IRC channel logs

2020-09-25.log

back to list of logs

<dannym>PotentialUser-52: I'd be all for it--basically back when I was looking at it stuff was moving too much (mono people said to just use dotnet core--but for which parts?); that was Mar 2019 ago.
<dannym>I have package definitions for dotnet-cli, roslyn, corefx, coreclr from that time--but seriously, I don't use dotnet enough to care about it anymore
<luis-felipe>rekado: I remember Sun's virual workplace (MPK20). But that was 3D and I don't know if it still exist.
***jonsger1 is now known as jonsger
<xelxebar>Waz up, Guix?
<drakonis> https://www.tweag.io/blog/2020-09-10-nix-cas/ relevant
<drakonis>content addressed storage nix is pretty impressive
<xelxebar>drakonis: Yeah, this is how I had guessed nix/guix worked before actually digging in.
<xelxebar>Want to see the proposal!
<drakonis>well, nix does that now
<drakonis>guix still does not have CA yet
<drakonis>not sure if there's plans to have it yet?
<drakonis>it doesnt seem to be something available globally
<drakonis>guix publish seems to have it
<drakonis> https://hpc.guix.info/blog/2019/01/creating-a-reproducible-workflow-with-cwl/ hmm
<drakonis>hmm, didnt know guix already had it?
<drakonis>seems to be unclear whether it is the default
<drakonis>hmm, that blog post uses ipfs to provide CA storage
<atw>I guix pulled and got "guix pull: error: You found a bug: the program '/gnu/store/…-compute-guix-derivation' failed to compute the derivation for Guix...Please report it..." but the underlying problem appears to be "substitute: guix substitute: error: TLS error in procedure 'handshake': The TLS connection was non-properly terminated." I'm assuming that doesn't need to be reported?
<atw>yeah, it was transient
<zacts>hello #guix
<guix-vits>Hello. Any way to achieve same without pre-baked file? The file need to be executable. https://paste.debian.net/1164599
<guix-vits>I'd looked in guix/gexp, but understand only how to use #:recursive? on single file.
<str1ngs>guix-vits: maybe you want to use plain-file?
<str1ngs>guix-vits: what do you mean by pre-baked anyways?
<guix-vits>Hello str1ngs. AFAIK, plain-file is for text-files (no executables).
<guix-vits>Pre-baked means, that i've an executable bash-script, that remains executable with #:recursive? keyword of local-file().
<guix-vits>I wan't to "declare" a file inside of scheme. So everything remains in one "config.scm" file, rather than being split to many files.
<guix-vits>Only WA i have is to create a login.sh in /tmp (from .scm), then chmod() it, then local-file() it. So it's "self-containing" config.scm. But i hope there is some make-executable-file(name contents) in Guix.
*guix-vits stops.
<g_bor[m]> [guix-vits](https://matrix.to/#/@freenode_guix-vits:matrix.org): you could try progam-file
<g_bor[m]>It won't be very nice though.
<g_bor[m]>Another option would be to have a look at plain-file, see what it does, and add a counterpart with chmod.
<guix-vits>g_bor[m]: thank U.
<tch70>hello guys
<guix-vits>o/
<tch70>I had some issues with compiling guix 1.1.0 on linux system with guile 3, where do I post a ticket?
<guix-vits>"Report bugs to: bug-guix@gnu.org." What issues?
<tch70>No bugzilla or an equivalent? It couldn't configure, and then it complained about define-json-mapping 'not precisely defined'..
<tch70>..I made it compile somehow, but have my doubts about the 'solution'
<guix-vits>tch70: Did U tried to install pre-compiled Guix, then `guix environment --pure guix`, then `./bootstrap && ./configure --localstatedir=/var && make -j`nproc` check` ?
<guix-vits>tch70: The closest to bugzilla is issues.guix.gnu.org, but it point to the same mainling lists.
<tch70>No, compiled from source.
<guix-vits>Can U try to compile the same sources using `guix environment --pure` tool? It just creates the 'environment'.
*guix-vits knows nothing more. Developers?
<tch70>But it would use the already compiled guix, right?
<tch70>..or should I download and use guix from the binary?
<guix-vits>tch70: It would use already compiled guix, to get (or if U switch OFF the substitutes, build) the tools needed to compile guix.
<guix-vits>tch70: U should nothing, just have a fun. Guix is great tool, toy, and so on.
<tch70>I don't quite understand the meaning of the procedure, but no harm in trying (later), I suppose. Thanks!
<testerman>Hola!
<testerman>IceDove in Guix is not working at all.
<testerman>XMP parser error
<testerman>Here is the paste log, [link to pornography spam removed]
*testerman is afk for now
<nly>sneek: botsnack
<sneek>:)
***apteryx_ is now known as apteryx
<janneke>sneek: later tell civodul: not sure what happened yesterday...but "it works" => http://ix.io/2yEJ
<sneek>Welcome back janneke, you have 2 messages!
<sneek>janneke, str1ngs says: Hello, I pushed a fix to emacsy where the console no longer spams a backtrace when using C-g . see https://git.savannah.nongnu.org/cgit/emacsy.git/commit/?h=wip&id=7469899cda1a000020515c6c86cc87fe0b253515 .
<sneek>janneke, str1ngs says: Hello, I pushed a fix to emacsy where the console no longer spams a backtrace when using C-g . see https://git.savannah.nongnu.org/cgit/emacsy.git/commit/?h=wip&id=7469899cda1a000020515c6c86cc87fe0b253515 .
<sneek>Will do.
<janneke>sneek: bag of botsnacks!
<janneke>str1ngs: very nice! emacsy is becoming really usable
<rekado_>again porn spam… removed it from the logs
<rekado_>and attempts to post porn spam to the mailing lists again
<wleslie>embeddable buffers?!
