<jgart[m]>nckhexen: Is what GNUtoo requesting possible through the debbugs control codes? <jgart[m]>Otherwise most people will never learn them <nckhexen>jgart[m]: Should be, with ‘tags NNN patch’. <nckhexen>If you want to waste money on an .io domain, it's your call, but if people need that to RTFM I have little faith in the result. <jgart[m]>I have unlimited domains at SourceHut pages <jgart[m]>I need that website. I'm making it for myself <jgart[m]>We need something flashy for the control codes <jgart[m]>Maybe a js editor with preview of what would happen if you send a particular debbugs control code <johnabs[m]>Okay, guix emacs issue: apparently I need to rebuild the epdfinfo program, but it get the error message: can <johnabs[m]>can't exec aclocal, no such file or directory. But autoconf is installed <johnabs[m]>Oh, I thought it was autoconf, that's what it complained about first, I'll try that, thanks! <jgart[m]>does geiser work bug free for anyone else? <TopExpert>unmatched-paren: actually there was a way to remove gcc from the gnu-build-system, it has an #:implicity-inputs? flag that doesn't add it <the_tubular>Any words on 1.4 ? I want to automate AMI creation and wonder if I should wait for 1.4 to be released or there's still some bugs to iron out ? <TopExpert>the_tubular: you should be able to automate it in guile, then just return an operating system definition at the end of the (((()))) <TopExpert>I doubt guix as a command line interface to aws cloud is in the works <TopExpert>but it's just guile, so it can do anything in the configs, including spawn AWS instances <TopExpert>it should support packaging in OCI (Open Container Format), that is interoperable with most clouds and can be uploaded as an AMI <TopExpert>mastering guile seems like a good idea; it is very easy to spawn cloud machines in a shell script so Guile should be nicer than shell <johnabs[m]>unmatched-paren: Actually, I found the emacs-pdf-tools guix package, and that worked just fine, but thanks for your help! <xd1le>antipode, just thought it was funny ***jesopo is now known as jess-o-lantern
<pkill9>mbakke: where is your website again? <sektor[m]>How do we feel about turning the speakup module on fr all kernels? I just realized that it may be easier to do that than to write an accessibility kernel for each version. <Cairn>I'm doing a bit of yak shaving over here. Should I make a package suffixed "-6" in order to not rename a qt-5 package? Or should I make the qt-6 version without the suffix and rename all instances of the other package to include "-5"? <apteryx>yep. Past commits usually include the commands used to do the renaming automatically, probably helpful here :-) <Cairn>I instinctively press tab when typing part of a name into a command like `guix build`. Takes a long while, so normally I C-c out of it. The C-c doesn't escape like I'd expect: my cursor ends up on the far-left column and I need to C-c again for it to redraw the prompt. Anyone else have this issue? <apteryx>or more easily from debbugs-gnu-bug 57931 RET -> C -> patch RET <Cairn>I'm using a nearly unconfigured bash. Should be easy to reproduce. <apteryx>seems a single C-c should give us a prompt <apteryx>our completion needs a good overhaul to be usable, in my opinion <Cairn>I don't know much about CLI development. Does that sound like an easy fix I should look for? Or should I just submit a bug report? <Cairn>Guix's CLI is... minimal. A feature I'd love to have is terminal resizing during, say, a `guix pull`. <apteryx>if you fix it, that's even better, and you mention the issue closed in the commit message. <Cairn>Once I shrink the window past a certain point, it stops redrawing the same line's percentage and starts just spitting out new lines over and over. <Cairn>It'd be nice for there to be an adaptable progress bar <Cairn>Anyway, yeah, thanks. I'll submit the bug. <Cairn>I'll bug report that too for good measure *apteryx has alias b='./pre-inst-env guix build' for when developing <Cairn>Sorry, what's the pre-inst-env part do? <apteryx>I typically want to build packages I just modified in my tree <apteryx>the pre-inst-env script setups the environment variables to use the guix from your tree <Cairn>I've just been setting GUIX_PACKAGE_PATH and writing temporary definitions in separate files. <apteryx>ah. I recommend working from the tree <apteryx>you'll have all the other packages available as examples <Cairn>Oh, I keep the guix repo open as well to use as examples. <Cairn>I just