<ArneBab>Is there a way to get Gnome to open *.jar files with java -jar by default? <nckx>Guix uses this mechanism to launch ‘foreign’ binaries through Qemu, see qemu-binfmt-configuration. You could adapt that to your needs. <spudpnds>Hello folks. When downloading content from ci.guix.gnu.org, sometimes it's fast, and I get like 2Mb/s, and sometimes it's really slow, and it's like 35Kb/s. Anyone know why there is such a difference? <nobodynogroup>NieDzejkob: I did yesterday, but i have done a lot of stuff since then including restarting the device so im doing it again <nckx>nobodynogroup: It sounds like ~/.config/guix/current isn't in your $PATH. If it isn't you can pull until you're blue in the face without making progress. <nckx>nobodynogroup: How did you install Guix? <nckx>If you used the installer script, it should have installed /etc/profile.d/guix.sh and sourcing /etc/profile should set the correct variables. <nobodynogroup>nckx: yeah i used the script, i just added path and infopath, i will try sourcing if this pull doesnt work <spudpnds>Does the guix "hint" about GUIX_PROFILE happen every time you run guix package -i? <spudpnds>I'm starting to think I've done something wrong, but I've like 3x checked, and I haven't. <ArneBab>nckx: I have a hard time seeing how this could be done quickly <nckx>spudpnds: It should be printed only when installing the package(s) added new search paths. Not every possible time. <ArneBab>ok :-) — I’ll delay that then. I need a way to launch a jar, just to show that it’s easy, but if it isn’t then using the terminal is easier. <nckx>ArneBab: I don't know what intimidates you most, but a quick test should be as easy as echoing ':ExecutableJAR:E::jar::/home/arne/jarwrapper:' to /proc/sys/fs/binfmt_misc/register, where ~/jarwrapper just basically runs java -jar "$@" (probably with some env vars to boot, but I don't know anything about Java). <nckx>s/intimidates/seems most tedious/ 🙂 <ArneBab>I hit that when checking whether running the freenet installer would work with doubleclick <ArneBab>right now it’s bash: /proc/sys/fs/binfmt_misc/register: No such file or directory <nckx>ArneBab: Did you load the module && mount binfmt_misc -t binfmt_misc /proc/sys/fs/binfmt_misc ? <seepel>Hi guix, I recently spent some time fixing the g-golf guile library package. I'd like to contribute a patch, but have a couple of questions. 1) g-golf only builds for Guile 2, should the package name reflect that? 2) Both the library, as well as the tests depend on guile-lib. Does that mean that guile-lib should be duplicated in both native-inputs and inputs? <ArneBab>no, I din’t — that‘s part of what feels tedous when you just want to do a quick check <nckx>Actually, it's part of %pseudo-file-system-types, so it should be mounted already. <nckx>ArneBab: I agree, but that's Linux for you. Yet another thing that systemd now handles on many distroes because the kernel offers mechanism, not policy… <nckx>(…and nobody else stepped up.) <daviid>seepel: please don't change the (GNU) G-Golf package name in guix, thanks <seepel>daviid: No problem, I see that some packages seem to and some don't, so I wasn't sure. <spudpnds>I'm having a hard time understanding why "guix package -i stow" is downloading mesa. <nckx>daviid: Does something rely on the name? <spudpnds>Is there a way to output a list of dependencies for a package. It seems crazy to me that a simple program like stow would need gtk+ and mesa, annd harfbuzz, and pango... <nckx>spudpnds: guix graph --path <package a> <package b> <nckx>You're right, there is no dependency on mesa. <nckx>(Nor on the other packages you mention.) <nckx>They're not being installed for stow. <spudpnds>Ok, maybe it just happened when I asked for stow to be installed. <spudpnds>I did install emacs-next just before this. <nckx>That depends on everything you mentioned above, although I don't understand why they'd be downloaded now. <spudpnds>Anyway, I'm not too worried about it. I'll pay more attention next time and if something is truly