IRC channel logs

2020-05-13.log

back to list of logs

<ngz>Hmm. I'm picky, but I think the manual contains wrong uses of @command{...}
<rekado>ngz: may be so!
<ngz>E.g., @command{man wpa_supplicant} should be @samp{man wpa_supplicant}. Ditto for @command{guix system init}
<ngz>I'm not so sure for, e.g., @command{guix package}. One might consider it to be a single command…
<raghavgururajan>Guixguy: That would the target commit's hash-string from the target git repo.
<rekado>The Texinfo docs say “When writing an entire shell command invocation, as in 'ls -l', you should use either '@samp' or '@code' at your discretion.”
<raghavgururajan>*would be the
<ngz>rekado: yes, and it also says that @samp looks better than code for entire shell commands. See (info "(texinfo)@samp")
<ngz>OTOH "guix package" is not an entire shell command.
<ngz>Hmm. I think it makes sense to use @command for Guix sub-commands, e.g., @command{guix gc}, but I suggest to switch to @samp for complete calls, e.g. @samp{guix gc --list-failures}
<Guixguy>raghavgururajan: Do I guix-download my git repo and use that hash?
<whk>lfam: I sent message to 41217-done@debbugs.gnu.org but didn't get an acknowledgment nor does the bug seem to be updated. Are special privileges required to use #####-done@debbugs... vs just #####@debbugs... when updating?
<raghavgururajan>Guixguy: You are trying to use it in your package definition correct?
<Guixguy>Yes
<raghavgururajan>Guixguy: OK. Then, https://guix.gnu.org/manual/en/html_node/origin-Reference.html#origin-Reference
<raghavgururajan>That is the syntax.
<raghavgururajan>the commit can be a tag or hash.
<Guixguy>raghavgururajan: Thank you. I'll read it over.
<raghavgururajan>Yw!
<Guixguy>raghavgururajan: Sorry, I'm lost. I don't see commit spoken about on this page.
<raghavgururajan>Guixguy: No worries! Could you send the link of the git repo you are trying to use?
<Guixguy>raghavgururajan: https://paste.debian.net/1146458/
<Guixguy>Here is my code with the repo in it
<Guixguy>raghavgururajan: Congratulations on your internship =]
<raghavgururajan>Thank you!
<raghavgururajan>So, you cannot use tags because the repo does not have one. https://gitlab.com/kozodev/guix/-/tags
<raghavgururajan>You can see hash mentioned for each commit in this page. https://gitlab.com/kozodev/guix/-/commits/master
<raghavgururajan>You can use one of them (probably the latest).
<raghavgururajan>In your package definition, replace v1.0 with a hash.
<Guixguy>I see that the hash updates when I do a git push and shows in the terminal
<Guixguy>Very cool
<Guixguy>raghavgururajan: Thank you again, I'll play around with this.
<raghavgururajan>You're welcome!
<civodul>rekado: GC is at 50% of the deletion phase
<Guixguy>OH, people tag their releases with a version on git
<reepca>we seem to only have man pages for some of the pthread functions with the man-pages package
<reepca>for example, we have pthread_create(3) but not pthread_key_create(3)
<reepca>or are those glibc-specific extensions...?
<civodul>reepca: no, pthread_key_create is POSIX i think
<civodul>dunno why these pages are missing
<reepca>I can find a brief overview in the glibc info manual
<civodul>ok
<civodul>so Cuirass should restart in 4h
<civodul>until then
*civodul -> zZz
<eta-expanse>is building a package from a file equivalent to pulling the same file from a channel and installing it? or does installing do extra stuff like with the profile?
<atw>eta-expanse: I think it's accurate to say that building creates items in the store, but nothing more: nothing is added to your profile, a new executable won't be available from the command line, etc. Installing is (I believe) equivalent to adding stuff to the store (if necessary) then adding it to your profile. Does that help?
***catonano_ is now known as catonano
<eta-expanse>atm: yes that clears it up. I was wondering why my package built but none of the binaries, man pages etc were available
<lispmacs>xelxebar: KE0VVT: all my font issues were fixed after running the command fc-cache -f -v
<atw>eta-expanse: I take it you did guix build -f ?
<eta-expanse>atw: yes
<eta-expanse>also explains why the propagated inputs didnt install
<gry>does someone here speak Russian?
<Gooberpatrol66>is there a rough ETA on when KDE will be added?
<xelxebar>lispmacs: Yeah, that sounds about right. Thanks for sharing
<sneek>Welcome back xelxebar, you have 1 message!
<sneek>xelxebar, bricewge says: You were right about not “mucking about with http_proxy”, using it will query DNS in clear. Better to use torsocks as you suggested.
<xelxebar>sneek: botsnack
<sneek>:)
<ecbrown>Gooberpatrol66: don't hold your breath. kde is massive and complicated
<whk>New user question: New 1.1.0 vm created from iso with xfce as window manager. The system updated running "guix pull" and then "sudo guix system reconfigure /etc/config.scm" (per manual). After reboot and login all font characters are replaced by square with four dots (character not available?). I assume this is a locale issue since between guix pull and the reconfigure I got "guile: warning: failed to install locale" but answer to <htt
<whk>ps://debbugs.gnu.org/cgi/bugreport.cgi?bug=40826> seems to imply that is currently normal. Am I missing a step?
<whk>I can get in via ssh if there is a command to fix?
<ecbrown>whk: one thing to try is fc-cache -rv
<ecbrown>but perhaps its missing fonts. (not an xfce user, just want to give you something to try while waiting for an expert.)
<ecbrown>the locale thing is (in)famous by now
<whk>ecbrown: Thanks I just tried doing the "guix install fontconfig and then fc-cache -rv" as the user and "fc-cache -rv" as root but no difference. Agreed on missing fonts for xfce after update. gdm is fine for the login screen and sets font path but I haven't found where my login xfce session is being logged yet.
<Guixguy>Hey all, I'm hoping someone recognizes this error and can tell me what I'm missing. I am trying to build a package.
<Guixguy> https://paste.debian.net/1146489/
<Guixguy>This is my package: https://paste.debian.net/1146490/
<whk>Is there a way to see the packages installed from the iso? If I run "guix package --list-installed" as user I only get the "fontconfig" from above and as root I get nothing. Trying to see if an xfce4 package is missing.
