<lfam>Just /etc/mail is what we usually do for this kind of thing
<lfam>If we had a sendmail service, we might generate config files and put them in the store. Then, when starting sendmail, we'd give it the path of the config file. We do this for openssh, for example
<lfam>It really depends on how sendmail works, but the store can't be used as a default location for config files that users will create
<Rovanion>Hmm, now the build process either fails because there is no folder `/etc/mail` or if I try to create it with `mkdir-p` that the build is not allowed to create it :/
<nixo_>civodul: yeah I like less output tool. When lzip was added, I looked at the output enjoying to see lzip being used, but that's it. I'll raise the verbosity for a week when zstd will be available :D
<tissevert>lle-bout: yeah, it seems like I have to try guix-explorer, it looks really promizing
<tissevert>civodul: I must be missing something very obvious : I just cloned guix-explorer, tried to run ./explorer /run/current-system/configuration.scm and I got a lot of warning about failed autoloads, then I'm told that the target itself is not found
<tissevert>I'm obviously on GUIX, GUILE_LOAD_PATH is obviously set and defined properly (to its default value)
<tissevert>my shell successfully autocompleted the end of «…figuration.scm», proving that it's not a typo, and the file does exist on my system
<leoprikler>does anyone here use guile-ac-d-bus? I'm trying to establish a connection, but it seems the argument to connect is invalid
<Rovanion>I may have asked this before but is it possible to know exactly which command is run in a certain phase? Looking at the source of guix it should be `python setup.py test` but that command fails when I'm in the failed build directory with environment-variables sourced.
<leoprikler>mhh, perhaps test is included through the setuptools shim?
<leoprikler>try formatting it like "python" "-c" setuptools-shim "test"
<Rovanion>On the command line in the build dir it says that setuptools is not defined: https://paste.debian.net/1191496/ If I replace the check phase in the package definition I get that the symbol setuptools-shim is undefined. Running `python test.py` in the build dir does work and finishes successfully. Replacing the check phase with `(invoke "python" "test.py")` fails on the other hand due to what I guess
<Rovanion>leoprikler: Yup, that did it. Now I have the same failing tests in both the guix build and build dir. What I really want now though is for `(invoke "python" "test.py")` to work just as well as `python test.py` does in the build dir...
<tissevert>what do you people use in practice for development ? I started by making «default.scm» builders in each project's folder and using guix build -f or guix build environment -l but then local dependencies stroke back
<tissevert>do you have a local (gnu packages *devel-or-something-like-this*) module that references all your local projects ?
<tissevert>jlicht: thank you ! ok, so if I understand what you said + the page of the doc you linked, you set somewhere in guix.scm that uses «dependency», you have something similar or equivalent to (define dependency (load "../dependency/guix.scm")) so thate the «dependency» package exists in the scope of the main package definition ?
<tissevert>hmmm the above approach (define dep (load "../dep/default.scm")) doesn't work here
<tissevert>the variable gets loaded: I get errors when introducing unguilish noise in ../dep/default.scm or when I use the variable dep somewhere inappropriate in my file like (name dep)
<tissevert>but when compiling for good, even though I see the corresponding ghc-package (this is for haskell packages) in the compilation lines, the builder finally yells that it couldn't resolve dependency for package Dependency (with the uppercase initial showing it's considering it from the haskell packages point of view)
<yoctocell>tissevert: The guile part seems to work fine on my end, it starts to download a bunch of haskell packages.
<yoctocell>Maybe the problem is with some haskell stuff?
<tissevert>maybe but as I said: it builds fine in a stateful environment (on another distro where Cabal works), my other self-contained haskell packages with no local-dependencies work fine using a similar setup, and, finally, defining those two packages in a big module with several (define-public …) work flawlessly
<yoctocell>Hmm, I'm afraid I don't really know what to do, sorry
<tissevert>don't worry, thanks for all the help you've already provided !
<everstone>Hi! I'm trying to get my audio volume, but pulsemixer says "connection refused". Am I maybe missing a group? Audio works fine otherwise
<lle-bout>tissevert: hey! did you succeed running it? try: "guix environment guix" first, then run the command in a subshell there
<sneek>Welcome back lle-bout, you have 2 messages!
