<dftxbs3e>samplet: http://dpaste.com/0GSDXXM - output of: "./pre-inst-env guix build --target=powerpc64le-linux-gnu bootstrap-tarballs --keep-failed 2>&1" - for some reason, running a second time after the first ranlib failure gives this result but it may be interesting.. uploading config.log in a sec
<switchy>is there a way to not use substitutions only for the final derivation, and not its dependencies? I want to build netcdf-fortran with gfortran 8.3.0, since it seems like the version I got (however that was) isn't built with that gfortran version, and so doesn't work...
<switchy>oh, I guess gfortran is defined as gfortran-5
<samplet>switchy: You could change the package definition to use gfortran-8 instead of the default gfortran.
<samplet>switchy: I think even using “--with-input=gfortran=gfortran@8” would do it.
<switchy>so if I do `guix gc -R $(readlink -f ~/.guix-profile)', that hdf4-alt built with gfortran@5 pops up -- the one built as a dependency of netcdf-fortran isn't there. I guess that explains that bit?
<reepca>although I actually just found out that I was in reality trying to backup a source release. It took me about 15 tries before I finally looked in the original directory and saw it didn't have a .git either.
<reepca>how exactly do search-paths work in builders? For example, if a package has as a search-path (search-path-specification (variable "FOOPATH") (files '("foo/"))), will the set-paths phase of builders just add the foo/ subdirectory of every single store item that has one and is either an input to the derivation or referenced by the inputs?
<reepca>I've got a strange situation where I'm in an environment that has a superset of the inputs necessary to build a package, but manually building it won't work because something isn't found. Meanwhile the proper build succeeds, I assume because something one of the inputs references (but that isn't an input) gets picked up by an environment variable.
<g_bor[m]>I was trying to modify our openssh service to change the activation snippet into a one-shot service to see if it helps the openssh does not start on boot problem.
<g_bor[m]>I am quite sure, that I am doing something wrong, but the current behaviour of shepherd/guix system reconfigure seems to be problematic when errors occur.
<g_bor[m]>I got a backtrace after the 'to complete the upgrade run...' message, then an error in procedure scm-error, wrong type argument, and a warning: while talking to shepherd: no such file or directory, and finally an unresposive vm.
<ruffni>got my guix sd to work. only question: browser of choice is icecat? is it possible to have js running? because i can't even chat with you here (via webchat.freenode.org) without js..... and even disabling all the plugins and scrolling through security settings still won't make it work. also icecat seems to not show numbers of any kind (sounds weird and believe me, it is). searching for bz2 results in bz (2 replaced by a noticably
<ruffni>ah yeah... stupid of me. thanks! freenode webchat works now :) but what about those numbers? has anybody else experienced such a thing? i can type a literal 2 into location-bar and theres just this wide white-space
<maddo>well common libraries detection doesn't really work well, since sometimes headers and js post-processing will make it fail
<rekado>a simple test with dune, however, simply produces no output at all
<rekado>we also found that we have to tell mpirun to ignore openib nodes when we want to run on omnipath only or else it will mix nodes and communication will obviously fail on those that have infiniband.
<g_bor[m]>I've found out why my previous diffs did not work.
<g_bor[m]>In the openssh-activation we have a define, and shepherd hangs on that, giving me a backtrace about definition in expression context, and the whole machine becomes unresponsible. I fixed it by expanding the definition inline, it is almost trivial, but this still leaves two open questions:
<g_bor[m]>1. I believe shepherd should simply fail to start the service in this case, print the stacktrace, and resume operation. Wdyt?
<g_bor[m]>2. It seems that somehow gexps do not compose nicely, or rather I failed to find the way to do this properly. Any idea how should I fix this definition in expression context error? This might be problematic if we have more complicated services, this restriction on defines does not seem to be ok. Wdyt?
<g_bor[m]>Another thing, changing the openssh-activation to a one-shot shepherd service does not solve the openssh not starting on boot problem.
<bgardner>Hey guix; so I finally got my postfix.scm to the point where it builds but it turns out Postfix has an interactive installer that looks pretty incompatible with Guix without some fairly radical voodoo that is over my head. Would it be beneficial to submit the postfix.scm as a WIP anyway for a smarter person to run with?
<g_bor[m]>bgardner: I believe so. There is some interest in getting postfix. I belive that the installer does not do much more than generating config files. If that is so, then we should provide a service to generate those.
<bgardner>g_bor[m]: Okay, I'll submit the patch in just a bit then. Thanks!
<g_bor[m]>ok, a bit more reading gives me this: we have to generate a main.cf, and then run make upgrade instead of make install.
<g_bor[m]>I will have a look at that makefile later :) sounds like fun...
