found this email thread where Ludovic wrote that the question marks are horizontal ellipsis not being displayed properly because 'guix' is not running under a UTF-8 locale https://issues.guix.gnu.org/55857. Can this also be related to my <mirai>I'm not entirely confident what/which approach should we take here, in fact I'm somewhat surprised this wip thing "kinda works" since I'm completely unfamiliar with how this #',... syntax sorcery works <mirai>this was done by "staring at the code hard enough" <lechner>oleander / for some reason, i think that message is normal or at least non-problematic. or maybe i told myself so because I saw it too <civodul>mirai: custom sanitizers in define-configuration sound like the most flexible approach longer term <civodul>i can help review the patch once you're confident <civodul>define-configuration is pretty hairy so i'll need to page that in :-) <apteryx>seems the build farm needs to be kicked again? <oleander>lechner: Hi, I figured that out, infact my question was more about the locale not configured properly according to what Ludovic wrote in that email thread. The message doesn't bother me much but I'm struggling with my terminal and browser since they are not displaying some characters properly, like the €. <mirai>civodul: so we're 'sanitizer 'serializer literals based approach then <oleander>I get question marks instead of the € symbol in my terminal too so I was wondering if the two things were somehow related. <civodul>side note: macro "literals" should be bound <mirai>though how should we handle the current status? right know the custom-serializer (which is being renamed into extra-args) is used for serializers <mirai>we want to preserve backward compat here <mirai>but we want to get people to start putting them behind (sanitizer ...) (serializer ...) <civodul>custom-serializer doesn't exist yet, does it? <mirai>it does, it's what's after doc <civodul>ACTION shouldn't engage in discussions after a week off <mirai>the original plan was to rename custom-serializer to extra-args, have extra-args bind to everything that comes after doc and see what do we get there <mirai>we'd want to deprecate having a bare custom-serializer at the bottom to have it inside a (serializer ...) "field" <mirai>the get-serializer procedure confused (sanitizer ...) at the bottom with a custom-serializer <mirai>about "> side note: macro "literals" should be bound", what do you mean by bound? <lechner>oleander / so i have the euro working in both terminal and browser (and i don't even live in europe). do you use X or Wayland? <jlicht>sneek: later tell cbaines: What would I need to do to get QA to start building one of my branches? Any kind of naming scheme, or something more involved? <apteryx>it'd be nice if guix worked out of the box on RH-derived systems (selinux) <apteryx>has someone ran 'guix pull' recently? did you see any news? <apteryx>(I'm asking because I seem to not see any myself...) <sneek>cbaines, you have 1 message! <sneek>cbaines, jlicht says: What would I need to do to get QA to start building one of my branches? Any kind of naming scheme, or something more involved? <cbaines>jlicht, which branch are you thinking about? <cbaines>for core-updates, I've been submitting some builds manually, which currently requires SSH access to bayfront (where the coordinator runs) <cbaines>I'd very much like to see this be easier for branches, but currently I'm pretty short on time to push this forward myself <jlicht>cbaines: It's the (freshly pushed) `node-18-updates' branch. <nckx>Something about some rando adding RPM support… <jlicht>since I added a patch-level version bump to the open patch series, but it's not being picked up by QA; I'd like to merge it if all lights are green, instead of letting it bitrot again :-) <gabber>is it possible to fast-track the integration of python-scapy (#60732)? i'm about 40% sure nothing is wrong with my patch and "just" needs a merge <mrvdb>How far off is 'guix system' to be available for powerpc64le? I'm running guix on top of an archpower kernel now and the mixed system is getting a bit hard to manage. <jlicht>apteryx: I saw the rpm news just now as well! <apteryx>great, thanks! I think it was that on my machine I pulled the bad news commit at first, and the fixed one didn't show up, because that news had already been consumed <gabber>mrvdb: AFAICT you should be able to build it? <mrvdb>gabber: could do indeed, time permitting. Was just wondering if an iso is just around the corner or still way off <gabber>depends on how you define "just around the corner"... you could just `guix system image --system=powerpc64le-linux your-config.scm` (or whichever architecture you were referring to before, not sure) <andreas-e>gabber: I will have a look at the patch later, it looks indeed straightforward, and as if there has