<mirai>afaik it's still relevant but pending review
<lechner>mirai / didn't you write that it would be superseded by your proposed changes to define-configuration?
<winter>jpoiret: speaking of current-guix, is it normal for it to never recompile unless you touch config.scm? i've tried editing/recompiling random modules and config.scm is the only one so far that's actually made it do something... guess it's something wrt to the channel build system, though it just seems weird that no other module changes (e.g. (gnu
<winter>build image)) would cause it to rebuild, no?
<NewUser-Basic-Qu>Damn...playing with desktop-services left me with a system that does not allow me to login directly and does not bring up netwoerk and ssh to login from outside. Any chance to chroot into it?
<iyzsong>oh, didn't know this chroot manual section, seems not too hard :)
<winter><winter> not even self-build.scm... <-- I edited channels.scm to change the path of self-build.scm, and that threw, so it's definitely reading it. I edited it to remove the "Computing Guix derivation..." message, but it's still showing. 🙃
<cel7t>I'm getting a ft2build.h no such file or directory error while building packages that rely on freetype2
<lechner>yeah, i'm going to ask the reviewer to write it. i understand neither the format nor the purpose. everything that's needed for a changelog elsewhere and to explain the motivation for the commit is there
<jpoiret>lechner: doesn't seem like the best patchset for busy reviewers then
<iyzsong>lechner: done, thank you. i'm not a good commit message writer, so i only add the missing title and file modification info.
<jpoiret>winter: did you commit the change to the build-self.scm file before testing it out?
<tux_life>Hi! I know how to edit a file in /etc via etc-service-type. Can this service also be used to edit files within a subdirectory, eg /etc/security/limits.conf? As? Should I indicate as filename "security/limits.conf" ?
<civodul>tux_life: hi! for /etc/security/limits.conf, there's a separate service that exists
<tux_life>civodul Thank you! Interesting... Now I'll see how it works.
<Guest74>I can't start EXWM with slim through my .xsession file. I have "exec dbus-launch --exit-with-session emacs --maximized --debug-init --load ~/.exwm" it starts, but I get some xelb authentication errors
<bjc>civodul: i'm looking through the new shepherd and while i'm still fitting it together in my head, i'm pretty sure you can get rid of the ‘(when (one-shot? obj) …)’ test in the service start method
<bjc>that case is already handled by the service controller when it gets the ‘*service-started’ message
<bjc>sending the additional ‘notify-terminate’ doesn't do anything. additionally, it races with ‘*service-started’. the race is annoying but harmless, since they do the same thing in the end
<civodul>bjc: oooh, it wasn't entirely clear to me but apparently removing that (when (one-shot? ...) ...) doesn't break anything
<sturm>mirai: thanks, I've seen the docs for how to configure system-level tasks, but didn't realise you could use it for non-system things too. Where would this config go? (presumably not done in a `guix reconfigure`)
<Kabouik>Does our libva package include what would be included in the libva-wayland2 package of Debian-based distros? I need to install dependencies for an Appimage (which does not include everything it needs apparently), and libva-wayland2 is one of them.
<Kabouik>libva 2.15.0 has email@example.com as a dependency so I would assume so
<winter><AwesomeAdam54321> If I do `git format-patch` and it adds lots of prerequisite patch IDs, is it better to remove them or leave them be considering that they're not required for the patch? <-- What do you mean by "prerequisite patch IDs?" format-patch should only include patches in the range you specify
<winter><jpoiret> winter: did you commit the change to the build-self.scm file before testing it out? <-- I did not, is that required? I assumed it wasn't because editing config.scm without touching Git caused things to get rebuilt.
<winter><winter> <jpoiret> winter: did you commit the change to the build-self.scm file before testing it out? <-- I did not, is that required? I assumed it wasn't because editing config.scm without touching Git caused things to get rebuilt. <-- I checked, and don't see any Git logic in current-source-directory, so that shouldn't be required, I think?