<leoprikler>I have a question regarding gettext: Guix claims it's GPL3+ (some sources under GPL2), but the manual claims "the [LGPL] covers the use of the gettext library" (https://www.gnu.org/software/gettext/manual/html_node/Using-gettext-in-own-code.html#Using-gettext-in-own-code). Who is right here?
<Brendan[m]2>I'm back to not being able to edit network manager connections as my user :/
<guix-vits>Brendan[m]2: IDK, but try add the user to netdev group?
<Brendan[m]2>i am in netdev
<Brendan[m]2>doesnt seem to help
<guix-vits>Brendan[m]2: As WA: Can U create new ones, and then edit them?
<Brendan[m]2>WA?
<guix-vits>Work Around
<Brendan[m]2>a work around isn't a solution
<Brendan[m]2>i can always edit them with sudo
<roptat>hi guix!
<guix-vits>o/
<apteryx>Brendan[m]2: when this happens to me, is because the connections wer created with the "make available for all users" check box, which is true by default.
<guix-vits>Makes sense, btw.
<alextee[m]> https://rendaw.gitlab.io/blog/55daefcf49e2.html#how-to-guix-for-those-who-don-t
<alextee[m]>This is great
<leoprikler>just a heads up, I think guile-build-system does not follow symlinks
<civodul>thought of the moment: Someone™ should write a blog post on how to use 'guix environment', 'guix build --with-git-url', and 'guix publish' to hack on stuff and publish binary to fellows
<sneek>Welcome back civodul, you have 1 message!
<sneek>civodul, janneke says: not sure what happened yesterday...but "it works" => http://ix.io/2yEJ
<civodul>janneke: yay!
<civodul>so now we need a CI jobset
<janneke>civodul: yeah, i guess -- someone who can write that and possibly someone who can debug it until it works
<janneke>;)
<roptat>hey civodul, yesterday I managed to get information about my network links, using netlink from guile :)
<janneke>wow, that sounded grumpy
<civodul>roptat: woohoo, well done!
<civodul>janneke: i can understand the grumpiness :-)
<civodul>yeah i'll take a look
<civodul>janneke: actually we could enable on 'master' but simply make sure the 'supported-systems' fields of packages known to fail is up-to-date
<roptat>so this week-end I'll try to add the rest of the rtnetlink protocol and we should be able to set addresses and other stuff from guile!
<civodul>things like Mesa, Qt, etc.
<janneke>civodul: yeah, you were an amazing help, but still, took us a week to run something that "just works" on our laptops
<civodul>yeah :-/
<civodul>offloading does not "just work" when one enables the childhurd service though
<civodul>we could "make it so" maybe
<roptat>and from there, we can expand, as netlink is not limited to rtnetlink (we could manage the firewall for instance :))
<civodul>roptat: yes, that sounds very cool
<civodul>we can at last have a proper static-networking-service too!
<roptat>yes, that's the first goal
<roptat>then improve the networking in our containers
<roptat>the two of which can be done with rtnetlink alone. the firewall needs netlink_netfilter
<roptat>I see there's also a netlink_selinux, but I'm not sure it can actually set policies (looks like it is only used to receive selinux events)
<civodul>neat
<civodul>looking forward to guile-netlink 0.1.0 :-)
<roptat>yeah... it's very low-level for now :)
<zimoun>civodul: I am polishing the Channel subsection move with your comments. Just to be sure to your reply: you mean “do not move ‘guix pull’ and ’guix time-machine’” for now, right?
<apteryx>guix publish! is this new? I've never noticed about it.
<efraim>civodul: tell me about ld-wrapper and GUILE_SYSTEM_COMPLETED_PATH, could it cause a segfault while bootstrapping guile-final?
<apteryx>ah, nevermind, I know 'guix publish' :-) but not via the command line.
<civodul>zimoun: yup!
<civodul>efraim: guile-bootstrap@2.0 sets it, so if you try to run a different Guile in that context, it aborts
<zimoun>thank you.
<civodul>efraim: the situation where it occurs is if you do --with-input=gcc=gcc-toolchain (with the yet-to-be-committed patch set)
<civodul>namely, ld-wrapper in that case runs on guile@3, but GUILE_SYSTEM_COMPILED_PATH is set in the environment
<civodul>very much a corner case
<janneke>civodul: yes, we should certainly find out why offloading doesn't work initially; we can't scale up the number of childhuds until we do
<janneke>i have a vague hope that for a cuirass job, it "just works", dunno
<civodul>janneke: yes, it just has to be turned on, but i'd like to avoid spawning hundreds of builds that are bound to fail
<civodul>that's why i wrote we should take a look at 'supported-systems'
<civodul>and delete "i586-gnu" from key things like Qt and Mes
<civodul>+a
<civodul>("Mesa", not to be confused with "Mes" :-))
<janneke>civodul: ah, right -- that makes sense; can we just add "hello" as a first job?
<civodul>janneke: to do that we need to create a separate jobset, but yeah, we can do that
*civodul goes to Cuirass /admin/specifications
*civodul ends up doing "sudo sqlite3 /var/lib/cuirass/cuirass.db"
<civodul>janneke: https://ci.guix.gnu.org/jobset/hurd-master
<mothacehe>civodul: I can take care of that if you want, I have been living in the sqlite3 interpretor these days
<mothacehe>oh too late
<civodul>ah :-)
<civodul>i've been living there for a while too, nice to meet you :-)
<civodul>that said, i never get it right on the first attempt
<civodul>so be ready to "sudo sqlite3"!
<mothacehe>haha, I might have a patch improving contention on Cuirass database, but more on that in a few days :)
<civodul>heh cool, thumbs up for everything you've been doing on Cuirass!
