<zimoun``>jackhill: it could be good to reopen the question. Some time ago, some discussions happened but without a policy we can refer, if I remember correctly. For example, I am not convinced the Debian ‘Policy’ is the right one.
<jackhill>zimoun``: I remember that there was a Debian discussion, but don't remember the outcome. Maybe this is something we coudl work on with other FSDG folks.
<roptat>civodul, should we add a link to the conference announcement in the topic here?
***ChanServ sets mode: +o civodul
<vagrantc>ok, i *think* i've identified what's wrong with the u-boot-installer ... seems to work fine when you pass it a device file, e.g. /dev/mmcblk1 but not if you pass a symlink to a device file ... e.g. /dev/disk/by-id/RELATIVELYUNIQUEIDENTIFIER
<vagrantc>guile does the writing of the various bits to the device directly
<vagrantc>or at least, that's what it looks like to me
<zimoun``>jackhill: from my memories, I sent an email to the FSF and to Software Freedom Conservancy about this topic. And they never replied to you. But I do not find such emails. So I am doubtful if I really sent them. :-)
<civodul>vagrantc: did you see b97b423e3f61c80d5877dadc95b3f316cd61788f ?
<vagrantc>i'm going to try a simpler use-case to just be able to have pinebook-pro (rockchip) and pinebook (allwinner) co-existing ... looks like the offsets should permit that
<vagrantc>but longer term would be amazing to implement the whole idea
<lispmacs[work]>does anybody here happen to know how to disable autocomplete in libreoffice calc? I couldn't get a response from #libreoffice. I've found some info in online searches, but it doesn't seem to apply (operating system difference?)
<lispmacs[work]>unfortunately offline help files are still not available in Guix libreoffice
<vagrantc>ok, managed to get offsets that allow me to install both pinebook and pinebook-pro system on a single microsd
<xelxebar>bdju: Have you tried the info manual? I find it a lot quicker to navigate than the webpage render.
<bdju>I'm not very good with info so I usually reach for the web version
<bdju>oh, right. so I was saying, in section 8.8.1, Base Services, there's an example of configuring adb. it assumes you're using %desktop-services and has you modify that service. I'm using %base-services. could anyone take a look and tell me how I'd write that section instead?
<bdju>I tried some things the other day and I just can't figure it out
<bdju>specifically the services block at the bottom of the example is what I need an equivalent to
<wleslie>do I need to import anything in particular to use this modify-phases/substitute* stuff? I took it from embedded but I can't see anything special there https://bpa.st/55UA
<vits-test>bdju: M-x info in Emacs is more intuitive. For example, C-s searches first a current node, then the next ones. Also it do not go to the next page with C-n and C-v: it's done with Space. M-x info, C-h m.
<bluekeys_>Hi guix. Do we have something to convert stl files to gcode? I would have gone for cura3d, but it isn't packaged.
<mothacehe>cbaines: no just this message in an infinite loop
<cbaines>mothacehe, Ok, are you using the package in Guix?
<cbaines>I'm not quite sure how you want the coordinator to operate, but there are a couple of different allocation strategies implemented, and there's also some options for the behaviour when sumitting builds (like ensure-all-related-derivation-outputs-have-builds?)
<cbaines>mothacehe, yes, the intent here is to build things, even if they already exist in the store, even if there's a substitute available
<mothacehe>ok I have to take a closer look to those strategies
<cbaines>So, if you submit 10 builds for the same derivation, it will be built 10 times
<cbaines>(which might be useful for testing reproducibility and whether it builds on a range of different setups)
<mothacehe>cbaines: I see, I'm concerned about multiple slaves building the same packages on different machines.
<cbaines>mothacehe, if you only submit one build for an output or derivation, that shouldn't happen
<mothacehe>even if the build is a build time dependency of another build?
<cbaines>yeah, there's some hacky code involved, but the agents work hard to try and have the daemon only build the single derivation they're asking it to build
<cbaines>it could do with improving, but the agent will try and get the guix-daemon to substitute the build inputs prior to asking it to build the derivation in question (assuming there are build inputs missing)
<mothacehe>cbaines: btw, it would be nice if agents could auto-register themselves so that we can deploy easily an agent on all berlin build machines
<cbaines>mothacehe, indeed, maybe the coordinator could just accept new uuid's, or maybe it could do some kind of trust on first use of a uuid/password
<cbaines>also, I'm not sure I'm set on using UUID's, they're just handled as strings, so maybe one could choose to use hostnames instead.
<davidl_>I have python-pip as a propagated-input in one package, so after installing that package python-pip shows up in guix package -I | grep python, however, pip only works if I install it via guix install python-pip - how come?
<civodul>roptat: BTW, is there a way to let translators know what the priorities are?
<roptat>I have a bunch of uncommited files in gnu/packages/aux-files (.dirstamp, ...), should I add them to .gitignore, or is it only me?
<lispmacs[work]>i built and started the guix/hurd vm using the cmd line two-liner in the recent guix blog post. I'm I supposed to have internet access from the VM automatically with that configuration and cmd, or is there more I would have to do?
