IRC channel logs
back to list of logs
<attila_lendvai>guix home reconfigure is stuck after "Finished updating symlinks.", but in a way that i can't even C-c it. <fnurkla>Is there a way to disable paging when calling guix subcommands such as "guix search"? <nckhexen>fnurkla: guix respects PAGER and GUIX_PAGER. <nckhexen>If it's set but empty, that disables paging. *nckhexen → 😴💤, good night Guix. <attila_lendvai>this is a sign that the project grew beyond what can be managed by the current number of maintainers and/or infrastructure <lechner>attila_lendvai: it might be worse with more bugs! <lechner>but i have also been sitting on some patches <attila_lendvai>maybe this package should be in a different channel, managed by a different set of committers, e.g. the maintainers of perl stuff <attila_lendvai>lechner, i don't need it, i don't even remember what i needed it for... but it should either be rejected or applied by now. <lechner>attila_lendvai: you need it to update ddclient <attila_lendvai>lechner, heh, right! :) i abandoned my pursuit of the ddclient update. <lechner>attila_lendvai: maybe, echo "i abandoned my pursuit of the ddclient update." | mail email@example.com ? <Kabouik>Anyone successfully using org-latex and the PDF export on Guix? <Kabouik>My tex files are created but pdflatex fails to create the pdf. It's missing ecrm1095 apparently, which seems to be in the texlive module, not packaged for Guix. But I find it weird that I'd be the first to hit this issue, I expect many Guix users to use org-latex <gabber>Kabouik: are you using the giant all-included latex or your own mix? <gabber>unmatched-paren: (require 'foo) doesn't work in your version either? <lechner>Hi, gocryptfs asks maintainers to build the executable via a Bash script, so that certain version numbers appear in the help output. (the build date can be overridden for reproducibility.) the binary also builds with the standard golang build system in Guix, but the help output is a bit uglier. what is the best way to handle this for Guix, please? <lechner>in guix the exact provenance is never in doubt. should i just patch out the line of output? <apteryx>for writing good wrappers, we'd need to have the list of native vs regular inputs, but unfortunately they all get joined as 'inputs' unless cross-compiling. Perhaps we should revise that? <abhicherath[m]>I guess the issue is that `cranelift-codegen` is memoized with (mproc "cranelift-codegen" #:version "0.68.0" #:repo #f …) , where repo is false <ennoausberlin>Good morning. I have a basic emacs-meow config now (shamelessly stolen from some gist). I wonder if I can define sequences (chords) with meow, e.g. <Space f f> to find-file.
***Dynom_ is now known as Guest6513
<jonsger>meh, Thunderbird in older versions could hide those join/leave messages. Sadly that feature is gone... <sneek>Welcome back jonsger, you have 1 message! <sneek>jonsger, nckhexen says: (file-name (string-append "thunderbird-" version "-checkout")) ; <-- here <hapst3r[m]>I am trying to install the guix package manager on top of my distro, but when trying to download the openPGP signature for the binary, it hangs eternally. Manually trying to fetch via `gpg --keyserver pool.sks-keyservers.net --recv-keys 27D586A4F8900854329FF09F1260E46482E63562` returns `unknown server error` <nckhexen>The SKS keyservers have been dead for (what feels like) years now. Which installation instructions are you following, hapst3r[m]? <nckhexen>The freeze seems to be on Savannah's side. savannah.gnu.org's not loading for me… <nckhexen>There was maintenance yesterday. Perhaps this is related.
***TopExpert is now known as phonymontana
<jonsger>nckhexen: a thanks, you would be a good addition to Guile's error messaging :) <hapst3r[m]>nckhexen: Thanks for the link. However, since it contains no user id (as per gpg message), it is skipped. <hapst3r[m]>I receive the same error message regardless whether I import manually or via `gpg --keyserver ...` <jonsger>is firstname.lastname@example.org mail delivery a bit slow ATM? Didn't get Mathieu's mail from yesterday neither Ludos reply from an hour ago :( <nckhexen>hapst3r[m]: Ah, right, forgot about that server's feature. <nckhexen>jonsger: Yes, has been particularly so for a few days no. They are aware of the problem. <nckhexen>They've tried a few things but can't see to find the root problem. <nckhexen>hapst3r[m]: ‘Eternally’ aside, I noticed that savannah will eventually serve the request, it just took over 6 minutes here. <hapst3r[m]>nckhexen: That's awesome to know, thanks! But which one have you used? the one I initially tried, the one you supplied, or the one by tobias.gr? <nckhexen>I left wget running, and it eventually returned after 6 minutes. wget is also what the installation script uses. There are 2 keys to fetch, so maybe it would have worked for you after ~12 minutes, or maybe I just got ‘lucky’. <nckhexen>The default wget time-out is 15 minutes. I've got a patched staged to reduce that just a bit, because to users this just looks like ‘freezing’, as you observed :) <hapst3r[m]>nckhexen: Oof, that's very long. Well then, I just fetched from your page (thanks for that "service", btw), hoping the rest of the installer works fine :) <nckhexen>The rest doesn't rely on Savannah, at least, so good luck! <hapst3r[m]>well, but `guix pull` (my first command after installation) works via savannah, no :D? <nckhexen>I've reported the issue to the Savannah folks, not much more we can do on that side. <hapst3r[m]>I appreciate how you subliminally added "attractive" to the mix ;) <nckhexen>It was odd to refer to myself as ‘trustworthy’, so I might as well make the most of it. <ennoausberlin>Maybe the slowdown is related to an attack on the fiber cable in Marseille yesterday, too. German news reported that <nckhexen>gnu.org is entirely US-based AFAIK (they lease from HE, or they have their own ASN, or both, I'm not sure), but still wow. <unmatched-paren>Wow indeed. I only knew about the *other* cable severage in Shetland; seems like the UK Guardian didn't report the severage in France...? <nckhexen>Oh, nice, great. My reasoning above was ‘doesn't WE connect to the US through the UK, not France?’ so there goes that reasoning. <unmatched-paren>nckhexen: to be clear, the outage is confined to Shetland, there's no problem elsewhere <nckhexen>Saying ‘oh good’ now would be exceedingly mean. <unmatched-paren>ennoausberlin: so, are they saying the incident in France was probably sabotage? <jonsger>nckhexen: what does HE mean in this contect? <ennoausberlin>unmatched-paren: Maybe the news exaggerate a bit, but they call it an "Anschlag", what I translate to attack <nckhexen>And they speak of a Täter (although I didn't read the rest, yet). <ennoausberlin>unmatched-paren: As I live now in Strasbourg, I at least can tell you, that I am affected somewhat, but for IRC my connection is fast enough :) <nckhexen>I was thinking of downgrade attacks, but I'm not paranoid enough to care. <civodul>nckhexen: that would be detected as well <mirai>doubt it looks acceptable to include in guix in this state <nckhexen>(NB, I don't mean downgrading from 1.3.0.) <mirai>unmatched-paren: missed it, I think I meant to put it under /var/lib <nckhexen>I guess ‘partial upgrade attack’ would be a better term here. <unmatched-paren>mirai: why use a hash table? an alist would be simpler and fast enough <unmatched-paren>you should probably construct the config file as a file-like and then put it in /var/lib/mympd/config, instead of writing it with an activation service <mirai>unmatched-paren: it looked easier to iterate over the alist via a hash-table <unmatched-paren>mirai: you could do ``(for-each (match-lambda ((field . path) ...)))'' <unmatched-paren>but that doesn't matter anyway imo, as you should probably implement this with mixed-text-file anyway <mirai>unmatched-paren: by constructing instead of using an activation service do you mean using the shepherd service to do it? <mirai>What's the difference between mixed-text-file and text-file*? When should I pick one over the other? <unmatched-paren>mirai: i'm not sure exactly, but the manual says "[mixed-text-file] is the declarative counterpart of text-file*." <mirai>I'm looking at special-files-service-type, so its a service that creates symlinks, is that right? <unmatched-paren>and mixed-text-file et al create said files and return objects that become their paths inside a gexp <unmatched-paren>all file-likes do that; such as ``origin'', ``local-file'', ``package'', ``plain-file'', ``computed-file'', etc <mirai>So I am to use mixed-files, etc. within the activation service to create and place the config files in the store <mirai>a "gexp" within special-service? as in `(("/var/lib/..." ,(mixed-text-file ....)) ? <unmatched-paren>mirai: activation-service-type is more for *mutable* state that needs to be set up before running the service <unmatched-paren>but if you have something immutable like a config, it's best to use special-files <mirai>(define my-service-type (service-type ... (extensions (list (service-extension special-files-service-type ALIST-OF-CONFIG-FILES <unmatched-paren>and of course the procedure would be extracted into a separate define... <mirai>ah right, service-extension is a one argument procedure <mirai>that is fed with either the default-value or the configuration record <unmatched-paren>well, the procedure that service-extension accepts is fed with that, yes <mirai>**the procedure that is passed for service-extension <mirai>One more thing, how can I patch and test existing services already in guix? <mirai>(/part of guix upstream that is) <unmatched-paren>sudo guix shell -D guix -- ./pre-inst-env guix system reconfigure CONFIG-FILE <unmatched-paren>if you want to use nginx modules, i believe you just extend nginx-service-type... <unmatched-paren>if you want to actually add a new feature to the service itself, use that command <unmatched-paren>if you try to use sudo ./pre-inst-env or ./pre-inst-env sudo, you'll get "module not found" errors <gabber>is it possible `guix home container ...` mode of operation is different from that in my home environment? i'm having issues figuring out whether my home-emacs-service can't (require) emerges from the guix or the emacs side of the issue <gabber>should i just reconfigure home to test my luck? <gabber>unmatched-paren: yes, it seems. have you tested yours only in a container? <unmatched-paren>hmm, could you try removing the init file and adding --eval "(print load-path #'external-debugging-output)'' <gabber>i checked the load-path variable (which apparently automatically inherits from EMACSLOADPATH) -- it seems correct. maybe emacs has issues resolving chained-together symlinks? <unmatched-paren>you know how when you build a profile, it says something like "Building Emacs subdirectories..."? <lxsameer>hey folks, does guix supports binary packages now? <lxsameer>unmatched-paren: a package that the source is not available, e.g linux firmwares <unmatched-paren>do you mean packaging binaries? the guix channel won't accept those, but you can write them yourself, though you'll have to do some patching of the binaries <gabber>unmatched-paren: i'm somewhat confused. in my solution the packages are added to the home-profile. subdirs.el is generated and seems to contain everything necessary. so i doubt this is the reason why (require) won't work. i've also tried (require) with the optional path -- but i still get a "No such file or directory" <nckhexen>lxsameer: Guix has always supported packaging arbitrary files, but it will never ‘support’ non-free software. <unmatched-paren>what's going on here? /gnu/store/9g8cvairk2x055nqsj0a6vv24pi0ljv2-shepherd-emacs.scm:1:1860: definition in expression context, where definitions are not allowed, in form (define (directory? directory) (and (file-exists? directory) (eq? (stat:type (stat directory)) (quote directory)))) <morganw>lxsameer: you can get some compatibility for existing binaries with the --emulate-fhs option. I've not used it but I think you can try in a shell with `guix shell --container --emulate-fhs $program`. <nckhexen>I've used it and it's just as wonderfully as you describe, in the common case. <zimoun>rekado_: since I know you use some “slow” computer and an Emacs user, have you switched to Emacs 28 with native-compilation? If yes, are you compiling AOT all your Emacs pacckages? <Kabouik>gabber last night I was using my own mix (or, actually the example manifest provided in the Guix manual), but I also tried the whole 4GB package about 10 days ago and got the same issue. At the time I thought it was just me mis-using it so I didn't ask any questions, but since then I read somewhere in a guide (which I can't find anymore) that converting tex to pdf in Guix was not as straightforward as in other distros (and hence was not detailed in <Kabouik>that guide), so maybe it's not just me <gabber>unmatched-paren: was that directed at me (the shepherd-emacs in the store thing)? <gabber>Kabouik: i have had issues converting tex to pdf with a minimal LaTeX installation but i'm fine with the full LaTeX suite. but i think i've had a similar issue when trying to go from org to pdf via tex <gabber>unmatched-paren: have you tried the (require) with the optional path too? <ae_chep>`guix shell` calculates some env vars, and then drops into a shell in which those are present. I only want to be given what the config is, I don't want to be dropped into a new bash prompt. Can this be done? (reason: I normally use a non-posix shell and want to add a native support for it) <nckhexen>(To be added after the ‘guix shell’ invocation, sans ‘?’.) <rekado_>zimoun: I use Emacs 28 and I noticed a lot of compliation messages when I first used certain features (like entering some mode). <ae_chep>But `guix shell pkg1 pkg2 --export-manifest > manifest.scm` still needs to be followed by a `guix shell -m manifest.scm`. So it is not there yet <efraim>does anyone have an easy way to compare the staging vs master branches for build failures? <unmatched-paren>ae_chep: why is it not there yet? you said "I only want to be given what the config is"; the manifest is the config <ae_chep>I need to know how the `env` is changed too <unmatched-paren>i believe the env is simply calculated from the search-paths of the packages in the manifest <zimoun>rekado_: thanks. I am asking because I have not been annoyed on my desktop machine but then on my poor laptop, the first time experience is not great. Not because of Guix but because upstream design, IIUC. Anyway, thanks. :-) <nckhexen><still needs to be followed by> Eh? IDGI. <ae_chep>yes, so would a PR that allows `guix shell --export-env` be welcome? <unmatched-paren>ae_chep: now, if you want the /etc/profile, you can do ``guix shell PKG PKG ... -- sh -c 'cat $GUIX_ENVIRONMENT/etc/profile''' <nckhexen>ae_chep: How does it differ from ‘-- env’? <ae_chep>nckhexen: I have not thought of that <nckhexen>Then you missed my message above and I am now unconfused :) <ae_chep>I saw the thing about "<" needing to be followed by a ">" <nckhexen>ae_chep: I meant ‘-- env? (To be added after the ‘guix shell’ invocation, sans ‘?’.)’ earlier. <ae_chep>I wondered what "IDGI" may stand for <nckhexen><foo said blah> is just my (but a common) way of quoting folks on IRC. <ae_chep>the `-- true?` `-- env?` followed by a `sans ?`. I noticed those but couldn't compose them together in my mind :/ <nckhexen>Well, we all got there in the end, and the true friend was the total confusion along the way. *gabber is happy to not having missed it <ae_chep>I quote people using double-quotes but can only go 2 levels deep. I similarly struggle when adding smileys inside paranthesis <ae_chep>with that out of the way, I'll use `guix shell -- env` to put something together <reyman>i have some problem to compile mutter-42.4, i'm alone in this case ? <reyman>In log i have : 78/107 mutter:core+mutter/wayland / wayland-unit FAIL 2.50s exit status 1 <apteryx>so... I've experimented with baking Class-Path in Java .jar, and it works, but has the following drawback: we don't fully control the CLASSPATH ourselves. Java starts the search for its classes using the environment CLASSPATH, and visits each Jar and expands their own Class-Path attribute in a deep first search way, as far as I can see. <apteryx>this means we can no longer easily "abuse" mixing java component versions that could sometimes work if the libraries mixed were backward compatible (as not controlling the CLASSPATH anymore an older version may be loaded first) <apteryx>it also currently bakes too many references, as there's no way to differentiate between inputs and native inputs when compiling natively, on the build side <apteryx>it breaks a few packages, either because of incompatible libraries loaded first, or by introducing cycles due to capturing the native inputs <nckhexen>ae_chep: I'm in the ‘:-) closes a parenthesis camp’ but I still struggle with ‘:-( )’. Even with the space it looks like a monkey. Please advise. <nckhexen>reyman: Can you share the full output (or as much as paste.debian.net allows, starting from the bottom)? That message alone isn't enough info. *nckhexen is building it though. <ae_chep>nckhexen: I resort to square brackets <ae_chep>(or I resort to ^.