ripgrep from within emacs <Cairn>But you're right, I should look back into working from the tree <Cairn>I'll read that section of the manual once this yak is fully shaved <apteryx>there's not much to it, really. you git clone, build it, then run guix as usual with ./pre-inst-env in front *apteryx is interested in adding language packs to libreoffice <Cairn>Oh are they not packaged yet? <Cairn>Seems like a pretty important task. I understand why you'd want to add them <kenran>Hey everyone! I'm looking forward to trying out Guix (the package manager only, at first), and specifically want to check out `guix home'. I've followed https://guix.gnu.org/manual/en/html_node/Installation.html and tried executing `guix home`, but I'm getting "guix: home: command not found". Is this a version issue, or do I have to enable it first maybe? <Franciman>oh i thought guix home was only available in guix system <kenran>I'm in a meeting in 2 minutes, going to write back after! thanks! (I've found an older thread mentioning having to `guix pull` before) <kenran>but guix pull fails: guix pull: error: derivation `/nix/store/dfh81c62m0lgmr4fsp9jb58c7zxx3xgs-guile-3.0.2.tar.xz.drv' has incorrect output `/gnu/store/b1sljij4vihpr5s485i25y7ljzba6j25-guile-3.0.2.tar.xz', should be `/nix/store/962l91hdpscd01p9rdsa8cizlcfs97rx-guile-3.0.2.tar.xz' <unmatched-paren>kenran: um, are you trying to use guix with the nix-daemon or something? <kenran>hum, that's indeed strange. I'm on NixOS, maybe I misconfigured the guix daemon. I'll investigate in an hour :) <unmatched-paren>Hmm, if I were to add a (local-file "foo.text" ...) to a package's inputs, would the label to refer to it be "foo.text"? <AwesomeAdam54321>How do I refer to a specific package output for an input in a package definition? <brendyn>also equivalent is (list glib "bin") but the first style is idiomatic <sneek>Yey! unmatched-paren is back <Kabouik>Anyone successfully using Jitsi with Icecat? <pkill9>i think hardwarr acceleration doesnt work for me in icecat <kenran>unmatched-paren: ok I'm back: since the first issue (guix home command not found) _might_ potentially be solved by using `guix pull`, I'll try getting that to work first. but I'm on NixOS and don't know much about how the daemon there work. Could it be that guix and nix share a socket or something? <fiesh>Kabouik: you need some special trickery, alas I'm not at my guix laptop right now. if someone has a backlog of this channel, I explained it a little while ago to someone, but I don't remember it exactly <fiesh>Kabouik: iirc you need an addon that sets the client agent to something jitsi recognizes and allows (Firefox) <kenran>as for what I've done to get that output above: I've started `guix` as `/var/.../guix` (no symlink yet), and so far I've started the guix-daemon with `~root/.config/guix/current/bin/guix-daemon --build-users-group=guixbuild` for the time being. I don't suppose any of this is an issue, but it might well be!? ***wielaard is now known as mjw
<Kabouik>Heh that worked fiesh, thanks a lot. I was expecting some deeper issue. <jpoiret>civodul: I'm in the process of writing a bug report for the grafting issue from yesterday <civodul>jpoiret: oh that was a grafting issue, i hadn't fully grasped the context <civodul>time for "make clean-go && make", comrades! <abrenon>(I hate when it doesn't build and I eventually have to type this without properly understanding why) <civodul>it's better than the alternative, which is to not get a proper error in case of ABI breakage and instead have things break randomly <rekado>I wonder how this affects (license #f) or custom licenses defined in channels. <kraai>unmatched-paren: Thank you for your help with packaging doctl; I've just submitted the patch. <mroh>rekado: yeah, (license #f) doesn't work anymore on my channel. <futurile>I need some help with profiles. I'm setting up a new machine, and have put some packages into manifest files (e.g `core-apps.scm` and `XWindows-apps.scm`. When I run `guix package --manifest core-apps.scm` guix removes some manual apps I've installed (e.g something installed with guix install ssh). Is there a way to tell Guix to ADD these apps to the profile, and not remove things already installed? <abrenon>mroh: ohhh : ( any idea how to replace this ? <futurile>For various reasons I want to keep the list of packages to install in separate files. But I want them all added to the default profile. <pkill9>futurile: I think this is possible <pkill9>you could possibly have