IRC channel logs


back to list of logs

<c107>`whereis guile` returns `guile-2.0`. I don't understand why `configure` reports its deficiency.
<c107>I can't `configure` Guix.
<c107>Why won't `configure` see my Guile 2.0 in `/usr/bin`?
<civodul>Hello Guix!
<civodul>new translation!
***sea`` is now known as sea`
<mark_weaver>civodul: I met with djbclark today. He very generously lent me his YeeLoong 8133, which contains the newer and much faster Loongson 3A processor. (apparently it's about 3-4 times as fast).
<civodul>woow, sounds great!
<civodul>so you'll put it to good use? :-)
<jxself>djbclark is seeing people in real life?
<mark_weaver>yes, of course :)
*jxself feels shockec
<jxself>er I can't spell...
<jxself>Well, I am happy for him then.
<mark_weaver>jxself: I think you probably know more than I do about that, but yes, he agreed to meet with me, and we hung out for several hours. It was quite pleasant! We get along very well.
<mark_weaver>apparently he still hangs out at Pika quite a lot.
<mark_weaver>apparently the 72-core MIPS box is a bit of a strange snowflake.
<mark_weaver>he seems to be under the impression that it's more meant for MPI stuff.
<mark_weaver>it might be better to just get a YeeLoong 8133 for use as a hydra build machine.
<mark_weaver>but first I have to figure out whether it's possible to run the thing at all without the non-free video blob.
<mark_weaver>He's also open to further investigation of the 72-core MIPS box, but apparently that will be a bit of a project.
<civodul>wouldn't a Fuloong be better as a build box?
<civodul>(YeeLoong is laptop and Fuloong is box, right?)
<mark_weaver>yes, that's right. Well, djbclark certainly has at least one Fuloong that he's willing to donate, but it's no faster than the YeeLoong I'm working on now. So that means around 22 hours to build gcc.
<mark_weaver>If there was a 3A-based Fuloong, that might be a good choice. I'm not sure off hand if such a thing is easily available outside of China.
<mark_weaver>I've seen a picture of a Loongson-based rack-mount server, but I've not seen such a thing easily available outside of China.
<civodul>twenty-two hours.
<mark_weaver>well, I'm judging solely by the timestamps of the build logs in /usr/var/log/nix/drvs/
<mark_weaver>and I should mention that although several of the gcc builds have taken about 22 hours, some others have taken closer to 18 hours. I'm not sure why the difference.
<mark_weaver>maybe because some are building the bundled gmp/mpfr/mpc, and others or not?
<mark_weaver>s/or not/are not/
<mark_weaver>anyway, a long time :)
<mark_weaver>actually, I think djbclark would be willing to donate multiple Loongson 2F-based boxes for this purpose, if that would be helpful.
<mark_weaver>He has a big stack of YeeLoongs lying around.
<civodul>mark_weaver: you could grep for "`build' phase succeeded" in the logs
<civodul>that tells how long it took
<mark_weaver>ah, indeed it does!
<mark_weaver>does that measure real time, or cpu time?
<mark_weaver>the reason I ask is because I've often been hibernating to disk during the builds.
<mark_weaver>(whenever I have to bicycle somewhere)
<mark_weaver>civodul: okay, looking at those, I see that the best build time for gcc-4.8.2 is 18.14 hours, and the best time for gcc-4.7.3 is 20.28 hours
<mark_weaver>and gcc-static-4.7.3 took only 16.37 hours to build.
<mark_weaver>heh, well, I'm glad to see it's much less. I still have to build gcc-static-4.8.2 in order to get the new bootstrap tarballs. that's the biggest remaining build.
<mark_weaver>so hopefully I'll have them in 2-3 days.
<civodul>cool :-)
*civodul -> zZz
<civodul>good night/day!