<xelxebar>Guixguy: Looks like an argument you're providing to string-append is evaluating to #f instead of a string
<Guixguy>xelxebar: The strange thing is I'm not using a string-append in my code
<xelxebar>Oh, right. I just looked at your package definition
<Guixguy>whk: If you guix search xfce, you could look at the dependencies it pulls?
<xelxebar>Guixguy: I'm not exactly sure of that error, but there are some issues with this package definition.
<raghavgururajan>whk: You mean packages installed system-wide (i.e. based on config.scm)?
<ecbrown>whk: those would be as root
<ecbrown>whk: as root, guix package -I
<xelxebar>Guixguy: file-name doesn't select a file out of a git repo, it tells guix what to name the checkout directory
<Guixguy>xelxebar: oh
<xelxebar>Guixguy: Second, supply the full hash commit when referencing git checkouts.
<xelxebar>Guixguy: Third, I'm pretty sure guix lint will complain about your embedded quotes in the synopsis and description. Try using @code{..} fragments.
<whk>raghavgururajan: Yes.
<whk>ecbrown:
<whk> root@guixFirst ~# guix package -I
<whk> root@guixFirst ~#
<whk>Guixguy: Thanks that shows dependencies. Still not sure what is currently running though. I just tried installing xfce package manually but no change.
<xelxebar>Guixguy: That said. I'm not sure how you'ure supposed to actually go about packaging up guix modules...
<raghavgururajan>whk: If you go the directory "/run/current-system/profile/bin". You can see list of programs installed in system.
<Guixguy>xelxebar: How do I find the full hash of a commit?
<xelxebar>git rev-parse
<xelxebar>That, or using the gitlab interface, looks like you can click on "commits" from the main page and clicking on the one you want shows the full hash in the url
<Guixguy>xelxebar: Ah! I was on the webpage, didn't notice the url
<Guixguy>Thank you =]
<whk>raghavgururajan: thanks I will look there
<xelxebar>Guixguy: Also, probably don't want to leave homepage empty. Mimicking other guix packages, you can just put the gitlab url in there if that's effectively it's only homepage
<raghavgururajan>Guixguy: You can also click the scriptpad icon next to the short hash version, which copies long hash version to clipboard.
<xelxebar>Also, for some reason it looks like you have a bare "uname-s" symbol sitting there at the bottom of your package definition.
<Guixguy>xelxebar: I am testing with guix build
<Guixguy>raghavgururajan: Thanks, that is good to know
<Guixguy>xelxebar: Good idea on the homepage being gitlab
<xelxebar>Guixguy: You might gain a lot more mileage out of trying to mimic other similar packages.
<xelxebar>I make extensive use of grep on gnu/packages
<Guixguy>xelxebar: I'll give that a shot. Thank you for your time and patience
<whk>To see the base installed packages: "guix package --list-installed --profile=/run/current-system/profile/share". Still doesn't fix my font issue but it allows me to compare installed. Worse case I will rebuild and compare before and after update. Thanks for the help.
<whk>correction "guix package --list-installed --profile=/run/current-system/profile"
<Guixguy>whk: There has been a lot of font talk lately in IRC. If you come back during the Euro day, you'll probably get more confirmation about it
<raghavgururajan>whk: No, I meant to manually visiting the directory.
<xelxebar>whk: When you say "all font characters replaced by..." exactly which programs/environments are they not showing? I assume you mean something like "everything in X"
<whk>raghavgururajan: Understood. I looked at the directory and installed programs. Then backtracked to figure out how to check the version of the packages installed on the base system by the iso since they don't show up with "guix package --list-installed" as either user or root. So far what I can't find is the log file from the X Server for my session. I don't see in in /var/log and am still looking to see if it is in /run/...
<whk>xelxebar: Font is fine on gdm login screen. After login all font characters (menu's, file manager, terminal) are the box.
<raghavgururajan>whk: Try this. Add font-dejavu to packages section of your config.scm and so system reconfigure. Dejavu is the fallback font for Xorg server.
<whk>raghavgururajan: trying
<Guixguy>xelxebar: Do you git clone guix core to grep gnu/packages?
<xelxebar>raghavgururajan: But if gdm is working fine, that probabaly means that the fonts on the system profile are fine, no?
<raghavgururajan>If things don't work after reconfigure, try fc-cache -rv again as root and user. Then reboot.
<xelxebar>Guixguy: Yeah. Isn't that how you're developing packages??
<raghavgururajan>xelxebar: Probably, but gdm (gnome packages) use font-canberra.
<raghavgururajan>which might be installed with gdm-service.
<Guixguy>xelxebar: No, I'm making them with my git repo on gitlab
<xelxebar>In that case, I'd first try setting your wm's default font to canberra, just to see if it's a missing font issue...
<raghavgururajan>Oh. the command `fc-list`should list all installed fonts.
<xelxebar>Also, you really shouldn't need to reboot after reconfigures unless you want to pick up a new kernel.
<raghavgururajan>But some services require reboot to restart.
<xelxebar>Or just a manual restart...
<raghavgururajan>Ah yes.
<xelxebar>Also, we're installing fonts, not mucking about with system services anyway.
<xelxebar>Also... doesn't guix already rebuild the font cache when installing new fonts? I bet there's a hook for that. Just guessing though.
<raghavgururajan>I think so. I saw that hook somewhere.
<whk>Adding font-dejavu /etc/config.scm and reconfiguring plus rerunning fc-cache -rv has font working in new terminal window. So I assume reboot will fix for all.
<raghavgururajan>But not sure.
<whk>Current fonts installed in base are...
<whk>$ guix package --list-installed=font --profile=/run/current-system/profile
<whk>font-abattis-cantarell 0.111 out /gnu/store/iahd565pyyamqsb4cz32lw793di15lkc-font-abattis-cantarell-0.111
<whk>font-dejavu 2.37 out /gnu/store/7y3lvk3xf4im8n44337mc6y0ccysvfia-font-dejavu-2.37
<whk>So tomorrow I will do a reinstall and check to see what font's were installed before doing the "guix pull" and reconfigure.
<raghavgururajan>s/canberra/cantarell
<whk>That was copy/paste from the command???
<raghavgururajan>Sorry, I meant to say that for my previous message.
<whk>Rgr. Looks ok after reboot. I am now going to destroy image and try again and see if I can capture the problem for a bug report. Thanks again.
<raghavgururajan>Welcome!