^ style of expressions, which make me seem more enthusiastic than I try to be) <gnucode>unmatched-paren: I am officially using seatd. Thanks! <gnucode>unmatched-paren maybe...I don't actually know. I'll send you my config <pkill9>so what exactly does seatd do? and what does it replace? <unmatched-paren>which itself seems to be a grotesque hackjob carefully extracted from systemd so that we could have logind on Guix <Kabouik>sneek: later tell gabber: Yes, I have only tried org-latex, i.e., going from org to latex to pdf. I am ashamed to admit I never really used LaTeX before, despite being in an academic field and knowing about it since my young student years, because unfortunately in my field almost no one uses LaTeX (and therefore I cannot collaborate with colleagues if I use it). Moving to emacs recently however led me to play with org-mode, finally, and it was an <Kabouik>opportunity to play with org-latex and exporting my first org documents to PDF. <unmatched-paren>pkill9: for it to work, it and its dependencies would all need to change from using elogind directly to using libseat, which is provided by seatd and abstracts over the differences between the two. <Kabouik>sneek: later tell gabber: Maybe org-latex requires some fonts or whatnot that is not packaged in Guix yet, even the full suite, and maybe just doing Tex to PDF without the org-latex requirements would work. <pkill9>libseat abstracts over the differences between seatd and elogind? <gnucode>unmatched-paren: I couldn't get the sway greetd thing to work. as far as I can tell, I am still logging into a console... <gnucode>I didn't copy the supplementary-groups field though.. <gnucode>I just added "seat" to my user's group <gnucode>well, (greetd-wlgreet-sway-session (sway sway-latest) that seemed to try to start sway, but fails. <gnucode>i would only see a black screen with a cursor. <gnucode>but just got rid of the greetd-wlgreet-sway-session, and it seems to work. <gnucode>I am also running a wayland only sway session. <gnucode>I was surprized to discover that lxterminal runs fine in sway <AwesomeAdam54321>Is there a way to include the sources of all the packages in a Guix System image, in the system image itself? <unmatched-paren>Next problem: the environment doesn't seem to be inherited by the emacs process. <gabber>yay! what did you do? which part of the env? <sneek>Welcome back gabber, you have 2 messages! <sneek>gabber, Kabouik says: Yes, I have only tried org-latex, i.e., going from org to latex to pdf. I am ashamed to admit I never really used LaTeX before, despite being in an academic field and knowing about it since my young student years, because unfortunately in my field almost no one uses LaTeX (and therefore I cannot collaborate with colleagues if I use it). Moving to emacs recently however led me to play with org-mode, finally, and it was an <sneek>gabber, Kabouik says: Maybe org-latex requires some fonts or whatnot that is not packaged in Guix yet, even the full suite, and maybe just doing Tex to PDF without the org-latex requirements would work. <unmatched-paren>gabber: The problem I have here is that my init.el moves the backups, auto-saves, and lock-files into $XDG_CACHE_HOME <nckhexen>unmatched-paren: Weird! So it's not unset. I'd check the environment of each process in the tree to see where it gets reset. Inheriting variables is not optional or opt-in. <podiki[m]>hooray, matrix bridge seems to be properly bridge-y again <Lumine>"Did you see that ludicrous display last night" <nckhexen>unmatched-paren: I think the most user-friendly way is (sudo) htop, with tree-view (‘t’) enabled, then hit ‘e’. <unmatched-paren>nckhexen: Ah, wait, so you're saying that the variable is set, but empty? <nckhexen>I think so? getenv returns #f if unset, no? *nckhexen really AFK, sorry. <gabber>unmatched-paren: "Value is nil if VARIABLE is undefined" says C-h f getenv <reyman>@nckhexen finally that works after retry... strange ... <csepp>could someone restart the upower build or something? i see it got canceled which made tracker_miners not build which made a bunch of end-user applications not build. <nckhexen>csepp: I don't see any cancelled builds but I restarted a ‘failed’ one. It's always helpful to include an exact link. <csepp>nckhexen: thanks! will do next time. cuirass is a bit confusing for me, haven't used it enough yet. <csepp>unrelated but does anyone have a *working* config for using Fish on Guix? i tried it but can't for the life of me figure out how to source the sh scripts of profiles. <csepp>didn't find any tips in the info pages or the mailing list. <singpolyma>unmatched-paren: sounds ominous. Guix breaks chsh? <csepp>i didn't i did a proper reconfigure. it's set up the same way i set up zsh on this machine i'm typing from. <singpolyma>unmatched-paren: chsh came first so I think it can't be the fault of chsh <csepp>i even created a fresh account instead of modifying my existing one. <abhicherath[m]>How long does it take roughly before a bug report to email@example.com gets a number? <unmatched-paren>csepp: no, you just shouldn't use an alternative shell with your user unless it's POSIX <civodul>abhicherath[m]: hi! if it's your first message, it can take several hours; after that it's fast <singpolyma>unmatched-paren: do you know what guix bug causes this issue? <csepp>unmatched-paren: ah. well that sucks. i'd count that as a bug tbh, bash is rather limited and Guix doesn't even ship a nice config for zsh. <csepp>i had to install the GRML one manually. <csepp>i guess that could work. i have to manually exec to Fish on postmarketOS as well. <unmatched-paren>singpolyma: the problem is the scripts guix runs on startup won't work with a non-bourney shell <nckhexen>abhicherath[m]: In normal circumstances, not long, but the mailing lists have been experiencing longer than usual delays. <nckhexen>There is no first-time delay for the bug or patch trackers. <csepp>welp i'm writing a bug report anyway because forcing bash is not great. at the very least zsh+grml should be supported and documented. <csepp>not literally, it's just that when i had to configure zsh i had to do a lot of trial and error because it was completely undocumented. (as opposed to, say, Arch.) <csepp>unmatched-paren: they do work, but setting it up as a login shell was not trivial. <csepp>yup, this is a good reminder that i should. <nckhexen>It will have a much better effect than a ‘bug’ report. :) <singpolyma>unmatched-paren: why are the scripts run with the user's default shell instead of being shebangend into the store? <csepp>it does fix the zsh case, but clearing up the status of all the shells in gnu/packages/shells.scm should be a documented TODO item. <csepp>singpolyma: they are sourced, not run. they have to set environment vars. <csepp>virtualenv solves this by having a bunch of activations scripts for all kinds of shells from bash to windows batch files. <singpolyma>Couldn't it set the