IRC channel logs
back to list of logs
<podiki[m]>you may need font-dejavu and to recreate the fc cache <orneb>Yes, I'm on Guix System thanks! Are you also familiar with this error that I get after starting Sway on tty1? [backend/session logind.c:511]User has no sessions [backend/session/logind.c:559] Couldn't find an active session or a greeter session <vagrantc>you might need to add an elogind service? <orneb>To run Sway I had to add export XDG_RUNTIME_DIR=“/.run/swaywm" to my .bashrc then mkdir -p $XDG_RUNTIME_DIR and chmod 0700 $XDG_RUNTIME_DIR, but I get those errors. Is this because no display manager is installed? <vagrantc>orneb: it's been a while since i regularly used sway on Guix System ... but launching from tty was pretty much the way i did it for maybe a year <vagrantc>i'm just browsing my old system config and picking out things i vaguely recall having to add <Maya[m]1>if you enable gdm, sway just works fine, ir has an desktop entry iirc <Maya[m]1>i actually used home-sway-service-type from rde, i cant recommend it enough <orneb>Thanks vagrant, I'll try with elogind and see what happens. <vagrantc>i tried getting the smallest configuruation possible, as i was running it on a pinebook pro and aarch64 substitutes weren't generally available reliably <orneb>:Maya I prefer not to run a display manager since I'm using an old laptop and it already takes a while to reach the login prompt despite having an SSD. 20s to unlock my luks drive, then I have to enter my passphrase a second time. I intend to enable the auto login for the user on tty1 as soon as I fix the Sway error. Was wondering if something like this were possible on Guix. https://wiki.archlinu <orneb>x.org/title/Dm-crypt/Encrypting_an_entire_system#Avoiding_having_to_enter_the_passphrase_twice <vagrantc>orneb: i doubt that's possible, because the initrd is in the public store, so would be a security issue for anyone able to read the initrd <vagrantc>i'm not sure how secrets are managed, but guix would have to get some more clever post-processing of files in /boot to make that possible... to avoid embedding the key in the store which is publicly readable by design <vagrantc>e.g. append the key to the initrd when generating the files in /boot rather than just linking to the files in the store <clever>vagrantc: on the nixos side of things, there is support to package up secrets when generating the /boot config files <orneb>vagrantc: understood, thanks! <vagrantc>yeah, if nixos has already done this, you might be able to port the idea to guix... <robin>i'd definitely prefer that to entering a long LUKS passphrase twice in two different keyboard layouts...(and no typos allowed at the grub prompt!) :) <robin>but i don't see an easy solution with guix as-is (haven't looked at nix's approach yet) <robin>(though frankly, i'd be fine with just fixing the keymap in grub and/or allowing multiple unlock attempts) <apteryx>yes, we should implement a retry mechanism for GRUB <jab>robin: I just use a keypass phrase that is numbers only. You can also use special characters too. :) <robin>jab, i'll probably try fiddling with grub configuration first, but that's a good workaround :) <robin>in theory grub could also use the systemwide keyboard layout (grub has some feature already for keyboard customization), but that's probably considerably trickier to implement <jackhill>spaethnl: the site is down. As I understand it someone needs to visit the datacenter to restart the server hosting it, and they're on vacation. <apteryx>fun, it looks like my xvnc-service-type is springing into life <spaethnl>jackhill: Hmm.. 'We can’t connect to the server at debbgus.gnu.org'. Is that URL correct? <spaethnl>Unfortunately those bugs didn't help. I'm trying to get my network printer working. I've added the printer, but when I attempt to print a test page via the CUPS web admin interface I get "Unable to locate Printer "BRW4CEBBD851047.local".". My understanding is that I had to add avahi to the services, which I've done, but now I'm not sure what might still be wrong. <spaethnl>Anyone know if there are additional things I need to configure to get CUPS to print to the network printer. My understanding is that even if it is able to discover it, that is different from being able to connect to it? <jackhill>I think this channel tends to me most active in the European daytime, so if you don't get an answer, and you're able to, I might try then, or email email@example.com <ncbfg36>Does anyone know of a good replacement for intel 7260 wifi card that works with linux-libre kernel? All I have laying around is an atheros card from a T500 thinkpad but it's the wrong size for this slot (T440p) <ncbfg36>jackhill: this laptop is running osboot firmware so whitelist is not an option. <jackhill>ncbfg36: super, that's great! Unfortunately, I don't have a great recommendation about what will fit, sorry <ncbfg36>jackhill: no worries :-). It seems to be a smaller form factor than regular mini pcie. Less wide and a little longer <ncbfg36>oh i'm just confused. It's an m.2 2230 slot. Do realtek wifi chips work with free firmware or do I need an old atheros chip only? <Idliketointerjec>Hello, guix! When I upgrading packages without substitutes guix-daemon is breaking. Now it is trying to build rust-1.39.0 and later for sure, as befor it will show error: corrupt input from socket. So again I have to restart guix-daemon and then again it will break. Anyone could say what is the problem? <Maya[m]1>orneb: i get that, i just know it works, yet this is a good reason not to use gdm
