*jgrant might try to nap again, a bit, before the GSD talk.
<jcca>Hi, Is there a way to inject distcc into guix? I mean DISTCCPATH/bin into guix PATH
<rekado>hmm, I just cloned the git repo on a different machine, ran ./bootstrap, ./configure, make, sudo make install, started the daemon, but when I want to build a package I get this error: "guix build: error: open-file: No such file or directory: "/gnu/store/vbcm1...-guile-bootstrap-2.0.drv"
<rekado>I already tried rebuilding after make clean, but to no avail.
<jgrant>It's mostly not being able to shut the ol' brain off, for good or bad. Usually it's just a lot of random blabber, sometime it's hyperfocused and clear thought. Kinda hard for me to discriminate to one side or another.
*jgrant operates pretty well on 4 hours of sleep though, really. It's until you get to that 2 hour range for a few days in a row -- that's where things are apt to get promblematic. :^P
<taylanub>is it fine to just take synopsis and description from Debian packages?
<mark_weaver>taylanub: I often look at Debian's synopsis and description, and there's no legal problem with copying them wholesale in principle, but in general theirs would not always conform to our conventions and requirements.
<mark_weaver>we have to comply with the FSDG for example, and we try to avoid stuff that sounds like marketing-speak, etc.
<mark_weaver>and we have some conventions that guix lint checks for.
*jxself imagines mark_weaver wandering off into the sunset
<DusXMT>Does anyone know why a cross compiler (or cross linker used by a cross-compiler) is having trouble finding shared libraries? The dynamic linker is set up properly, I can see the correct value if I edit xgcc, and the path with the libraries is even added to the compiler's search path, so I don't know...
<DusXMT>Manually specifying them on the command line makes it compile without problems
<DusXMT>Specifying the glibc library folder with --rpath makes it work... hmm...
<mark_weaver>DusXMT: you might try running the cross-gcc to link a simple test program with the -v option, so that it shows exactly what options are being passed to the subprograms (including the linker)
<mark_weaver>and compare that with what happens with one of our cross compilers that works properly.
*mark_weaver looks at the relevant part of the GCC source that would be used on Hurd to find its dynamic linker.
<DusXMT>I'm trying, I'm just building a linux cross compiler
<mark_weaver>DusXMT: also note that the C preprocessor macro being patched in the other cases (GLIBC_DYNAMIC_LINKER*) doesn't match the macro used in gcc/config/i386/gnu.h (GNU_USER_DYNAMIC_LINKER). did you fix that too?
<DusXMT>mark_weaver: I've fixed that as well, the dynamic linker is found
<zykotick9>trying a qemu install of GSD and running into an issue at step #2 (i believe) - the "deco start cow-store /mnt" resulting in "impossible" https://bpaste.net/show/b9741c19cdd9 /mnt is mounted, and a file can be touched in there.
<mark_weaver>civodul is really the person to ask. I'm just muddling through in his stead.
<DusXMT>Looking at native binaries, the libc and libgcc paths are in their runpaths...
<DusXMT>If one was to remove these, what would happen?
<DusXMT>that's a silly question probably, the libraries simply wouldn't be found...
<DusXMT>I think I know what's happening now: on the Hurd, libc needs other libraries in order to function, ie. libmachuser and libhurduser. In order to find these, it needs to have the runpath set. This problem doesn't occur on GNU/Linux, as libc works on its own
<DusXMT>This problem occurs when building libgomp, a library of gcc
<jgrant>zykotick9: Can you list the commands you ran in-regards to creating that partition and mounting it to /mnt?
<zykotick9>jgrant: "fdisk /dev/sdb" created one partition, then "mkfs.ext4 /dev/sdb1". there is currently a lost+found AND a /tmp/guix-inst directory on it.
<zykotick9>jgrant: sorry, the mount was "mount /dev/sdb1 /mnt"
<mark_weaver>I've never tried installing Guix in a VM, so I'm afraid I would not be of much help here, but I can tell you that it's trying to call: mount ("/gnu/store", "/.ro-store", "none", MS_BIND | MS_RDONLY, NULL);
<mark_weaver>are you using a kernel we provided? maybe it lacks some features needed here?
