<apteryx>rekado nckx '/dev/sdk2 100T 129G 100T 1% /' on berlin *apteryx sighs of satisfactoin <parakesh>i just wanted to download a build image of Guix on a Hurd basis, but all the build images return a 502 bad gateway error <parakesh>the linux based images work, but not the hurd ones <parakesh>unmatched-paren: thank you for the hint, at least im not alone :) <parakesh>does it look like a "could be fixed any minute" thing or a "check again in a few days" thing? because that determines the rest of my day :P <parakesh>Aye. Fair. then i will check again in a bunch. <civodul>apteryx: yay, that's a pleasant thing to see, well done! <nckx>sharlatan: Could you check (and fix :) your connection? <gnucode>well I've got my server installed with guix, but I cannt seem to be able to update it. <gnucode>reconfigure is failing with invalid cert. <gnucode>does guix 1.3 have problems updating to the latest version ? <gnucode>hmm. guix system describe says that I am on generation 2 from Jul 21 2022.... <shcv[m]>I found an "and/l" helper in the web.scm file I was using as a reference for a service I'm working on. It looks like it tests a variable, and if truthy constructs a list with the value bound to '<>. Is there something like that in a utility library, or should I just copy its definition? <shcv[m]>gnucode: I had cert problems once due to the system clock being too far off <gnucode>I actually fixed it just now by doing a # guix pull. <gnucode>shcv[m] I have just copied definitins before. <gnucode>I know for a fact that fold is defined in several different places in the guix source code. :) <gnucode>but it's a 5 line definition. who cares? <gnucode>also does guix home manager let you initialize git repos ? <shcv[m]>No idea. It would be nice to have some helpers for things like that though <shcv[m]>I'm starting to work on moving my configs to guix home, so I'll probably be working on things like that. Doom emacs will need the git checkout feature you're describing I think <apteryx>sneek: later tell civodul thanks! hats off to rekado for the iDRAC access, which proved its usefulness here, and to nckx, who fixed the bootloader and completed the migration, sparring us a great deal of downtime <muradm>congrats with issues.guix.gnu.org is back! <muradm>however it seems there is some issue with timing? i just sent a reply to 56859 for instance, but issues.guix.gnu.org shows that it happened "40 minutes ago".. may be it is for me only?.. <muradm>ah it uses time of composing, not the time of received ***rgherdt_ is now known as rgherdt
<sudognu>hii, I'm trying to install guix on my machine but am having problems with system init <sudognu>...-lonux-libre-gnu.tar.xz.drv failed to produce output for ...-linux-libre-...-gnu tar.xz <emacsomancer[m]>Any chance of someone updating the mu(4e) package to the 1.8 version? (I'm using a janky local build, but it's, well, janky.) <abrenon>sudognu: any useful log message you could paste and share ? <sneek>civodul, you have 1 message! <sneek>civodul, apteryx says: thanks! hats off to rekado for the iDRAC access, which proved its usefulness here, and to nckx, who fixed the bootloader and completed the migration, sparring us a great deal of downtime <abrenon>sudognu: also, is your system trying to compile linux-libre ? have you disabled substitutes ? <sudognu>abrenon: seems like it couldn't download the libre kernel <abrenon>ahhh, so perhaps temporarily no substitutes ? <abrenon>that's too bad, I wanted to update today ! : ) <sudognu>I erased my artix partition for this "-w- <abrenon>long ago I erased an artix to put guix too, and I lived to tell the tale ; ) <sudognu>oh also, it tried to download linux-libre from many places, not just guix ci <sudognu>I think moving back a few commits could do the trick <civodul>"guix weather linux-libre" is green for me, with a commit from yesterday <abrenon>mine is a bit old so I'm time-machining my way to check for the latest commit <civodul>sudognu: what does "guix describe" report? <abrenon>ahh, perhaps an old install medium for each there are no substitutes anymore <abrenon>I can confirm that the weather is sunny too with the very latest commit (I git pulled to obtain the commit's hash right before checking) <sudognu>i installed from 1.3.0 first but then used the weekly iso <abrenon>ok, so not that, but what does "guix describe" report ? <sudognu>might be still better than gentoo x3 <sudognu>abrenon: I'm still checking the weather <abrenon>Oo is your network bad ? it's not supposed to take very long: "guix weather linux-libre" <sudognu>I'm checking the entire weather "-w- <sudognu>I redirect guix.gnu.org to tropin's ip, and it doesn't work perfectly <sudognu>but when I set the url to ci.guix.trop.in it works correctly. <abrenon>well that's certainly good