vars before executing the shell? <nckhexen>That's how I recommend people fix this, by making some kind of bash ‘boot loader’ stage before executing an optional more exotic shell. <gnucode>how many people use guix system on power9 ? I'm guessing less than 5 people. <gnucode>actually is guix system ported to power9 ? <nckhexen>I vaguely remember somebody offering access to their own machine which might have been a Raptor? <gnucode>I just saw the guix devel email about potentially disable power9 support.... <gnucode>I'm tempted to donate $1000 towards buying a physical raptor machine. <nckhexen>There should be another POWER9 node in the workers list (‘jade’), I don't know why it's not. <nckhexen>…so I hadn't seen the suggestion to remove powerpc64le support for the release. <podiki[m]>should we have maybe a "guix survey"? to get a feel of what people are running and use? <podiki[m]>though maybe server logs would tell us how much various substitutes are used for different architectures <podiki[m]>there are often some guesses around "well this package was broken forever and no one said anything, so maybe no one is using it on that architecture" <podiki[m]>I think it would also be interesting in of itself to know what the guix user base is like <gnucode>podiki[m]: I have noticed that gnome-clocks is no longer working for me on sway... <nckhexen>I don't think gnucode's estimate was far off. Chickens, also eggs. <cbaines>gnucode, I have a Power9 machine that's used for Guix things (a Talos 2 system) <cbaines>so I don't think there more of a need for hardware for that specific architecture, but the hardware available needs putting to better use *nckhexen reboots guixp9 to get a kernel update in. <gnucode>cbaines: do you have guix system on it? How is the experience on it ? <civodul>nckhexen: guixp9's back! thanks! :-) <cbaines>gnucode, no guix system yet, the Guix grub package doesn't build for powerpc64le-linux <cbaines>gnucode, so yeah, it's running Debian, but that single machine is easily able to keep up with building everything for powerpc64le, and I don't even leave it on all the time <cbaines>practically the biggest issue is that it gets stuck having built 1000's of things, and then is slow to upload them <nckhexen>Yeah, lack of hardware isn't the issue here, it's (if I summarise the mail fairly) the current requirement that the single machine has 100% uptime or Cuirass freaks out. <zeta_reticuli>Hi Guix. When I install GuixSD manually after "guix system init /mnt/etc/config.scm /mnt" some packages reports errors like this "guix system: error: corrupt input while restoring archive from #<closed: file 7f5781b11bd0>". Does anyone know how to fix it? <nckhexen>Extra hardware would be *nice* to deal with backlogs (such as we have now, but it shouldn't be bad) or extra development branches, but I really don't think it's the blocker some people might (justifiably) think it is from following along here. <AwesomeAdam54321>zeta_reticuli: I think this might be caused by an network connection interruption <gnucode>cbaines that's a bit of a bummer. It would be cool to get Guix system on power... <cbaines>I think efraim mentioned some approach to me about how Guix builds grub, but I forget what it is <nckhexen>civodul: The reboot was just opportunistic, BTW, the real issue (in 2 senses of the word) was that the SSH tunnel needs to be run in a screen session, or at least that's what I did. <nckhexen>So now I have 2 screen windows running 2 vital services :) <cbaines>nckhexen, is the VM on the berlin wireguard VPN thing? <zeta_reticuli>AwesomeAdam54321: ok, thank you. I'll try enter "guix system" again <nckhexen>cbaines: It *should* be, but it wasn't reachable. I'm looking at that now. <nckhexen>cbaines: I don't understand why guixp9 would be particularly problematic, to the point that it justifies disabling the guix jobset. Do you have an idea? <nckhexen>Do the (also off-site) aarch64 machines drop out just as frequently, and one just always happens to be reachable in practice? <nckhexen>Throwing hardware at *that* seems suboptimal. <nckhexen>cannot build missing derivation ‘/gnu/store/daii3asls4h4z7cnq8abk4s29cbnzx44-emacs-auctex-13.1.5.drv’ 🤔 <cbaines>nckhexen, I think it's that just trying to compute the derivations to build for the guix jobset requires building derivations for all the supported systems <cbaines>so when those builds can't happen, the Cuirass evaluation fails <nckhexen>I've copied that .drv from berlin to guixp9 and restarted the emacs-auctex build, just to see what happens. <unmatched-paren>next issue is that dependencies of the emacs packages aren't being found <unmatched-paren>so i'll need to add each package's transitive propagated inputs to the list of packages <lechner>nckhexen: You probably don't use the gnu.org equipment but, if you ever do, Hurricane Electric is five miles from my house! <nckhexen>Oh nice. I've used their (free) services in the past. And they sent me a T-shirt. <unmatched-paren>Hitting Super+E under home-emacs-service-type finally pops up an emacsclient window with my entire configuration loaded! :) <unmatched-paren>Before I do that, should I try adding AOT compilation of early-init.el and init.el? <nckhexen>lechner: If they're not as likable as their image on the Internet suggests, I don't want to know. <nckhexen>unmatched-paren: It's up to you, but that's just as welcome as a separate improvement. Whatever gives you the most energy. <gnou_lib`>Hello, I recently installed the Guix system with i3 on my PC, what happened is that I had a problem when I wanted to change the i3status configuration, I could not find the configuration file anywhere, could someone help me? <lechner>nckhexen: they are great in my community! i obtained three radio licenses on their premises, and they host many good free software companies and organizations, such as gnu.org and rsync.net, although they do charge <lechner>plus, their tunnelbroker service was invaluable for me for years <unmatched-paren>Very interesting: If I prepend the EMACSLOADPATH and