IRC channel logs
2024-01-25.log
back to list of logs
<apteryx>jackhill: does it not share its version with webkitgtk? <apteryx>it could, but it uses its own source, for $reasons <apteryx>we should probably have it inherit it directly from webkitgtk <apteryx>lilyp: I doubt berlin takes 10 h to build webkitgtk. I'd expect more like 20 min <apteryx>not sure why it still hasn't happened though :-/ <apteryx>ok, it's doing it for the last 600 s <bienjensu>Is there anything special I need to do after sending a patch for the first time? <mange>I don't think so, but your first posts to a list might be held up in a moderator queue for a bit. Is there a particular reason you're asking? <Gooberpatrol66>why does guix download a huge number of packages (~7G) when i run guix home reconfigure, and then delete them when i run guix gc? <Gooberpatrol66>either they're unimportant and shouldn't be downloaded, or important and shouldn't be deleted <mange>First guess: grafts? Does it print a bunch of messages about grafts while it's working? <mange>One way to test it would be to see if running the reconfigure with --no-grafts ends up downloading less (at least, in the sequence "reconfigure, gc, reconfigure"). That might mean running some more outdated libraries during the test, though. I wonder whether the --gc-keep-outputs and --gc-keep-derivations daemon flags would also help save the redownloads. <adanska>Hi Guix. I'm having some trouble with the asdf build system; I cant seem to get the tests to run. <adanska>theres a file called 'lem-tests.asd' in the tree that should be picked up... and yet nothing happens <adanska>:p i guess the tests are just not covering what i want <rekado>Gooberpatrol66: can you give us more details? The behavior depends on the packages. Some tools are needed for processing build hooks, but 7G seems a little excessive even for that. <efraim>I think I have rust fixed on core-updates, just waiting for the build to finish <janneke>and even better, the rest of the rust stack is unaffected :) <efraim>I'll probably cherry-pick that commit to the rust-team branch so I don't have to worry about the merge conflict later <efraim>I'm taking a look at clang-runtime-13 now <efraim>clang-runtime-14 fixed, now on to 13 <efraim>I'm going to go down the line to see the earliest clang-runtime I can fix with just that patch and then push the updates <sneek>civodul, you have 1 message! <sneek>civodul, rlb says: any idea offhand why the recent signal thread changes might cause the standalone tests to hang on test-guild-compile with the parallel test harness? May be repeatable for -j4 and greater here... <efraim>rust is finished building on berlin <janneke>ACTION lost again racing the build farm :) <efraim>janneke: do you have ssh access to the build farm? <civodul>i was looking at evaluations that don’t terminate on berlin <civodul>one reason was that several were waiting on gnutls-3.8.3 and for some reason that build didn’t complete and the lock was held by one of the ‘cuirass evaluate’ processes <civodul>preventing others from making progress <civodul>specifically it’s /gnu/store/r6903q7vsapjmg3f8adisx0bkq5ackv9-gnutls-3.8.3.drv (aarch64-linux) that failed to complete <civodul>let’s see if it works better this time <janneke>efraim: ah yes, only access to overdrive seems to have been revoked/removed <janneke>ACTION has problems building qemu-minimal without doc <janneke>it builds but then fails with something like builder for `...qemu-minimal-8.1.3.drv' failed to produce output path `/gnu/store/xxx-qemu-minimal-8.1.3-doc' <efraim>do you have a phase that needs to be adjusted? <janneke>i removed 'move-html-doc, but it even `compress documentation' succeeded already <janneke>it's as if (outputs ..) just won't be overridden, adding a mkdir ../doc instead as a workaround for now <efraim>It looks like ldc currently only builds on x86_64. I'm going to see about moving it to clang-15 so we can keep our "standardize on the odd versions" schema going and reduce compiler proliferation <efraim>I'll see about qtwebengine@6 later <jpoiret>apteryx: yeah, I saw that pkgconf had some upsides. I'm worried about compatibility though <jpoiret>but also pkg-config and static outputs don't mesh very well i'd say <efraim>well that's fun, imagemagick and imagemagick/stable produce the same derivation on core-updates so inkscape fails to build since inkscape DOES refer to imagemagick, and therefore also imagemagick/stable <janneke>efraim: i was just wondering about that <lilyp>Looks like imagemagick isn't that stable on core-updates :) <efraim>I'm going to modify imagemagick/stable to not install the documentation, and I guess remove some inputs <efraim>nope, nothing