IRC channel logs
2026-05-14.log
back to list of logs
<bjc>janneke: i think i have a solution to our problems <bjc>it is a hack, but it should work <bjc>in fact, is working =) <bjc>and as a bonus, we have better log output now. no idea what was blocking that before but it's fine now <bjc>if you put that at the top of runsystem a lot of stuff goes pretty smoothly <bjc>i'll put this on the pr <dajole>Man, I kinda miss the comprehensive texlive package. Trying to compile an old .tex file and having to add two dozen texlive- packages gets old fast. <apteryx>how can I close an issue with emacs-forgejo? <apteryx>dajole: make sure to write a manifest.scm for it <apteryx>some guidance in (info "(guix) Using TeX and LaTeX") if you hadn't read it yet (tlmgr info something) <apteryx>csantosb: I saw your change about the bug-reference-mode in .dir-locals.el; do you know how to cause it to open in emacs-forgejo? <apteryx>I wonder if I could support both the old regexp for legacy debbugs issue and newer codeberg ones <apteryx>csantosb: actually I think it's not even necessary to setup bug-reference-bug-regexp and bug-reference-url-format anymore; there's implicit auto setup done if these are not set and codeberg is in the handled list, see (info "(emacs) Bug Reference") <apteryx>and evaluate (bug-reference--setup-from-vc-alist) to see how codeberg is handled by default <janneke>sneek: later tell bjc: you got tmpfs going, lovely! <apteryx>csantosb: actually I think we need it because the auto setup from vc will choose debbugs if your repo is hosted on savannah; and this can happen in ourcase <apteryx>ACTION is trying to figure out how to correctly configure bug-reference-setup-from-irc-alist <apteryx>OK, it works, but you need to clear any global bug-reference-url-format and bug-reference-bug-regexp for bug-reference auto-setup to kick in <thanosapollo>apteryx: try "h" to view all keybindings you've set. By default you use "x" to toggle close/open, I have a script to view forgejo links using forgejo.el but using jabber.el for now. (Not sure what the rest of the convo was) <thanosapollo>actually... having it as a custom part of jabber.el is not ideal, could add a custom hook for browse-url or smt similar to use forgejo.el globally for any buffer if it's one of the configured forgejo-hosts. Will investigate <apteryx>thanosapollo: hi! good to know about 'h' and 'x' to toggle close/open state. Perhaps we could use an already existing emacs-forgejo entry-point to jump to an issue? forgejo-issue-view looks like perhaps workable as a procedure to call from bug-reference <apteryx>debbugs uses debbugs-browse-url, which takes URL, extracts the bug number from it using debbugs-browse-url-regexp, and call debbugs-browse-function on the bug number <thanosapollo>apteryx: check forgejo-view-browse-url, I think what you want is wiring it up to browse-url dispatch <thanosapollo>should be simple, I'm not at home atm, but will def do that when I'm back <csantosb>apteryx: so all fine ? links work properly from scm and erc buffers for me <csantosb>Remains how to combine bug-reference and emacs-forgejo <csantosb>And maybe, how to replace M-x C-c RET by the usual bind <csantosb>I did something similar to you, this is why <thanosapollo-tes>apteryx: check latest on master, just set forgejo-browse-url-integration to t and bug-reference-setup-from-irc-alist <apteryx>great, you also reused forgejo-hosts, that's from where I had started drafting it <apteryx>i was just reading the code; will try it now <csantosb>I'm trying to disable it now, to fall back to previous behavior <csantosb>Not a big deal for me, I prefer that one by far, but the question will arise soon or later <apteryx>thanosapo11o: for debbugs, integration is done via the bug-reference-mode-hook or bug-reference-prog-mode-hook; maybe that could help with teardown <thanosapo11o>csantosb: you mean it should fall back to opening in the web browser or something using #'forgejo-view-browser-url-teardown <apteryx>there's an example in (info "(emacs) Bug Reference"), it looks like: (add-hook 'bug-reference-mode-hook 'debbugs-browse-mode); I guess that's nice in that if you disable bug-reference-mode, it turns off automatically <csantosb>thanosapo11o: yes, the flag just switches between opening in the user default browser, or redirecting to