<rekado>Blackbeard[m]: “guix package -i mu” <civodul>weird thing: my Emacs gets inter-letter spacing wrong today <civodul>the space in between letters is slightly too large <civodul>and i don't think i changed or upgraded anything <civodul>looks like it's a problem with "Liberation Mono" specifically <efraim>update to fontconfig? or was that on staging? <rekado>civodul: do you have /var/cache/fontconfig/ ? <civodul>it seems to have been around for a long time <rekado>After I removed it the fonts in a Java application went from really ugly to really nice. <civodul>it spawned a new Emacs to make sure, but Liberation Mono still looks ugly there <civodul>maybe it's just a signal that it's time for me to change fonts <efraim>I just saw the lxqt patch(es) is from pantherx <civodul>if you interrupt it and start again :-), you should get substitutes <civodul>well, to be on the safe side, you can do: guix pull -n && sleep 300 && guix pull (protip) <civodul>and you're still not getting substitutes? <civodul>commit 835dada6e is definitely built form armhf-linux <civodul>so you could try specifically that commit <efraim>I pushed an update to my ietf channel <efraim>I'd have to check if I'm legally even allowed to work for them ;) <civodul>i don't think we've seen anyone from PantherX on our channels? <efraim>I'm guessing they're responsible for the Iran hits on the balancer <roptat>civodul, I tried with the specific commit, but now I'm building /gnu/store/ynhkall01n6v75m6l0ajyq4bv23i9gfp-module-import-compiled.drv <civodul>roptat: did you try to wget its output to see if ci.guix.info has it? <roptat>I don't have wget nor curl, and it's going to be a pita to build them :/ <roptat>oh, I could do it from another machine, right? <roptat>not sure what I should wget though? <efraim>IIRC wget is "smaller" than curl <efraim>yeah, curl lists openssl and gcc-7 <jonsger>civodul: I still get paid from SUSE for working on Guix (okay kind of) :P <roptat>so how did it build guix-modular? <civodul>from an x86_64 Guix, so it didn't need to build this "module-import-compiled" <roptat>it seems to be needed for compute-guix-derivation *civodul feels like he's /msg'ing a bug tracker :-) <civodul>roptat: yay! we're following along :-) <roptat>-bash: /run/current-system/profile/bin/guix : /gnu/store/253gbbg95axyhf18wd5fwjdwya7vqzwn-bash-minimal-4.4.23/bin/bash : mauvais interpréteur: No such file or directory <roptat>nothing in /run/current-system/profile seem to be working anymore :/ <civodul>roptat: you were just running 'guix pull', no? <civodul>that doesn't modify /run/current-system <roptat>I interrupted it because it was going to build everything <civodul>hmm so did the store get corrupted or something? <roptat>well, I was running nvim in another terminal, and now nvim doesn't start, and that comes from the system <roptat>but I can still run git from root's profile <roptat>I thought it was safe to interrupt guix anytime, the goal was to rerun it again and let some time for the substitutes to be baked <civodul>does dmesg show anything that would suggest a file system corruption? <roptat>that would be sad because the SD card is pretty new <efraim>hmm, now mesa doesn't build on aarch64 on staging <roptat>/var/log/messages doesn't show anything suspicious <roptat>(using mapfile to read files since I can't use cat) <rekado>civodul: thanks for the report; I’ll ask Arun about this. Could be a problem in guile-email. <rekado>civodul: thanks for your singularity bug reports and immediate fixes! ***rvgn-net is now known as rvgn
<civodul>rekado: re Singularity, did you test it on Guix System back then? <rekado>civodul: I think I only tested on Fedora back then. <civodul>they've rewritten the whole thing in Go anyway <civodul>BTW, i tested a CPU-intensive multithreaded program packed with -RR <civodul>i guess we pay the PRoot price at every futex(2) syscall <efraim>fixed mesa on aarch64 and armhf on staging <civodul>i hadn't thought about this "detail" <civodul>so after all, maybe we should try the LD_PRELOAD hack as well <rekado>that would allow us to limit interception to file system related syscalls, right? <civodul>yes, and technically we'd just interpose on a couple of libc functions; we wouldn't use ptrace at all <civodul>it's also more fragile because of that, but maybe "good enough" in many cases <civodul>it's sad that we have to rely on piles of hacks to work around OS weaknesses <apteryx>Q: how do I update a custom profile? It seems I can't 'guix package -u . -p /path/to/the/profile' <apteryx>hmm, I swear last time I tried it wouldn't work. It does now. Case closed :-) <roptat>so the system on my