to see here. all is broken <efraim>14 hours for rust-1.75 on riscv64 on the rust-team branch <efraim>imagemagick and imagemagick/stable are now two separate derivations <efraim>and inkscape and inkscape/stable both still build <efraim>well, now inkscape builds, inkscape/stable always built <jobol>hi, how to have a different place than /gnu/store? symbolic link refused. A mount bind might be possible but can I avoid it? <rekado>jobol: using a different location other than /gnu/store means that you cannot get *any* pre-built binaries from our servers. <rekado>you’d have to build *everything* from source, from the ground up <rekado>jobol: one of the reasons is that binaries embed the absolute file names of libraries and other files <rekado>so if you wanted to use a different location you’d have to either rebuild everything (so that your preferred location is embedded) or rewrite *every* file’s references to /gnu/store with a location that is of the exact same length. <rekado>if you *cannot* write to /gnu/store you may want to look into running Guix inside a separate mount namespace. <rekado>this means that all your Guix-installed packages would *also* need to be executed inside of this separate mount namespace. <rekado>another way is to use “guix pack -RR” from a machine where you *can* write to /gnu/store and transfer the generated bundle over to a machine where you don’t have Guix (and thus also don’t have /gnu/store) <lilyp>maybe someone with a cloud to spare could run Cuirass preconfigured to have /opt as the store path <lilyp>As we all know, /opt is the right place for everything :) <efraim>I'm giving up on libetonyek on core-updates for now <janneke>efraim: what about libical, i'm getting: error: in phase 'build': uncaught exception: <efraim>I haven't gotten to libical, right now I'm working on node-10 <janneke>ok, i'll see if -keep-failed will produce more info <efraim>looks like I fixed the test, version string mismatch with zlib being 1.3 and not having a major.minor.patch version string <janneke>well, with --keep-failed it built, eh, "yay" :-/ <janneke>ACTION also isn't a big fan of brittle software and transient errors <efraim>I forget the package, I had one where I had a cached failure one way and a cached failure if I marked that test as XFAIL <janneke>client/bluetoothctl.rst:342: (WARNING/2) Cannot analyze code. Pygments package not found. <janneke>it was weird, beceause there was already a python-docutils as a dependency <efraim>I wonder if you only need to add python and not also python-pygments <efraim>oh, now I'm back to python-urwid :/ <efraim>I'll do that later, time for lunch <janneke>ACTION tries with removing python-pygments <janneke>hmm, `guix shell' doesn't work in my fresh core-updates hurd vm :( <janneke>root@guixydevel ~/src/guix/hurd-team# guix shell <janneke>guix substitute: error: times: Invalid argument <janneke>substitution of /gnu/store/ygqc1k3yxrsmw9qr5z04b26lamkfyk2a-bash-static-5.1.16 failed <rekado-web>janneke: looks like your commit to maintenance.git broke authorization <rekado-web>your key has to be added to the authorizations file first <rekado-web>I'll force push after the deleting the commit, then add your key, and restore the commit <janneke>ah, i figured my key was already there... <janneke>rekado-web: thanks, i'll look at your fixed commit <rekado-web>I wonder if I can delete the branch and recreate it. <hy-im-sofie>hello everybody, i'm not sure if my question belongs to this channel. Please let me know if this is not the place for such a question. I'm new to guix and i want to write a system definition for my raspberry pi. For me this would need to include some flatpak apps and i don't find information on how to code this into the system definition. I'm able <hy-im-sofie>to setup flatpak and install apps by hand but i want to solve the complete setup of my system inside the definition. Is someone willing to help a beginner? <rekado-web>there is no integration between Guix and flatpak. I don't think we offer a flatpak service that automates installation of flatpak applications. <lilyp>Is there a way of using match-record to select a package from a list by name or should I rather code around that? <hy-im-sofie>Maybe this is the reason i can't find any info on how to to such a thing. In the past i used a bash script to setup apps, preferences, keymaps and so on and i just assumed this would be possible using guix/guile as well. <civodul>lilyp: hi! how do match-record and package lists relate to each other? :-) <lilyp>it's a list of packages and I'd use filter with a match-record-lambda if that's possible <lilyp>sorta like (match-lambda ((= package-name needle) #t) (_ #f)) <civodul>this