<oswald_t>If yes, i am playing with guix distribution, and I was wondering which packages are installed by default when a user is created ? Because, if I've well understood, each user has their own packages
<oswald_t>[thanks for your answers, sometimes I bug :) ]
<lfam>Like I said, the level of care depends on the package
<dftxbs3e>lfam, I think at some point we should even have GNU Guix Data Service do this for us and then gives us a patch as an URL to curl | git am -s in and double check then push
<rekado>for CRAN + bioconducter package updates I use ./etc/committer.scm a lot, but I also check the resulting commits. I wouldn’t dare push the results automatically. The code just isn’t smart enough to be relied upon.
<benoitj>does anyone see ripgrep not building anymore?
<zimoun>dftxbs3e: well, if you want to practise your scheme, instead of writing a Bash, you should try to write a Guile script. Then using GUIX_EXTENSIONS_PATH, you can have your own and do: “guix update osip”
<lfam>There are different types of distros. One example is pypi, which is just a clearinghouse that redistributes code. Another example is Debian, where they are extremely careful about who can update what, and in what circumstances. It's expected that package maintainers read the upstream code and understand it. Guix can be in the middle.
<zimoun>dftxbs3e: yeah but from my point of view, it is different to have personal helpers (even shared as etc/committer.scm) and full automation
<dftxbs3e>zimoun, not full automation but helpers yes, helpers that do things and tells you what you should do as human
<dftxbs3e>so as a packager for example you just have to sit and follow a well-tested process, e.g. refresh + commit automated then prints a message: check the project home page for news before committing
<zimoun>roptat: Hi! I am building camlboot and I am a bit confused because I am not able to parallelize the build (options -M or -c) and the Makefile seems a good candidate. So it takes ages. :-) Well, 4-6h had been reported, one on a laptop. What is your machine?
<davexunit>is it possible to add supplemental groups to the nginx user that the nginx service creates?
<davexunit>I need it to be a member of another group so it can have read access to files I want it to serve
<davexunit>alternatively: does anyone have a working cgit+gitolite setup? mine broke at some point because of file permissions.
<davexunit>I thought I had figured out a way to do it by making a new service type that inherits from nginx-service-type and changes the account service extension to use a modified set of accounts, but that just results in an error about there being 2 nginx services. presumably because the cgit service extends the official nginx-service-type
<Whyvn>I followed the manual for the postgres-service-type, but it seems to no be able to connect to the server when I try to do anything with postgres. I get the error "psql: error: could not connect to server: No such file or directory" I tried deleting the /var/lib/postgres/data directory and reconfiguring and restarting the service, but still same results.
<rekado_>dftxbs3e: committer.scm is a little primitive; it doesn’t yet attempt to summarize changes to the “arguments” field; someone looking for a fun project could add support to list the names of the changed/added/deleted build phases.
<dftxbs3e>rekado_, it's always impressive I think :-) - I tend to think the reason Scheme was used in GNU Guix is also to operate on code like this, read it easily, transform it easily, generate it easily
<aidalgol>Why is code for the step "Source etc/profile to augment PATH and other relevant environment variables" split into two command invocations, with the first setting an environment variable, instead of just doing it all at once?
<aidalgol>More importantly, for Build Environment Setup, should I set max jobs to the number of CPU cores I have, or can jobs run subprocesses in parallel?
<aidalgol>And is there any reason not to symlink the systemd unit files to /etc/systemd/system/ instead of copying them?
<nlyy>what's better, (for-each guix-download (map license-uri licenses)) or package with all licenses?
<kondor>Hello folks, just a question: i kust got a brand new ssd drive ~1Tb in size to dedicate to /gnu/store . Can I simply `dd` my existing /gnu/store onto that new drive, then mount it and all will be fine?
