IRC channel logs


back to list of logs

<jgibbons[m]>Does anyone here use guix professionally?
<Ribby>hey hey hey, I got the guix os correctly running! :)
<Ribby>yes, and full ethernet internet connection!
<Ribby>turns out that the fault lies in the setup of the usb's memory and file system. I wanted to do a partitioning table and space, this is for writing multiple os into their own specific partition.
<Ribby>I can even use the gnu grub to custom boot a os partition! Too bad, I can't figure out if I messed something up because I was going for gold, then I gone broke!
<Ribby>people say that os may have their own unique boot installer and their own different way to formulate partitioning table and partition space.
<Aurora_v_kosmose>We don't have support for JS build systems, do we?
<nckx>There's a node-build-system. I have no idea what it's good for.
<nckx>Ribby: Congrats! :)
<nckx>Good night, Guix.
<Aurora_v_kosmose>nckx: Oh you're right. Ah... I was looking in the wrong file. node.scm is where all node-using packages are, not javascript.scm
<Ribby>Do you guys know how to remove memory cache and temporary files in the xcfe desktop environment? I can do that on a gnome desktop environment.
<apteryx>how can I copy the *content* of a directory to the current directory?
<apteryx>(in Guile)
<iyzsong-w>apteryx: (guix build utils) module has a 'copy-recursively' function.
<apteryx>ah, thank you, I had forgetten that copy-recursively acts like 'cp -rT'
<apteryx>fun, this worked:
<apteryx>weighs 1.7 MiB instead of 144
<apteryx>(selenium 4.1.0)
<iyzsong-w>cool :o
<Ribby>Do you guys support efibootmgr? EFI variables not supported?
<Ribby>EFI mode?
<Ribby>Oh, it's in UEFI boot.
***iyzsong-www is now known as iyzsong-w
<Ribby>I want to uninstall efibootmgr and efivars, but I have to pull, package -u, and upgrade.
<Ribby>You guys have mirrors for pull, package -u, and upgrade?
<kitty1>Yo UwU
<char[m]>Is there a way to ask guix why a store item exists? like what profile or installed package needs it.
<Aurora_v_kosmose>Hm, to convince a dev to no longer depend on jquery or embark on the fool's errand of packaging jquery...
<Aurora_v_kosmose>Hmmm... there's like... 17 matches in the project for '$('. I could probably patch jquery out myself.
<char[m]>Aurora_v_kosmose: Although there is not a proper jquery package, several R package for web interfaces have jquery. maybe you could look at the way they do it.
<jgibbons[m]>Does `guix system search` search in all channels?
<jgibbons[m]>Or is it limited to the default "guix" channel?
<Aurora_v_kosmose>char[m]: Most of those packages seem to be built around grabbing the blob off the site in a manner controllable by those languages/libs.
<Aurora_v_kosmose>It's barely one step off simply vendoring.
<char[m]>Aurora_v_kosmose: Yeah, definily not ideal.
<PurpleSym>sneek: later tell civodul: I don’t have access to a powerful machine or good internet connection right now, so I cannot work on it at the moment. The branch wip-python-pep517 would have to be rebased and failing packages need to be fixed.
<sneek>Got it.
<Aurora_v_kosmose>char[m]: :/ Seems like the bootstrap situation is somehow even worse.
<Aurora_v_kosmose>And unlike jquery, the package's dependency on it seems much more tightly bound.
<Aurora_v_kosmose>I'll set that aside for today and come back to it later & see if I'm not simply misunderstanding somethign.
<Aurora_v_kosmose>bootstrap, jquery, w3.css... Yeah.
<Aurora_v_kosmose>I think finishing that patch will require packaging those dependency chains. It's too tightly tied to them.
<raghavgururajan>Hello Guix!
<iyzsong-w>hello raghavgururajan!
<whereiseveryone>jgibbons[m]: guix searches in all channels you're subscribed to
<whereiseveryone>but it doesn't report news from all channels or atleast, I haven't been able to get it to do so yet
<blake2b>hi guix
<blake2b>something seemingly simple that i've never been able to figure out how to do with guile is to reload a module that i'm working on
<blake2b>from what I understand, at the REPL it should be:
<blake2b>(guile-user)> ,reload (my module)
<blake2b>or: ,reload (current-module)
<blake2b>but that never seems to work. any tips on what I'm doing wrong?
