<nckx>checking for chroot... yes <nckx>dstolfa: Pushed! Follow the commit message format of 36bfd64213d805c1757178ccf2fb205002cd059c exactly. <nckx>Including no self-S-o-b. <nckx>I checked the change log but found nothing suspicious. <nckx>Update: the bonkers !HAVE_CHROOT #error was transient, which is stupidly unreassuring. <nckx>Despite the ‘… yes’ above, the first run actually wrote ‘/* #undef HAVE_CHROOT */’ to nix/config.h. The second run immediately after that wrote ‘#define HAVE_CHROOT 1’. <nckx>I'd report it as a bug but what's the point of collecting ghost stories. <civodul>it was the texlive packages crashing :-/ <cbaines>jackhill, ah, it does look like the bordeaux key should be there. I can have a look, or would you like to send a patch? <dstolfa>nckx: ah i see, i misunderstood the commit format, my bad :) <dstolfa>sneek: later tell nckx: after the ipsec.conf/ipsec.secrets service gets pushed, i'll start working on swanctl support, aiming for a simple, but working configuration to start with, and extending it as needed so that it's not overengineered and risks getting out of sync with upstream <dstolfa>sneek: later tell nckx: also, it might make sense to move strongswan out of networking.scm into vpn.scm while we still can (i doubt anyone was using it since no service existed and not a single auth worked) <bdju>got a kicad build failure <bdju>log too big to paste most places it seems <bdju>`make: *** [Makefile:166: all] Error 2` and `command "make" "-j" "4" failed with status 2` are the last two lines of the log in case that helps at all <itismyfirstday>I'm not sure if I can help, but you'll need the lines before that with the actual error (usually in the last set of lines, but may be up further, look for a clear "error" message) <sneek>Welcome back itismyfirstday, you have 1 message! <bdju>itismyfirstday: http://ix.io/3qpd here's from the first instance of Error (including a little before it) to the bottom. does this have the important part? <itismyfirstday>first log shows too old of an SSL library; not sure where it is getting openssl from, but guix has the needed version <itismyfirstday>second one not sure, somehow a "history_file" is referenced somewhere? that looks odd to me <jackhill>sneek: later tell cbaines thanks for confirming that it should be there. I went ahead and submitted at patch: #49109 <itismyfirstday>munksgaard: I believe I figured out the issue you were having with that haskell package, same I was seeing. -package-db has to be passed to runhaskell, there's already a package that does that in a custom build, but probably better if that is done more automatically by haskell-build-system. I patched locally and it worked for me, will report on the <Retropikzel>Hi, I'm trying to install guix with the graphical installer and everything else works nicely but grub wont install. I think it would if I could pass it "--removable" as argument in the install script. However I could not find a way to do this from the manual. How can I do it? <jackhill>hmm, I wonder if the new substitute server warrents a news entry <raghavgururajan>leoprikler: Will you be around this afternoon? I wanna discuss with you some polkit stuff. <leoprikler>I'll probably be around, but just as a disclaimer, I'm not a polkit expert <sneek>cbaines, you have 1 message! <sneek>cbaines, jackhill says: thanks for confirming that it should be there. I went ahead and submitted at patch: #49109 <cbaines>jackhill, thanks for the patch, I've gone ahead and pushed now <muradm>hi, any one knows were to look for solution on this? <cbaines>is your problem with paste.debian.net? I see "Your requested paste entry '1200323' could not be found" <cbaines>muradm, assuming you're talking about the GPU hang stuff, that's most probably a linux thing rather than a Guix thing. You could try using different linux versions though? <bricewge>muradm: Which kernel version are you using and which CPU are you using? <muradm>cbaines: i want to see some bug entry to know at least which versions to play with, i get this lemur mabye a month or two ago, since 5.11 i had this issue <muradm>bricewge: Linux nomad-lp1 5.12.9 #1 SMP 1 x86_64 GNU/Linux <muradm>00:02.0 VGA compatible controller: Intel Corporation UHD Graphics (rev 01) <bricewge>What about the CPU not the iGPU? "cat /proc/cpuinfo | grep 'model name'| uniq" <muradm>bricewge: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz <bricewge>Arf, I have 11th gen too working but with some drivers I haven't submitted yet <muradm>yes, X11 + i3 + emacs nothing else I use <muradm>sometimes it survives for a day or two, sometimes a couple of times a day <muradm>I suppose it has something to do with that new shared model, but just speculating <bricewge>Try switching the X11's intel driver option AccelMethod to SNA, Guix default to UXA <muradm>I didn't switch to wayland/sway for now <bricewge>Is it