<mbkamble>Hi. What is the correct command sequence to shutdown "user shepherd" after completing "guix home reconfigure ...." before doing a desktop session logout. Presently after logout and login, the shepherd process and all sessions under it (eg gpg-agent) from the old session are still active when I login. This there are multiple "user shepherd" and <mbkamble>user services after several desktop logout and logins. <mbkamble>Should I just do a kill -9 PROC_ID_OF_USER_SHEPHERD before logout? <unmatched-paren>I'd have thought they'd target places like Reddit, Discord, possibly Matrix; more popular networks and sites whose residents are perhaps more susceptible to this kind of thing... Apparently not :) <PotentialUser-50>Hello. Got Guix on my pinephone. Trying to install emacs-next-pgtk, but the whole phone crashes when building. Very new to Guix, but I think it may have something to do with the phone being low-spec. Is there somewhere I can specify the -j2 make flag or equivalent? <yewscion>mbkamble: Does `herd stop shepherd` not work for this use-case? <mbkamble>thanks yewscion. ithat works. t was as simple as that? follow up questiosn: "herd status" shows "root" as one of the services. What does "root" service mean or do? <yewscion>mbkamble: AFAIK "shepherd" is just an alias for "root" in the above (or vice-versa). It refers to the shepherd process herd communicates with. <whereiseveryone_>PotentialUser-50: Seems like pinephone pro's specs might be better suited for Guix. I would have tested this if it weren't for my PPP being donated to The Salvation Army... <singpolyma>whereiseveryone_: upstream guix thinks that guix-history channel is the answer to that <singpolyma>Maybe there are packages relying on specific ruby versions so they keep them for now ? <whereiseveryone_>I'd prefer to get all my history from guix's master branch but maybe I haven't warmed up to guix-past yet <singpolyma>I'd also like that, but they want to be more like Debian than nix I think :) be "a distro" with a known good set that is tested together and not much else <singpolyma>Which is totally valid. I'm not having, obviously the people doing most of the work get to decide :) <xelxebar>It's not great, but `EDITOR=cat guix edit` effectively achieves that. <whereiseveryone_>singpolyma: Maybe become one of the people doing most of the work and then propose the idea... <xelxebar>Exactly. It's not precisely what you want, but you could pipe to less and then search for the string "define-public nnn" <xelxebar>Which isn't what you want at all, probably :P <xelxebar>What are you trying to achieve, exactly? <xelxebar>I mean, what are you trying to achieve with that? <xelxebar>`EDITOR=vis guix edit` does'nt work as expected? <whereiseveryone_>In vis, if I'm in another file I can pipe in the package directly where my cursor is <whereiseveryone_>so with vis open I want to do exactly the following in vis' command mode on the current line that I'm at: `:<guix cat nnn` <xelxebar>I see. You're thinking of copying in the define-public form of some package as a template or something? <whereiseveryone_>That would bring in the lines for the nnn package where my cursor is currently at very conveniently <whereiseveryone_>would be cool to do the same for any other guile symbol (think language server protocol) <xelxebar>Yeah, that use case isn't super convincing :P <whereiseveryone_>why limit it to opening a guix package in your editor but prohibit opening a package to standard output? <xelxebar>EDITOR=cat achieves that already. What you're asking for is entirely different functionality that extracts just the define-public form of a package definition. <xelxebar>It's not obvious that this would be generally helpful. The isolated definition has module dependencies, and might even be using helper functions defined earlier in the file. <whereiseveryone_>I'd be happy to first just implement it and put it in a channel or library <xelxebar>Yeah, I'm not entirely sure how `guix edit` is sending line number information. You should be able to grab that somehow. <whereiseveryone_>I'm just interested in making this work for my use cases and if others want it then we can upstream it <xelxebar>Try this: `EDITOR=echo guix edit nnn` :D <xelxebar>The edit commands looks to be specifying line number in the file like `+<line-number>`. You might be able to hack something together quikly with a custome "EDITOR". <xelxebar>Hrm. `vis $(EDITOR=echo guix edit ...)