<ams-hash[m]>So if config.scm has the install base packages. How do i edit what the base packages actually are
<ams-hash[m]>(I want to remove afew of them)
<ams-hash[m]>How do I install the nix package manager on guix
<AwesomeAdam54321>ams-hash[m]: You can do `guix install nix` to install it
<ns12>Hello, I am surprised that "guix remove ..." sometimes requires a download of various files. Why is that? If I am only removing a package, why do other packages need to be downloaded?
<jpoiret>ns12: since you'll have to run profile hooks, you'll need the software that's used in them
<jpoiret>when modifying profiles, you always run profile hooks
<ns12>jpoiret: What are "profile hooks"?
<jpoiret>basically, they're additional things generated in a profile based on what packages are in it
<jpoiret>like the man pages db, gtk pixbuf loader cache, etc.
<jpoiret>and to generate these, you need some software usually, that may not already be in the store
<ns12>jpoiret: Thank you. That answers my question.
***iyzsong-www is now known as iyzsong-w
***iyzsong-www is now known as iyzsong-w
<abrenon>yo guix
<efraim>tracked my sddm/pam issue to the expat graft :/
<EndOfAnEra>so I'm unable to evaluate a module with a package definition with a simple hello world, it looks like a bug
<EndOfAnEra>Here's a screenshot
<EndOfAnEra>"unknown file:#f:#f: encountered raw symbol in macro output in subform socket of (current-location-vector)
<EndOfAnEra>There's no "socket" subform anywhere I can find
<EndOfAnEra>is it something with my set-up with geiser+emacs+guixSD repl
<jpoiret>can you load it with a standard guile repl?
<jpoiret>btw, guixSD is now just guix system
<EndOfAnEra>jpoiret: you mean guix repl?
<EndOfAnEra>1 sec, I'll google how to load the file
<jpoiret>just (load "file.scm")
<EndOfAnEra>yeah it loads
<jpoiret>then it seems like a geiser issue
<jpoiret>well, rather, I think it's a geiser interop problem with the Guix-specific guile things
<EndOfAnEra>ok, most likely, I'm mixing packages from both Doom Emacs, and guixSD-installed "guix install emacs-..."
<jpoiret>can you show the backtrace with geiser?
<jpoiret>you should have entered a new prompt in the REPL
<jpoiret>you can type ,bt in there to get the backtrace
<EndOfAnEra>jpoiret: here's the backtrace
<EndOfAnEra>looks like a syntax parsing issue
<jpoiret>ohhhh, maybe geiser doesn't support guile 3.0.8 yet
<jpoiret>there was a change with `read` i think, that also caused issues with g-exps
<EndOfAnEra>typical, I run into all issues in the world preventing one from even starting to develop
<jpoiret>I personally don't think geiser is required for guile dev
<EndOfAnEra>I really need autocomplete and such
<jpoiret>you'll find geiser autocomplete to be pretty lacking
<jpoiret>it will for module variables, but even for simple syntax transformers I don't think it works well enough
<jpoiret>it's no DrRacket
<EndOfAnEra>jpoiret: do you use just plain scheme modes?
<jpoiret>I think so
<jpoiret>only need syntax highlighting and smartparens (rainbow delimiters too)
<EndOfAnEra>when learning, you need fast round-trip too, a key combination to evaluate in repl, and then it's all try try try ad infinitum
<EndOfAnEra>will look into the geiser configuration, thank you
<jpoiret>I have a terminal on another workspace in which i just often do `guix build -f mypackage.scm`
<jpoiret>by the way, your package symbol shouldn't be a function
<jpoiret>it should be (define my-hello ...)
<jpoiret>unless there's a reason you'd want it to be a procedure
<EndOfAnEra>jpoiret: thought I'd use this module in a separate file, from the system or home config
<jpoiret>otherwise you may run into a nasty error
<EndOfAnEra>use-module, then call the function providing the package declaration
<EndOfAnEra>I understand with encapsulation and exports, but it has to be done right, otherwise it's all one large hacky file
<jpoiret>but you could just export the variable itself, and use it like (profile (list my-hello)) for example
<jpoiret>you can export variables too, not just procedures
<EndOfAnEra>sorry, for us that try to develop professionally, we can't accept half-solutions, need modules etc
<jpoiret>that's how the packages in gnu/packages/*.scm are defined and how you use them
<jpoiret>modules can export variables is what I'm trying to say
<EndOfAnEra>I'll look into making it work, thank you so much for confirming, I thought I was doing something wrong
<GuiGuiGuix>Hello I face some troubles with ungoogled-chrome 98.0.4758.102-1 and teams (needed for my job). I've tried to install the previous version with guix package -i ungoogled-chromium@97 but doesn't find it. What's the best way to use the version 97 thanks
<jpoiret>you'd need to use time-machine
<jpoiret>i'd try `guix time-machine --commit=c334b7c52fe77b68a90b23fbac5c9de7337e607b -- package -i ungoogled-chromium`
<jpoiret>(that commit is the one right before the update to 98)
<jpoiret>although the package will get updated again if you do `guix package -u`
<jpoiret>for cases like this, you can use a separate profile with something like `guix time-machine --commit=c334b7c52fe77b68a90b23fbac5c9de7337e607b -- package -p ~/.chromium-profile -i ungoogled-chromium` and start chromium from there
***taylan2 is now known as taylan
<tribals>Hi there!