emacs-forgejo <thanosapo11o>wouldn't work properly with jabber.el though. I want to intercept the https:// as well for now. I will take a look closer at how debbugs work and add proper defcustoms for it <apteryx>and if you look at the implementation of debbugs-browse-mode, it hooks its browse-url function before 'browse-url-browser-function via `add-function' <apteryx>and debbugs-browse-mode is a minor-mode you can toggle on or off anywhere; might be handy to have also a minor forgejo-browse-mode that matches it <csantosb>The main drawback I see with emacs-forgejo: it is so good that people will start introducing forgejo:... links in comments 😀 <apteryx>(the debbugs-browse-mode is from the emacs-debbugs package) <apteryx>I guess whether yo want to offer an auto setup or document the hook to add, but debbugs does the later, and I guess that may be more idiomatic? <thanosapo11o>it sounds better, add-funciton :before-until on browse-url, remove-function when disabled. No global state mutation. Will look at how debbugs does it <csantosb>Enjoy: (add-hook 'magit-revision-mode #'bug-reference-mode) <apteryx>csantosb: thanks, I didn't have that one <apteryx>and then the readme or documentation can read like that in (info "(emacs) Bug Reference"), i.e. (add-hook 'bug-reference-mode-hook 'forgejo-browse-mode) and (add-hook 'bug-reference-prog-mode-hook 'forge-browse-mode), and it should just work? <csantosb>apteryx: The "c" for comments on changes seems fixed now too <thanosapo11o>wait... We can enable bug-reference per project, like ;; see #8597 and work with emacs-forgejo. Didn't know that <thanosapo11o>actually you do not need dir-locals.el since I already resolve this, just (add-hook 'prog-mode-hook #'bug-reference-prog-mode) and it just works. <apteryx>I thought this: ((options->transformation '((with-branch . "emacs-forge=master"))) emacs-forgejo) should give me the latest emacs-forgejo to put in my home config, but it seems not! <apteryx>not nice that it's silent in case it does nothing <efraim>try `((with-branch ... instead of '((with-branch <apteryx>I was using emacs-forge instead of emacs-forgejo <apteryx>now I could see 'retrieved commit 1257a3de5a45b696446930cb87523646a4295119' <apteryx>csantosb: ((options->transformation '((with-branch . "emacs-forgejo=master"))) emacs-forgejo) <apteryx>where the options-> thing is from (guix transformations) <efraim>"I applied the transformation and it did nothing" would be an improvement <apteryx>I'd expect some error; is there a valid use case for "do a transform that is a no-op, silently" ? :-) <efraim>I apply all the transformations to all my packages, I wouldn't actually want them all to give messages each time I reconfigured my home <efraim>I'm afraid to check my git log to see how many months that took me to work out correctly <apteryx>OK! I see; in this case yeah, perhaps a warning, or a flag to make it safe <efraim>we already made builds quieter for 'guix upgrade', it seems like it could fall into the same category <loquatdev>How do I get the name of a file-like object? Not inside a Gexp. <loquatdev>Say I have (define %some-file (plain-file "file.txt" "contents\n")), is there a way I could do (some-proc %some-file) to get "file.txt" as a string? <identity>loquatdev: try (plain-file-name %some-file) <adanska>loquatdev, if u look in guix/gexp.scm, u can see in the exports theres a procedure `plain-file-name`. I strongly suspect this is what you are looking for <identity>well, it takes approximately 5 minutes to see a message and find a solution <loquatdev>I'll expect 5 minute responses going forward ;) <loquatdev>I'm assuming if I were to pass a computed-file then I would need something like computed-file-name <loquatdev>I was thinking there would be some catch-all function for file-likes but after thinking about it that doesn't really make sense since not all file-likes have names until their derivation is build. <loquatdev>Thanks y'all, not sure where your timezones are but have a good night <apteryx>thanosapollo: was able to reconfigure my home with emacs-forgejo from master branch; and bug-reference integration seems to work fine, tested with my ERC snippet <thanosapollo>thanks apteryx, FWIW added bug-reference-mode support for jabber.el as well. <csantosb>Thanks to emacs-forgejo, I disabled all codeberg email