IRC channel logs


back to list of logs

<joshuaBPMan>And where can I learn more about what to do if one disk fails and how to fix it with a new disk.  I am using libreboot.
<joshuaBPMan>I suppose most of my coding stuff is already in git data lossage not likely there.
<lfam>For backups, I use the borg program, and back it up to a home server, as well as a paid account at
<lfam>I've never set up RAID (multiple disks) but if I were to, I'd probably use the built-in RAID tooling from btrfs
<joshuaBPMan>How much do ya'll suppose a RAID home server would cost?
<jlicht>+1 on both borg and
<pineapples>raghavgururajan: Pardon the influx of messages but I've finally managed to get the embedded fonts to "work". Passing "-DDESKTOP_APP_USE_PACKAGED_FONTS=OFF" to #:configure-flags did the thing: the client's UI elements are now using the embedded OpenSans font. Speaking of which, do you have any clue as to why stickers are not rendered? I reckon disabling the emojis might be the culprit here
<lfam>joshuaBPMan: There is a huge range of prices. You could get a simple multi-disk enclosure and do the RAID in software, and that would cost a couple hundred dollars
<lfam>You could get something like an Odroid HC4 which can hold two disks, for less than 1 hundred
<lfam>Or you could pay more for a NAS that handles RAID for you
<lfam>You could also get a more capable home server than the HC4, and pay more accordingly
<leoprikler>pineapples: you can try enabling the emojis and watch everything go boom
<leoprikler>on a less joking note if you find out why it fails to load emojis without that patch, it'd be very helpful
<joshuaBPMan>lfam: thanks.
<pineapples>leoprikler: I'll see what I can do
<pineapples>lfam: Do you own an Odroid HC4?
<atkka>hello, I'm doing my first guix system install, I've encountered some issues, I have been able to work around the networking and keymap issues, currently the install has hung on downloading libseccomp-2.5.0, its been on 77.3% for an hour
<atkka>is there a way to restart it without going through all the previous steps again?
<atkka>I can switch ttys and ping, so networking is up
<ryanprior>atkka: I don't know how much of the previous work you'll have to do if you restart. I know Guix can get hung up on transient network issues and need to restart though.
<atkka>ryanprior: thanks, I fear I'll have to restart
<ryanprior>I hope it doesn't set you back ages! Fingers crossed or you & I'd be interested to hear how it goes.
<atkka>thanks, I think I'll do a 100% manual install this time, the installer was very buggy for me
<atkka>I had to set up networking and keymap manually, the installer couldn't detect ethernet or wifi, I was able to connect manually in another tty
<atkka>also loadkeys was unable to load the colemak keymap
<atkka>not complaining, just FYI
<atkka>I can make bug reports after some more testing
<lfam>pineapples: No
***iyzsong-- is now known as iyzsong-w
<pineapples>leoprikler: Have found the culprit. The package "qtimageformats" was missing from telegram-desktop's inputs. Everything is now working as expected
<atkka>are there server issues at the moment? I'm still trying to install guix system from the live usb, I'm only downloading at 11KiB/s, this is the third attempt today and it keeps hanging
<apteryx>atkka: bug reports about your finding would be welcome yes!
<atkka>apteryx: I will report them if there aren't too many hoops to jump through.
<atkka>currently downloading from at 8KiB/s, any ideas?
<raghavgururajan>> liltechdude‎: raghavgururajan: How can I interactive see output from (assoc-ref %build-inputs "foo")?
<raghavgururajan>I am not sure about the context of what you are trying to see and/or do. Can you eloborate please?
<lfam>atkka: Can you share the URL of one of these slow downloads?
<lfam>Or, at least the name (including the hash)
<apteryx>atkka: to report issues, you can simply email, with a descriptive subject. You can consult current issues at or
<apteryx>atkka: from North America the download speed seems healthy: wget reports 8.54MB/s.
<atkka>thanks apteryx, I will see if there are any open reports and if not I'll report the keymap issues and networking issues
<atkka>hmm, that's strange
<liltechdude>Again question about variables and user. Can package define variables that user could modify? For example, about game engine, G-engine-package declare $G-GAME-HOME someting like this: G-GAME-HOME=/gnu/bla-bla-bla/<G-engine-package>/share/games/ and package G-game-on-this-engine modify in to G-GAME-HOME=/gnu/bla-bla-bla/<G-engine-package>/share/games/:/gnu/bla-bla/<G-game-on-this-engine>/ ?
<apteryx>atkka: and yes, restarting your machine (well, just doing 'guix system init' a 2nd time) means restarting from scratch IIRC.
<liltechdude>I probably understand how to do this on password-store example
<apteryx>liltechdude: if the resources come with the package, you can simply hardcode the variable values to their exact location (using substitute* or configure options if available).
<atkka>apteryx: I'll have to retry again, should I use 1.2.0 or the latest snapshot?
<apteryx>So far you've tried with 1.2.0?
<apteryx>Given you've always tried that route 3 times without a good outcome, I think it could be worth it to try the latest snapshot.