<janneke>civodul: nice, no failures yet ;-)
<mothacehe>thanks, hope it will end up with noticeable improvements :)
<mothacehe>performance improvements I mean
<terpri>oh nice, icecat is jumping to firefox 78. i'd prefer 81 ;) but that should be a big upgrade
<civodul>janneke: zero failures but... "evaluation 16812 registered 0 new derivations"
<janneke>ooh
<civodul> https://ci.guix.gnu.org/eval/16812/log/raw
<civodul>not very informative
<civodul>i think i did something wrong
<janneke>hmm
<civodul>ah
<civodul>handling of the "hello" subset in (gnu ci) is questionable
<civodul>why (string=? system (%current-system)) ?
<civodul>too bad vc-annotate doesn't follow file renames
<civodul>janneke: if you have a checkout of "master", can you remove that 'if' and likewise for (list) below?
<civodul>otherwise i'll do it later
<apteryx>civodul: oh, is this a vc-annotate limitation rather than a git limitation?
<apteryx>it annoys me as well, but it predates my use of vc-annotate.
<civodul>apteryx: prolly there's an option to turn it on and it's in fact a civodul limitation
<civodul>i'd like to know
<terpri>enhanced tracking protection, block autoplay, webrtc improvements, spidermonkey baseline improvements for ~10% speedup of baseline code, picture-in-picture video, less annoying notification popups, a global page zoom setting, also lockwise which could be very interesting but which i suspect icecat will disable
<terpri>unclear whether icecat removes all the interesting firefox accounts/sync stuff like lockwise (password management), or just hides it...one can self-host the servers, so there's not a freedom problem with it, more of a UI problem
<civodul>terpri: maybe a privacy issue if the defaults are to call home
<terpri>possibly
<terpri>the addons situation also seems pretty unfortunate, you have to just know to go to addons.mozilla.org and check the license metadata of each addon (of course the good ones are all free), the icecat "alternatives" were pretty broken last time i tried them (which, to be fair, was probably a couple years ago)
<civodul>yeah well, there are guidelines that are seem reasonable to me given the freedom-first goal
<terpri>i wonder if mozilla would allow a license-filter parameter for addons, both for icecat and for firefox users who care that
<civodul>but it'd be nice if icecat had a real addon finder
<terpri>addon search*
<guix-vits>terpri: Try #nomad-browser, forget IceCat. Scheme thru the Web.
*guix-vits hides
<civodul>good point :-)
<terpri>ahh, a webkit frontend, not something with its own engine
<terpri>nomand and next both seem fairly promising
<guix-vits>terpri: recently str1ngs done something to allow both (at different buffer each) webkit and qtwebengine.
<terpri>the latter of which is essentially blink
<terpri>i track nomand and next, but i'm still going to stick with mozilla until the bitter end, simply to prevent an apple/google-controlled web engine duopoly
<civodul>the "bitter end" seems to be approaching fast
<terpri>nomad*
***rekado_ is now known as rekado
<terpri>that seems unclear, they dismissed a huge number of employees, but are also cutting...basically r&d projects and hyperfocusing on firefox afaict
<civodul> https://www.jwz.org/blog/2020/09/this-is-a-pretty-dire-assessment-of-mozilla/ and the cited article suggest they're cutting on Firefox development
<civodul>dunno
<terpri>and are at least trying to find revenue streams other than google-search-by-default contracts (mullvad partnership for VPN, pocket -- where the server is nonfree, but github kremlinology suggests that it's because the team is too small and overworked, maybe they'll think of more)
<terpri>they spun off rust (which must have been expensive to develop) and, iiuc, moved servo people into the gecko team
<terpri>they should definitely stop paying their executives so much, though
<jackhill>I hope they suceeed. Currently for me, WebKit offers the best balance for advancing my software freedom by being the easiest to build, easy to create custom browsers around, use of gstreamer, and requiring the least patching to makde FSDG-free.
<jackhill>on the other hand, it is probably easier to fingerprint…
<wleslie>this post is so jwz. love it.
<wleslie>"in my humble but correct opinion" yep
<civodul>yeah :-)
<terpri>VPNs -- trustworthy ones like mullvad -- make perfect sense from a privacy point of view, at least for americans. they're no silver bullet, but they keep your ISP from spying on you and can be combined with other privacy technologies like containers (there is a separate-container-per-tab extension that works fairly well, for instance)
<terpri>(commenting on a point in the article)
<civodul>i think the article has a point: browser fingerprinting doesn't rely on IP addresses
<terpri>yes, hence my comment about other privacy tech being necessary (which firefox promotes with container addons; containerization is also not a perfect solution, but it helps)
<terpri>(icecat is terrible for privacy in practice, incidentally, if you ask panopticlick)
<terpri>also, in terms of failed projects...certainly there were missteps
<terpri>but compared https://killedbymozilla.com/ to https://killedbygoogle.com/ , and consider how many of the mozilla-killed projects have a future outside mozilla
*civodul unhappy: Out of memory: Killed process 730 (.emacs-27.1-rea) total-vm:12095704kB, anon-rss:11313176kB
<rekado>helm again?
<civodul>dunno, something else i guess
<civodul>i'll have to profile again
<civodul>it happens after several days
<rekado>:(
<rekado>I haven’t seen any crash since upgrading, but I have had to restart Emacs as I was playing with my GPG settings.
<apteryx>civodul: no issues here, but no more helm
*rekado uses ivy
<apteryx>i made the switch as well. I was growing annoyed at not being able to easily override the find-file when opening /gnu/store items.
<civodul>yeah perhaps i should switch but i'm unsure what i'll miss
<civodul>like git-grep right from helm-find-file
<civodul>apteryx: oh that one too!
<apteryx>these are basic things easily replaced
<civodul>i've accumulated too many annoyances of that sort in my workflow in general
<apteryx>counsel-git-grep
<civodul>ok but these are different key bindings?