been a problem with the QA service; it should not have taken such a long time. <andreas-e>Two quick comments: The commit message is not perfect, and we avoid marketing speech such as "powerful". (I will adapt it, no worries.) <andreas-e>(I just need to build chromium first, so it could take a while :-)) <gabber>andreas-e: thank you very much! yeah, sorry about the wording - it was straight up copied from their projects' side (: <jlicht>cbaines: asking the obvious, but the sorting is descending based on issue number then? <andreas-e>Sometimes that is the most difficult part :) I have packaged software of which I did not have the faintest idea what it did, but it was needed as an input for something I was interested in :) <cbaines>jlicht, yep, maybe it needs to be by the date when the latest series was received or something else. <andreas-e>Why do passing issues cause additional work? Are they not just set aside? <mrvdb>gabber: 'guix system image -t iso9660 gnu/system/install.scm' is what I was looking at, install that in a vm for a bit to test stuff before I image it onto the talos machine <cbaines>there's not that much additional work, but the branch will be deleted (and maybe recreated after 6 to 7 weeks) although I'd like to reduce that period to make the testing more representative <gabber>mrvdb: exactly, but make sure to pass --system after the `image` part, otherwise you will build (i think) the same architecture as you're already on <cbaines>I would increase the number of patches processed, but there's a backlog in processing revisions on data.qa.guix.gnu.org so it wouldn't have an immediate effect anyway <mrvdb>gabber: i am on powerpc64le so in my case, it's ok :-) <gabber>mrvdb: i will attempt the same thing on my riscv64 <mrvdb>gabber: stuck on failures building edk2-tools for me. <gabber>mine does also not look too good - libsegfault fails with cross compiling <mrvdb>not 'around the corner I guess' ;-) <gabber>and locally i had troubles with "guix substitute" dying unexpectedly <mekeor[m]>how can i find out which package offers a file called libstdc++.so.6? <andreas-e>Sounds like the C++ compiler; but I did not see it in gcc-toolchain. <mekeor[m]>searching packages by file-names is not possible, right? <andreas-e>No. But someone had started a website with their personal packages and files. lechner? <oenone>If any installed program depends on it, you can search in /gnu/store <andreas-e>They come from gnu/packages/gcc, but are not publicly visible. <andreas-e>They end up in /gnu/store/...-libstdc++ and are then probably used only by g++ internally. <ericson2314>and then time to convene Nix-Guix store standardization body :) <ericson2314>or maybe just the nix team is working well and everything is less controversial :) <apteryx>the build farm is way behind, it seems; shall we restart the cuirass service? <apteryx>it's not even completing evaluations, per ci.guix.org.gnu <futurile1>ACTION thinks 'ahoy' is a much better greeting than his 'evening all' <drakonis>ericson2314: have you learned scheme yet :V? <efraim>apteryx: however the packages start getting built again, yes, I'd suggest restarting cuirass (or something :)) <nckx>If nobody does it I'll do it when home. <nckx>At least CI is still evaluating core-updates. <nckx>I haven't built anything granular enough to notice the tiny increments and feel cheated. <nckx>But I know in my heart that they are there. <unmatched-paren>(actually, i'm downloading supertuxkart right now, and i can confirm that it does exist) <unmatched-paren>The evil bovine-people who control the world don't want you to know, that's all. <gabber>i'm looking for the bug that fails libsigsegv on riscv64 - the first one returns 139 in test stackoverflow1. i am looking at the sources right now - but i fail to understand where this 139 (or hex 8b) is supposed to come from. how can i invoke gdb for that specific build (i built the package with --system)? <apoorv569[m]>For a font package using a zip package where do I get the (sha256 (base32 hash from? sha256sum package.zip ? <nckx>You could derive the value from ‘sha256sum’ but I strongly recommend not bothering. <nckx>unmatched-paren: I just learnt about the ‘goat men’ mythos today, so this is surprisingly relevant to my lived experience. <apoorv569[m]>does guix hash does like first get sha256 then covert that to base32? <lebowski>hi guix! i have a problem. i'm trying to write a daemon in python and run it with herd. so... in my daemon i call some utils (such as awk and grep). i've added this utils to my system config as system packages and used (substitute* ) in my daemon package definition to call this utils from store. but when i try to run my daemon it cannot use these utils (for example - /gnu/store/l77gnq45qfig4kh4y8j259h0ndqxfgx5-xrandr-1.5.2/bin/xrandr <lebowski>cannot be used in my