<apteryx>With some luck you'll get fresh substitutes that'll download quickly.
<atkka>ok, i'll download and try the latest snapshot, thanks for your help
<apteryx>Best of luck!
<apteryx>That's the most difficult part in my experience :-)
<lfam>It seems that the gnuradio is failing to build (test failure)
<apteryx>Once you have one working version of Guix it's easy to roll-back when something goes wrong.
<lfam>I mean, the gnuradio package
<lafrenierejm>Is anyone available to help with this line in a CMakeLists?
<lafrenierejm>> target_include_directories(udevmon PRIVATE "/usr/include/libevdev-1.0")
<lafrenierejm>Regardless of whether libevdev is included in inputs and/or propagated-inputs, the build fails on a #include of libevdev.
<lafrenierejm>And I'm assuming it's caused by the above.
<apteryx>hmm, so it's adding /usr/include/libedved-1.0 as as search path for the header files, which doesn't exist on Guix
<apteryx>you need to patch this to the location of the libevdev/include directory
<lafrenierejm>apteryx: Right. That agree my reading of the file, at least.
<apteryx>or add some pkg-config based find directive
<apteryx>this would be best, but it's a patch for upstream
<apteryx>the headers are at /gnu/store/kcch36d5hhivrzpd3y4n7pnzxzs5jzgc-libevdev-1.9.1/include/libevdev-1.0, if you want to try the patching route.
<apteryx>(under include/libevdev-1.0 of the libevdev input)
<liltechdude>It is possible to run program already with option? e.g `ls --color' when `ls'
<liltechdude>i.e declare aliases in installation script?
<apteryx>on my Guix System there's a Bash alias for it: alias ls='ls -p --color'
<lafrenierejm>apteryx: I'm attempting to patch the path in this recipe:
<lafrenierejm>apteryx: That, in turn, gives this CMakeLists.txt in build dir:
<lafrenierejm>apteryx: Oh, wait. I think that's working. The remaining errors are in the other build targets, not in udevmon.
<apteryx>there are also other include_directories being added in that CMakeLists.txt
<apteryx>I tihnk you could do (substitute* (("/usr/include/libevdev-1.0") (string-append (assoc-ref inputs libevdev) "/include/libevdev-1.0"))
<lafrenierejm>Ah. Yeah, that would be a better substitution.
<lafrenierejm>That allows all of the builds to succeed. :)
<apteryx>perhaps useful (I use this a lot when I fight with CMake) is the info manual that comes with it if you install cmake in your profile
<lafrenierejm>Oh! That would probably be very useful.
<apteryx>info cmake; then most commands/directives are indexed so you can do 'i target_include_directories RET' and it should bring you ta the right place
<apteryx>(else hit ',' a couple times to go to the next hit)
<atkka>apteryx: question, when using the installer one of the first steps is to choose your language and keymap, shouldn't the new keymap/layout be used for the rest of the installation?
<apteryx>It'd make sense, yes! Isn't it the case?
<lafrenierejm>apteryx: Many thanks for the help!
<atkka>user setup, passwords, luks passphrase are all in US qwerty
<atkka>after selecting my language and keymap which is not qwerty
<apteryx>lafrenierejm: :-) you can thank the awfully long rust bootstrap
<atkka>this happened in 1.2.0 and on the latest guix system which I am trying to install now
<apteryx>OK! Seems a valid bug to report to me, if it ain't already!
<atkka>ok, just making sure
<raghavgururajan>> I used to be able to garbage collect gigabytes in less than a minute, but now it takes several minutes
<raghavgururajan>lfam: I notied it too. Mine also is SSD.
<lafrenierejm>Is there a way to build a single executable using cmake-build-system? I'm wanting to do the equivalent of for, but the #:target argument in Guix seems to specify architecture.
<raghavgururajan>sneek, later tell pineapples: Glad that the configure-flag worked. Regarding stickers, I am not sure.
<sneek>Will do.
<apteryx>lafrenierejm: you mean a static binary?
<lafrenierejm>apteryx: No, I just want to output exactly one executable per recipe.
<lafrenierejm>In particular, right now I want to build only the udevmon executable.
<apteryx>What is the build system producing right now?
*apteryx goes off to bed
<lafrenierejm>apteryx: It produces all 4 executables that are CMakeLists `add_executable`.
<apteryx>shouldn't the package distribute them? We usually leave it to upstream to make this kind of decision.
<Aurora_v_kosmose>"~/.guix-profile/bin/ld: cannot find -lz" what package would contain what I'm missing?
<lafrenierejm>apteryx: Ah. Sure, I'll leave as-is and just note them all in the description, then.
<wleslie>I have a (substitute*) that stumbles on a file that seems to be in latin-1ese
<wleslie>how should I handle such a file
<raghavgururajan>lafrenierejm: There are two way I could think of: [1] Patch the CMakeLists.txt to remove the sepcific `add_executable` line(s). [2] After 'install phase, add a 'custom phase to remove the specific executable(s) in {output}/bin.