to know in any case, thanks for checking this use-case : ) <sudognu>abrenon: do you think it's good enough for a bug report? <abrenon>I'm not sure a system can be assumed to work in a context where DNS is bypassed to lie about the resolution <abrenon>I mean, I also think it's a cool trick and I resort to it time and again, but when it doesn't help, I consider it my responsability to understand why it didn't help rather than assuming the service I'm trying to tweak should work <abrenon>it's still weird that with the redirection, weather is able to find the substitutes, but system install isn't though <sudognu>abrenon: still, its' inconsistency is weird... <abrenon>but there are many parameters that could be out of guix: perhaps he simply publishes several sites at the same IP, and, with the wrong hostname, the webserver redirects you to a different virtual host where there are no substitutes ? <abrenon>yeah, if you nslookup the URL into an IP and query that, you reach a clone of the guix.gnu.org site <abrenon>in any case, there is a bug in guix, there's quite some work remaining to do to show that despite that dirty hack it's still supposed to work <civodul>rekado_: it might be of interest to you (i think you were trying to use those ports in a non-blocking context) <sudognu>civodul: I didn't use tls, just plain html <rekado_>civodul: thanks, this will come in handy! <civodul>i wonder why openconnect propagates gnutls & co. <civodul>efraim: do you happen to remember the rationale for 0d7f282b0295f9b2edcc960d25d53e69c81f08dd ? <mroh>Love, Peace & Guix! Good morning! ***wielaard is now known as mjw
<Cairn>Or at least the initial git pulling part. <Cairn>Ugh. "build of /gnu/store/*-pkill9-free.drv failed" <civodul>Cairn: hi! that must point to a build log, and it's probably an issue in this specific channel <sudognu>Linux takes long to build, how do I make get a custom kernel here? <Cairn>Oops. -there seems to be a small issue in the channel. Just frustrating, since I'm only setting it up to try something out. <Cairn>Little problems at every step of the process. <pkill9_>kabouik: hello, I'm returning to guix so I will fix my repository and fhs compatibility tool, which I actually want to turn itno a script that "installs" it rather than in the system config so you aren't forced to rebuild all the fhs stuff when you reconfigure <pkill9_>kabouik: I haven't updated that repository in atleast a year so no surprise it doesn't build, also doesn't bu8ildf or me ***pkill9_ is now known as pkill9
<ncbfg36>what is the syntax for setting kernel arguments in config.scm? It's not entirely clear to me from the docs. Something like (operating-system (kernel-arguments "nosmp")) gives me errors <abrenon>and what does the errors say ? isn't a list expected given the plural form of "arguments" ? <ncbfg36>wrong type argument in position 1 (expecting empty list): "nosmp" <mroh>ncbfg36: try (kernel-arguments '("nosmp")) <ncbfg36>apologies I have no networking on laptop at the moment irc is on my phone <abrenon>isn't the message suggesting an *empty* list weird though ? why does it say an empty list and not just a list ? <ncbfg36>mroh: that did the trick. It was expecting the preceding "'" before the kernel args in parens <abrenon>well congrats ! you've just learn a syntax shortcut to make a list <apteryx>civodul`: hello! any thoughts on the newly introduced use of *unspecified*? It introduced some breakage because of not being a serializable type. <apteryx>(in the context of services when using 'define-configuration') <abrenon>well there's the additional property of stopping evaluation but I was coming to that <unmatched-paren>such a shortcut is implemented in some other lisps like janet and clojure, but not lisp <mroh>abrenon: maybe it says empty list, because a list is at least an empty list? idk. <ncbfg36>I was trying (kernel-arguments (list ("nosmp"))) but that didn't work <abrenon>no, I mean things generally work for any data of a given type, requiring a particular value is really upsetting <abrenon>ncbfg36: which is why it is important to understand that quoting / evaluation thing <abrenon>in guile, arguments to a function are passed directly after the name of the function, within the (…) <abrenon>list is just a function which returns a list containing all its arguments <abrenon>by default, when guile sees a list, it will try to evaluate it: its first element is expected to be the function, and the others its arguments <unmatched-paren>Procedures can be thought of as accepting a single list containing the arguments <abrenon>so when you write (a b c d), you're creating a very short-lived list containing a, b, c, and d, but which guile is going to reduce immediately by trying to apply a on the arguments b, c and d <abrenon>should "a" not be a function, you're doomed : ) <unmatched-paren>the `read` procedure, used on a port, reads a single s-expression from the port and returns its quoted form <abrenon>this is why a function such as "list" is required, in this environment where pretty much everything is a list <ncbfg36>oh okay. So long form would be (kernel-arguments (list "nosmp"))? <abrenon>but it has an additional subtlety ! the long form with (list …) lets guile evaluate the code inside <abrenon>so if "nosmp" was in a variable named, say, kernel-arg <unmatched-paren>"everything is a list" is a pretty good design because it's both elegant and allows macros to be implemented very easily <unmatched-paren>classical macros (which are created with define-macro in Guile) are like procedures that return ASTs <abrenon>because this quote simply stops evaluation <abrenon>so '(kernel-arg) is a list, alright, but it doesn't contain "nosmp", it contain a symbol, which is a kind of ultra-cool token-thing in LISP, it's just an abstract named concept 'kernel-arg <abrenon>it's still different from a string, you can think of it as a kind of infinite on-demand enum type <ncbfg36>oh i see. so because everything is a list even if i'm using a single argument i need to declare it as a list with a single entity <abrenon>ncbfg36: in this case I think it's rather that the type expected at this place in the declaration is a list <pkill9>does anyone here primarily use guix as a package manager on archlinux/gentoo/slackware? and if so do you have any problems with the applications integrating with the rest of the system? <unmatched-paren>(symbol->string "foo bar baz") allows you to create symbols that normally would not be possible <abrenon>it could handle simple strings, but this has apparently not been implemented <ncbfg36>as a strange side effect of setting "nosmp" it now shows gnu graphics while booting <ncbfg36>haha yes that also isn't what i had intended. I just wanted to turn off hyperthreading <mroh>maybe worth a bug report. <ncbfg36>I was under the impression that this is what "nosmp" does. Not disable all but a single core <muradm>works for me locally, however there are few ancient emacs-xyz packages which probably may break at runtime <muradm>new mu brings significant performance improvement and responsiveness <nckx>ncbfg36: Linux displays a Tux (patched to a gnu + Freedo in linux-libre) per on-line CPU, I guess you have too many otherwise :) <rekado_>muradm: I’ll take a look at the patch. <nckx>muradm: Oh thanks! Mu's dog slow here. <rekado_>muradm: do you really need coreupdates as inputs? <muradm>rekado_: coreutils you mean? yes they '/bin/rm' command into source, and to properly map it, i needed input <muradm>if there is better way to depend on /bin/rm that can be changed ofcourse, any suggestion? <muradm>nckx: keep in mind you need re-init :) <rekado_>you’ve got two instances of patching /bin/rm <rekado_>once with (which "rm") and then again with search-input-file <muradm>rekado_: yes one for tests, ad-hoc (which "rm") as it was before <muradm>rekado_: actually you may be right, it is used in TempDir destructor only, and it is used in tests only <muradm>so probably that code will never run at runtime <muradm>i can switch to (which "rm") i think ***civodul` is now known as civodul
<nckx>muradm: I do that weekly now, so it can't get worse. Not sure if placebo but that seems to help (for about said week). <rekado_>not sure you even need coreutils as an input, though. *nckx says hullo to everyone, then leaves. <civodul>apteryx: hi! do you have pointers re breakage due to *unspecified*? <civodul>it feels like that change was made long ago <apteryx>civodul: the only thing that I've found failing so far is the jami-service-type, whe the jami-account objects have fields taking the *unspecified* value (was not exposed by the jami-provisioning system test). I added coverage for that situation in https://issues.guix.gnu.org/56799#16. <two[m]>unmatched-paren: sorry, i meant i want to use it in a shell <unmatched-paren>If you want your Guix package to be reviewed and merged quickly, pretend it's an Emacs plugin ;) <rekado_>substitute* with a string is equivalent to substitute* with a single-atom list <rekado_>I actually have coreutils in the inputs again and use search-input-file for the run-time reference to /bin/rm <rekado_>by default the build environment has coreutils – but is it a native input? <rekado_>when cross-building mu we want to keep a reference to the target architecture’s /bin/rm <muradm>rekado_: last update have no coreutils and two files to patch for /bin/rm (this is why list is needed) <two[m]>unmatched-paren: getting the same error <unmatched-paren>two[m]: this looks more like a cmake-related problem than a guix-related <two[m]>/gnu/store/cxxn5zy4vfaknlvpfg4k3q4vygrdsik7-profile is not my guix environment <unmatched-paren>if it isn't recognising the compiler as a compiler, there's probably something wrong <two[m]>s/guix environment/$GUIX_ENVIRONMENT/ <two[m]>there isn't a compiler there, it's looking in the wrong folder <muradm>rekado_: you are right at build time rm is there already, however it is not used at runtime, only for tests, so no need to specify it as input <two[m]>CMAKE_PREFIX_PATH is /gnu/store/cw89h7ibafmyd5gwdfg54mz1zx3g8b7m-profile/ in env <kabouik>Hey pkill9. Thanks for your answer, and sorry for having pinged you a million times in that discussion, I didn't mean to. Great to know that you'll update it. Again, as I said, I don't fully understand what fhs is used for yet (I guess it helps with software that expects the usual GNU/Linux filesystem?). I have already removed it and your channel from my system.scm and reconfigured, but if you make it into a script and it doesn't <kabouik> need to be into the system.scm anymore, that'll be great. <kabouik>Still very interested into making Appimages (at least one appimage) work on Guix, so I'm still following your repository anyway. Appimages are one thing on my Guix to-do.org created yesterday (my first org-mode file. :')). <gnucode>civodul: looks like a pretty nasty bug! <PurpleSym>Are there any pointers for fixing Qt applications failing to build during 'qt-wrap? <muradm>civodul: i don't observe accumulations of sshd on my servers <civodul>muradm: and you're running shepherd 0.9, without a firewall protecting port 22? <rekado_>muradm: applied with commit 7e3234cb709076c8f4514f8b9320a0585dd9c958 <PurpleSym>unmatched-paren: Ah, okay. The traceback is different, but I’ll try to `guix pull`. <muradm>civodul: it says "i'm 0.9.1, but i don't listen on 22" :) <PurpleSym>unmatched-paren: No, I tried twinkle and lyx so far. Both are broken. <muradm>civodul: we need fail2ban and fail2ban-service-type ;) <civodul>muradm: heh, not listening on 22 is a good idea, but it means your machine is less likely to be hammered :-) <unmatched-paren>PurpleSym: You should probably send that information to the issue, rename it, and reopen it <civodul>muradm: oh, thanks for the heads-up! <abrenon>unmatched-paren: PurpleSym, yeah it seems nheko is affected too <apteryx>hmm, did I break all qt5 packages inadvertently? ^^' <apteryx>it seems to be because there are no 'qtbase' argument passed to the wrap phase <apteryx>so (package-name->name+version (strip-store-file-name qtbase)) fails in qt-utils.scm, line 148 *apteryx looks at twinkle <apteryx>some idea: perhaps affecting packages setting arguments explicitly. <apteryx>perhaps the default #:qtbase goes away for some reason <dgcampea>What's the proper way to add an extra 'origin' object to the inputs of a package? If I read the docs correctly, <dgcampea>'modify-inputs' only concerns about 'package' objects and '10.7 Invoking guix style' only talks about 'package' objects as well. <dgcampea>Also on the same topic, '9.2.1 package Reference' has a note that the old input style that associates a label to the object is deprecated but, <dgcampea>supposing I have 2 extra 'origin' git objects 'A' and 'B', both with a 'dir/foo.pp' file, how can I refer to 'A's 'dir/foo.pp' in particular? <Cairn>pkill9_: Looking forward to your repo working again. It'll be useful to me. <apteryx>adding #:qtbase ,qtbase-5 to twinkle build arguments fixes the build... so that confirms the hypothesis: all packages explicitly setting arguments do not have an implicit qtbase argument and hence the wrap fails fails. hm <kabouik>So if I want to build third-party software from git, like https://github.com/d99kris/nmail, what is the best way in Guix? Ultimately I'd be happy to make a package but that'd be only for a selection of software. For those things I found on reddit.com/r/commandline, I would prefer just trying them and compiling them for myself with less efforts than a scm. <kabouik>I understood that guix environment --ad-hoc is a good way to build stuff without cluttering my system, but as far as I understand it requires a package name <muradm>kabouik: guix environment <pgk-name> --ad-hoc will drop you into build environment of <pkg-name>, guix environment --ad-hoc <pkg-name> will drop you into empty environment with <pkg-name> available for user <kabouik>So I'm a little lost as to how to build something and then using `make install` when this probably requires using /usr/bin or `~/.local/bin` <gnucode>unmatched-paren the afternoon there? hmm. You must be accross the pond. <kabouik>But <pkg-name> doesn't have to exist in a channel in the first place? <muradm>kabouik: normally you don't "make install" under guix, what you do is you prepare a package in some or another way, and