IRC channel logs


back to list of logs

<vagrantcish>i guess i could try reproducing in guix environment ...
<maav>vagrantcish: you should get the environment too when you use -K option, don't you?
<maav>I mean /tmp/guix-build-package-.../environment-variables
<maav>there you can use guix environment (or directly that file) and run the tests manually
<jonsger>cbaines: thx, I get
<vagrantcish>ah, --keep-failed ... i always use longhand options :)
<cbaines>jonsger, I saw some test failures, but that looks quite clear. hdf5 was changed recently
<jonsger>vagrantcish: thats good, it's easier to remember and if your too lazy your autocomplete is bullshit...
<vagrantcish>i've only ever poked at the source directory after using --keep-failed ... didn't know about the environment-varaibles stuff
<maav>i prefer short options for interactive commands as i don't like to type even the extra dash, long options for scripts because i want to remember what i wanted
<maav>i discovered the environment-variables file by luck... and really was a lucky day :)
<jonsger>cbaines: so I should open a bug and set Ludo to CC?
<cbaines>jonsger, hmm, do you have what derivation you tried building?
<cbaines>the error you got looks weird: "The HDF5 header files used to compile this application do not match"
<cbaines>"Headers are 1.10.7, library is 1.8.21"
<cbaines>I don't get why the library version is old...
<cbaines>I did try building it, and I'm 95% the way through what I guess is the build phase
<PotentialUser-29>cbaines: ty for the support earlier. I got it working. though I am still a bit unsure about the record syntax at work here since I never touched guile before. can a record be shadowed with "new values" per se so that the end result is old record with "new values" overriding where applicable?
<jonsger>it's /gnu/store/mpy3f0gqilqavd2qyhkr7im5drw7jqcl-qgis-3.12.1.drv but I'm working on so I added `("python-future" ,python-future)` as input
<jonsger>cbaines: its in the check phase :)
<maav> it matches with out versions nonetheless, as there is hdf5-1.8 and hdf5 (1.10)
<cbaines>jonsger, ah, check phase just started, so I guess I'll find out soon...
<maav>it might be a duplicated reference through two dependencies?
<cbaines>PotentialUser-29, you're welcome :) I'm glad you got it working. With some of the records used in Guix, there's some syntax introduced (define-record-type*) that allows one record to inherit values from another, which sounds a lot like what you describe
<cbaines>jonsger, maav is probably on to something: guix graph --path qgis hdf5@1.8
<cbaines>that says it's the netcdf package that's probably the issue
<jonsger>cbaines: doesn't look so interesting to me...
<cbaines>it pulls in hdf5-1.8, because of a test failure with 1.10
<rekado>I’ve been working on plymouth integration
<rekado>not finished
<PotentialUser-29>cbaines: tyvm for the link again. Will have to learn more guile in the future guess :) I want to use guix system/sd on my work machine and with your help above I am 50% there. a bigger hurdle is getting a scala dev env (which I need for work) I guess till scala 3 drops in december the compler wont make it into guix perhaps I can make due with dock
<PotentialUser-29>er envs for a while though :o)
<rekado>I still have a stash of changes somewhere
<cbaines>jonsger, it looks like netcdf hasn't been updated in Guix for 4 years now, so updating it may fix that test failure
<maav>rekado: if nobody takes it after the release, ping me and I'll take a look
<cbaines>jonsger, I didn't even make it to the error you got locally, I got the flaky test failure instead: 101 - qgis_meshlayerrenderertest (Child aborted)
<cbaines>PotentialUser-29, great, good luck :)
<jonsger>cbaines: I got that as well
<jonsger>scroll down
<cbaines>hmm, my output ended with that
<cbaines>unfortunately I ran guix build --log-file, rather than guix build --no-grafts --log-file, so I lost the output
<cbaines>and I've closed the terminal
<cbaines>maybe it stops on the first failure, and you just happened to get both failures at roughly the same time
<jonsger>maybe, I'm building with cores=12 :)
***nuka is now known as niko
<maav>cbaines: i'd test to put hfc-1.8 in qgis' inputs, and i'd open a bug report ifthat works, as these two versions seem to be incompatible
***jess is now known as jess-o-lantern
<maav>jonsger: you... you must know the envy I feel right now ;)
<cbaines>maav, well, trying to go forward and update netcdf is better long term, although that's maybe a good short term approach
<jonsger>maav: it seems you are using qgis, aren't you?