<efraim>wleslie: use the substitute* code from inxi in admin.scm as an example, it changes the encoding so that it should work
<wleslie>oh cool
<wleslie>setting the default port encoding didn't fix it (I now get "cannot convert narrow string to output locale" 84 #f #f)
***apteryx is now known as Guest71302
***apteryx_ is now known as apteryx
<wleslie>ended up dinking the source (there were only 3 problematic characters)
<mothacehe>hey zimoun!
<liltechdude>game engine succesfully patched
<liltechdude>now, how to export variable in global profile?
<leoprikler>What do you mean by "export variable in global profile"?
<leoprikler>If you want the variable to refer to e.g. ~/.guix-profile/share/game-engine/games, you'll have to add a search-path to the package description of game-engine.
<liltechdude>like that? (native-search-paths
<liltechdude> (list (search-path-specification
<liltechdude> (variable "INSTEAD_GAMES_PATH")
<liltechdude> (separator #f) ;single entry
<liltechdude> (files '("share/instead3/games")))))
<liltechdude>i copy & paste it from pass recipe, but after installing var not appear in shell
<liltechdude>after reboot nothing happened
<leoprikler>for the record, if you use instead3 it should probably be INSTEAD3_GAMES_PATH, but okay
<leoprikler>can you paste your package description?
<leoprikler>Nice Github username btw.
<liltechdude>leoprikler: what I wrote wrong in recipe?
<leoprikler>Currently checking, hang on.
<leoprikler>yep, path was wrong
<leoprikler>Apply this on top of master and you should get the correct behaviour.
<leoprikler>A few notes: Environment variables ending in _PATH usually denote colon separated variables, e.g. PATH=/bin:/usr/bin/:/run/current-system/profile/bin:...
<leoprikler>Since your patch only adds support for a single directory, it should probably be INSTEAD_GAMES_DIR, INSTEAD_GAMES_ROOT or something like that.
<leoprikler>You may also want to tidy your patch up a little and not commit any whitespace noise.
<leoprikler>The synopsis should probably be just "Simple text adventure interpreter", while the description should be at least one sentence describing what it is/does.
<liltechdude>the most problem now it is understand to apply patch >.<
<liltechdude>how to*
<leoprikler>Do you have a local checkout of the guix repo?
<leoprikler>In that case, "git apply" should do the trick without actually committing anything.
<liltechdude>but i want to do this in emacs (vc or magit) if possible
<leoprikler>C-x M-g A?
<leoprikler>oops, C-x M-g w
<liltechdude>my previous interaction with recipe was with something like "private channel"
<leoprikler>Private channels are neat for stuff that's hard to add to the main Guix repo, but if you do that simply because you haven't found something yet, you'll end up in a pretty rough spot.
<wingo>civodul: were source properties ever an important overhead in guix compilation? on the front-end side i mean
<wingo>just remembering that they were changed somewhat in 3.0.5 i think, perhaps improved, dunno
<civodul>wingo: hi! i think weak hash tables in general are quite expensive, and source properties were a source of overhead
<civodul>i don't have a simple bench for that, though
<civodul>i just remember noticing it via C-level profiling
<leoprikler>liltechdude: btw, the license for instead is license:expat, but it also has a bit of zlib-licensed stuff
<wingo>the commit was fwiw
<liltechdude>leoprikler: should I ask author about proper license?
<leoprikler>nah, it's fine if you do (list license:expat license:zlib) with comments in the package description
<civodul>wingo: nice, it does look cleaner this way, with fewer back-and-forth conversions
<leoprikler>license:zlib applies to some sdl2 extensions
<wingo>bigger change is that the "filename" is inline instead of in an alist
<wingo>there was an attempt to cache the last filename, to avoid excess consing, which probably mostly worked, but a hack
***rekado_ is now known as rekado
<civodul>wingo: oh i see, i had overlooked that
<civodul>longer term we should steal tricks from GCC's libcpp
<liltechdude>how can I change channels.scm for pull changes from local guix checkout?
<liltechdude>simply set (url "~/guix") does not work
<civodul>liltechdude: easiest way is to run "guix pull --url=$HOME/src/guix"
<jimmy>hello! :) i need to install debian pkgs. i got dpkg but when i run it i descover that i have read only file system. i got a problem or is just a guix system configuration?how can i fix it?
<civodul>liltechdude: note that tilde expansion only happens in the shell, not in your channels.scm file
<civodul>so you would need to write "/home/USER/guix"
<leoprikler>liltechdude: you can use a file:// url, but if you just want to test something in your local checkout, it's better to use ./pre-inst-env
<jimmy>sorry,could you explain me more? what kind of comand should i type?
<civodul>jimmy: you cannot directly install Debian packages on Guix System (you're on Guix System, right?)
<civodul>dpkg is indeed available as a Guix package, but i don't think it does much
<civodul>probably the easiest way is to find the package you need in Guix proper
<civodul>or in a channel
<civodul>or, as a last resort, to package it
<jimmy>i got the dpkg.... but when i use it it tells me the file system is read only
<jimmy>i would love to pack it..but i am afraid i m not good enought....could you explain me where i can find other pkgs?
