IRC channel logs
2023-10-29.log
back to list of logs
<nckhexen>Hmm, we're missing some input, somehow. It works outside of a --pure environment but not within one. <nckhexen>Ah, I didn't add nss-certs, that's probably it. <jaeme`>nckhexen: I was experimenting with --pure environments, it's very finicky in there. <jaeme`>Which would come first ~rust-findshlibs-0.10~ or ~rust-find-crate-0.6~? <nckhexen>rust-find-crate-0.6 by any rule I'm aware of. <jaeme`>nckhexen: Figured, I'm currently sending a patch series but my email service is ratelimiting me lol. <jaeme`>I forgot to renew my email fooey. <jaeme`>or maybe I'm hitting 'yes' in git-send email too quickly. <nckhexen>Rate-limited mail is a strange concept to me, but you're certainly not the first. <redacted>Is doing `guix system reconfigure ./guix-config.scm' a sane thing to do on a foreign distro? <redacted>Is that just not a thing that would work (I would expect it not to work)? <redacted>(I'm having some trouble getting Guix installed on a VPS, and I'm wondering if I can convert the foreign distro the VPS has installed.) <singpolyma>redacted: if you do that on foreign it will try to take it over. Some bits ofkthe foreign will be left over and ymmv <redacted>When doing a clean Guix install on a VPS, I get errors about mismatched hashes. Apparently things aren't building the way they're expected to. What are some reasons that might be? <Lumine>redacted: if the hash mismatch involves libre-linux-deblob-check, that was patched just earlier today, I had that happen <cnx>I'd like to interest you in the patchset adding pipewire home service <cnx>have been anticipating it for a while now <Kolev>Say I have everything I need installed, and I go a year without updating. Will I be OK? <cnx>as long as you don't connect to the internet <fossquen>How is the $PATH variable defined? Can I add to it in an idomatic way? I have two packages defined in my home-configuration.scm, emacs and htop. The binaries for these two are available at .guix-home/profile/bin, and they run just fine. However, htop and emacs are not callable without a full path every time as it appears that .guix-home/profile/bin <fossquen>is not included in $PATH. Please help. I've been lost on this for a few days. Thanks! <Kolev>jpoiret, I'm using latest to install. We'll see if it works. <lechner>okay, finally the bot is a Guix service on my always-on equipment! <jaeme>Guess that ends my late night guixing <vivien>The earliest job is from September 1st though, are we sure they will all get built sometime? <vivien>apteryx, I lost the URL for the GNOME release engineering team report, could you paste it again? <adanska>ive been lurking on the gnome-teams issue tracker haha! maxim posted it on issue 66689 <adanska>thank you btw vivien for all of your hard work :))) <voroskoi>guix lint complains about tabulation, but there is no etc/indent-code.el anymore, what should I set in emacs to do the right thing here? <vivien>I forgot that the information was so close to me <lilyp>voroskoi: you can M-x untabify <voroskoi>lilyp: maybe i am doing it wrong, but nothing happens <lilyp>unless you highlight the tabs, I'd guess you wouldn't see it <fnat>Has anyone been having issues with EXWM after updating to 0.28 recently? It no longer starts on my system, I'm now in the process of bisecting init.el to see what's going on <fnat>Unfortunately things fail very quietly, no error is raised <fnat>I noticed it yesterday, after a reconfigure; my previous reconfigure must have been the day before yesterday <lilyp>hmm, I'm not acutely aware of grand changes in Emacs within that timeframe, but I don't know every commit either <fnat>Thanks lilyp, I seem to be making some progress with the bisection <fnat>I'll report my findings as soon as I manage to fix it <fnat>So far, I've narrowed it down to the :requires section of my use-package EXWM configuration <fnat>ok, found the atomic change that breaks things; if I add "windower" to ":requires" then EXWM will (silently, with my setup) fail <fnat>nothing to do with EXWM then, I think <fnat>just an unfortunate interaction of various parts, possibly it boils down to me misusing use-package <fnat>anyway - nothing guix related, sorry for