notifications. This is a huge improvement. <apteryx>ACTION returns to their forgejo-service-type <thanosapollo>do we have any guix members in the forgejo development team? <csantosb>What do you mean by forgejo devel team ? <hwpplayer1>As far as I can remember I was able to run Firefox or other standard apps inside it but it started with scratch buffer <apteryx>hm, depending on mysql is not enough to have it ready by the time a user uses it <adanska>loquatdev, none of that dynamic dispatch nonsense in my guile please! <adanska>though man i really miss CLOS... common lisp is amazing <identity>hwpplayer1: not sure what exactly you are asking, but you can start graphical programs (or any programs, really) with M-& <hwpplayer1>It was a small screen display and I couldn't see the line at the bottom <identity>hwpplayer1: oh, i think that was fixed in a later version; it is probably because you set the font size *after* activating ‘exwm-mode’ <sham1>Hm, apparently emacs-forgejo doesn't like my TOTP setup on Codeberg <sham1>I'm currently writing a ticket fwiw <csantosb>hwpplayer1: once you have your install, open a term M-x ansi-term, guix pull, reconfigure your home system, restart <hwpplayer1>I am running guix pull, then how to reconfigure my home system ? <hwpplayer1>If Hurd is stable than can we say only GNU but not GNU/Guix/Hurd ? <apteryx>I think a GNU Hurd system could be called just a GNU system, instead of GNU/Linux, yes. <tusharhero-xmpp>> This document describes GNU Guix version 0.0-git, a functional package <hwpplayer1>I think it should be GNU the official package manager GNU toolchain and Hurd the official kernel <hwpplayer1>package manager Guix kernel Hurd and GNU toolchain all they are official GNU projects <hwpplayer1>in the meanwhile I prefer AGPLv3-or-later for my code <apteryx>it's a good choice yes, especially if your software is designed to be used remotely via a server <csantosb>hwpplayer1: There is an issue with Guix 1.5 and Emacs/EXWM, this is why you need to guix pull newest commits, where the problem is fixed <apteryx>did you provide the config file at the end? <csantosb>sudo guix system reconfigure /etc/config.scm <hwpplayer1>sudo guix system reconfigure /etc/config.scm running <csantosb>This is it, you reconfigure your whole system, not only your home configuration <csantosb>I had to go through the same as you, so I just remember the pain first time <csantosb>Once you reboot, things work as expected <fanquake>I'm trying to use a shell like "guix shell --load-path=/path/to/dir --manifest=/path/to/dir/manifest.scm --manifest=/path/to/dir/manifest2.scm" <fanquake>This works fine if I execute it from in /path/to/dir (--load-path can also be —load-path=.) <fanquake>but it doesn't work from anywhere else, just throws "hint: Did you forget `(use-modules (blah))' or `#:use-module (blah)'?" <fanquake>Feel like I'm missing something obvious here <Duongph2098>Hello guys. How do I create a bridge interface br0, assign a static IP address to it, and set my main ethernet interface (say, eth0) as it's slave? (The guix way) <bjc>if you're using a desktop install, you'd configure it through network manager <Duongph2098>bjc: Yes I'm using a desktop install, but I'm trying to do with static-networking-service-type <bjc>i don't see any straightforward way of doing it, though i may be missing something <Duongph2098>bjc: Do you have a working configuration? Can you share please? <bjc>not in guix, no, sorry <bjc>i thought i had this set up on old systems but my configs for that don't have it either <bjc>i think i probably did it through network manager, but it was a long time ago <Duongph2098>Correct me if I'm wrong: doing this through network manager means using nmcli commands, right? <apteryx>Duongph54: not very well integrated yet I think; you can write one shot shepherd service that run the commands you need at boot <efraim>bdunahu: sorry, I didn't check the bug tracker before pushing my update <sneek>Welcome back efraim, you have 1 message! <efraim>i'll take a look at double-checking the URL in a bit, right now I'm fighting a python/rust package <bjc>Duongph54: there is ‘network-manager-configuration’ which should let you do it declarativele <apteryx>oh nice, that's new enough that I didn't know about it :-) <apteryx>is it complete enough to set up a bridge