<sneek>lle-bout, raghavgururajan says: I am revising the chart to comply with upstream GNOME 40.
<lle-bout>raghavgururajan: aaa superb! thanks a lot!
<tissevert>I really just wanted to prompt other people to check I wasn't doing completely bad stuff, it seems you do things differently but aren't shocked either by my first attempt so I guess it's somewhere in between : )
<lle-bout>tissevert: I still can't run guix-explorer on my system configuration but I run it onto examples, the reason is that my system configuration depends on several third party channels somehow I can't manage to bring them into scope fully (specification lookups on packages included in these channels don't work)
<tissevert>everstone: I wanna say «polkit» here 'cause I have similar weird permission errors (with NetworkManager) here with a very default setup and correct groups so maybe it has to do with polkit
<tissevert>lle-bout: ohhh, I didn't see I needed that, thanks !
<tissevert>hmmm… no luck ^^' still doesn't see the file for some reason
<lle-bout>tissevert: I use GUILE_LOAD_PATH="/path/of/cloned/channel/git:$GUILE_LOAD_PATH" that works for some but it seems specification lookups still don't have that packages for those channels.
<lle-bout>tissevert: I suggest you can investigate the polkit issues still with some of the example configurations in gnu/system/examples in the guix repo
<lle-bout>tissevert: no rush and no need for promies - I think guix-explorer can be useful as a general resource to learn more about GNU Guix service composition which is very useful for solving ANY such issue
<lle-bout>tissevert: you may try civodul's suggestion here.. to use guix repl -- explore.scm .. just why havent I thought of it before heh..
<everstone>tissevert: I think polkit didn't help, but it maybe did. When I tried just running "pulseaudio", I saw an error related to my default.pa config. I moved it, and now it works!
<tissevert>yoctocell: ok, interestingly I get a very similar error trying to define two local bash packages, each one with a simple script containing a printf, and the «A» script calling the «B» script
<lle-bout>seems we can't give arbitrary expressions to guix repl like with guile -e
<civodul>lle-bout: something like adding (guix-explore (cdr (command-line))) at the bottom of the file
<lle-bout>civodul: thanks didnt know about "command-line" variable, useful, like my previous attempts it says this: guix repl: error: Usage: explore FILE
<lle-bout>Rovanion: generate numbered patches using git format-patch or use git send-email and submit all a single issue, also look at using -v2 -v3 etc. arguments, the number corresponds to the revision number of the patchset
<lle-bout>civodul: so final command, thanks a lot! From your original repo: cat explore.scm | cat - <(echo "(guix-explore (command-line))";) | guix repl -L$(guix build guile-syntax-highlight)/share/guile/site/3.0 -- /dev/stdin /run/current-system/configuration.scm
<lle-bout>woops well that command doesnt work because relative path isnt working correctly
<raghavgururajan>apteryx: Did you test it with linphone account or other SIP account. The former didn't work, but the later did. This for missing rootca.pem. For missing /lib, multiple codecs we not loaded for any case.
<lle-bout>tissevert: what are you speaking of exactly here?
<everstone>Is there a way to make my operating-system.scm source a helper .scm file? E.g. I define (my packages) for use-modules there
<tissevert>nothing in particular, just the spectacular (and scary) «file not found» I got on my first attempts
<tissevert>the tool will certainly provide meaningful insights when I start digging but I don't really have time at the moment so it doesn't really matter
<lle-bout>tissevert: for explore.scm? I don't find it that surprising that since there's no guix.scm file it's harder to reproduce civodul's dev environment
<everstone>Figured it out! Needed to change guile load path and move some files around
<jeko>Hey ! I want to create VM on-demand to host pair-programming session inside Emacs to learn Guile with friends. First time I try such thing, I am not used to VM stuff. So I would like to ask if I am going to waste my time (I feel it is pretty straight forward to ssh into the VM from the outside)
<tissevert>hmmm, true, I've been believing a bit too blindly «oooh it's guile code, it must be blessed by Guix reproducibility properties» when it has nothing to do in fact
<nckx>How are you starting GIMP, what happens, how does it tell you etc.
