IRC channel logs

2021-08-10.log

back to list of logs

<iskarian>sneek, later tell civodul: I haven't had a chance to test it yet, but the fix looks good (for the medium term anyway; I'm not happy with Guix not being able to present everything that will be downloaded at once, and would very much like to fix that long term)
<sneek>Got it.
***Kimapr9 is now known as Kimapr
***Kimapr96 is now known as Kimapr
***Kimapr5 is now known as Kimapr
***Kimapr6 is now known as Kimapr
<Fare>Can I generate a manifest from whatever it currently installed?
<apteryx>--export-manifest
<apteryx>(for the 'guix package' command)
<podiki[m]>It'll even capture your transformations you may have done to packages, which I just found out and it's very cool
<Fare>Thanks again!
<the_tubular>Can you explain that podiki[m] ?
<iskarian>the_tubular, I believe they mean that if you installed a package with, for example, --with-latest or --with-input, it puts those in the manifest as well
<the_tubular>Ohh, I didn't knew that!
<the_tubular>I gotta learn how to make use of guix vm
<the_tubular>I need to create 2 vms, and I'd like them to run guix
<podiki[m]>the_tubular: yup, what iskarian said. I installed some with --with-latest or used a git version with branch instead, another with a patch, all saved in the manifest if you export
<the_tubular>That's good to know!
<podiki[m]>you can also then pin with "inferiors" to make sure you capture an exact version too (this is all in the cookbook)
<podiki[m]>very handy, time for me to set up a bunch of manifests and profiles to organize my package list
<podiki[m]>(so the manifest will have some guile code to save the transformation to be applied to package install)
<Fare>how do I determine which package contains a given binary, say bluetoothd or pacmd ?
<Fare>(trying to connect to my bluetooth headset)
<Fare>ahem... if I try to run /gnu/store/a900m4ql85lj7y2p10n42yp2gpdw53pi-bluez-5.55/libexec/bluetooth/bluetoothd I get the error: D-Bus setup failed: Connection ":1.737" is not allowed to own the service "org.bluez" due to security policies in the configuration file
<Fare>how am I supposed to start it? guix says it started it, but no process is running and trying to run manually leads to the error above
<Fare>am I missing some polkit configuration?
<Fare>herd: failed to start service bluetooth
<Fare>dmesg shows some firmware being successfully installed.
<Fare>Aha, I see in /var/log/messages: Aug 9 23:15:15 localhost bluetoothd[13051]: src/main.c:parse_controller_config() Key file does not have group “Controller”
<Fare>if I add a [Controller] section it complains about missing fields, but most importantly in the end: Aug 9 23:21:49 localhost bluetoothd[13486]: src/main.c:main() Unable to get on D-Bus
<the_tubular>Sorry I know nothing about bluetoothd, I should learn it though
<Fare>this looks more like some kind of missing or broken dbus policy
<Fare>however, I don't see any significant discrepancy wrt what's in nixos's /etc/bluetooth or /etc/polkit-1/ or /etc/dbus-1/
<Fare>though /etc/dbus-1 has a lot of crap I don't understand
<bsturmfels>hi folks, under "Application Setup", the manual recommends running "nscd" on foreign distros. Is it ok to run the foreign distro's version, or does it need to be Guix's version?
<Fare>aha, so I had to restart dbus so it would read its configuration file which would enable bluetoothd to connect to dbus... and restarting dbus killed the X session. Oh well. I'm back with working bluetooth.
<podiki[m]>bsturmfels: I don't know the difference, but I ran the foreign version (and that seemed to make sense to me)
<podiki[m]>I don't think you'll be running services from guix on a foreign distro, since that is through the system reconfigure part of guix
<podiki[m]>(and using herd)
<bsturmfels>podiki[m] thanks
<Fare>Now, that my headset is connected, I'm trying to tell pulseaudio about it, but pacmd tells me: "No PulseAudio daemon running, or not running as session daemon." Yet there is a pulseaudio daemon running... owned by gdm (!)
<Fare>and if I try to start my own pulseaudio, of course, I get rejected.
<podiki[m]>Fare: I think pulseaudio also didn't see devices at first for me? but now works (and I didn't manually add the service or anything....) sorry
<Fare>podiki[m], do you know what you changed? Just rebooting?