<DusXMT>mark_weaver: Well, this is the otuput I got when trying to compile the conftest.c file it choked on when runnign configure for libgomp (part of gcc, ld-wrapper isn't available yet): http://paste.lisp.org/display/145508
<mark_weaver>zykotick9: see what I wrote above. to debug this, we need to figure out why that system call is failing.
<DusXMT>(Note: both libmachuser.so.1 and libhurduser.so.0.3 are in /gnu/store/pv4n9fhyjsv8ghl34n1fnvqirvp4lajs-glibc-hurd-cross-i686-pc-gnu-2.18/lib/)
<zykotick9>mark_weaver: insanity is trying the same thing and expecting different results, BUT i tried command again, 1st time "fuse: bad mount point '/.rw-store': No such file or directory" 2nd time it seems to have worked? or at least, no error output?!?! thanks all.
<mark_weaver>DusXMT: I don't know where we add the rpath for glibc, but whereever that is, I guess we need to add more rpaths for those other libraries that glibc depends on. alternatively, it might be better for our glibc package to copy those other libraries into it.
<mark_weaver>we should ask civodul what he thinks before committing to a particular strategy though.
<DusXMT>mark_weaver: at that point, there's no rpath added yet, and they're in the same folder as libc.so
<davexunit>did anyone happen to stream ludo's talk today?
<jgrant>davexunit: Nope, the whole room didn't have streaming until after if at all.
<davexunit>jgrant: not surprising. thanks for the update.
<jgrant>I stopped waiting after 35 minutes into the 45 minute talk, after a whole night of general annoyance. :^P
<jgrant>davexunit: We only have to wait 10 weeks, to 4.5 months now. :^)
*jgrant wonders if it'd be easier to define a trivial-build-system for these font-league packages ... they have 0 build instructions. I need to be able to extract each one in the desired end directory, wun mkfontdir in said extracted direcory, and just genrally run fc-cache.
<taylanub>this test suite uses ps(1). should I add a dependency or disable the test suite?
<grasshopprWhoppr>A couple things desired in the installation: Colemak and wireless instructions.
<jgrant>grasshopprWhoppr: Yeah, we need loadkey or something for colemak. I can write a basic tutorial for wpa_supplicant if you want. :^P
<jgrant>The alternative keyboard thing is relavitely understood as an issue, me thinks. I was tasked with adding a feature request to GNU's bug tracker on this issue, haven't yet -- though it's on my list.
*grasshopprWhoppr is reading the arch wiki, but one of us could submit steps for manual.
<jgrant>grasshopprWhoppr: Where you you want this guide to be, for Wireless? On the Libreplanet, or actually shippy on the install image?
<jgrant>It probably wouldn't be a bad idea to dedicate tty2 to a relatively indepth install guide with general methods to set up nicities like network configuration, keyboard, etc. Then move the stardard manual to tty3. The "Begineer Guide" for Archlinux seems pretty popular and helpful in-route to stepping people through a foreign environment with relative ease.
<jgrant>grasshopprWhoppr: I have no decsions on this matter, for I have no real power of influence though -- might be something you should ask civodul about at somepoint here, or on the ML.
<jgrant>Personally, I'm all for making the installation instructions more expansive -- maybe just not in the cannonical docuentation.
<grasshopprWhoppr>jgrant, I'll do that. The arch/parabola instructions start with one page that, incidentally, say how to loadkeys and connect, and then, you continue through the web.
<jgrant>grasshopprWhoppr: You're refrering to the begineer guide, no?
<mark_weaver>jgrant: running fc-cache within the package build is no good, because it will not have access to the user's home dir or any other fonts the user has installed. fc-cache needs to be run in an environment where all the user's fonts are available.
<jgrant>Well, mosh is complaining when I try to connect (having problems staying connected with pure openshh) of "/gnu/store/xxxxxxxxxxxxxxxx-mosh-1.2.4/bin/.mosd-real: Did not find mosh server startup message."
<mark_weaver>we have only one ssh-compatible service at present, and that is 'lsh-service'
<jgrant>mark_weaver: I must not be clear, on the partion that has Archlionx on -- I am installing Guix to boostrap install GSD. On the Archlinux install that I'm using to bootstrap, the sshd daemon needed to be restarted for the system to realize that mosh was installed.
<jgrant>So it wasn't an issue with Guix on my local box, but Archlinux that I'm using to install Guix on my VPS.