<tribals>How to install package's all outputs in manifest? What is the syntax to do that?
<tribals>May be I need to specify package multiple times?
<ngz>Hello. I have been encountering the following error lately, when trying to update some packages. Does that ring a bell?
<ngz>gnu/packages/docbook.scm:479:40: In procedure inputs: error: %texlive-latex-listings/deprecated: unbound variable
<ngz>It looks like a problem in dblatex, but I see nothing wrong there.
<nouun>I'm trying to debug building a package, is there a way to stop guix from deleting the tmp directory so I can inspect it when the build fails?
<Guest7657>Anyone in this chat use Hurd on main machine?
<jpoiret>tribals: yes, you'd need to specify (the-package "out") (the-package "lib") and so on
<ngz>nouun: Sure, use -K flag for guix build.
<ngz>-K stands for --keep-failed
<allana>Hi guix! I'm fumbling through writing my first system service. herd fails to start it, which is almost expected. Are there any kinds of logs that I can get from herd that explain any reason why a service can't be started?
<civodul>allana: hi! logging is not a strength of shepherd, but hopefully you'll find info in /var/log/messages
<sneek>Welcome back civodul, you have 1 message!
<sneek>civodul, PurpleSym says: I don’t have access to a powerful machine or good internet connection right now, so I cannot work on it at the moment. The branch wip-python-pep517 would have to be rebased and failing packages need to be fixed.
<civodul>PurpleSym: we can have it built on ci.guix, would that help?
<civodul>perhaps rebasing should come first
<allana>civodul: thanks. No luck for me though, will carry on :-)
<PurpleSym>civodul: I don’t have enough time to rebase that branch right now unfortunately. Maybe in a few weeks.
<civodul>it'd be great if the burden was shared among several people
<PurpleSym>Sure, that would be nice. If anyone wants to go ahead, please do.
<PurpleSym>I can answer questions here or by mail with some delay.
<nouun>ngz: Thanks, that's exactly what I was looking for.
<ngz>civodul: Hi! I have been encountering an error lately, I may be related to a recent commit of yours. Here it is: gnu/packages/docbook.scm:479:40: In procedure inputs: error %texlive-latex-listings/deprecated: unbound variable
<ngz>Does that ring a bell somehow?
<ngz>It may*
<florhizome[m]>I'm trying to use guix home on a foreign distribution; i used guix home import to import the existing files. But everything related to the shell doesn't seem to work; critically the env vars aren't set up, for example.
<florhizome[m]>Could this be, because the bash configuration files are symlinked from the store, and so they can't be executed?
<civodul>ngz: it should be fixed by "rm gnu/packages/docbook.go && make"
<jpoiret>383 unread mails after not looking at them for 1 week and a half, oops
<civodul>ngz: it's a consequence of the texlive package deprecation and subsequently d93bb18386cbd936c2cf35d013c60f33c5a25010
<civodul>jpoiret: it's terrible!
<ngz>civodul: Oh, interesting. Thank you. Yes, I thought so (about the commit) but couldn't decypher the error message.
<civodul>my email setup is in flux due to an ongoing migration from one server to another, so i'm even more likely to lose email than usual
<iyzsong>florhizome[m]: with guix home managed bash, its: ~/.profile -> ~/.guix-home/setup-environment, maybe need to remove ~/.bash_profile if it exists.
<jpoiret>florhizome[m]: default store permissions are r-xr-xr-x
<iyzsong>remove both ~/.bash_profile and ~/.bash_login, and relogin (or bash -l for test)
<apteryx>hi Guix! I've updated our python-selenium package to 4.1.0; its test suite needs two JavaScript files nowhere to be found other than its binary wheel. They are free software, but minified. Is it acceptable to fetch them from the binary wheel and use them just so the test suite runs?