daemon, but it is installed and can be called from my (or root) profile. mb anyone know how to resolve this. <gabber>lebowski: you don't need to add the service's dependencies to your system :) <gabber>what is the exact error message? maybe paste.debian.net it? <nckx>apoorv569[m]: Yes (like the ‘sha256sum’ does with hex), but with a twist: the alphabet used it not the most widespread one, but a Nix-specific one. <nckx>lebowski: I don't really understand the problem, but them being installed or not has no effect on what you're trying to do. That's the whole point of patching. <nckx>If /gnu/store/l77gnq45qfig4kh4y8j259h0ndqxfgx5-xrandr-1.5.2/bin/xrandr but your daemon fails to invoke(?) it, it's likely that the exact error matters. Could be related to permissions, for example. <jackhill>hmmm, `guix weather ungoogled-chromium` reports that substitutes are availabe, but trying to upgrade a profile still results in a local build of building /gnu/store/wv1kr7vjlxjlaj5rvi2qfzjj787kn9a8-ungoogled-chromium-109.0.5414.119-1.drv... <lebowski>nckx: so there is not way to cal utils in python script? <nckx>lebowski: I don't understand. <jackhill>guix build ungoogled-chromium fetches substittues fine <lebowski>i mean, is there way to call from script (which packaged as herd service) any cli utils? <nckx>Yes. Patching like you're doing. <nckx>Hence why I say you should pay more attention to the error, not just ‘cannot use’. <apoorv569[m]>the softwares are mostly my own stuff custom build of DWM ST DMENU and all <zamfofex>jackhill: I remember running into a similar issue with Wine before. Where ‘guix build’ would substitute fine, but ‘guix install’ would try to build it from source. <lebowski>nckx: is there an example of any herd service which is written in something like python or bash and calls cli utils? <nckx>Almost certainly, but I don't know all services('s languages) by heart. You seem to be missing my point though, which is that patching your services' package to invoke /gnu/store/l77gnq45qfig4kh4y8j259h0ndqxfgx5-xrandr-1.5.2/bin/xrandr *works* and is the correct solution. You need to investigate why it's not working, for example by finding the error message. <apoorv569[m]>For openpgp-fingerprint thing needed for channel authentication I basically need to sign my commits with GPG right? and give my public key in place of openpgp-fingerprint? <lebowski>can i envoke any shell during the build to debug? <nckx>If you share more details it might be possible to help you, but currently there's not enough info to say anything more. It's not even clear when it's failing—above it sounded like run time, now you're implying build time? <lebowski>okay i need to rewrite my stuff to get more info, will be back <nckx>apoorv569[m]: Basically, if you're the only committer. More accurately, in (make-channel-introduction C (openpgp-fingerprint F)), F is the fingerprint of the key used to sign commit C. C is basically the user's ‘trusted commit’ anchor. If you have multiple channel committers, you can't use just any other authorised key. <lechner>mekeor[m] / please have a look at juix.org. i am looking for additional people who upload their store maps for comparison and better search <lechner>andreas-e / please have a look at juix.org. i am looking for additional people who upload their store maps for comparison and better search <mekeor[m]>lechner: wasn't there an attempt to implement this as a cli feature? iirc, there was a mailing list thread about it... <andreas-e>lechner: I bookmarked it for later treatment. With all the core-updates compilation, my store is filling up very quickly... <mekeor[m]>is it stupid to try to run node.js version 18 just by downloading it and running it inside guix-shell--container? that's where it get the libstdc++ related error. any ideas? :) https://paste.rs/Rcc <lechner>mekeor[m] / yes, but my approach is different. my solution, like so often, is social rather than technical. for example, you can see that someone has 'gocryptfs' but it's not on any of the substitute servers <andreas-e>How does it work? If I type gocryptfs into the search field, it finds nothing. Nor for libstdc++. <lechner>andreas-e / i'll take provisional builds, too. The nature of SQL abstracts very quickly from what you think is junk <unmatched-paren>andreas-e: i think you need to type the full path after /gnu/store/.../ <lechner>andreas-e / please try %glibc% or %/gocryptfs <andreas-e>Interesting. Indeed I am not familiar with SQL patterns. <graywolf>Out of curiosity, how does (local-file ...) that is later placed into /etc/... work when booting into previous configuration? Is the state of /etc restored on boot? Or will I end up with new /etc and old guix? I see `populating /etc from ...' message during reconfigure, so I'm unsure if that also happens automatically on booting into older