EMACSNATIVELOADPATH to the default environment, it doesn't workp <nckhexen>sneek later tell gnou_lib`: At first glance, I don't see any Guix code to generate an i3status/config. Is it possible you've just been using i3status's built-in defaults? If so, creating ~/.config/i3status/config would suffice. <unmatched-paren>But if I do (append (list ...) (default-environment-variables)), it works. <nckhexen>lechner: Oh good, that's a relief (and of course they'd charge). And same, they were my only gateway to the IPv6 Internet for years. I filled in their T-shirt form on a lark, expecting never to hear from them again (or a polite ‘uhm, dude, you live in Europe’ e-mail), but nope. Still got mine. <nckhexen>Apart from unlocking some advanced settings in the tunnelbroker IIRC. <jackhill>nckhexen, rekado_: looking through my old system generations, my certificate problem seems to have occured between 582b1f626f351d0c519c973ba3c49d1c270200bf and ec7ba6ae5308faba68181f3fffc7d115d37cd1d4 *unmatched-paren opens guix/build/emacs-build-system.scm to see how this whole AOT thing works <nckhexen>jackhill: Could you share the problematic URL again? <nckhexen>If someone catches gnou_lib when they reconnect, could you quickly send them a logs.guix.gnu.org link? They might not be aware. <jackhill>yw. I mean, it's my site, I could have done something wrong there too, but I don't see it, and ssl labs is happy with it. <nckhexen>jackhill: You mean a particular commit within that range, right, not any commit within that range? (That would be too easy…) <lechner>jackhill: it says " This page intentionally left blank." <unmatched-paren>jackhill: if it means anything (i didn't see your original problem) the site loads for me <nckhexen>jackhill: IIRC several Guix System users connected just fine at the same time, including me, so it's not your server (unless you're running behind some CDN joint). <nckhexen>OK, make that *all* Guix System users, probably. <jackhill>nckhexen: I mean it works with 58… and doesn't with ec… I haven't biscected yet. <jackhill>lechner: yeah, I'm have cert validatation problems <jackhill>clearly I have offended the computer gods *nckhexen builds the 582b1f6 VM of fail. <lechner>jackhill: with wget (from scrollback)? <jackhill>lechner: yes. I first noticed with epiphany, but have been using wget for testing <lechner>nckhexen: ERROR: The certificate of ‘jackhill.us’ doesn't have a known issuer. <jackhill>lechner: yep, that's what I see too. Thanks for checking <nckhexen>openssl s_client -port 443 -host jackhill.us <gnou_liber>nckhexen: Yes, I am using i3status's built-in defaults. <jackhill>The cert I'm using was generated with acme.sh with let's encrypt and provided to Debian apache with the cert, key, and fullchain concatinated together. <gnou_liber>I disconnected from the channel because I was having problems connecting to the channel on ERC (Emacs). <nckhexen>gnou_liber: OK, then you'll have to create & populate the configuration file yourself. <pkill9>apparently the internet contributes to the greenhouse effect as much as the aviation industry <pkill9>well, current infrastructure used by and/or for the internet <nckhexen>jackhill: I'm not familiar with Apache. Is that the recommended method? I've only ever done that for ZNC, which gives you an indication of how ‘hmm’ it is IMO. <gnou_liber>nckhexen: I have two other questions: 1. How do I uninstall the Network Manager applet? 2. How do I change the sound server from pulseaudio to pipewire? All this in the Guix system. <jackhill>lol. I'm not sure about recommended, but it should work fine. <lechner>certbot is the reference, i think, but the certificate is fine <nckhexen>pkill9 might be able to help with Pipewire. Be aware that it is not fully supported in Guix yet; there is absolutely no (use-pipewire? #t) setting that will magically make it work. <pkill9>gnou_liber: i had pipewire working, n3ed to set it up again <jackhill>I use pipewire manually. What I do is start pipewire, start wireplubmer, and then start pipewire-pulse. If your desktop environment starts pulseaudio automatically by connecting to its socket, then you can race killing pulseaudio and starting pipewire-pulse <apteryx>cwebber: that job posting reads excellent, good luck! <cwebber>and yeah we use guix for all our stuff at the spritely institute too <lechner>jackhill: on an affected system, curl reports curl: (60) server certificate verification failed. CAfile: /etc/ssl/certs/ca-certificates.crt CRLfile: none <nckhexen>gnou_liber: For the applet, you can't ‘uninstall’ it, it's provided as part of %desktop-services. I don't know how to remove it using modify-services. <unmatched-paren>Hmm. Seems as if emacs doesn't provide a way to express (byte-compile-file INPUT OUTPUT), only (byte-compile-file INPUT). <gnou_liber>nckhexen: I understand, thank you for your help. <nckhexen>gnou_liber: You could ask on firstname.lastname@example.org. <pkill9>gnou_liber: basically just by stopping pulseaudio,then running pipewire,then pipewire-pulse and wireplumber <pkill9>i did have pipewire running in a home service <pkill9>try doing pkill pulseaudio && pipewire <pkill9>then pipewire should run before pulseaudio restarts <pkill9>what i need is to make a guix system service that starts my user shepherd <nckhexen>lechner: Do you have an /etc/ssl? I don't, on --commit=582b1f6. <nckhexen>It's a dead link. I assume that's not a bug in my VM, but checking just in case. <gnucode>unmatched-paren: are you helping the guy make that emacs home service ? <gabber>unmatched-paren: it does? would you mind to share? <gabber>i'd love to have a look and try! <unmatched-paren>Gah, just when it was all going so well, I tried logging out, logging in on tty3, and ``emacsclient -nw''... the socket refused the connection D: <lechner>nckhexen: yes, i do. i am on commit e0b414fc599c2d9092dfa57455f035cbedb7810e, i think, but also have the other non* repo activated <gabber>does your home service stop on each logout? <nckhexen>No point in debugging this if my VMs aren't being built right. <lechner>nckhexen