a regression? Did it work correctly before? <muradm>bricewge: it should be modesetting driver, not Intel, I don't remember when I last time loaded specific driver under, modesetting was working fine, do you suggest trying intel specific? <muradm>I suppose it was always problem on this laptop since 5.11 <bricewge>You are already using the intel kernel driver i915 AFAIU <nckx>Good morning children of the Guix. <sneek>Welcome back nckx, you have 2 messages! <sneek>nckx, dstolfa says: after the ipsec.conf/ipsec.secrets service gets pushed, i'll start working on swanctl support, aiming for a simple, but working configuration to start with, and extending it as needed so that it's not overengineered and risks getting out of sync with upstream <sneek>nckx, dstolfa says: also, it might make sense to move strongswan out of networking.scm into vpn.scm while we still can (i doubt anyone was using it since no service existed and not a single auth worked) <soheilkhanalipur>What should I change in config.scm to configure an ethernet connection to modem as PPPOE? <ecbrown>soheilkhanalipur: i would look in nmcli/nmtui. i think pppoe could have a lot of stuff like auth <ecbrown>i have seen out there cases where a networkmanager config is created and then supplied inside of config.scm <soheilkhanalipur>There is something unknown about my connection; I connect properly for first time, but after a few seconds, I disconnect from Internet (not Ethernet connection). But I do not know what causes it… <ecbrown>are you saying that your wireless disconnects, but ethernet holds connection for long times? <soheilkhanalipur>I have an Ethernet connection, but the Internet connection is down after a few seconds. <ecbrown>i suspect if you lower mtu to 14xx <---- i don't remember the numbers instead of 1500 <soheilkhanalipur>ecbrown: If you know solution, please help (I am also a newcomer, please explain more easily 😊) <nckx>(Not a joke in extremely poor taste.) <nckx>Yeah, but 1488 is more frequent than others IME <nckx>But yes, ideally you'd calculate it based on actual data but that's going to be hard to explain very simply. <ecbrown>well, i hope my xx's accurately expressed my uncertainty <nckx>I thought you were grasping for a specific if forgotten number. <ecbrown>soheilkhanalipur: i don't connect over pppoe so as to have encountered specific pppoe problems. question: can another gnu/linux distro actually work? <ecbrown>i would like to figure out if this is a guix problem or perhaps a networkmanager issue <nckx>soheilkhanalipur: Do you know the name of your gateway interface? It's one of each first word listed by ‘ip a’. <nckx>That command will also mention ‘mtu XXXX’ where XXXX is probably 1500 now. <soheilkhanalipur>ecbrown: I also worked properly with GuixSystem three weeks ago! This problem occurred suddenly!! <nckx>You could try to lower it with ‘ip link set dev <the interface name> mtu 1400’ and see if that helps at all. <nckx>dstolfa: Done. Update your bookmarks. <roptat>nckx, I think you just broke (gnu packages vpn) <roptat>gnu/packages/vpn.scm:179:16: warning: possibly unbound variable `curl' <nckx>roptat: Thanks. I ran ‘make’ & ‘guix pull’ but didn't get any errors 🤔 <cbaines>make won't show errors, just warnings: <cbaines>gnu/packages/vpn.scm:179:16: warning: possibly unbound variable `curl' <cbaines>gnu/packages/vpn.scm:180:15: warning: possibly unbound variable `gmp' <cbaines>gnu/packages/vpn.scm:183:19: warning: possibly unbound variable `libsoup' <cbaines>but guix lint and guix build should show the issues very clearly <roptat>nckx, yeah, but "guix build strongswan" doesn't work anymore :) <nckx>cbaines: That was useful, thanks! I wasn't sure how to reliably check that (gnu packages multiprecision) was really unused now. <cbaines>nckx, if you're talking about the warnings, just be careful, since the only way to ensure guild actually runs for all of the modules is to delete all the .go files. <cbaines>so if you want to check if a module is unused, run make clean-go, then make, and then check the output for warnings <nckx>That's exactly what happened cbaines. <nckx>I don't pretend to understand how that can happen at a Guile level. <cbaines>and even then, if removing a module, I'd maybe run the derivation linter for all packages *nckx ran this unholy monstrosity earlier and now their bash history is forever cursed: guix lint $(guix package -A . | grep gnu/packages/networking.scm | cut -f1,2 | sed 's/\t/@/') <dstolfa>nckx: better than what i did before i found out about `guix hash` <nckx>Please tell me you reverse-engineered nix-base32 and computed the hash in bash. <nckx>I mean, I asked, but I was expecting you to lie. <dstolfa>i even had a tr '[:upper:]' '[:lower:]' in there <dstolfa>then did a lot of stuff, ended up in xxd and tr <dstolfa>near the end i thought... maybe there's a better way and then i found `guix