<maav>cbaines: sure, sure, it's the best option, i mean just to check :)
<maav>jonsger: not at all, i was just learning what do these software packages do :)
<maav>jonsger: but the issue seem obvious from the message you pasted, as the versions matched
<jonsger>yeah, I just need to build some kernels so I don't have time for qgis yet :P
<maav>nah, just grub 15 times today, but with 1 core, hehehe
<cbaines>maav, is one core all you have, or all you have to spare?
<maav>cbaines: i have two, like GB of RAM :)
<cbaines>that is quite a small amount
<cbaines>I had to find what was memory leaking on my desktop a few days ago, as all 64GB of swap on my system was used up...
<cbaines>I think it turned out to be my music player
<raghavgururajan>mpv or vlc?
<cbaines>Lollypop in this case
<raghavgururajan>I see.
<cbaines>I updated the package just in case something has been fixed, and then just carried on
<maav>well, two can be a good amount, for example i also have two hands and they are ok, as far as it works... :)
<cbaines>it was just the last think I expected to be leaking memory!
<maav>also, i don't have to wait until i fill up 64GB :P
<raghavgururajan>It's in World category of GNOME. So might me not matured.
<raghavgururajan>*not be
<maav>well, good luck with that cbaines and jonsger, i'm going to sleep and my couple of cores have to finish their reconfigure :-)
<jonsger>cbaines: I had serious memory leaks while listening to music via deezer on firefox, but it was on openSUSE. I don't have that issue on Guix system :)
<cbaines>jonsger, I don't know if you were looking at updating netcdf as well, but I've built 4.7.4 locally with hdf5-1.10
<cbaines>I haven't tested if any of it's dependencies still build though
<vagrantcish>diffoscope was getting too far behind ... finally cleaned out a lot of those broken test suites...
<mroh>yay! I was looking at that a while ago, but it ended in another (unfinished) git stash, as usual ;)
<vagrantcish>and lazy me forgot to check reprotest.... fixing...
<joshuaBPMan>interesting ---> points to the actual hurd wiki. That seems new...
<xelxebar>Hrm. Seeing this; however, I am pretty sure my guix-daemon is running the same version of glibc as my user...
<apteryx>cbaines: I managed to find what caused local-file to fail :-)
<apteryx>the problem occurs when the entry for Guix in the GUILE_COMPILED_PATH contains a trailing slash
<apteryx>vagrantc: thanks, sounds good!
<apteryx>hmm... 'make release' -> make[3]: *** No rule to make target 'po/doc/guix-manual.pot', needed by 'distdir-am'. Stop.
<Gooberpatrol66>How do I change a package definition?
<Gooberpatrol66>I want to change an input for a package
<xelxebar>Gooberpatrol66: Actually, you don't need to change the definition, check out the --with-input=PACKAGE=REPLACEMENT option for guix build/package.
<xelxebar>Gooberpatrol66: If you need to change a recipe or something, then you probably want to clone the git repo and work directly from there.
<apteryx>Gooberpatrol66: but if you do want to hack on the Guix sources, as xelxebar said clone the git sources, build them following the manual instructions, then ./pre-inst-env guix edit the-package
<Gooberpatrol66>I want to be able to install the changed package
<Gooberpatrol66>the --with-input flag gives a "unknown package" error
<Gooberpatrol66>guix package: error: --with-input=jack=jack2: unknown package
<xelxebar>Gooberpatrol66: What package are you trying to modify? I'm not an expert with guix stuff yet, but I can take a crack at it.
<xelxebar>Haven't actually used an of the --with-* options yet.
<Gooberpatrol66>xelxebar: qjackctl
***daniel is now known as Guest31084
***Guest31084 is now known as daniel-molina
<xelxebar>Gooberpatrol66: Building now, with this command: guix build --with-input=jack=jack2 qjackctl. Will report what happens.
<Gooberpatrol66>xelxebar: does that flag work with guix package, or only guix build?