<apteryx>you can set it to the same keybinding if you like. The suggested one in info ivy is: (global-set-key (kbd "C-c g") #'counsel-git-grep)
<terpri>firefox send? free software, self-hostable (i might add it to hcoop, i found it very useful when it was available). firefox os? still lives on as kaios, with the kaios technologies working with mozilla still. thimble? projects could be migrated to anil dash's glitch. shumway obviously became irrelevant once adobe flash became irrelevant. sunbird was more or less rolled into thunderbird. minimo was obviously simply
<terpri>made obsolete by smartphones. killing camino was maybe a bad decision... firebug was simply better than venkman and now firefox has built-in devtools again...the rest is ancient history
<apteryx>(that's another gripe I had with helm; their re-invented help reader instead of just offering a normal info manual)
<terpri>firefox os/b2g was probably the most expensive mistake, if i had to guess
<civodul>apteryx: oh Ivy has an info manual, that's really cool
<terpri>otoh, if it had worked out better, it would have given them a google-independent revenue stream...
<apteryx>There are some things I miss, mostly out of habit with helm, but overall I feel I regained control of Emacs with ivy (it gives me regular editable minibuffer!).
<terpri>s/with the kaios/with kaios/
<mroh>civodul: I think, last time you posted a mem profile indicates that this time gnus is eating your mem?! How big is your .newsrc.eld? Do you have evals in your gnus group preferences or something like that?
<civodul>mroh: .newsrc.eld is 1.6M, but i think it was eieio eating memory, for the Gnus registry
<mroh>ah
<civodul>that one is 3.6M tho
<rekado>I needed this one to become friends with Ivy: (define-key ivy-minibuffer-map (kbd "<return>") 'ivy-alt-done)
<terpri>i suspect that the dipping usage numbers include mobile usage, where there are...obvious problems (if it was wrong for microsoft to bundle internet explorer with windows in 2000, why isn't it wrong for google to bundle chrome with android in 2020?)
<rekado>the above lets you hit RET to add a directory name to the accumulated file name on find-file
<terpri>(and of course, apple allows only webkit on iOS, so firefox for iOS is...not really firefox)
<terpri>jwz clearly hasn't tracked firefox development over the years as firefox is no longer "kind of crappy" imo, post-quantum, and re: "They have an entire building full of people. What do all of those people do???" most of those people were working to make gecko not-crappy
<terpri>but i think he's correct with his final assessment: "[...] in my humble but correct opinion, Mozilla should be doing two things and two things only: 1. Building THE reference implementation web browser, and 2. Being a jugular-snapping attack dog on standards committees."
<civodul>+1
<terpri>although they will need to do *some* revenue-generating activities to do #1, or else they can't survive to do #2
<terpri>(mullvad partnership is an excellent idea imo, and pocket would be...unobjectionable if the team took the time to free the server code. i also wonder how many users would simply donate to firefox development through liberapay or whatever, a la wikipedia)
<terpri>(VPNs are overhyped but mullvad is one of the few actually good ones, and while they aren't an anti-website fingerprinting solution they are an anti-ISP-spying solution, which matters at least in the US)
<terpri>to give an example of how screwed up things are in the US, using the default ISP servers, visiting a nonexistent domain will not fail, but pull up a page of advertisements (and some low-quality search results for the domain name so it can be marketed as a "helpful feature")
<terpri>widespread tls makes spying on content harder, but still leaks some metadata
<civodul>bandali: did you try "guix install emacs-next --with-branch=emacs-next=master"?
<civodul>(or other Emacs hackers)
<terpri>and hypothetically, some people might use programs like bittorrent to download (and simultaneously upload) gnu/linux ISOs, and may not want their ISP to know which distros they are publishing ;)
<terpri>anyway, those are my thoughts on mozilla, apologies if i was a bit verbose; i'm personally invested in gecko a bit more than other engines (having been the primary person implementing BigInt), and quite concerned about the effects of an engine duopoly and WHATWG being controlled by apple, google, and microsoft alone
<civodul>terpri: it's good to read the views of someone's who's in the first circles around Firefox!
<terpri>if funding dries up for my friend and client's gamedev tools project, i might try to go work for mozilla, depending on how dire the situation looks (that would leave my friend in a tough spot, as i'm the only other skilled programmer working for her, and wouldn't be able to help much with liberating the project -- but she has job options in the proprietary game-engine world -- saving mozilla is probably more
<terpri>important than liberating one interesting product and setting up a single hacker with a free-software career)
<terpri>but that's contingency plan $n$ of $m$
<str1ngs>it would be nice if Mozilla had a reusable widget because right now I don't think xulrunner counts. The other options are webkitgtk and qtwebeninge or cef. the last two are basically blink which does not help the monopoly issue.
<terpri>yes, making gecko more embeddable could be valuable. right now spidermonkey is the only component that can be readily used on its own, afaik (gnome-shell uses it -- although an old version clearly, because 2n+2n doesn't work ;))
<luis-felipe>I just installed nomad 0.2.0 and somehow it conflicts with Emacs. For example, now Emacs logo GNOME's dash is nomad's.
<luis-felipe>*Emacs logo in the GNOME dash
<andreas-e>Hello Guix!
<terpri>i believe brave wrote a blog post or something about their challenges with reusing gecko (which is initially what they wanted to do, i believe)
<str1ngs>luis-felipe: Hello, that is very strange. let me see why that might be happening.
<clemens3>aeh, saving mozilla?!!! they are already in the hands of the big tech companies who are happy mozilla rotbloating along..
<andreas-e>I got a bug report for one of my software projects on ppc64, with the suggestion to debug it on qemu-binfmt.
<andreas-e>But I suppose that since this is not a supported platform for Guix, our qemu-binfmt service will not work for this?
<mroh>civodul: emacs-next is not usable/broken on current master. see #43277 or something like `guix environment --pure emacs-next -- echo $EMACSLOADPATH`
<str1ngs>luis-felipe: does logging out of gnome-shell and logging back in help at all?
<luis-felipe>str1ngs: I didn't try that yet, I wanted to mention the issue here first.