hash` <dstolfa>well, at least something good came out of that! <nckx>It was accurate when I updated #guix, the other channels had explicitly not yet decided. Could you submit a patch? <nckx>oriansj: Otherwise I'll update it but it won't be today. <tschilptschilp23> * Hi! I'm just trying to compile guix and keep receiving a warning about 'help2man missing on your system' in the final stage of make. I'm actually in an environment invoked by: <tschilptschilp23>guile-git guile-json make autoconf automake gettext texinfo graphviz strace git``` <tschilptschilp23>I've already placed help2man first, but I guess the order does not have any effect, as it doesn't change the output. Does anyone have an idea what could be going wrong here? <tschilptschilp23> * Hi! I'm just trying to compile guix and keep receiving a warning about 'help2man missing on your system' in the final stage of make. I'm actually in an environment invoked by: <tschilptschilp23>```guix guile-gcrypt gnutls guile-sqlite3 guile-zlib guile-lzlib guile-avahi \ <tschilptschilp23>guile-git guile-json make autoconf automake gettext texinfo graphviz strace git``` <tschilptschilp23> * Hi! I'm just trying to compile guix and keep receiving a warning about 'help2man missing on your system' in the final stage of make. I'm actually in an environment invoked by: <nckx>Thanks for posting actual text instead of dodgy Matrix links that many people won't click. You code looks fine, it's just text, don't try to format it with weird quotes. <nckx>Another one of those warnings that's been around so long that I've grown to ignore it. <nckx>It also looks bogus, tschilptschilp23: help2man *is* found, and run, but it fails. The build system doesn't get this and just thinks ‘failure? Must be missing, better print that it's missing’ whilst it's not. <nckx>But you shouldn't need any of that. <nckx>help2man is found just fine with ‘guix environment guix -- make’ alone. <tschilptschilp23>nckx: I usually try to use pastebins, but thought the environment command would still just fit with the triple-backticks ... <nckx>What's the point of the triple backticks? <nckx>That just looks busy and confusing. <nckx>Then we know it's not missing. <nckx>The help2man hint is *also* bogus (guix-daemon --help outputs to stdout, not stderr) so I wonder what's going on. <nckx>I need to go and will look closer this evening. You can report a bug if you want to but I'll make a note of it as well. <nckx>Bogus errors tickle me :) <tschilptschilp23>nckx: thanks for looking into that -- I just saw that the first line differs, though I'm not sure if that's of interest for you, the warning itself is exactly the same: <tschilptschilp23>i guess i will reduce the --ad-hoc list and try "make check" afterwards ;) <vivien>Hi, is it possible for my guile library to symlink dependent modules in its output? I’d like to avoid propagated-inputs, because it creates incompatibilities between what guix considers the latest version of a propagated input and the package variable with the shortest name. <vivien>(specifically, I can’t install in the same profile both what package is resolved by the name "guile" and the variable guile, same for other packages) <maximed>sneek: later tell jackhill: It targets core-updates (so G-exps can be used everywhere and the patch doesn't have to deal with 'canonicalize-reference') though <lispmacs>hi, the last guix blog post immediately brought a few questions to mind: (1) what is the algorithm for whether you get a substitute from one server or the other? and (2) if a build process fails to eliminate native cpu instructions (a bug I've reported a few times) will there be a difference between builds I get from one server vs the other? (3) is there already any mechanisms in place looking for differences betwee <lispmacs>n one build vs the other? (4) how does this work with guix challenge? <lispmacs>might need a follow-up post if I'm not the only person wondering about these things <rekado_>re 2: yes, that is possible, and yes: that would be a bug <yoctocell>lispmacs: re 1: i think the order of the of substitute urls decides which substitute urls is preferred. if you call guix-daemon --substitute-url='foo bar', it will try to fetch from foo before it tries bar <lispmacs>well, I've been using --no-substitutes and --no-grafts for a while so maybe I don't need to know :) <yoctocell>lispmacs: (guix store) defines %default-substitute-urls, ci.guix.gnu.org comes first <yoctocell>so i guess ci.guix.gnu.org is preferred by default ***sneek_ is now known as sneek