<podiki[m]>oh maybe I manually installed pulseaudio too? (I see it listed as installed, don't remember though)
<podiki[m]>maybe there is a pulseaudio service in desktop-services, but ifyou don't have pulseaudio installed...
<podiki[m]>but yeah, maybe a reboot. I use pavucontrol, for what that's worth
<Fare>there is indeed a pulseaudio service there, and I suppose that's how gdm got to start it
<Fare>rebooting... see you!
<the_tubular>Could someone help me make a VM using guix. I've read on guix WM and I'd like to pass the VM a few NiCs, how would I go about doing that ?
<apteryx>the VM produced with 'guix system vm' is started by a QEMU script
<apteryx>you could study the script and plug your extra interfaces. Unfortunately the QEMU command line is rather obtuse. I usually refer to 'info (QEMU)'.
<apteryx>the_tubular: ^
<apteryx>you could also produce a VM disk image (using 'guix system image'), and import this into some more user friendly frontend such as virt-manager, which should allow you to do this, I believe.
<fsg>Is it possible to run 'guix system' for creating a container on a non guix OS?
<fsg>I would like to try the distro on a container rather than a full VM
<abrenon>hello guix !
<leoprikler>fsg yes
<leoprikler>I think the only commands you can't run are init, reconfigure, list-generations, delete-generations and rollback
<leoprikler>though perhaps I misses some generation command
<fsg>leoprikler: nice, thanks for the info
<the_tubular>Sorry apteryx, completly missed your pings. guix system image looks very good, I'll have to get more info about that
<civodul>Hello Guix!
<sneek>Welcome back civodul, you have 1 message!
<sneek>civodul, iskarian says: I haven't had a chance to test it yet, but the fix looks good (for the medium term anyway; I'm not happy with Guix not being able to present everything that will be downloaded at once, and would very much like to fix that long term)
***nf__ is now known as nf
<civodul>does someone remember the story with the Python test_multiprocessing_fork test that would never complete?
<civodul>i was looking at adding Python 3.6 to Guix-Past and stumbled upon that
<leoprikler>should there not be a fixed version somewhere in the git tree before the 3.7 update?
<efraim>I would package the latest 3.6 python and not the version we had, hopefully whatever it was is fixed in that version
<civodul>i took 3.6.5 and its patches as i found it in the Git history
<civodul>but that test hangs
<civodul>i think that's the latest 3.6.x no?
<muradm>hi guix
<abrenon>there I go, having after thoughts of transforming a plain assoc of known opam repositories into a function
***muradm` is now known as muradm
***srk_ is now known as srk
***Lightsword_ is now known as Lightsword
***irfus- is now known as irfus
***avoidr_ is now known as avoidr
***daviwil_ is now known as daviwil
<surfit>I'm trying to build a historic package that has a patch in the historic source-tree, but not in my current guix install -- how can I point guile's patch path at my local patches?
<surfit>reference https://pastebin.com/mzX0L4Rn
<raghavgururajan>Hello Guix!
<MysteriousSilver>\o
<raghavgururajan>Anyone having USB 3.0 ports and running Guix System, with UAS/UASP (https://en.wikipedia.org/wiki/USB_Attached_SCSI) enabled? The command `lsusb -t` should reveal the information.
<muradm>in (substitute* <file> ...) how to match whole line? should this work? (substitute* "hello.c" (("^ switch = 1;$") " switch = 0;"))
<muradm>while in hello.c there are switch = 1; with 4-8-12 spaces in before, i want to replace the one with 4 spaces only
<muradm>it does not
***ggoes_ is now known as ggoes
<leoprikler>muradm, i think in this case writing a patch might be the easier thing
<muradm>leoprikler: writing patch is hard :) can't add patches to my current system
<muradm>at least i don't know
<leoprikler>uhm, is this inside guix source or is it for a personal channel?
<muradm>i don't have personal channel, i do like "sudo -E guix system -L ~/.config/guix reconfigure ..."
<muradm>managing a channel is a bit of overhead
<muradm>pam_env.so does not provide conf file to add more complex variables
***lukedashjr is now known as luke-jr
<MysteriousSilver>not a guix question, but what would the ideal format to archive files? (in terms of minimal requirements/dependencies for extracting)
<MysteriousSilver>s/the/an
<abrenon>what's needed for the ocaml repl to get readline to work ?