<AwesomeAdam54321>winter: My workflow if I made a new branch for each addition I made would be to stash my unrelated uncommitted changes, switch, make the commit(s), switch back to HEAD and pop
<Kabouik>Is config.scm the right way to add a user to a group in Guix? I created a "shadow-input" group and have a `(supplementary-groups '("wheel" "netdev" "audio" "video" "kvm" "libvirt" "shadow-input"))` line, which was working so far for other groups listed there, but guix system reconfigure says "error: supplementary group 'shadow-input' of user 'kabouik' is undeclared". I created the group using sudo groupadd.
<AwesomeAdam54321>I haven't tried such a workflow though,right now I just commit all my changes on top of HEAD
<winter>Kabouik: You need to declare the group with user-group first, I'd assume.
<Kabouik>Thanks winter! Is there a documentation on that? I don't know how to use it.1~
<apteryx>the issue would be that the locale package corresponding to the language your system is using gets "copied" or referenced by your IceCat user profile and fails to follow Guix-provided icecat updates.
<Kabouik>Thanks winter. I'm checking it out in the manual. It seems (user-group) goes directly into (operating-system) but I'm having trouble positionning it for now, I'll iterate until it works (:<)
<civodul>apteryx: hi! maybe not related, but i could not get IceCat locales to work, maybe due to stale date in ~/.gnuzilla?
<apteryx>mirai: it starts, but if the language pack referenced by the profile is stale, it could have been garbage collected by 'guix gc' and then the UI could show up mostly blanks
<lechner>civodul / Hi, some time ago we discussed whether 'load' behaves exactly as if the code were in that place in the source file, but that is not so. Some syntax constructs (or something else) in a new Guix service (cachefilesd) behaves differently. Most notably, the define-configuration mechanism does not allow for fields to be set when 'load is used. Is this known, or perhaps worth documenting?
<jlicht>hey mirai! I just read up on your bus-factor notes. If you still need an educational example for a PEG-parser; the .mailmap file as used by git
<jlicht>(If we allow for a simplified expression for parsing "email", at least)
<mirai>jlicht: I wrote a crappy SGML catalog parser using PEG (to generate the missing catalog.xml for docbook 4.1.2 at #61015)
<apteryx>nalaginrut: you could use a --with-git-url transformation if it's commited; if it's not commited you'll need --with-source, or some manifest/package file that does something fancier to avoid copying unnecessary .git files
<apteryx>nalaginrut: but personally I'd simply use 'guix shell -D your-package' and build it "manually" using './configure && make', as that's more efficient if you're constantly changing files/working on your package
<gabber>how can i (successfully) run WireShark to capture packets from my interfaces? wireshark somewhat refuses to be run with sudo and simply adding `wireshark` to my user groups does not seem to be the way to go ("supplementary group 'wireshark' of user 'gabber' is undeclared")
<winter>has anyone ever seen an issue where a substitution successfully occurs, but the next time it needs to be used, it's downloaded again? (but usages after that don't require it to be downloaded again, just the second time...)
<Guest40>I was reacting to " the manual could improve in chapter 12.6 a bit, i guess". To me it seems pretty much fine, but I did read the 12.1 first, so I already saw an example of using (user-group).
<gabber>Guest40: would you agree that it might make sense to add a reference to the (users) and (groups) clauses in 12.6?
<mirai>Anything stopping Guix from using ZSTD to compress the derivations?
<mirai>or just using ZSTD for compression in general
<Guest40>gabber: Probably. Quick (very) glance through other pages in that section indicates that most topics do reference the operating-system field they belong into. mapped-device seems to be only other exception.
<agnem57>Would this channel be the right place to ask for guidance on writing service types? Does there exist other channels that would suit such discussion better?
<Guest40>mirai: I mean, zstdcat is longer then zcat :P
<mirai>Guest40: have zcat intelligently pick what decompressor to use
<Guest40>Currently guix uses gzip right? Is there a comparison somewhere? I would assume zstd to be better compression, with higher CPU usage (for compressing)? I'm unsure what is more constrained resource for the CI (disk space vs CPU).