` should be exactly the same as `EDITOR=vis guix edit`. <whereiseveryone_>`bat -r $(EDITOR=echo guix edit nnn | awk '{print $1}'):+20 $(EDITOR=echo guix edit nnn | awk '{print $2}')` <xelxebar>+1 for every open parens -1 for every close parens. When your sum reaches 0, you've reached the end of the s-exp <whereiseveryone_>` bat -r $(EDITOR=echo guix edit nnn | awk '{print $1}'):+32 $(EDITOR=echo guix edit nnn | awk '{print $2}')` <xelxebar>So you basically just want to feed `+n:+32 <file-name>` to your editor, right? <whereiseveryone_>would you write that as a script that parses the output of `EDITOR=echo guix edit nnn`? <xelxebar>Yeah, grabbing the entire form is good programming fun. Would be good for you to try. <xelxebar>Your input is two strings `+<n>` and `<filename>`. Your target output is a string that contains the entire `package` or `define-public` s-expr. <whereiseveryone_>The part I need to compute is the offset from `+<n>` till the end of the package? <whereiseveryone_>since bat allows me to give an offset instead of exact absolute line number <vagrantc>grrr. my reproducibility patches for itpp and discrover were only partly effective. <vagrantc>those were supposed to be my test cases for trying to patch texlive to do the "right" thing all that time <xelxebar>whereiseveryone_: Try this (http://ix.io/41yq). If you put that in ./guix-cat.sh, then `EDITOR=./guix-cat.sh guix edit hello` I believe does what you want. <gnucode>xelxebar: is there any plan to use ix is guix? eg: paste.guix.gnu.org <xelxebar>It's simply the only pastebin that cli-friendly by default that I know of and, thus, my personal preference. <PotentialUser-29>I follow this to partition, and modify config.scm for the system, but it's not working <gnucode>sneek: later tell nckx that I have updated my opensmtpd-records patch for a version 2 that includes texinfo documentation. Version 3 of the patch will try to format the code to guix's prefered scheme format. <ncbfg36>I am having trouble installing guix to LVM volumes in a luks encrypted partition on a system running libreboot firmware. This is the output of running "guix system init /mnt/etc/config.scm /mnt" https://paste.c-net.org/GrievousTrunks and this is my config.scm adapted from bare-bones.scm https://paste.c-net.org/LacedDesign . Looking at search results I'm assuming that i've made some error in my config.scm <ncbfg36>but i'm struggling to work out what's going on. I've booted installer, opened luks volume, installed lvm2, activated and mounted LVM volumes. Am I missing something obvious here? <ncbfg36>Since the last mentioned ice-9/boot-9.scm raises wrong type argument have a made an error in the bootloader section? <ncbfg36>raghavgururajan: thankyou for looking into this. The two sources you mentioned are the ones I referred to when setting up. I will look over your config.scm to try to identify any significant differences. I am new to guix so it's all a bit mysterious to me <raghavgururajan>ncbfg36: Found it. The syntax for `swap-devices` has changed after those guides were made. *raghavgururajan should update those guide ... <ncbfg36>raghavgururajan: Okay now it throws the error "error: swap-space: unbound variable \n hint: Did you forget a 'use-modules' form?" <ncbfg36>I can't find the "(swap-space " mentioned in guix docs <jpoiret>ncbfg36: you're using the installer image, right? <jpoiret>if so, it doesn't have the swap space changes <raghavgururajan>ncbfg36: Ah, are you using the latest-release image or stable-release image? <silicius[m]>Weird, I couldn't update python-pymdown-extensions to 9.5 because guix cannot download it during guix build <ncbfg36>jpoiret: raghavgururajan: yes i'm using the 1.3.0 iso. <jpoiret>then you don't need to use the swap-space things <jpoiret>you'll have to update your config.scm for them though after having installed <ncbfg36>okay. so I should revert to the old syntax? <ncbfg36>I tried commenting out swap-devices completely and I get that error result again <jpoiret>you'll want (swap-devices (list "/dev/mapper/GUIXVG-swapvol")) for now <ncbfg36>jpoiret: okay thanks. So now i'm dealing with the original error. <ncbfg36>I've started with the bare-bones.scm located on the install iso and just added what i need to have the simplest starting point. As far as I can tell the syntax is correct and the brackets are all closed off in the right place. <jpoiret>ncbfg36: your (file-systems (append (list ... %base-file-systems))) isn't parenthesized properly <jpoiret>it should be (file-systems (append (list ...) %base-file-systems)) <jpoiret>you should use an editor that highlights pairs of matching parentheses <ncbfg36>jpoiret: ahhh I thought the final bracket after %base-file-systems))) was supposed to match the leading bracket (file-systems <jpoiret>Note the difference between both of my examples though <ncbfg36>by my count it's correct then. I tried what you suggested and it threw an error about missing closing parentheses <jpoiret>You're adding %base-file-systems inside the (list. ..), whereas it should be the second argument to append <jpoiret>I'm not saying you should remove a closing parenthesis <ncbfg36>jpoiret: OH! haha yes I see. I'm using nano so just jumping between bracket pairs. Bracket highlighting would be very useful here <ncbfg36>I tried "modprobe dm_mod" but that didn't help ***Edward123 is now known as EdwardIII
<ncbfg36>okay so using '(file-system-label "swap")' worked. It's now downloadimg from repos. Thanks so much for your help jpoiret :-) <ncbfg36>raghavgururajan: can I ask which system you are using libreboot + guix on? <ncbfg36>raghavgururajan: oh cool! I'm working on a T500. I'm waiting on parts from ebay to modify it for a quad core QX9300. Was slightly concerned about spending money on such an old system <sneek>nckx, gnucode says: that I have updated my opensmtpd-records patch for a version 2 that includes texinfo documentation. Version 3 of the patch will try to format the code to guix's prefered scheme format. <nckx>ncbfg36: You do the BGA work yourself? <silicius[m]>I found the source of the problem (I think). pymdown-extensions uses hatchling instead of setup.py to build and install, which is not currently present in guix (and most likely it is not supported by pypi too) <ncbfg36>nckx: Oh I see what you mean. I just found guides online. Need to snap a few pins and solder a wire to make a bridge. Doesn't seem to complicated <nckx>silicius[m]: I don't understand. I was just trying to reproduce your report and (pypi-uri "pymdown_extensions" "9.4") is perfectly valid. <nckx>How would setup.py affect that. <nckx>ncbfg36: Oh. I thought you were actually replacing the CPU yourself. <silicius[m]>nckx: The problem was not pypi-uri, pypi just does not have this build, it's 404 <nckx>Which build? As noted, I tried 9.4 with success. <silicius[m]>ok, I see the problem. pymdown-extensions works for <9.3, pymdown_extensions works for later versions <nckx>ncbfg36: Oh, wow, you have an actual *socket* on those things. I had no idea. <silicius[m]>yep, it worked. Though it still can't be build since it uses hatchling to build at this version and it's not in guix afaik <ncbfg36>nckx: that's what I love about old thinkpads. Can replace just about everything <nckx>silicius[m]: Let's hope hatch(ling) doesn't build using itself. <f1refly>how does one use lxc on guix? I want to build my own calyxos to include magisk using a debian container, but I don't know how to enable network access <silicius[m]>nckx: I made so many branches with new packages and fixes just to package hydrus, it's getting out of hand <f1refly>I don't think it's mentioned in the guix manual either <nckx>ncbfg36: I wish they could eat just a bit more RAM (but then I lament the same of the 230s). <ncbfg36>nckx: yeah I think the T500 maxes out 8gb. I'm very keen on getting a frame.work laptop when they ship to Australia <nerthus>feels like there's more effort spent on sending stuff to Mars than Australia the way I hear ozzies banging on about shipping on pretty much every item hehe <ncbfg36>nerthus: haha well we pay a huge tax on tech compared to USA <nckx> I've signed up for their 'let me know when you bother shipping here' mail, although I haven't decided yet. <nckx>(This is just regular Europe, too, not dropbear-eating snake island.) <nckx>I guess !US == out of luck. <nerthus>well even in Europe you get the fun of digging through on what Americans mean by "we ship to EU!" <nerthus>ok... do you mean EU as in Europe or European Union, cuz sometimes countries like Serbia are just exempted <nckx>Extremely Western Europe. <nerthus>honestly when they say we ship to EU, I just parse that as "we ship to UK" <ncbfg36>frame.work ship to UK, france, germany, austria, ireland, netherlands, <nckx>Importing stuff from the UK now costs about as much has having it shipped across the Atlantic anyway. <ncbfg36>some people use forwarding couriers but that messes with your warranty <nckx>ncbfg36: I am literally surrounded by those countries. On all sides. Hah. <nerthus>so