<abrenon>it looks like it's called rlwrap
<abrenon>I wonder why that isn't directly included with ocaml ?
<roptat>hi guix!
***jackhill_ is now known as jackhill
<abrenon>hi roptat
<roptat>abrenon, just saw you sent a new version, thanks!
<roptat>again for the commit message, make sure to end your sentences with a full stop ;)
<abrenon>ohhh : / sorry
<abrenon>I was kinda confused by this first line thing (I hadn't seen it ended up in a very broken Subject)
<roptat>it's ok, if it's the only issue, I can fix it before I push
<abrenon>so I suppose you are refering to the description along each modified path ?
<abrenon>hmmm even the first one
<abrenon>ok
<abrenon>sorry about that
<roptat>yeah, even the subject should end with a dot
<roptat>in the documentation, you should use two spaces between sentences
<roptat>a known repositories -> a known repository
<abrenon>will fix
<roptat>it's not clear from the doc that you can use --repo multiple times
<abrenon>"This can be used several times and […]"
<roptat>also "The list searched always includes @code{opam}" doesn't feel right, but I'll need help from a native speaker to make it sound good (or tell me it's fine)
<roptat>bah, nevermind ^^'
<abrenon>you're right, it sounds bad even in french
<roptat>how could I forget that sentence 30s after reading it ^^'
<abrenon>it sounds like a weird translation of (::["opam"])
<abrenon>and I forgot to re-check my code with the auto-indenter
<abrenon>and given the amount of modification in the importer itself, I'm pretty sure there will be a diff
<abrenon>it's really tiring to have to review its output by hand and discard most of it
<roptat>maybe "The list of repositories always includes @code{opam}, so calling this importer with @code{--repo opam} is redundant."
<roptat>you can keep "although valid" if you like it
<abrenon>what's weird is it really looks like a search path but it's not really a search path
<roptat>oh, and I think you use the wrong word order whenever you talk about options or files: "the sub-directory @file{packages/}" should be "the @file{package/} sub-directory"
<roptat>yeah, I think you can just call it "the list of repositories"
<roptat>Add the given repository to the list that will be searched -> Add the given repository to the list of repositories in which packages will be searched for?
<roptat>it's maybe not super important, the main message is there, but it's better if it sounds nice :)
<abrenon>what about the translation ? will it all happen on weblate ? or must I find a way to edit the fr.po ?
<roptat>no, it'll all happen on weblate after we push
<roptat>now for the code, I see you used (cache-directory #:ensure? #f), why not ensure?
<abrenon>hmmm also, I think I didn't use the right textwidth for the documentation : (
<abrenon>I have no idea, I think I copy-pasted this part, I don't properly understand what it does
<roptat>I guess it's fine because you mkdir-p the directory if it doesn't exist
<abrenon>yeah but if there's no reason for to do it, we should do it
<podiki[m]>does anyone have any examples of modify-services for dbus? or examples of how they use dbus-service (seems like the modify form is different than others, and I wrote legal scheme but didn't get new service files added to dbus....)
<abrenon>I suppose breaking a @code in the documentation ?
<abrenon>or other, btw
<abrenon>I mean any @<tag>{…}
<roptat>you can have a line break in the middle
<abrenon>sorry for asking, but since I need to break lines and indent by hand, I was wondering if I could stuff the begining of a @code{ before the new-line
<roptat>podiki[m], if you only want to add to dbus, you can extend it instead of modifying it. I use (simple-service 'dconf dbus-root-service-type (list dconf)) for instance
<roptat>abrenon, yes
<abrenon>is there a way to render the documentation locally and see what's the output of the texi I'm typing ?