<xelxebar>Gooberpatrol66: The manual says that it works with package as well
<xelxebar>What command did you run that gave the error
<Gooberpatrol66>xelxebar: guix package -i --with-input=jack=jack-2 qjackctl
<xelxebar>Gooberpatrol66: jack-2 != jack2. Maybe that was the problem?
<Gooberpatrol66>xelxebar: I also tried guix package -i --with-input=jack=jack2 qjackctl
<Gooberpatrol66>it seems to work with guix build
***sneek_ is now known as sneek
<rekado>you shouldn’t need to rebuild packages with jack2
<rekado>jack has the right API already
<rekado>if you want to load jack2 at run-time just put it on LD_LIBRARY_PATH to swap out jack
<xelxebar>What are references/referrers of store files?
<xelxebar>We have --requisites that (recursively) gives the items a package depends on. I guess that --references and --referrers do something like the reverse, showing reverse dependencies?
***apteryx is now known as Guest16909
***apteryx_ is now known as apteryx
<wleslie>does anyone happen to have an m4 script they like that detects openssl and works on guix? thinking I might use the one from lighttpd
<cbaines>apteryx, I'm glad to hear you figured out the local-file issue :)
<divoplade>Hello, I would like to speak to #guile, but every time I try to send something I get "#guile :Cannot send to nick/channel"
<divoplade>Am I doing something wrong?
<cbaines>hmm, seems like you're in the channel
<cbaines>I'm unsure why it wouldn't work
<vagrantc>does it block unauthenticated nicks?
<divoplade>cbaines, I can read you there
<vagrantc>divoplade: you may need to identify to nicserv
<divoplade>vagrantc, how do I do that? I'm on hexchat, but I can switch to another client if it's easier
<vagrantc>divoplade: /msg nickserv help
<vagrantc>divoplade: first you have to register, then you have to identify, etc...
<jlicht>hey guix!
<cbaines>jlicht, o/
***vlorentz1 is now known as vlorentz
<user_oreloznog>Hello maav!
***daniel is now known as Guest9569
<maav>hi, user_oreloznog, i didn't see it until now :)
<maav>haven't seen*, my grummar :-P
***Guest9569 is now known as daniel-molina
<mfg>I want to run guix-build directly from the repl but it seems i'm using the wrong format for my options... How are those passed?
<mfg>Ah now i got it
<efraim>(guix-build "arg" "arg" "arg")
<efraim>returns undefined though
<mfg>have you imported guix scripts build?
<mfg>i thought the arguments were a single string that's why it failed (gave no error though)
<leoprikler>nah, all procedures are defined as (proc . args) for command-line usage
<efraim>it also works with environment
<efraim>(guix-environment "--ad-hoc" "fontconfig" "--" "fc-cache" "-frv")
<mfg>thx leoprikler and efraim :)
<maav>hi efraim, and sorry the overload of wip code in the glibc locales :)
<efraim>not a problem :)
<xelxebar>mfg: Cool. What are you working on that gets you using guix-build in the repl?
<mfg>xelxebar: i want to find out why it's impossible to use a local path to a git repository in a package definition but it's no problem when i invoke git-fetch directly...
<mfg>but i guess it's just that i'm using it wrong :-D
<xelxebar>Oh. I have been toying with some ideas that want to do that. Didn't realize it might be problematic!
<mfg>pretty sure it's something obvious ;-)
<mfg>ah and i didn't know there are multiple git-fetches, the one that i tried was in (guix build git), the one used seems to be in (guix git-download)
<mfg>what i don't understand though is: how do i attach a repl to the running guix-daemon? as this is where the functions i want to inspect get executed... or is it possible to start a second one?
<leoprikler>mfg: you mean like guix repl?
<mfg>leoprikler: i'm pretty certain i don't understand things correctly. But now i think i know what i did wrong... git-fetch wasn't executed because i had used another link in between which has the same repo and therefore guix didn't try to redownload it. Thus the breakpoint on git-fetch doesn't fire - lel.
<mfg>but thanks for reassuring that guix repl is what i want to use :)
<roptat>hi guix!
<maav>hi roptat!