<jimmy>what is guix proper?? what you mean find in a channel?
<leoprikler>(Guix) (proper), meaning a proper part of Guix.
<leoprikler>i.e. you find it somehow using `guix package -s`
<leoprikler>"in a channel" means someone else already packaged it in their local collection, but has not (yet) added it to Guix proper
<ekaitz[m]>jimmy: what are you trying to package?
<ekaitz[m]>I may be able to help you package it
<jimmy>i am not really trying to package it... i was asking some question about dpkg. because i would like to install some debian application
<jimmy>i dont know how to package something... i am not good enought
<sneek>Welcome back pineapples, you have 1 message!
<sneek>pineapples, raghavgururajan says: Glad that the configure-flag worked. Regarding stickers, I am not sure.
<jimmy>i found dpkg on proper guix.. but i cannot make it work
<leoprikler>Well, yeah, but we're not talking about dpkg here, we're talking about "packages, you'd install through dpkg".
<leoprikler>IIUC parts of dpkg can be used to make packages developed on Guix machines available to Debian users (don't quote me on that).
<leoprikler>But what's more likely you're trying to do is getting some .deb and trying to install it in Guix.
<jimmy>a ok.... sorry .. i was thinking to try anbox
<leoprikler>Which can't possibly work.
<pineapples>leoprikler: May you add "-DDESKTOP_APP_USE_PACKAGED_FONTS=OFF" to the "telegram-desktop" package recipe's #:configure-flags? We need it for the embedded fonts to work
<jimmy>and Gcstar...a pkg to manage movie and other collection..maybe you can just advice a different pkg taht make the same
<leoprikler>pineapples, what are "embedded fonts" in this context? wouldn't that disable the fonts, that come packaged with telegram?
<ozark>Learn Packaging: [porn spam removed]
<ozark>jimmy pineapples ^
<pineapples>Fonts bundled with telegram's dependency, lib_ui: No, quite the opposite — with -DDESKTOP_APP_USE_PACKAGED_FONTS set to ON Telegram would yield a message that the fonts bundled with it have been disabled. I was confused by this myself at first, but the OFF value is actually the one we want
<pineapples>uhh, is that a bot and a phishing URL?
<ozark>You were asking about packaging in guix right?
<leoprikler>but why a mega URL?
<leoprikler>Seems weird
<ozark>Oh its a talk
<ozark>FOSDEM I think. I had it in my device.
<leoprikler>I think it'd be better to point directly to FOSDEM, but fair enough.
<leoprikler>pineapples, do you know where in the CMake that decision is made? It seems weird
<pineapples>leoprikler: As in at which point the default value of this configure flag is set by CMake?
<leoprikler>as in where the inversion between ON and OFF takes place
<pineapples>I'm sorry but I don't know. I have little experience with CMake. I'll agree this seems rather strange, yes
<leoprikler>Fair enough.
<leoprikler>Can you shortly paste the output of "tree $(guix build telegram-desktop)"?
<leoprikler>Perhaps that gives some better insight.
<raghavgururajan>leoprikler, pineapples: {tdesktop-source}cmake/options.cmake
<mdevos>sneek: any sneekmail for me?
<mdevos>sneek: botsnack
*mdevos bye (busy)
<raghavgururajan>> leoprikler, pineapples: {tdesktop-source}cmake/options.cmake
<raghavgururajan>* {tdesktop-source}/cmake/options.cmake
<leoprikler>Interesting, it is apparently set to the value of ${dont_bundle_fonts}.
<raghavgururajan>pineapples, leoprikler: Here,
<pineapples>Thanks, I couldn't have possibly found it by myself. I don't see why it should be in a separate repository
<raghavgururajan>Yeah, their suff are very scattered.
<raghavgururajan>ozark: What is your problem?
<raghavgururajan>ozark: It was totally uncalled for. Please read,
<jimmy>can someone tell me were can i fount a channel to install anbox and gcstar pkgs?? or suggest other pkgs in guix proper that make the same job???
<leoprikler>raghavgururajan: Btw. why do we recursively check out Telegram now? Does that not defeat the point of the stuff we've done earlier?
<leoprikler>jimmy: I don't think anbox is so easily packaged in Guix, given that it *requires* snaps to work for some reason.
<raghavgururajan>leoprikler: For telegram-desktop? It has lot of submodules.
<leoprikler>Yeah sure, but we've already split them up neatly before? Why revert that? Seems kinda backwards.
<raghavgururajan>Oh you mean in v1?
<raghavgururajan>* v2
<raghavgururajan>leoprikler: No problem, I'll do a v17.
<jimmy>ok...and what about Gcstar??
<leoprikler>Hang on, still investigating
<jimmy>is possible in some other way to run android on guix? maybe with qemu??
<jimmy>ok! sorry.. thankyou very much!!!