funny I'll see if I can reproduce it. <spudpnds>Oh, I think in between I did a "guix gc" <spudpnds>(I'm a noob, so just reading along with the manual and trying things) <nckx>Ah. That & grafts & regenerating your profile probably do. <coalquiercosa>hi guix! I'm attempting a guided graphical install on a thinkpad, but when I get to the partitioning step it doesn't show any disks D: <nckx>coalquiercosa: Sorry to hear that! Does your ThinkPad use an NVME drive? There's a known bug with those, but I think it's different from what you describe. It's night in Europe and Sunday mostplace else. Please send a report to bug-guix at gnu.org if nobody answers. Include as much information about your system/storage as possible (you can switch VTs with C-M-F2). ***jonsger1 is now known as jonsger
<dissoc`>how do I get the package version in a phase within arguments? ***catonano_ is now known as catonano
<sneek>Welcome back raghavgururajan, you have 1 message! <sneek>raghavgururajan, NieDzejkob says: you could try regenerating the autotools files, cf. the bootstrap phase <pkill9>does anyone know how the guix import scripts generate package inputs <pkill9>it's kinda confusing, as I need to generate the variable name, which is unquoted, and also generate the quote of the list <brettgilio>Hey, so I am working on an emacs package that uses the gnu-build-system. What I am trying to achieve is bytecompilation of a subdirectory of that emacs package. This is usually achieved using the #:include keyword for emacs-build-system, but as those emacs-build phases are called manually im not sure how to achieve this. <apteryx>brettgilio: often I find that using the gnu-build-system doesn't bring much (it's often a shallow Makefile infrastructure that ends up byte compiling all the elisp files, which the emacs-build-system handles more directly). <brettgilio>apteryx: Yeah, thats what I feared. It's the emacs-telega package which has a pretty sizable recipe. When I originally wrote it, I used the gnu-build-system because the package has several C components that need compiled and knows how to use the test phases without much intervention. <apteryx>brettgilio: ah, right, those are tricky. <apteryx>In those times, you can borrow some build phases from the emacs-build-system. <apteryx>combining both. It's tedious, but there are various examples already. <brettgilio>@apteryx: Yep, thats exactly what i'm doing. But the emacs-build phase itself only bytecompiles the root directory. I've tried to chdir to the subdirectory, but it just recompiles the root directory <brettgilio>Theres a contrib subdirectory that needs to hit too <apteryx>it's a bit strange that their build system doesn't handle the byte compilation? <brettgilio>im a co-maintainer so i can make changes if needed <brettgilio>the makefile does just list the root directory for being compiled <brettgilio>blackbeard: Did you know you can get syntax highlighting and inline org mode rendering in telega? :) <brettgilio>blackbeard[m]: Join our development group on telegram @emacs_telega <brettgilio>apteryx: I think i figured it out. I just replaced the emacs-build phase with a manual invokation of make compile <apteryx>mbakke: couldn't work out inkscape this weekend, I'm still winding down some ruby work or c-u. Did you have a date on mind to merge staging? <raghavgururajan>In git, is there a way to apply a patch and commit it, in one command? <bavier1>assuming the patch was made with 'git format-patch' <raghavgururajan>sneek, later tell nckx: I want to see if a build failure is reproducible. Could you try to build gtksourceviewmm on guix.t.gr? REPO: Guix Savannah ; BRANCH: wip-desktop ; COMMIT: 3000f52653ab08179696c57a560a3efe2dd9c338 <sneek>civodul, you have 1 message! <sneek>civodul, NieDzejkob says: jboy asks: is ludovic still developing the guix jupyter kernel? it appears to be broken currently. ***sputny1 is now known as sputny
