<atw>"No such file or directory: "my-root"" <atw>I'm testing out Agda and agda2-mode and when I compile an Agda program, it seems like Agda is trying to write into the store. Has anyone seen behavior like this out of a Haskell package before? ***kelsoo1 is now known as help
***help is now known as Guest44760
***Guest44760 is now known as kels007
***kels007 is now known as kels001
***kels001 is now known as kelsoo1
<pmikkelsen>hi guix, what is the best way to upgrade a system with not so much ram? a normal guix pull justs runs out of memory <pkill9>pmikkelsen: Guix requires atleast 1GB of RAM I think <pkill9>atleast for installing it first time, and `guix pull` involves compiling <pkill9>sorry that probably doesn't help much <pmikkelsen>pkill9: allright, I just wanted to know if it was possible to do a guix pull on a better machine and then maybe rsync it or something but its okay <iyzsong>pmikkelsen: one way is to run 'guix pull' on a machine with enough RAM and then 'guix copy' the 'guix-latest' store to the target machine, symlink it to ~/.config/guix/latest manually. <dddddd>Hi, I just did a "binary instalation" on an arm64 system that has no network available (I can copy files using flash drives). Is there any supported method, for situations like this, to install new programs as _binary_? (something apt-offline alike?) <dddddd>Let's say midnight commander... I see (big) closure files at hydra, maybe I can use them? How? <efraim>Kind of, on one machine you can run 'guix export --recursive midnight-commander > mc.nar' and on the offline system 'guix import < mc.nar' <efraim>Unfortunately I haven't been able to build from the 0.14.0 release tarball to guix on aarch64, so for the moment upgrading a new installation is a bit hard <efraim>AFAIK there are zero substitutes (compiled binaries) for aarch64 for the 0.14.0 release, on starting about a month or two later <dddddd>hmm, I think I read about export/import indeed. It's an option if you have another guix system around (using the same arch?) which is not the case right now. <efraim>Same arch would be best, otherwise you could try with --system=aarch64-linux <dddddd>Any idea about using that closure file? <efraim>Other than CaviumX machines armhf binaries should run on aarch64, but currently the aarch64 guix-daemon can't build armhf binaries <rekado_>I’m trying again to fix bug #22533, but the changes break a bunch of tests. ***danialbehzadi1 is now known as danialbehzadi
<rekado_>bah, I’m getting way too many test failures in Python after the change. In seemingly unrelated tests :-/ <atw>rekado_: Idris' missings deps are "aeson >=0.6 && <1.2, trifecta ==1.6.*". I'm going to try fixing that. Should it be difficult? <rekado_>I remember that Idris was very picky about these dependencies. <rekado_>You may not be able to just tell it to accept other versions. <rekado_>built python-minimal with the changes to the bytecode compiler, but I had to disable a bunch of tests that I really think shouldn’t fail. <rekado_>but with this change python-six builds reproducibly. <rekado_>python-sip also builds reproducibly now. <civodul>rekado_: oh you're tackling .pyc reproducibility on core-updates? <rekado_>I just wanted to see if I could increase the number of reproducible packages that our pipeline uses. <civodul>a nice side-effect of this paper :-) <rekado_>it also showed me that r-minimal is no longer reproducible again … :-/ <efraim>'flatten' is defined several times, i assume guix/glob.scm is now supposed to be the master version <efraim>civodul: I'm working on openntpd, changing to match-record showed a mistake I made <janneke>civodul: thanks for your initrd-module efforts! <janneke>this should make installing on less-standard systems lots easier <efraim>before I forget again, in hdf-java in maths.scm, there's an 'automake-1.15' hardcoded, which I can't change to (version-...) and check on my aarch64 board <atw>rekado_: according to LTS Haskell, upgrading Idris to 1.2 ought to solve aeson. Trifecta is in the cabal file for Idris 1.0 but LTS Haskell does not list it as a dep. I will upgrade Idris removing trifecta and see what happens. <axg`>Hello everyone, so I'm trying to install the legacy python on my system. under the search two packages of `python` under versions 2.7.14 and 3.6.3, and I cannot figure out how to install python2. There is no flag under guix package to specify version of package or anything like that. Why do those two packages have to have the same name? <axg`>Alright, so I'm dumb