IRC channel logs
2023-03-12.log
back to list of logs
<bjc>i forget, does guile have tco? is that why (let loop …) is so popular? <two[m]>tco is required by scheme standard <bjc>i thought i remembered guile having some issue with it because of c interactions. glad to be wrong, though <gnucode>jpoiret: the last time I asked about on the mailing list about guix one day supporting the OpenBSD kernel, I was told that supporting multiple libcs would be hard. <sleepycat>what about the possibility of stripping away guix's dependency on specific glibc features that don't exist in libc? <bumble[m]>would someone here recommend a command for uploading files somewhere to easily share? <bumble[m]>here after doing guix pull and sudo guix system reconfigure, guix home fails always with "failed to load /path/to/guix.home.scm" does anyone know why this might happen? <bumble[m]>the full and correct path to the guix home scm file was given <Guest40>Having guix on mac would be nice, however I get the technical difficulties and that it simply is not a priority :) <bumble[m]>I see there was an error in the home file... but guix home reconfigure command does not surface this and instead prints file not found <bumble[m]>using "(build-system gnu-build-system)" did seem to install the font file because my system is able to render thai characters using the font, however I don't think it is correct as the font does not appear in ~/.guix-home/profile/share/fonts <bumble[m]>I'm interested in completing this as a learning experience. How would I see what was generated by ./configure and make when installing the package? for developing a package like this does one need to install fontforge and other tools in a sort of test environment? how should one follow what the package install sequence is doing? <mirai>bumble[m]: how did you install this <bumble[m]>without knowing anything else perhaps one could install fontforge then run ./configure, make and other commands ... and use what is learned there to update the package scm file and re-install until the fonts show up in... <bumble[m]>@mirai "guix package --install-from-file=guix.pkg.fonts-thai-lwg.scm" <mirai>are you sure the thai script you're seeing is using that font? <bumble[m]>tbh I'm _certain_ about... I think the font is being rendered by qutebrowser and I can try to use qutebrowser's web-inspector to see which font-family is actually used <bumble[m]>I'll try to find out specifically what qutebrowser inspector is showing <bumble[m]>the reason I thought it was using the font is, thai characters appeared to look nicer after installing the package, probably I confused <mirai>does this render to you? --> สวัสดีครับ <bumble[m]> * tbh I'm _not certain_ about... I think the font is being rendered by qutebrowser and I can try to use qutebrowser's web-inspector to see which font-family is actually used <bumble[m]>it will take me a little bit of time to load the element in qutebrowser inspector... one moment... <bumble[m]>inspector shows "font-family: Roboto, sans-serif" <iyzsong[m]>you need switch to the combine tab, see the final font used <bumble[m]>@iyzsong thank you okay adding more info here... <bumble[m]>in the "Computed" tab, with the element focused, the bottom of the tab shows: <bumble[m]>waree is one of the fonts provided by that package <iyzsong[m]>it may be Waree, fc-match -v 'Waree' will output the font path and other things <bumble[m]>@iyzsong thank you it is listed by fc-match I would paste here but output is long enough it might be rude to paste <bumble[m]>file: "/home/bumble/.guix-profile/share/fonts/opentype/thai/Waree.otf"(w) <bumble[m]>the packge was installed more than once, using both (build-system gnu-build-system) and (build-system font-build-system) and not sure which one would have generated the file in .guix-profile/share/fonts... <next4th>bumble[m]: have you unpack the source tarball to check? 