<zimoun>apteryx: is it possible to exclude them from the installation? Just keep them for running the tests ?
<apteryx>oh, I hadn't check, it seems they got installed, bad
<apteryx>I'll delete them and see if the package runs normally without them
<efraim>can we skip those tests?
<apteryx>the files in question are selenium/webdriver/remote/getAttribute.js and selenium/webdriver/remote/isDisplayed.js
<apteryx>Expat licensed
<efraim>probably better for the mailing list than IRC but I've started feeling like we should have "blessed" versions of jquery and some other javascript libraries that we can use to patch in rather than pretending some packages don't have them bundled
<efraim>are they unminified in the git repo?
<jpoiret>efraim: wdym by "pretending some packages don't have them bundled"?
<efraim>jpoiret: find /gnu/store/ -name '*jquery*'
<jpoiret>oh, do you mean just unbundling?
<jpoiret>that'd be cool but there might be compatibility issues wrt versions
<jpoiret>i'm no web expert though, those territories scare me :)
<apteryx>efraim: sorry, I may have missed a reply, but were you saying these getAttribute.js and isDisplayed.js are from jquery?
<efraim>apteryx: no, I meant in general I'd like to package a blessed jquery and replace a bunch of copies of it
<efraim>apteryx: I was wondering if upstream in python-selenium there's an un-minified version of those two javascript libraries
<apteryx>OK! I agree that's a worthy goal (like any unbundling in Guix)
<apteryx>efraim: it's nowehere to be found in the git checkout!
<apteryx>I guess their Bazel build system fetches them from sowhere on the net
<apteryx>I haven't tried understanding their Bazel build system yet
<allana>to continue and hopefully gain some help with writing a service for the first time, I am writing a service that uses "make-forkexec-constructor/container". Herd fails to start the service, and I am curious to know where I can look for any errors. As civodul suggested, I checked /var/log/messages, and I also looked at /var/log/containerd.log hoping to see something related to my service. I did not. Anyone have experience here debugging
<allana>system services? I'm afraid that I don't have a clue where to look.
<efraim>I fight bazel every couple of months for tensorflow
<efraim>It's been a while since I last wrote a service, I'd start by using a minimal os image with guix system vm. IIRC guix system container doesn't like make-forkexec-constructor/container, too many layers of containerization
<allana>For example, I'm even curious to know what my gexps evaulate to, and I don't really know how to find out. Still feel pretty new to guile and scheme.
<efraim>often I'll either throw some of the code into a guile repl to see what comes out or check out the files after they've been built
<efraim>you can cat the .drv files to find the 'out', and then then look in there
<allana>efraim: thanks for the tip! I struggle here, even in geiser/guile repl, even with ,use (guix), I can't jump-to-definition or look up documentation for the guix stuff. I must be doing something wrong
<apteryx>efraim: this commit shows the files are probably "generated" by yet another build system, "buck":
<apteryx>Web feels like the Far West to me
<apteryx>where each file get to have their own build system ^^
<efraim>ugh now I'm getting build failures for python-minimal@3.9.9 with expat@2.4.5
<apteryx>seems if I use rake and un the 'prep' task, I may be able to generate these .js files as part of the build (c.f.:
<florhizome[m]>> <> I'm trying to use guix home on a foreign distribution; i used guix home import to import the existing files. But everything related to the shell doesn't seem to work; critically the env vars aren't set up, for example.
<florhizome[m]>> Could this be, because the bash configuration files are symlinked from the store, and so they can't be executed?
<florhizome[m]>or anyone using guix home on a foreign system...
<acrow>I use guix home on both foreign and SD systems but I'm no expert.
<acrow>I would ask if you have set up local-environment-variables simple service?
<AwesomeAdam54321>Would it be a good idea if all shepherd services had their stdout sent to specific files?
<AwesomeAdam54321>It would be an elegant solution to the logging problem since all the service logs could be found in /var/log/services/* for example
<jpoiret>AwesomeAdam54321: there's already #:log-file (
<jpoiret>but yeah, it's an optional argument
<florhizome[m]><acrow> "I would ask if you have set up..." <- No I’m doing it with the environment–variables field of bash configuration
<acrow>Are you using the home-bash-configuration?
<acrow>The biggest problem I recall from my conversion to guix home was that the init did not correctly parse aliases in my bash_profile. I had to manually do that.