would bind ‘needle’ to the package name, so not quite what you want <civodul>better to just do (find (lambda (p) (string=? (package-name p) needle)) lst) <lilyp>oh, no worries, I'd actually hardcode the needle <snape>Hi, what is the use of a local channel? <snape>To me a channel was meant to be shared but I've seen something about channels for local checkouts in the list <snape>is it for more than just testing? <lilyp>snape A channel is best shared, but you can keep it to yourself as well. In that case, it's just the same as a local git repository of package definitions :) <snape>lilyp, yes but what's the point when you can just add guile code in your, say, config.scm <snape>why would anyone want to have to "guix pull" their own channel <lilyp>it may boil down to preference and separation of concerns <snape>I believe some newcomers think setting up their own channel is the usual way to run their own package code <snape>(like the one I'm trying to help on the list) <snape>but the reality is much simpler <ennoausberlin>snape: I still use a private channel for my packages. I started that way, but I have to switch at some point in the future, because pulling can be very time consuming, if there are a lot of changes in the guix main channel at the same time. <snape>ennoausberlin, so what I do is I just add my own packages in my home.scm/manifest.scm, works very well <snape>and not time consuming at all <ennoausberlin>snape: I have to see the syntax to adapt this way. Right now I use define-public, I guess it was different for manifests <snape>ennoausberlin: you can use define, no need for define-public <snape>and then: (home-environment (packages (append my-packages (specifications->packages my-specifications))) <snape>my-specifications is a list of strings as you would see in a usual manifest.scm, and my-packages is a list of packages (as in "(define package ...") <ennoausberlin>snape: I will try to activate my scheme-fu, but I fail often on debugging. Thank you <snape>ennoausberlin, and in the manifest.scm, which starts with specifications->manifest, well you just need to transform it into (packages->manifest (append my-packages (specifications->packages specs))) <ennoausberlin>snape: I used manifests before to setup for a given use case, but I never used own package definitions inside it. But I will give it a try <apteryx>re inkscape, we should use this opportunity to update it <apteryx>perhaps we need to comment the disallowed-reference for now <apteryx>and add a comment about re-enabling it when 'imagemagick' is bumped <apteryx>ACTION fixed the rust phase ordering problem <apteryx>jpoiret: if RHEL 8 is on pkgconf, I doubt it'd be too painful to migrate to it <apteryx>just a pkgconf-as-pkg-config, with a bin/pkg-config -]. bin/pkgconf should work <apteryx>jpoiret: but perhaps the low-risk, easy change would by -Ddefault_library=shared passed in default configure-flags of meson-build-system <apteryx>then generated .pc file would not have a ton of libraries in Requires.private <user363627>When installing guix on a foreign distribution (via the shell installer script) where are the manpages installed? <user363627>It seems with the binary installation i get the info pages but not the manpages. If I do "guix install guix", then I get the manpages as well, but the version of guix I get via " guix install guix" is different from the one available to the root user <sneek>I think I remember nckx in #guix one month and 1 day ago, saying: Happy Christmas all y'all.. <distopico>Hi, someone kwnos hoy can I translate this `options snd-hda-intel id=HDMI index=1` into kernel parameter, I know that with a single options is like `snd-hda-intel.index=1`, but what about 2 options, is possible `snd-hda-intel.id=HDMI.index=1` ? <apteryx>it looks like tE CSS doesn't refresh on a page reload there either <apteryx>I used the timestamp trick on the file name <apteryx>ah, I think the javascript has the same flaw (cached to old version) <apteryx>so the button doesn't currently work <rekado>apteryx: works for me. Good job! One problem: the button can only be clicked. It cannot be navigated to by tab, as it is not a real UI element. <apteryx>that's also true of other buttons, I think? <rekado>no, the message download button is a link, so it works out of the box <rekado>you can add a tabindex=0 to the new button to make it selectable <rekado>you’d also need to make it possible to trigger it via keyboard, e.g. to respond not only to a “click” event. <apteryx>OK! I'll test this and send a v3, thanks for the feedback <janneke>ACTION just built a fully functional hurd vm (or so it seems for now) from guix, that's with glibc-2.38 <apteryx>efraim: I've pushed