cubietruck is broken and unusable... last time I had an fs corruption I was able to recover from a less recent guix that wasn't corrupted in the store, but this time it seems that I only have one version of guix in the store and it's corrupted <apteryx>civodul: perhaps you've reached the level where bugs are fixed with a simple 'why can't you' ;-) <roptat>this is the end of my record: 3 years without breaking a system, thanks to guix :) <roptat>(I used to break them far more often, so that's great :D) <civodul>roptat: Guix can't fight file system corruptions :-) <lsl88>civodul: did you think that I was not going to be around? <efraim>hmm, is there a way to convert an aspell dictionary to a hunspell dictionary? <rekado>lsl88: did you overcome your “guix pull” problem? <nckx>lsl88: How did your ‘guix pull’ adventure end? *nckx is on public wifi, tunnelling through DNS, latency funtimes. <lsl88>civodul: when I say that I love my community and that I am proud of it, it is true. <lsl88>rekado: it did not finish yet, it is still guix pulling since yesterday's lunchtime? early afternoon? (of my timezone), maybe your 17:00 <rekado>lsl88: what is it doing right now? <lsl88>rekado: I added manually the substitute url to the command, then I will add it as the default, and reconfigure the system to have my swap on. My teenager computer will be a proud guixSD machine <efraim>what programs do we have for translating text? I tried offlate but it seems it only supports hunspell atm <lsl88>rekado:it is running tests, test 1117 right now <efraim>i mean for doing it myself, other than editing .po files by hand <rekado>lsl88: of what package? “git” maybe? <rekado>I’m asking because by now you might already be getting substitutes. <lsl88>rekado: it has just passed that test part :) and it is checking stuff, like C++\ <rekado>that might be a configure phase of some other package now <rekado>are you pulling a particular version of Guix? Or just the latest? <lsl88>checking-configuring, and like that. In almost all cases the checking is yes. <lsl88>rekado: the last one, only changing --substitute-urls <roptat>efraim, oh you're using offlate <3 <rekado>have you authorized the key for ci.guix.info? <rekado>because you really should be getting substitutes. <roptat>you can send me a merge request if you want to use aspell ;) <lsl88>rekado: so you mean that my machine is not using it because of the uptime? <roptat>efraim, you might want to try poedit, and I think there's an emacs mode (but I don't use emacs) <rekado>lsl88: I mean it looks like your machine isn’t using substitutes from ci.guix.info. <rekado>if you’re compiling everything for a day or more this seems like a reasonable conclusion. <rekado>because ci.guix.info has lots of substitutes. <lsl88>rekado: but that means that it is not using substitutes at all_ <rekado>that’s why I asked: did you authorize the public key for ci.guix.info? <rekado>if that’s not authorized you get no substitutes at all. <lsl88>rekado: no, because I was already running the guix pull command <lsl88>oh, the hell, thought that at least I was going to use hydra instead <nckx>lsl88: I linked you to a .pub key yesterday. Did you not authorise it? <nckx>lsl88: ci.guix.info is the only server that builds substitutes for ‘guix pull’. <lsl88>nckx: I was already pulling, so I was planning to add it afterwards. Thought "ok, I have already trhown the command, then I will add it properly so that I don't have to use the option again@ <nckx>Oh. Had I know that I would have told you to ^C immediately :-) <efraim>roptat: I created a guix.scm file for offlate, but unfortunately we only have hunspell dictionaries for english and french <lsl88>rekado: what do you mean? but isn't it a transaction? <lsl88>nckx: didn't get it, so then I will cancel it now and add the key <rekado>lsl88: completed builds end up in /gnu/store; they won’t be removed when you abort “guix pull”. <rekado>only unfinished builds are removed. <rekado>it’s transactional in that you won’t even get a half-pulled Guix. <nckx>lsl88: Oh well. Nothing lost but time. <lsl88>nckx: I tried a wget and got 404 <nckx>Indeed. It's a symlink and cgit doesn't handle that well. <Blackbeard[m]>what packages options do we have on guix to send email from emacs <lsl88>nckx: but there is no plain, that is what I don't get <nckx>cgit doesn't handle symlinks, is all. <lsl88>got 404, so went to the browser <nckx>lsl88: Yes. So I posted the symlink target above. <lsl88>now I have to add it with archive, and set it the default one <lsl88>uhm, I downloaded the berlin one <nckx>The last link was to show that ci.guix.info is a link. It points