<thomassgn>my wireless usb thingies stopped working some time back, so I keep booting into a progressively older system generation to have internet. I saw today there is a ath9k-htc-firmware package, do I include this in my config somehow to get wifi back? <nckx>raghavgururajan: Revert commit e2bb2bdd01a21d20f769dcc34eae06bfb47b15fd. <sneek>Welcome back nckx, you have 1 message! <sneek>nckx, raghavgururajan says: I want to see if a build failure is reproducible. Could you try to build gtksourceviewmm on guix.t.gr? REPO: Guix Savannah ; BRANCH: wip-desktop ; COMMIT: 3000f52653ab08179696c57a560a3efe2dd9c338 <nckx>thomassgn: I was responding to your <do I include this in my config somehow to get wifi back> question earlier. <thomassgn>oh, allright. Do you have any idea how to fix my wifi problem though> <nckx>You could try adding ‘(kernel linux-libre-5.older)’ to your system .scm. <nckx>thomassgn: If that's too old for you, you can use an inferior to get the (presumably) 5.x version from the generation that does work for you. <thomassgn>nckx: Nice, thanks. I see theres a patch in the discussion you linked with a fix dated 9 days ago. <thomassgn>might have a look at trying it -- though I havent used guix' package transformations in ages... :) <nckx>Oh, you can try applying it of course. I have no idea what your skill level is. ***lukedashjr is now known as luke-jr
<civodul>mbakke: thumbs up for your work on Ganeti! <rekado_>“build.xml:79: The archive commons-io-2.5-SNAPSHOT.jar doesn't exist” <nckx>raghavgururajan: OK, I thought it was a local error but since you weren't sure about reproducibility I thought I'd ask. <nckx>raghavgururajan: ‘Java related errors’ which? <nckx>Oh no! civodul deleted guix.git! <nckx>We're not movin' to GitHub. <nckx>raghavgururajan: It's back if you add a / at the end. I don't see any good reason for cgit to be so picky. <civodul>so i had someone who almost wanted "guix pack --entry-point='/bin/sh -c'", which doesn't work <civodul>not sure if it's worth supporting arguments to the entry point *nckx opens a refreshing ice-cold can of worms. <nckx>raghavgururajan: Oh that. See my ‘Revert’ comment earlier. <nckx>You updated the (correct) tarball hash with that of an HTML page. <nckx>So if there's something to be fixed, it needs to be fixed by tweaking the mirror list (again). <rekado_>i wonder if more java things are broken now, because we’re installing to a maven-compatible location instead of the old share/java or lib <nckx>raghavgururajan: After reverting that commit, if you get any hash error again, please post the full error output (including all URLs) so we can identify the misbehaving mirror. <raghavgururajan>But there was also non-hash related error regarding that java package <raghavgururajan>nckx: For a package that I am packaging, I get "error: glib.h: No such file or directory", despite having glib as input. *raghavgururajan says: paste.debian.net does not work. Throws "do not spam" error. <nckx>raghavgururajan: poly2tri-c is looking for include/glib.h, while our glib provides include/glib-2.0/glib.h. Probably because the former is quite old. <nckx>Because you're still using the old hash, Guix is still using the HTML page as a source tarball, and untarring it (predictably) fails. <nckx>I hope that glib error doesn't mean that it requires glib 1.x. <nckx>I was just about to say the same about the .pc file. <nckx>So glib@2 changed the location of its headers since 2012? OK… <nckx>raghavgururajan: Sec, I'm trying something else. <jonsger>trying to update rust-cargo-c, but it wants cargo 0.45. I wonder if thats provided by rust:cargo? <nckx>raghavgururajan: https://paste.debian.net/plain/1157141 - two unrelated changes: setting NOCONFIGURE so we can still use upstream's autogen.sh (which needs which), and setting include paths to find glib.h and glibconfig.h. <nckx>Now I get ‘*** No rule to make target 'cdt.c', needed by 'cdt.lo'. Stop.’