then you can use it via system profile, some other profile or in default profile or else <muradm>kabouik: no, it does not, you can create your own locally <muradm>or you may have your own channel <unmatched-paren>kabouik: not only you don't "make install", most of the time you *can't* <muradm>unmatched-paren: sudo make install will :D <kabouik>Hum, but then if we take the nmail example above, it requires libxapian, libetpan, cyrus-sasl which may or may not be available as prexisting Guix packages, and if they are not, that means I have to build them too. Without appending a `make install` step after their compilation, nmail won't find them, so I guess I would need to make packages for all dependencies in the first place <unmatched-paren>It'll probably also break if you update a dependency of the `make install`ed package <muradm>kabouik: then you recursively track all dependencies and make packages for them <muradm>and then use them as an either inputs/native-inputs/propagated-inputs <kabouik>That's what I feared (I mean I understand why it's like that, but with my skill level, it'll make it really hard and engaging to try third-party software; and at the moment I use a lot of them) <kabouik>Yes the dependencies here are not so unusual <muradm>kabouik: not so hard, practice with few packages, and you get used to it <kabouik>But stil having to package things even for just trying them makes it more difficult than I was hoping <kabouik>When dependencies are already packaged <muradm>you have hundreds of examples already to copy paste from :) <kabouik>I even wrote one myself muradm some months ago but I can't even understand how I came up with it now, months later. :p <kabouik>And still, writing a .scm chunk is not exactly trivial even for those who know what they're doing, compared to just executing commands from a readme.md <kabouik>I mean I kind of understand what I did specific to nnn, but what I don't manage to understand/remember is the actual Scheme parts: use-modules, could never remember the inputs and build-system, etc. <unmatched-paren>also: (define-module (...)) (use-modules ...) should be replaced with (define-module (...) #:use-module (...) #:use-module (...) ...) <kabouik>But that is my lack of Lisp knowledge, I don't know what # or ' or : do for instance. In use-modules I don't understand why we sometimes pull gnu and some others guix, and then what is `gnu packages`, and then why we use `gnu packages ncurses` and not just `gnu ncurses`. This is all due to not mastering lisp I believe. <kabouik>So I do get that the learning curve for someone like me will be steep, and if I manage to hold on, I will learn a lot; but going through packaging and writing my own .scm file for every git thing I like, with all the trials and errors and long `guix pull` every time will be quite demanding <podiki[m]>the module naming scheme follows the file structure <podiki[m]>so (gnu packges ncurses) since in the guix src: gnu/packages/ncurses.scm <kabouik>(Can't do it right now, my guix machine is at home and I realized I can't ssh into it from WAN) <podiki[m]>or just guix build -f file.scm (assuming all you need is in the file and the last line is just the package name to build) <unmatched-paren>Basically, the gnu/guix split is the split between "guix the program" (guix) and "the default guix channel" (gnu) <unmatched-paren>-L . is far better than -f FILE.scm, because it allows you to modify multiple files without pulling <kabouik>Is it permanent (as in adding a private channel in my channels.scm), or temporary and ignored for future guix pulls? <podiki[m]>but for just a package with new deps in one file, it works (I haven't had to use -L for example, in all the packaging I've done) <jackhill>why does `make clean-go` report that ./guix/build/po.go is stale, but then it gets rebuilt with another make? <kabouik>The gnu/guix split is still not clear to me. In the package I linked, I don't understand lines 3 to 7. Are these guix subcommands? <efraim>civodul: openconnect probably propagates gnutls & others because they're listed in Requires.private <kabouik>This is why I'd be curious to know if -L. is temporary, as opposed to the private channel in channels.scm which affects each guix pull <unmatched-paren>e.g. the build systems, communication with the daemon, "compilation" of packages to derivations <unmatched-paren>e.g. it defines all the packages, services, etc using the stuff from guix <kabouik>I don't know how one can come up with knowing they'd need a (guix utils) in the package.scm for example <podiki[m]>you don't have to, you try something, guix will complain about a missing module, you add it :-) <podiki[m]>most of the time I let the errors guide me, unless I know for sure where something comes from <kabouik>I