***wielaard is now known as mjw
<attila_lendvai>something needs to be done for a flatpak app to be able to open files in my home... some dbus or xdg magic, but i cannot find it in my notes. any help, even search keywords, would be appreciated! <vivien>attila_lendvai, isn’t the job of portals to give access to files? <attila_lendvai>vivien, yep, i'm honing in on it with searches. i think i must install xdg-desktop-portal-gtk and probably xdg-desktop-portal-wlr, too. but they are installed... :/ *attila_lendvai tries an update *attila_lendvai notes that if there was a proper wiki, he would have recorded it there, and found it quickly next time under the flatpak entry <nckx>There are several wikis IIRC, so I don't believe you :) <attila_lendvai>there's no guix wiki, where i can search for something and only get guix related stuff (this is a moot point here though, because i suspect that this issue is not guix specific) <nckx>So if you mean ‘competently-hosted wiki’, I'm forced to agree, but with out sysadmin track record we won't be competing any time soon. <nckx>‘GNU Guix — we probably still have a home page?’ <nckx>attila_lendvai: I think that if you're not using one of the two Great Wiki Distros (I'm still counting Gentoo, although they seem to've waned) you're automatically out of luck :) <Maya[m]1>I have managed to get gdm fingerprint authentication!! <Maya[m]1>a pam module was missing, i’ll clean it up and submit a patch <nckx>I was just wondering if our fprintd service still worked. *nckx got a working fingerprint reader. <Maya[m]1>i just don’t know if i set it up correctly so i’ll submit it open to suggestions <Maya[m]1>(correctly as in securely, it works but i don’t want to compromise guix user’s safety) <mbakke>why do some rust packages have the same rust inputs in #:cargo-inputs and (inputs ...)? <mbakke>oh, looks like they won't get propagated to dependent packages unless they are in inputs? <unmatched-paren>because Cargo expects every input in the dependency tree to be in the registry at build time <mbakke>I had one package complain that the requirements of rust-dlib (a #:cargo-input) was not satisfied; when I added an (inputs ...) field to rust-dlib with the same packages as #:cargo-inputs the build succeeded <mbakke>bizarre indeed; other packages such as rust-wayland-server uses a mix of #:cargo-inputs and (inputs ...) *mbakke has no idea how that all works <apteryx>a more concrete question: is someone using XDMCP with GDM or another one of our login managers? <Maya[m]1>is there anywhere a guide for define-configuration? <Cairn>I've added a quirk in /etc/libinput and when I run `libinput quirks list --data-dir /etc/libinput /dev/input/event17` it shows my modified value <Cairn>When I remove the --data-dir flag though, it just shows the default. I can't find anywhere in the docs that suggests I would need to do anything to get my quirk applied other than add it to the dir, but I'm not sure how it's supposed to recognize it. <Cairn>Is it just always watching /etc/libinput for changes? Seems bizarre <q-posev>Hello everyone. I am packaging new library with Guix and encountered an issue. So the software has an Autotools build system in place, but the source code is generated, which means that the typical autogen.sh includes one additional line which parses the output of the `git tree command` and generates the Makefile.am file. Building from the source <q-posev>code distribution works like a charm. However, when I attempt to build from the fixed Git commit, I noticed that actually there is no `.git` folder in the /gnu/store/......-checkout/ folder which is produced by Guix before the `unpack` phase. Because of that, my code generation fails. So my main question is: is it possible to force Guix to keep the <q-posev>`.git` folder after cloning and checkouting my Github repository? I suppose that this might be a common use case since some programs include the Git hash info in their metadata, but I could not find anything on the web. Thank you in advance! <nckx>I don't think that's possible. <nckx>Programmes that try to embed git hashes are just fixed not to do so. <nckx>Not sure what you mean by ‘git tree command’, so can't give more specific guidance, but the thrust will always be ‘patch the build not to try to run git at build time, 's bad’. <nckx>s/embed/sniff out/, because there's nothing stopping you from specifying a ‘commit’ in the Guix package and hard-coding it during the build. It's only the ‘let's run git assuming we're in a git repository’ that's not supported. <q-posev>I see, thank you! We parse the output of the `git ls-tree --name-only -r` command to get a list of files to process since we do not want to hard-code it. Even if running git at build time is a bad practice, I was surprised that the `.git` folder is removed by Guix. Is there any practical reason for that?