<lfam>Is it decided that the event will be based on BBB?
<roptat>more or less, but if you can offer a better solution, we can discuss it
<lfam>Nope! I don't know any free software solutions in this space, but BBB seems good
<lfam>I have some solid technical experience with live AV streaming and can offer advice to presenters, etc
<roptat>so, the goal is to have pre-recorded talks, so people can watch them in advance and come with more questions than in usual conferences (because, hopefully, they'll have more time to realize they didn't understand anything :p)
<mfg>is it possible to define inside a package how many cores hould be used for the build?
<roptat>the day itself will have live Q&A and discussion sessions
<roptat>mfg, there's #:parallel-build and #:parallel-tests (or -check, can't remember), but that's only choosing 1 or multiple
<lfam>That's a good idea roptat. It's tedious to just watch videos being played in a chat room
<mfg>roptat: thanks, this is what i was searching for :-)
<roptat>so, I guess if you have good advice for producing a video, that'd be great to have, same with advice on live Q&A
<lfam>I'll speak up on the mailing list discussion. In general, I'll wait for specific questions or misunderstandings. I don't really have a general "how to produce a video" spiel
<katco>i haven't ventured too far away from defining packages yet. can someone point me in the right direction for programmatically fetching source into the store from a SCS (e.g. git, hv, fossil, etc.). i.e. for git, i've gotten as far as this, but i'm not too sure what to do with it: `(git-fetch (git-reference (url "https://github.com/digitalocean/godo.git") (commit "HEAD")) #f "")`
<jlicht>nefix: you have a naming conflict as well: git-configuration-user refers to the field accessor in <git-configuration>, but seems to also refer to the record-constructor for <git-configuration-user>. Not sure if that actually matters in practice. The field error from earlier also seems to come from the fact that dont wrap the `name' and `email' in a <git-configuration> record.
<roptat>katco, you want that for defining a package, or something else?
<roptat>katco, that returns a procedure that wants a "state", so I think you can pass it (open-connection) from (guix store)
<roptat>katco, which fails with "In procedure bv-length: Wrong type argument in position 1 (expecting bytevector): """, so try passing #f instead
<nefix>jlicht: now I'm getting 'wrong number of arguments' (after changing 'config' for 'user' and using '(user (git-configuration-user ...))'
<roptat>katco, that gives you two values: a derivation and the store connection. You can use them later in build-derivations, from (guix derivations)
<roptat>there might be a better way to do it though
<jlicht>git-configuration-user refers two things in the snippet you posted: the record constructor you want, but also the accessor for the `user' field in <git-configuration>. You should rename one of them.
<katco>roptat: i'm not there just yet, but i think i will be when i can focus (helping my kiddo with virtual-school). so before i lose you: thank you so much! this looks like the thing i probably want.
<bluekeys_>I've tried adding google noto which i think aims to be a complete fontset. I've also dne a fc-cache -rv. I've installed emacs-all-the-fonts. It's doom modeline that is playing up. f27b is an example of a character that I'm missing
<tibbe>bluekeys_: I'm using doom modeline with font-adobe-source-han-sans, font-hack, gs-fonts, font-dejavu, font-gnu-freefont as installed font packages (quite a bunch, don't know which are required) and it works fine...
<bluekeys_>I wanted to use cura3d today, but we don't have it. I checked and it's lgpl or agpl. I don't remember which. I want to try and work on a pacakge for it. Could I do it by adding my own channel?
<bluekeys_>tibbe gthanks. I'll install them one by one and reboot. Lets see. Will report back here.
<tibbe>bluekeys_: Weird. I just have (doom-modeline-mode 1) after I setup doom-vibrant theme. With the font packages noted above and all-the-icons, all-the-icons-dired in my profile. What icons aren't rendered for you? Maybe I'm just not using them
<zimoun`>When ’foo’ has several outputs “guix build foo --no-grafts --check -K” keep only the last output. Is it expected/reported? Do I miss something?
<bluekeys_>It's all fixed now. I mentioned a few moments ago a particular missing icon, but I can't say now, I'm guessing I hadn't run m-x all-the-icons-install-fonts
<tibbe>zimoun`: As far as I am informed: If the package was already built, use --check. If the package wasn't built before, use --rounds=2. It's a bit weird
<katco>ok, well you've at the very least advanced me quite a bit. tyvm!
<roptat>actually, I think what happens with with-branch and co, is that the checkout is cloned outside the store, so the hash can be computed, then copied to the store
<roptat>it makes sense, because if you download directly to the store, two different versions would get the same hash, despite having two different content
<roptat>so, for git at least, you could use the same thing: a git-checkout. for others, you'd have to define your own type and gexp compiler
<zimoun`>apteryx: do you remember an issue with “guix build --check -K” keeping (/gnu/store/…foo-bar-check) only one output?
<katco>roptat: that makes some sense. but i was hoping to not have to write anything for all the SCSs and instead take advantage of what's already there. but it sounds like that might not be possible... hm