generation. <lechner>actually, the wildcard isn't quite working the way it is supposed to, i don't think. the Pg docs say zero more characters so %/bin/gocryptfs does not presently work (but %bin/gocryptfs does). many improvements are neeed, and i have the experience to imlpement them. it's just a proof of concept to gauge public interest. plus it's written in Guile, which is currently a bit hard for me (to be published shortly) <lechner>andreas-e / i currently create a map with a script. it writes a bunch of files into an empty folder. if you give me access to those files, your content will show up on the website ten minutes later <lechner>submission online is one of the things i am working on <lechner>my hope is to assist anyone working on packaging, or merging things like core-updates <andreas-e>So would you just like us to send you the outcome of "file -type f" from /gnu/store? <lechner>andreas-e / no. give me a few minutes please, and i will post here <lebowski>nckx: okay, need more time. thank you for tips. will debug it tomorrow... <nckx>All right, I hope the extra info helps debug it :) <lechner>then create an empty folder somewhere and change into i <lechner>then execute guix shell -m ../juix.org/manifest.scm (or wherever it is located) <lechner>then GUILE_LOAD_PATH=/absolute/path/to/juix.org/git/scm:$GUILE_LOAD_PATH ../git/scm/catalog/client/inventory.scm (or whichever is the relative path) <lechner>That process may take an hour on rotational media, but is probably faster on SSD. <lechner>I'll accept those files in any archived form, such as tar.xz or whatever <lechner>the machine type of x86_64 is presently hard-coded, so please tell me if your base system is something different. i have not given any thought to cross-compiled files. perhaps my focus on the base system is an error <elevenkb>hey y'all's i installed guix with the script thingy on a foreign distro but I can't access the definition of the `guix-publish.service` for systemd. <elevenkb>it isn't in the location given in the manual. <lechner>the script creates a JSON file for each folder in your store (and ignores plain files, the metadata of which tends to be solely of local value). It also computes the "output hash" via guix hash (which in my view should potentially be part of the path names in the store) but no hashes of individual files in order to maintain a modicum of privacy and plausible deniability <andreas-e>In procedure stat: Datei oder Verzeichnis nicht gefunden: "/gnu/store/9pps4b60d7nl9wv6jmwvfqccmmfnszd9-jsoncpp.patch" <andreas-e>The file is a soft link to something that indeed has disappeared. <zamfofex>elevenkb: Which of the directories is missing from the file name in the manual? <elevenkb>zamfofex: ` ~root/.guix-profile/lib/systemd/system` <zamfofex>What files are there in ‘~root/.guix-profile/lib/systemd’, then? <elevenkb>everything? the whole directory `.../lib/systemd` is missing. <lechner>andreas-e / i have not seen such dangling links but will of course look into it. you are the first person other than i to run that script. you are also far more experienced with Guix and likely to do advanced stuff. please forgive the errors for now. <elevenkb>zamfofex: for now, i'll just modify the one in the guix source <mirai>colourized diff format for easier reading <lechner>andreas-e / if i do not respond here within a few minutes once you make the files available for download, please send the information to felix.lechner@lease-up.com. today is a federal holiday in the US, and i will go on a walk with my family shortly <andreas-e>lechner: Not doing anything advances, just running a chromium build... I need to do something else now, enjoy the family! <elevenkb>hey i get an error: `invalid option name/value` when running guix pull. <elevenkb>have no idea what could be causing it, because my channels file is the same as always. <mekeor[m]>ACTION is guix-pulling --branch=node-18-updates to test node version 18… <nckx>I just got a mail that my DNS server is up (again), so that's good. <nckx>elevenkb: Doesn't look familiar, but any more context to that error? <andreas-e>lechner: After "guix gc -D the_dangling_link" your script starts running. But I will stop it now: I have >100000 directories in /gnu/store, and each of them creates a json file of 1kB to 10MB size. This risks to fill too much of my home directory :) <gnucode>nckx: what's the benefit of running your own dns server? <elevenkb>weird yah, just `guix pull` already gives me that error <lechner>andreas-e / no worries, maybe another time! <elevenkb>even without any `-C` flag even though I usually use that. <andreas-e>The 10MB file corresponds to a checkout of openjdk, which has about 82000 files. Referencing them one by one and using a json output file is just too much. <mekeor[m]>gnucode: you could use it for dns-based adblocking, like pi-hole <lechner>i am not sure i have ever seen a 10 MB file, though. may latex <mirai>mekeor[m]: that's a dns resolver <andreas-e>If possible, it would be good to filter out the fixed output derivations, aka source. It would be interesting to keep only the built packages. <lechner>andreas-e / i plan to switch to cwebber's syrup as soon as i figure out how to use it <nckx>gnucode: I think we two have discussed this before :) ‘Fun’—that special brand of Fun®. <lechner>andreas-e / there is a lot of metadata about store folder i would like to have! right now i cannot even figure out the package name <lechner>i would also really like to know which derivation they came from, i.e provenance <andreas-e>Maybe it would make more sense to tie it in with the Guix Data Service, which has knowledge about derivations/ <nckx>elevenkb: But there's no surrounding (con)text? <TristanCottam[m]>> <@mekeor:matrix.org> gnucode: you could use it for dns-based adblocking, like pi-hole <lechner>andreas-e / all nice ideas. at this point, i cannot even link to packages.guix.gnu.org or to CI <lechner>i built and maintained the current version of lintian.debian.org, which does all those things and more <gnucode>DNS based pi-hole. nckx are you actively doing that? <apteryx>efraim: I kicked cuirass in case you hadn't seen form guix-sysadmins <nckx>ACTION falls asleep face-first on keyboard; by sheer power of coincidence banging out the letters ‘Good night, Guix’ in the process. <gnucode>unmatched-paren: you got blog post 2 out!? Geez the last one added 30+ points to my IQ! <unmatched-paren>gnucode, gabber: if either of you are planning on reading it, could you leave me a sneek message with your feedback? i'm concerned that it may not be comprehensible for people who haven't already understood monads fully <unmatched-paren>much of the time i spent writing it was either (1) procrastinating or (2) trying to figure it all out for myself :) <gabber>unmatched-paren: bold assumption of you to think i hadn't understood monads ;) <unmatched-paren>agh, i should have added an extra conditional to that sentence, i suppose :) <gabber>i will read it for sure - but probably not tonight. if i find anything especially noteworthy for you to feed back, i can gladly do so, but i have some confidence in our (guix project, that is) review process <mwette>Thanks for that blog on monads. super nice! <gabber>i really liked your last post, especially the hands-on approach! <unmatched-paren>when i started it, i didn't understand the topic very much, so i did have to spend a lot of time trying to figure it all out <unmatched-paren>but now that i (necessarily) have some degree of comprehension of the subject matter, i can't really judge it from the perspective of someone who doesn't <unmatched-paren>that looks interesting! i'll have a read, and then i'll have to go zzz :) <gabber>IIRC monads are one of the basic topics in functional programming - which was taught at uni.... so the term doesn't really scare me <unmatched-paren>writing an introduction to Guix monads is particularly annoying, though, because (1) Guile isn't a pure language, so there's not the obvious "IF WE WRAP THE SIDE EFFECT IN A MONAD AND PLUG OUR EARS AND SAY LALALALALA REALLY LOUD IT STOPS EXISTING" use case and (2) there aren't many monads implemented in Guix to demonstrate with <unmatched-paren>so i ended up first demonstrating the construction of a 'maybe' type, without talking about how it's a monad, and then showing how it benefits from monadifizication ;) <unmatched-paren>but it took way longer than i'd hoped because i simply couldn't think of a way to explain things at first <mwette>unmatched-paren: I appreciate the contribution. Thanks. <elevenkb>yah, the error is in some C++ code in guix, so i'm hoping that I can bootstrap into a `guix pull` by using the source code. <gnucode>unmatched-paren: what do you mean by guile not being a "pure language"? <unmatched-paren>it's possible to do stateful and io operations without sneaky monad trickery <unmatched-paren>which deprives guile of what is probably the clearest use case for monads <elevenkb>yah literally all my commands don't work. <elevenkb>even `guix shell hello` doesn't work still get the same error: ` <elevenkb>okay, I ran `guix system switch-generation` and `guix home switch-generation` and at least guix commands work for me now. <wdkrnls>Dear Guix, can you explain what exactly is an implicit input? <sneek>wdkrnls, you have 1 message! <sneek>wdkrnls, gnucode says: that "I have never been able to figure out how to do offloading..." <wdkrnls>I feel like I have an empirical idea of what it could be, but I'm not sure how to actually test it with Guix. <wdkrnls>When I run ag --scheme 'implicit-input' against a guix git clone I see some references to it, but no lovely (define (package-implicit-inputs package) ...)