the noise <lilyp>fnat you can do windower-autoloads maybe? <fnat>lilyp interesting, what do you mean exactly? <fnat>the way i've been having it until yesterday was: <fnat>a (use-package windower) section, followed by a exwm section <lilyp>when I use use-package for optional stuff, I write :requires <the-package>-autoloads <fnat>the exwm section "required" windower via the :requires keyword <fnat>the change that might have broken things is me adding a keybinding section in (use-package windower :bind ...) <lilyp>yep, that delays the initialization <lilyp>I think you can force it to be eager tho <fnat>super, i'll look into that <fnat>something that the autoloads thing might help with? or is it unrelated <fnat>sweet! this helps massively, ty so much lilyp <lilyp>it might help; instead of checking that the feature is loaded, you just need the package's autoload <lilyp>which guix already processes before init.el <fnat>oh and (again guix unrelated) is there a way to make emacs more talkative in cases like duthils <lilyp>you can --debug-init, but that probably won't work well for exwm <fnat>yeah, no i tried that and it didn't help <atuin>'(with-store %store (run-with-store %store (lower-object s)))' I am trying to create a derivation for a service (s there is a service) but i get gexp-input-error. What's the way to derive a service? <lilyp>atuin I suppose you have already tried easier stuff like gexp->derivation? <atuin>ohhh I didnt indeed :D, let me try <atuin>but the service is not a gexp right? <atuin>yeah, that throws a wrong type argument, I guess i am missing some step in the middle to get from service to the gexp <tux1c>has anybody here successfully built %hurd-default-operating-system from gnu/system/hurd.scm? <tux1c>im trying to simply execute `guix system vm config.scm`, config.scm returns %hurd-default-operating-system, but guix complains about "'term-tty6' requires 'host-name', which is not provided..." but a host-name is obviously provided in the default configuration <fnat>Anyone who has special access to guix-patches and that could help me debug an issue where my emails to the list seem to be dropped <fnat>I know there's a moderation-related delay, but I don't think that explains it as some email never shows up on the ML <fnat>I suspect new email addresses may be gray-listed and moderated separately, but this is an address known to the list already <fnat>I tend to think I might be doing something wrong with git-send-email, but then again my msmtp logs would seem to indicate that the email has been sent successfully <fnat>Unfortunately I don't have direct access to the SMTP server (as in access to the logs) <nckhexen>fnat: The moderation delay generally affects only new addresses (and separately, for each list). All moderation queues are currently empty. If you have a copy of a message that never arrived, we can ask the FSF sysadmins about it. I don't think any Guixer has privileged access to the gnu.org mail server. <nckhexen>Or even unprivileged access for that matter. <efraim>jaeme: the rust-team branch has rust 1.73 <fnat>nckhexen: yeah, thanks, I'll try and contact the ML sysadmins then - I noticed I'm actually not subscribed to guix-patches, so that might make it more likely for me to be caught by some antispam... <nckhexen>(I don't think so, but then nor do I have an answer.) <fnat>yes, I have the original message and the local msmtp log <nckhexen>Can you share the message with me? At least I can check whether the address hasn't been accidentally blacklisted in Mailman. <nckhexen>Re: ‘try’: I'm happy to interface with them if you'd rather not or suspect you'd have technical troubles. <fnat>nckhexen: hey thanks, that's very kind of you, I've actually sent them an email just now (them = guix-patches-owner@gnu.org, as found on the ML page); of course I'm happy to share the email and the various bits of info with you - shall I DM you here or do you prefer email <johnwind>trying to use bcc, but python just couldnt import it, why is that ? <vivien> Purism is doing a lot of things with source-only projects that are used as git submodules <vivien>What should we do with that? Package the projects, invent new install rules to install the source? <vivien>Or