to the berlin key. So all is well. <nckx>Since changing the defaults will only take effect *after* you've reconfigured. <apteryx>python fails on core-updates because of some SSL issue (2 tests failed: test_httplib test_urllib2_localnet). Known issue? <lsl88>nckx: there is sth I find weird. I did cat the .pub file <apteryx>efraim: OK! I can try bumping those on core-updates to see if it fixes my problem <nckx>lsl88: You mean inside ‘(public-key (ecc …’? <lsl88>and when I inspect /etc/guix/acl is the first one that appears <nckx>lsl88: Then it's already authorised. <lsl88>but being the first one in acl, I just ran guix archive --authorize, does that mean that it is the default one? <apteryx>efraim: we're already on 2.7.16 and 3.7.2 on core-updates <nckx><nckx> lsl88: Does your /etc/guix/acl contain 8D156F295D24B0D9A86FA5741A840FF2D24F60F7B6C4134814AD55625971B394 ? <nckx>So you must have added it? <nckx>The order means nothing. <lsl88>and I have another entry that is an rsa, guess I didn't have it before <nckx>The order in --substitute-urls="first second third …" is what matters. <nckx>lsl88: You'll notice that the entries in …/acl don't even have a server name. They are just keys. Guix has no idea which one belongs to which server. <nckx>And as you've learned in the case of berlin. and ci., multiple servers can serve substitutes signed with the same key. <lsl88>and how do I tell guix, use this one by default? I am reading the manual <nckx>Guix tries the substitute-urls in order, then validates whatever it downloaded against *any* key in …/acl. There's no such thing as a default key. <nckx>lsl88: And no such thing as a default server, either, really. Just an ordered list. You can call the first one the ‘default’ but that's not really correct. <lsl88>nckx: ok, so then just importing it, and now using the --substitute-urls just fot this pull, but later on not specifying it at all will make it work? <nckx>lsl88: New versions of Guix will use ci.guix.info by — eheh — default so all you have to do is ‘guix pull (done!); guix system reconfigure --substitute-urls="https://ci.guix.info …" /etc/….scm’. <nckx>Then reboot (so you get a nice new kernel) or just restart the daemon and it will use ci.guix.info. <nckx>lsl88: You can verify with ‘ps aux | grep [g]uix-daemon’ or similar. <nckx>It will print the currently used ‘default’ --subsitute-urls. <nckx>Or, well, it does here, I should say. <nckx>It doesn't on another box… <apteryx>nckx: I thought ci.guix.info is used by default on a Guix System. <nckx>apteryx: Not if that system predates it. <lsl88>nckx: it is already downloading, then when reconfiguring will check with the community if I am adding everything fine (the substitutes and the swap that does not load by default) :) <nckx>By ‘downloading’, do you mean substitutes? Huzzah! <nckx>(If you mean ‘the sources to build qt’: not huzzah.) <lsl88>and document, at least for myself, so that I remember the steps. Maybe a new video comes from this? :) <lsl88>nckx: I believe that sth is wrong, it starts saying compiling... <lsl88>nckx: when it started, it said the download and built, but now it is compiling... <nckx>lsl88: Hard to say. Compiling a few things isn't abnormal. <nckx>lsl88: Add ‘--dry-run | less’ to your ‘guix system reconfigure … …’ command. It will tell you what will be built and what will be downloaded without doing so. Might be enlightening? <nckx>(No need to cancel the running command to do that.) *nckx sighs. Somewhere long ago, some genius invented SSH and thought ‘this should time out in the default configuration if someone doesn't type for 6 seconds.’ <lsl88>nckx: no, it ends successfully (echo $?) but when I run guix --version it still shows 15 <nckx>lsl88: And in a new session/terminal? <nckx>IIRC some profile path changed between .15 and now. <nckx>Your shell might be remembering the old one. <lsl88>nckx: I rebooted and yeah! it doesn't show 16 but commit 598.... <nckx>Mine shows guix (GNU Guix) c0d43f62232507b441edf28c3f051921dfb04c2a <lsl88>but i through ie guix describe and worked <nckx>lsl88: 598c2bbc5312f5d25c52015714343d4da8e3e7b0 ? That's an hour or so old. Congrats :-) *lsl88 is going to have lunch. 