jackhill: curl also reported using gnutls * GnuTLS ciphers: NORMAL:-ARCFOUR-128:-CTYPE-ALL:+CTYPE-X509:-VERS-SSL3.0 <unmatched-paren>I looked at their patch, and (I don't intend to be rude or patronising; I'm aware that it's their first patch, and it's a pretty good first patch) I wasn't too impressed (it doesn't handle launching emacs --daemon at all, for example) <jackhill>lechner nckhexen interesting. I'm working on a bisect currently <gnucode>unmatched-paren gotcha. How do you manage to launch emacsclient in sway? Do you have a keybinding for it? <unmatched-paren>gabber: It should stop at logout, given that it's a home service and thus, afaics, tied to a session? <gabber>do many people use emacs in a non-daemon way? if so we should consider that as an option to the home-service <gnucode>ahhh. gabber I guess I use emacs in a non-daemon way. hahaha <gabber>unmatched-paren: so my guess is it doesn't nicely stop or not start after login <gnucode>unmatched-paren: I gave up rolling my own config...I'm on doom emacs now. <unmatched-paren>i'm pretty sure it stops cleanly, since i don't just kill it to stop the service, i do ``emacsclient --eval (kill-emacs)'' like you do <gabber>and the bug doesn't reproduce anymore? *unmatched-paren gonna log out again to test *unmatched-paren going to restart, to see if it works ootb after that <kaelyn>Has anyone else encountered this error building email@example.com for i686 on core-updates: "/tmp/guix-build-file-5.42.drv-0/file-5.42/src/.libs/file: symbol lookup error: /gnu/store/5vac93xqlc3zq6jhdg097k0jvw0g1y22-glibc-mesboot-2.16.0/lib/libpthread.so.0: undefined symbol: h_errno, version GLIBC_PRIVATE"? <unmatched-paren>gabber: I think this confirms your hypothesis about it not cleaning up something when logging out... <unmatched-paren>So, logging out, and jumping to tty3, then doing emacsclient -nw, now prints ``*ERROR*: cannot open file: /dev/tty3''... <unmatched-paren>And after that I get the system-error when restarting the emacs service. <unmatched-paren>I don't see anything special in make-kill-destructor that would make this work... <gabber>ah, you mean using emacsclient among different consoles? aren't there login-specific XDG envvars? <gabber>does that actually work when you're using emacs without your home-service? <gabber>aha! i guess that's a fd.o feature <unmatched-paren>Without even installing any emacs packages into the profile, except for emacs itself :D *unmatched-paren git format-patch \o/ <gnucode>unmatched-paren: should I go ahead and tell the other guy working on the emacs home service... to work on something else? <unmatched-paren>gnucode: It seems like they've been working on that for a while, I don't really want to make them feel as if all their work was for nothing... <gnucode>yeah fair... I'd rather know if someone else completed my opensmtpd-records though.... <gnucode>that's the only thing I've been working on. :( <gnucode>I honestly have no idea how you have the dedication to create as many patches as you do. :) <unmatched-paren>i want to start working on some of my own projects, but actually getting things started is a chore, and it's much easier to contribute to an existing project like Guix <nckhexen>jackhill lechner: You're missing /etc/ssl/certs/DST_Root_CA_X3.pem , right? *nckhexen got called away, sorreh. <gnucode>unmatched-paren what sort of projects do you want to work on ? <jackhill>that file seems to be in mjmpb4k2g21p7hyx9zq57p9xymbl16ac-nss-certs-3.71 <jackhill>but maybe not in 1klwvqm3njp070h982ydcix1gzf2zmdl-nss-certs-3.81/ <gnucode>unmatched-paren somehow I doubt that. I hear tell that you have GC code that will trash unused objects at exactly the right time. :) <gabber>unmatched-paren: to your statement before mine. it's fun! <tribals>Is there a way co get os configuration from runnig OS made from unknown source file? <jackhill>Presumably nss-certs removed it for a reason, but why does Let's Encrypt still use it? I guess that isn't a Guix problem <jackhill>or maybe it's something acme.sh is doing wrong that certbot doesn't do. Either way, not guix's problem… <nckhexen>Oh, that was little more than a guess. There was 0 time between my last peek at this earlier and now. No thought took place. <nckhexen>And the bad commit has ISRG_Root_X1.pem so that ain't it. <nckhexen>(And it's not missing from ca-certificates.crt) <nckhexen>‘dynamic linker name not known for this system "x86_64-linux"’ <tricon>the running config is at: /run/current-system/configuration.scm. what are you hoping to do? <jackhill>nckhexen: somehow I have the memory that openssl looks at one and gnutls looks at the other. Not sure though. <nckhexen>Yes, the above error is what I got when trying to install curl to check that (I don't know my way out of a wget paper bag). <tribals>tricon: I'm trying to use it as base point for my own configuration, but I don't know much how to write configurations <lechner>nckhexen jackhill: my affected system also lacks the DST certificate, and a working system has it but i am not sure how that is related <tricon>tribals: happy to help where i can. it can feel daunting at first, but i recommend initially looking at it as a "configuration language (DSL). as your understanding of Scheme/Guile increases, you'll see the programming underneath it and understand how you can accomplish various tasks. <lechner>nckhexen jackhill: my affected system also ships ISRG_Root_X2.pem in addition to the X1 available on both, although i am not sure how that is related, right now either <tricon>unmatched-paren: your avatar on that site gives me anxiety. <gnucode>unmatched-paren: would your emacs home service work with say doom emacs ? <gnucode>I guess I'll just have to test it... but I haven't actually migrated to guix home yet.... <tricon>unmatched-paren: hah! i hadn't seen that one. perfect. <tribals>tricon: the most challenging part for me is how to write configuration and do not loose access to remote machine <tricon>tribals: understandable. i've had to remote to the console of a VM before for that reason.