<user_oreloznog>roptat: the robot is almost pleased about my :)
<user_oreloznog>roptat: he said not a member of the French team yet, and now: Hubert Lombard has been successfully subscribed to the Translators mailing list.
<user_oreloznog>so now, maybe i must wait a little more to make it good...
<user_oreloznog>Or I'll send a mail to a coordinator
<erkin>Hi Guix! I have a working PC now. :-)
<civodul>hi erkin, good! :-)
***jonsger1 is now known as jonsger
<erkin>What's the recommended way for sending a patch for fixing multiple minor errors in documentation?
<erkin>I figured it might be too lazy to just say "fixed some errors" in the message. :-P
<jonsger>erkin: if they are in different chapters and of different kind it's maybe better to make one commit per fix
<jonsger>erkin: gives you maybe some inspiration how to write the commit message :)
<erkin>This seems like a good idea: Adding description of fixes for each one in the message.
<erkin>Thanks jonsger
*kozo[m] sent a long message: < >
<jayspeer>hey guys, how do I get gnome to see my bluetooth device? I've managed to use using the bluetoothctl
<jayspeer>follow up question is: is there a pulseaudio-bluetooth package? I can't seem to find it
<apteryx>civodul: do you think should be fixed at the level of Guile? it seems odd that a trailing slash ends in the %load-path or %load-compiled-path. If I compare with Python, it seems to normalize the entries in its sys.path in a consistent manner.
<jonsger>cbaines: you don't have your guix-patches thing running anymore?
<cbaines>jonsger, I do, but the domain names might have changed
<cbaines>I'd start at Patchwork
<cbaines>then there are links in the patches
<apteryx>when using git send-email, --in-reply-to should look like '', or '<>' ?
<apteryx>according to man git-send-email, the later.
*apteryx seems to recally it wouldn't work, but tries again
<buenouanq>in guile scripting on guix sd, what do I use for the full path to the interpreter? #!/usr/local/bin/guile obviously isn't the ticket
<mroh>buenouanq: /run/current-system/profile/bin/guile
<buenouanq>mroh: thank you very much. Can you tell me more about this path and what goes there and how it should be used?
<divoplade>buenouanq, If you use #!/usr/local/bin/guile, and make a guix package out of your code, it will automatically be replaced with the appropriate executable name
<divoplade>Once installed as a package.
<lfam>I'm a bit out of the loop — is it still okay to update the Git package directly on the master branch?
<sneek>Welcome back lfam, you have 1 message!
<sneek>lfam, civodul says: re linux-libre 5.9, no comments from me other than "thanks"!
<lfam>sneek: botsnack
<maav>lfam: guix refresh --list-dependent git says "Building the following 108 packages would ensure 241 dependent packages are rebuilt", so if its the sum it should go to staging, if the number that counts is the second one can go to master :)
<lfam>I guess that all of us who used to update Git immediately got tired :)
<maav>lfam: right now i don't have machine to compile these 351 packages, it's getting really hot just with one (new) test for check-system... :)
<lfam>It's okay, maav, that's what the build farm is for :) I'm building the latest Git now and will push shortly
<maav>cool! :)
<jlicht>does anyone know of a package that uses the guile scripting facilities of GNU Make?
<PotentialUser-92>Why is Guix System, a rolling release distro, still using an old version of GNOME from 2019? Isn't GNOME 3.38 stable enough?
<lfam>PotentialUser-92: It takes a lot of work to update the GNOME ecosystem, and nobody has completed that work yet
<ryanprior>PotentialUser-92: a couple potential reasons. First of all, it's work to update everything and maybe nobody's done the work yet. Second, we do have a stable release cycle for packages with lots of dependencies, which might include parts of GNOME.
<PotentialUser-92>Ok. Thank you.
<ryanprior>Are you interested in updating some GNOME packages? I gave a talk about packaging GTK apps for Guix at GUADEC & I'm happy to help out anybody who's interested
<erkin>They left. :-(
<rekado>raghavgururajan: how far is your work on Gnome? Do you think there’s already something we could merge from wip-desktop soon?
<raghavgururajan>rekado: Yes, these ( can be merged to master. I was asked to split the patches. The first 5 is ready.
*raghavgururajan pings dannym
<rekado>raghavgururajan: excellent!
