IRC channel logs
2020-03-09.log
back to list of logs
<sirgazil>But someone on the mailing list seems to have a patch for Thunar. <mroh>I send a patch to 25201@debbugs.gnu.org is that enough or do I need to do more to get that fixed? <NieDzejkob>mroh: When adding a new file, add it to gnu/local.mk too. Also, I don't think the patch file name should contain the bug number. *sirgazil is frustrated (time to take a break) <nly>does anyone use guitarix? the tuner seems to be not worknig <guix-vits>in Foreword of sicp package: "(at most {1 {1\over2}} feet)." is displayed. Is this a bug (seen both in M-x info and info)? <NieDzejkob>Blackbeard: does the error look expected? You can label the package as x86-only like xf86-video-intel <Blackbeard>NieDzejkob: it succeeded with I686, x86_64 and aarch64 :) <Gooberpatrol66>jackhill: it appears guix-devel requires a membership to post. how can I gain access so I can ask for help with the package I wrote? <NieDzejkob>Gooberpatrol66: it doesn't require membership, it's just that the first post needs manual approval <guix-vits>NieDzejkob: a typo is in Manual#Contributing : http instead of https for contributor-covenant.org <NieDzejkob>guix-vits: Thanks! Do you want me to add a Reported-By line in the commit? <Sleep_Walker>I'm having problem with running `guix pull` from clean 1.0.1 installation - is it known problem? <Sleep_Walker>build of /gnu/store/yc1dgyl6cx0h4irf4x9c99y39333akm9-guix-package-cache.drv failed <Sleep_Walker>Generating package cache for '/gnu/store/qcfldax4bq5cc3ypx8cby9p2zzhya9v2-profile'... <Sleep_Walker>(exception unbound-variable (value #f) (value "Unbound variable: ~S") (value (nss)) (value #f)) <guix-vits>Sleep_Walker: is this a "Graphical" or "Manual" installation? *guix-vits with Russian accent "ahh foreigner :)" <guix-vits>civodul: Sleep_Walker can't pull on fresh install on-top of openSUSE, above; is this a known issue? <Sleep_Walker>civodul: I was just checking if there is some known problem - it fails on `build of /gnu/store/yc1dgyl6cx0h4irf4x9c99y39333akm9-guix-package-cache.drv failed` with pretty short log http://ix.io/2dNo <Blackbeard>the readme says to only use 'make' and 'make install' <NieDzejkob>Blackbeard: is that all output? I'd rather you posted too much than too little <guix-vits>in your first post: "all warnings treated as errors" <NieDzejkob>You can replicate that with (add-after 'unpack 'werror-begone (lambda _ (substitute* "Makefile" (("-Werror") "")))) <civodul>Sleep_Walker: if you have channels apart from the official one, the error could come from there <Sleep_Walker>civodul: thanks! I believed I wipe whole installation before reproducing but I indeed forgot this one <guix-vits>Blackbeard: if compiler notices something "unusual" but not "wrong" within the code, it can emit an "warning". Good programming practice is "treat all warnings as errors"; so this flag is exist. <Blackbeard>ok now I have another problem that I think arch is solving too <Blackbeard>mkdir: cannot create directory ‘/usr’: Permission denied <Blackbeard>make DESTDIR="${pkgdir}" BINDIR="/usr/bin/" install <NieDzejkob>you can't install to /usr, you need to install to the output <NieDzejkob>in your case, you want no DESTDIR and BINDIR of (string-append (assoc-ref outputs "out") "/bin") <NieDzejkob>yeah, though it's usually preferred to get outputs as a real argument with (lambda* #:key outputs ...) <NieDzejkob>Blackbeard: before you go, could you post the new package definition and error? <Blackbeard>NieDzejkob: oh I was ssh-ing into my desktop and I already turned it off :( <jonsger>guix-vits: what is the excact problem on openSUSE? <guix-vits>Sleep_Walker: ^ (jonsger: idk, some pull error.) <Sleep_Walker>I thought I wiped whole guix installation and left channels file in - and it was my source of trouble <janneke>haha, my efforts to create a smaller qemu footprint are somewhat feeble, qemu has not been ported to the Hurd <civodul>you should talk to youpi about that, youpi commits to both <janneke>ah, good -- i tried to build everything natively, to avoid solving cross build problems :-| <janneke>hmm, what's with this `-pc' in i586-pc-gnu that is sometimes mandatory (for cross-mig, apparently) <PurpleSym>How can I resolve package conflicts? guix is telling me to upgrade the conflicting packages at the same time, but that’s what I’m doing (via a manifest). <guix-vits>PurpleSym: Can you share which packages is in conflict? <DamienCassou>is there a way to generate a value for the PATH variable from a list of