<leoprikler>Gcstar seems discontinued and I can't download its source either.
<jimmy>can you suggest an alternative to those pkgs?
<leoprikler>According to, Tellico (not yet packaged) is an alternative.
<jimmy>is not just for kde??
<jimmy>what about griffith?
<leoprikler>Also discontinued.
<jimmy>ok...seems i have no options....
<jimmy>what about instead to use nix pkg manager?? this will give me the possibility to add some pakgs i need from the nix repos??
<leoprikler>Perhaps Kodi (packaged) is similar enough?
<raghavgururajan>pineapples and jimmy: Forgot to mention; don't visit the link posted by ozark. It's a spam.
<leoprikler>jimmy: nixpkgs, does seem to have a more up-to-date gcstar
<leoprikler>It points to
<vldn>hey :) is it now possible to load unfree firmware in the linux-libre kernel? Like a self-developed kernel module for some µC and shit :D latest news i found was from 2017
<vldn>(and whoever said it is a feature that none other modules could be loaded is wrong, the developer stated that it's a bug thats difficult to fix, don't want to block freedom)
<pineapples>raghavgururajan: Thanks for the heads-up! I have been wary of it until the very end, and it seems that this attitude has paid off
<jimmy>kodi looks perfect! thankyou very much!!!
<raghavgururajan>pineapples: Np!
<leoprikler>vldn: I'm not sure how to do this with linux-libre exactly (it seems you can perfectly load kernel modules in Guix already), but if you really do need some non-free kernel stuff, there is always the channel, that we do not speak of.
<vldn>yes i know, but i don't want to compile a whole kernel everytime
<leoprikler>You might want to take a look at the zfs-service-type currently under development by raid5atemyhomework.
<pineapples>vldn: As far as I know linux-libre has the ability to load unfree firmware, such as Intel microcode or AMD GPU blobs, removed
<pineapples>doesn't have*
<pineapples>oh wait nvm
<pineapples>The first iteration of the sentence above is correct
<raghavgururajan>civodul: Can you set user-mode -v on the nick ozark. (PS: It's the porn spammer again). I am not able to /msg you for some reason.
***ChanServ sets mode: +o civodul
<civodul>raghavgururajan: how do i do that actually?
<civodul>thumbs up mothacehe!
<civodul>that's really great news
<raghavgururajan>civodul: I think `/mode nick -v`
<mothacehe>thanks civodul :)
<jimmy>can someone tell me if is possible to install .deb file on the guix system?
<leoprikler>Yes, I can, the answer is no.
<stikonas[m]>Well, dpkg won't find dependencies
<jimmy>may be possible in the future?
<stikonas[m]>Well, if you install all deps, your guix will turn into Debian :D
<stikonas[m]>Most package managers are not designed to be used on foreign distro
<jimmy>ok...sorry for my silly question.. im not very good at computing...
<stikonas[m]>No, not silly, it is normal question
<apteryx>mothacehe: woohoo! Congrats for the grant :-). That's exciting news!
<rekado>mothacehe: this is most excellent! Congratulations!
<jimmy>ok...instead you can tell me if is possible to emulate or run in a sand box tha android system?
<leoprikler>I found, which seems to have some pointers towards running Android in qemu.
<mothacehe>thanks apteryx & rekado!
<zimoun>mothacehe: Cool!
<civodul>mothacehe: the reporting and monitoring bits sound great to me! it can make a real difference for users
<civodul>mothacehe: in other news, you should be able to connect to the OverDrive now
<leoprikler>be warned though, that's a very weird web page
<zimoun>mothacehe: what is “WCAG audit“?
<mothacehe>civodul: yes, NGI also offers support for security & accessibility audits which will be welcome too :)
<pineapples>Them again?
<civodul>bah, i don't know how to block that
<civodul>i tried "/mode -v" but i can't do that
<civodul>if someone know better, lemme know!
<mothacehe>zimoun: this is an audit regarding Cuirass website accessibility.
<civodul>mothacehe: nice! that'd be most welcome
<civodul>yeah i saw you mentioned the accessibility guidelines, that's great
<civodul>did you mean it comes "for free" with the grant?
***civodul sets mode: +zr
***civodul sets mode: +b ozark!*@*
<mothacehe>civodul: yes the audit is mandatory and comes for free! I was able to log to overdrive1, thanks :)
<zimoun>is it possible to use ’match’ and a pattern as ’values’?
<civodul>zimoun: 'match' matches a single value
<civodul>you were asking about multiple-value support, right?
<civodul>not possible, you'd have to first wrap those values in a list or something
<zimoun>I am trying to fix the recursive importers. And sometime it returns values and sometime only one value.
<civodul>ah that's a terrible kind of bug
<civodul>it's hard to track
<zimoun>I have tried “call-with-values” with list but I am not sure it does the job
<zimoun>but it does not seem the recent commit adding “version“, the bug existed before
<civodul>try this: (call-with-values (lambda () (values 1 2 3)) list)
<zimoun>it is what I tried… well, I will track again and see why it is failing (probably between my keyboard and my chair ;-))
<zimoun>Aside, the gnu importer as a “bug“ see ‘guix import gnu do-no-exist’ and I have a pacth. It is about the ftp-error incorrectly handled. But, it is not possible to add something in the test suite, right? The network is not available in the test suite, right?
