IRC channel logs
2024-01-13.log
back to list of logs
<below_turtles_>how to activate a cuirass spec? Clicked the deactivate specification button... <roran>freakingpenguin: can pass work on android? <freakingpenguin>roran: There's an app on F-Droid called Password Store. It can integrate with open keychain if you have a hardware key. <roran>Very interesting. How does losing said key end up for users, usually? <freakingpenguin>Badly if you don't back up the private keys to a flash drive or paper printout during creation. <roran>Sorry, below_turtles_; I'm just a dummy, so I dunno. <roran>freakingpenguin: so you CAN extract those keys. i've wondered a lot about that <freakingpenguin>roran: You generate the keys off the device, back them up, then move then on the device. Once it's on the device you can't get them back off. <roran>Aha. Does the device have some sort of "yeah now get rid of those" mechanism/ <freakingpenguin>I think that depends on the device? I use a yubikey 5 which supports creating a "fake" pin that wipes the GPG keys. <roran>very nice; thank you freakingpenguin <guestinator>hi guix, guix pull lists the channels in my channels.scm and succeeds, but afterwards guix describe only shows the guix channel. what might be happening? <freakingpenguin>guestinator: Does guix describe show the time when you last run pull? I also see missing channels if I pull as a normal user and describe as root fwiw. <guestinator>it appears to be happening when I execute a symlink to ~/.config/guix/current/bin/guix, but not when I execute it directly (well, “directly” as that is itself a symlink to the store). which is weird <guestinator>particularly weird since pulling grabs all the channels just fine <guestinator>seems like guix might look for channels.scm using a relative path based on $0 sometimes? weird <lilyp>apteryx: #~'() is equivalent to #~(quote) <apoorv569>Wasn't catch2 compile issue fixed? I just pulled but still having compile error. <jpoiret>heh, funny comment around the package definition for Perl: <jpoiret>“Yeah, Perl... It is required early in the bootstrap process by Linux.” <jpoiret>it just built fine for me on x86_64 but I don't have the same derivation as you <jpoiret>you could try to find the output of the test that's failing earlier in the log, that would probably help <apoorv569>For `guix` channel I'm on, c0b303aaa3d6154acbe054120d11467eb98e6d33 according to `guix describe` <jpoiret>apoorv569: since those are tests for a testing framework, it's the tests that are being tested that mostly fail. Seems there's one actual failing test, but I can't pinpoint it :p <jpoiret>apoorv569: huh, it seems that it's the test "uniform_floating_point_distribution is reproducible - float" <apoorv569>jpoiret: I see. So its something to do with my system? <jpoiret>their documentation seems to say that their implementation of it should be reproducible on all hosts, but they might not have checked thoroughly enough <jpoiret>no issues open about it though. If you can reproduce manually (ie. by building it yourself manually), they might be interested in a bug report. <jpoiret>`guix shell -D catch2` rather, to have all dependencies, then try to build the latest tag from their repo <jpoiret>if the same test still fails, then it might be an actual problem <jpoiret>i haven't seen any changes to that part of the code since 3.1 but I might be wrong <apoorv569>jpoiret: I don't know how to build the latest tag.. like what do I run? `guix shell -D catch2` then inside the shell what to run? <apoorv569>guix shell -D catch2 then guix build catch2 seems to be just fine.. also I don't see the compilation it downloads substitute I guess. <freakingpenguin>catch2 is broken specifically with --system=i686-linux. Don't think the patch made it in yet. <freakingpenguin>Also I'm pretty sure shell -D would be if you wanted to build it outside of guix, I don't think? it matters to guix build <jpoiret>freakingpenguin: yeah, the goal was to reproduce it manually without going through `guix build` (upstream maintainers wouldn't really like having a guix build bug report) <jpoiret>freakingpenguin: do you have a ref for that patch? is it a guix patch or an upstream patch? <janneke>so, on core-updates, git-minimal has <janneke> #$(this-package-native-input "bash-minimal") <janneke>the Hurd cross-build fails like this <janneke>%exception #<&invoke-error program: "make" arguments: ("-j" "8" "V=1" "SHELL_PATH=/gnu/store/arwrkg9qpp8fcr7q80zccl3hhaf9b944-bash-minimal-5.1.16/bin/sh" "TEST_SHELL_PATH=/gnu/store/ifmrcim21rv0k0crpsn3fw0mj0hwn3jl-bash-5.1.16/bin/bash" "NO_INSTALL_HARDLINKS=indeed") exit-status: 2 term-signal: #f stop-signal: #f> <janneke>where file $(readlink -f /gnu/store/arwrkg9qpp8fcr7q80zccl3hhaf9b944-bash-minimal-5.1.16/bin/sh) gives: <janneke>/gnu/store/arwrkg9qpp8fcr7q80zccl3hhaf9b944-bash-minimal-5.1.16/bin/bash: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked, interpreter /gnu/store/kp95j1hpn27dhdf8dfi8ghjqfc94w305-glibc-cross-i586-pc-gnu-2.38/lib/ld.so.1, for GNU/Hurd 0.0.0, stripped <podiki>A catch2 fix was reported here but personally haven't had a chance to try <rekado>janneke: does #$ vs #+ make a difference here? <HikariNee>The GHC package on Guix has been out of date for a while now, GHC moved to 9.4 for LTS, while looking at the source I saw guix still packages the extremely old versions like 4.4 and 8.8, surely its been long enough to consider dropping those? <HikariNee>I would love to update the package but GHC is too large for my computer to compile without freezing <rekado>hibiki: these extremely old versions have been added with great pain <rekado>the old versions of GHC are