much for being in the Benelux, ayy <nckx>If they deliver to Luxembourg I'm sending a mail. <nckx>'Did a Belgian hurt you, and would you like it to happen again.' <nerthus>funny enough, they might open up a Luxembourg HQ "for its central location" wink wink <nckx>We have those. We compensate. <nerthus>ofc, gotta keep your 4 governments running <ncbfg36>I wish I lived in Europe. Australian govt wants us to have crap internet and store all our personal data in plain text with some legalese saying only the authoraties are allowed to access it <unmatched-paren>ncbfg36: I suppose the legalese is the only part that could pass as encrypted. <ncbfg36>unmatched-paren: depressingly accurate <nckx>ncbfg36: Ah, crime must be against the law in Oz. Clever. <nerthus>no, following the law is a crime in Oz <ncbfg36>literal quote from former PM regarding backdooring encryption "The laws of mathematics are fine and good, but the only laws I concerned with are the laws of Australia" <nckx>There are context within which that's a reasonable and intelligent (if risky) thing to say. However, my limited knowledge of Aussie ex-PMs suggests it was no such context. <nckx>I mean, US legislators *have* tried to legislate the laws of mathematics, which is worse. ***wielaard is now known as mjw
<ss2>I just noticed: pulling Guix' channel from a cloned location does not through this stale warning. <oriansj>nckx: very true, Kansas did legally define PI as the number 3 for a while there <oriansj>To which the only defense against bad government is software development distributed around the globe; so until there is a single world government we can route around that damage <dhruvin>I've been trying to fix the issue for some time with no luck. Any help is great! :) <dhruvin>The error says unsupported manifest format, but I haven't given it any manifest file. So, I believe this is something internal. <rekado>dhruvin: my guess is that you’re using an older version of Guix on a newer profile <rekado>the profile manifest format has recently been bumped to allow for intra-manifest references <rekado>commit 4ff12d1de7cd617b791996ee7ca1240660b4c20e <dhruvin>Thanks rekado. Is it possible that `guix system image` command may be generating an image with older guix installed in it? <ncbfg36>raghavgururajan: I hadn't heard of OSboot. The reason I went with T500 (years ago, only just found the time) was reading that it's the last generation that could remove intel ME completely. I'm not super paranoid about it but I thought "why the hell not". The T440p has much more useful specs though. <nckx>dhruvin: I think the guix is always older, being the 'guix' packaged in the generating guix version (i.e., 'guix show guix' vs. 'guix describe'). <dhruvin>IIUC, when I'm doing `guix system image ...`, I get the `guix show guix` one installed. I need latest guix in that image, so I run `guix pull` inside a vm of that image to install latest guix (which is the `guix describe` one, as you mentioned). <dhruvin>Am I doing something wrong here? As in is there another way to have latest guix installed with `guix system image ...`? Or some other way to update that old guix to latest in aforementioned image? <nckx>I do not see anything wrong, this is all expected. I'm not aware of a built-in way to directly use the 'outer' guix as guix inside the image without pulling. I don't know if the versions are sufficiently apart to be responsible for your problem, just pointing out there will always be a difference as it's currently designed. <dhruvin>Okay. I'll see if the versions are too apart to cause this issue, and if they are, how to package somewhat newer version of guix with `guix system image`. Thanks. <sneek>denna[m] was in #guix -34 minutes and -13 seconds ago, saying: unmatched-paren: Ah i see rde is not exactly guix. .... <civodul>does that mean that dannym *will* be here in 34 minutes? <nckx>sneek broke into the time machine again. I told you not to keep botsnacks in there. <nckx>dhruvin: There's a script (I forget what it's called, sorry) to update it, because it's a minor pain to recover from botched updates, but code-wise it's a bump like any other. <jpoiret>there are many issues with trying to provide the current Guix for a system image <jpoiret>we'd need to keep the sources around in the store somehow <jpoiret>i've been tinkering with adding posix_spawn (it's nearly ready for prime-time, i'm squashing the failed tests right now) and the different interactions between the guix package