<civodul>tests should not depend on external network services, in general
<zimoun>ok, thanks.
<apteryx>oh, package-closure seems expensive
<apteryx>I'm using it in a fold-packages and it takes ages
<civodul>yeah, don't do that :-)
<apteryx>what's currently the way to fold-packages filtering on inputs?
<apteryx>e.g., has X but not Y in dependency graph
<apteryx>Last time I came up with this:
<apteryx>but I though there must be a readily available way to do this; seems a common use case to use fold-packages
<apteryx>that pasted version is quite fast
<pineapples>mothacehe: Are both the accessibility and security audits mandatory?
<civodul>apteryx: maybe compute the closure of the referrers of X, that of the referrers of Y, and take the difference?
<mothacehe>pineapples: yes they are
<civodul>you can use the (guix graph) API for that
<apteryx>civodul: will peek in (guix graph), thanks.
<apteryx>For the record, here's a working example with that 'package-all-inputs' procedure:
<civodul>the general idea in (guix graph) is to build query procedures that are then O(1)
<civodul>so there's an initial cost, and then it's fast
<civodul>it should work well for the kind of query you describe
<apteryx>OK! Looking at it, it's probably a bit intimidating for non-seasoned Guix hackers; it requires leaving the package/inputs facilities and thinking in terms of graphes (edges, nodes, etc).
<civodul>agreed, it's not newcomer friendly
<civodul>actually, we could leave it to external tools
<civodul>like, you export what you want with "guix graph"
<civodul>and then use another tool to perform graph queries
<civodul>Dot experts know how to do that in its language (i don't)
<civodul>there's also the Cypher backend of "guix graph", but i've never used it
<civodul>and then there are graph viewers/tools, but i think we don't have them packaged
<civodul>so... room for improvement
<civodul>i know that is good for viewing and querying large graphs
<mdevos>Would it make sense to move the "mkdir-p/perms" that is defined in some places to (guix build utils), and would that lead to a world rebuild?
*mdevos tests the latter
<rekado>zimoun: all importers should return multiple values
<apteryx>civodul: Seems a bit weird to think that Guix cannot be used itself for this kind of query
<rekado>if they don’t then they should still return more values with the second being an empty list.
<apteryx>can't memoization magic be applied to package-closure? So that no matter how many times you call it, it computes things once and operates at reasonable speeds?
<rekado>zimoun: the idea is to return not just the package but also other inputs that still need to be imported
<rekado>zimoun: so if an importer cannot get that information it should just return (values the-package-definition '())
<rekado>once all importers do this we can make assumptions about the return values
<apteryx>Tulip 5 'looks' powerful
*apteryx wonders what makes package-transitive-* much faster that package-closure
<mdevos>yep, world rebuild. I see ‘gcc-mesboot’, ‘make-3.2’, etc. appearing in ‘The following derivations will be build’
<civodul>mdevos: yes, that'd be a patch for the "core-updates" branch
<zimoun>rekado: yes, it is just that there is inconsistency between what is returned, sometimes it is ’values’, sometimes not. So basically, the ’lookup-node’ in ’recursive-import’ from guix/utils.scm can fail with an ugly backtrace. Anyway, I am going to send a draft for a fix; if I miss something or misunderstand, explanations in the patch submission :-)
<mdevos>civodul: ok, I'll write a patch, though it may take some time before it's on master
<civodul>mdevos: yes, typically a couple of months before the branch is merged
<mdevos>I see ‘mkdir-p/perms’ doesn't create the directory and set the permissions atomically; there's a tiny window where a freshly-created directory has the permissions that would be expected from the umask. Is this something to be concerned about (and to be fixed in the patch)?
<apteryx>rust bootstrap seems broken on core-updates (rust-1.20 fails to build)
<apteryx>So none of GTK/GNOME stuff can be built there :-/
<civodul>mdevos: it's a good idea to be concerned about this, yes :-)
<civodul>in general, given that changes in (guix build utils) take time to trickle in, we should be extra cautious about interfaces and implementation details
<civodul>BTW, anyone has comments on --with-latest?
<civodul>otherwise i'll just push and we'll see :-)
<apteryx>civodul: Would the upstream-source-compiler work with any kind of upstream file? (VCS checkout, tarball, single file, etc?)
<str1ngs>hello daviid just going over your e-mail in regards to <gdk-event> LGTM. I just had to switch to the new !event-type method vs !type. I'll test some more with removing gdk as a dependency.
<sneek>Welcome back str1ngs, you have 2 messages!