<terpri> https://brave.com/the-road-to-brave-one-dot-zero/ is the closest thing to an explanation i can find
<str1ngs>luis-felipe: can you try that see if it helps?
<luis-felipe>str1ngs: Yes, but later. I'm in the middle of work.
<str1ngs>luis-felipe: you could try with a test user. and then just switch users would to avoid logging out.
<terpri>(i guess "graphene" was like electron for gecko or something...and browser.html might have failed, or might be a wip, but XUL is on its way out afaict)
<civodul>mroh: ah ok
<terpri>clemens3, they are in a strange position, to be sure. i'm sure google's happy to have them exist with a small market share as a shield against antitrust suits
<luis-felipe>str1ngs: A test user? I'm in the Guix System. Don't you have to reconfigure the whole system to add new users?
<str1ngs>I lost faith in w3c when they though this was a good idea. https://gpuweb.github.io/gpuweb/
<str1ngs>luis-felipe: it might be best to wait till you have time to logout and back in.
<clemens3>terpri: don't get me wrong, i chose firefox over chrome and safari or edge any day, but that project has been a sh*t sh*w since they went live 2 decades ago immediately rewriting the whole browers, giving us such bloat presents like rust, revriting and dropping apis and email clients and reader this and that..
<clemens3>not opening up the whole thing for easy scripting in batch mode
<clemens3>using some arcane non standard compression to save the history
<terpri>but MoFo is still a nonprofit with a specific mission, and they cannot legally deviate too much from that mission, afaik (IANAL but served on the board of a nonprofit consumer coop for several years)
<clemens3>very closy soft attitudes..
<clemens3>likely both on incompetence and with purpose
<terpri>yes, it's been a bit of a mess all along
<clemens3>and of course the money comes from big tech as said, so no suprise, firefox is just the controlled opposition for internet dominance
<str1ngs>luis-felipe: if you find any other issues with nomad let me know. also here is a more recent screenshot https://www.bufio.org/images/2020-09-25-090634_2048x2110_scrot.png . I'm currently working on header bar support.
<clemens3>or just compiling the bastard takes a phd work..
<str1ngs>clemens3: CEF is just as bad as firefox for compiling
<str1ngs>maybe even worst.
<clemens3>yeah, chrome for sure..
<dannym>andreas-e: It depends on what you mean. The qemu itself is built on your host platform, so if you have an executable and want to have qemu binfmt without guix, it would totally work
<clemens3>tops it even
<clemens3>but still you need rust to compile firefox, and you can't compile rust with gcc only
*civodul posts patches to record/replay package transformations: https://issues.guix.gnu.org/43614
<andreas-e>dannym: I see, but what I would really like is more or less a virtual machine, where I could compile my project and debug by adding "printf" here and there or using gdb.
<andreas-e>I am in the process of reactivating my account on the gcc build farm, so this could be the solution.
<zimoun>civodul: record transformation is nice. It probably breaks my script attempt for from profile/manifest approximate manifest.scm+channels.scm, but that’s fine. :-)
<str1ngs>andreas-e: gcc112.fsffrance.org is good for power8
<str1ngs>andreas-e: gcc135.fsffrance.org for power9
<civodul>zimoun: manifests are unaffected, you should be fine :-)
<andreas-e>str1ngs: Thanks! I will have a look once my ssh keys are accepted.
<zimoun>not profile/manifest, I guess. Well, some properties should be added, I imagine. :-)
<janneke>civodul: pushed native-system restriction as db2785cd86ee420039ca44bba898151a4f9bbf2b to master
*janneke hit an unexpected rush hour in the physical universe, earlier
<janneke>eh, lifting of native-system restriction, of course
<terpri>civodul, very nice! i haven't used that feature (i tend to just create custom packages w/'inherit' for that kind of thing) but it sounds pretty useful
<terpri>in fact, depending on how powerful package transformations are, i might not have to make so many custom packages :p
<civodul>janneke: thanks!
<civodul>terpri: yeah, it's also useful for quick tests and for the Scheme-averse folks
<janneke>civodul: i'm pretty sure i typed that when working on cuirass integration, i remember disturbingly little of that time
<str1ngs>hello janneke, just a FYI I pushed a small fix to emacsy https://git.savannah.nongnu.org/cgit/emacsy.git/commit/?h=wip&id=7469899cda1a000020515c6c86cc87fe0b253515. this should fix the console backtrace spam when using C-g
<janneke>str1ngs: yeah, great! sneek told me ;-)
<str1ngs>janneke: great, sorry for the noise :)
<janneke>that agenda-spam was terrible
<janneke>no, i'm happy to hear it ;)
<str1ngs>I suspect this was overlooked because debug-on-error? was probably set to #t most of the time.
<janneke>ah, right
<civodul>janneke: hmm, still zero: https://ci.guix.gnu.org/eval/16815
<civodul>i must be missing something
<luis-felipe>str1ngs: I logged out and back in, and the problem persists. When I launch Emacs, it is identified as nomad in the dash. And the Emacs menu next to the GNOME Activities menu says nomad instead of Emacs.
<luis-felipe>Also, in the whole list of applications, Emacs has its Emacs logo, and nomad it own logo, and nomad's logo is marked to indicate that there is an instance open, while the Emacs logo does not have the mark... And it should.
<str1ngs>luis-felipe: thank you for reporting this. I have an idea what might be causing this. will look into it some more.
<andreas-e>Is there an update on how to get a hurd vm? The procedure of the blog post does not work anymore.
<andreas-e>(You see that I am finally catching up with April fool's jokes.)
<str1ngs>janneke: I still need to look into the agenda spam. it happens when you kill the message buffer at certain times.
<janneke>str1ngs: i see
<str1ngs>janneke: also I've yet to figure out how to call a interactive command outside of the context of a keybind. closest I have come is ((colambda _ (execute-editing-command))) but even that is not right. this works but its a nasty hack (begin (emacsy-key-event #\x '(meta)) (emacsy-tick)) . right now it's not possible to call an interactive command from say a GTK signal.