though? <Duongph54>However it doesn't mention the part about setting the real network interface as the bridge's slave <Duongph54>Because in my current setup, I need to set my real ethernet interface as the bridge's slave for it to work <bjc>apteryx: it has ‘extra-configuration-files’, so i'd assume so <apteryx>I TInk this may be the first time I see this one: suspicious ownership or permission on `/gnu/store/n9lgdziv85mdzsqckdkcc0r0fj07dphw-mysql-9.7.0'; rejecting this build output <cdegroot>Is anyone aware of a Guix config for Medley Interlisp? I have a feeling it's not a simple thing to build but I really want to be able to pull and build the latest and greatest :) <apteryx>actually I think I've seen that one in torcs; there's a 'fix-permission phase there to rectify that <ieure>cdegroot, Not sure what you mean by "config." I found a channel with some packages for it, but they're outdated and need work to build against the current version (Interlisp changed their build system since the packages were made). <ieure>I forget where I found it, though, and don't see it at a glance. Somewhere on Codeberg. <ieure>cdegroot, The main Interlisp package doesn't build, but the underlying VM does, though not correctly, which is the source of the downstream build failure. <ieure>cdegroot, They have this weird setup where there are three different binaries for the VM. One with X11 support, one with SDL support, and a special one for initializing the loadup (which is X11 only). The downstream package is failing because it's not using the special-for-loadup VM binary. <ieure>cdegroot, The other packages I found lumped all three binaries into the same package, which I personally do not care for. I think the correct fix here is to have the init be either a hidden package used for the loadup phase of the Interlisp build, or a secondary output of the maiko package. <efraim>can someone restart the substitutes service on bordeaux? <cdegroot>ieure: I don't know what the default word for a Guix package definition scheme program is, I guess :) <cdegroot>Anyway, thanks for the extra info - wasn't aware about the package split. I really like the idea (big Smalltalk'80 fan, same concept) but the version I used years ago was very limited w.r.t. screen resolution (resulting in a bad screenshot in my book lol) so I'm hoping that's fixed now and I'd like to install it "natively". So I guess three VMs it is with some AdvAdvGuix for ensuring the loadup script gets that one version but the <cdegroot>package pulls in X11-or-SDL (coin toss, I guess?). <cdegroot>(if it's anything like Smalltalk'80, all it needs is a raster device/frame buffer to draw on so I don't think you'll be able to spot the difference) <ieure>cdegroot, The init VM is only called to build the final Lisp image, it has to be a native-input of the interlisp package, but should not end up in the package closure. Might have to donk around with the build to make sure no references leak, shouldn't be too big a deal. <ieure>cdegroot, What I don't have a good solution for is having the image be functional without a hard dependency on one of the SDL/X11 implementations, without having redundant image builds for each. <ieure>I'm kind of inclined to only package the SDL version, since it *seems* like it should work on X11, Wayland, and raw Linux framebuffer. <cdegroot>I mean, I don't care for the latter two (you will have to pry X11 out of my cold, dead, hands ;-)) but SDL should give the best bang-for-the-buck. <ieure>But I haven't run either, so I don't know if it works as well or not. I'm kind of assuming the SDL stuff is a modern addition, and the actual legacy VM code targeted X11. <ieure>cdegroot, We are in violent agreement. Ability to run a weird old Lisp GUI directly on the framebuffer has a certain appeal to me, though. <cdegroot>I do not disagree :). There was a hack called "SqueakNOS" that booted Squeak straight on top of the metal and it was quite neat (very nice to have your TCP stack in your application code, including debugger and whatnot) <yelninei>janneke: glibc 2.43 is looking a lot better wrt gnulib problems <cdegroot>I'll try to see whether I can take your code across the finish line, thanks for the leg up (well, multiple legs looking at the snippet ;-)) <yelninei>so far it is one xfailed gnulib test that now passes, a few coreutils tests newly hang, but this might be solved with the coreutils update and/or making sure the /hurd is also on glibc 2.43, but nothing like lake year <yelninei>ill try to build a little more just to be sure <PotentialUser-18>Is there a way to have a multiline `substitute*`? Just checking before I implement it for my usecase. <mwette>PotentialUser-18: For that, I have been working on coding somethign like ed (the unix line editor) in Guile for that purpose. What is your approach? <dajole>`guix locate` often times misses direct hits and I often find myself having to use `find /gnu/store -name...