use the project tarballs with submodule included and not worry about it? <cbaines>vivien, regarding the log data.qa.guix.gnu.org queue, it is going down, but quite slowly <cbaines>there's plenty of things to do too speed up processing jobs, but I disabled computing derivations system tests this morning, and that should save around an hour per revision <lilyp>vivien: there are some packages like gnulib that should give you a hint as to how you can reuse them <vivien>lilyp, the purism tarballs include them by default, and their project says they should be used as submodules <vivien>feedbackd is the project *using* submodules, and libcall-ui is an example of submodule <vivien>feedbackd -> gmobile, calls -> libcall-ui <lilyp>there used to be an example of this in gnome itself, but it seems to have gone <lilyp>found it: gnome-recipes still has one <redacted>> You found a bug: the program '/gnu/store/[hash]-compute-guix-derivation' failed to compute the derivation for Guix (version: "[hash]"; system: "x86_64-linux"; host version: "1.4.0"; pull-version: 1). <jpoiret>redacted: there should be a path to a log file <redacted>There is none. I've double checked that I didn't miss it. <redacted>Might not matter if I can just upload an image to the VPS. <yaslam>Hello, I am trying to create a package and I need to use propagated-inputs, but it does not install anything into the system profile, am I using it wrong, thanks. http://ix.io/4Kgl <nckhexen>Description of the week award. Aside: you can, and I would, write (replace 'build (lambda …)) instead of (delete 'build) (add-after 'configure 'no-wait-build-anyway (lambda …)). <nckhexen>I'm going to have to do my annoying ‘define ‘does not install’ please’ thing again. Your code looks fine. <ryblade>grub is stuck at boot after upgrading the kernel to 6.5.9, after selecting guix in the grub menu the menu goes away but the background graphic remains. been that way for several minutes. anyone know what could be wrong? <nckhexen>I don't think it's stuck, just unable to display. <yaslam>nchexen: Thanks for the advice, I have changed it. I have no idea why the program doesn't see qtsvg. <ryblade>it still accepts ctrl-alt-del to reboot so it isn't proper hung <nckhexen>Then it's likely just waiting for your passphrase. <ryblade>and it's a thinkpad, interesting. not the same model as the issue, but still... i'll try entering my passphrase <nckhexen>yaslam: I'm not familiar enough with Qt to know, but propagation is seldom the answer. I suspect that this is something that is normally handled by a qt-build-system phase. <ryblade>yeah, haven't done anything custom with the fw <nckhexen>OK, that's definitely new. A regression either in upstream Linux or our configuration of it. <ryblade>ran fwupd in another live distro so it's up to date, that's all i've done with the firmware since buying this <nckhexen>And an indication that my efforts to add back Coreboot framebuffer support were not relevant to fixing this issue. <ryblade>drat, sorry if this makes more work for you <nckhexen>Eh, I *wish* I had more work, because that would imply I knew where to start & what to try :-/ <nckhexen>I have a ThinkPad that would probably be affected, but it would take some effort/time to install a stock Guix + Linux-Libre to find out. <ryblade>if there's any info i can grab off this T480 for you that would help, let me know <Altadil>nckhexen: I’m also seeing this issue, let me know if there is any info or testing I can provide to help. <nckhexen>ryblade: Could you post a quick note to that bug (mail to <bug#> at debbugs dot gnu dot org) mentioning your model & fact that it's not corebooted? Sounds like it might be useful. <ryblade>nckhexen: i have my email set up on another machine, but i will boot it up soon (as in today) and get on that for you. <ryblade>is it also weird that i'm asked twice for my password? /boot is unencrypted on my system (the rest is encrypted) and i'm using /boot (rather than /boot/efi) as my ESP. <ryblade>if that isn't expected behaviour i will mention it in my email <nckhexen>yaslam: Stepping back a bit: ‘guix shell --pure -f /tmp/4Kgl.scm -- prismlauncher’ doesn't complain about a missing QtSVG. It draws cats. SVG-looking