14:30 here and be back soon <lsl88>nckx: thank you, but Ricardo and you helped a lot :) <lsl88>and I still have to reconfigure it for the swap stuff and you told me to add the subsrtitutes url there too? <nckx>lsl88: No, ci.guix.info is included in the new Guix. <nckx>Nothing to be done unless you set SUBSTITUTE-URLS explicitly in you OS configuration, which is unlikely. <nckx>Swap should be trivial to add. Enjoy your lunch. <pinoaffe>what should I change / what module do I need to include to define sway? <mikadoZero>I am using Guix System. I am trying to switch my shell from zsh to bash. I have removed zsh from my system configuration, remove the shell section form my users in the system configuration, run system reconfigure and restarted the computer. When I open a new terminal it is still zsh. How can I complete my switch back to bash? <bavier>mikadoZero: do you have a shell set in /etc/passwd? <mikadoZero>bavier: In /etc/passwd at the end of the : seperated lists for my users there is: /gnu/store/ia34lsanp33vaf32yhw2nnpnm2pc2g1y-zsh-5.6.2/bin/zsh <lsl88>nckx: you know, I faced this issue when I upgraded in my foreign distro too, but guix keeps using hydra, unless I add the --substitutes.... <nckx>lsl88: Have you reconfigured and restarted the daemon/rebooted? <nckx>(Reconfigured with the new Guix, of course.) <lsl88>nckx: uhm, only rebooted the system, but tried solving the locales issue <nckx>lsl88: So your daemon is still ’old’? Then you'll have to keep using --substitute-urls until the new daemon is running. <lsl88>nckx: i guix pulled with my user. <nckx>lsl88: pull doesn't update your system. <lsl88>and the weird thing is that even with my user, I am still asked to run guix gull (even that is the version I showed you I have installed) <nckx>‘guix pull’ gives you a new Guix command that ‘knows about’ all the new packages. You still need to install them, either with ‘guix package’ or ’guix system reconfigure’. The latter will update your daemon (which must then be restarted). <nckx>lsl88: Hm. I don't use guix pull so am conditioned to ignore that message, no idea what it means, sorry. <nckx>lsl88: Why don't you reconfigure now if you haven't done so after ‘guix pull’? Then you'll be running the newest Guix System, at least. And your daemon will use ci.guix.info automatically. <nckx>No time like the present. <nckx>(You'll need to use --substitute-urls= once more.) <bavier>mikadoZero: you need to adjust your user-account shell in the system config then <mikadoZero>bavier: I have this in my user-accout (shell #~(string-append #$bash "/run/current-system/profile/bin/bash")) but after a system reconfigure it still shows zsh as the shell in /etc/passwd. <rekado>lsl88: make sure to use ’guix’ from ~/.config/guix/current/bin <bavier>mikadoZero: bash is the default, so you could remove the "shell" field altogether <nckx>mikadoZero: Your're also appending /run/current-system/… to /gnu/store/…-bash-…. That is always bogus. <mikadoZero>bavier: I have removed the shell field, saved, system reconfigured. zsh is still the shell for my users in /etc/passwd. <nckx>It's either (string-append #$bash "/bin/bash"), or simply "/run/current-system/…/bin/bash". <nckx>(This will not fix your problem, but it would never have worked.) <bavier>mikadoZero: you may also need to reboot or restart the service that manages that file, not sure <nckx>The lazy Windows-esque approach to ‘which service would that be?’: reboot… *nckx apologises profusely for even suggesting that. <mikadoZero>bavier nckx: I will try rebooting the computer. I will be right back. <lsl88>rekado: yes, that is what guix points to <nckx>sneek: later tell mikadoZero: I just reconfigured my system with a new shell and /etc/passwd wasn't immediately, so it's not something wrong with yours. <nckx>sneek: later tell mikadoZero: s/immedately/& updated/ <apteryx>nckx: there might be a dependency on another service which cannot be restarted? <nckx>apteryx: It might not be a shepherd service at all. <mikadoZero>I have rebooted and shells that I open are still opening with zsh and not bash. <sneek>mikadoZero, you have 2 messages. <sneek>mikadoZero, nckx says: I just reconfigured my system with a new shell and /etc/passwd wasn't immediately, so it's not something wrong with yours. <sneek>mikadoZero, nckx says: s/immedately/& updated/ <mikadoZero>I am trying to get bash as the shell as I was having trouble with zsh and guix environment. <nckx>mikadoZero: Did you remove (shell …)? <mikadoZero>nckx: Yes (shell ...) is no longer in (user-account ...) <kmicu>Hi pinoaffe, Sway was added 11th this month so it’s possible you need to use guix pull to pull recent changes to make sway available. <nckx>mikadoZero: If you have no mention of (shell …) in your system configuration, run guix system reconfigure, reboot, and /bin/passwd still mentions zsh… well, that simply ‘can't happen’. 