. <nckx>We should look into why PKG_CHECK_MODULES([GLIB], [glib-2.0 >= 2.28]) doesn't set up CFLAGS later. <nckx>raghavgururajan: Watch out when setting CFLAGS in autotools projects: setting it as a make-flag overrides the default, happily undoing all of ./configure's hard work. <nckx>Still doesn't build. I have to go AFK. <nckx>raghavgururajan: Meanwhile, could you briefly explain why we need this? I'm worried about the maintenance burden of an otherwise unmaintained library from nearly a decade ago, and this wasn't a promising start. <civodul>grr i get crashes with Xorg for some combinations of external screen/resolution/rotation <civodul>rekado_: did the introduction of the maven build system break things? <civodul>i didn't know about the share/java change <rekado_>civodul: not sure if it’s the maven build system or just an upgrade to commons-io <efraim>I've been thinking some about elogind and systemd user services, 'loginctl enable-linger USER' works to prevent logging out killing your processes, but it doesn't start them at boot-up. my impression is elogind doesn't want to grow that functionality <nckx>raghavgururajan: copy-recursively. <nckx>Note that unlike install-file it doesn't imply mkdir-p. <efraim>we/i could create a start-user-shepherd service, give it a list of users, create/remove /var/lib/elogind/linger/USER files and launch shepherd for each user, as that user, using a one-shot service <raghavgururajan>For the love of all that is lovable, why do some projects are named short+common. <raghavgururajan>efraim: I usually search for packages on other distros like debian and arch. This time they don't have <efraim>I wasn't sure if netmrg was it or not *raghavgururajan did a GitHub search <pkill9>I want to make a function called 'install-file-recursively' now <pkill9>that would work too, although it would be more consistent with 'delete-file-recursively' <pkill9>which has a 'delete-file' counterpart <pkill9>although 'copy-recursively' isn't consistent with it, since it's counterpart is 'copy-file' <efraim>I've tried to 'copy-file-recursively' a number of times <nckx>Not me! …because I grep for every single time. *nckx sides with #:recursive?. <nckx>raghavgururajan: What needs mrg? <apteryx>I'm asking in advance, but when we change the expiry of our GnuPG key, we must re-upload to Savannah, right? Anything else that needs to be done? <mbakke>civodul: thanks, I'm pretty happy with it! And thanks for making such a nice and intuitive gexp interface :-) <mbakke>apteryx: dunno, I also have staging patches in my queue that I haven't gotten around to, maybe target the weekend? <civodul>apteryx: yes, re-upload to Savannah, and that's it <civodul>no need to update the copy in the 'keyring' branch, though it can't hurt either <apteryx>civodul: OK, great! Thanks for the info. <mbakke>hmm, the new chromium is pretty sluggish, I think something is wrong...somewhere. <nckx>raghavgururajan: OK, as long as the UI is more than a proof of concept, example, or toy (in 2019 it was all three). <apteryx>nothing new for me on the machine I'm tying this from. It seems to do its own swapping to disk pre-emptively, even with 1 GiB of remaining memory, and slows down to a crawl. As it as "only" 8 GiB of RAM, that occurs rather often. <apteryx>mbakke: about chromium being sluggish ^ <apteryx>still I rely on it for Jitsi (audio often has issues in Icecat, I think because it's based on the older Firefox ESR) <pkill9>i thought handling external processes in guile was really cumbersome and difficult, but it's so easy <pkill9>by default it runs the command in bash, so you can use pipes and other bash stuff <pkill9>scripting will be much nicer in Guile I think <NieDzejkob>jonsger: cargo 0.45 is in rust@1.45:cargo, I think <NieDzejkob>mbakke: I recall that semi-recently, some chromium binaries on some distro were built without optimization by accident. maybe this is related? <rekado_>pkill9: one thing you could try to improve this is to build up an S-expression directly instead of working with strings. <mbakke>NieDzejkob: possibly, do you remember which distro? I suppose I should try toggling the "is_official_build" flag which enables all Chrome release optimizations (I've been toggling the individual flags manually, but not recently) -- unfortunately that means making Clang install a cfi_blacklist.txt and using LLD for linking. <rekado_>pkill9: I’m speaking of packages->package-inputs <pkill9>rekado_: yea i would rather do that, but i didn't know how, because I want to generate the package name and variable, but put that in a quoted quasiquote <rekado_>if you want I have a few more comments <NieDzejkob>mbakke: why is using LLD for linking a bad thing? <mbakke>NieDzejkob: idk, I'm just unfamiliar with it :-) ***stikonas_ is now known as stikonas