<abrenon>I saw make update stuff but I didn't find the corresponding output
<roptat>you can "make doc/guix.info" or "make doc/guix.html" and view the result with info or your browser
<abrenon>also apparently I didn't replace the old description, I left the description of the names… : (
<podiki[m]>roptat: thanks! related question, confused a bit over docs vs code in where it will pull the dbus service files from. comments seem to say package need to provide them in `/etc/dbus-1/system.d`, but codes looks to be from `/share/dbus-1/\\.service`
<abrenon>wow, this version of my patch was complete trash actually
<roptat>podiki[m], I see dconf doesn't even have an etc directory, so the code is right: share/dbus-1
<podiki[m]>roptat: okay, bug in the doc strings then (as I thought). and possibly a package I'm trying needs modification, as it puts them in `/share/dbus-1/services/` I think (not at that computer currently)
<roptat>that's the same directory as dconf, that should be right
<roptat>the find-files can look for files in subdirectories
<podiki[m]>for context this all comes back to xdg-desktop-portal for things like clicking links in flatpak apps; I've seen it come up in the IRC logs but never with an answer of anyone actually having it work
<podiki[m]>ah
<podiki[m]>okay, let me try with the dbus-service line you gave me, not sure what I did, certainly didn't pull in the packages service files
<roptat>(haven't seen the code, but I expect something like (find-files "share/dbus-1" "\\.service") which means find any file named *.service in share/dbus-1
<roptat>abrenon, what's this "(('remote #t) #f)" case?
<podiki[m]>you are right, it is in a find files
<abrenon>I realised that my repo-type had to return something that wasn't a proper uri
<abrenon>I mean, get-uri can fail
<abrenon>which is why I had a throw, which became a warning, because actually there's no use failing for a mere fishy repository
<abrenon>the others may be enough to find the package
<abrenon>that's what happens l. 142 in the or
<abrenon>I would have found it more intuitive to return #f, but apparently this is what (warning …) returns
<abrenon>so, since this can't be confused with any other case, instead of wrapping the warning inside a begin to return #f, I matched against this case in get-opam-repository
<abrenon>this change goes along with the replacement of map by filter-map in opam-fetch
<abrenon>which used to believe it could trust all its repository specs, but since I know accept one to fail, the mapped list must be filtered
<abrenon>(which in turn justifies why that case in get-opam-repository returns #f)
<roptat>I see
<abrenon>maybe a 'bad-repo would be more explicit
<roptat>maybe you could add a comment about that?
<abrenon>wouldn't a symbol be clear enough ? I like code to document itself
<roptat>sure
<podiki[m]>roptat: thanks, that pulled in the dbus service file. xdg-desktop-portal still doesn't seem to work, but now at least I think it is set up
<roptat>podiki[m], maybe you need a reboot?
<podiki[m]>does anyone here use xdg-desktop-portal? we have the package and it has been updated, so I assume it works somehow. I'll keep trying and if not hit guix-help
<podiki[m]>roptat: yes I did, since dbus needed a restart. I see it gets the service file from the package. but doesn't find the handler it wants for the message it needs to handle (eg open a url). maybe needs some more setup for guix
<roptat>abrenon, oh and on the CLI help line, maybe add "can be used more than once"?
<abrenon>while writing the documentation I realized I had erased coq-released and friends
<abrenon>I briefly considered using a function instead of an assoc for know-repositories
<roptat>I think the assoc list is fine
<abrenon>but it seemed shaky and less satisfying conceptually
<abrenon>about the help message, would you add that on a separate line or within the parentheses ?
<roptat>on a separate line, outside the parentheses
<abrenon>are option messages expected to be proper sentences ?
<roptat>you'll need to match indentation too
<abrenon>I see no punctuation
<roptat>no
<abrenon>no capitals so I'd risk a no
<abrenon>: )
<abrenon>the bad thing from not working on branches but sending "all-in-one" patches is it's hard to follow if all issues have been properly fixed
<abrenon>I think I'll remove repo-type, it's just a match after get-uri
<abrenon>no it's just plain ugly
<podiki[m]>related note roptat, should we be providing configuration tips in package descriptions? I saw a couple that gave a code snippet for adding the service provided by the package, or other notes. I think it would be useful, even something just like "activate this service by adding it to your dbus-service" (with details in manual or cookbook)
<podiki[m]>or maybe even some metadata in packages so a `guix package` command can say "you've added packages with dbus components, you may want to modify your system configuration to use them" (or on removing, to remove them)
<roptat>sure, that sounds useful
<roptat>although, you don't want to install the package just to provide dbus services, you only need to add it to the dbus service
<muradm>hi guix
<muradm>suppose i what define service-type
<muradm>which can be used multiple times
<muradm>but multiple instances share same extensions
<muradm>in service-type i add (service-extension etc-service-type my-service)
<muradm>my-service makes single file that is shared between multiple instances of my-service-type
<muradm>currently i get "duplicate: my-service.conf in /etc"
<roptat>maybe you could have a single service that gets extended, and runs multiple instances, but extends etc-service-type only once?