🤔 <nckx>mikadoZero: So ‘grep /etc/passwd zsh’ returns the nonempty string? <nckx>Something's rotten in your state of Denmark. <mikadoZero>nckx: Yes `grep zsh /etc/passwd` returns the lines for both user with the last item in there colon seperated lists being zsh's path. <lsl88>but it asks me to run guix pull and package -u and starts with derivations <charlag>Hey everyone! I have a stupid question again. I'm trying to package a crate but something's wrong with my setup. I cannot import any gnu or guix modules. I cannot find any info where to get them or how to check that they're installed. Thanks in advance. <nckx>lsl88: Ignore that recommendation for now. <lsl88>nckx: I cancelled it, will run it again <nckx>lsl88: ‘$your_guix_reconfigure_command --dry-run’ will print a nice list, grouped by what will be done. <nckx>A few local builds are unavoidable. For example, there will never be a substitute for your exact profile, or info/man page directory, or… *kmicu bets on KDE Framework 5. <lsl88>nckx: I don't get it very well, I think it is rebuilding everything in the store, is that possible? <lsl88>I only added the swap partition to the config.scm file <efraim>nvm, it seems it can't find kwayland <nckx>lsl88: What did --dry-run say? <nckx>Paste.debian it if you want. <lsl88>nckx: sorry, so I stop the reconfigure? <nckx>lsl88: --dry-run won't do anything… <nckx>Why not just leave it running for now? <lsl88>nckx: shall I run my full command with dry run? <nckx>mikadoZero: I really don't know what to say. If zsh is nowhere in your operating-system declaration, it shouldn't be in /etc/passwd, period. <pinoaffe>kmicu: thanks for the help, it still doesn't work though <nckx>lsl88: I generally mean ‘try now‘ when I suggest trying something ;-) <kmicu>pinoaffe: is the error message still the same about unbound variable or do we have now something different? <nckx>But you can let the reconfigure run in the background while you do. Unless it makes your system unusable. <kmicu>pinoaffe: could you share your ‘guix --version’? <lsl88>nckx: sorry, not feeling well, didn't get it <lsl88>I will copy the paste debian in a sec <charlag>I'm not sure if my message is visible or I screwed up with IRC again somehow... <kmicu>(No rush. Guix’s immutable but can be frustrating at times 🤭) <bavier>charlag: message seen. You have guix installed already? <nckx>charlag: It's not you. Just no-one who knows the answer. <charlag>Thanks. It's not to nag, just happened to me once. I have guix for sure but importing it from Guile doesn't work <kmicu>Hi charlag. Your message is visible. Could you provide more details though? <nckx>lsl88: No sorry needed, just wanted to avoid another day-delay :-) <bavier>charlag: make sure GUILE_LOAD_PATH and/or GUILE_LOAD_COMPILED_PATH are set properly <charlag>bavier: thanks. Is there any doc describing where it should point to? <kmicu>That looks good lsl88 Only trivial packages to build. <lsl88>nckx: so what I was getting was fine? <nckx>lsl88: Yes! ‘Building’ a .drv is a trivial operation. <bavier>charlag: usually something like "$prefix/share/guile/site/2.*" <bavier>charlag: that, or start guile up with guix's pre-inst-env script (if you installed from source) <mikadoZero>nckx bavier: Thanks for the help. I am still working on it. <nckx>lsl88: For example, you'll see that you need to ‘build’ …-qemu….drv, but that …-qemu (no .drv) will be downloaded. The latter is the actual package that would take ages to build locally. The .drv is just a text file. <lsl88>I have always thought it was for updating your system: ie adding a service <nckx>The daemon we're trying to upgrade is part of the ‘system’. <nckx>‘guix’ on the command line just talks to it. <nckx>lsl88: There's no real difference. Updated packages need to be built. New packages need to be built. <nckx>Sometimes a ‘system reconfigure’ takes seconds, if you just changed a small detail and there's no new software to install. <nckx>Other times it can take… longer 😒 <lsl88>ok, so I was right with my command, that not only will update the packages but also make the changes like the swap partitionn <lsl88>nckx: yes, the uptime was of more than 9 hours yesterday <nckx>‘guix system reconfigure’ == make it so. <nckx>Uptime means how long your computer has been on. I hope you've had uptimes higher than that… 🙃 <kmicu>So when we will rename it or add the alias ‘guix make it so’? <charlag>bavier: something works but not guix modules. E.g. I installed chickadee and I am able to import it, but not (guix packages) e.g. <pinoaffe>kmicu: I still have the same error message, my guix version is 0.16.0-1.c845323 <pinoaffe>this is my first time trying guix, I just grabbed the iso online, followed the instructions to the best of my abilities but ran into this issue <kmicu>Ah, thank you for that info. In such case run ‘guix pull’. You current Guix