<civodul>nckx: x.509 certificate expiration notice for guix.gnu.org! how do we handle that? :-)
<civodul>i'm often in the middle of wondering what to do next these days ;-)
<lfam>After not being very active for a while, I've forgotten a lot of knowledge about what kinds of changes trigger rebuilds. Specifically, I want to update git and git-minimal, which does change the derivations of packages using git-fetch. But it doesn't change their output. Will this cause us to rebuild the world? Or is it okay-ish for master?
<lechner>lfam / how about creating a QA jobset for your Git change and looking at the rebuild list for an answer?
<lfam>Yeah, we could do that, definitely. As I alluded to in my message to you about Go, new Cuirass jobsets "should appear in a day or two". Adding them is a manual process that I'm not totally familiar with yet
<cbaines>lechner, I don't think there's such a thing as a QA jobset?
<lechner>cbaines / thanks! just making sure. i did not know what i was looking at. sorry about the confusion
<lfam>This brings me back to a wishlist feature: the ability for a package to forbid run-time references to itself. So you could add a property to git-minimal/pinned like "disallow-references" if a package kept a reference to it, the build would fail
<cbaines>winter, I'm guessing a bit, but I think the change reverted in 8a9bf794e184934e1432f25f4954117d4b46f655? was bigger in scope than this version change being discussed
<cbaines>unfortunately the data service is timing out
<civodul>lfam: do you know #:disallowed-references?
<cbaines>hopefully in the future we'll be able to build all the packages for patches such as this one (currently the limit is 200 affected packages per architecture, so this affecting ~500 isn't a candidate for being tested)
<civodul>cbaines: how 'bout raising the limit (to 1,000?) and adding more hardware behind qa.guix?
<cbaines>civodul, I think that's definately the direction to go in
<winter><cbaines> winter, I'm guessing a bit, but I think the change reverted in 8a9bf794e184934e1432f25f4954117d4b46f655? was bigger in scope than this version change being discussed <-- Right, as discussed above, the pinned Git was introduced to fix this.
<winter>lfam: Are there currently packages that depend on pinned git at runtime, is that why you bring that up?
<lfam>I don't know. But references tend to "sneak in" over time as upstream build scripts and tooling change
<lfam>We want to avoid suddenly being surprised that a package is using a vulnerable copy of Git
<gabber`>mirai: i've added the wireshark group (and am now part of that) but (not so surprisingly) i still cannot capture traffic. should i give `setuid` a try?
<lfam>I can't remember a specific example, but I think it's happened
<apteryx>winter: the reason 8a9bf794e184934e1432f25f4954117d4b46f655 caused lots of rebuilds is because it touched git-minimal/fixed (now git-minimal/pinned), by stripping #t from one of the phases (this was a change of mine)
<apteryx>and git-minimial/fixed inherits from git, so if you touch something else than source you're probably changing it
<agnem>Does anyone have an implementation of a service type for running interception-tools' udevmon with the interception-dual-function-key package? I'm struggling a bit to get store paths properly resolved into my configuration when passed as a string to extend my udevmon service.
<Guest40>Would anyone have a simple example how to build a package with patch applied? What I want to do: I want to take existing guix package, apply few patches (all in .patch files) and build the result. I don't see anything like patches field in the (package) definition, so if someone would have an example how to do this ,that would be great.
<mirai>I don't know if guile supports sexp comments
<Guest74>Guest40 last time I had the same question it was not possible; basically fixing upstream or using string substitutes was the answer
<lechner>well, i might have set the link the other way. does this violate the store layout?
<winter>lechner: Nothing is violated. The interpreter of every Guix-generated ELF is an absolute path to the store's ld.so, but running ldd on the binary uses the host's ld.so, which causes the *displayed* path to be incorrect.