<Whyvn>is anyone successfully running postgres on guix system? I have the basic configuration for the manual. Leaving all the defaults how they are. In the pg_ctl log it says listening on IPv4 address 127.0.0.1, listening on Unix socket /var/run/postgresql/.s.PGSQL.5432 but when I sudo -u postgres -s /bin/sh and createuser --interactive it says it cannot find file or directory in /tmp/.s.PGSQL.5432, why would it use /tmp instead of /var/run
<cage_>hi! i just submitted my patches to include a new package (and many libraries ;-)) to guix, i would like to thanks all the channel for the help!
<jackhill>Has anyone else experienced ungoogled-chromium-wayland occasionally crashing sway? It seems to happen mostly when creating a new chromium window. Unfortunately, I don't have much better information than that as I can't get it to happen consistently.
<jackhill>Or any advice on getting more information about it? I could open a bug, but I'm not even sure if it's a Guix problem, and that's all the information I have.
<zimoun>civodul: indeed recursive? #t for SWH will be cool! The SWH fallback needs some love. :-)
<jfred>lf94: IIRC for some pastes it doesn't, maybe middle-click pastes? (it's been a long time since I've used it, but I remember definitely accidentally pasting multiline stuff into a channel when I did)
<simonsouth>siraben: Yes, unfortunately that keypress is gone. But I'll know this for next time.
<lf94>jfred: they probably fixed that, because i've never had it happen once
<yjftsjthsd>You know, I used to literally run a keylogger on myself. At the time, I did it to get keyboard stats because I was interested in alternative layouts - it's easier to justify using a new layout if you can show that it's more efficient for the actual things you have actually typed - but you've just made me realize that it would sometimes help to
<yjftsjthsd>figure out what keybind I just accidentally triggered too:)
<simonsouth>Yep. I think I accidentally pressed C-m instead of C-v after scrolling back to the top of the log.
<simonsouth>IT seems rcric will "helpfully" resend the contents of whatever line your cursor happens to be on when you press Enter.
<Noisytoot>pretzel: I meant merge it into Guix, not merge it with another bug report
<balance>Hello! Please help with next question (already asked but can't find info how to do that).
<balance>How to do detached header with luks (no lvm) on guixsd? When i do "guix system init /mnt/etc/config.scm /mnt" a error occurs "/mnt/etc/config.scm:22:22: error: no LUKS partition with UUID '11111111-aaaa-2222-bbbb-333333333333'" (here using fake UUID). Read manuals but did not find an example with detached header.
<balance>Read about `info guix "Initial RAM Disk"` but cant find how to put header and key (for the purpose not to enter luks password twice) to initrd? and how to write correct /etc/config.scm for the case described above?
<nckx>There is no built-in support for that, so why it's certainly possible (‘everything's possible’ with a Turing-complete configuration system) you'll actually have to create the Guix LUKS Detached Header System.
<nckx>It won't be as simple as adding a file to the initrd (there's no code that will do anything with it) but you can customise the code that invokes cryptsetup.
<nckx>sertified: That's because the error message is misleading.
<nckx>Don't you get the same error when you invoke it manually?
<sertified>nckx: I get the same error when run ./codium
<sertified>I've tried $ source codium but it says it cannot run binary file
<nckx>Binaries compiled for other distributions hard-code (by necessity) the location of the linker-loader, usually ‘/lib/ld-linux.so.2’. This does not exist on Guix. Unix being unix turns this into a useless error message; but it should really be ‘/lib/ld-linux.so.2: no such file or directory’.
<nckx>You can use patchelf to edit the binary itself to change this hard-coded location, or create a symlink.
<balance>nckx: Thank you for answer! How to customise the code that invokes cryptsetup, or where to read about that?
<sertified>Noisytoot: No. And I need to run this particular VSCodium since my professor pre-installed a Lean prover extension, which only exists for VSCode. Not to mention the Lean library that is specific for this class.
<pkill9>though what i want doesn't need this actually i think
<bavier[m]>Noisytoot: right, I was just hoping there might be some full sentences is upstreams docs that would also satisfy `guix lint`s desire to have sentences begin with a capital letter.