somethign int he manifest that reads what is in the profile and add those packages to the manifest, if they aren't already written in the manifest <abrenon>futurile: that's the expected behaviour, the -m option tells guix package to have *exactly these packages*, it's not a "loop on add" <pkill9>does the guix command allow multiple --manifest options? <pkill9>I think you wanna generate a new manifest containing additional packages from the guix profile <pkill9>there's different ways of doing htis I think <futurile>yeah - I see that's the expected behaviour. Is there a way to 'add appliations to a profile by reading from a file' :-). I mean I suppose I can do something ugly like shell redirection, but that seems broken. <pkill9>futurile: are you willing to use a separate script to achieve it? <futurile>pkill9: guess so - seems bizarre that you can't add multiple apps to a profile from multiple files <abrenon>I don't think you need shell redirections, a simple for loop in your favourite shell would do the trick <rekado>a manifest is used to declare the exact state of a profile <pkill9>futurile: you can specify multiple manifests on the command line <rekado>if you want to combine multiple manifests you can concatenate them <rekado>or you can build different profiles and set their search paths <abrenon>yeah, as long as you need imperative features over declarative ones, it's natural that the shell would be the appropriate tool <pkill9>so all you need is a script with `guix package --export-manifest > /tmp/profile-manifest.scm ; guix package -m /tmp/profile-manifest.scm -m <new-manifest>' <futurile>rekado: because I think it's a fairly common use-case, to "add a list of apps to my system" <pkill9>note that it will update all existing packages too thp <futurile>abrenon: sure get your point - though the cli is an imperative system - anyway, not trying to upset anyone, just thought it was something that people would do with a package manager. I'll go the script route. <abrenon>not with *this* package manager : ) but I think we've all reassured you that a little scripting was perfectly ok here <abrenon>florhizome[m]: there's no proper concept of uninstallation since this would be inherently imperative <rekado>you just get a new profile generation according to your specifications <brendyn>whats the point of this let (let ((x obj)) (and (pair? x) (every license? x)))? <florhizome[m]>Actually i wanted to have a command to remove a set of packages from a manifests file <florhizome[m]><florhizome[m]> "My patch Just doesn't appear..." <- I sent it to the debbugs address via git-sendemail. Twice now. 🤔 <unmatched-paren>sometimes issues.guix.gnu.org can take a while to pick up on new patchsets <unmatched-paren>but an ack will tell you the bug number, and you can go to issues.guix.gnu.org/NNNNN without waiting for it to appear on the front page <rekado>issues.* syncs everything from debbugs continuously <rekado>if it’s on debbugs it will be on issues.* within a minute or two. <PotentialUser-28>hey everyone, methinks that `cl-trivial-custom-debugger` was removed from `lisp-xyz` quite recently. <antipode>Look up the reasons for why it was removed. <antipode>If the reason was simply 'there are no dependents', you can re-add it. <antipode>(and preferably submit your new package definition as well) <antipode>brendyn: I assume to check that 'x' is actually a list of licenses <antipode>(it technically checks for a pair, not a list, but possibly usually close enough) <florhizome[m]><unmatched-paren> "florhizome: have you not..." <- After the first Mail, yes <brendyn>antipode, but why let x instead of just refer to obj <antipode>brendyn: To not evaluate 'x' multiple times. <antipode>... OTOH there is (eq? #f obj) etc above ... <nashdidan[m]><florhizome[m]> "I experienced this, too" <- I experienced this, too <nashdidan[m]>I'm using guix home to manage everyting that's not part of the bas system install, so when I need to update I run guix home reconfigure ~/.config/guix/home.scm which does not remove anything installed manually. I add changes I want to keep eventuall to home.scm. <antipode>(remember, it's a macro, not a procedure). <antipode>Looks like some small optimisations are possible ... <nckhexen>florhizome[m]: It will eventually appear. Don't keep sending dupes. Debbugs isn't clever enough to spot those. <rekado>I wonder if we could optionally bypass “Compute guix derivation…” on “guix pull” by looking for a recent build of the guix derivations on