cats. Although I didn't check whether they are. <nckhexen>ryblade: Nah, that's expected. Being asked only once (and it working) would be weird. <nckhexen>Guix System does not currently support booting unencrypted files, not without manual scripts/modifications. <nckhexen>(You'll notice that your ‘unencrypted boot’ doesn't in fact provide anything needed to boot the OS, i.e. the kernel & initrd.) <nckhexen>GRUB needs to decrypt your /gnu/store to load them, then later Linux needs to decrypt the same thing to mount it as /, and there's no sharing of secrets between the two. <ryblade>right on, i noticed on guix the kernel and initrd are in the store and not copied to /boot as some other distros do. <nckhexen>Some of us do so ‘manually’ (scripted), but I think we all use ad hoc scripts. <nckhexen>(GRUB can't read my /, so I'm in a similar situation, even though it's not encrypted.) <yaslam>nckhexen: the issue is that certain icons are missing in settings and when you click "Add", and when I did "guix shell qtsvg@5" then the icons would come back. <nckhexen>ACTION disappointed by bitmapped cats. Cats are vectorial by nature, everyone knows this. <nckhexen>yaslam: I'm trying a version with cmake & qt build systems smashed together, but my laptop is heavily loaded. <weary-traveler>hm is there a way to disable 'check' build phase on some of the dependencies when building a project using guix? <nckhexen>You can define a package variant with #:tests? disabled (either by using inherit, or using package transformations, see the options->transformation example in the manual). Note that an untested package differs from a tested package and will have a different hash, affecting all its dependents. <nckhexen>It's still weird having a paste site use the correc^W^Wmy preferred colour scheme. <nckhexen>yaslam: Welcome back. Borrowing from qt-build-system did the trick, I think. <sneek>Welcome back yaslam, you have 1 message! <sneek>Welcome back peanuts, you have 1 message! <vivien>lilyp, gmobile does not have meson rules to install anything, so I can’t use it as a stand-alone input for feedbackd <NewGuixUser>rekado I've tried a few different commits to see if there's one that fixes the matplotlib issue and haven't had any success. Do you think writing my own package for matplotlib and setup_scm_git_archive would help? <vivien>Should I add rules to pack everything in a static library, and install the library and the header files? That seems to be quite a stretch <vivien>Or maybe I should add a rule to install the whole source tree in /share? <nckhexen>(I'm assuming there's a reason you can't just use (package-source gmobile) instead, which sounds preferable.) <lilyp>A shared library would be preferable to a static one <lilyp>I sadly can't inspect the source without pulling it (thanks, Gitlab) <lilyp>thus I don't know which ones need to be submoduled and which ones don't – hence the generic statement <lilyp>still, it sounds like upstream should put in some work to make their subprojects proper subprojects :) <vivien>Did I AGAIN send a WIP patch that I did not try to build? <lilyp>vivien: It's fine, I had a little bit of fun getting it to work halfway <yaslam>nckhexen: I have a question, why does the version for the inputs use dash and not @? Thank you. <yaslam>Like for example the input is qtsvg-5 but when you install it you use qtsvg@5 <nckhexen>Ah. ‘Convention’, AFAIK. There's no reason you couldn't (define qtsvg@5 …). <sneek>peanuts, you have 1 message! <sneek>Welcome back peanuts, you have 1 message! <nckhexen>RavenJoad: Legalities-wise (is that what you meant?) I don't see anything missing. Although you do switch between Raven and Karl in different places. <graywolf>Hi, quick question, what is a good way to basically use (local-file), but replace specific things with store references? I basically want mixed-text-file, but for a amount of content that I would prefer in a standalone file. <RavenJoad>nckhexen: Yeah, legalities was really the important thing. I must have missed my name in a few places. <voroskoi> (service home-files-service-type `((".tmux.conf" ,(local-file "tmux/tmux.conf")))) <voroskoi>this puts .tmux.conf in ~/.config/.tmux.conf, but i would like to put