<sneek>str1ngs, daviid says: please track/read 'today's logs - 2020-01 13th - for what i just wrote to tohoyn, and please try and let me know if nomad works fine with this new 'version' -
<sneek>str1ngs, daviid says: thse patches also fixes the get-tls-info bug you add, re-run your code and let me know, but here, the following snipset - - displays ;; #t #<<g-tls-certificate-gnutls> 7fd0bfa0c5c0> () - when button clicked ...
<apteryx>civodul: also, is it possible to add an index for '--with-latest'; I find it nice to be able to get to any option with ease in the doc.
<civodul>apteryx: it's a compiler for <upstream-source> records, which currently only deal with tarballs
<civodul>apteryx: re the index entry, we don't have any index entry for command-line options, do we?
<civodul>i don't see any in C-h i
<raghavgururajan>leoprikler: Would you be able to review Telegram-CLI (#45954) as well?
<leoprikler>Currently in a meeting, but Telegram-CLI requires Telegram to be merged first, right?
<raghavgururajan>Oops sorry. Ah yeah.
<apteryx>civodul: we don't but I think it'd be great to
<apteryx>I'm not sure if there's something for this in Texinfo, or if we'd just add @cindex for them.
<apteryx>I know other GNU manuals have handy indexes for commands such as the GNU Bash manual.
<apteryx>well it's not for options, but for commands such as [, ], for, etc.
<apteryx>oh, it's the Emacs manual which has them
<apteryx>try i then -- TAB, you'll see its available options
<raghavgururajan>civodul: It seems #guix now only allow registered users to join. I think the command `/mode #guix -v ozark` caused unintended effect.
<raghavgururajan>sneek, later tell civodul: It seems #guix now only allow registered users to join. I think the command `/mode #guix -v ozark` caused unintended effect.
<sneek>Got it.
<zimoun`>rekado: about the importers, the issue is that main of them return #f when something is wrong instead of (values #f ’()); so then ’recursive-import’ is broken because the ’lookup-node’ is waiting to unpack ’values’. Patch#45984 shows that replacing by ’(values #f ’())’ fixes the issue; especially bug#44115.
***zimoun` is now known as zimoun
<civodul>mothacehe: you might have lost connection to the OverDrive because i'm experiencing networking issues
<sneek>civodul, you have 1 message!
<sneek>civodul, raghavgururajan says: It seems #guix now only allow registered users to join. I think the command `/mode #guix -v ozark` caused unintended effect.
<civodul>raghavgururajan: how can it be reverted?
<civodul>sorry for being an eternal IRC noob
<liltechdude>how to fix this?
<mothacehe>civodul: noted, thanks!
<mothacehe>looks like the guix-master evaluations are stuck :(
<mothacehe>the guix-daemon child process associated to the evaluation is reading fd 4 undefinitely
<apteryx>civodul raghavgururajan It's the r flag I think
<apteryx>We have to remove that flag if we want unidentified users to be able to join (
<civodul>mothacehe: fd 4 is the client i suppose, right?
<civodul>it may be that the client is "doing something else"
<civodul>so you should strace the client and see what it's doing
<mothacehe>yes it's a socket apparently, doesn't help much :(
<civodul>check the client PID in "guix processes"
<civodul>(it's while doing this kind of investigation that i ended up writing "guix processes" :-))
<mothacehe>client (evaluate) is polling on fds 3 7 12
<mothacehe>mhh, maybe i should strace from the start a master evaluation then
<nefix>Hello! Can I run an AppImage inside Guix?
<civodul>nefix: not sure, you should give it a try
<nefix>civodul: it doesn't work, I'm currently getting some progress using NixOS's
<apteryx>civodul: should we '/mode #guix -r' to restore non-authenticated IRC users?
<civodul>apteryx: sure, i just didn't know what to do :-)
***ChanServ sets mode: +o civodul
***ChanServ sets mode: +o apteryx
***civodul sets mode: -r
<civodul>done! (right?)
<apteryx>looks ok!
<civodul>the script above is worth a look
***ChanServ sets mode: -o apteryx
<civodul>damn, even that i don't know how to do
<apteryx>via /MSG ChanServ DEOP #guix yournick
***ChanServ sets mode: -o civodul
<civodul>ah thanks :-)
<apteryx>I'm just stealing commands from Tobias' cheat sheet ;-)
<apteryx>I can never remember any IRC stuff myself.
<civodul>i guess i need that cheat sheet too! where is it? :-)
<apteryx>should be in your INBOX anytime now
<civodul>cool, tx
<apteryx>you're welcome
<apteryx>I'm trying to debug the rust bootstrap... Starting at rust 1.29 built from mrustc 0.9. Currently fails with: output/rustc-build/rustc_binary "-C" "linker=/gnu/store/hjfcp3a70xb9psn9llvpikpwxhwldz89-gcc-7.5.0/bin/gcc" "-Z" "force-unstable-if-unmarked" "-L" "output/target-libs" "src/libcore/" "-o" "output/target-libs/libcore.rlib" failed with status 1 -> error[E0601]: `main` function not found
<apteryx>in crate `lib`
<apteryx>If someone has ideas, let me know!
***Anonymous____ is now known as Anonymous__
<roptat>hi guix!