<lle-bout>nckx: on startup it says: "GIMP requires the GEGL operation "gegl:introspect". This operation cannot be found. Check your GEGL install and ensure it has been compiled with any dependencies required for GIMP." in some GUI dialog
<lle-bout>Bumblehorse: use a paste service like paste.debian.net next time, and yes sorry about that, try: "while true; do guix pull && break; done" - there's actually fixes available for this but you need to update
<nckx>yoctocell: I use a trivial bash script that builds the output of lle-bout's command.
<PotentialUser-60>nckx: I upgraded my user profile using a manifest ... i also tried to remove gimp from that manifast and put it into my system declaration ... both lead to the same results .. I run it by typing gimp into my awesome launcher ... or by typing it into alacritty (?)...
<lfam>You can also use `guix graph` in certain ways, for example with the reverse-package method
<nckx>PotentialUser-60: Thanks. Weird. I can invoke the latest ‘guix install gimp’ just fine (as ‘$ gimp’), and even launch it directly from the store (/gnu/store/.../bin/gimp); it fails only in a --pure Guix environment.
<nckx>That's what the output of refresh -l means, after all.
<raghavgururajan>lle-bout: No worries! Never mind about rebase, I've added the php fix patch.
<nckx>I'd say it's a bug if it outputs anything unbuildable :)
<nckx>(By nature, I mean; of course the build can fail.)
<lfam>nckx, yoctocell: I also recommend adding --no-grafts to the build command, unless you need the grafts for this
<lfam>Otherwise, it might use about double the disk space
*nckx adds --no-grafts to almost everything don't tell mother.
<lfam>I use it liberally, but never for `guix package` :)
<lle-bout>civodul: somehow I received 4 copies of your reply to "terrible ux emacs upgrading in guix"
<lfam>nckx: Changing the subject, do you know if anyone is, like, "in charge" of the web presence? I'm wondering who to talk to about the build failures of the website and manuals
<lfam>If you check the recent mcron.log on Berlin, you'll see what I'm talking about
<PotentialUser-60>nckx: strange, indeed. I removed it from the manifest. pulled the latest guix. Did guix install gimp && gimp same results. Not that I'm surprised but you never know :)
<lfam>nckx: I'm considering manually updating the symlink that points to the devel manual
<lle-bout>nckx: using this: "env | while read line; do echo $line; ./pre-inst-env guix environment --pure --ad-hoc gimp -E $(echo $line | grep -Po '^.+?(?==)') -- gimp; done" - which preserves environment variables one by one, every of them fail, so it must be a combination of two or more
<apteryx>raghavgururajan: I'm puzzled as to why I can't reproduce any linphone-desktop failures here? any tip?
<raghavgururajan>apteryx: Can you try removing linphone folders in .config .local and .cache
<raghavgururajan>Also, if you run from terminal, you'll see the things related to missing /lib and codecs not found.
<apteryx>yes, but these are just warnings and they are actually enabled (the codecs)
<raghavgururajan>apteryx: Once, clearing the .dot-files, try logging into linphone account using assistant. You'll see could not send request. Then apply only rootca part of my patch and re-login. login will succeed and you;ll get pop-up to download codec.
<nckx>*: honestly, I might just punt and revert gegl. I'm in no mood for a sometimesbug today, and the old one did the job.
<lfam>I think we only have a single Mathieu these days, right?
<lle-bout>cbaines: or how to observe every day what you do related to GNU Guix
<cbaines>lle-bout, I have a few Wekan boards that I use for various things (Guix Data Service, Guix Build Coordinator and a general Guix one) but they're not public currently
<cbaines>There's rough roadmaps for the Guix Data Service and Guix Build Coordinator in their respective Git repositories
<cbaines>and then there's the more goal like perspective on what I'm trying to do, so improving the patch review process, the tooling for operating a substitute server, and the thing I'm looking to start around security and trust in Guix