<dstolfa>would it make sense to implement a git hook that first requires the package to build in CI and then push to master? every time i update my system on a laptop that's sitting on my lap i feel like i have a 50% chance of being boiled alive by something (right now that's qemu) <itismyfirstday>anyone knowledgeable about gobject, cairo, gir, and library paths? (it is painful, and I'm stuck on libobject-cairo not being found, seems it needs absolute path in gir which is not happening) <Antlers>Currently trying to figure out if I can use inferiors to downgrade to srt-1.4.2, since srt-1.4.3 has a failing test on aarch64-linux, and has been failing it's CI tests- I'm still ~pretty~ exceedingly naïve about guix, but frankly expected something like dstolfa's suggestion to be standard when pinning packages is such an involved process; you don't push broken packages? <Antlers>And disclaimer, I could be totaly mislead about the state of the srt package, and am not asking for help with it; ds's point still stands regardless <leoprikler>I think we're rather working on downstream solutions, such as channels-with-substitutes-available <dstolfa>leoprikler: that works too, just something that doesn't boil me alive <leoprikler>Developers are encouraged to test other architectures as well, but most live in x86 land still, the solution (emulation) doesn't really hold up all too well and feels like wasting resources (especially on weaker machines) <leoprikler>It's also difficult because you need to push to some branch to know whether CI will fail and most people who want something "more stable than master" actually have differing opinions on how exactly that ought to be implemented <leoprikler>Meanwhile the rollback mechanism ensures that you can just wait it out in case that stuff breaks, so yeah… <dstolfa>leoprikler: in my case, all i want is guarantee of substitute being available so i don't have to literally hold my laptop in the air while it updates a package <leoprikler>hahaha, you might want to make it a habit of typing `guix weather' then :D <dstolfa>i did, but there's too many packages i use :( <dstolfa>i checked linux-libre, ungoogled-chromium, icecat and those, but i missed qemu <leoprikler>guix weather ought to take manifests though, does it not <dstolfa>leoprikler: it does, is there any way to actually show which ones are not available? <dstolfa>right now it's just saying 129/131 available <dstolfa>not much information on which aren't <dstolfa>alright that will do for now, but would still be nice to have (maybe a separate branch or channel) that ensures that it is available :D <leoprikler>I don't think there's an upstream solution to that <leoprikler>Besides the need to evaluate many user manifests to find particular working sets, the hashes would change for each dependant package. <leoprikler>therefore the substitutes would need to be built again <dstolfa>leoprikler: i remember talking about this to lfam, i think the simplest idea was to just have a branch that lags a few hours behind master so that substitutes always exist <dstolfa>where always is... a little bit less of an assertion and more of a fuzzy thing :) <leoprikler>I don't think it's quite that simple when we talk about stuff like package X breaking on aarch only <dstolfa>yeah, of course not. i'm just talking about mildly improving the state of substitutes right now for weaker devices <dstolfa>this wouldn't solve any of what Antlers brought up, it would just boil me less often if i didn't check guix weather :) <leoprikler>another downstream solution would be to simply wait six to twelve hours between guix pull and guix package :D <leoprikler>One bug discussing the channel configuration thingy is #47929 <pkill9>can guix be fixed so it knows everything that will be downloaded from the beginning? <itismyfirstday>how does one find a library path in a package build? e.g. the path to an input dependency package lib? <raghavgururajan>itismyfirstday: Build with `-K` argument and the let the build fail. -K keeps the build-dir. Go to that dir and open the 'environment-variables' file. There you will find library paths. <civodul>pkill9: i think we discussed it before and the short answer is "no" <itismyfirstday>raghavgururajan right, but what I mean is I think I need to patch a package where it needs the path to the library from one of its dependencies.....I guess this is what propagated-inputs and profiles are for? <raghavgururajan>itismyfirstda: Ah, you need (string-append (assoc-ref inputs "input-name") "/lib") <itismyfirstday>are there any examples (didn't see in the manual) for using guile snippet to patch in a package? I'm guessing it is pretty straight forward guile string replacement, but wouldn't mind some samples to follow <itismyfirstday>raghavgururajan great, thanks! that's just what I need, and was just searching through the source to see examples as well *vagrantc could've sworn vagrantc saw itismyfirstday a couple days ago... <itismyfirstday>haha yea, not so much first day, more like a marathon week deep diving into guix and haskell packages (and now gobject...) <vagrantc>always approaching things with fresh eyes can be a good thing :) <itismyfirstday>just one of those