<muradm>roptat: could you point to such example?
<roptat>I can't think of any
<muradm>in guix source base
<Guest69>Hi
<roptat>hi Guest69 :)
<Guest69>How to use intel-vaapi-driver package with xorg-configuration procedure?
<podiki[m]>roptat: ah good point! well then in the package description would be the most useful to make more common. then when a user searches for something and sees it, they can use that info (I think I just proved how someone new to a package would have no idea)
<podiki[m]>since this is already done in some, would be good to keep an eye out and update descriptions. assuming I get xdg-desktop-portal working, I can submit that easy patch
<roptat>podiki[m], thanks!
<podiki[m]>in general I think we need more examples, service config especially. there's some good ones in the manual and cookbook, but I think this would help users know where to look
<Guest69>The XORG_DRI_DRIVER_PATH variable is set inside the xorg-wrapper procedure, which includes [mesa]/lib/dri. How do I extend that variable to include [intel-vaapi-driver]/lib/dri, inside set-xorg-configuration?
<podiki[m]>Guest69: it already should appear in LIBVA_DRIVERS_PATH
<roptat>Guest69, it looks like it's not possible to change that mesa in "(string-append #$mesa "/lib/dri")"
<podiki[m]>if it should also be in XORG_DRI_DRIVER_PATH then I think the package definition should add that native-search-paths (see gnu/packages/video.scm for the package)
<roptat>podiki[m], I think we're talking about the generated config in (gnu services xorg)
<podiki[m]>oh
<roptat>we could extend xorg-configuration to also have a field for the default mesa and xkbcomp, and use them instead of the hard-coded packages
<roptat>or maybe it's not needed there, but I'm no expert on this
<muradm>roptat: from <service-type> record (compose is a function Any -> Any
<muradm>what should be the return type of it
<roptat>muradm, so what if your service is a list of configurations, and compose is append? then you can extend with a list of configurations (one per instance)
<roptat>compose takes the service value, a value you want to extend the service with (can be a different type) and returns the new service value
<roptat>for instance, nginx's compose takes the service value (an nginx-configuration record), an extension value (an nginx-server-configuration) and returns a new service value (an nginx-configuration)
<muradm>looks like i'm getting it now, (compose concatenate) will make a list of configs for example, (extend will flatten than list of configs to single config, and then it will be passed to my (service-extension functions...
<abrenon>roptat: what about a completely different approach to the documentation of --repo ? it's too close to the implementation
<abrenon>users don't care that there's a list to represent the repositories searched
<roptat>ah sure
<abrenon>what about something that would go like: "By default, packages are searched in the official opam repository. This option lets you add…"
<roptat>sounds good
<abrenon>thanks for your feedback
<JorgeTern[m]1>Hola,estoy tratando de crear otro perfil de usuario pero no lo encuentro en la guia, alguna ayuda
<abrenon>Hola JorgeTern[m]1
<abrenon>me parece que la opción "users" en el config.scm quiere una lista
<abrenon>pues deberías poder añadir otros ahí ¿ no ?
<abrenon>¿ no funcciona ?
<JorgeTern[m]1>En la terminal ?
<abrenon>ahhhh
<abrenon>no, en guix se hace declarativamente, editando un archivo especial que describe el sistema
<abrenon>¿ has leido http://guix.gnu.org/fr/manual/es/html_node/Cuentas-de-usuaria.html#Cuentas-de-usuaria ?
<abrenon>(y esa pajina describe lo que pasa dentro de (users …) en una configuracion completa, tal como describida en http://guix.gnu.org/fr/manual/es/html_node/Uso-de-la-configuracion-del-sistema.html)
<JorgeTern[m]1>Gracias,voy a leer para ver como lo creo.
<abrenon>de nada
<abrenon>(no toda la gente aqui habla castillano, tendras mas respuestas si puedas escribir tambien es ingles — pero que si no puedes, no te preocupes)
<JorgeTern[m]1>Que paquete usas para traducir ? es dificil para mi
<abrenon>no utilizo nada
<abrenon>(pero no pensaba que mi castillano se había vuelto tan mal ^^ lo siento)
<abrenon>si te cuesta escribir en ingles no importa mucho, hay otra gente quien le escribe, pero no todas las personas aquí
<abrenon>roptat: I found a very good reason for me to use #:ensure? #f also I don't remember having thought about it so this is probably pure luck that I used it
<roptat>oh ?