<Guixguy>xelxebar: Thanks for your suggestion. I am cloning the core project now
<xelxebar>Guixguy: No worries. BTW, if you're developing guix modules, is there any particular reason not to just develop directly in-tree? Doing so, you could add a module, make, and then do ./pre-inst-env guix repl and have your module available right there.
<Guixguy>xelxebar: Mainly because I don't know how to do that =P
<xelxebar>Guixguy: Well, just clone the repo and develop away. Then to build, just start a guix development environment with "guix environment --pure guix" and run make
<xelxebar>Since you are starting from a pristine repo, you'll need to run ./bootstrap && ./configure --localstatedir=/var before make
<xelxebar>Anyway, there are actually some good videas explaining how to get started with guix development: https://guix.gnu.org/videos/
<Guixguy>Excellent!
<Guixguy>xelxebar: Thank you very much. This is going to be a tomorrow project. Night Guix
<xelxebar>Night!
<warp>is there a difference between (assoc-ref %outputs "out") and %output ? I'm seeing packages that use both
<xelxebar>warp: Well, the full answer is in guix/devirations.scm around line 1282
<xelxebar>But AFAICT (assoc-ref %outputs "out") looks to be equivalent to %output
<warp>xelxebar: thank you, I'll take a look
<KE0VVT>lispmacs: Are you on a foreign distro?
***rekado_ is now known as rekado
<dissoc3>has anyone used the hackrf package in guix?
<dissoc3>in the package description it has some further instructions that im a little confused about
<xelxebar>dissoc3: I've not used it, but what are you confused about in particular?
<guix-vits>xelxebar: "(services (modify-services", 100%
<xelxebar>What is guix-vits?
<guix-vits>xelxebar: `guix show hackrf`
<guix-vits>it's the only (potentially) confising part.
<dissoc3>i dont understand how cons hackrf just adds any udev rules?
<dissoc3>(rules (cons hackrf (udev-con...)))
<xelxebar>guix-vits: Oh. I thought you were a bot or something. heh
<xelxebar>dissoc3: Yeah, I agree that configuration template is confusing. I assume that udev-configuration-rules is bound to the udev configuration and hackrf is bound to the rule for hackrf
<xelxebar>Looks like the packages puts the udev rules under /lib/udev/rules.d, so perhaps you could use file->udev-rule to convert that and stuff it in your udev rules?
<xelxebar>The guix manual on Base Services might be helpful, since it gives some examples of munging udev configuration
<xelxebar>dissoc3: Sorry I can't be of more help
<efraim>I'll be at my computer in a few minutes, I can take a look at it and try to make a code snippet
<guix-vits>aren't '(rules' in '<udev-configuration>' accept <package> (as comment suggests in base.scm)?
<mroh>good morning guix!
<bricewge>dissoc3: I added a helper function in Guix to help such use case. Just add the following to your service list “(udev-rules-service 'hackrf hackrf #:groups '("dialout"))”.
<divansantana>With the latest update, my flatpak packages are not working. bwrap: execvp xdg-dbus-proxy: No such file or directory
<divansantana>Seems a xdg-document-portal is needed
<divansantana>Seems a xdg-document-portal service is needed
<rekado>divansantana: looks like we should just propagate xdg-dbus-proxy
<efraim>oh i see there's already a snippet in hackrf
<divansantana> https://github.com/flatpak/flatpak/issues/1359
<divansantana>They suggest systemctl --user restart xdg-document-portal
<guix-vits>efraim: so <udev-configuration> accepts list of <package> to (rules ?
<divansantana>or xdg-document-portal --replace . I can't find xdg-document-portal binary or service.
<efraim>guix-vits: I'm not sure, but I assume the snippet is correct.
<bricewge>guix-vits: You are correct
<divansantana>How could I install / downgrade just the flatpak package without reverting the whole system upgrade.
<rekado>divansantana: you can use inferiors
<rekado>or guix time-machine
<divansantana>rekado: Thanks. Let me read the manual.
<bricewge>divansantana: Looks like the systemd service in xdg-desktop-portal are only “D-BUS Service” so maybe we don't need them and just adding xdg-desktop-portal to the dbus-root-service-type will be enough.
<divansantana>bricewge: I'll test a patch :)
<bricewge>divansantana: I spoke too fast, there are indeed services in there that should be run as shepherd services.
<dissoc3>hackrf works with adding that snippet in the config. I was just confused how adding hackrf to the rules actually worked
<dissoc3>confusing I guess because I've added some other udev rules by manually writing out the strings to put in rules
<bricewge>dissoc3: The description in misguiding too “To install the hackrf udev rules, you must add this package as a system service via modify-services”
<bricewge>It should be “you must extend 'udev-service-type' with this package”.
<dissoc3>i see. thanks for clarifying
<guix-vits>+1
<dissoc3>it worked fine. i was just confused and I was going to write a similar package for a similar device (bladerf)
<lprndn>Hello Guix!
<bricewge>Helllo lprndn !
<bricewge>How is it going? I saw you asked about account-service the other day, did you managed to fixed it?
<bricewge>Also since core-update lightdm as some issue with fonts, all character are replaced with boxes.
<bricewge>Maybe I should pick your last revision of the patchset.
<lprndn>bricewge: Fine and you? I more or less solved it with a simple wrapper. I opened an issue (41207) in case someone has a better idea.
<lprndn>hum...
<lprndn>I'll test again on latest master.
<bricewge>lprndn: Good thanks. #41207 looks ok, I'll give it a try.
<lprndn>bricewge: also it needs a second patch you can find in the lightdm patch serie. I shamefully forgot to add an extension to polkit...
<lle-bout>test
<bricewge>lle-bout: Hey there!
<bricewge>lprndn: It may be usefull to have a new patchset revision since Ricardo has merges some patch in the series there are some conflict when applying the patch IIRC.
<rekado>I’m still working on lightdm, by the way
<lprndn>bricewge: you mean the polkit patch?
<lprndn>rekado: hey! If you need anything, I'm here!
<bricewge>lprndn: No I mean bed695aa94cd85800ec2c6296fb2d13e7ac29133 and the 4 commit preceding it
<rekado>I’m just slow :-/
<bricewge>lprndn: You'll find out when rebasing on master I guess
<lle-bout>bricewge: hello 😄 - was testing if ERC reconnected properly after suspend.