***daviid` is now known as daviid
<lechner>build.bash is located in the root of the source repo <unmatched-paren>(lambda* (#:key import-path #:allow-other-keys) (invoke "sh" (string-append import-path "/build.bash"))) <lechner>unmatched-paren: thanks! the import-path is not enough, though <lechner>sh: github.com/rfjakob/gocryptfs/build.bash: No such file or directory <unmatched-paren>lechner: you might want to just write a src/(import-path)/VERSION file containing #$version i guess <nckhexen>jackhill: lechner: Back again. The presence & absence of DST_Root_CA_X3.pem is definitely involved in this, whether or not it shouldn't be. <nckhexen>jackhill: Uhm, you said SSL Labs was happy with your server (silly of me not to check): ‘Chain issues: Incorrect order, Extra certs’ <nckhexen>Which again points to this weird concatenation you're doing. Are you sure that's legit? <unmatched-paren>so, just patch that bit away i guess, and use GITVERSIONFUSE=#$(package-version go-...-fuse)'' <lechner>unmatched-paren: also, why is it building with go 1.17.11 instead of 19 <nckhexen>jackhill: Even so, can you try without it? Debugging an incorrectly configured server is not what I signed up for 😛 <unmatched-paren>lechner: the latest go available in guix is 1.19, but the ``go'' variable used in builds is 1.17 <unmatched-paren>because if we changed ``go'' it would trigger mass rebuilds of the go ecosystem <nckhexen>jackhill: You're serving this bad certificate yourself, no wonder I couldn't find it in Guix :) <nckhexen>jackhill: Specifically, stop sending X1, and you should be good. That is Guix's (well, nss-certs') job, not Apache's. <vivien>Why does self-contained-tarball in (guix scripts pack) use gexp->derivation and not computed-file? I wish I could use its output as a gexp input <nckhexen>unmatched-paren: Not counting x86_64 just for fun! <vivien>I’m trying to populate a flatpak repo with a pack of my application, and I need a self-contained-tarball for that <vivien>Maybe I could find all transitive inputs and take a big directory-union of that <vivien>(no it would not work, I need a self-contained thing) <vivien>Can I use a derivation as a gexp input? I think I could by running it, get the output file name in the store, and convert it to a local-file, but it feels very un-guixy <jackhill>nckhexen: thanks, sorry about that, let's see if I can fix it <vivien>The manual says I can do #$obj when obj is a derivation, but I get "invalid G-expression input"… <nckhexen>jackhill: I'm not annoyed at you for saying ssl labs was ‘happy’, I'm annoyed at myself for not verifying that. I understand that the big green ‘A’ (or whatever it was) made you think so. <nckhexen>‘Yey, your site is secure! …on old systems that happen to shadow the bogus root you ship, but who's counting.’ <jackhill>nckhexen: yes, also, I've fixed it to the non-expeired root now, but I still have "incorrect order, Extra certs". Works thought! <lechner>unmatched-paren: hi, what does this mean, please? go list -m: not using modules <lechner>jackhill: was this an issue with acme.sh? <unmatched-paren>lechner: i'm not sure, i think it means we're using GOPATH instead of the new "go modules" system, so ``go list -m'' can't possibly work? <jackhill>lechner: yes, at least since I had been using the acme.sh installation for some time, it was defaulting to the expired chain. I fixed it with `./acme.sh --set-default-chain --preferred-chain "ISRG" --server letsencrypt` <jackhill>and thanks to guix for surfacing the issue before someone not-me complained :) <lechner>jackhill: Your site says " This page intentionally left blank." <lechner>unmatched-paren: the network access seems more serious <jackhill>lechner: on purpose. Maybe some day it will say something else. <lechner>jackhill: no problem, but why would anyone complain about it not serving correctly? <jackhill>lechner: oh, you mean because it's blank? I do have some non-crawlabe pastbin type stuff there, and other sites with the same configuration <lechner>jackhill: sorry, just being facetious! good to see you here, btw. i think a read a bunch of your stuff in or around #haskell <unmatched-paren>(apply (assoc-ref %standard-phases 'build) #:import-path directory arguments) <vivien>So maybe I understand the problem: self-contained-tarball does not return a derivation, it "monadically"-returns a derivation, so I have to do #$(run-with-store (open-connection) (self-contained-tarball …)) instead. I’m not sure, because I don’t understand exactly how I should create the second argument ("profile") of self-contained-tarball. <vivien>I tried "make-manifest" but it seems to fail <unmatched-paren>I had to mess around a wee bit with substitute* to patch some executable invocations. <unmatched-paren>The 41st is for updating tcell-v2 to 2.5.3, the 42nd is for adding tcell-term. <vivien>It was: $(run-with-store (open-connection) (mlet* %store-monad ((profile (profile-derivation (packages->manifest (list package …)))) (tarball (self-contained-tarball "the-pack" profile))) (return tarball))) <vivien>The pack is suspiciously large though <vivien>That’s a bit much than what I expected for guile and gtk <lunabee>are go cli tools also packaged under the go-domain-package convention or just libraries? <nckhexen>(The go-read-this-very-long-package-name-ill-wait style names, I mean. <lunabee>any reason guix import uses that format then? <nckhexen>…actually I can't really think of plausible exceptions so s/Try to //.) <nckhexen>lunabee: Because it can't know. Guix import should not be used without human supervision & loving correction. <lunabee>wait, avoid go-package-name in general or for only tools?