see; that might already be what I did there, plus inspiration from others' packages <podiki[m]>or, you could look up (with seaching, geiser, etc.) where something is defined and then add the module so you have it <podiki[m]>yup, errors plus lots of searching/copying/pasting is a good way to learn <unmatched-paren>also, since this is a package derived from nnn, you should use (inherit nnn) <kabouik>Yeah actually I used the branch there unmatched-paren because I wanted the package to reflect every future commit (hence the -git name of the package). I knew there was a hack to skip the hash verification, so that was the plan, but I got away from my Guix machine before figuring it out <podiki[m]>don't forget there are package transformations, you can just do e.g. 'guix install nnn --with-git-url=nnn=/path/to/git --with-branch=branchname' if it will just build without modifications, you don't need a new package <kabouik>Well then I need a nmail.scm (might be in reach for me), then a mullvadvpn.scm (much harder as it uses systemd and I already can't master it, much less Shepherd; and it's non-free so I won't discuss it here), then appimages, then .scm files for a million other CLI tools I use here and there. I have a lot on my plate. :> <kabouik>Thanks for the detailed explanations unmatched-paren by the way <kabouik>Good to know podiki[m], that seems useful too for things that are already packaged (I could fork them and apply my patches, git push, then guix install from my git url) <kabouik>So basically `guix install nnn --with-patch=/path/to/my/no-easy-delete.patch` would do? <podiki[m]>if you don't need to actually change a build process but modify sources/patches, guix transformations is really useful <apteryx>OK, I have a fix for #:qtbase, but perhaps we could instead keep everything implicit and use search-input-file on inputs/build-inputs. <apteryx>that'd be better for rewriting business <kabouik>What happens if I install nnn from the current guix channel with my patch (--with-patch) and then in two months guix pull and guix package -u? <kabouik>Forgive me if I misunderstood but I didn't get why you mentioned the nmail License in the first place (we were not chatting about the license); it's just a naive question <kabouik>No problem, at least I understand now :p <unmatched-paren>There's probably some way you can connect to it with a free application though <podiki[m]>kabouik: I think it would be replaced with the regular version? I'm not sure, I use manifests (see manual/cookbook) and there you can include the transformation so it will be used when you update with the manifest <rekado_>FWIW you don’t need to package anything to build it from source when using Guix. <unmatched-paren>(Also, a nonfree VPN is probably more damaging to your privacy than no VPN, but I digress.) <rekado_>just use ‘guix shell …’ to enter an environment containing gcc-toolchain, pkg-config, and whatever you need to build the thing, then run ‘make’ as you would elsewhere. <kabouik>Yes, with openvpn configs most likely, but their client is convenient for server hopping and avoiding georestrictions (more than having 40 different openvpn configs) and for renewing a subscription (it uses a no-account system, so you have a unique code, and without the client I'd have to save it somewhere and copy-paste it in the payment form, while the client eases that up) <kabouik>(which is convenient since I usually renew my subscription from a phone or from my HTPC which has no keyboard) <apteryx>jackhill: I've sent some email about fixing the collaterals from introducing #:qtbase <apteryx>the fix is simple, but I'd like someone to have a quick look to avoid a cascade of fixes :-) <jackhill>apteryx: ok, I'll be on the lookout. I can confirm the qutebrowser fix works <apteryx>yep, but there were many others with a different problem (not label related but default argument value related) <jackhill>ah, ok, I guess I don't have that many Qt applications installed to haven noticed on my own <kabouik>Oh, so with `guix shell` I could just compile all the binaries I need and then copy them somewhere else? If their function (not compilation) depends on something that is already packaged, I can install those packages out of the guix shell. If those deps are not packaged, I'd need to build them too though, and then I'm not sure how to make them communicate together <kabouik>My guix machine is not here and I have no space left on my current laptop to install just the guix package manager, so I can't play with it now <unmatched-paren>Be aware that if the libraries the binary needs are updated, it'll probably break <unmatched-paren>I was just looking through Qt packages and seeing whether they build <gnucode>So I just mounted a usb stick to ~/usb. mount /dev/sdb1 ~/usb. Now it says that ~/usb is owned by root. When I try to chown ~/usb to my regular user, it is NOT letting me. Why? That seems very odd. <unmatched-paren>heavily annotated and amended version of your package that does not use the latest git master <unmatched-paren>to turn this into the git version, you could just replace (commit ...) with (commit "HEAD") <jackhill>apteryx: I don't have a feeling about whether using the explicit #:qtbase arguement is better or not <apteryx>I guess for now it'll do. If it becomes a problem we can revisit. <apteryx>any other qt5 packages that were broken to try outside of twinkle? <jackhill>I'm trying to run screen from a relocatably tarball pack on AlmaLinux 9 host, but it error with "getpwuid() can't identify your account!" Ideas? <mroh>apteryx: Im trying some qt apps I use. So far mumble and kiwix-desktop fails in wrap phase, some fix as for qutebrowser seems to work. <jackhill>apteryx: yes, produced with `guix pack -f tarball -RR -S /bin=bin screen` <apteryx>mroh: OK. There are two distinct issues. One is the label change from "qtwebengine-5" to "qtwebengine". The other is Qt packages lacking an implicit #:qtbase when specifying args themselves. For the later, I have a fix at the level of qt-build-system.scm <efraim>the 'qt-wrap phase failed for me on a custom qt-build-system package <apteryx>I'm testing another package then I'll push it <apteryx>efraim: pushed (4905b5b83904366d068bde899aae15288cc1adcb) <efraim>apteryx: thanks. also the #:qtbase thing worked ***Dynom_ is now known as Guest7652
<PurpleSym>apteryx: I didn’t realize you were Maxim. Anyway, thanks for the 'qt-wrap fix! <apteryx>PurpleSym: eh, no worries. Fixing what I break is the least I can do :-) <yuu[m]>hey y'all can link me your nice guix configs, ideally those with btrfs filesystem <apteryx>mroh: any remaining Qt packages to fix (of the earlier label problem type) ? <apteryx>the rdv-jami-guix rendezvous point is back online, available for anyone in the Guix community to use (it had been down for a rather long while -- post shepherd 0.9 and *unspecified* issues :-)). <yuu[m]>efraim: apteryx unmatched-paren thanks!! <minikN>Hi. I'm using mixed-text-file to save a shell script to the store, I then retrieve its path with (file-append my-file). But how can I make it executable? <jab>hey guix, so I have my guix server up and running at home. I'll find out in about 8 hours, if I can ssh into it when I am not at home. <unmatched-paren>which is like mixed-text-file, except the gexp contains instructions to create the file, instead of returning a string <unmatched-paren>(computed-file "foo.sh" #~(begin (call-with-output-file #$output (lambda (file) ...)))) <minikN>So this is similar to what I would use in add-before, add-after when writing package definitions <unmatched-paren>i use this to strip comments from waybar.json before putting it in the store <unmatched-paren>i used to have one for transpiling my neovim config to lua using `fennel`, but i dropped that in favour of emacs+evil <minikN>thanks unmatched-paren, although I should feel like this should have been documented better. <minikN>Yeah. It's "there". But that's it ;) ***mark_ is now known as mjw
<jackhill>apteryx: hmmm, is there a qtwayland for qt6, or is the separate package no longer needed? <apteryx>not every qt6 packages have been added yet, so it's probably something to do <mroh>apteryx: the qt packages I use build with your fix. Thank you very much! <sudognu>hmm, can I contribute to substitutes somehow? <sudognu>and can it contribute to ci.guix.gnu.org? <sudognu>cool that there are substitute servers and repos hidden around the web... <jab>unmatched-paren: do you have the server code for tobias' substitue server? <jab>That would be pretty cool to set up a guix service for such a thing to make it really easy for people to set up substitute servers... <jab>unmatched-paren: gotcha. :) <jab>maybe I could write a cookbook entry for how to set up your own substitute server... <jab>nckx: why does your substitute server say that chromium is malware? What about ungoogled-chromium? <sudognu>civodul: aight x3. someday I'll make a public repo with substitutes... <blake2b>its actually quite easy, if you have a linode running guix its pretty simple <jonsger[m]>jab: I could send you the config of my substitution server via private chat if you are interested :) <jab>jonsger[m]: that would be welcome! <jab>blake2b: I do have a linode running guix... ***theruran_ is now known as theruran
***kir0ul4 is now known as kir0ul
***f1reflyylmao is now known as f1refly
***alanz_ is now known as alanz
***eichin_ is now known as eichin
***dmj`_ is now known as dmj`
<jab>I guess I didn't realize that there was a guix-publish-service-type.... ***clever_ is now known as clever