and the guix pull guix and their respective build systems have been annoying <jpoiret>it makes it harder for someone who hasn't stared at that mess for long enough to understand which modifications end up where and when <jpoiret>i think one of the blockers for a change with how we currently handle guix-in-guix would be to compartimentalize the daemon (and now the extensions also), that way a new commit in guix wouldn't cause guix-daemon and guix-extensions to be rebuilt <jpoiret>otherwise that'd be wasted CI time, and bad `guix pull` substitution *dhruvin is looking for that script in guix source <nckx>It might not be in a separate file. Start with 'guix edit guix', I think there's a comment there. Else where would I have read it... <jpoiret>civodul: i'm modifying (guix inferior) to use the glibc posix_spawn instead of primitive-fork, and everything seems to work well right now, but I seem to have broken the "&inferior-exception, legacy mode" test and i don't really understand how it's supposed to work right now <jpoiret>you've written that test, that's why i'm pinging you! it claims that open-inferior-pipe would run "guile" and not "guix repl", but the implementation of open-inferior-pipe says the opposite <jpoiret>to make matters worse, stracing the test shows that the execl() done by open-bidirectional-pipe errors with ENOENT (because bin/guix doesn't exist in the repo anyways) <jpoiret>(glibc posix_spawn is able to directly report ENOENT to the parent rather than letting the child spawn and erroring, that's why with my changes the error becomes apparent) <reyman>Hi there, i need some help to repair my guix, i'm outside, i cannot even rollback, there is something that blocked just after login, i try with old generation without success <reyman>so i suppose i could try to repair using chrooting from live cd <jpoiret>chrooting is hard to get right, you'll have to mount the store properly too <reyman>@jpoiret the system freeze just after text login appear <jpoiret>for your rollbacks? did it work before upgrading? <jpoiret>can you focus a grub entry, press 'e' and add 'nomodeset' to the linux line parameters <reyman>ok thanks, i will try and i'm go back (i have only one machine) <reyman>i delock my disk, but i need some advice to found the correct log file <reyman>@jpoiret in my message log i found the problem with gnome gdm <reyman>Jul 4 17:52:07 localhost vmunix: [ 34.144077] NET: Registered PF_ALG protocol family Jul 4 17:52:07 localhost elogind[408]: New session c1 of user gdm. Jul 4 17:52:07 localhost gdm-session-worker: GLib-GObject: g_object_unref: assertion 'G_IS_OBJECT (object)' failed <jpoiret>yeah, it's possible that GDM locks up because KMS doesn't work properly <reyman>i just add nomodeset at the end of grub line <reyman>yes everything goes well, but i try to remove gdm login using some exemple of transformer of dominicm <reyman>and probably i miss something in the config ... <reyman>just to be sure, i try to add nomodeset after "... modprobe.blacklist=x,x nomodeset text" <reyman>is there some tutorial/info to chroot a guix encrypted <jpoiret>yes, seems like it's a fine place to put it <jpoiret>it's very weird though that you're not able to boot previous generations <reyman>yes, i don't know why ... this is the first thing i try to do <jpoiret>can you remove /var/lib/gdm and mkdir it again? <reyman>@jpoiret, you right, this is that <reyman>i remove, reboot with an old config and it's work. *rekado rebased wip-guix-log <dhruvin>PurpleSym: I have similar issue with sourcehut CI. See logs of last couple of hours for context. <PurpleSym>dhruvin: So `guix pull` is broken? I successfully pulled today on my local machine 😕 <dhruvin>guix pull from an old guix seems to be broken. <dhruvin>guix, inside (gnu packages package-management), is v1.3.0. Which doesn't seem to be able to understand manifest v4 (something that has been recently updated). <dhruvin>This is my understanding of the situation. Not sure exactly what's wrong. <PurpleSym>I’m fetching the latest nightly guix installer though. <dhruvin>I don't know whether the nightly installer comes with guix more recent thatn 1.3.0. Can you check it using `guix --version`? <PurpleSym>I’m seeing paths including guix-1.3.0-27.598f728 in there, so it should be the most recent one. <dhruvin>git show 598f728, shows a commit from May 20. <PurpleSym>Ah, right. And the manifest change occured later, so the current `guix build guix` cannot read manifests from `guix pull`’s guix. <PurpleSym>And voila: `guix shell guix -- guix describe` fails. <rekado>maybe we should update the guix package in gnu/packages/package-management.scm <rekado>I can’t do it now, but maybe in a couple of hours I could take some time <PurpleSym>Does it have any implications for `guix pull`? <rekado>if you use “make update-guix-package” successfully in a clean working directory with only published commits nothing should break. ***Xenguy__ is now known as Xenguy