<rekado>jonsger: have you been able to confirm this problem? https://issues.guix.gnu.org/41138#1
<jonsger>rekado: no. I'm running IRC in icedove for serveral days. I don't see CPU spikes. But thunderbird can sometimes be a little bit fishy. The only solution is to restart. This was also the case for my on openSUSE from time to time
<rekado>jonsger: would you like to comment on that issue?
<jonsger>sure
<rekado>thanks!
<lprndn>bricewge: yeah, I see... Is it possible it's because they're already on merged on master?
<lprndn>* merged on masted
<bricewge>lprndn: Yes rekado already merged the package related patches of your patchset
<bricewge>sneek: later tell sirmacik How did you installed network-manager-vpnc? I think ti should go in network-manager-configuration-vpn-plugins field https://guix.gnu.org/manual/devel/en/html_node/Networking-Services.html#Networking-Services
<sneek>Will do.
<bricewge>sneek: botsnack!
<sneek>:)
<jonsger>rekado: I guess its not a bug at our side, more a general thunderbird hick up...
<sammich>hey, when i run pinentry-gnome3 i get 'No $DBUS_SESSION_BUS_ADDRESS found' and it runs curses pinentry, any idea how to fix this? I have dbus in my services in the system config
<wingo>sammich: i think that means your session (!= system) didn't start dbus, somehow
<wingo>if you are in gnome it should have started dbus and everything
<sammich>ooh ok i see, im using sway started from the tty
<abralek>Hi, How do you guys maintain your channels for elpa for example. Do you just import packages and store them somewhere?
<abralek>I interested in a location. Do you follow the same rule guix use gnu/packages/emacs-xyz ?
<abralek>Or maybe there is a general channel for elpa somewhere?
<rekado>abralek: I don’t maintain my own channel for these things. Everything I use I try to add to Guix proper.
<jonsger>abralek: if something is FSDG compatible, it's the best to submit to upstream (us). It's the least maintenance burden for us all
<abralek>I see. So basically guix import and then send it to the upstream
<YannD>Hi everybody, dumb question : I saw https://issues.guix.info/38636 : gfortran-XX , like gcc-XX are now hidden packages, and it's preferable to use gfortran-toolchain. But, for the moment only gfortran-toolchain-7.5 is available.Should we have (define-public gfortran-toolchain-9 ... in commencement.scm ?
<sammich>hm, i ran with 'dbus-run-session sway' and now that the variable is defined, but i get 'No Gcr System Prompter Available' :(
<rekado>YannD: yes, probably
***amfl_ is now known as amfl
<YannD>OK, what is the best way to send patches ? Should I make a bug issue ? (I'll see on doc if there is a contribution guide for dummies :) )
<sammich>ah, i got it - i needed the `gcr` package
<rekado>YannD: there’s a Contributing section in the manual
<YannD>rekado: reading it
***dddddd__ is now known as dddddd
<jonsger>rekado: how can I obtain thunderbird-sources from icedove? like the "comm" folder
<divansantana>Am I doing something wrong with time-machine command? Trying guix time-machine 6064ae811d -- install flatpak to install the old version of flatpak.
<lle-bout>divansantana: I don't think you can mix and install packages from multiples versions of GNU Guix in the same environnement (default), you could spawn a dedicated environnment though
<divansantana>nevermind, being stupid. guix time-machine --commit=04dc98291cb3cbbf871652c2ccb32c6064ae811d -- install flatpak
<lle-bout>jonsger: guix build -S icedove ?
<lle-bout>divansantana: ah ok, if that works! 🙂 Never really used time-machine.
<jonsger>lle-bout: that reults only the icecat source. but i found a solution
<lle-bout>jonsger: does it not give you the sources *after* patching as written in the docs?
<lle-bout>there's also; guix build --sources=all icedove
<lle-bout>or s/all/transitive/
<divansantana>Yeah, so guix time-machine --commit=04dc98291cb3cbbf871652c2ccb32c6064ae811d -- install flatpak fails with http://paste.debian.net/1146541
<divansantana>lle-bout: perhaps I should try in another profile?
<rekado>jonsger: unfortunately, you can’t because it’s just a native input
<rekado>I’d very much like to have composite sources in Guix
<jonsger>rekado: not a problem
<jonsger>there are ways around :P
<rekado>then we can get rid of all those native inputs and the build phases that copy them.
<rekado>divansantana: please report this as a bug by sending mail to bug-guix@gnu.org
<lle-bout>divansantana: hmm, no I think here it's the bug in the actual commit revision
<lle-bout>it's a bug *
<lle-bout>rekado: what would composite sources be?
<lle-bout>divansantana: there's no hope in backporting fixes to old GNU Guix versions that contained older package versions though
<lle-bout>(officially)
<divansantana>rekado: K, thanks.
<lle-bout>you could probably try instead writing a package for a newer version of GNU Guix
<rekado>lle-bout: Guix has a patch and quirks mechanism to retroactively fix bugs
<lle-bout>rekado: ah, so you can patch on top of commit revision "ranges" if used with time-machine..?
<divansantana>guix time-machine --commit=5de2246f408743c1a253e0fa7dfe902ec2a05355 -- install flatpak works. Which is a more recent commit :)
<rekado>re “composite sources”: currently the “source” field only accepts a single origin. If you need more you need to add origins to native-inputs and copy them before building. I’d like to put those extra origins in the “source” field.
<divansantana>It works! And that command is awesome.
<lle-bout>rekado: ohh.. well.. the current "proper" way is to maintain precomposed tarballs out-of-Guix?
<rekado>no, we use native-inputs instead
<rekado>and I think that’s ugly
<divansantana>deleting a profile as simply as rm -rf ~/.guix-extra-profiles/some-profile ?
<rekado>don’t use -rf
<rekado>this is just a link
<lprndn>bricewge: On the lightdm's font issue. Adding font-dejavu works but not font-gnu-freefont... I don't know what to think about it...
<bricewge>I glimpsed over conversation about font issue since core-updates merging so lightdm is probably not alone with such issues
*jonsger wonders if someone of the Guix community crowdfounded a MNT Reform laptop
<rekado> https://twitter.com/dustyweb/status/1259933776068706305
<pkill9>i would like that laptop, but i don't want to spend 500+ euros
<pkill9>:(
<pkill9>oh, it's 1300 euros, eek
<rekado>yeah, I also can’t afford it
<jonsger>dustyweb: nice :)
<jonsger>pkill9: its 1000 euro if you assemble it yourself (DIY kit)
<jonsger>the MNT Reform has the same SoC as the librem5 :P
<mothacehe> janneke: I tried building wip-hurd-vm but Guile 3 fails with the following error: https://paste.debian.net/1146552/. Something I'm doing wrong?