<civodul>hey roptat!
<roptat>since there were no recent comment on, I'd like to go forward and push the changes, and publish the blog post as soon as possible
<paulj>Evening all! I have just made a new installation of guix on a Lenovo X270 - looking forward to learning more. It'll make a change from 16 years of Gentoo! Depending on how I get on, I may migrate all my systems over.
<roptat>to summarise: we're switchig to weblate for translations. The changes replace mentions of the TP with mentions to weblate, and the blog post explains the translation mechanism for the website.
<roptat>are we good? Am I missing something?
<roptat>paulj, welcome :)
<paulj>Thank you! In time I hope to be able to contribute.
<mothacehe>paulj: nice! did you use the graphical installer?
<paulj>I did - it worked flawlessly. Everything seems OK, with the exception of the wifi (which I expected). I now need to work out what should be installed at system level, and what should be at user level. It's quite a different paradigm to that I am used to.
<mothacehe>paulj: great to hear, that's sadly not true for every new user :p
<mothacehe>regarding package you probably want to have most of them at user level
<paulj>I think the lenovo laptops have been good for linux for some time, which is why I have it. I'll follow your advice for the package installation!
<paulj>I would like to get coreboot on it, but it looks like this isn't yet working for the X270.
<mdevos>Could someone take a look at mkdir-p/perms in %dovecot-activation in (gnu services mail)? The hardcoding of "/var/run/dovecot" seems rather suspicious to me, but perhaps it's actually intended (idk)
<lfam>It's surprising but some of my "live profiles" have gone missing
<lfam>I'm seeing "/root/.guix-profile/etc/profile: No such file or directory"
<lfam>The final referent of the symlink chain from ~root/.guix-profile to /gnu/store is indeed not there
<lfam>Here are the generations of the `guix pull` profile:
<lfam>This is not the profile that is broken, but I assume some change in guix caused this
<lfam>I filed a bug
<joshuaBPMan>Hey guix people.
<pineapples>leoprikler: Do you remember when I shared my concern regarding Telegram possibly not being able to run under Wayland? I mentioned that a number of packaged Qt programs cannot run under Wayland even though they should be able to. Now I know why that is the case — "qtwayland" is missing from their inputs so they were built without Wayland support in the first place.
<leoprikler>Ah, but everything works fine for Telegram, because it has qtwayland, right?
<pineapples>Yes, that's correct
<efraim>i wonder if kwayland should propagate qtwayland
<pineapples>It would make perfect sense in my opinion
<pineapples>On the other hand, kwayland is not a dependency of all Qt-based software
<pineapples>On the second thought*
<civodul>lfam: could it be that you ran guix-daemon from a checkout with the wrong localstatedir setting?
<lfam>civodul: I never run the guix-daemon from a checkout. It's not part of my workflow
<lfam>I do a lot of work from the checkouts, though
<lfam>So, it's possible I misconfigured the checkout once, but I definitely didn't run the daemon from there
<avalenn>what does "ice-9" mean in all core libraries of Guile?
<ft>avalenn: — "There, he learns of a substance called ice-nine, ..."
<avalenn>One book to add to my reading list perhaps.
<avalenn>And thanks for the instructive mail from Andy Wingo.
<roptat>civodul, should I ask the TP admins to disable our string domains?
<zimoun>civodul: v3 about hint sent! Thanks for the review, really improve. :-) What is missing is to replace all the args-fold* by parse-command-line. Does it make sense?
<civodul>roptat: yup i guess so; is weblate fully ready?
<civodul>zimoun: i'd have to check, i'm not sure what differences there are
<zimoun>difference, habits? ;-)
<paulj>Could I clarify a point on package installation? If I install a package with dependencies, do I automatically have access to them? For example, installing nyxt installs sbcl as a dependency, but it doesn't seem to be accessible in a shell. Do I also need to explicitly install sbcl if I want to use it?
<apteryx>paulj: only if the input was a propagated input
<apteryx>normally we prefer using normal inputs as this prevents collisions in a profile and composes better
<paulj>Sorry - you're going to have to decode that a little for me! My understanding of what you have said: if I want access to sbcl, I should directly install it. I guess this also means that if I remove nyxt, sbcl won't be garbage collected.
<apteryx>Guix will keep tracks of things it refers to
<apteryx>so if you only installed nyxt, which requires sbcl (not propagated), you'll have nyxt along sbcl under /gnu/store, but only nyxt will get installed to your user profile.
<apteryx>If you later uninstall nyxt and run 'guix gc', the sbcl will be garbaged collected along nyxt, since nothing refers to it anymore.
<paulj>... and when I installed sbcl afterwards, it set the links up so I can access it directly. At least that is how I understand it!
<paulj>I am interested in setting up dwm, but as this normally means editing the source, I need to read more about how this should be done.
<paulj>I am sure I came across something about how this can be done in the handbook, but I need to go back and find it.
<bavier[m]>ewww... 207MiB tarball, expanded has 562MiB of test data.