<apteryx>zimoun: OK, I now know the javascript minified sources are usually generated from other sources at build time by Bazel, and needed at run time
<florhizome[m]><acrow> "The biggest problem I recall..." <- I‘m not sure what actually gets parsed of .bashrc and .bash_profile, tbh.
<acrow>IIRC, guix home puts a layer of guile between /etc/profile and whatever shell you choose to run. Or, of course, just entirely replace the shell with a repl. But there is a tricky bit in the bootstrap because in the process of replacing ~/bash{_profile,rc} you cannot fail to preserve the GUIX environment variables and there is danger of re-writing them after setting them.
<acrow>the tricky bit is in restarting bash after guile (the default config).
<civodul>hey samplet
<samplet>civodul: hi
<sneek>Welcome back samplet, you have 1 message!
<sneek>samplet, zimoun says: do you know the status of IPFS and SWH? Especially regarding Disarchive and Disarchive-DB.
<civodul>samplet: how's Disarchive going? i saw the thread a couple of weeks ago on swh-devel and the perspectives sounded quite exciting!
<samplet>civodul: I was pretty happy to stand up a prototype (it came together quickly). I think the new design is clear/good enough, but I’ve been unable to work on it for the last few weeks.
<samplet>Right now, I’m trying to clean up some Gash-Utils stuff so I can cut a release and subsequently finish my clean up of commencement.scm.
<civodul>samplet: oh, that's another nice endeavor
<zimoun>apteryx: ah, the JS story is always really ugly. )-:
<jpoiret>i've been out of the loop for the past week or so, has there been some progress on the next release?
<jpoiret>i'll resume working on (current-guix) soonish™
<samplet>I think the JS story is getting markedly better. For instance, I learned about this at Guix Days:
<civodul>jpoiret: not that i'm aware of, but it seems we're in a better shape now than a month ago
<civodul>i'm willing to give a hand and get that past the finish line!
<samplet>Spot checking those packages, it seems like if we could bootstrap a few test frameworks, we could have a pretty respectable collection of nice JS packages.
<jpoiret>i think it'll be just some minor tweaks so i should be able to finish it myself, but thanks!
<phf-1>Hello Guix! I'm writting a small tool in Guile to get a hang of it. I was wondering: is there a module to pretty print guile code? A better one than `pretty-print'? For example, my output is:
<phf-1>But course, I would like something closer to:
<phf-1>I'm using `(pretty-print datum #:width 80 #:max-expr-width 80)' for now, but does something better exist?
<apteryx>phf-1: no, I think that's the best we have
<phf-1>apteryx Ok, thanks!
<florhizome[m]>if I put an echo „test“ at the end of my .bash_profile , where should I be able to see that be printed?
<phf-1>florhizome[m], try `bash --login'
<phf-1>florhizome[m] in other words, when bash is called as a login shell.
<florhizome[m]>ah ty
<florhizome[m]>Ok we got some errors there
<florhizome[m]>Don’t seem to be escaping right
<florhizome[m]>Also, I get a warning about guile failing to install locale but that’s since always...
<phf-1>florhizome[m] "Also, I get a warning about guile failing to install locale but that’s since always..." What do you mean exactly ?
<florhizome[m]>‚ guile: warning: failed to install locale‘
<civodul>florhizome[m]: see
<civodul>samplet: re Gash-Utils, i wanted to propose the inclusion of code based on as a replacement of the existing tar code, which seemed less flexible
<civodul>do you think that would make sense?
<attila_lendvai>is something like a conference happening? it became quiet on the list.
<samplet>civodul: I do like that code better (particularly the fold-style interface). I probably won’t do it this release, though, because testing the full bootstrap takes ages and swapping the tar implementations could cause very time-consuming problems.
<samplet>I’m hoping my next test will be the last, but I fear that some recent changes in Gnulib will cause problems.
<civodul>samplet: yeah, probably not the best time to change this
<civodul>are you starting off the wip- branch that janneke had put together a while back?
<florhizome[m]><civodul> "florhizome: see https://guix.gnu..." <- Weird I had the locales installed in guix homes profile but the env var pointed to my normal profile...
<florhizome[m]>ok so bash ––login seems to work.
<florhizome[m]>It moreso seems that nobody is actually running it
<samplet>civodul: No, I’m not on that branch. I want to take a look at it, though. (I know janneke has been busy with other stuff lately, so maybe I could help out there.) Anyway, I think my work will be mostly independent of that other branch.