*janneke looks
<janneke>mothacehe: that's without civodul's cross-diff?
<janneke>i just tried: cf9deff16a squash! system: Use 'hurd' package in label.
<janneke>./pre-inst-env guix build guile --target=i586-pc-gnu --no-grafts
<mothacehe>janneke: yes on that very commit, without modifications.
<janneke>does that work? (fwiw, i'm using guix environnment guile3.0-guix)
<mothacehe>Yes this command works
<mothacehe>I test with the same environment as you
<janneke>mothacehe: we have been backing-up a few commits; to fb120a69a8 services: hurd: Use activation-service, hurd-etc-service.
<janneke>...but on both, building the hurd vm does not work
<janneke>i haven't seen the guile build problem, though
<janneke>does the guile failure log show something interesting?
<janneke>(i haven't been building with --no-graphs here, only on a newer branch, rebased on master)
<mothacehe>It strange because both `./pre-inst-env guix build guile' and `./pre-inst-env guix build guile --target=i586-pc-gnu guile' work.
<janneke>"computer, tell me what's different in human language"
<mothacehe>Haha! Starting the build again with -K, see you in 45 minutes :)
<lprndn>Hum does anyone succeed at using any font from font-gnu-freefont?
<lprndn>For me it doesn't appear in LibreOffice, has no effect in inkscape...
<bricewge>lprndn: Incriminating commits may be 405c0c947cfd59d7bfb81052545cd635970d2d0c or 5483a2d0a913fe533744699e9ef5757c6e3f6983
<lprndn>bricewge: I tried disabling the fontconfig-file-system when debugging lightdm and it didn't work so I would go for the second one. Should I send a bug report?
<bricewge>lprndn: It would be useful to report it.
<mothacehe>janneke: I have more info, see: https://paste.debian.net/1146566/.
<mothacehe>gzip uses lstat which returns -EOVERFLOW
<mothacehe>According to man lstat(2): " This error can occur when, for example, an application compiled on a 32-bit platform without -D_FILE_OFFSET_BITS=64 calls stat() on a file whose size exceeds (1<<31)-1 bytes."
<mothacehe>And 'file /gnu/store/jn0cm31xx0kpxbc924fjskbxnxjhyznq-gzip-mesboot-1.2.4/bin/gzip' returns '/gnu/store/jn0cm31xx0kpxbc924fjskbxnxjhyznq-gzip-mesboot-1.2.4/bin/gzip: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), statically linked, not stripped'.
<pkill9>how often is the linux kernel updated with security patches? e.g. is their repository regularly updated? Or do they only update in releases?
<janneke>oh my, an old gzip in the x86 bootstrap
<janneke>that's...unfortunate
<janneke>so, this is guile-final from commencement?
<janneke>i still don't understand how i don't see this
<ecbrown>is there a way to "trick" guix offload into utilizing a machine that has a load > 2? although it has a load, it's still a ton faster than building locally
<ecbrown>i had tried jacking up speed, but no luck
<mothacehe>janneke: It's not guile-final because ./pre-inst-env guix build -e "(module-ref (resolve-interface '(gnu packages commencement)) 'guile-final)" works.
<andreas-e>ecbrown: As I understand it, the load is only taken into account when comparing different available machines. If offloading is available, it will not build locally. Maybe you could tweak the field parallel-builds in /etc/guix/machines.scm?
<lprndn>bricewge: done
<bricewge>lprndn: Thank you!
<ecbrown>anreas-e: thanks, introducing parallel-builds with a largish number gives a normalized load average, and the vm's are utilizing the offload machine
<Guixguy>pkill9: Thank you for that guix time-machine for flatpak :D
<Guixguy>sneek: botsnack
<sneek>:)
<katco>i'm hacking on a `copy-build-system` package, and the `validate-runpath` phase is failing with an empty `RUNPATH`. does `copy-build-system` not do `ld-wrapper`?
***dongcarl7 is now known as dongcarl
***stefanc_diff_ is now known as stefanc_diff
<civodul>hey katco!
<civodul>katco: i think copy-build-system really just copies files around
<civodul>it doesn't compile anything AIUI
<katco>heya civodul :) that's my understanding as well, but it can still take inputs, so i was hoping it could sort the paths somehow
<mbakke>hm, cuirass seems to have stopped updating, presumably because the 'core-updates' branch went missing
*mbakke removes the specification
<mbakke>probably also need to restart cuirass?
<mbakke>restart was not necessary
<civodul>mbakke: yup, restart unnecessary
<civodul>katco: oh, i guess if you just copy .so or executable files around, and they passed validate-runpath before, they should still pass it after?
<lampilelo>why would guix download a lot of packages after invoking "guix install googletest"? it doesn't have dependencies afaik
<lampilelo>if i didn't do "guix upgrade" after "guix pull" is it possible that install command upgraded the packages before installing googletest?
<Guixguy>Hey all, what is scheme's equivalent to (with-open-file..)?
<janneke>lampilelo: try: guix refresh --list-transitive googletest
<janneke>does that match better with what you saw?
<lfam>Hey efraim
<lampilelo>Guixguy: with-input-from-file perhaps
<janneke>Guixguy: Guile has with-output-to-file, with-input-from-file
<reepca>Guixguy: call-with-{output,input}-file
<Guixguy>janneke: reepca: Thank you
<lfam>sneek: later tell efraim: I was wondering about libgit2 changing from git-fetch to url-fetch. How was libgit2 being fetch previously? Do we bootstrap libgit2 somehow?
<sneek>Got it.
<lampilelo>janneke: no, it pulled wayland for example
<lampilelo>but that command showed it depends on 68 packages, e.g. ncurses and I don't think it uses it
<mbakke>lampilelo: guix has "profile hooks" that are invoked when you use 'guix install', 'guix remove' etc, I guess those are what were pulled in?
<mbakke>lfam: git-fetch uses 'git-minimal', not guile-git
<lampilelo>mbakke: how can i check it? i don't have any custom configuration for guile
<lfam>sneek: later tell efraim: mbakke clarified for me that git-fetch uses git-minimal and not guile-git, so libgit2 wouldn't be involved
<sneek>Okay.