a refined rust fix; I had a test failure I needed to skip (it used the publish procedure) <janneke>wait, we don't have pretty-printed builders, it's emacs doing! <apteryx>janneke: ah, I thought you had contributed that change or someone <apteryx>emacs-guix is able to do that yes, but I don't think it does so by default for -builder files <janneke>apteryx: "me didn't do or configure nothing" <janneke>i just happened upon pretty-printed builders, while walking through .drvs <apteryx>their extension is registered by the mode <apteryx>rekado: any idea what's the event corresponding to hitting the enter key on the button? <apteryx>no, it works on the top message, but not the others, for reasons <apteryx>ah right, I need to register the events for every object <apteryx>since there can be more than one button <apteryx>I want to use a class, not a unique button id <rekado>janneke: yay! Thanks for repairing the hurd vm! <weary-traveler>i'm trying to debug the test environment for a package (i.e., the environment within which the tests for a package are evaluated). trying to "echo $var" in the test-command simply omits '$var' (as opposed to its value). how do i inspect the environment variables? if i have a way of doing something like a printf that outputs to the build log it'll help <janneke>rekado: yes, i got quite some help from apteryx and efraim on core-updates too <janneke>rekado: did you get maintenance sorted out? i'm still a bit puzzled by the breakage, the commit that removed my account changed nothing else than what i put back... <weary-traveler>okay the info page on debugging build failures has some relevant information <rekado>janneke: I think we may need to ask the savannah admins for help here. <rekado>I pushed a new branch “master2” for now. <janneke>rekado: just reverting my commit won't help? <janneke>ah, .guix-authorizations; the commit in itself was OK -- i get it <janneke>crap, i missed 56d25466656ae699915b297ee9c8776decfddbf2 <janneke>(and failed to enable the commit hook on my "new" machine too) <apteryx>the commit hook is enabled by just 'make' nowadays <rekado>janneke: no worries. Since it’s only the maintenance repo, a hard revert is not quite as tragic <rekado>I just don’t know who exactly to talk to in order to get this fixed. <janneke>rekado: maybe you can ask on #savannah here? <Groumf>Hello guys. I don't really get the use of use-package-module in system definitions. The classic example from the documentation shows it with screen. I don't see the benefit of putting screen there. <Groumf>gnu.scm seem to indicate it uses package-module-hints. But I see none of these in the screen.scm <ulfvonbelow>(use-package-modules screen) is equivalent to (use-modules (gnu packages screen)), it just saves you a bit of typing <ulfvonbelow>in the manual's example, it's present because it has to be loaded in order to get the package bound to the 'screen' variable <ulfvonbelow>yep, pretty much an include, although IIRC it will also prevent duplicate reloading <Groumf>damn I read a lot about guile at this point but I can't say I am comfortable with it yet <sham1>I really wish that there was a Guix Home service for configuring user git settings. Having to configure the stuff the "old-fashioned way" just feels sad when other stuff can be done with Scheme code. Oh well, that elbow grease can be used later on <ieure>sham1, Seems like a pretty reasonable project if you want to pick it up! I just dumped my git config into SCM next to home.scm and have it install the file. <sham1>Yeah, I'll have to look into it at some point! <sham1>If nothing else, it'd help to be able to use Guix at work as well as at home <civodul>you can chime in in that thread an say how much you’d like to see it happen :-) <sneek>I think I remember iyzsong in #guix 15 days ago, saying: apoorv569: likely '(default "")'. <civodul>sham1: maybe you can adopt the patch and send a new version there? <apteryx>rekado: i'm a bit confused by my data-tooltip state getting corrupted when I repeatedly click the button quickly <apteryx>how are multiple click events handled on an html element? do they replace older ones? do they run "concurrently"? something else? <apteryx>(in my handler, there's a 3 s timeout that reset the tooltip to its original value) <apteryx>if this was true what I'm seeing shouldn't happen <apteryx>ah, setTimeout breaks the atomicity of a message <PotentialUser-37>Hi there! in a container what is the machine-id? There is no /etc/machine-id file... <PotentialUser-37>Problem solved! I guess I can just expose the /etc/machine-id of my machine :) <bumble>I would like to get the equivalent store directory path inside a scheme function, how can I do this?