'gnu-build-system' will run './configure; make; make install', 'font-build-system' will scan and install font files (ttf, otf, etc.) (i'm same iyzsong, just get up..) <bumble[m]>@next4th the source tarball is downloaded and unpacked. I don't know how to verify gnu-build-system will run the configure make commands <bumble[m]>the unpacked source does not contain .otf or .ttf files that I can find <next4th>okay, it seems that gnu-build-system does works, i checked with './configure --prefix=/tmp/abc; make; make install' and see files in my '/tmp/abc' <bumble[m]>so does this mean the package basically works? <bumble[m]>sadly, I did not learn very much about guix from this... still do not really know how to package anything <bumble[m]>I would be interested to submit the package upstream <next4th>then you need to send a patch to guix-patches@gnu.org, see "Contributing" section in 'info guix' for details... <bumble[m]>I'm not sure if my thank you reached you, but I have audio in qutebrowser now from pipewire as well <next4th>yes, i got it, happy to be helpful here :) <bumble[m]>reading the contributing section per your advice and cloning the savannah guix repo now <bumble[m]>hoping it is not too big not a lot of space on this machine <next4th>my checkout have 1.2G... and it surely can't build on a 2G mem + 2G swap machine, I think 8G memory is needed to hack on guix now.. <bumble[m]>ok thanks for the details this machine has 8G memory and about 90G of space so should be OK <Guest40>What is the story regarding with guix and cgroup2? <Guest40>It seems that by default it's in hybrid mode (v1 and /unified), is it possible to use only v2? <next4th>not sure, maybe elogind need cgroup v1? <Guest40>elogind should be able to run on v2 only as far as I can tell from looking at its source code <Guest40>But guix explicitly mounts the v1, so I wonder if it's used by guix for something or if it could be changed to v2. <next4th>i guess it can be changed to v2, when i run the system without elogind, it works fine without cgroup fs. <Guest40>Oh, that's nice. I guess I will add this to my todo list and send a patch later. The current setup is not exactly podman friendly and too me a while to figure out how to run it, so would be nice if others will not have to spent the time on it. <bjc>are you running it rootless? <bjc>did you have to add /etc/sub[ug]id? <Guest40>Yeah, but also `elogind.legacy_elogind_cgroup_controller=1` kernel param, that one took me long to find. <bjc>ah. i didn't know about that one. maybe that's what i'm missing <bjc>if you're so inclined a write-up with how to get it going along with whatever patch you submit would be nice <Guest40>Well first patch will be to tweak the comment above the podman package definition (the one mentioning cgroup2 mount) to also include this option (is the comment the best option? Is there no "post install message" or something in guix?). <Guest40>I'll look into the switch into cgroup2 but that will likely not be exactly trivial to figure out how to do nicely, so that will take some time. <bjc>there isn't a post install message, no. there is a cookbook, though, and sometimes you can find useful instructions in the official docs <bjc>cgroups2 was where i got hung up, since i thought it was required. maybe it is for newer podman? i started messing around with seatd/greetd, but i never got it worked out before i had to move on to something else <Guest40>podman (afaik) works fine with v1 and v2, but it cannot handle the hybrid hierarchy (/sys/fs/cgroup/unified) <Guest40>Which is what guix (well, elogind) creates by default with the configure flags guix is using <bumble[m]>@Guest40 if your approach to using podman and guix is documented one day I would be interested to read that <Guest40>bumble[m]: I will post the patch soon-ish, but there is not much to document (once I know what is required). tl;dr: create /etc/sub{u,g}id with desired values using etc-service-file (mine contain `USER:100000:65536`), install podman, and boot with elogind.legacy_elogind_cgroup_controller=1. Boom, `podman run --rm docker.io/library/hello-world` does <Guest40>(that is when executed as regular user, meaning rootless podman works fine) <bjc>Guest40: i'm getting: WARN[0000] "/" is not a shared mount, this could cause issues or missing mounts with rootless containers <bjc>Error: open /etc/containers/policy.json: no such file or directory <lilyp>jpoiret: sorry, just a committer, not someone with admin rights :( <rekado>cuirass download URLs no longer work <jpoiret>ahhhh, the joy of waking up and finding out gnumach successfully built :D <jpoiret>sneek, later tell gnucode: Ah, you're right. I was thinking more about guix the software instead, but if we don't have a libc for macos then it's gonna be hard building any packages either :p <arjan>is there any way to write a PID file during shepherd service start for programs that do not support it? <mirai>what's the difference between maintainer/admin-rights and commit access <mirai>arjan: what are you trying to do? <arjan>I want to send the process of my service a signal to reload certificates in a certbot deploy hook <mirai>is this process in question 'nginx'? <arjan>no, in this case it is a shepherd service I define myself, but the program it runs does not support writing a PID file <mirai>arjan: I'll get back to you on this in a bit <mirai>I was about to write something that might be of use here <davidl>Hi, I have a bluetooth headset and when I manually change the profile to HFP from A2DP, it works fine to use with the microphone in the web-browser. However, I want it to automatically switch over to HFP - as needed - which it does on a MacOS X laptop I have. How can I make the switch happen back and forth automatically on Guix? <davidl>rekado: thx for sharing this: (simple-service 'my-dbus-extensions dbus-root-service-type (list bluez-alsa)) - it helped me be able to finally use my bluetooth headset microphone (although it doesn't switch profiles automatically yet, still need to use pavucontrol). <jpoiret>mirai: maintainers have more responsibilities than committers <jpoiret>davidl: do you use pulseaudio or pipewire? <angry_vincent>can i build all packages without X11 in same fashion as Gentoo ? sorry for a straight analogy <jpoiret>well, you can, but it's going to be quite complicated <jpoiret>I don't think you'll be able to get most things building <jpoiret>using package transformations you can remove arbitrary inputs of packages, but then the build phases also have to be adapted most probably, as well as configure flags <jpoiret>also, what is X11 for you? there are some x11 libraries which are now also used under wayland <angry_vincent>i do not see any X11 library used by wayland on my current box <jpoiret>angry_vincent: by that i meant thinks like libxkbcommon <jpoiret>mirai: I agree that it's not documented anywhere in the manual/on the website i think <jpoiret>could probably go in that reworked contributing section, i'll add that to my draft <jpoiret>civodul: i'm trying to selectively apply some patch to gcc if the target is hurd (we need to patch libstdc++). How can I pass the patch to a phase? The phases are sexps, and the inputs have the new style so I can't just add a patch in the native-inputs with a name <jpoiret>I can't just transform it into g-exps cause otherwise gcc-boot0 which inherits from it will have to transition as well, and same for the new input style <jpoiret>mirai: btw, with-imported-modules outside and use-modules inside <mirai>does shepherd always require root privileges even for things like "herd status?" <jpoiret>yes, the socket is only accessible to root <bjc>it does if you're looking at the root shepherd. ‘herd status’ as a regular user will look only at that user's shepherd instance <mirai>I'm hesitant about writing a script that checks if some shepherd service is running and then making http requests <mirai>it just doesn't sound "right" <jpoiret>probably there could be some more extensibility added to shepherd <jpoiret>i mean, to the shepherd service definition in guix <jpoiret>i think for now the only thing you can add is services, but it could make sense to add hooks to event handlers so that users can run their own code as well <bjc>why not add a shepherd service that runs the http script, which depends on the other service? <mirai>I'm writing a "heartbeat" mcron-service-extension for a shepherd service that occasionally misbehaves <bjc>this is getting a little wonky, but you could write a dependent shepherd service which installs the mcron service on ‘#:start’ and removes it on ‘#:stop’ <mirai>bjc: that could be an option (drawback being that it litters herd status output) <mirai>how does that unregistering work? <bjc>i dunno, i haven't thought that far ahead ;) <ekaitz>hi all! i'm trying to use Kawa with javafx and I'm having issues building it <ekaitz>the kawa we distribute looks like doesn't have the javafx support added by default <ekaitz>did anyone manage to do it before? <ekaitz>also, we don't provide the full javafx in any package, is there any reason behind that? <lechner>Hi, what are the gexp equivalents for %current-target-system and hurd-target? please? <lechner>thanks! what do i do with the asterisk in the coreutils input, please? <lechner>and %current-system in native-inputs stays as is, right? <jpoiret>well, not the asterisk itself, but rather what it's trying to <lechner>i only see coreutils and coreutils-minimal on packages.g.g.o <euandreh>I rebooted my server, but now the activation phase is failing. How can I make Guix continue booting and ignore the activation exception? <euandreh>The activation is