<str1ngs>should be (execute-extended-command)
<mbakke>andreas-e: I think you can use childhurd-service-type, see https://guix.gnu.org/manual/devel/en/guix.html#index-childhurd
<andreas-e>Thanks! But that looks much more complex than the independent VM image advertised in the blog post.
<janneke>str1ngs: oh, those "commands" are synthesized, right?
<janneke>something too clever there, i'm afaird
<str1ngs>janneke: right, they get wrapped in a trampoline. which is not so bad in terms of safety. I just can't figure out how to reproduce the same context that emacsy-key-event and emacsy-tick have.
<zimoun>civodul: piouf! TeXInfo is not friendly! :-) Well, do you think that Substitutes could go one level higher, similarly to Channels?
<civodul>s/TeXInfo/Texinfo/ :-)
<civodul>i think menus are the main issue
<civodul>not sure we want to move "Substitutes" higher, do you think it'd make sense?
<raghavgururajan>Hello Gu...Gu...Gu...Guix!!!
<civodul>howdy raghavgururajan!
<zimoun>menus are; and I have not found how to fold/unfold chapter/section/…, as it is possible with Org.
<raghavgururajan>rekado: Looks like one of spam attempts went through, #43606 (https://issues.guix.gnu.org/43606). :(
<raghavgururajan>civodul ^
<zimoun>civodul: I do not know. Now, it makes sense to me, but I have read / navigated on the Package Management chapter a couple of times these days. I will re-give a look at the info and html (because they “look” different :-)).
<mroh>I don't know how to start... I would like to help working on guix more, so I'm thinking about getting commit access, because often I have the impression that posting (simple) patches generate even more load on the committers. I think, I need 3 people vouch for me. mbakke said he would (months ago. Hope it's still the case ;). Perhaps 2 more committers are reading this and like to vouch for me (mike@rohleder.de)?
<andreas-e>raghavgururajan: But maybe the bug report is valid nevertheless?
<mbakke>mroh: Happy to vouch for you (again). Especially with your nice new Guix apparel. :D
<raghavgururajan>andreas-e: FWIW, yeah. But the same issue as already been reported.
<andreas-e>The link does not work anymore, so it would be enough to close it as a duplicate then.
<raghavgururajan>?
*raghavgururajan is afk
<mroh>mbakke: hehe, thank you ;)
<apteryx>andreas-e: it's simple as long as you run a Guix System :-)
<apteryx>e.g, you add '(service hurd-vm-service-type)' to your list of services, guix system reconfigure, then can access it with: ssh root@localhost -p10022
<apteryx>This is the GNU Hurd. Welcome.
<janneke>andreas-e: i believe the bare-hurd.tmpl example holds a minimal recipe in a comment
<janneke>andreas-e: https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/system/examples/bare-hurd.tmpl
<janneke>a new blog post on the Hurd is somewhat overdue..
<andreas-e>apteryx, janneke: Thanks for the explanations! A blog post is always welcome :)
<janneke>andreas-e: yw, thanks for the interest :-)
<raghavgururajan>andreas-e: Did you mean the issue link or spam link?
<andreas-e>The spam link does not work any more, so you can simply close the bug report if it is a duplicate.
<raghavgururajan>andreas-e: It still works. Anyway, nckx closed it.
<raghavgururajan>> apteryx‎: This is the GNU Hurd. Welcome.
<raghavgururajan> 😍
<andreas-e>apteryx: kex_exchange_identification: read: Connection reset by peer
<andreas-e>Connection reset by 127.0.0.1 port 10022
<andreas-e>Is there anything else I need to do?
<apteryx>no, it works as it for me here. Perhaps run ssh with -v to debug what it's struggling with
<apteryx>as is*
<andreas-e>No need to set up an ssh key somewhere?
<andreas-e>It fails very early on. Right after printing "Local version string SSH-2.0-OpenSSH_8.3".
<apteryx>which guix commit are you on?
<andreas-e>67ad1e148cdca053c7a96a090209832a8a5ca178 for the system
<andreas-e>I tried a "hurd restart hurd-vm", and starting fails.
<apteryx>mine's a bit dated: f5163902d7c3cfed5a97033f38e92d7158b19fa8 (Sep 15)
<andreas-e>Now it did restart.
<andreas-e>But I also fail to ssh.
<andreas-e>I do not need to reboot, do I? (Very unguixy. But who knows.)
<apteryx>I'd try just to rule out something silly
<andreas-e>Now it works. Very weird.
<andreas-e># uname -a
<andreas-e>GNU childhurd 0.9 GNU-Mach 1.8/Hurd-0.9 i686-AT386 GNU
<andreas-e>Amazing!
<apteryx>:-)
<apteryx>you can thank janneke for that work
<raingloom>(psst. anybody wanna check on my Yggdrasil patches? it has docs now!)
<andreas-e>How do I get network access in the childhurd? I want to "guix build hello"!
<DrimysWinteri>hello
<sneek>DrimysWinteri, you have 1 message!
<sneek>DrimysWinteri, apteryx says: guix system search . | recsel -e 'location ~ "base.scm"', or look directly at the gnu/services/base.scm source file.
<DrimysWinteri>I'm trying to get 'mako' to work, but it keep saying "Failed to connect to user bus: No such file or directory"
<Philipp[m]1>Hi, just installed GuixSD for the first time on a X200. Very nice! :-) How can I configure a VPN client with network-manager-openvpn on a per user basis?
<davidl>Philipp[m]1: looking at help-guix email archives it looks like people are struggling.
<davidl>Philipp[m]1: Personally I ended up creating my own ovpn-service based on openvpn-service, but that is not per-user specific. Your best bet might be to use the network-manager vpn stuff via Gnome.
<Philipp[m]1><davidl "Philipp: looking at help-guix em"> thank you, will have a look there.