dependencies? I would like to use `wrap-script` to setup the PATH of a bash script. I could use the PATH that is current while the package builds, but this contains more things than necessary (e.g., GCC) <civodul>dftxbs3e: i got a bounce from your email address ("spam detected" :-/) <NieDzejkob>DamienCassou: Yeah, you could do something like (wrap-program the-program `("PATH" ":" prefix ,(map (lambda (dep) (assoc-ref inputs dep)) '("foo" "bar" "baz")))) <guix-vits>PurpleSym: IDK, but entries for python-mistune has different hashes. <PurpleSym>The only difference is in mistune.cpython-37.pyc and it’s literally the hash of its own /gnu/store directory that’s different. <NieDzejkob>I'm trying to set up offloading onto my Debian machine, but ssh with a command directly on the commandline doesn't run a login shell, which means it doesn't read /etc/profile.d/guix.sh which sets the relevant env vars. As a perhaps related problem, I get a protocol error from guix offload test: https://paste.debian.net/1134102/ <NieDzejkob>is there a way to update /usr/local/bin/guix on a foreign distro? the equivalent on Guix System would get updated on a reconfigure <civodul>NieDzejkob: you would run "guix pull" as root <civodul>since /usr/local/bin/guix points to root's profile <guix-vits>PurpleSym: did you customized anything in python-mistune, or mistune.cpython-37.pyc is packaged inside of jupiter-zpid? <NieDzejkob>civodul: oh, I thought I was running some old version because the commit ID is different, but it's just that someone pushed a commit between my pull as user and as root <NieDzejkob>PurpleSym: Maybe there's a python-mistune in both Guix and your channel? <raingloom>if i need multiple source files for a package, i just use a list of origin objects, right? <civodul>raingloom: 'source' cannot be a list, but you can add extra origins as inputs *rekado_ started packaging ROOT *rekado_ waits for that custom LLVM build that’s part of the ROOT build… <PurpleSym>guix-vits: No. NieDzejkob: No, it’s pulled in from the main channel. <guix-vits>NieDzejkob: is hash of propagated input dependends on "caller"? <guix-vits>both use (propagated-inputs with python-mistune <PurpleSym>That shouldn’t be a problem, both *could* use the same version of python-mistune. <guix-vits>could, but somehow they don't; if "caller" can change output of propagated-inputs? idk. <NieDzejkob>PurpleSym: Could you try `guix environment --ad-hoc python-jupyterlab@1.2.6 jupyter-zpid@1.0.0`? <guix-vits>btw, both packages are new: i'd recently pulled, and guix search returns '' for both... <NieDzejkob>PurpleSym: it would also help if we could see your channel <PurpleSym>NieDzejkob: Strange, `guix environment` works. <PurpleSym>And `guix environment --ad-hoc -m <manifest>` works too. <civodul>hello PurpleSym! good to see you here :-) <rndd>hi everyone! could anybody help? i wanna try i3wm on my laptop. so, i dont know how to change my config.scm to install and apply i3... where i can found information about possible functions and arguments fo config.scm? <rndd>NieDzejkob: thank you! found DE declaration examples in manual but there is nothing about window managers =| <guix-vits>btw: if one running `reconfigure` as not-root, is "Permission denied" error is OK? <NieDzejkob>rndd: you only need a login manager or whatever it's called, like gdm or slim, and the window manager, either listed as a package or installed into your user's profile <NieDzejkob>guix-vits: Permission denied is the expected result of running reconfigure as non-root. <NieDzejkob>it won't break anything, but it also won't have the effect it would if you reconfigured as root <civodul>i'm working on that bug but i don't know how long it'll take <jonsger>using some dedicated machines at Hetzner should be much cheaper for them... <civodul>also Guix/Nix would allow them to share build resources efficiently <civodul>as opposed to having lots of mostly idle machines <civodul>you know what, i started going through my guix-devel backlog yesterday and didn't manage to get very far <anadon>civodul: Before I leave for a bit, last Thursday and Friday I ran into failures for `guix pack --format=squashfs ...