failing from some activation code I've written, so I know I can ignore it <jpoiret>you can't really do that unfortunately <euandreh>Unfortunatelly, no. All dozens of previous versions have similar code <euandreh>The issue is that I changed a directory to become a symlink to another disk <euandreh>But it seems that disk isn't available yet at this time of the boot process <euandreh>"/srv/www" is now "/srv -> /mnt/production/www" <euandreh>Could I get a bash prompt from the Guile REPL? <jpoiret>it is possible but you have to know your fair share of guile <lechner>are you trying to mount, or simply replace the symlink temporarily with (mkdir-p "/srv/www") <euandreh>This is just temporary, to get the system up and working <lechner>i personally would go with option (2). mirai has delayed mounting figured out. will post some of it here shortly <euandreh>oops, Ctrl-D when inside ,bournish quits the repl <euandreh>kernel panic after that :) reboot and try again <lechner>you may also be able to avoid some of your troubles by mounting a partition holding "www" onto /srv/www (instead of "srv" on /srv) as then the folder exists early <euandreh>(delete-file "/srv") followed by (mkdir-p "/srv/www") created the directory, but I still got a panic after that <euandreh>I think I missed another path or symlink <euandreh>hmm, actually not. After the reboot, it worked <euandreh>lechner: OK, so for fixing it you sent me a link to the definition of the inetutils package <euandreh>lechner: does that manual call to mount(8) not conflict with the (file-system ...) definition of /acct? <lechner>please note the (mount? #f) in file-systems <mirai>lechner: networkingmanager patch has landed in guix <mirai>you can simply use 'networking now <lechner>euandreh / do not listen to me. let's ask mirai what the best way is now <lechner>mirai / 'networking in file-systems or in the one-sot? <euandreh>mirai: if you scrollback a few minutes, I just had an issue where some activation code snipped I wrote was calling (mkdir "/srv/www"), but "/srv" was a symlink to a different disk that wasn't mounted <euandreh>mirai: so you were cited regarding having worked on delayed mounting <euandreh>have you tried some solutions to this issue? <mirai>lechner: you can remove that 'networking-online workaround <mirai>euandreh: so iiuc the hang is being caused by a missing mount at boot time <mirai>filesystems that aren't guaranteed to be present at boot time must be needed-for-boot? #f <lechner>they end up in the early guile repl, actually <lechner>boy, do i dislike the "they" as a modern pronoun <mirai>the delayed mounting can be done in two ways depending on the scenario <euandreh>oh, I don't think I need any delayed boot, actually <euandreh>just the needed-for-boot? option, I think <lechner>don't you still have to mount manually, though? <euandreh>IIUC, when needed-for-boot is #t guix will mount it for you, before running any activation code <euandreh>as far as I can tell, that's all that I need <mirai>lechner: it will still mount even if needed is #f <euandreh>it is attached via SAN, but it looks local to the OS <euandreh>thanks lechner and mirai for the help! :) <lechner>Hi, is it allowed to mix old-style and new-style inputs? <user`>Hi I am having problems with compiling the Linux kernel with guix. <user`>I did a guix system reconfigure /etc/config.scm <user`>I am a beginner with guix so it is totally possible I just missed something obvious... <jpoiret>lechner: no, it's not possible unfortunately <jpoiret>also, what's the output of `guix describe`? <lechner>when using as is, i get "error: file-append: unbound variable" presumably because file-append is not available on the build side. <lechner>when using named fields, i get "error: name: unbound variable" <mirai>lechner: it was changed recently <mirai>it really doesn't support named fields <mirai>this one uses plain define-record-type <lechner>mirai / thanks! i think the gexp in the example is still not right, tough <mirai>there's no gexp in the example <lechner>Hi, with the core-updates process changing towards feature branches, what is now the right way to file changes that affect a lot of packages, please? (This is for an update of inetutils, which rebuilds around 3,500 packages) <mirai>the service does look a tad ugly though <lechner>maybe you can get your hands on it before a lot of people adopt this strange version <mirai>I'd prefer someone who actually uses that service to pave the new path <mirai>I don't see myself using it so I'm not in