<jonsger>NieDzejkob: ah, okay. I cheated around :P <pkill9>rekado_: yes I would like to hear your comments <rekado_>pkill9: you don’t need the (begin …) after (let …) <rekado_>pkill9: you get cleaner code by replacing (> (length args) …) with (match args …) <rekado_>each of the match clauses can then bind the arguments <rekado_>this way you’ll also be able to get rid of (list-ref args 2), which isn’t very pretty <rekado_>(if this that #f) can be replaced with (and this that) <rekado_>like in (if user-input (specifications->…) #f) <brown121407>Hi! Can you guys help me understand what exactly does #:skip-build? do for rust packages? Don't they need to eventually be built too? How else would you install them? Is it some kind of CI-only instruction? <pkill9>rekado_: where can i find documentation on (match)? <pkill9>I search and just find regex match structures <sneek>zimoun, you have 2 messages! <sneek>zimoun, apteryx says: wow! that's a lot of interactions to account for <sneek>zimoun, apteryx says: I'm glad it's reproducible to others at least! <zimoun>apteryx: interestingly, on my machine --max-jobs=2 --cores=1 fails and --max-jobs=2 --cores=2 passes then --max-jobs=2 --cores=4 fails. <nckx>brown121407: Nothing CI-related. Such Rust libraries are ‘built’ only as part of the package that depends on them. They can't be assembled into anything meaningful on their own. E.g. <https://paste.debian.net/plain/1157189>: there's no concept of /bin/addr2line or libaddr2line.so, so what would the build system build? Nothing, it's only there to be a source library for others. <nckx>zimoun: Huh. Reproducibly? <nckx>zimoun: So by fails above, you mean you get the same ‘all build users in use’ error? *zimoun commutes; back in 1h <dustyweb>... whose username on irc I have forgotten even though we still talk on XMPP ;) <nckx>dustyweb: Was it some variant of zerodaysfordays? <nckx>grep says no: just tsarfox (duh). ***terpri_ is now known as terpri
<brown121407>Is there any reason we have Rust packages that passed 1.0 that still include the minor version in the variable name (such as rust-serde-1.0)? ***xMOPX is now known as XMoPX
<efraim>we haven't gotten around to renaming them yet <brettgilio>Hey all, would somebody mind reading my email I sent about 10 minutes ago regarding qemu-binfmt? <nckx>brettgilio: I was just going to ask which command(s) caused that error. <brettgilio>nckx: ./pre-install-env guix build --system=armhf-linux swi-prolog <brettgilio>can replace swi-prolog with lean, and get the same issue *civodul hasn't seen the message yet <brettgilio>Agree. I've been staying at home with my kid playing daddy daycare. It's been great fun. <brettgilio>But, fingers crossed, I am getting some job applications back out there so hopefully something will come around :). *nckx wanted to test a native build but the aarch64 boxes aren't responding >_< <civodul>perhaps you can open a new bug for that one *civodul is grateful the Xapian search in Mumi works so well & thanks rekado_ + Arun :-) <brettgilio>civodul: Sure. You want it under a new bug number, right? <apteryx>civodul: have you seen the issue about std::round not available from cmath? Not to disrupt you from what your are doing, but I thougth you'd find it "interesting" :-) <mbakke>apparently it can be worked around by compiling cmake with "-D_FILE_OFFSET_BITS=64" <apteryx>mbakke: to me they are different, but perhaps related? <apteryx>the one I linked to is that 'round' is not