things where you fix problems to be lead to other ones, eventually finding out there are some long standing bugs/idiosyncrasies in some things <itismyfirstday>but I think I'm nearing the end, hopefully ending up with a couple of easy patches to fix the problems I ran into, which should help some haskell packages at least <euandreh>Given that Guix uses gettext so heavily ("find -type f -name '*.po' | xargs wc -l" says ~1.2M), have we started to hit some of its limitations? <raghavgururajan>itismyfirstday: I think it would be better for folks to help, if you could describe the problem with the paste of package definition. :) <euandreh>As in: has it been limiting for us somehow? <vagrantc>other than the occasional bug in the .po files preventing something from building correctly ... but that can happen anywhere in the codebase <civodul>euandreh: gettext scales well to much larger message catalogs <euandreh>leoprikler: from where I stand, I find gettext to be the best in class. I encounter a lot of ad-hoc poor solutions, and I've never actually seen people using catgets anywhere. I want to ponder this view of mine with shortcomings that gettext may present and I'm not aware of. <euandreh>vagrantc: sure. Not specifically about such details, but message catalogs in general, or how gettext's view of languages and translations helps or hurts. <euandreh>civodul: sorry, I didn't quite understand what you mean. Is that "scales too* well" as in "it does a great job"? <vagrantc>that said, localization is one of those things i've always found important but the (free) software world caters too much to my native language for me to notice much <leoprikler>euandreh: "scales well to", not "scales too well" <muradm>hi, i have (mapped-devices ... (mapped-device .. (type luks-device-mapping))) (file-systems ... (file-system (type "btrfs") (mount-point "/.swap") ...)) (swap-devices '("/.swap/swapfile0")) basically, swap file on luks encrypted btrfs volume. there seems to be no relation between (swap-device ...) and mounted file systems, so on every boot "swap-/.swap/swapfile0" fail to start <euandreh>Also, I'd guess that doing l10n work for french for example is more straightforward than doing it for mandarim. Is that true? <muradm>any way to make swap-device depend on mounted fs? <leoprikler>the universal truth about l10n is don't split stuff into units smaller than sentences <leoprikler>whether French or Mandarin doesn't really matter at that point then <euandreh>leoprikler: that's interesting. Does that come from accumulated experience? <euandreh>Or is this just a general well-known rule of thumb? <leoprikler>Relying on English grammar is simply a bad decision <civodul>euandreh: in short, it scales well :-) <leoprikler>You won't even find it in other European languages like German or French <itismyfirstday>raghavgururajan unfortunately it is a huge set of packages as it turns out, I see 1300 lines of package definitions (via import from hackage)...mostly smooth sailing besides something that someone recently reported an issue on, and now the gobject stuff (fixed it I believe) <leoprikler>or rather you'll find cases in which your assumptions break <euandreh>leoprikler, vagrantc, civodul: ty, always a pleasure! <dstolfa>sneek: later tell raghavgururajan: how well does gnome40 work in guix right now? what are the main issues blocking it from being merged? <dstolfa>looks like sneek fell asleep, even bots need rest <leoprikler>Whereas if you're given the full structure and context of a sentence – even if sentence structure works different in your target language – you as translator can assemble the sentence as you need. <leoprikler>But gettext is flexible enough to sort those things out even with GUI applications <leoprikler>[though IIRC Guile gettext does not have contexts] <raghavgururajan>dstolfa: The work has been stalled for now, as I am trying to figure out a way to rebuild stuff in a efficient way. My X200T is melty. <euandreh>leoprikler: nice, I'm starting to use gettext a bit more for some projects, and I like to hear you mention flexibility when talking about it <muradm>not at least now, gnu/services/base.scm:2153 could have mount point lookup instead of empty list <dstolfa>raghavgururajan: oof, yeah building gnome on an x200t is not the simplest way <leoprikler>I find it funny, because Ren'py is like "so, we heard you like Python, so we put Python into your translation system so that you can Python while you translate", and I'm here developing my own VN engine saying "lol, just use gettext" <muradm>and gnu/services/base.scm:2163 one more condition on file-system mount-point prefix somehow <ngz>Hello. I have a question concerning new substitute server. I'm on a foreign distro, and I authorized bordeaux's key. Do I need to add --substitute-urls option in guix-daemon.service, too? I think not, but I'd rather be sure.