version is from 2018-12-05. <kmicu>(Guix pull will fetch all latest additions—including sway.) <charlag>I failed to install Guix by hand twice and then installed though the Fedora COPR. May be it is useful to someone. Works fine so far. <pinoaffe>kmicu: i did run guix pull, that did not help <kmicu>pinoaffe: let me check on a live system based on the same iso. *pinoaffe always mixes up front and back slash after writing too much latex *kmicu is building /gnu/store/…-guix-system.drv 89% <kmicu>(That last 10% takes 90% of the time xD) *nckx has been building Guix for the past 1.5h. <nckx>All these productive citizens. <civodul>what's up with people spending days building Guix? <civodul>looks like everybody is waiting for Guix to build today :-) <kmicu>That common building gives a sense of Community. Cool. Cool cool cool. <nckx>Guix: a primitive form of hazing. <nckx>*GNU's advanced form of hazing, apologies. <nckx>Are you also building Guix rockandska. <kmicu>pinoaffe: did you ‘guix pull’ as your user or as root (via su or sudo)? <rvgn>Blackbeard[m] Same here :( It takes long time but gets finished. <apteryx>Blackbeard[m]: maybe a 'not enough ram' issue? <pkill9>it would be good for `guix publish` to be able to report build failures, so if you try to build a package and the substitute server has already tried but failed to build it, then guix will just error out saying that the subsittute server reports that the build fails <rvgn>Blackbeard[m] It failed for me twice repratedly a while back. So I deleted package generations and did garbage collection. Then I pulled guix and re-tried update guix again. Then it worked. <apteryx>ah, then more likely a 'not enough disk space' issue ;-) <rvgn>apteryx You could be right <rvgn>Anyway, how to see whether guix package -i/-u is getting source package and building (or) getting pre-built binaries (substitutes) and installing?? <rvgn>Is there an argument that can passed to tell guix to use subsitutes? <rockandska>nckx: not really, only commit once and was just a little modification <nckx>rvgn: Guix uses substitutes by default. <apteryx>rvgn: you could try adding the GNU mirror at mirror.hydra.gnu.org <rekado>pkill9: this has been discussed on the mailing list a few weeks ago. <kmicu>rvgn: you can add --dry-run to see whether a package will be downloaded or built. <rockandska>not sure if it could be done easily, but, is there any way to install a specific old "native" guix package version ? a kind of search on previous guix commits without adding additional definitions ? <pinoaffe>kmicu: my process: boot guixsd iso, run dhclient -v, mount /dev/sdb5 /mnt/guix mount /dev/sdb4 /mnt/guix/boot, herd start cow-store /mnt, guix pull, gnu system init /mnt/etc/config.scm /mnt <pinoaffe>so I ran guix pull as the root user of the livecd <rockandska>'guix package -s bash@4.3.3' works only for the actual repository commit, doesn't it ? <kmicu>pinoaffe: could you execute ‘guix package -A sway’? What is returned? <kmicu>pinoaffe: thank you for additional info. I assumed you already installed the stock config. Now I understand you are during installation process and have root shell. <pinoaffe>kmicu: guix package -A sway prints nothing <kmicu>pinoaffe: and at the same time ‘guix --version’ prints? <kmicu>pinoaffe: Are there any error messages after ‘guix pull’? It lookes like nothing is updated. <pinoaffe>when I run guix pull again, it says building from channel .... .git 1c4d419, computing guix deriv, nothing to be done <rekado>are you running guix from ~/.config/guix/current/bin? <pinoaffe>i'm running guix from /root/.config/guix/current/bin/guix <kmicu>So it pulls 1c4d419 but guix is still on 0.16.0-1.c845323. It looks like LiveCD has different behavior than real system. <kmicu>(This is why I always try to install stock config and then adjust system to my liking.) <kmicu>(Cuz even when we fix that unbound sway error then Wayland can be broken anyway.) <kmicu>nckx: any ideas why ‘my process: boot guixsd iso, run dhclient -v, mount /dev/sdb5 /mnt/guix mount /dev/sdb4 /mnt/guix/boot, herd start cow-store /mnt, guix pull, gnu system init /mnt/etc/config.scm /mnt’ doesn’t update guix and make ‘sway’ visible? <civodul>commit c845323 is from Dec. 2018, whereas Sway was added a week ago <kmicu>civodul: we execute guix pull though. <civodul>pinoaffe: you might need to type "hash guix" in the shell, or to explicitly type ~/.config/guix/current/bin/guix instead of just "guix" <nckx>Well, for one, you're mounting /dev/sdb5 to /mnt/guix instead of /mnt as is customary, without adjusting the cow-store location accordingly. ***Guest27710 is now known as sturm