<florhizome[m]>when would bash ––login be run normally? Maybe I miss that in my greetd config
<florhizome[m]>or should guix home do that automatically?
<acrow>florhizome: I think guix home starts bash which then reads it.
<florhizome[m]>hm apparently it doesn’t
<acrow>forhizome: .bashrc and .bash_profile are linked from the store where guix home placed them after some massaging. You are suggesting that bash is started wo any initialization script?
<acrow>BTW -- reading that issue 42162 is fascinating. I had only been guessing what disarchive was before reading this. Metadata and serialization and hashing are so intimately connected and have historically been done so many different ways.
<florhizome[m]>acrow: Yes
<florhizome[m]>when I execute bash ––login, the stuff in my .bash_profile gets set up.
<florhizome[m]>so what seems to be missing is something running that command
<florhizome[m]>.bashrc gets eval‘ed though
<acrow>.profile should be set up too. Then the scripts usually chain together .profile -> .bashrc -> .bash_profile (if login). or similar YMMV but bash is going to go through it's init path sequence to find what it can.
<acrow>Oh, something else. The .bash{_profile,rc} are in your ~/src/guix-home now, IIRC.
<acrow>I think you have to move them out of $HOME, identify them in your home-configuration.scm and then guix home re-establishes them from the store. Otherwise guix home will overwrite the files you are telling it to read to create those seame files. It's a bit of a race.
<acrow>guix home init also had a role to play. It may be that your config does a verbatim copy but init makes some modifcations. IDR.
<lfam>Is anyone else having trouble with certbot renewals?
<florhizome[m]><acrow> "I think you have to move them..." <- hm ok I might try that
<lfam>I'm also struggling to update the certbot / acme packages beyond their old-ish version. The check phase crashes with this:
<lfam>TypeError: don't know how to make test from: <acme.client._ClientDeprecationModule object at 0x7ffff5d49f10>
<acrow>It may be that guix home init modifies and moves them for you but you have to recall that in your home-configuration.scom
<florhizome[m]>acrow: how would I know that?
<acrow>Inspection is the only way I know.
<acrow>Our manuals are excellent but the guix home portions have not yet had the time to fully mature.
<acrow>I'm learning that sometimes the best documentation is the issue under which the functionality was birthed.
<acrow>Finding the guix home issue would be a good read. Which I have not done.
<lfam>My certbot issue was that my ISP blocked port 80 on their end. So nothing wrong with Guix stuff
<acrow>Every time you run guix home reconfigure your .bash{rc,_profile} will be relinked to whatever you've put into the /gnu/store.
<acrow>It is a feature.. establishing version controls that allow rollback etc. It does require a change of mindset regarding your management of shell scripts and whatever else you tell guix home about.
<Christoph[m]>Can guix be used with PureBoot? There is no pureboot package, but I'm not really sure if my question even makes sense.
<florhizome[m]>acrow I don’t know how this connects to the issue anymore...
<florhizome[m]>my .bash_profile simply doesn’t seem to be evaluated during the login process.
<apteryx>I was in some community channel, and their bot prints: "Hi, this is Ed Manlove, community member! Just a friendly nudge that "guys" is a bit gender-specific, and we'd prefer something that would include everyone, such as "Folks" or "Everyone" or "All." Thanks!
<apteryx>which is nice!
<apteryx>sneek: perhaps you can learn that trick?
<lfam>Perhaps, I am sympathetic to the goal
<lfam>It can be annoying to have the bot talking often
<apteryx>it prints this only when someone mentions "guys", or something
<unmatched-paren>hello, guix
<unmatched-paren>how would i create a package that inherits from another one, but replaces some inputs with others? i know how to do it for old inputs but not for new
<lfam>You "know how to do it for old inputs but not for new"? Can you explain what you mean by that?
<lfam>In any case, check out modify-inputs:
<unmatched-paren>i know how to modify alist-based `(("NAME" ,pkg) ...) inputs, but not list-based (list pkg ...) ones
<lfam>Ah, the old and new styles!
<lfam>I think modify-inputs should work for you
<ngz>I find it a bit unfortunate that some forms in modify input are strings and others packages. I understand why it is so, but it looks a bit strange. `append' and `delete' sound symmetric, but they are not (they couldn't anyway). Maybe an explicit `delete-by-name' would have been better, if only to avoid this symmetry misinterpretation.
<ngz>It is clearly documented in the manual, but I can't help being surprised when looking at some package definition.
<acrow>florhizome: Here is how I declare my simple one:
<unmatched-paren>what's wrong with this labelled input?