it in ~/.tmux.conf <efraim>home-files-service-type goes straight in ~/ home-xdg-configuration-files-service-type goes in ~/.config/ <voroskoi>efraim: it does not for me, trying again <the_tubular>Anyone is able to run exwm inside a nested wayland session ? <rekado>NewGuixUser: I haven’t had any time to look at this today, but I’d investigate why matplotlib insists on using setuptools_scm_git_archive and why it claims that it isn’t there. <lucypoo>hello there, has anyone using `package-input-rewriting` encountered an issue where it seems to have no affect before?, maybe a clear of the cache but i am hesitant to do that currently as a lot is riding on my cache and a lot seems dependant on the store item i wish to rebuild <lucypoo>providing a name transformation function seems to trigger a rebuild of the software stack, maybe will resolve, but will be a long time <NewGuixUser>rekado: I manually deleted setuptools_scm_git_archive's drv and tar.gz from my /gnu/store/ and guix now complains that it can't find the drv when trying to build matplotlib. So that seems to indicate to me that it was looking for it. I am a little puzzled why it didn't recreate the files though when it saw they weren't there. <rekado>oh… you shouldn’t ever delete stuff from /gnu/store <rekado>if ever you need to remove stuff from there use “guix gc -D” <rekado>the assumption is that the only thing to ever add or delete stuff in /gnu/store is guix-daemon <rekado>it has a little database that must be consistent with what’s in /gnu/store <rekado>you can try to repair your /gnu/store with “guix gc --verify=repair,contents” <lucypoo>nckhexen just in reference to items in /gnu/store <lucypoo>i guess so, but it seems the behaviour is somewhat undefined as `package-input-rewriting` is not functioning as expected <nckhexen>Deleting random unrelated files won't fix the bug. <nckhexen>What are you trying to do, and how, and what's the expected effect? (I don't see it in the scrollback, but do say if it's there.) <lucypoo>i define a new python version, then use `package-input-rewriting` to create a function which replaces the old python version with the new, then i apply this to the `python-setuptools` package <nckhexen>The bug will be in there, or in the assumptions it makes. <lucypoo>however the store isnt unrelated, i was considering the deletion of the exisiting built package <nckhexen>I have to start an hour-long drive, but I'll be back. <nckhexen>But to your point: removing store items ‘to force a rebuild’ is a (bad) habit from other distributions. Guix has many bugs but boy does it know how to hash things and put them in the same bucket each time 💯 <lucypoo>ahhh, so assembling code, one thing of note is that it works when using cmd line package transforms <vivien>lilyp, I’m trying to debug the sip test error in gnome calls. The test suite fails, but if I open a shell in the kept failed directory, and set up the environment variables, then I can build the test fine. I see one difference: the test uses 192.168.1.*** in the shell, and 127.0.0.1 in the build. <lilyp>vivien: hum, there appears to be at least one networking test among them then; you won't get networking to work in the build container <lilyp>check if there's a flag to disable those; otherwise try to identify the one that requires networking and disable that <NewGuixUser>This is bizarre. I've done a clean build and see matplotlib still complaining about setuptools_scm_git_archive. When I grab that derivation from the matplotlib derivation I see that it does exist and was built in my /gnu/store. The last thing I can think of is that somewhere along the line a version isn't being specified properly because I see <NewGuixUser>"ERROR: Could not find a version that satisfies the requirement setuptools_scm_git_archive (from versions: none)" in the matplotlib build log. <lilyp>NewGuixUser: ahh, that's a nasty one – which commit are you at? <NewGuixUser>lilyp: I'm using v1.4.0 on master and 1e25b23faa6b1716deaf7e1782becb5da6855942 from guix-past to get python3.8.5 <NewGuixUser>I've tried changing up the commit on guix master and it hasn't helped, so it must be related to pulling in an old python <lilyp>hmm, you could try ((package-with-explicit-python (delay python-from-guix-past) "python" "past-python") python-matplotlib) <NewGuixUser>lilyp: Yep, I've already got a function that does that courtesy of rekado <lucypoo>ooh i just stumbled upon the explicit python func <weary-traveler>is there a way to tell 'guix build' to ignore 'check' phase of some/all dependencies? <lilyp>NewGuixUser: you might for whichever reason be missing python-setuptools-scm as well <NewGuixUser>weary-traveler: If you look at the package-with-explicit-python function in my paste right before your message there's an example of deleting the check phase with a transformation <NewGuixUser>lilyp: I don't think so. Based on matplotlib's drv it's looking for "/gnu/store/jblkvvmk1mc1b26b4imwifvm3nrbhr10-python38-setuptools-scm-6.3.2.drv" which I have and I also have the corresponding build directory <NewGuixUser>so python38-setuptools-scm has a "/lib/python3.8/site-packages/" which contains code, however python38-setuptools-scm-git-archive-1.0 has a "/lib/python3.8/site-packages/setuptools_scm_git_archive/" which only contains __init__.py and __pycache__ <NewGuixUser>so maybe setuptool_scm_git_archive never actually got built <weary-traveler>NewGuixUser: i see. so will have to define a package transformation that selectively deletes 'check from #$phases as you have. will need to figure out some guile i suppose. thanks <weary-traveler>i did post a message on help-guix earlier today regd this, but i'm guessing it's waiting in queue to be reviewed since it was my first post to the mailing list <NewGuixUser>or maybe that's the intended behavior? I just did a normal "guix install" of setuptools_scm_git_archive and get the same empty directory, just for python3.9 instead of 3.8 <attila_lendvai>nckhexen, did you have a chance to look whether you can find your luks/cryptsetup --allow-discard patch for guix? <lucypoo>nckhexen: so i wasnt replacing the python used by the python build system, so it seems `( options-transformation '( ( with-input. "python=python@3.12.0"))` replaces the build system python and `( package-input-rewriting '( ( "python" . python-3.12)))` does not <f1refly>what would be the best way to automatically mount an sshfs automatically no a laptop? <f1refly>I have a hetzner storage box and would like to have access to it from my laptop, but it should be able to notice if its offline or online and mount/unmount the sshfs appropriately <NewGuixUser>So I can use time machine and install matplotlib just fine, it's only when I try to substitute in python3.8.5 that it breaks. <kone>GUIIIIIIIIIIIIIIIIIIIIIIIIX. <lucypoo>NewGuixUser: perhaps youre issue is similar to mine, i could only replace a python version with options->transformation <lucypoo>as it needed replacing in the build system <lucypoo>`package-with-explicit-python` wasnt working for me, but perhaps i was using it wrong, had to copy the def into my code as it is private in guix/build-system/python.scm <NewGuixUser>Several iterations ago on this manifest I had been using with-input to replace python, but was running into issues that needed package-with-explicit-python to be solved. I guess I'll try using both at once and see what happens. <rekado>NewGuixUser: I’ll try patching matplotlib next <rekado>the problem is not with setuptools-scm* but lies deeper <rekado>it will also fail to find numpy when given the opportunity <NewGuixUser>rekado is this what you're trying to patch? "(setenv "SETUPTOOLS_SCM_PRETEND_VERSION" #$version)" <NewGuixUser>I wonder if that version isn't getting set properly so it isn't being found <rekado>we only see this problem with python 3.8, which includes a certain version of setuptools, which may behave differently from that in 3.9 <rekado>that env var is only set to make setuptools-scm work fine even though we don’t have a .git directory <elevenkb>is there any way to step through how a simple package like hello is built? <sneek>Welcome back peanuts, you have 1 message! <NewGuixUser>rekado So is the thought that the provided setuptools with 3.8.5 is incompatible with the setuptools-scm and git-archive plugins provided in v1.4.0? I'm reading up on setuptools now to try to get a better understanding of how these pieces all fit together