<lfam>I guess we could reevaluate that setup if we wanted, now that libgit2 is releasing tarballs
<lampilelo>mbakke guix*
<lfam>I feel like it's okay, though
<mbakke>lampilelo: 'guix size $(guix build googletest)' will list all runtime dependencies of googletest
<lampilelo> mbakke: that's more like it: glibc, gcc, googletest, bash-static. so i guess these 176 MB of packages were unrelated
<lampilelo>so why guix install would pull them?
<Guixguy>lampilelo: Thank you
<Guixguy>lampilelo: janneke: reepca: (open-file "test.txt" "w") creates a file for output
<lampilelo>Guixguy: but you have to close it explicitly afterwards, with-output-to-file will do it for you
<Guixguy>lampilelo: oh right, thank you
<lampilelo>Guixguy: note that with-output-to-file takes a thunk as a second argument, that means you have to wrap your code inside a lambda or a function that takes no arguments
<Guixguy>lampilelo: OK, thanks =]
<mbakke>lampilelo: because the guix profile hooks use GTK+ (if you have any GTK+ applications in your profile), which means Guix needs to download GTK+ (and its dependencies such as Wayland) to update your profile for things like icon-theme.cache
<Guixguy>lampilelo: (with-output-to-file "test.txt" (lambda () (display "world")))
<lampilelo>mbakke: mhm, so if i'd upgrade beforehand it wouln't do that?
<mbakke>lampilelo: yes, because then the because the prerequisites would already be in your store
<lampilelo>Guixguy: yes, and you can use (current-output-port) if you want to specify the port if something needs it, 'display' uses it be default so you don't have to tell it
<Guixguy>Great =]
<lampilelo>mbakke: ok, that clarified it for me a bit, thanks
<peanutbutterandc>Hello there
<lfam>Howdy
<peanutbutterandc>:)
<peanutbutterandc>I was watching a few videos on hurd and I was interested in just talking about it... but #hurd doesn't seem all that active right now
<peanutbutterandc>Does anybody here have a good (free) book they'd recommend to someone who wants to understand operating systems?
<lampilelo>Guixguy: alternatively you can use (call-with-input-package "file.txt" (lambda (port) (display "foo" port))) if you don't want to overwrite your current-output-port
<lampilelo>Guixguy: i'm not thinking clearly, what i meant was "call-with-output-file"
<lampilelo>input-package? wtf?
<peanutbutterandc>I have a question... regarding hurd... it appears that (from a few videos at least) the hurd developers have a "meh" attitude about the mach microkernel. It is more like "it is what it is", and that they aren't all that thrilled... how do the developers who are porting guixSD to hurd feel about mach here?
<peanutbutterandc>Not that I understand the entire thing or anything. Just fanboying.
<peanutbutterandc>lol I have a guix hurd running on qemu and I'd like to poke around with it some more but I have no idea what to do
<peanutbutterandc>the struggles of being a noob
<lfam>peanutbutterandc: Those people do chat here. I'd hang out for a while to get a response
<guix-vits>peanutbutterandc: If you've some time, can you please test if guix-with-hurd VM image works good _without_ kvm?
<peanutbutterandc>lfam, I see. It's not technical. Just wanted to talk about hurd just so...
<peanutbutterandc>guix-vits, It doesn't for me. I've tested it out. And failed. It just gets frozen...
<peanutbutterandc>guix-vits, On the plus side, I'm running guix-hurd on gnome-boxes :D
<peanutbutterandc>funny, we've got python3 python2 perl on guix-w/-hurd. But not guile. :D
<peanutbutterandc>guix-vits, Is there anything else I can poke around with on guix-hurd that might prove to be useful to the devs too?
<olivuser>hello guix
<olivuser>does it make sense to use emacs and geiser in packaging?
<olivuser>I am currently trying to modify an existing package and was wondering if I could, for example, test just a specific phase in a build process?
<guix-vits>peanutbutterandc: idk, i'm not developer. Thanks, those weren't frozen for me, but "echoing" of my inputs were broken. Pity this machine of mine not support kvm.
<cbaines>olivuser, because of the different environment in which the packages are built, you're probably going to need to build the package to test even a specific phase.
<peanutbutterandc>guix-vits, Hmm... for me, I am on a foreign distro (elementary os) and used guix-installed qemu to start it. I had to manually add myself to kvm group and `chown root:kvm /dev/kvm` and it worked. I don't know what was wrong...
<pkill9>I would have thought that guix would have --fallback enabled by default
<olivuser>cbaines, okay, thanks for the info! but then, in which cases is geiser useful around guix?
<pkill9>since that's what it effectively does with all packages if there are no substitutes available, fall back to building them locally
<cbaines>pkill9, --fallback relates to when a substitute should have worked, but there was some problem. Like it failed to download due to the network, or a locale issue meant it couldn't use the substitute.
<cbaines>olivuser, I haven't used it myself, but I guess it can be useful to navigate around the package defintions
<pkill9>yeah, but i would have thought by default it would just build locally instead
<pkill9>because as far as the user is concerned, it just means a substitute isn't available
<pkill9>but maybe I'm just bikeshedding
<reepca>it's a good default in places with stable networks, because it can ensure the user is aware of any new connection issues, but with potato internet it can be irritating. Of course, the most potato of internets will produce no error whatsoever, and will simply wait forever.
<peanutbutterandc>reepca, It is a good default to not have --fallback with potato computers too. I for one am glad that guix doesn't --fallback by default unless I explicitly tell it to. :)
<efraim>lfam: I'm glad it got cleared up. I forget about git-fetch too sometimes :)
<sneek>Welcome back efraim, you have 2 messages!
<sneek>efraim, lfam says: I was wondering about libgit2 changing from git-fetch to url-fetch. How was libgit2 being fetch previously? Do we bootstrap libgit2 somehow?
<sneek>efraim, lfam says: mbakke clarified for me that git-fetch uses git-minimal and not guile-git, so libgit2 wouldn't be involved
<efraim>sneek: botsnack
<sneek>:)
<lfam>sneek is a good bot
<nojr>hello, how can I find the `examples/` directory in the Guix's Guile installation? I tried looking in `.guix-profile/share`
<cbaines>nojr, assuming you want the examples directory containing the example configurations for different systems, perhaps try ~/.config/guix/current/share/guile/site/3.0/gnu/system/examples
<cbaines>unless you have guix as a package installed in your profile, it won't be there. Usually a different profile, managed through guix pull is used for guix itself.
