IRC channel logs


back to list of logs

<mark_weaver>well, after several failed attempts, I finally built a new kernel for the YeeLoong 8133 that works and does what guix-daemon needs, and am now bootstrapping Guix on the Loongson 3A.
<mark_weaver>these builds are going a *lot* faster
<civodul>Hello Guix!
<civodul>mark_weaver: i'm downloading the new bootstrap tarballs!
<civodul>great that the process eventually completed ;-)
<civodul>welcome back a_e!
<a_e>Thank you!
<civodul>brand new stuff at
<civodul>thanks, mark_weaver
<mark_weaver>hi civodul!
<mark_weaver>I noticed that core-updates is still using gcc-4.7.3 as the default compiler.
<civodul>because it needed the new bootstrap gcc to switch to 4.8
<civodul>so i'll be trying that now
<mark_weaver>I tried reverting the commit that reverted GCC 4.8 support, but the build of gcc-cross-boot0 still failed.
<mark_weaver>(when using the new bootstrap tarballs)
<mark_weaver>specifically, the configure script in the bundled mpc fails to find the built bundled mpfr.
<mark_weaver>the reason it fails is because the GCC build system is passing a bad option to mpc's configure. it's passing --with-mpfr-include=/tmp/nix-build-gcc-cross-boot0-4.8.2.drv-0/gcc-4.8.2/mpfr/src/src
<mark_weaver>notice the "src/src" at the end, which is wrong. there should be only one "src" there.
<mark_weaver>so mpc can't find mpfr.h, because it's looking in the wrong directory for it.
<mark_weaver>I guess I should mention that this failure happened on MIPS, but I doubt that matters. the problem didn't happen until I reverted back to gcc-4.8.2
<mark_weaver>civodul: will you build bootstrap tarballs for x86_64 and i686 that have the same date?
<mark_weaver>(same date in the URL on, that is)
<mark_weaver>(the GCC build system also passes a bad '--with-mpfr-lib' option, with the same mistake)
<mark_weaver>djbclark: btw, after several failed attempts, I finally built a new kernel for loongson 3a that works properly and does what guix-daemon needs, and started using it to bootstrap guix.
<mark_weaver>djbclark: along the way, I moved /home of LOonix to the root partition, and installed vanilla Debian Wheezy on the partition that used to be /home. that's the host system that I'm now using to bootstrap Guix.
<mark_weaver>(so LOonix fully intact and still boots by default)
<mark_weaver>*is fully intact
<mark_weaver>I seeded the Loongson 3A machine with the /nix/store from my Loongson 2F, which contained almost everything that currently builds on the 'loongson' branch. Everything I've tried seems to work perfectly on the 3A.
<a_e>mark_weaver: This sounds all very good!
<mark_weaver>hi a_e! welcome back :)
<a_e>As I understood, I had better wait some more before rebuilding guix on the loongson?
<a_e>Until the new bootstrap binaries are activated?
<mark_weaver>right, probably so.
<mark_weaver>I already have a private patch to do just that, but then there's the other problem of upgrading to gcc-4.8.2
<mark_weaver>which apparently still has some snags.
<mark_weaver>it probably makes little sense to build from core-updates now until gcc is upgraded.
<mark_weaver>I must say, for someone accustomed to the performance of a modern intel machine who wants to get away from intel, the Loongson 3A laptop is a serious contender. it's vastly more performant than the 2F.
*civodul is back
<civodul>mark_weaver: re buildling the x86_64/i686 bootstrap tarballs, yes, i'll do it
<civodul>same date
<civodul>i have the MPFR hack in a stash too
<civodul>mark_weaver: so you have a 3A now? cool!
<civodul>djbclark: thanks for your support!
<mark_weaver>yes, djbclark kindly lent me his 3A machine.
<viric>is 3a the quadcore?
<viric>is it fast?
<viric>good :)
<mark_weaver>it's quite powerful enough to make a decent development machine.
<civodul>is it widely available?
<mark_weaver>well, it's available from
<mark_weaver>I don't know what you mean by "widely" available.
<civodul>that's what i meant :-)
<civodul>i was wondering if one had to travel to China and hide it in the luggage :-)
<mark_weaver>it's on sale for a significant discount until Nov 18
<civodul>so it doesn't need any non-free blobs?
<mark_weaver>the 3A machine *does* require a blob for the video, unfortunately.
<civodul>argh, ok
<mark_weaver>I still have yet to fully investigate how functional it would be without any blobs.
<mark_weaver>I vaguely recall that there's a blob in the (otherwise free) BIOS.
<mark_weaver>however, it seems that it can be used with a deblobbed linux.
<viric>are bzr revisions specified by an integer?
<jxself>Much nicer than hashes.
<viric>are those unique?
<viric>per project
<jxself>You mean can two separate projects have a revision #60? Sure.
<djbclark>mark_weaver: sweet re 2f/3a compat etc
<mark_weaver>djbclark: indeed! debian wheezy, while far from optimized, works quite well on the 3A.
<mark_weaver>although for some reason sound doesn't work.
<mark_weaver>I have yet to figure out what's different about LOonux that allows the sound to work properly.
<djbclark>mark_weaver: this is the new n64 debian build that they just finished recently?
<mark_weaver>no, I'm currently using stock debian wheezy.
<mark_weaver>I wanted to stick to stock stuff as much as possible, in the hopes that it will teach me what needs to be done on Guix.
<mark_weaver>but in terms of seeing what the hardware can do, that's certainly a good one to try at some point.
<viric>I'm testing filegive with ngrok... Amazing, it fakes the SSL certs!
<viric>ah or I'm wrong. :)