meant to be used to bootstrap more recent versions of GHC <rekado>so we will not drop them unless a different bootstrap can be found <freakingpenguin>Those package definitions look terrifying, props to everyone who figured them out. <singpolyma>Hopefully a new GHC fixes the show stopping bug that keeps me on time machine to older guix for now on my haskell projects <singpolyma>OTOH so thankful for time machine making my life possible right now <mwette>I'm working on spec for several related packages. To compile B, B needs compiled A, cmpiled A needs files from B's source. What's the way to do that? Do I need to create a source-only package for A? <mwette>How about (native-inputs (package-source ,A)) ? Lemme try. <freakingpenguin>singpolyma: Even ghc-next (v9.4.4) is too out of date? It also looks like 9.4.8 was released and ghc-next can be built with the new revision. Haven't finished running tests though. <jpoiret>some testsuite failures with our new upgrades 🙃 it takes ages to rebuild <jpoiret>it takes longer than rust at least, but I suspect that's mostly because the testsuite is not run parallelized by default <jpoiret>ah, apparently having it on causes test failures <HikariNee>that feels good to know, thanks for your effort :) <freakingpenguin>podiki: re. catch2, if switching to clang isn't desirable I found the build succeeds with "-DCMAKE_CXX_FLAGS=-msse2 -mfpmath=sse". <jpoiret>I can't remember if SSE2 is considered ok <jpoiret>we have packages that explicitely disable SSE2, and some that explicitely enable it, so there is no consensus on this <freakingpenguin>I wonder if the clang fix works because clang might assume sse2 on 32-bit builds by default. <panosalevro>hey all, can someone help me with guix home? after installing packages via `guix home reconfigure`, i cant run any of the packages. it can only find the packages after `source /home/user/.guix-home/profile/etc/profile`, but i don't understand why the profile isn't sourced automatically. last time i asked, people suggested that i include a home-bash-service-type (which i was missing), but i still <panosalevro>can't run any of the packages without sourcing the profile manually. <Franciman>after you run guix home reconfigure. Do you start a fresh shell? <panosalevro>Franciman: well i tried starting a fresh shell, switching to tty, rebooting.. nothing helped <graywolf>As far as I can tell the installation image uses grub-bootloader. How would I go about switching that to generate a install medium with uefi instead? <mekeor>graywolf: i'd guess: write a operating-system declaration, inherit from the install-image's os-declaration, generate an iso from that <graywolf>mekeor: I mean, I can do that (I already do for unrelated changes), but I not sure how to configure the bootloader field. If nothing else, it will require an extra partition, and I am not sure how to specify the partition layout to guix system image invocation... <efraim>i'm getting compile errors on guile-2.0 on powerpc32 :( <graywolf>Ok, follow up question: Why is grub-efi-removable-bootloader not considered an EFI bootloader? Are there technical limitations I am not aware of? <Guest14>How can I test git send-email but not actually send the mail to test my config? <efraim>civodul: I was looking at bug28211 about using guile-2.2 for grafts and I tried guile-3.0 and got a segfault <efraim>GC-marker-7[6947]: segfault at 2010 ip 00007ffff7e19bc7 sp 00007ffff3ad0c30 error 4 in libgc.so.1.5.1[7ffff7e12000+1c000] likely on CPU 19 (core 9, socket 0) <Guest14>Does Sourcehut host websites like Github pages? Would explain why the website is down <graywolf>Yeah I would assume it is hosted on the same infrastucture (the git-send-mail.io is made by Drew deVault, owner of sourcehut) <Noisytoot>I think the tome4 package is non-free. See COPYING-MEDIA in the tarball it downloads ('All the medias located in all the "/data/gfx" folders are granted to use with the Tales of Maj'Eyal game only.'). This was reported a few years ago, but it was decided that it was allowed in Guix because of the non functional data exception in the FSDG: https://issues.guix.gnu.org/28193. I think this is wrong. If the game does not run without it, how is it <Noisytoot>non-functional? Also, it restricts freedom 1, because if you modify the source code, you cannot legally use the non-free media that it does not run without. <rekado>FSDG says: “For example, some game engines released under the GNU GPL have accompanying game information—a fictional world map, game graphics, and so on—released under such a verbatim-distribution license. This kind of data can be part of a free system distribution, even though its license does not qualify as free, because it is non-functional.” <rekado>the non-functional part is undisputed. <rekado>the only question is whether it is fine for the license to restrict use in other games <Noisytoot>I don't think it should count as a verbatim distribution license if it restricts use <rekado>the FSDG only mention permission to copy and redistribute for commercial and non-commercial purposes <Noisytoot>so non-functional data under a license that allows copying and redistributing for commercial and non-commercial purposes but does not allow use would be fine? <lilyp>that'd be a weird license if ever I saw one <peaceful_elion>Howdy, folks. Anyone have any advice as to how I could get an Emacs package with debug symbols? I don't see one in the repos and I'm not sure how I might compile my own. <gnucode>Hey friends, I'm planning on hosting a happy belated new year's party for the Hurd. Mostly we are going to be talking about the Hurd, but if you want to drop by feel free!