<apteryx>Philipp[m]1: I use nm-connection-editor for that. The openvpn-client-service seems system wide (not per user).
<apteryx>you still need to add the required openvpn plugin to the network-manager-service-type.
<civodul>all this got me playing with the childhurd on my laptop
<Philipp[m]1><apteryx "you still need to add the requir"> Thats what I am missing, I guess. Still find the configurations langugage quite unfamiliar. Could you give me a hint on how to do that?
<civodul>janneke: what do you think of these additions to the manual: https://paste.gnome.org/pqlna9va7 ?
<civodul>andreas-e: hallo! perhaps the text above can be useful to you? ↑
<civodul>this thing is indeed truly amazing
<andreas-e>civodul: The text is nice, but thanks to apteryx I got beyond that (except that the ssh connection started by being reset by peer, and ended up working). Now I wonder how I get network into the childhurd. "ping IP" loses all packages.
<civodul>ah awesome!
<civodul>but i think ICMP (ping) is not supported through qemu, is it?
<apteryx>not via its default 'user networking' facility
<civodul>right
<civodul>inside the childhurd i can do "telnet www.gnu.org http"
<apteryx>there are scripts shipped with qemu to setup a bridge easily, but these need to be setuid-root
<civodul>oh ok
<DrimysWinteri>what is the correct way to give mako dbus access?
<zimoun>civodul: argh! fight with s/TeXInfo/Texinfo, I understand why menu was out of sync. Well, hope everything is fine. I did my best for the commit message.
<andreas-e>civodul, apteryx: Okay, so no ping. But I only tried it since "guix build hello" failed when trying to download https://ci.guix.gnu.org/nar/lzip/rkfwng8xg400kjq7mzdl8lq78nqvzdcl-bash-static-5.0.16.
<civodul>zimoun: cool, i'll check it out... not tonight though :-)
<andreas-e>(In any case, the VM is too small for hello: "warning: at least 626.0 MB needed but only 196.7 MB available in /gnu/store")
<civodul>andreas-e: you can configure its disk size though!
<andreas-e>I know. I just found it funny that my VM is too small for hello :)
<zimoun>civodul: yeah the diff is terrible ;-)
<civodul>andreas-e: that's because there are no substitutes (yet)
<civodul>but in the meantime, i can offer "hello" as a service if you want
<civodul>i hear some say that it's really "hello as a software substitute"
<civodul>oh well
<andreas-e>"HASS"? German for hate.
<civodul>oh, that's telling
<civodul>:-)
<andreas-e>I am lost. If there are no substitutes, why is ci.guix.gnu.org advertising them? If ci is not advertising them, why is my "guix build hello" command trying to download them? Something is off.
<andreas-e>Hello, civocul!
<civodul>Freudian slip? :-)
<civodul>so i guess ci. only has substitutes for tarballs and the like
<andreas-e>guix/ui.scm:2115:12: In procedure run-guix-command:
<andreas-e>Wrong type to apply: #f
<andreas-e>substitution of /gnu/store/gla1r3nmxdz8rsx9g3bignq70src3p7s-bash50-015 failed
<andreas-e>substitution of /gnu/store/41bxkx3ynds2xsd6szxgvchzr76rlmyq-bash50-016 complete
<civodul>"Wrong type to apply: #f" is terrrible
<andreas-e>LOL
<andreas-e>Freudian slip; hm, no, just fatigue
<civodul>heheh
<andreas-e>All this spam is getting me confused.
<andreas-e>Anyway, fewer things are more exciting than compiling hello on the hurd!
<civodul>that's for sure!
<str1ngs>when you just want to say hello to the chicken, but you get the whole GNU barn instead :)
<civodul>:-)
<andreas-e>janneke: I am still confused about the persistent hurd image at the end of the documentation. The "image" field is clear. But why the "options" field set to "--hda"?
<andreas-e>janneke: So disk-size is in bytes, memory-size in MB? That is confusing.
<librecat[m]>firefox 68 ended support so icecat beta rebased on 78 !!
<librecat[m]>sorry got too exited there
<andreas-e>For the persistent hurd vm, this part does not work: (image (const "/out/of/store/writable/hurd.img"))
***teythoon1 is now known as teythoon
<andreas-e>"guix system reconfigure" blows up.
<andreas-e>(image "/out/of/store/writable/hurd.img") compiles, but might not do the correct thing.
<andreas-e>Adding "const" thunks the value, right?
<janneke>andreas-e: ah weird units for sizes, yeah sorry ;-)
<rekado>(const "hello") is a procedure that always returns "hello", no matter what arguments you give it
<janneke>andreas-e: possibly you found a documentation bug; we changed things around a bit with options
<andreas-e>So '(const "hello")' is a thunked '"hello"', no?
<andreas-e>janneke: Okay, thanks. So what do I write? (image "/some/path")?
<janneke>andreas-e: intially, i had options deault to '("--snapshot" "--hda")
<janneke>we thought that was a bad idea, and moved --hda hardcoded into the command
<janneke>see https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/services/virtualization.scm#n946
<janneke>i'll update the documentation
<andreas-e>Well, I know nothing about the hurd, so maybe I am making mistakes. I tried to set disk-size to 5GB, but "df -h ." still shows only 1.5GB.
<janneke>andreas-e: so if you use (options '()) then --snapshot is gone and you have a non-volatile childhurd
<andreas-e>Okay. I think the "--hda" is fine.
<andreas-e>What does not work is (image (const "...")).
<andreas-e>I get a horrible backtrace about monad, gexp, and so on.
<rekado>andreas-e: the difference to a thunk is that a thunk takes no arguments, whereas a “const” procedure takes any number of arguments (including none)
<andreas-e>rekado: Okay, I see.
<civodul>(const x) = (lambda whatever x)
<janneke>what are you trying to accomplish by setting the image thunk?
<janneke>"<civodul> janneke: what do you think of these additions to the manual: https://paste.gnome.org/pqlna9va7 ?"