<acrow>florhizome: You will see that guix home applies this input into a template.
<unmatched-paren>i label it because there's a phase that manipulates it
<unmatched-paren>the parentheses match up correctly...
<lfam>What do you mean what's wrong with it?
<lfam>Does it not work?
<lfam>Anyways, the label usually goes before the value
<lfam>Check git-manpages
<unmatched-paren>I saw somewhere in the irc history that with new-style inputs you do (list pkg "name"); i'll try find it
<lfam>Oh, okay
<lfam>So, you still didn't answer my question. What *is* wrong with it?
<unmatched-paren>guix build: error: /home/paren/code/guix/gnu/packages/dlang.scm:102:2: package `ldc@0.17.6' has an invalid input: ("_" (#<origin #<<git-reference> url: "" commit: "ldc-v0.17.6" recursive?: #f> #<content-hash sha256:15jzs38wanks2jfp2izzl7zqrp4c8ai54ppsgm8ws86p3sbbkmj8> () 7fbd36fbbb40> "phobos-src"))
<unmatched-paren>no idea what's wrong with it
<unmatched-paren>hm, can i download the logs of the past few days somehow so i can grep through them?
<lfam>Do you have an example of a package that uses this labelling mechanism with the new style?
<lfam>I can see that the "phobos-src" string is not part of the (origin ...) s-expression. It's just hanging out there in the list
<lfam>So, the syntax looks incorrect
<lfam>However, without more context, I can't tell what list this is. Is it the list of inputs?
<florhizome[m]><acrow> "florhizome: Here is how I..." <- Yeah I tried to do it both by declaring it all with the plain–file function or calling local–file
<lfam>unmatched-paren: There's an example here:
<lfam>Oh, my bad
<lfam>That's not an exapmle
<lfam>It's an example of something else
<lfam>If it were me, I'd use the old style for now, in order to get the code working
<unmatched-paren>lfam: This is a list inside the list of inputs
<unmatched-paren>(list (list (origin ...) "phobos-src") ...)
<lfam>It would be great to find an example of it in the codebase
<unmatched-paren>yeah, i'll have a look
<lfam>That is, if your goal is to use the new style. If your goal is just to add this input, you could use the old style
<unmatched-paren>oh, hm, (list (list pkg "name") ...) is actually for getting specific inputs...
<unmatched-paren>i'll try reversing it
<unmatched-paren>nope, still invalid
<nckx`>(inputs (list (list package "foo") ...)) often written as `(,package "foo") is used to refer to package's foo output. Are you sure you're not mixing memories?
<unmatched-paren>yeah, i just remembered that
<unmatched-paren>but (list "name" pkg) doesn't work either
<unmatched-paren>i'm downloading all the logs from 2022-02-01 to today so i can ripgrep through them
<lfam>unmatched-paren: Your use case may not be served yet by the new style
<lfam>The handful of packages I know that do this are still using the old style
<unmatched-paren>ah, i'm remembering a message about outputs from 2022-02-27
<unmatched-paren>The source that i'm trying to use as an origin input is the phobos standard library for D (for LDC), but actually these libraries are also available as submodules of LDC, so i can actually just replace them with a (recursive? #t) in the git-reference
<unmatched-paren>not sure why they were inputs in the first place actually...
<unmatched-paren>and so begins the long process of `clang++ ...`
<unmatched-paren>oh, actually `g++` not `clang++`
<maximed>unmatched-paren: maybe for unbundling reasons?
<sneek>Welcome back maximed, you have 1 message!
<sneek>maximed, xelxebar says: Thanks for the pointers. You've at least given me some traction. Cheers.
<unmatched-paren>funny thing is, these packages were untarred and `mv`ed to exactly the same place where they would have been if they had been cloned with submodules
<unmatched-paren>with no further manipulation or anything
<unmatched-paren>really weird
<maximed>Are git submodules pinned?
<maximed>If not, then it might be for reproducibility.
<maximed>Perhaps it's ‘half-unbundling’ to make clear where the bundled dependency comes from?
<maximed>(and hence, whether it is up-to-date-ish or not)
<maximed>Also, I don't think that the SWH fallback currently supports recursive? yet
<unmatched-paren>that's fair
<Aurora_v_kosmose>I fixed up a few packages' tests & mailed in the patches. It uses fixes I found when I was attempting to package clog.