ci.guix.gnu.org. <nckhexen>lechner: I don't see your question answered, but apologies if dupe (hah): the daemon never runs as your user. If that means you're not running the daemon, you're probably safe, but why risk it... <PotentialUser-28>it seems that my issue has to do with how i had two different versions of guix installed, one from `guix install guix` <PotentialUser-28>when i would run `guix pull` then `guix describe` the one from the install script would be update to a newer commit. <nckhexen>florhizome[m]: I wish I could say that's unheard of... <rekado>PotentialUser-28: you should never do “guix install guix” <dgcampea>how can I confirm or check that certbot is being automatically executed? <nckhexen>rekado: I've come to a point in my life (mellow with old age?) that I think we should just (if (and (string=? command "install") (string=? package "guix")) (error "no.")) rather than keep hand-wringing about the ‘right’ way to fix it. Would that have any chance of being accepted? <nckhexen>(if (and foot gun) refuse) is just sensible. <rekado>we can even just mark the package as hidden <rekado>just needs to be carefully checked that this doesn’t break anything <rekado>(“guix” would no longer be found with fold-packages, I think, nor with specification->package) <florhizome[m]><nckhexen> "florhizome: I wish I could say..." <- Well this explains a lot :/ quite uncomfortable :// <nckhexen>rekado: Would that not break ‘guix shell guix’? <nckhexen>* -D guix, all this time and I still haven't internalised newthink. <nckhexen>But, I would also not want to break ‘guix shell guix’, that's veering away from ‘obviously never correct’ again. <nckhexen>However, reluctance shared. It took me a while to see the light. <lechner>dgcampea: aside from inspecting file change times and certificate expirations, you could perhaps check recent renewals with certbot certificates <kenran>I managed to fix my NixOS/Guix problem, but I don't actually know what fixed it. I reinstalled twice, now manage guix-daemon.service and gnu-socket.mount via my NixOS configuration, and rebootet. All of that might have something to do with it :) <kenran>After the `guix pull` I can also use `guix home` now, so time for experimenting! <brendyn>dgcampea, i mentioned something similar years ago <xd1le>for the record I'm using Guix in NixOS and doing something similar, I mostly just copy pasted my config from somewhere <xd1le>kenran, ^ there it is if helps you or anyone. Seems to work for me. <xd1le>I prefer Guix System but still using NixOS as a remnant of maintaining some nix packages I guess. <xd1le>people have tried to package Guix for Nix but I guess there issues <GNUtoo>apteryx: thanks, I'll look into it <GNUtoo>nckhexen: jgart[m]: thanks too for the infos <GNUtoo>Thanks, I now have the manual to also close bugs, rename them, etc <rekado>civodul: I’m looking at your license predicate thing. It’s fun! <rekado>I always found recursive syntax definitions intimidating, but the pattern is not uncommon. <rekado>I wonder if we could express this recursion in a higher level macro; this is like some sort of syntax fold, transforming the input syntax into neat little buckets that are processed in the final clause of the macro. <sneek>Welcome back civodul, you have 1 message! <sneek>civodul, dsmith-work says: One of the problems of writev (IMHO), is how to deal with incomplete writes. You might need to modify an iov entry. A bit icky. <civodul>there's a partial implementation of it in Guile <sektor[m]>Now to go learn how to write Shepherd services. <jorge[m]1>Another option to make Guix pull that accumulates my download? This is because I have a very very slow connection. <unmatched-paren>sektor[m]: you'll want both a record or configuration and a variable bound to a (service-type) record <unmatched-paren>your service-type needs a (name ...), (description ...), and (extensions ...) field <unmatched-paren>the first is a symbol, the second is a string, and the third is a list of service extensions <unmatched-paren>using service-extensions, you can add more files to the dotfiles service, add packages to the profile service, and add services to shepherd <unmatched-paren>so, you can write (service-extension shepherd-service-type SERVICES) <unmatched-paren>these accept a (documentation ...) field (string), a (provision ...) field (list of symbols that are provided by the service), <unmatched-paren>(start ...), a gexp that is used to run the service, e.g. a (make-forkexec-constructor)-created object <unmatched-paren>and (stop ...), another gexp used to stop the service running, e.g. a (make-kill-constructor) <unmatched-paren>sektor[m]: have a look at gnu/services/mcron.scm (scroll to the end and read upwards) for a fairly simple example <gnucode>hey guix, I am getting closer to having an opensmtpd-service with records patch v2. I am considering adding in some tests. Would it be lame to have the tests check for sanity on the configuration? For example I have code that makes sure people don't use conflicting options, and warm the user (and error out) if they try. <unmatched-paren>though i'm not sure whether they'll work if you use them in a service's code... <gabber>i'm trying to cross-compile an operating-system definition for aarch64-linux-gnu. i get an `gnu/build/linux-modules.scm:257:5: kernel module not found "ahci"`. what am i missing? i am trying to get going with a minimalistically simple configuration <f3n1x>ahò . I've installed the emacs-guix package in my Guix system. Yet, when in emacs i go for instance invoking the 'guix-installed-packages-...' command i'm getting Starting Guix REPL ... No prompt found! . What am i missing ? <f3n1x>(let me add that a Guile REPL prompt is opened in another parallel buffer, though ^^ ) <unmatched-paren>f3n1x: emacs-guix is a little buggy, as i'm sure you'll have noticed <f3n1x>umh... i confirm ! unmatched-paren . What a pitty ... as long as it looks like a very enjoyable tool ? May i report the issue somewhere ? <gabber>unrelated question: is it possible that Guix's GDM won't start unless there's an active network link? <lechner>gabber: gdm (or elogin) definitely mess do something with my network after twenty minutes or so, but i have not seen an issue without a connection. (on my first boot, for example, i did not have network i do not think.) anything is possible, though <gabber>i'm just curious -- i'm currently fiddling wich a machine's networking setup (which doesn't fully work yet). so booting ends in a console prompt. as soon as i `sudo dhclient <interface>` the screen switches to the usual GDM login screen <gabber>anyhow -- is there an example for dhcp-client-service type floating around? i have included it (via modify-services %desktop-services) but the interface doesn't seem to actually try to obtain a lease. <lechner>gabber: gdm can take a few minutes to start. are you just witnessing that delay? <gabber>lechner: it's possible -- though this is on a very powerful and fast machine (12th gen i9) -- i haven't noticed similarly long delays in my slower setups. maybe it's a coincidence, but as soon as the machine has a valid IP gdm starts <lechner>gabber: hi, is 127.0.0.1 not a valid IP? <gabber>lechner: sorry, i meant something like "address-resolving uplink" ;) <lechner>gabber: i have to say sorry for being cheeky <gabber>np, i guess i should take a break (: <xd1le>lechner, are you sure you mentioned the correct timestamp? <lechner>xd1le: sorry, it was 16:30 i think, and thank you! <sneek>cbaines, you have 1 message! <sneek>cbaines, unmatched-paren says: sorry for the long delay, but I've finally sent a rebased and slightly amended version of the greetd-wlgreet-sway-session patch to its issue :) Thanks for reviewing it! <xd1le>lechner, ah yes he means psp ***mark_ is now known as mjw
***Dynom_ is now known as Guest3203
<unmatched-paren>Ugh, I was gonna make a gammastep home service, but it turns out it doesn't play nice with my display... <sektor[m]>unmatched-paren: will have a play with this once I can get my qemu issues sorted. <sektor[m]>In reference to the service tips you posted earlier. <sektor[m]>Kinda hard to write an espeakup service for a system that isn't guix. <Grimpper>Please, could anyone that knows Guile check this error and give me a hint on what could be the problem? Thanks <the_tubular>TopExpert, Sorry for the very late answer, but how would I do this ? From the docs I read, AMI needs a VMK image made from Vmware, am I wrong on that ? <Grimpper>For context it hapens when autocompleting a guix command on Eshell <Grimpper>I see. And what do you reckon, that import should happen at guile package level or at the emacs-guix package? <Grimpper>sad. It's a really cool package. And I got it to somewhat work pinning old versions of it's dependencies but this one I cannot figure out <GNUtoo>daviid: for my database issue, I've tried to add C.UTF-8 and I failed, but I also read in the libc manual that the C/POSIX local should be there on all computers. After updating Guix