***stikonas_ is now known as stikonas
<Cairn>Ok, package question. I see that my active libinput package has its own version of /etc/libinput: `/gnu/store/yfmgjhh5qhgni8i47p7yyy313sb2b9z4-libinput-1.19.2/etc/libinput/local-overrides.quirks` <Cairn>It looks like libinput's only accessing this file. <Cairn>So how, in my Guix configuration do I modify this file with my custom value? <Cairn>I know I'm not supposed to touch /etc on its own, I can't get libinput to recognize the file even in ~/.local/share/libinput. So what's the guix method of adding the lines I want to /etc/libinput/local-overrides.quirks? <antipode>Cairn: Doing a quick "git grep -F libinput", Guix does not make any libinput configuration files. <antipode>So it looks like etc-service-type is your friend here. <antipode>(Though possibly libinput is looking in /gnu/store/.../etc, as the etc directory has not been overriden in the libinput package definition) <Cairn>Yes, it's looking in the /gnu/store/.../etc. <Cairn>I can tell because it lists the directory when running `libinput quirks list --verbose` <mbakke>since 'librsvg' is (now?) a staging-level package, I'm not sure 'librsvg-bootstrap' is necessary ... even though it clocks in at 3k dependents <Cairn>And it doesn't list my temporarily created /etc/libinput/ file <Cairn>Is that a packaging issue? Should libinput not include a /gnu/store/.../etc file? <Cairn>And is there someway I can replace it? <antipode>It is important that the package includes /gnu/store/.../etc files, such that the defaults are available (as defaults, or as a basis for a customised configuraiton) <antipode>I guess that the problem is that it unconditionally uses /gnu/store/..., even if /etc/libinput exists <antipode>It appears that libexec/libinput/libinput-replay actually looks at /etc/libinput <antipode>(I don't think it's actually run though, since it as an unpatched shebang -- if it does, there are some purity issues) <antipode>To be clear, I am not aware of a method to tell 'libinput' to look in /etc. <Cairn>Should I make a post in guix-devel about it? <antipode>Sounds more like bug-guix@ to me, OTOH there might be a method we simply didn't know about. <Cairn>Ah ok. Well I'll mess with it a little bit. <Cairn>I'm curious if using a .local file will solve it and maybe I just haven't placed it there <Cairn>Oh, is issues.guix.gnu.org down? <nckx>q-posev: I don't think it's reproducible, and it's generally huge compared to the source, and to any value it would bring — we've come this far without it… Those are just the first 3 possible reasons that popped into my head. <Lumine>Cairn: it has been down for a week now <nckx>A project refusing to even build without downloading the full commit history being ‘bad practice’ is just my opinion. <nckx>Cable-plugging people are on holiday.
***ChanServ sets mode: +o nckx
***nckx changes topic to 'GNU Guix | https://guix.gnu.org | issues.guix.gnu.org is down: use https://bugs.gnu.org/NUMBER instead | videos: https://guix.gnu.org/blog/tags/talks/ | paste: https://paste.debian.net/ | Guix in high-performance computing: https://hpc.guix.info | This channel's logged: https://logs.guix.gnu.org'
***ChanServ sets mode: -o nckx
<Cairn>Does anyone have a good example of etc-service-type in their config? Trying to test adding that file to /etc so I can be thorough for my bug report <Cairn>I see that libinput doesn't have a service-type, so should I just make a custom service type instead of modifying? <nckx>I've heard that some particularly advanced distros even package it. <raghavgururajan>By the way, I didn't get a chance to read the maillist thread regarding teams. Are they meant to be self-nominated? <nckx>Add yourself, or ask to be. *raghavgururajan is hearing voices in their head: "desktop" "gnome" "messaging" "xmpp" ... <Lumine>I'm hearing voices too but they're telling me to eat <nckx>If this turns into a philosophy flame war I *will* eat someone.