and I found the python@2 flag. <axg`>However, it uninstalls python3 and replaced it with python 2, how do I get around that? <pkill9>you can't have both versions installed to the same profile <pkill9>altho someone said they were gonna work on it so that you can <pkill9>but currently you can't because it relies on the PYTHONPATH variable to find the libraries i think <axg`>oh damn that sucks, alright, thanks <atw>axg`: re your much earlier question about TRAMP: I've started using it more and I believe it's specifically file name completion that can be slow. technomancy seemed to suggest that it does synchronous network calls. <axg`>atw: thanks for getting back to me! Do you mean that only while trying to do autocompletion it should slow down? I haven't played around with tramp mode, but haven't been able to replicate it recently. When I experienced the most obvious problems, I had more than one tramp sessions open so maybe it somehow compounded. Haven't tried nor replicated it though as of yet. <atw>axg`: my tests were far from scientific. I'd ask on #emacs. You could also look into elisp profiling and figure out exactly what the bottleneck is. <axg`>And another question I had was with regards to the st (suckless terminal). There is pretty much no reason to have it in the repo since all configuration is done manually and involves compilation. Would the proper approach be installing the executable in like $HOME/bin as to not mess with the system or making a custom personal package that would use the user's profile? <axg`>atw: I will check it out. Everything has been running smoothly though, since I haven't used tramp too much, maybe sessions left open for too long cause some kind of problem on my system. Next time I see the bug, I'll try to isolate it and bring it to #emacs. <atw>axg`: custom package, I'd say. A user can define their own derivative package with local sources. IMHO Guix addresses this perfectly <efraim>axg`: without looking at it specifically, it would be much easier for a user to have a custom version of st in their GUIX_PACKAGE_PATH if one is already defined inside guix, and then to make their changes in relation to the guix version <axg`>so the inherit option huh? I gotta look into the packaging closer, I'm still pretty oblivious as to what guix makes easy <atw>Yeah, what efraim said. Being able to easily define my own custom-compiled packages was one of the big draws of guix for me <efraim>on my TODO list is to make a local variant of parallel where I rename everything to gnu-parallel so I can install moreutils at the same time <efraim>i'm pretty sure I have to move the binary and man files and run some substitute* magic on the man file and that should be most of it <atw>how can I figure out what haskell packages are bundled with our GHC (8.0.2) ? <atw>Ah, guix environment ... -- ghc-pkg list <thorwil>is that (still) accurate? i'd like to set up a minimal wayland solution. maybe with weston <thorwil>(services %desktop-services) implies slim-service, right? ***danialbehzadi1 is now known as danialbehzadi
***danialbehzadi1 is now known as danialbehzadi
<civodul>i've restarted 'guix publish' there after gathering backtraces to understand the deadlock <atw>Thanks! I was able to --fallback <pkill9>how do i add patches to an inherited origin? <rekado_>pkill9: take a look at libtiff/fixed in gnu/packages/image.scm <civodul>hey mange, i just pushed 2 out of 3 patches <mange>Yay! I don't know how I unstaged forking-service.sh, so thanks for pulling the previous version. <mange>The 0.5 second timeout is interruptible. The select call returns when a socket is ready, or after 0.5 seconds if none are ready. <mange>Like, (select (list sock) '() '() 0) will have more overhead than (select (list sock) '() '() 0.5), because in the first case it won't actually wait for a socket, it will poll the sockets. In the latter it will wait on the sockets and come back into guile when one is ready, or after 0.5 seconds. <pkill9>does anyone else find that 'st' just closes with message "child finished with error '256'" <pkill9>the window closes to soon, it's like a flicker <pkill9>note: I'm running Guix on a foreign distro <pkill9>that's the only message it returns with <verisimilitude>I'm having, perhaps, a similar issue with Emacs; it won't launch at all. <rekado_>pkill9: I’m using a custom st variant (different colours) and it works just fine. <pkill9>i wish the error provided more information lol <pkill9>oh well, I'll just use it on my native distro