it worked though, so it was probably a bug or something like that. <jpoiret>GNUtoo: this locale doesn't exist yet in our glibc version iirc <jpoiret>the C.UTF-8 locale was only added very recently <daviid>GNUtoo: ok,but i think i did suggest C.utf8, did you try tht? <daviid>though again, not an expert (at all) on these matters, i think it was recommended somewhere ... <daviid>then that is weird, if your guix env 'locale -a' has it, postgress should accept it (?) *GNUtoo retries just to be sure <GNUtoo>Well it works, it was just to report what made it work *GNUtoo is also curious when he helps people on how things get solved at the end <daviid>GNUtoo: my very old postgres exp is that you need PGCLIENTENCODING and PGDATESTYLE set and exported (if te later date style differs from te encodin <GNUtoo>building locale 'C.utf-8 => [error] cannot open locale definition file `C': No such file or directory <daviid>GNUtoo: to complete, i was building postgres manually at the time, and had and still have these vars set and exported: <GNUtoo>building locale 'C.utf8'... => [error] cannot open locale definition file `C': No such file or directory <GNUtoo>But the C locale is supposed to already be there anyway <GNUtoo>And it works without trying to add a C UTF-8 locale <daviid>ok, i thoyght the C locale is ascii only <GNUtoo>On Parabola I managed to generate a C UTF-8 though <GNUtoo>That stuff is strange, anyway it works now <daviid>yeah, ok, i'll leave this to guix experts .. *GNUtoo will just leave it as-is since it works <GNUtoo>I'll try to fix other issues instead, like try to see if i686 works fine without the patches I use <ric342[m]>Hi what is guix equivalent of buildFHSUserEnv <Cairn>I don't think there is one yet <Cairn>IIRC it's a work in progress <ric342[m]>Is there any other way to run programs that need to see dependencies in normal layout? <Cairn>ric342: I've run into this trouble as well. I'm not sure. <ric342[m]>and other things that are probably not specifically packaged for guix <ric342[m]>They appear to be using buildFHSUserEnv in their shell.nix for openwrt <ric342[m]>Along with the dependencies, which seems easy <Cairn>Have tried writing it yet? Might just need a patching phase or two to get around the FHS thing <Cairn>Could have just been easier to use it on Nix, but not impossible to go without <ric342[m]>I may have to use on nix until I am good enough with guix <ric342[m]>It probably isn't too hard to do on guix but considering I know very little about guix it would probably take me days haha <Cairn>Totally understand. I also feel that way, although I'm just choosing to waste those days 😅 <Cairn>I'm attempting to update qtserialport to work as a qt 6.3.1 package. It has a phase to set an absolute path for udev so that it can understand guix paths. I can't seem to find any issue with this that would have been broken by updating, but I'm still getting a cmake error that libudev can't be found. <Cairn>Anyone with qt experience happen to be around? I'll post to guix help if not <two[m]>it it possible to retry pull without recomputing derivation? <two[m]>hm i see `--keep-failed` is available <Cairn>How's it failing? What's it say? <two[m]>Cairn: TLS error in procedure 'write_to_session_record_port': Error in the push function. <two[m]>this is only produces by guix pull, but most of the time <Cairn>Does it mention a specific package or anything like that? <two[m]><two[m]> "hm i see `--keep-failed` is..." <- no, that doesn't work... <Cairn>How long has it been happening? <two[m]>Cairn: Computing Guix derivation for 'x86_64-linux'... / <two[m]>substitute: updating substitutes from 'https://bordeaux.guix.gnu.org'... 0.0%guix substitute: error: TLS error in procedure 'write_to_session_record_port': Error in the push function. <two[m]>guix pull: error: `/gnu/store/...-guix-command substitute' died unexpectedly <Cairn>Can you try using only ci.guix.gnu.org for just one guix pull? <Cairn>Probably unrelated: just curious <florhizome[m]>nckxhexen turns out I am getting a TLS error: „TLS is required, but was not offered by host debbugs.gnu.org“ when sending my patch ***jackhill is now known as KM4MBG