<nckx>dhruvin: make update-guix-package is what I meant... not a script. <dhruvin>nckx: Yes, I followed your advice and found build-aux/update-guix-package.scm script you suggested. <nckx>Yes, I was right after all \o/ <PurpleSym>I can’t update the guix package, because the testsuite fails at tests/services/telephony.log right now. <nckx>That sounds familiar; Jami? <nckx>Those tests will be fixed but 'later'. <PurpleSym>`ERROR: In procedure %resolve-variable: error: parse-dbus-reply: unbound variable` ***mark_ is now known as mjw
<PurpleSym>Ah yeah, there’s more and I’m running out of time. Maybe just xfail them? <nckx>Frankly, if you patch the script/target to skip that test, go ahead. <nckx>Well, shouldn'tfailbutcurrentlybroken, yes. <nckx>More unrelated tests not in telephony.scm? I've got to go, but those should be assessed separately. ***GNUtoo_ is now known as GNUtoo
<nckx>reyman: Thanks. We should keep an eye on whether it arrives. <nckx>The comment form has been broken before... <rekado>has it ever been fixed? I thought it had been disabled pending a review of why it broke… <jpoiret>I was looking at updating the package as well for some other changes I'm doing and yes, those tests are still broken <jpoiret>i think the better way would be to just remove the test contents in tests/services/telephony.scm, rather than drop it from the Makefile.am and forget to re-add them once they're fixed <jpoiret>the tests take a very long time locally for me <tissevert>I'm finally looking at guix home and I can't help but wonder: where does it fit within an operating-system declaration ? <tissevert>does one really have to generate a system, then boot and login to said system, before being able to configure their home ? <unmatched-paren>i suppose you're thinking of `sudo guix system reconfigure` running `guix home reconfigure` for all the configs specified? <tissevert>no, I'm more wondering how I'm going to make a live CD with a set of predefined packages / config for the user, when I can't alter that RO system after the install <tissevert>I was almost sure I had seen a "packages" field in the user accounts description, and thought maybe a whole home-config had been added since <tissevert>I was distressed to find I couldn't even set the packages : ) <unmatched-paren>i guess we could have a (configuration ...) field in <user-account>. it'd certainly be possible <unmatched-paren>(configuration (home-environment ...)) or (configuration (local-file ...)) ; can be replaced with any file-like object <yewscion>Hello all, I am trying to play a MIDI file on my Guix install, but when I load the FluidR3Mono_GM.sf3 file into timidity the sound is /incredibly/ distorted. Is there something I'm overlooking? I am used to using .sf2 soundfonts, but I'm happy to see a freedom-respecting format exists instead, and if I can get it working I hope to use that going forward. <morganw>If anyone happens to use Common Lisp, how would I work around this problem when requiring a library? "ASDF could not load esrap because OS-FILE-ERROR(EROFS): Read-only file system." <tissevert>yewscion: sorry I'm not on my guix system right now but I'm willing to test if I have the problem too, I intend to use guix to make music at some point <yewscion>tissevert: Awesome, thanks. Let me know if You don't; Could very well be my config. <yewscion>morganw: I just started SLIME on my Guix after installing cl-asdf and sbcl-esrap, and got T as the return value when I tried (asdf:load-system :esrap). Is there a more complex command You are running, so I can test it? <morganw>It seems to only break when trying to recompile the library, when I use clisp as the lisp. If I use SBCL then it seems to work. <morganw>I wonder if clisp is actually "unsupported" outside of bootstrapping because it looks like it is trying to compile paths that have "sbcl" in them. <morganw>(I presumed that cl-esrap was portable and sbcl-esrap would have been specific, I have both packages installed) <yewscion>Ahh, that would be unfortunate. I rarely fall back to clisp, but