<whk>Bug #41241 "xfce4 default font display broken after guix pull; sudo guix system reconfigure /etc/config.scm" submitted ( https://debbugs.gnu.org/cgi/bugreport.cgi?bug=41241 ). I will be on for a while if somebody is available to debug.
<nojr>cbaines: There are additional examples in the ‘examples/’ directory in the
<nojr>cbaines: There are additional examples in the ‘examples/’ directory in the
<nojr>cbaines: "There are additional examples in the ‘examples/’ directory in the Guile source distribution."
<nojr>(info "(guile) Hello Guile!")
<reepca>nojr: unpack the result of 'guix build --source guile' and look in the examples/ directory.
<reepca>that's what it means by "the source distribution"
<cbaines>nojr, I've found that text in the Guile manual. I guess this relates to Guile then. I don't believe those files are included in the Guix package.
<cbaines>You can also browse them on the web here http://git.savannah.gnu.org/cgit/guile.git/tree/examples or clone the Git repository
<nojr>cbaines: thank you very much
<sirmacik>anybody tried to work with some guile inotify library?
<sneek>sirmacik, you have 1 message!
<sneek>sirmacik, bricewge says: How did you installed network-manager-vpnc? I think ti should go in network-manager-configuration-vpn-plugins field https://guix.gnu.org/manual/devel/en/html_node/Networking-Services.html#Networking-Services
<sirmacik>bricewge: I got it working eventually prolem was with something gnome related (polkit probably), fixed it by temporarly installing gnome and adding connection from it
<whk>btw is there a preference on png's in debbugs emails? I thought about attaching screen snapshots of problem but didn't want to clutter the system.
<sirmacik>how should I approach creating package for this module? https://github.com/ChaosEternal/scheme-inotify
<rekado>whk: it’s no problem to attach PNGs
<whk>Rgr. I will add an update to make it easier to see.
<mbakke>whk: thank you for submitting the report, similar problems have been discussed here on IRC but it's good to have an issue to document things in
<jonsger>meh. I wanna fix existing bugs, not discovering new ones :(
<lampilelo>is there a simple and/or good way to integrate PATH, C_INCLUDE_PATH, LD_LIBRARY_PATH, etc. from guix with the host system?
<civodul>sirmacik: it's a library for Chez Scheme (not Guile), and it's actually not free software (there's no license)
<whk>mbakke: np. Still getting a feel for level of detail to provide. I don't see a difference between the two that would cause the problem but don't know enough yet to dig down further on this one.
<whk>jonsger: :-) agreed
<sirmacik>civodul: and this one? http://files.sph.mn/u/software/sourcecode/guile-inotify/
<civodul>rekado: we're at 3.4T on berlin, and we were at 3.1T earlier today--not bad!
<mbakke>whk: I just replied to it with some information about the underlying cause, no good solution though :/
<sirmacik>civodul: I meant that url https://github.com/ChaosEternal/guile-inotify2
<sirmacik>guy wrote two inotify libraries
<civodul>2nd one is also non-free
<civodul>1st is written in C
<rekado>civodul: yes, very good!
<rekado>civodul: thanks for your help!
<civodul>yw!
<civodul>sirmacik: overall, you'd be better off writing it, or getting the second one freed
<civodul>it's just a handful of functions
<sirmacik>ok, thanks (;
<rekado>if you have opened a bug in the past, please consider visiting issues.guix.gnu.org to see if it’s still open
<rekado>you can search your issues with something like this:
<rekado> https://issues.guix.gnu.org/search?query=submitter%3Arekado+is%3Aopen
<whk>mkbakke: Thanks I can use the workaround of adding fonts-dejavu for now. So far I have just been working my way through the manual learning how guix does things compared to gentoo, debian, and centos. I am currently compiling my list of noob questions to ask if I don't figure it out.
<civodul>rekado: is there a URL to view specifically "guix-patches", or "guix", or "guile"?
<rekado>we are not including Guile at all.
<rekado>for guix-patches vs guix I could add a search prefix package:
<rekado>in the past you could view *any* issue, as long as it was hosted on debbugs.gnu.org; since we no longer use Guile Debbugs that’s no longer possible.
<civodul>ah so even a direct URL to a "guile" issue no longer works?
<civodul>re patches, actually one can go to: https://issues.guix.gnu.org/search?query=tag%3Apatch+is%3Aopen
<rekado>yes, because we don’t index Guile issues
<rekado>I can change that.
<rekado>indexing is just a tad slow, so I limited it to just “guix” and “guix-patches”.
<cbaines>I guess there could be another deployment of mumi for Guile :)
<rekado>yes, it’s configurable
<davexunit>civodul, sirmacik: regarding inotify, I have a pure guile wrapper available for snarfing: https://git.dthompson.us/starling.git/blob/HEAD:/starling/inotify.scm
<civodul>good, thanks davexunit!
*civodul pushed more issues. links to https://guix.gnu.org/contribute
<davexunit>it could be made into its own project but uh... what would the name be? that obvious one is taken.
<civodul>rekado: should the main page have a button to report a bug/send a patch?
<civodul>or is it too can-of-wormy?
<jonsger>can someone try `xdg-open mailto:guix-devel@gnu.org` on Guix system?
<Guixguy>lampilelo: Thanks again. I have added (call-with-output-file) as a comment under it so I don't forget
***jb` is now known as jbnote
<jbnote>hi there, i'm trying to package autofs. I'm almost there but validate-runpath is failing:
<jbnote>/gnu/store/p4db2aax6d1pcqw5v4j2fcm7xqxzaggy-autofs-5.1.6/lib/autofs/lookup_program.so: error: depends on 'libtirpc.so.3', which cannot be found in RUNPATH ("/gnu/store/fa6wj5bxkj5ll1d7292a70knmyl7a0cr-glibc-2.31/lib" "/gnu/store/01b4w3m6mp55y531kyi1g8shh722kwqm-gcc-7.5.0-lib/lib")
<jbnote>i've specified libtirpc in the inputs, though, I don't quite understand if the compilation process is to blame or if I somehow broke the RUNPATH which should be validated against.
<civodul>hi jbnote
<jbnote>any lights on this would be appreciated :)
<jbnote>hi civodul
<civodul>it could be because the build system doesn't explicitly pass -L... -ltirpc
<civodul>or maybe because it tries to pass its own -Wl,-rpath flag
<mroh>yay, autofs!