<civodul>rekado: re gfortran, i agree, they should be matching
<civodul>g_bor[m]: indeed, you cannot compose gexps in arbitrary ways, but that's the same with macros: they let you generate invalid code
<bgardner>g_bor[m]: That postfix.scm hsa been submitted to guix-patches, thanks again
<joshuaBPMan>OpenDocument export failed: Buffer is read-only: #<buffer styles.xml>
<terpri>joshuaBPMan, i get a similar error with a customized package for emacs 26.1 on guix system (typing C-c C-e o o in an empty org document -> "OpenDocument export failed: Buffer is read-only: #<killed buffer>")
<reg[m]>Hi everyone. Yesterday I tried Guix for the first time and I'm really excited about it. I'm facing one issue, though: when I choose to encrypt the disk during installation, the system just won't boot. The Grub menu shows up, then disappears and I'm stuck on the Grub background image. Has anyone else seen this? I'm using the x86_64 GNU Guix System 1.0.0 iso on a Lenovo T430 laptop. The problem doesn't occur without encryption.
<affinespaces>I just ran into a stacktrace when trying to setup wifi in the installer, sent the report over to the e-mail address a few minutes ago. will try without encryption too to see if that works
<g_bor[m]>vagrantc: yes, that would be a nice addition :)
<reg[m]>g_bor: I'm not sure, I just used the default partition scheme (tried both with and without separate home folder).
<reg[m]>g_bor: I mean the default partition scheme as proposed by the graphical installer.
<affinespaces>I get the stacktrace (for a different issue) whether or not I use encryption or the separate home partition
<g_bor[m]>reg: ok, I will have to have a look around. I did not notice the issue, but I did not do this using the graphical installer yet.
<g_bor[m]>If someone with some hands on experience is here, then please notify us :)
<reg[m]>g_bor: Thanks! If there's anything I can do, let me know.
<tao[m]>I was just running through all the packages Guix doesn't have to decide if there's anything I can'ti live without. It was going fine until I hit `fzf`?!
<g_bor[m]>yes, I have also seen several sorting related commits, not one on this page.
<sirgazil>g_bor[m]: By the way, if you are planning to install on a real machine using the graphical installer, I recommend waiting for the 1.0.1 release. I found some bugs in the installer that made things hard.
<g_bor[m]>I am only doing a test install on a vw to have a look at this encryption problem reg reported. Do you have anything related to that on your list of bugs?
<sirgazil>g_bor[m]: Not to encryption, no. Also some of the bugs I exprerienced using the graphical installer in a real machine were not visible in a VM I used for testing.
<nckx>How can I ‘guix system reconfigure’ without immediately applying all changes? For some reason guix really likes to restart X for me… :-/
<nckx>(Hence killing itself before it can install the bootloader.)
<g_bor[m]>reg: I just did a fresh install with encryption in a vm, and I can't reproduce the problem there. Also, sirgazil told me that some things behave differently on bare-metal and vm, but I don't have a spare computer to test on right now.
<reg[m]>g_bor: Ok, thanks for checking! Is there any way I can provide you with logs or anything? If so I'd be very willing to.
<nckx>tao[m]: Yes! Flatpak on Docker on Nix on Guix on GNU on Linux is the way of the future!
<g_bor[m]>reg: are you getting dropped to the GRUB command line?
<retiform>How do you get opengl applications to work on non-native systems?'
<vagrantcish>gdm is still running ... but maybe not after a reboot
<nckx>reg[m]: OK, so it's not the obvious (c on another key), but maybe you're being hit by something similar to my <https://debbugs.gnu.org/cgi/bugreport.cgi?bug=35583> (which I haven't investigated further, sorry). Seems unlikely™ that the keyboard not working has anything to do with your decryption troubles, but there have been stranger bugs.
<terpri>rekado, might be a guile error, from return_unused_stack_to_os in libguile/vm.c
<nckx>I've installed AbiWord and it doesn't crash, but I get high-frequency flickering artefacts in the dark grey area surrounding the page that I can't screenshot but make the application impossible to use without going mad.
<mfg>Hi, i just finished installing Guix 1.0. I then ran `guix pull && sudo guix reconfigure /etc/config.scm`. I noticed some warnings about /root/.config/guix/current not existing, which is right because i did not use sudo guix pull. But the Documentation says that sudo guix reconf... should use the User which runs that command, so this warning seems like a mistake or the Documentation is wrong ?
<nckx>mfg: You can check that hypothesis with ‘sudo which guix’.
<mfg>okay that seems correct it uses the guix from my home folder; Why the warning then ?
<mfg>Hm okay. I don't know Debian that good either :)
<nckx>The ‘building’ you see when you run ‘guix pull’ is just compiling the Guile Scheme code that these definitions themselves are written in, which does take much longer than other package managers. That will hopefully improve some day.
<mfg>well it's just 1.0 there will be time to optimize things, i guess :D
<nckx>mfg: Most sane package managers have a separate ‘update the view of the world’ and ‘install a package from that world’ commands. Installing the latest thing at every new invocation might sound nice but really doesn't work in practice.
<nckx>Source: all ‘package managers’ that try to do this. 😛
<str1ngs>yeah, everything has its own name. ie substitute vs binaries :P
<nckx>I guess it's not a problem in practice but, for example, I just had to check whether ‘-u’ in Guix stood for ‘update’ or ‘upgrade’ because the two are so close. It would take me a few invocations on Debian to learn which is which.