tune with the audiences desires <jpoiret>lechner: I'd say send it to the ML, it's a core package so should be CC'ing the core team as well <jpoiret>but core-updates is a bottleneck right now so you probably won't have any answers there <lechner>Hi, should issues have a tag that indicates when they were elevated to (or merged into) a feature branch? <user`>jpoiret: x86_64 is my architecture <lilyp>Wow, github just completely broke for me on Epiphany <angry_vincent>jpoiret: technically what i meant is disable of X11 support, where it can be, when using Wayland, it is not too complicated, though not straightforward for some applications. i have done this, my question was if guix allows custom packages, etc <lechner>angry_vincent / Hi, we have custom packages, but i don't think they work they way you expect <lechner>not to my knowledge. you can copy the package definition and change it, but our functional model means that you will locally rebuild all consuming packages (which may not be that many for user-facing apps) <lechner>what is the purpose of your endeavor, please? <lilyp>lechner: it's just raising js errors on and on <lechner>lilyp / sorry to read that. it may ruin the remainder of your day <lechner>angry_vincent / just copy the package definition into your config, change the variable name, and use that <angry_vincent>is there a way of tracking reverse runtime, buildtime dependencies? <jpoiret>angry_vincent: we don't have a notion of runtime or buildtime dependencies <jpoiret>guix detects whether build outputs retain references to their inputs, and only those inputs are kept around <jpoiret>you could call those "runtime" dependencies if you will but it's not something you have to write down in Guix, it's taken care of for you <jpoiret>you can use `guix size` to see how big package closures are, ie. all of what has to be installed alongside a specific package, and `guix graph` to find out all package dependencies <lechner>angry_vincent / you won't have any issues if you define and use custom variable. the rebuilding will only happen if you run a custom version of Guix in Git after modifying packages there <PotentialUser-95>hello, is it possible to run on GUIX the scanner HP laserjetpro mfp m125nw? I tried with the instructions of the manual, with "scanimage -L" my scanner is visible, even simple-scan sees it, but when I try the scan gives me: impossible to connect to the scanner. can anyone help me? <PotentialUser-95>lechner / yes: result: device `hpaio:/usb/HP_LaserJet_Pro_MFP_M125nw?serial=CNB7GD3H74' is a Hewlett-Packard HP_LaserJet_Pro_MFP_M125nw all-in-one <PotentialUser-95>scanimage: open of device hpaio:/usb/HP_LaserJet_Pro_MFP_M125nw?serial=CNB7GD3H74 failed: Error during device I/O <lechner>PotentialUser-95 / sorry, I have no experience with HP's All-in-One app (hpiao:) <atuin>Hi! I have a question: I'm trying to replace syslogd with rsyslogd. I created a vm and a container using "guix system" but neither of them have the /dev/log file. When I create the same vm or container using syslogd the file (/dev/log) is there. Any idea what can be happening? how can just changing one service affect the creation of the socket? <NewUser-Basic-Qu>hi...most probably qemu related question but seeing it on y guix-vm. Am currently connected to all available ttys withs same user. Would expect that "who" or "w" command shoes me all my logins but it only shows my ssh connection to the VM... <jpoiret>lechner: can you send the paste of your changes here? <lechner>lint also complains that the package is not tracked on Disarchive or Software Heritage. should we switch to importing from Git? <angry_vincent>is there a collection of packages ( scm's ) to look how they defined? <lechner>you can use grep or packages.guix.gnu.org will provide a clickable link <davidl>jpoiret: Im using pulseaudio. I tried making changes to my default.pa file but it didn't fix the auto-switching mode issue - at least not so far. I suppose I should try pipewire? Im on Sway VM by the way - not sure if that means I need to do something extra manually. <jpoiret>lechner: you can't have new style and old style inputs in the same package definition I think <rekado>jpoiret: it is possible. We have some packages where native-inputs are old-style (due to custom labels) and regular inputs are provided as a plain list. <jpoiret>i wonder if lechner still has the same error as before <jpoiret>maybe you could post it again here so that everyone has a chance to see it <bdju>whoa new progress bars in