` on CentOS7 related to "guix/scripts/pack.scm". Adding the "--no-recovery" option helped one kind of failure from replicating on the command line, but the guix command still fails. <civodul>anyone knows if there's a bug report for what anadon writes about? <dftxbs3e>civodul, hey, what do you mean? I don't spam <civodul>dftxbs3e: free.fr bounced my message saying "spam detected" <dftxbs3e>civodul, oh... well I can't do anything about that :-/ - free.fr is my isp <civodul>"to bounce" is intransitive in the real world, but you see what i mean <civodul>there's been abuse related to GNU, so perhaps that's the aftermath <dftxbs3e>civodul, sure :-) - well in the mean time, email me with lle-bout at zaclys.net <dftxbs3e>civodul, oh.. I think I misunderstood. Is it related to: "You have been unsubscribed from the Guix-commits mailing list"? <dftxbs3e>Maybe an RSS feed would be more appropriate for this mailing list <civodul>probably messages from guix-commits@gnu.org kept bouncing, so mailman eventually unsubscribed you <dftxbs3e>civodul, makes sense. I'll email the postmaster. <civodul>anadon: i was saying i haven't read the discussion above <civodul>but! if anyone else beats me at it, that's even better :-) <anadon>I had to reboot my system -- kernel updates and time change. <dftxbs3e>civodul, sent. Hopefully they're not too busy or spammed to even read and handle the situation. <dftxbs3e>I'm not sure postmaster@free.fr is even read. <dftxbs3e>With millions of customers.. and dozens of millions of email addresses over the years, it's got to be hard. <dftxbs3e>civodul, maybe considering that not many people know how email actually works and what postmaster even is, they wouldnt receive that many emails. <rekado_>hmm, the tensorflow team removed what little cmake support there was. <rekado_>it can only be built with bazel now. <rekado_>which means it cannot be built with Guix <anadon>What would be involved in adding the bazel build system? <guix-vits>anadon: Bazel written in Java; Finding some Java-hackers may be involving, probably. <anadon>Gross. Why is it that as time wears on I more and more dislike Java? <anadon>I suppose the first step would be to package bazel. <janneke>anadon: maybe because you didn't dislike it enough initially ;) <guix-vits>"I read a book about Java, and found it an elegant further development from C. But I have never used it." -- Guess Who? <nckx>civodul: Nope. anadon: Any progress debugging the singularity? <anadon>Alas, none. I'm more working on Guile fluency before I dig in and make a mess. Also going through my calc textbook, next draft on a paper, and actual work. <rekado_>Packaging Bazel is a serious multi-month project. <rekado_>many Java packages cannot be built directly from source as they depend on some pre-built jars. <rekado_>and then actually using Bazel: how could we make it *not* download very specific versions from the internet? <rekado_>we would not be able to reuse any of the packages provided by Guix. <guix-vits>:trollface: but Java's platform-independency! <rekado_>the Java culture is the much bigger problem. <guix-vits>TensorFlow and Bazel are both from Google; logical, btw. Anyway in meantime, i'm suppose one can create /usr/local ..? <rekado_>it’s just that newer versions are going to be difficult to provide via Guix. <rekado_>and no, installing this manually is rarely an option <guix-vits>window.location.replace("/versions/master/./install.html" + window.location.hash); <anadon>Still better than Gradle, I suppose. <anadon>Should someone open a ticket with bazel about building offline and in a way compatible with guix? <rekado_>anadon: I already asked them if there’s a lighter variant of Bazel that doesn’t depend on 40+ jars. They said no. <rekado_>they did not seem to follow my claim that it’s very difficult to build from source when nobody builds jars from source. <rekado_>so… maybe someone could discuss this with them, but I prefer not to do this myself. <bavier>I'm so happy we have a 'nyancat' package now <civodul>anadon: some of the developers were present in past R-B summits <civodul>so they are somewhat receptive, but maybe not enough <anadon>Reading over the FAQ, I don't think they actually understand the concept. <anadon>The whole "prove it" part of reproduciblilty seems missing. <civodul>well Bazel itself is a good tool for reproducible builds <civodul>it's not very different from Nix/Guix AIUI <civodul>but when you use a pre-built binaries, there's no build to reproduce, so you're fine :-) <NieDzejkob>I'm still having problems with offloading onto my Debian machine <NieDzejkob>I made my bashrc load /etc/profile.d/guix.sh and now "ssh z guix repl" doesn't complain about locales <NieDzejkob>I also made sure I'm using the most recent guix daemon on both machines <NieDzejkob>and I'm still getting the error: guix offload: error: failed to connect to `#<input-output: channel (open) 7f6509c11940>': Protocol error <NieDzejkob>it's weird, the #<input-output: channel> is supposedly the result of (build-machine-name machine) <rndd>hi everyone! does anybody here use dwm with guix? <anadon>NieDzejkob: I'd suggest something if I knew. Sorry :/ <NieDzejkob>After some print-debugging, I determined that the problem is in remote-daemon-channel's call to open-remote-pipe* <rndd>NieDzejkob: maybe. so my question is about modifying config.h (that is the basic way to configure dwm - the idea is about recompiling whole dwm when you want to change, for example, frame fize). So, i need to put my config.h inside project source and recompile how i can to do it with guix ? <NieDzejkob>wait, does dwm go into the system config or user's profile <rndd>NieDzejkob: no, just made this package as system in config.scm <NieDzejkob>if that works I might submit a patch to provide a function for this <rndd>NieDzejkob: that's cool, but how i can write in scheme -> download my git repo with dot files and copy config.h from here? =/ <rekado_>rndd: you probably wouldn’t do this in a system configuration. But if you absolutely want to do that you could use guile-git, or spawn any process (such as ‘git clone’) with ‘(system* …)’ <NieDzejkob>rndd: better solution: put the config file in your git repo <rndd>ok, thanks for suggestions <jackhill>Blackbeard: I suppose if it were easy we wouldn't need to collaborate on a project to wrangle it all :) <rekado_>if only there was a set of tools to automate most of the common steps. We could perhaps call this ‘autotools’. <rekado_>the most annoying packaging tasks I encountered were always those that didn’t stick to conventions or tried to reinvent them. <Blackbeard>I wanted to send package definitions on the weekend <Blackbeard>But I think having more games will be a nice addition <Blackbeard>I can't think of much 'work' stuff to package. Everything I need is already there <Blackbeard>Here I am packaging games when people already have needs hahah <bavier>Blackbeard: to be fair, many of the entries on the wishlist are for things that are difficult to package at the moment <mbakke>nckx: the libinput update causes 700+ rebuilds (see libinput-minimal) <nckx>Is there any way to track these things yet? <nckx>I have a file with ‘things that guix refresh -l lies about’. It's on another machine. <nckx>mbakke: Re: your thing, I thought of that exact thing when I saw the original thing this morning. No, I thought to myself, you're being silly and paranoid. <NieDzejkob>is there a way to get all packages that inherit from a specific package? <nckx>Not AFAIK. That's the problem. <cbaines>building /gnu/store/jis9853b5djlf6fxgfqhyd5vffwplmla-module-import.drv... <cbaines>while setting up the build environment: executing `/gnu/store/50f18mlbj915vydqmm6nyvkh2b23vmf0-guile-2.2.6/bin/guile': No such file or directory <nckx>NieDzejkob: I've suggested that (and have an ancient implementation rusting around somewhere). But it's not a pretty concept and Ludo' wasn't wild about it. <nckx>NieDzejkob: Not by design. <nckx>NieDzejkob: As in, I won't say ‘it works!’ because I don't know, but it ‘should’ 😉 <mbakke>I wonder if Guix is smart enough to track the rebuilds per origin. <mbakke>Unfortunately this does not work: guix refresh -l -e '((@ (guix packages) package-source) (@ (gnu packages freedesktop) libinput))' <nckx>Hm. Interesting approach. <NieDzejkob>nckx: well, from my short experience, it doesn't :/ <mbakke>"expression does not evaluate to a package" <NieDzejkob>you will see my frustrated calls for help if you scroll up a screenful or two <nckx>NieDzejkob: What happens? <nckx>Sorry I was just summoned here by libinput, haven't read any backlog. <NieDzejkob>the TL;DR is that guix offload test fails in assert-node-can-import with guix offload: error: failed to connect to `#<input-output: channel (open) 7f35de0400a0>': Protocol error <mbakke>NieDzejkob: what does 'guix offload test' say? Can you ssh to the target machine using the credentials from /etc/guix/machines.scm ? <mbakke>NieDzejkob: is guile-ssh installed for the offloading user, and GUILE_LOAD_PATH set up? <mbakke>NieDzejkob: indeed... I looked at my offloading setup, and the 'guix-offload' user has only three things in ~/.guix-profile: guile, guix (!), and glibc-locales. <mbakke>NieDzejkob: it also sources ~/.guix-profile/etc/profile from .bashrc <NieDzejkob>hmm, I'll try installing gui... oh of course that's it <NieDzejkob>the source code I was reading was