<janneke>can you share that paste again / this does not exist :-/
<janneke>ah, you want to supply your own image, yea
<janneke>*h
<janneke>andreas-e: so that's probably (const ".../disk-image")
<civodul>janneke: does that one work: https://paste.gnome.org/phz1ldqjr/nbrgoe/raw ?
<civodul>or maybe https://paste.gnome.org/phz1ldqjr/nbrgoe/
<civodul> https://paste.gnome.org/phz1ldqjr seems to work for me
<civodul>this is confusing
<andreas-e>janneke: The documentation seems to imply that to have a persistent VM, you need to provide "options" and the name of a file, via "image", where the VM would be stored. Maybe I misunderstood?
<apteryx>civodul: note that gnome pastes have a default expiry of 30 m. It may be cached on your end but already expired?
<janneke>civodul: terrible, none of them work for me -- what could i be doing wrong?
<andreas-e>They also do not work for me.
<andreas-e>I mean the pastebins.
<janneke>andreas-e: ah, we should make that more clear
<janneke>no, you can have a persistent VM by just not using --snapshot
<janneke>in fact, i think that civodul and i kind of broke the (const ".../image") scenario: currently the hurd-vm-service mandates that the image starts a secret service :-(
<janneke>andreas-e: thanks for checking the paste-bins
***ae is now known as Guest92136
<janneke>i believe that mothacehe was working on/dreaming about a units library, to help clean up all different units and converssions use all over the place
<andreas-e>Sorry, my machine froze and I needed to reboot.
<andreas-e>Now I am having bigger troubles:
<andreas-e>Everything back to normal; "herd status" complained about a missing socket.
<janneke>ah yes, the documentation states that (image ..) is necessary --hmm, i don't get it
*janneke got some nice bug reports here to think about...
<janneke>andreas-e: ah, i'm slow today: yes, of course you cannot have a writable image in the store!
<janneke>'doh
<andreas-e># herd start hurd-vm
<andreas-e>herd: exception caught while executing 'start' on service 'hurd-vm':
<andreas-e>In procedure connect: Connection refused
<andreas-e>I do not think that this is related to hurd-vm; something is off with my shepherd. I had similar problems earlier today with docker.
<janneke>bah
<apteryx>for me the real deal would be being able to share the host stare with the VM in a writable manner (via the host guix daemon), in a writable manner. IIUC, right now I need to define a large, self-contained VM image that has its own store?
<andreas-e>janneke: I see; but then I first need to copy the store image to another place in the file system, and not just give the name of an empty file?
<andreas-e>apteryx: That would not help that much since the hurd VM and the Linux machine would not share anything, I suppose.
<janneke>yes, copy it, make it writable -- the documentation could give a few more hints there
<andreas-e>janneke: How to find it also, since its name is printed only when building it during "guix system reconfigure".
<andreas-e>I will reboot and come back; this shepherd is worring me.
<civodul>janneke: https://web.fdn.fr/~lcourtes/pastebin/childhurd_doc.html
*civodul got tired of pastebins, installed & configured emacs-scpaste
<janneke>ty!
<apteryx>civodul: nice!
<janneke>civodul: very nice addition, thanks!
<janneke>we should fix the (options --hda) bug that andreas-e found, and could be more clear about why in-store images can only be volatile, and how to create a valid writable image
<janneke>e.g, you /must/ add a secret-service...
<janneke>civodul: it would be great if you could add those while you're at it, otherwise i'll do that tomorrow
<andreas-e>janneke, civodul: That would be much appreciated, but without any hurry.
<apteryx>andreas-e: it'd at least share the space, without having to care much about the VM disk size.
<janneke>andreas-e: ah sure thanks for the reminder!; i rather meant: i'll do that not-today ;)
<civodul>i'm not sure i followed the --hda issue
<andreas-e>civodul: If I understood correctly, you need (options '()) instead of (options '("--hda"))
<janneke>right, we used to have --hda inside options by default, but moved that hardcoded into the command line
<janneke>but the doc has: https://git.savannah.gnu.org/cgit/guix.git/tree/doc/guix.texi#n25544
<civodul>ah ok
<janneke>so it should just advise: (options '())
<andreas-e>apteryx: Indeed, that would be an argument. But I think that for having a shared store, the childhurd daemon would also have to go through the outside daemon, so that the latter can update its database. "guix gc" would also be thorny, from inside or out. A separate image looks much safer.
<apteryx>using GUIX_DAEMON_SOCKET it should be possible to talk to the host guix-daemon via ssh
<janneke>hmm, would that work, share a file system between hurd and linux?
<apteryx>We already do it read only
<andreas-e>Time to leave. Good night!
<apteryx>I don't see why read/write wouldn't be possible. I think the issue was with the guix-daemon having to be shared, but given this is a service, we can set it up the way we like
<civodul>janneke: pushed the doc bits!
<civodul>we could automate 90% of the offloading, the missing bits being /etc/guix/{machines.scm,acl} on the host, which are not managed declaratively
<janneke>civodul: great, thanks
<civodul>i still don't understand why https://ci.guix.gnu.org/jobset/hurd-master gets stuck at 0
<civodul>i tried "make cuirass-jobs" with the same params and it works
<janneke>nothing is different really, still different outcomes...hmm
<janneke>tricksy computer
<civodul>very likely PEBKAC
<civodul>i'll check with a clear(er) mind tomorrow
<janneke>or picnic
<civodul>or picnic, who knows
<civodul>:-)
<janneke>yeah, thanks a lot again for today
<civodul>yeah, sorry for being this slow :-)
<civodul>night!
<janneke>night!
<shcv>has anyone installed guixsd on a raspberry pi?
<buenouanq>I have a Beagle Bone Black I've been meaning to put it on...
<shcv>getting it on the PINE64 sopine would also be interesting, but I don't actually have those yet