<civodul>the "ld wrapper" in Guix automatically adds -Wl,-rpath for libraries found in the library search path (-L, etc.)
<jbnote>OK. the build process for this is horribly broken in other instances, so i'll dig into it and patch.
<civodul>yeah, maybe check the build log for anything suspicious
<jbnote>Thanks a lot. The community is amazing, be it by mail or on IRC.
<lampilelo>Guixguy: np
<janneke>... or be it IRL -- i very much agree
<janneke>civodul: so...i got the xattr pipe translator setup working; we only need to patch two kernels and get this cross-build thingy going :-|
*janneke looks into adding a setxattr syscall instead of going through attr's setxattr
<whk>jonsger: what package is xdg-open in? I get from a new iso install of 1.1.0 with xfce command not found.
<jonsger>whk: xdg-utils
<civodul>janneke: wait, i have that in a stash i think :-)
<whk>jsonger: Tries to launch in "Mail Reader" but I don't have one installed at this time though
<janneke>civodul: ok, good -- no hurry then
<civodul> https://paste.debian.net/1146621/
<civodul>then (setxattr file "gnu.translator" command)
<janneke>yes, ty
*janneke combines our patches -- it seems both have complementary bits right
<civodul>cool :-)
<janneke>mine has some things wrong, at least
<jonsger>whk: hm okay
<janneke>user.translator="bah bla pfinet" => https://paste.debian.net/1146625/
<janneke>civodul: ^^ combining it now works
<janneke>of course, you need to patch your linux to set gnu.translator :-)
<civodul>janneke: yay!
<civodul>why patch linux?
<janneke>civodul: your linux prolly does not allow setting "gnu.XXX", only "user".XXXX https://lists.gnu.org/archive/html/bug-hurd/2020-05/msg00016.html
*janneke 's linux, does :-) ^^^
<janneke>civodul: wdyt, should i add getxattr and add a test?
*janneke opts for "maybe later"
<janneke>i'll push to master once it works for booting the hurd
<civodul>janneke: oh!
<civodul>and you already have a Linux patch!
<civodul>woow
<civodul>i thought one could use any name
<civodul>anyway, thumbs up!
<civodul>but that means we can't use it in practice, except in a VM with the right kernel
<janneke>yea, and a (trivial) hurd patch -- /and/ "it works"
<civodul>or, with mke2fs/libext2fs is they let us do that
<civodul>neat
<janneke>well, we just need to upstream this patch, then
<janneke>;)
<janneke>no idea how that works, a review by "someone" who's done some linux work could help
<civodul>janneke: that's not me! :-)
<janneke>maybe youpi, dunno
<civodul>ah yes
<civodul>though i suspect youpi is busy :-)
<janneke>says who? :-)
<civodul>heh
<whk>Still going through the manual but is there a way to do the equivalent of pxe boot+kickstart to install (or to create an install iso with a default build)? Currently the best I can think of is boot by iso, go into manual install scp over a script & config.scm and then run.
<civodul>whk: yeah currently you have to copy over the config.scm locally
<civodul>i think g_bor[m] looked at automating that but i'm not sure what the status is
<codemac>question - if libreoffice suddenly goes off compiling on install instead of downloading a pre-built binary - how can I find out why?
<civodul>janneke: for setxattr, could you add a simple test in tests/syscalls.scm?
<civodul>mostly to catch "obvious" problems, especially when porting
<civodul>(could be like the tests for mount/umount)
<g_bor[m]>civodul: there is a little bit blocker. I will add a small patch this week, if it has not been dealt with since then.
<civodul>hey g_bor[m]!
<civodul>nice
<g_bor[m]>It is that the command execution is blocked for a key press.
<civodul>ah, i see
<g_bor[m]>This is on the system init.
<civodul>are you using the protocol used by the tests?
<mbakke>O_o optimizations-for-level, did we have this capability the whole time? :-)
<civodul>codemac: presumably it just means that your substitute server(s) haven't built it (yet)
<civodul>mbakke: hmm yes :-)
<g_bor[m]>Not yet I will look into that.
<janneke>civodul: sure, easiest would be to add that together with getxattr?
<civodul>janneke: ah yes, why not
<civodul>no rush anyway!
<g_bor[m]>I was just driving the backend directly.
<rekado>civodul: yes, I can add a way to submit bug reports via the web interface. It’s the same mechanism as that for commenting on existing issues.
<janneke>otherwise, how/what to test...
<civodul>janneke: for mount we just test a basic error case
<janneke>i made sure the hurd boots with this, so "it works for me"
<civodul>but at least it shows the ABI is right
<civodul>mbakke: the big difference will be with what wingo did on Guile master
<janneke>right, i'll have a look
<g_bor[m]>What is missing is custom partitioning. Right now it just does the all in one partition receipe.
<rekado>the new compiler performance numbers seem magical
<codemac>civodul: I would presume that, but can I check it? Whenever I upgrade packages on my laptop, I don't want to even attempt compiling some packages
<rekado>around 780M (down from 2G) memory, and seconds instead of minutes for module compilation
<codemac>civodul: like guix package -i --only-substitutable? or even only `guix pull` for what has been completely built?
<rekado>codemac: you can view build status on ci.guix.gnu.org, but it’s messy. You can check with “guix weather”
<mbakke>rekado: woooow, thanks for the numbers.. Can't wait to have this in Guix :-)
*g_bor[m] -> zzz
<rekado>I’ll be running “guix pull” just for fun then
<mbakke>heheh
<tsmish[m]>Is freehdl broken for everyone or just me?
<rekado>tsmish[m]: https://ci.guix.gnu.org/search?query=freehdl
<rekado>tsmish[m]: I also don’t see any bug report about it.
<rekado> https://issues.guix.gnu.org/search?query=freehdl+is%3Aopen
<tsmish[m]>build log: https://paste.debian.net/1146635/
<dissoc3>freehdl build failed for me
<tsmish[m]>I don't see anything in my patches that differ from master and prevent building it.
<tsmish[m]>Also, it seems it isn't available as substitute.
<rekado>tsmish[m]: could you please send a message to bug-guix@gnu.org with a copy of the build log?
<tsmish[m]>rekado: I will, thanks.
<tsmish[m]>rekado: How much of build log to attach?
<rekado>if it isn’t too big you could just attach all of it as a file