invoking guile -c to open the connection <NieDzejkob>guix offload: error: unknown error while sending files over SSH <NieDzejkob>how do I make it print the whole exception and not only the message <NieDzejkob>I ran it in the REPL, the exception doesn't help <NieDzejkob>even though `ssh z guix repl` was working before <mbakke>Huh, I thought you did that already? <NieDzejkob>I thought "huh, there's already a guix in the ~/.config/guix/current profile, I don't need that, I'll try with just guile for now" <NieDzejkob>so, idk why, but adding guix to the profile made guix offload test say "guix offload: '192.168.1.50' is running GNU Guile 3.0.0" <NieDzejkob>before guix in profile: using guile 3.0.0; after: using guile 2.2.6 <NieDzejkob>if my master node goes to sleep during a long offloaded build, will that be a problem? <NieDzejkob>or can I suspend it and rest assured that icecat or whatever is compiling on the build slave? <mbakke>NieDzejkob: I believe the builds will get killed after some timeout. <NieDzejkob>hmm, okay. What about --keep-failed? I remember reading that it will automatically imply --no-offload <NieDzejkob>Is there a way to make it keep the build directory at least on the slave? <Blackbeard>How can terminus or any other font can be added to the installer <Blackbeard>When I installed Guix I had trouble with the fonts being a small <Blackbeard>Not the installer per se, sorry the USB because I used the console <mbakke>NieDzejkob: there is no way to pass the --keep-failed flag to the remote (in fact --keep-failed implies --no-offload) <mbakke>I occasionally ssh to the remote and 'guix build -K /gnu/store/...-foo.drv' if I don't want to build and investigate locally <NieDzejkob>okay, I guess I'll just start the most painful builds through ssh <NieDzejkob>speaking of which, rust@1.40 is doing some shenanigans that makes ld-wrapper generate an incomplete rpath <NieDzejkob>/gnu/store/g7w8qzw94cd5snmhwx9ankpx3axy0m4d-rust-1.40.0/lib/rustlib/x86_64-unknown-linux-gnu/codegen-backends/librustc_codegen_llvm-llvm.so: error: depends on 'librustc_driver-1040000d6b8afc7f.so', which cannot be found in RUNPATH ("/gnu/store/g7w8qzw94cd5snmhwx9ankpx3axy0m4d-rust-1.40.0/lib/rustlib/x86_64-unknown-linux-gnu/codegen-backends/../lib" "/gnu/store/nsrym6zn3yb4390fazx1gq8rg0m8dhkd-llvm-8.0.0/lib" "/gnu/store/2plcy91lypnbbysb18ymnhaw3zwk8pg1- <NieDzejkob>gcc-7.4.0-lib/lib/gcc/x86_64-unknown-linux-gnu/7.4.0/../../.." "/gnu/store/ahqgl4h89xqj695lgqvsaf6zh2nhy4pj-glibc-2.29/lib") <NieDzejkob>the offending .so is in /gnu/store/g7w8qzw94cd5snmhwx9ankpx3axy0m4d-rust-1.40.0/lib <mbakke>NieDzejkob: weird! If the build system respects LDFLAGS, perhaps you can try adding that directory through 'LDFLAGS=-Wl,-rpath=...rust-1.40.0/lib' ? <nckx>Blackbeard: I've already added it, but you need to build your own installer (it's not in 1.0.1) and add fbcon=font:TER16x32 to the kernel command line. <nckx>However, if you use a USB console, it's the ‘host’ machine that sets the font so I don't understand that line and have chosen to ignore it. <nckx>Blackbeard: It is, but it turns out that it won't affect the installer because the installer knows best and hard-codes a console-font service and font (2 different ones, even: LatGrkCyr-8x16 on TTY2 and lat9u-16 on the others). <mbakke>I never remember how to cancel a Cuirass evaluation, I should write it down somewhere. <rekado_>mbakke: better even: let’s add it to the admin interface <Blackbeard>I can understand the hard coded when starting, I guess it is hard to use a good font for every display out there <nckx>‘setfont -v latarcyrheb-sun32’ works. <nckx>Terminus is only available as 16, but this was about size, not prettiness, right? <nckx>Blackbeard: Anyway, setfont -h will print the font directory as its last line. You can list consolefonts/ yourself to see the available fonts. Some 22px options there if 32 is overkill. <mbakke>rekado_: that's the best kind of writing it down <mbakke>nckx: re libinput in particular, making the inheritance more visible could be a UI improvement. E.g. with a comment, or changing the inheritance order. <raingloom>is there a way to pretty-print a derivation? it's rather hard to debug them when everything is on a single line. <nixo_>Hi guix, how do I get the path of a plain-file in the system configuration? <sneek>nixo_, NieDzejkob says: FYI, you forgot to --signoff on the nheko patchstack you pushed. I tried to find an option to signoff by default, but failed miserably, so, uh... be more careful next time ;)