<raghavgururajan>I will be splitting remaining patches on #42958 today.
*raghavgururajan is in "lets do it" mode
*rekado starts a new wip-texlive branch with fixes to the modular TeX Live
<maav>rekado: kudos! :)
<jonsger>cbaines: it would be pretty handy if the branch name would be derived from the issue number :)
<cbaines>jonsger, it's derived from the Patchwork series number
<cbaines>I guess the issue number could be used, but that would get a little complicated where you have multiple series for a single issue
<jonsger>cbaines: I know, but it would be nice so I can checkout it more easy
<morgansmith>is there a way to do package transformation to a manifest somehow? I tried guix package -m manifest.scm --with-branch... but that didn't work. It works with package update and package build, but not manifest
<leoprikler>you'll have to write your transformation as scheme code into the manifest
***amiloradovsky1 is now known as amiloradovsky
<morgansmith>oh no. that sounds tricky. Do I gotta do the weird gexp transformations that I still don't know what they do?
<leoprikler>I'm not quite sure, but I think most transformations should actually be exported somewhere.
<leoprikler>If not, you can still copy the relevant code portion from `guix build`
<leoprikler>or just import options->transformation from (guix scripts build)
<leoprikler>That should work even better than the command line, since you can specify it for just the package you need or the entire manifest at will.
*leoprikler → afk
<morgansmith>has anyone done this before? I'm a little lost...
<vagrantc>can linux-module-build-system be passed an argument to use a different kernel version or kernel variant?
<vagrantc>i ask as the wireguard-linux-compat package is building against 5.9.1, and it's only needed on versions < 5.6
<vagrantc>seems silly to keep it around if it defaults to building where there's already an in-tree module
<civodul>morgansmith: no gexps involved :-)
<civodul>depending on the transformation it can be quite simple
***jtmar is now known as jamestmartin
<civodul>the manual shows package-input-rewriting for instance, which corresponds to --with-input/--with-graft/etc.
<civodul>though it could probably use more examples
<morgansmith>what section of the manual are we talking about?
<morgansmith>oh, probably 8.2
<morgansmith>Oh dear. I think I'm in a rabbit hole. I might have to read a lot of code to do what I want to do :P
<civodul>heh :-) what kind of transformation do you want to do?
<morgansmith>just a simple --with-git-url
<morgansmith>but it needs to propagate
<dustyweb>hm, if someone has a minute
<dustyweb>I could use help debugging something that's er... very nerdy
<morgansmith>dustyweb: just shoot. We're all interested in fun problems
<dustyweb>trying to package lojban-mode for emacs :)
<dustyweb>however, when trying to build emacs-lojban I get:
<dustyweb>Compiling /gnu/store/2hgmjkw6m1id4gr9zcg7qgig9wj758mv-emacs-lojban-0.23/share/emacs/site-lisp/lojban.el...
<dustyweb>Symbol’s value as variable is void: lojban-c-rgx
<dustyweb>I see it's defined on
<dustyweb>I also see a defconst for it in the downloaded file
<efraim>My pine64 thinks its 2115 again, I might need to implement a fake-hwclock service to set the date at startup
<vagrantc>when it leaps past 2038, it gets hard to set it right.
<vagrantc>haven't seen that problem for a long time
<vagrantc>it'd be nice to have a fake-hwclock service that at least sets the date to the date of the system generation, to avoid backsliding into the abyss from which ntpd refuses to leap even if you insist
<vagrantc>efraim: in other words, looking forward to your service :)
<civodul>morgansmith: --with-git-url is like using (git-checkout ...) instead of (origin ...) in the package of interest (meaning that the checkout happens on the client side)
<dustyweb>well I submitted it to guix-patches
<dustyweb>maybe someone will have an idea
<civodul>morgansmith: and then, if it's deep down in the graph, you can use package-input-rewriting to do the substitution
<civodul>i hope that makes sense
<morgansmith>civodul: ok so I completely ignored everything you said and made a procedure using options->transformations but the procedure I made wants a "store"? idk what that is
<morgansmith>maybe I should look at the package-input-rewriting function...