it's good to know it's usable if I need/want it. <yewscion>morganw: I ran `guix shell --pure clisp clisp-esrap clisp-asdf bash` , then ran clisp. In clisp, I ran `(require "asdf")` and then `(asdf:load-system :esrap)` and it warned about SET being used but otherwise returned normally after compilation. <yewscion>I can upload the log somewhere, if You think it'd help. <yewscion>Shot in the dark: You don't have Your `common-lisp` cache dir in guix-home, do You? I can see that causing this problem; I had a similar issue when setting up a emacs with my custom.el file. <tissevert>unmatched-paren: is there a place for feature request ? opening an issue sounds a bit too much for just a simple idea like that <tissevert>any pointer as to where to start from to implement such a thing ? <tissevert>I suppose wiring a config to a (configuration …) field like you suggest wouldn't be the hardest part ^^ <morganw>yewscion: Thanks, it works for me with your 'guix shell' example too. I do have all of my packages installed with 'guix home', but I haven't specifically done anything with a the cache directory. Is just installing all of these packages with 'guix home' enough to cause a problem? <tissevert>oh, ok ! thanks ! I'll try and open one describing this idea in the next few days <morganw>yewscion: it seems to work when I remove the sbcl-esrap package. I'm not sure how it ends up trying to load that though. <morganw>Is it actually unsafe to have packages installed for different implementations at the same time? <nckx>rekado: I'd disabled it a good while back with a stern 'don't re-enable this lightly' comment, and I see it has been re-enabled, presumably not lightly. <nckx>But I don't see a reply. :-( <civodul>nckx: i fixed the mumi/mailutils issue a long time ago, later re-enabled it in berlin.scm, which was deployed later, and here we are :-) <nckx>Well, I don't know for sure. *unmatched-paren wonders whether they've asked for a review for their aerc patchset too many times <nckx>Is this a statement of intent? Maybe. <nckx>It didn't sound like it but I can't be 100% sure. <nckx>reyman: Please let us know :-) <civodul>"Throw to key `mailutils-error' with args `("mu-message-send" "Cannot send message" #f (4132))'." <nckx>Yes yes. It's hard enough to type ASCII without mistakes here :) <nckx>I don't blame you btw. It's fragile by nature. <civodul>(mu-strerror 4132) => "Process exited with a non-zero status" <nckx>unmatched-paren: The lack of results suggests trying a different approach, not that I know which. I'd review it, I like mail, a lot, but you know. <civodul>"/var/mumi/mumi-mailer.rc: contains secrets and therefore must have no more than user read/write permissions" <nckx>That further reduces the fun factor. <unmatched-paren>Although it's only (searches for 'gnu: Add aerc' in aerc) 34 go packages plus aerc! :P <nckx>Modern packaging is fine. <unmatched-paren>That's probably tiny compared to what the eventual 'gnu: Add hugo' patchset will look like <nckx>Well, once my overpriced fan & bucket o' thermal paste have finished their grand tour of every continent except the right one, unmatched-paren, I will at least have a look, despite non knowing a thing about Go. <nckx>I think they are currently on an Arctic cruise. <nckx>Thanks civodul! That was fast. <nckx>Any chance their message was logged/queued? <unmatched-paren>nckx: Thanks for the offer :) Obviously if you aren't interested in the package you have no obligation to review it, I don't want to be pressuring anyone. <f1refly>is there a way I can tell guix that I want to build an android system or do I have to look up all the required packages myself and possibly put them in a package.scm? <unmatched-paren>umm ;; XXX Bundles jquery and several other .js ‘libraries’ — problem? <civodul>nckx: looks like previous messages were lost :-/ <unmatched-paren>otoh, if you try to package jquery you'll come out wishing you'd never been born. I think we can live with it. <yewscion>morganw: Hmm, I don't think guix-home should cause an issue in that case. I added sbcl and sbcl-esrap to the guix shell command above, and I was able to reproduce this error. Seems like a bug to me, but I'm still looking for where the source is. Might be worth making an issue, if You are willing to do so, on issues.guix.gnu.org <yewscion>IIUC, it seems like the inclusion of both SBCL and CL libs are causing SBCL libs to be prioritized, at least in this case. May be an issue with a load path? <nckx>civodul: How did you