redefined inside the 'std' namespace, as if the declarations inside cmath were not being honored <apteryx>so the errors says something along the lines 'round' undefined in the 'std::' namespace <apteryx>I've scratched my head a bit on the issue but couldn't get to the bottom of it so far. I suspected #include_next as it bit us in the past and cmath makes use of it. <hendursaga>So nvm doesn't work on Guix yet? And the version of node to install isn't even LTS? Odd. <jonsger>hendursaga: I think that's due to packaging node/nvm stuff isn't the funniest thing in the world... <hendursaga>jonsger: I suppose. nvm isn't a package though, it's a small little script. I'm getting the same error I get whenever I try executing, say, an AppImage <apteryx>hendursaga: you are more than welcome to become the npm champion of Guix! :-) *jonsger did update node from 10.19 to 10.21 today. It's a first step... <nckx>A small script of ~4000 lines of bash. <mbakke>hendursaga: Node 10 is LTS. Newer versions of Node depend on Typescript, which we don't have yet. <nckx>hendursaga: Which error message do you get? <mbakke>yes, we are running out of time fast :-) <civodul>mbakke: oh! so it's the thing we're all eventually hit by! <civodul>i wonder what silly thing cmake is doing here <civodul>it's not /proc i think, it's not sizeof(void*) either <momozor>Hi. I'm trying to install guile 3.0.4, but running guix install guile@3.0.4 says guix install: error: guile: package not found for version 3.0.4 <momozor>running guix show guile also not lists guile 2.x <momozor>I did both after executing guix pull FYI <nckx>momozor: What does ‘hash guix; type guix’ say? <momozor>nckx: guix is hashed (/usr/local/bin/guix) <nckx>momozor: And after running ‘. /etc/profile’? <momozor>nckx: guix is hashed (/home/momozor/.config/guix/current/bin/guix) <nckx>I'm curious though: did you just install guix (in this log-in session)? <momozor>nckx: I just did install it a few minutes ago. <momozor>nckx: Anyway, yeah it does fix my issue. Thanks. <nckx>Do you still have that terminal open? There should have been an instruction to log out & back in as the very last line. <nckx>If there wasn't, I'd love to know. <momozor>nckx: I've already closed it sadly. Probably I missed that very instruction as I was in a hurry. <jonsger>hm. pdflatex feels pretty slow on Guix <terpri>hendursaga, appimages don't really work on guix just so you know (flatpak's the closest appimage-like system that works with guix afaik) <nckx>Meanwhile, 4000-lines-of-sh nvm does absolutely nothing, printing absolutely nothing, successfully, no matter how I invoke it. ☹ *nckx pokes it some more. <nckx>terpri, hendursaga: Yeah, Guix doesn't exactly prioritise running binary blobs. We don't deliberately prevent it, but don't expect any to run easily. <terpri>tbf, some of those blobs are free software that are difficult to package for guix directly. understandable that it's not a priority though <terpri>flatpak support covers the most important ones from my perspective (tor browser, regular firefox for sites that break under icecat due to User-Agent differences, etc.) <mbakke>rekado_: there are a few R packages in the patch tracker, do you think you can review them in the near future? I know nothing about R and am not comfortable with it, but can try if you're busy. <brettgilio>mbakke: Re: the workd around. That should be applied to the cmake build system directly, not to the packages right? <brettgilio>I'm not entirely sure what that flag does, I havent had a chance to investigate it yet. At the doctor currently. <alextee[m]>mbakke: ah carla just had a new release a couple of days ago if you're still working on patches! <alextee[m]>not sure if we do release candidates, but that's the name of the tag <xoxoboii>hello, how i can replace linux-libre kernel to default linux? my wifi card no work with libre kernel <vagrantc>you have to find that on the black market :P <alextee[m]>xoxoboii: change your wifi card instead! after sending an angry email to whoever made it