<Noisytoot>"Apart from being the fastest CommonMark-compliant Markdown parser implementation in pure Python, mistletoe also supports easy definitions of custom tokens. Parsing Markdown into an abstract syntax tree also allows us to swap out renderers for different output formats, without touching any of the core components."
<Noisytoot>Something else could be faster on a different computer
<Noisytoot>"Please avoid marketing phrases such as “world-leading”, “industrial-strength”, and “next-generation”, and avoid superlatives like “the most advanced”—they are not helpful to users looking for a package and may even sound suspicious. Instead, try to be factual, mentioning use cases and features. "
<sneek>nly, mwette says: with fix to nyacc ffi-helper I was able to generate a guile wrap for libfuse. If you want to try I would need to send it to you as oto big (at 200 kB) to paste. I'm guessing it would need some front-end work to make useful.
<dongcarl>Does anyone know how hard it'd be to have `guix environment` accept derivations as the argument in addition to packages?
<dongcarl>I've run into this several times now, where I have derivations which are not searchable fail, and I'm unable to use `cd /tmp/blah-0 && guix environment <...blah.drv>` to start debugging the build
<pkill9>does anyone know how to use `build-package` from guix/scripts.scm?
<zimoun>does the package have a symbol in the manifest?
<zimoun>dongcarl: it does not answer your question, but something like ’guix environment -L foo -e '(@@ (a module name) blah)'’; assuming you add a module definition in your manifest and the custom package is defined by a symbol.
<zimoun>and I agree that maybe passing a derivation to “guix environment” could be useful for debugging. :-)
<dongcarl>zimoun: Oh yeah for sure I can hack around it... It's just so nice to be able to copy paste the derivation path and do `guix environment`, just like how `guix build` can handle it, you know?
<dongcarl>Does anyone know the difference between x86_64-linux-gnu and x86_64-pc-linux-gnu?
<leoprikler>i think it's historical, though some might say hysterical :)
<paulj>Evening all! I am working through the packaging example using the GNU hello package, but I have come across an unexpected error. According to the result of "guix download mirror://gnu/hello/hello-2.10.tar.gz", I have "No space left on device". In this paste - https://paste.debian.net/1187723 - you can see I have run guix gc, then show the disk free as >70% of disk space, and the error message from guix download. Have I missed
<nckx>OK, that's the culprit (it's common: ext4 has a design flaw in that it hashes directory entries and... just dies when the hash table gets ‘full’). Try ‘tune2fs -E "hash_alg=tea" /dev/foo’ to select a different hash algo, or ‘tune2fs -O "^dir_index" /dev/foo’ to disable it completely if that doesn't help.
<nckx>I think that works on on-line file systems...
<civodul>roptat: i wonder if we could turn the miniml compiler into a proper Guile compiler front-end, while we're at it :-)
<paulj>nckx: Seems to have worked - no error message
<roptat>I don't think it would help the bootstrap, but could be fun
<civodul>Guile has this "compiler tower" that makes it easy to add a language front-end
<nckx>The trade-off is that scanning directories is now (theoretically at least) slower. Not much we can do, short of reorganising /gnu/store to use, e.g., /gnu/store/aa/aaaa...-style hacks to work around one broken file system.
<civodul>roptat: yeah, fun is what i had in mind :-)
<paulj>Should I have used a different file system other than ext4? If so, which would have been the preferred option?
<civodul>roptat: anyway congrats to the whole team! it's great that Gabriel & co. got into this
<nckx>paulj: I don't really know. People seem to be successfully using btrfs. We should probably add a hint about ext4 to the manual... this has happened several times by now.
<paulj>I am on the manual page at the moment, and it shows using ext4 when setting up the partitions,
<nckx>paulj: You definitely didn't do anything wrong; it's rare enough that I'm sure plenty of people run Guix on ext4 without a care. Maybe it depends on the exact contents of your store. Maybe there has to be a collision. I don't use ext4 myself.