<abrenon>no wait I'm mistaken, the ensure only covers the cache-directory part sorry ^^
<abrenon>the actual …/opam/<hash> folder must not be created as soon as its name is generated, because on the first run, that would make it created after the archive
<roptat>yeah, I understand
<abrenon>but that's not relevant here
<abrenon>since only the "root" cache-directory is affected
<abrenon>so I think I can safely remove this #:ensure
***avoidr_ is now known as avoidr
<abrenon>hmmm it looks like (and=> …) is the kind of stuff I could've used in all my repository plumbing
<JorgeTern[m]1><abrenon> "(pero no pensaba que mi castilla..." <- jaja todo esta bien,gracias nuevamente por la ayuda,casi siempre encuentro quien me colabore en castellano.
<abrenon>: )
<abrenon>por ejemplo, un paquete no se hubiera equivocado en "castellano" ^^'
<abrenon>creo que estoy demasiado cansada
<abrenon>well good evening everyone then
<abrenon>roptat: thanks again for your patience and help
<vivien>Hello, there’s something I don’t understand with pattern matching. I wrote this: (match "http://sdfqsdf" ((? string? (= string->uri (? uri? x))) x)) to match a string that can be parsed as an URI and use the URI as binding. But why are the predicate in this order (string? left and uri? right)? I thought it would be the opposite: match anything that’s an URI after apply string->uri to anything that’s a string.
<roptat>wow, that works?
<vivien>I think so, but I’m not quite 100% sure
<roptat>yeah, it does, I'm amazed
<vivien>Although this orders make sense too, because (? uri? x) expresses that the x binding is an URI
<vivien>(which is what I want)
<vivien>this order*
<roptat>I think what happens is that it matches the strings with (? string? pat) and matches the string with pat, which is (= string->uri pat), and that matches (string->uri str) with pat, which is (? uri? x), and binds x to the uri
<roptat>from the description in the manual, you're not using it in the expected way: "(= field pat) a ``field'' of an object"
<vivien>Yeah, but look at the example
<vivien>There no "force" field
<roptat>but a field of an object is just a procedure
<vivien>Ooh
<vivien>No, force is not a field
<qzdl>hi guix \o
<vivien>It’s just a function for lazy evaluation
<roptat>you're right, it's already not doing what the description talks about ^^
<vivien>So the description should be "match anything, apply the function, and match the result with pat"
<roptat>I learned something interesting, thanks :)
<vivien>I’ll bug it
<vivien>But first I’ll ask for confirmation about the behavior ^^
<roptat>looks like ci is having troubles building core-updates... https://ci.guix.gnu.org/build/349274/log/raw
<roptat>cannot build missing derivation ?/gnu/store/b5jay6r5yw8ac2ir6ng8qyi91sj6lyr7-tar-1.34.drv?
<roptat>I don't know if that's it, but I see a few "guix substitute: warning: 141.80.167.131: connection failed: Connection refused"
<roptat>same here https://ci.guix.gnu.org/build/349271/log/raw
<roptat>cannot build missing derivation ?/gnu/store/4ik9gb8aipfvi0lvcvcpa32r2ndcfl6h-xz-5.2.5.drv?
<roptat>that's pretty low in the dependency graph
<qzdl>how can I debug the build of a package that is consistently causing OOM errors?
<qzdl>ram isn’t even being filled, and swap is untouched
<cehteh>building fails?
<cehteh>it prolly tries to preallocate or map some huge chunk of memory
<qzdl>the process gets killed, along with my x session
<qzdl>yeah it’s just hard to know where/how/why
<qzdl>could you repro if I post a paste?
<cehteh>i am not som much used to guix, maybe someone else here can, i just know how/why things could faul even when you dont see it
<vivien>roptat, https://paste.debian.net/1207174/
<cehteh>how much ram/swap does that machine have?
<qzdl>32gb ram, 8gb swap
<roptat>vivien, nice :)
<qzdl>unless there’s a catastrophic memory leak I don’t think that’s the issue
<cehteh>anthing constraining the memory? ulimits/container/pam ?