***Dynom_ is now known as Guest7333
<nckx>Cairn: Thanks for the bug report. <Cairn>nckx: New at this; hope it's enough info to work with. =) <nckx>podiki[m]: Sure, or graphics or game-dev... <podiki[m]>that includes a lot more than one header; looks like lots of tests and stuff. I was just going to package the header (plus the extension one), since I have no idea how to navigate the tests <davidl>sneek: later tell mbakke: how is it going with the ganeti-instance-guix pull request? It would be nice to have it merged and an updated guix package on guix master.
***mark_ is now known as mjw
*apteryx muses about a "*unspecified* considered harmful" blog post <podiki[m]>so this mikktspace library, it is just a .h and .c file, but think it just gets used in other projects rather than compiling as a library <podiki[m]>e.g. for a package I'm working on the makefile wants to compile the .c to a plain .o to be used <podiki[m]>I've made mikktspace a plain origin definition for now, other (better) options? <tribals>I'm trying to use guix as a provider for cross-compilation toolchains. It is a big mess to get "right" toolchain, especially when cross-compiling for ARM, so I'm trying to make my life easier a bit, with help from Guix. I know that guix support cross-compiling "natively", meaning that any package can be built for supported target with ease. But I can't figure out: could this be used for getting a toolchain to build non-guix software? <tribals>For example, if I need to cross-compile for arm (32-bit), could `gcc-toolchain` package be used? <tribals>Does it mean that there is no mentioned arm target? <lilyp>podiki[m]: I think providing a compiled package might still be useful. You could patch your makefile to replace mikktspace.o with -lmikktspace for example <podiki[m]>tribals: I don't have answer for you other than that shows what archs the package itself can be built for <podiki[m]>lilyp: so then for mikktspace I would just invoke a cc-for-target with -o file.c, and put that in #$output/lib? <podiki[m]>or rather use some cflags to build the .so you mean? <lilyp>don't forget to also add the header in #$output/include <tribals>podiki[m]: thanks, I was not aware of that manual section <podiki[m]>lilyp: do you have any examples off hand of this manual .so building in a guix package? I'm guessing I want cc-for-target and some flags from the gnu-build-system? <lilyp>not really, but there's a few examples of the contrary, i.e. pulling an origin into another package <lilyp>does this thing really have no makefile or other thing to speak of? <podiki[m]>right, thought of using additional origin in inputs too <silicius[m]>finally sent my hydrus network patches. Those waited for so long I had to update it two versions up <nckx>(substitute* … ((" [0-9]*B actual") " [0-9]*B actual")) <lilyp>nckx: hang on, what are you doing there? <nckx>The test suite happens to grep for a hard-coded number, which differs from reality (pending investigation, but the actual values are correct so I suspect fs shenanigans). To neuter this check, I patch in the same regex I used to find that number, and because it's in a grep it all happens to work out. <nckx>(At the risk of over-explaining the joke, but oh well.) <lilyp>Oh, okay, so it does have an actual use case. <lilyp>btw. do we have a fix for fontpens yet? <nckx>It might not make it into the final patch but yes. <podiki[m]>what flags should I use in this manual build of mikktspace? "-O2 -g -fPIC -shared" reasonable? (and -o libmikktspace.so) <nckx>I mean, it's just a docstring rounding error, the hard part isn't fixing it but what a Correct fix for that looks like. <nckx>To me, putting an example of round(return_5_point_3()) = 5 or whatever in a docstring is silly. <nckx>podiki[m]: I guess? This sounds like a wheel that many distros would have had to solve before you. <nckx>Does, e.g., Debian not have a Makefile you can snarf in its entirety? <podiki[m]>I found only one sorta package for this library, I think everyone just grabs the 2 files in their project and uses it <nckx>It doesn't look *un*reasonable, no. <podiki[m]>and since that accounts for 100% of users and their uses... <nckx>I don't actually understand the comment below that. <nckx>I don't use fontPens of course. <lilyp>I think the point is to reuse the doctests from approximateCubicArgLength (?) <lilyp>but I think that just hides the issue, because numeric errors in floating points can still happen <lilyp>so imho your round was the right suggestion <nckx>lilyp: I can't read it that way but you may well be right. <nckx>lilyp: Actually I think both suggestions are suboptimal, but I don't have a better one. <nckx>It probably doesn't matter. Hardly anyone will read those strings, ever. <podiki[m]>alright, think I'm just one package description away from a polished haxe