guix?? <jpoiret>civodul: I have some patching code that only changes the derivation if the target is hurd. Well at least I thought it was working, but now I'm still getting rebuilds of Python if I change the hurd/gnumach/dde sources <jpoiret>and btw I don't think we can build the latest hurd, only from last august, otherwise we need a newer libc and we'd need to also carry such a selective patch <Guest74>is it possible to globally graft package? Like that each package that includes the old one is instead using the replacement? <jpoiret>but I don't think that's what you want: you need to modify the package definition itself to graft <jpoiret>graft really are useful only to avoid world rebuilds, not to customize packages for yourself <Guest74>hhm, I just need to replace it on the whole system since without it does not have necessary files <unmatched-paren>it returns the output path correctly but it doesn't seem to actually build derivation <jpoiret>unmatched-paren: you need an mbegin in the body of the lambda <jpoiret>seems like it swallows it without erroring, from a quick reading of the macro <unmatched-paren>jpoiret: ah, so it's (mlet* (BINDINGS ...) BODY-EXPR) rather than (mlet* (BINDINGS ...) BODY-EXPRS ...)? <jpoiret>this could be fixed quite easily tbh <jpoiret>it's just a matter of adding an (mbegin ...) in the definition of the macro <civodul>jpoiret: hi! great work! we can postpone upgrades to the latest Hurd and DDE, that's kinda secondary <civodul>as long as we have MiG + Mach/Hurd headers + libc, we're pretty good <jpoiret>civodul: but doesn't that mean we won't be able to build hurd anyways? <jpoiret>hurd/mig/gnumach all upgrade in lockstep anyways <jpoiret>so if we can't build the newer hurd now, I don't think we will be able to later either <jpoiret>I'm trying to get rid of mig/32-bit as well, which I think is possible (and I've been told so) <jpoiret>I can stash the "working" patches I have to build up to hurd (non inclusive) <ellysone[m]>is the hurd ready for daily usage? how does it feels? <jpoiret>x86_64 and SMP support is in progress <jpoiret>it definitely feels ... special :) I've never used any niche OS so I might be biased though <lechner>ACTION has dreamt about using the hurd since 1993 and really cannot wait! <bjc>i've been waiting 30 years, what's another decade? <civodul>jpoiret: normally there's some relatively loose coupling between the Hurd and libc/Mach <lechner>i'm more worried it might be another thirty years <civodul>libc has stringent requirements: it needs certain interfaces from the Hurd (and possibly Mach) <civodul>whereas the Hurd doesn't require much from libc <jpoiret>civodul: right, but there's a september commit in hurd for pfinet that requires glibc-2.37 <jpoiret>oh, you mean that we might just be able to use an older hurd with a newer gnumach? <jpoiret>in any case, I'm getting pretty familiar with it so I don't think I have much work left anyways <jpoiret>the hardest parts should be sorted out <civodul>i don't think pfinet uses struct ifrtreq, does it? <civodul>that sounds more like a POSIX/Linux compatibility thing, no? <civodul>but pfinet_getroutes exists specifically so libc can expose those ioctls <civodul>which in turn probably satisfies net-tools and things like that <lechner>Hi, why are there no QA results for #41180 after three days, please? are they needed for a reviewer to approve a new service? <lechner>this one was co-authored by mirai, if that encourages anyone to take the plunge <cbaines>lechner, there's no branch been produced, so it's likely there was a problem when applying the patch <cbaines>unfortunately this isn't reported yet <lechner>did i forget a base commit (even though I have that in my git config)? <cbaines>I think it's just that the issue is an old one which isn't considered for testing <cbaines>now that the data service doesn't have a backlog, I'll look at getting more issues processed <lechner>okay, it does not add much for new services anyway. the emphasis here should be that the data model is okay since it's somewhat hard to change later. it's a simple daemon with five options <lechner>but anyway, with mirai's sign-off even the most hesitant reviewer should feel comfortable to push the button <mirai>you're overestimating my abilities way too much :) <lechner>one has to reach for anything with those time-strapped reviewers. just like my bag of popcorn. a few days ago, i ate kettle corn, which is caramelized and salty, and not very healthy. on the bag, it said "whole grain"