<morgansmith>wait, am I going to just have to create my own package that inherits the old package but has a new input? That's gross. I don't want to do that. I wanna use something like the transform-package-source-git-url function
<lfam>vagrantc: I think it would be nice if wireguard-linux-compat was based on 5.4, since that version will be around for a very long time (through 2025)
<civodul>morgansmith: yeah you can use options->transformations and pass #f as the first argument, it should work, but don't repeat it to anyone
<morgansmith>but what is store :'(.... I've been jumping through so many files and I cannot for the life of me figure it out
<morgansmith>oh it worked. I finally built a manifest with a transformation
<morgansmith>but how am I supposed to do it?
<civodul>we need to remove the 'store' argument from that procedure and properly document it
<civodul>then people will be able to use it in their manifest
<civodul>the other option is to use package-input-rewriting & co.
<civodul>it shouldn't be difficult, but how it maps to a specific transformation option is less obvious
<morgansmith>does using package-input-rewriting require me to pass it packages? that doesn't sound like it helps me since I'm trying to modify a package...
<civodul>yes, as in the openssl/libressl example in the manual
<morgansmith>so you want me to just make a completely new package using inherit, then use package-input-rewriting to replace the old package with my package? Seems really overkill
<morgansmith>I mean it sounds really useful. But overkill in my specific situation
<PotentialUser-38>Hi! I am wanting to set my PATH in guix sd/system but the only way I found so far is setting it in my bashrc. Unfortunately that does not cut it in a few use cases :o. Anyone have any tips as to how I could set the PATH (or other env variables) in preferably my system config.scm?
<morgansmith>I put it in .profile but it depends on a lot of things. Your shell should source .profile fine in the tty, but your desktop environment won't.
<morgansmith>You can sym link .profile to .zprofile or something? there's some filename that most desktop environments do source
<morgansmith>I just make sure to source my .profile in my .xsession, but that only helps you if you use a .xsession
<morgansmith>I don't think there is a way to set environment variables in your system config.scm. There is possibly a way to do it for a certain profile though
<morgansmith>It might require making a package definition just for that though
<PotentialUser-38>morgansmith: well that was what I was afraid of, :)
<PotentialUser-38>I can make everything work in the editor but if I am in a gui (only really applies to emacs but Im in emacs 90% of the time) then yea problems :o
<PotentialUser-38>can always find workaround of sorts but they feel like just that. workarounds
<morgansmith>That's odd. What are you trying to do?
<morgansmith>I'm in emacs 24/7 btw
<PotentialUser-38>well I just want to alter or load env variables for my user in a consistent way, but currently all I have is xsession and shell-rcs :)
<PotentialUser-38>I guess getting to use xinitrc would be nice also but they make that hard also :o
<morgansmith>ya, took me forever to get xinit up and running. if you want to check out how I did it
<morgansmith>It's not easy though
<PotentialUser-38>ok tyvm
<morgansmith>ya look at .config/guix/config.scm and then the last line of .profile
<vagrantc>lfam: exactly what i was thinking ... the 5.4 being long-term, and actually useful ... but maybe requires some reworking of the linux-module-build-system and whatnot
<vagrantc>and, on that note
*vagrantc waves
<PotentialUser-38>morgansmith: ok nice that looks like something I can try. it will be much more consistent then what I have currently. also I might borrow some of your emacs guix dev env config for emacs also if you dont mind :)
<morgansmith>Please read the disclaimer at the top of that file :P half of it doesn't work
<morgansmith>the emacs-guix package freezes my emacs a lot and often just doesn't work (idk why)
<morgansmith>It's licensed using the unlicense so you can plagiarize all you want :) I hope you find something nice in there
<PotentialUser-38>morgansmith: ok that mirrors my experiences with emacs-guix then :) I thought it was I who couldnt set it up :)
<morgansmith>If you can figure out how to work it, I'd love to hear it :P
<rekado>the new wip-texlive branch is now open
<rekado>I have a few more packages that I’ll push there in the coming days
<rekado>but I already welcome contributions to fixing our modular TeX Live packages
<PotentialUser-38>morgansmith: I doubt it will happen today. Il let you know if I find something out. so far the best "discovery tool" Iv found is guix edit <package> heh :)
<jlicht>nice work rekado!