<nckx>I don't know if that would cause this failure but it does mean that cow-store is effectively useless since it's still COW'ing to RAM. <kmicu>pinoaffe: IIUC it looks like guix is not updated in shell after ‘guix pull’ cuz cow-store is not enabled on ‘/mnt/guix’. <nckx>Oh, and you're guix system init'ing /mnt too… <nckx>kmicu: That's not quite what I said… <nckx>kmicu: That wouldn't cause shell hashing. <kmicu>pinoaffe: could you try again but with ‘herd start cow-store /mnt/guix’ before ‘guix pull’? <rekado>better not “guix pull” at all before system init. <rekado>just build a minimal system first. <kmicu>Let’s see where is the problem. (Alas I don’t have a spare hardware to test it; I already have Guix on everything ;) <rekado>we have substitutes for released versions, but we cannot guarantee anything for the current version of Guix that you’d get with “guix pull” *nckx returns coffeeless. <nckx>kmicu, pinoaffe: Even better would be to just drop /mnt/guix and use /mnt everywhere. There's no reason to deviate from the manual in that way and it's the correct way to use /mnt anyway ;-) <nckx>Doing it this way is a one-way ticket to typotown and that only makes this more confusing. *kmicu always recommends using defaults first and then tweak the system. *nckx would suggest starting anew at this point, suck though it may. <rockandska>Is there any helper to search for old revisions of package (with the associated commit) to using it with "git package --with-commit=" please? <nckx>rockandska: N[ot that I no of, n]o. <rekado>rockandska: this would be neat, though. <rekado>Blackbeard[m]: what selinux patch? <rekado>there’s an SELinux policy for the daemon, but I know for a fact that it is incomplete. <rekado>if you’d like to help you could install the policy and run the daemon in permissive mode, and then report the audit log. <rekado>Blackbeard[m]: I have Guix on Fedora on one workstation. <rekado>(because the Guix system does not yet support LDAP account authentication; I’m working on it right now) <rockandska>does the git repository with the definitions is fully cloned in the store and could be used at a search "database" for the commit corresponding to a targeted version by doing "git grep" on it ? <rekado>you’d need to set it to permissive mode <rekado>or else you won’t get an audit log. <rekado>then run the daemon and do “guix build hello” <rekado>this should give you enough errors to report :) <pinoaffe>nckx: ah yes, I'm actually using /mnt rather than /mnt/guix, the typo was only present in the irc message <pinoaffe>kmicu: yeah I'm setting up a basic i3wm system now, let's hope that that works <nckx>pinoaffe: Oh well. Wild geese are tasty. <kmicu>pinoaffe: if you are using /mnt then could you try civodul’s suggestion and execute ‘hash guix’ and check ‘guix --version’? <rekado>Blackbeard[m]: please send it to guix-devel@gnu.org <nckx>pinoaffe: Nothing important, sorry. <kmicu>pinoaffe: and ‘~/.config/guix/current/bin/guix --version’ too. <kmicu>(Though not doing any guix pull during installation is still recommended.) *nckx is still building Guix. *kmicu ’s finished building Guix long time ago cuz retro core2duo laptop (with libre wifi) has Samsung Evo SSD. *nckx has a Samsung Evo SSD sitting right next to this laptop MOCKING ME with its nonbootiness. <nckx>Oh, knot@2.8.0 is out and nobody told me. Something to do. <civodul>nckx: you're making this distro the most up-to-date one there is out there <civodul>well, omitting things like wip-gnome3.30 :-) <nckx>roptat: Did you get my mail about ACME DNS validation? I'm happy to help set it up (my experience is limited to the native configuration file format). <rvgn>nckx apteryx kmicu Thanks! Btw, are substitutes and sources stored in same repo? <roptat>nckx my server is having some troubles, so maybe not <nckx>roptat: …troubles with that? <roptat>Don't know, it's broken in multiple ways, but it oas mostly working until today <nckx>rvgn: Yesss… although there's no ‘repo’ per se, it's just the /gnu/store of the build farm (and a cache). <roptat>I was too busy installing guix on its replacement though <nckx>And then there's the content-addressed mirror stuff that only handles sources. <nckx>roptat: OK, well, let me know if there's anything (else) I can do to help. <roptat>So I'm too sure how to set up that challenge so we don't have to manually change the dns everytime <nckx>kmicu: Sources are downloaded by the build farm and served as regular substitutes though! <nckx>kmicu: I think it's the ‘classic’ distro use of ‘repo’, i.e. a mirror network of packages on the 'net. Like the ‘Debian repos’. That's how I read it