patch! looking pretty clean now after this debundling of those little packages (minimp3 and mikktspace) <shcv[m]>how important is the url entry in a channel declaration? <nckx>There is something sinister about that question. <nckx>shcv[m]: …what do you mean, exactly? <nckx>The obvious answer of course being ‘yes’, but you clearly have something in mind :) <shcv[m]>eh, I just noticed in the "Package modules in a sub-directory" example that there was no url declaration <shcv[m]>and on my channel, since I use it for local dev, it complains that the local version is likely a "mirror" of the public repo <nckx>Right, that's the main use case. To warn users about that. <nckx>(Because the url field is authenticated.) <shcv[m]>I'm also curious what the best way to declare a channel that optionally depends on local and public versions of another channel is. I was thinking I'd just put two versions ('public' and 'local') of the channel in the same repo <shcv[m]>idea being, I'll have make each of my packages their own channel, to simplify development, and then have an additional 'meta channel' that merges all of the other ones to be easier for others to use
***spaethnl is now known as Guest1901
<shcv[m]>hmm. If I define a package in the same repo as the source, how should I reference the source files? if someone pulls the channel does it download everything in the repo? is that a bad idea? <apteryx>is there an easy way to get the message ID from Gnus? <apteryx>nckx: I just sent a trivial v2 for 56799 (2 lines changed)
***rgherdt__ is now known as rgherdt
<apteryx>I don't understand what the "file-append, output" is testing; is it possible to use file-append with specific outputs? <Guest1901>I'm trying to get my network printer working. CUPS is able to discover the printer on the network for me to add, but after I've added it, when I attempt to print a test page via the CUPS web admin interface I get "Unable to locate Printer "BRW4CEBBD851047.local".". My understanding is that I had to add avahi to the services, which I've done, but now I'm not sure what might still be wrong. Anyone know what I addition
***Guest1901 is now known as spaethnl
<nckx>Guest3605: I have that same problem, really annoying. Cancelling and then restarting the job works around it. <nckx>I use that printer about twice a year so not worth fixing for me, personally, I'm afraid. <nckx>Also a Brother, although that's probably coincidence. <spaethnl>Hmm, canceling and restarting does not fix it for me :/ <apteryx>I'm confused by the Vinagre VNC client connection window; there's a 'Host', fair enough. Then there's also 'Use host [ ] as a SSH tunnel'. If I put a host there, what do I put then in the main 'Host' field? <shcv[m]>apteryx: it looks like file-append produces a record containing 'base and 'suffix that is then processed by the gexp expander in a way that uses the output. So, that test is to verify that the output appropriately affects the results <shcv[m]>you can look at file-append-compiler and the surrounding code in guix/gexp.scm for more details <apteryx>shcv[m]: thanks! isn't it invalid to use more than 2 arguments with file-append? <apteryx>OK, no the signature is: (file-append base . suffix), so it captures whatever 2+ args into the suffix list <shcv[m]>probably just an intermediate proxy that it connects to first, then tries to connect to the main host <apteryx>these strings are then string-concatene'd in file-append-compiler... seems a dubious feature <apteryx>I guess it could be useful if you were to format the string, but reckon 99% of examples in the source code only provide 2 arguments to file-append :-) <shcv[m]>they aren't "just" concatenated - since each is a separate object, they could involve stuff that gets processed by the gexp compiler / expander <shcv[m]>actually nvm, it looks like only base is expanded <apteryx>I don't think so, 'suffix' is expected to be a list of strings, as mentioned in guix/gexp:666 <shcv[m]>how do I declare a package with local source? <apteryx>OK, I'm connected to my remote machine via VNC, but I get a black screen. I was expecting to see my fancy SlIM login manager. <apteryx>shcv[m]: you can use the git-checkout procedure for the source <shcv[m]>the package reference says that any "file-like object" can be used in the source field, and it even mentions local-file, but I'm having trouble getting that to work. It sounds like it wants a single file, not a directory... <apteryx>you could always use fetch-url with a local uri such as file:///home/src/your-project <apteryx>but if you use git the <git-checkout> option is easier/better (from (guix git)) <shcv[m]>I'm using `(local-file "../<dir>" #:recursive? #t)` <jackhill>I gvfs-mount provided by any package? Is that a deprecated tool? I'm trying to mount smb:// volumes from the command line without root