<unmatched-paren>this could all be avoided if i was able to bootstrap ldc with gdc, but when i tried it gave me eldritch cmake messages
<unmatched-paren>i don't think there's anybody who enjoys debugging cmake problems
<unmatched-paren>so i gave up on that endeavour :)
<unmatched-paren>especially since the gdc team seems to be planning to rewrite parts of it in d, so it won't be bootstrappable anymore
<unmatched-paren>oh, excellent, a build error spanning a hundred lines, how fun
<unmatched-paren>what's the default label for a new-style input? is it the package's (name)?
<unmatched-paren>it looks like it is...
<gnucode>howdy guix!
<civodul>unmatched-paren: yup, the default label is the package name
<unmatched-paren>maximed: turns out the dependencies are not pinned, but that reproducibility doesn't really make sense, since somebody could remove stuff from the web at any time, no?
<unmatched-paren>if unpinned submodules mess up reproducibility, that would too, wouldn't it?
<unmatched-paren>s/reproducibility doesn't/reproducibility argument doesn't/
<unmatched-paren>except software heritage... yeah, that argument does make sense actually :)
<unmatched-paren>hmm what should i do now?
<unmatched-paren>hang on, just realized that even though ldc-bootstrap-0.17 needs to use old style inputs, the next ldc-bootstrap completely overrides it anyway, so it's fine
<Aurora_v_kosmose>...okay that's weird. Did git-email do something didn't like, or I did mess-up the git send-email invocation...
<civodul>Aurora_v_kosmose: it looks good, but there's a trick when sending patch series so that all the messages appear under the same issue
<Aurora_v_kosmose>Hm, nope, I used the exact same. I think it the 10 might be something it didn't like.
<civodul>exact same as?
<Aurora_v_kosmose>civodul: Ah, that is indeed a step I failed to do. Both times. The previous patch series I sent was sent the same way, but debbugs caught it anyway.
<civodul>somebody must have merged them after the fact
<civodul>Debbugs is not this smart ;-)
<Aurora_v_kosmose>Oh. I see. I was completely blind-sided.
<civodul>that's okay, it's a bit "surprising"
<Aurora_v_kosmose>Well, now I know to avoid repeating that. I'd assumed it had noticed the "in-reply-to" that git send-email added to all the follow-up emails.
<civodul>ah yes, i guess it could do recognize that, but i don't think it does
<Aurora_v_kosmose>Unfortunate. Oh well, I'll avoid a repeat anyhow.
<ngz>Hmm. Does Guix have a "lpr" command in some package?
<unmatched-paren>my gut says that sounds like something from CUPS because it starts with `lp`
<Aurora_v_kosmose>I know cups has its own version of that..
<unmatched-paren>yeah, it's from cups
<unmatched-paren>i just checked with `guix shell cups` then tried to tab-complete `lp` and `lpr` was one of the options
<ngz>It does. Thank you Aurora_v_kosmose and unmatched-paren.
<gnucode>hello, I am trying to disable substitues...
<gnucode># guix-daemon --no-substitutes
<gnucode>like the manual says is giving me an warning message...I suppose I can configure the guix daemon in my config too.
<gnucode>I think I figured it out... (use-substitutes? #t) in the guix-service-type.
<lfam>I figure you'd use #f to disable substitutes
<unmatched-paren>LDC - the LLVM D compiler (1.28.1):
<unmatched-paren>where can i find the commit message standards for guix?
<unmatched-paren>i can't find it in the manual
<lfam>In Submitting Patches, it's described as the ChangeLog format, with a link to the GNU Coding Standards, unmatched-paren
<lfam>Basically, the title imperatively describes the effect of the change. Then, you describe the changes in terms of code, by file and variable
<euandreh>hi guix, fellow programmer learning about sysadmin here. Here's a question about logging: IIUC, the default Guix System installation comes with "syslogd" which handles syslog(3), logger(1) and similar logs and puts them in /var/log/messages, but also "rottlog" which is basically a cronjob to rotate logs. So from the syslog decoupled architecture, my program sends logs to whoever is listening,
<euandreh>syslogd listens to whoever is sending, but there is no default analyser for whatever is store in /var/log/messages
<euandreh>is that correct?
<civodul>euandreh: hi! roughly yes
<civodul>syslog is able to filter messages though, according to its configuration file
<euandreh>civodul: yep, I saw that in the default configuration.
<euandreh>other than grep/sed/awk and friends, any suggestions or preferences of a log analyzer or analyzers?
<civodul>euandreh: no, or at least not that i'm familiar with
*civodul pushed months-old patches