***KM4MBG is now known as jackhill
<florhizome[m]>I tried to turn off the smtpencryption in my gitconfig but that didn’t do it <two[m]><Cairn> "Can you try using only ci.guix...." <- worked! may be bordeaux having bad tls, or just one of the cases where it's lucky <two[m]>thank, you, will do the arg again next time i update guix <Cairn>You might be lucky not to have to use the flag eventually. Maybe just for a little while <nckhexen>(There's only one x, I just happened to see your message by chance, florhizome[m]) <cbaines>I don't know of anything wrong with the bordeaux.guix.gnu.org TLS setup, that error could just be a networking thing <cbaines>unfortunately it doesn't really mean anything to me <nckhexen>florhizome[m]: <I tried to turn off the smtpencryption in my gitconfig> Uh… why is your Git talking directly to the mail server? Are you sure it is? That would be unusual, at least. <lechner>Hi, what is the threshold for recognizing custom substitute servers, please? Is it running 'pull' twice or reconfiguring the system twice, or a combination thereof? <nckhexen>It should not be necessary to do anything twice. <nckhexen>florhizome[m]: But to the actual issue: even in 2022, requiring TLS from all receiving SMTP servers is caveat emptor, and you will get bitten. <nckhexen>lechner: I can only guess at what the real issue is, but are you restarting the daemon in between? That's who actually decides which server(s) to query. <lechner>nckhexen: does 'guix pull' restart the daemon? <lechner>Also, I get "Connection refused" when querying port 8080 of my custom substitute server even though the publish job is running in shepherd <nckhexen>lechner: No. It does not touch the daemon so there is no need. <nckhexen>Guix pull has nothing to do with applying configuration changes. It fetches updates to Guix. <lechner>nckhexen: restarting the daemon did something. now i get something on localhost, although on port 80 (and not 8080). i also get nothing from the outside on either port https://paste.debian.net/1256577/ <lechner>nckhexen: i think i just have to adjust my guix-publish-configuration <morganw>Does anyone happen to be using stumpwm on guix? When I am trying to configure it there appears to be some functions which are missing entirely. i.e. format-with-on-click-id. I'm wondering whether that is a packaging error or whether I'm doing something wrong. <lechner>morganw: i use stumpwm but i'm not a very sophisticated user <lechner>nckhexen: yeah, that always help, although your advice to restart the daemon was probably indispensable <morganw>lechner: Technically it works for me until I tried to add something to the mode-line. It looks like the built image was created without loading all of the source files. <lechner>morganw: i also see "unbound" in the lisp eval <lechner>nckhexen: yay, it's working! i just had to drop the 's' from the https due to a lack of local certificates <morganw>lechner: Thanks for checking. It isn't just me then... <nckhexen>Pretty sure guix publish doesn't do HTTPS. <lechner>jpoiret: regarding your laptop remark from september 29, i have 'guix publish' working <lechner>nckhexen: some of the substitute servers do! <Andronikos>It is possible to disable spectre migrations in Linux. Would you recommend that? <lechner>nckhexen: i have the evil one listed with https <nckhexen>Do you mean nonguix? Please don't call other projects evil. I doubt that means they are running a guix publish with https support/patches. Probably an nginx or other proxy... <nckhexen>(Call them evil here, I mean :-). Do as you will elsewhere!) <two[m]><Andronikos> "It is possible to disable..." <- тщ <two[m]>> <@Andronikos:libera.chat> It is possible to disable spectre migrations in Linux. Would you recommend that? <mroh>Andronikos: try "mitigations=off" as kernel boot param. <unmatched-paren>nckhexen: isn't that risky? or is it only risky for servers or something? <florhizome[m]><nckhexen> "Who is actuall producing that..." <- see above ^^ <nckhexen>Git doesn't deal with any of that, nor can it control or affect any of it. <nckhexen>Mandatory TLS and Spectre mitigations are a soothing balm for the paranoid. <two[m]>may be fine if you're only running git on the machine, but they may also run browsers <florhizome[m]><nckhexen> "Git doesn't deal with any of..." <- git talks to msmtp which talks to the provider (about the setup) <florhizome[m]>I don’t really want to change the provider setup for this… but it seems I may need to? <nckhexen><git talks to msmtp which talks to the provider> Yep. <nckhexen>So only the provider can change that policy, or you can change your provider, but little else. <florhizome[m]>what is weird to me, is that sending normal mail is fine, but sending a patch via git send–email doesn’t work <florhizome[m]>also I’m pretty tired, if I’m not replying any more I fell asleep