fix it? Structurally? <yewscion>The system-wide equivalent of ~/.config/common-lisp is /etc/common-lisp, and because asdf doesn't have the concept of which implementation it is running on, it loads those files even under clisp, which overrides the registry for asdf in clisp to be that of sbcl. <morganw>yewscion: Thanks for tracking down the problem. Do you think it is a guix issue or does asdf not provide any mechanism which would allow for more dynamic switching? <gnucode>nckx: I updated my opensmtpd-records patch to include texinfo documentation. I think the documentation looks somewhat good. <yewscion>morganw: I would test, but I don't have a machine with another OS handy. That said, I can't remember ever having an issue with this on other OSes, and I've always had both installed... Though of course, I was using quicklisp for that. <nckx>Thanks gnucode! I'm sure it looks great, I'll take a look when I have my PC back. <nckx>I'll try to port my own server to it. <yewscion>I would guess that, because of the generalities introduced by GNU Guix, this is a Guix-centric problem. In most other systems, You would point asdf to different locations based on Your lisp-specific package manager or Your own development process. <yewscion>It /should/ be fairly straightforward to have each implementation load an implementation-specific directory for their packages in an init file, though I've not done that. I used to specify, for instance, (:tree "Documents/") in my own home directory to add my own asd files to the system; I wonder if You could specify a specific one per each implementation? <nckx>unmatched-paren: I was kind of curious when I saw your 1st mail what you expected it to do. <unmatched-paren>nckx: I'm not sure why I expected aerc to download all the messages in the thread and display them (so that i could properly reply) when I replied to the thread <unmatched-paren>Now awaiting a reply in #aerc regarding how to import a mbox file... <nckx>Aerc must be magic if it can do the former :) <reyman>@nckx i repost the comment to say, yay, rm /var/lib/gdm solve the problem for me :) <nckx>reyman: Thanks! Sorry, again. <nckx>unmatched-paren: Well, I'm still not clear exactly what you expected to happen. Debbugs would send you backdated copies of all previous messages to any thread you reply to whilst unsubscribed? (If that sounds thick, that's just me, not you.) <nckx>It's technically 'possible', but.... <nckx>It's not NNTP, not that I've ever used NNTP. <nckx>I've never heard of a list that worked that way. <unmatched-paren>nckx: I should look into the internal workings of email. Yet another item for the Things ( Is Hoping To Look Into But Probably Never Will Bother To list :) <yewscion>morganw: Successfully worked around it with an init file for clisp. <gnucode>nckx: The code as is, probably needs to tweak it's styling to fit into Guix's style guide. Also, you can definitely port your server to use it, But the code doesn't really work for creating opensmtpd filters. at the moment. <nckx>There's no standard way for a mail client to fetch mbox archive from a list, although you could write a plugin/script that does it for mailman, or try to make all MLMs use standard URLs, ... That would still be a Web 'side channel' though, in se unrelated to any true e-mail protocols. <nckx>unmatched-paren: ^ musings <morganw>yewscion: would it be possible to upload the init file as an example? <yewscion>Yup, working on it now. Was reading info on compiling clisp with such an option already enabled, but perhaps that is for another day. <nckx>Not even plain strings like "/run/current-system/..."? <nckx>Because that's all I use now. <gnucode>In order to specify (opensmtpd-filter (name "dkim") (proc-exec (mixed-text-file "dkim-sign" opensmtpd-dkimsign "various options" "or command line flags")))) <gnucode>I need to get opensmtpd-filter-configuration to support gexps...which it does not. <gnucode>it might work for plain strings like "/run/current-system/..." let me check. It's been a while since I toyed with the code. <nckx>I think it should just take a file-like object, no? <nckx>Well, not in a position to review, I said, so I shouldn't try :) <gnucode>It definitely should just take a file like object. :) I'm just not certain how to write (define opensmtpd-filter-configuration->string ...) if the filter record contains a gexp. <gnucode>actually...I might have just had an epifany... <morganw>yewscion: That is great. Thank you very much for all of your help.