<nckx>Yes, we recommend the file system that sometimes breaks.
<nckx>People were discussing changing that to btrfs about a year ago, guess it never happened or not completely.
<paulj>I'll stick with this at the moment, and consider changing to btrfs at some point in the future. I don't know anything about btrfs yet (like, is it stable!).
<paulj>Should be easy to start again on this system - I have worked hard to make sure everything is installed through manifests, so redoing it should be straight forward.
<paulj>Looking at the wiki page, seems like Brtfs is well used and no issue :). Maybe I will test out my ability to hose the system and re-install again...
<nckx>I've opened a bug report to remind myself to add a note about ext4.
<paulj>Thanks. I didn't answer your question between my posts above - yes, changing the hash algo to tea worked.
<nckx>Ah good, that probably avoids a performance hit completely.
<paulj>In that case, I will continue on my journey to create a package for ts, and look to play with file systems later.
<rekado>Madalin and I revived one of the build nodes (hydra-guix-128) today. Networking was never correctly set up and the serial console baud rate was also misconfigured, so we always postponed fixing it.
<rekado>I’m now reconfiguring it with “guix deploy”
<rekado>is there something else I should do to connect it to cuirass?
<civodul>rekado: nice! if you use the same config as the other build nodes, it should have the cuirass stuff too
<vagrantc>but notably, one of the other ones in .cache were up to date as well ...
<vagrantc>run into this periodically and it's a bit confusing exactly what branches need to be updated when and where...
<zimoun`>yeah, currently dot is unusable in practise. And I am annoyed by my iterations with “grepping” in the dot. And, I am in the mood about graph: I would like to be able to visualize what happens when an input is added/removed. Or what is the graph for pkg1 and pkg2 compared to pkg1 alone. etc.
<vagrantc>dftxbs3e: what i find confusing is it clearly successfully authenticated commits by you in the past
<vagrantc>would be nice to get rid of the skip-dex-test-with-missing-procyon phase in diffoscope and get that fixed upstream ... but i think the current substitute call might match multiple tests ... need to test which tests are affected for sure before pushing upstream
<rekado>zimoun: you could load the whole graph into d3 as JSON and then do without a server.
<dftxbs3e>projects that release security patches without making a release are really annoying
<vagrantc>or package procyon for guix ... but last i tried that some firm tugging of hair was involved
<rekado>“guix deploy” for the revived build node failed
<rekado>Throw to key `match-error' with args `("match" "no matching pattern" ("none" "/sys/kernel/debug" "debugfs" () #f #f #f))'.
<lfam>I think that, traditionally, this was considered good practice. Distros like Debian won't update the package anyways, but only patch old versions, so upstreams learned that it wasn't important to release for security updates
<rekado>I tried mounting the debugfs manually, but I still can’t get the service to start
<dftxbs3e>lfam, qemu's handling of it and grub2 recently is extremely annoying
<dftxbs3e>would be nice if we could annotate individual problems and say why they are false positives so they don't appear in the list
<lfam>If it is useful, sure. I'm not familiar with that site
<dftxbs3e>lfam, it's actually a really great site! I have NVD rss feeds as well and many others but that repology Free Software site does really lots of work we could use, a bit like what GNU Guix Data Service could do in the future
<lfam>Taking the first package as an example, 389-ds-base
<lfam>Is there somewhere to click that explains what is wrong
<lfam>The thing to remember is that using the newest version of every program will not create a working operating system. We have to do a lot of work (largely automated with the system tests) to validate things, and that is what "distros are for", which is sometimes forgotten by the anti-distro single-language people
<lfam>Usually, we do something like you describe for several months. We just update everything we can think of on the core-updates branch. It's totally broken. Then we fix it, which takes another few months. And only then is it useful
<dftxbs3e>I understand, what I meant is that master isnt really a downstream to core-updates