<qzdl>no constraints there
<cehteh>sounds odd indeed
<vivien>I understand how it can be thought as the "field of an object"
<cehteh>qzdl: guix build has a --log-file option
<cehteh>maybe that helps
<qzdl>I’m using -K at the moment to keep everything around
<qzdl>bit of a stupid inquiry, but how can we build a package without supplying the sha? i want to remove the patches I’m applying, but of course it will result in a different sha
<maximed>qzdl: The patches aren't included in the SHA calculation. More specifically,
<maximed>first the source is ‘built’ (i.e., downloaded) (and the hash is checked)
<maximed>then the patches are applied to that
<maximed>(I assume you're referring to the 'patches' field)
<qzdl>okay seems i need to brush up on the manuals more
<qzdl>yep exactly
<muradm>i'm adding a rust package, looking at other's commit messages, they are like "gnu: rust-some-package: Add rust-some-package 0.1.1"
<muradm>why not "gnu: packages: crates-io: Add rust-some-package 0.1.1"
<muradm>or "gnu: packages: Add rust-some-package 0.1.1"
<muradm>is there commit message guidelines?
<maximed>qzdl: fwiw, I don't remember it being mentioned in the manual that the hash calculation doesn't include the patches
<maximed>muradm: maybe "gnu: crates-io: Add rust-some-package" or even "gnu: Add rust-some-package"
<maximed>muradm: I don't it matters all that much, just make an attempt to be consistent
<leoprikler>muradm: fill-column is at 72 for commit headers
<muradm>leoprikler: did i violate it?
<leoprikler>No, it's to answer your question regarding gnu: packages: yadda yadda
<leoprikler>we only use gnu: or guix: (or doc:, etc) as vague hints – there's a full changelog in each commit anyway
<muradm>ah ok, yes i remember, just wandering what is the correct shape of comment
<roptat>mh, the peg parser is not very fast...
<roptat>I'm trying to use (guix build po) instead of our sed commands, because I had some troubles with fuzzy entries before, so I'd like to have a real parser rather than a sed rule
<roptat>but it's taking 75s to parse po/doc/guix-manual.fr.po for instance (or any other guix-manual po file)
<roptat>it's much faster with guix-cookbook, since it's a lot smaller
<roptat>but it works "translated 866 cross-references in 'doc/guix.fr.texi.tmp'"
<raghavgururajan>System doesn't boot after recent pull and system reconfigure. The boot stucks at "Error in finalization thread: success". I saw these warnings while system reconfigure, https://paste.debian.net/plain/1207184
<roptat>hm, down to 4s, which is much more reasonable ^^'
*lfam tests fix for qtwebkit on core-updates
<roptat>I had a look at ant-bootstrap on core-updates, but I don't understand. It's complaining the build directory already exists, so I set an option to make it use build-bootstrapped instead. Then, it complains that build-bootstrapped already exists, which is not the case (checked with -K)
<roptat>I don't see any obvious change on the java side either between core-updates and master...
<lfam>This is trivial, but are you sure that the current working directory is what you expect?
<roptat>the message is /tmp/guix-build-ant-bootstrap-1.8.4.drv-0/apache-ant-1.8.4/build.xml:558: Unable to create directory as a file already exists with that name: /tmp/guix-build-ant-bootstrap-1.8.4.drv-0/apache-ant-1.8.4/bootstrapped-build
<roptat>so I know exactly which directory it tries to create
<lfam>Hm
<roptat>I see this message comes from the test dir.isFile()
<roptat>and the file name is printed as dir.getAbsolutePath
<roptat>so it's really the same directory being tested and printed
<lfam>Remember that Guix sometimes fakes the path so that it always appears to be /tmp/guix-build-foo-version.drv-0 from within the chroot. Even when the 0 is actually another number. It's probably not causing trouble here but it's worth checking if that's the issue
***Kimapr0 is now known as Kimapr
<lfam>It would be really remarkable if that was the problem
<roptat>I only have one directory named /tmp/guix*
<roptat>and it's -0
<roptat>but even in the previous case, build was a directory, so isFile() should have returned false, not true
<leoprikler>does Java isFile() return true on directories then?