` directly. I read the manual section and tried to repopulated locate's db, but that hasn't changed things. Am I missing something? <dajole>I'm also aware of the glob flag. So e.g. `guix locate -g '*kde-agent*'` doesn't find anything, but `find /gnu/store -name '*kde-agent*'` finds tons of matches. <bjc>how hard would it be, i wonder, to get a guile that required no devices <bjc>a bootstrap guile that can talk to mach ports <janneke>bjc: regtur is thinking/working(?) on a guile that talks to mach ports <bjc>it'll be nice to write translators in guile <PotentialUser-18>mwette : I worded it weirdly - I just need to remove stuff from a small file, and my current idea is very simple: read&write the file with `(ice-9 textual-ports)` and match with `(ice-9 regex)`. Not gonna implement something crazy, but maybe it would be cool to have an option for `substitute*`to work on the whole file if need be - I just sadly <graywolf>Hi :) Is there a way to guix pack a dynamic list of packages? guix pack -m requires a file, in guix pack -e the expression must evaluate to a single package (list of packages nor manifest work) <graywolf>Best idea I have is to write a temporary manifest into /tmp with the content I need, which... works but was hoping for more elegant solution <kaelyn>graywolf: I don't know if this would work within the context of your use case, but you should be able to specify multiple packages in a single "guix pack ..." command line, both by name and through multiple "-e" expressions. <loquatdev>How can I debug `static-networking-service-type`? It's failing. The system reconfigures just find but I get `service networking could not be started`. Checking the status with herd yields no information or logs. <loquatdev>I verbatim copied the example from the manual and swapped out the ips for my own. <graywolf>Is anything useful in /var/log/messages ? <loquatdev>graywolf: There's a scheme backtrace... ` 1. &netlink-response-error: 17` <mwette>PotentialUser-18: got it - I'll post a note if/when I get mine (mostly) working <janneke>bjc: interestingly, also all kinds of messages come through with your tmpfs, nice <bjc>janneke: yeah, i don't have a perfect explanation for that yet <cmiller>hmmm, i am installing guix 1.5.0 and it spams looking for substitutes on bordeaux 0.0%, though I have seen a try with ci with 100%. why is it still trying bordeaux <bjc>possibly for days i think? <cmiller>I am just freshly installing the system. Why doesn't it timeout and choose ci? <ieure>cmiller, Guix currently lacks the ability to notice that a substitute server is unresponsive. <cmiller>ieure: Ahhhhh. Okay make sense. I assumed naturally that this was the point of having multiple endpoints. <bjc>on build or reconfigure or whatever. so it doesn't check bordeaux constantly <cmiller>bjc: Currently the 1.5.0 installer is running. How would I stop it it and change it? <bjc>from the installer? oof. i don't know <ieure>cmiller, The two servers point to different build setups which can have disjoint substitutes, so it's more for substitute coverage than fallback. <cmiller>ieure: Ah I see. Now everything makes sense <ieure>It *should* handle these issues way better. But it currently doesn't. <bjc>i don't know what's going on that's taking it down for so long, but i can't imagine it's good <cmiller>I think it fixed itself. It now says "no route to host" for bordeaux and I can see all the packages it is downloading from ci. Though it is slow <bjc>janneke: i can use the same trick to get /servers/socket/1 working before fsck btw <bjc>which means guile can control fsck <bjc>guile can mount dev. that's very exciting, indeed <bjc>i'm going camping tomorrow. i wonder how much i can get done before then