anyway. <nckx>roptat: If ‘not’: don't worry, DNS automation's as well-supported by certbot as webroot etc. You just have to configure the name server to accept updates, but all those in Guix can, IIRC. <pinoaffe>kmicu, civodul: doing guix pull, then specifiying path to .config..... seems to have worked <kmicu>pinoaffe: glad it worked. (So 'hash guix' wasn’t enough?) <nckx>Er… adding a ;; comment to /etc/guix/signing-key.pub makes Guix clients backtrace? <nckx>pinoaffe: I'm very curious what ‘command -v guix’ returns it that shell if you still have it open. <nckx>Although not bad considering our contributor base. <pinoaffe>nckx: the shell is currently busy initializing the system, but when it's done I'll tell ya <nckx>jonsger: The Nix folks do some mad automated importing and it seems to work, too. <nckx>pinoaffe: & hackage & probably npm or something to explain their place on that graph. <rockandska>"guix package --with-commit=bash=455859a50f88f625d13fc2f304111f02369b366b" doesn't seem to work <jonsger>we are behind Void Linux, Nix, Arch/Manjaro/Parabola, Slackware, Fedora Rawhide, Tumbleweed, Debian Testing/Unstable, Ubuntu 19.04... <nckx>rockandska: I haven't tried --commit= yet (it's newish), but does it specify the guix commit… or the bash commit (like a --with-source= for git)? <nckx>jonsger: TBH I don't think we can any meaningful conclusions from that graph besides inspo that we're all right and can do even better. <rockandska>the command i just gave return an error, but can't get if the format is correct or not, just followed the documentation <nckx>rockandska: I think it would build bash from bash commit xxx if bash were fetched from git. <jonsger>nckx: yeah, that was just reading out real, big repos out of the list. Minus stuff like CPAN, which haves of course 100%... <nckx>And my preconceptions about the AUR are somewhat tempered. <rockandska>nckx: ok got it, it needs to be a reference for the bash git itself and doesn't work if the origin don't use git-fetch <vagrantc>so, i've got the c201 working with The Guix System reasonably well ... but xf86-video-intel fails to compile (which i don't really need) and I'm not yet savvy enough to figure out how to install a minimalist desktop system <rockandska>so there is not an easy way to install a package from a commit coming from the guix repository itself <nckx>jonsger: I also don't push an update if it breaks even a single dependent package (and I hope that goes for everyone, minus the unavoidable mistake). Non-functional distributions provide less reassurance there so we might even have a slight handicap there (which is, of course, good for users). <vagrantc>how do you build just the xf86-video* that are needed for a particular machine? <nckx>rockandska: Not in that mix-&-match manner, I don't think. You'll have to guix pull that commit, then install bash, then be sure to guix pull back to the future. <nckx>Which indeed isn't great. <nckx>I wish I hadn't seen that :-( <rockandska>nckx: that's the part i'm missing in guix since all the data with the package history are somewhere :/ <rekado>then again, most of the CRAN/Bioconductor packages in Nix are not usable. <nckx>rockandska: The main goal of Guix (if I may speak for it) is to provide a single well-working (and ideally tested) graph of packages at any point in time. Not a generic mechanism to build bash@3.7. Not saying you're wrong for wanting to (it's cool and we can probably pull it of) but it's not the main goal. <jonsger>Ignoring all this list, I think we are overall quite especially if we look on our resource (compared to e.g. what SUSE has...) and the hard things we do (bootstrapping) <rockandska>nckx: indeed, but will be really useful to repro build env easily. Here, i was looking to an easy way to test some script on difference bash version, and use guxi as an "Bash version manager" since no one exist. <rockandska>Having the version of bash i want to test my script against easy as 'guix package --with-commit=bash=xxxxxx' could have help a lot <nckx>jonsger: Yeah. The more I click around on that confounded site the better I feel about our freshness. <nckx>We're actually… quite fresh. <rekado>rockandska: we recently gained a feature like that <rekado>rockandska: you can do “guix build glibc --with-git-url=glibc=git://sourceware.org/git/glibc.git --with-branch=glibc=release/2.25/master”, for example <rockandska>that's nice, but it doesn't provide the same functions as 'inferiors", it's more like something like that i would imagine on command-line <nckx>Ah, --with-git-url= was what I missed. <rekado>yes, it’s not for viewing the world through an older Guix lens.