<leoprikler>hmm, at least according to the docs it should not
<roptat>no, there's isFile() for files and isDirectory() for directories
<roptat>also, looking at the documentation, it uses S_ISDIR after checking the existence of the file
<roptat>(from classpath-bootstrap)
<roptat>oh, jamvm is buggy
<roptat>so, here is a test java file (called Test.java): https://paste.debian.net/1207193/
<roptat>after making it public, I enter an environment for ant-bootstrap: ./pre-inst-env guix environment ant-bootstrap
<roptat>then I build it with: CLASSPATH=$GUIX_ENVIRONMENT/lib/rt.jar jikes Test.java
<roptat>and I can run it either with "java Test" -> "no" or "jamvm Test" -> yes
<jonsger>hello guix
<roptat>interestingly, jamvm always says yes, whether the file doesn't exist, is a file or is a directory
<jonsger>the icedove update will come later this night, for icedove 91 I need to wait on Icecat...
<muradm>refactored v2 seatd/greetd services is out http://issues.guix.gnu.org/49969
<iskarian>Are grafts always handled locally, or is it possible to get a substitute for a grafted package?
<roptat>jamvm doesn't have an implementation for File though, so it might be that it's using classpath whereas java uses its own implementation
<roptat>oh https://issues.guix.gnu.org/issue/36685
<roptat>it's the same error, I found it in a comment in classpath-bootstrap
<roptat>oh no, it's not the same issue, but so close!
<roptat>mh... I tried the obvious, commenting the next line as with the patch from last time and that didn't work, but printing worked, I'm so confused
<muradm>lfam: what do you think about #49969?
<drakonis>nice.
<drakonis>muradm: is it a drop-in replacement for elogind?
<drakonis>rather, can i replace the service with it and everything will continue working as expected?
<muradm>drakonis: not actually, if you're heavily relying on elogind (like gnome) it will break things i suppose
<drakonis>ah i see
<muradm>loginctl won't work for instance
<muradm>you want seatd if you are minimal i3/sway like
<muradm>but i'm not sure of course, i don't use those, i derive from %base-services rather than %desktop-services
<muradm>seatd/greetd are for those who go minimal all the way
<muradm>and probably anti-systemd'ers
<muradm>:)
<muradm>btw, as far as i see from sway codebase it seems that logind support will be dropped
<muradm>drakonis: here are my services for instance https://paste.rs/3Wa
<iskarian>roptat, I'm curious: what did you do to speed up your peg parser?
<raghavgururajan>lfam: The commit 189003c83ee04e1819245c861c1ba31597db537d is causing this (https://lists.gnu.org/archive/html/help-guix/2021-08/msg00032.html). Any ideas?
<zacchae[m]>Is there a reason guix home hasn't been pulled to the main repo?
<roptat>iskarian, removed some cases that I ignored anyway, to make only one case
<podiki[m]>zacchae: I'm guessing it is going when core-updates-frozen is for the next release?
<podiki[m]>but I don't know
<zacchae[m]>oh is there a schedule for that, or just whenever it happens?
<zacchae[m]>thanks for the info
<podiki[m]>again I don't know, but my guess is once core-updates-frozen builds successfully at least. someone who knows can chime in
<podiki[m]>question: if you've added to %default-substitute-urls in system configuration, then `guix weather` should use all those urls by default, no? I only see it querying the default 2
<muradm>(source ... (patches (list (local-file "some-patch.patch")))) is possible?
<lfam>raghavgururajan: Hm, I don't see how that commit could break booting
<muradm>when i do that, patch is applied, but rust package fails to compile
<muradm>something with checksums? what does patch-cargo-checksums do?
<muradm> https://paste.rs/1I5
<roptat>podiki[m], right, maybe you need to restart the guix daemon?
<podiki[m]>I had changed it many configs/reboots ago..... bug?
<roptat>share your config?
<lfam>raghavgururajan: Those warnings from the reconfiguration look suspicious. I don't think I've seen them before
<roptat>podiki[m], do you get substitutes from the other servers you configured, when you don't use weather?
<podiki[m]>roptat: sure one sec. it follows the manual for modifying desktop services in substitute section
<podiki[m]>the relevant part: http://paste.debian.net/1207200/
<podiki[m]>in guix weather.scm: `(define %default-options
<podiki[m]> `((substitute-urls . ,%default-substitute-urls)))`
<roptat>ah, that might indeed be an issue with guix weather
<podiki[m]>roptat: yes, can see eg guix build and system reconfigure checking the urls at least