IRC channel logs

2013-11-07.log

back to list of logs

***sea``` is now known as sea`
<civodul>Hello Guix!
***civodul changes topic to 'GNU Guix --- http://gnu.org/s/guix/ --- 0.5 is in the works, and you can help!'
<civodul>here we go :-)
<mark_weaver>well, what a damn drag. after approximately 21 hours of building gcc-4.8.2 (it takes about 22 hours on my machine), my new kernel locked up, so I have to start the gcc build over from scratch.
<mark_weaver>and I have at least two more of those to build before the bootstrap tarballs are built: one static and one not static.
<mark_weaver>so the bootstrap tarballs are still several days away, I
<mark_weaver>I'm afraid.
<mark_weaver>(my old 2.6.39.1 is rock solid, but the m4 test suite does some obscure test of readlink("") that doesn't return the expected error code on my old kernel, so I have to use the new less stable 3.10.15 kernel sometimes too)
<viric>even using 2.6.39.1 kernel headers?
<viric>I can't say I ever had stability problems with the fuloong
<mark_weaver>the kernel headers are irrelevant. the issue here is what error code 'readlink("")' returns
<viric>other than the ext4 thing, and a crash-on-boot
<mark_weaver>viric: what kernel are you using?
<viric>3.6 or 3.7, I can't remember
<viric>I don't turn it on since months
<viric>But I used almost all kernels since 2.6.39
<mark_weaver>have you use 3.10.x?
<viric>2.6.36 even
<viric>n
<viric>no
<mark_weaver>okay, so maybe the problem is recent.
<mark_weaver>I guess I'm going to have to get access to the YeeLoong's serial console, so that I can debug this.
<mark_weaver>as things stand now, I have zero information about where the problem is. Nothing is printed, even if it happens when I'm using the text console.
<mark_weaver>and the problem is hard to reproduce. I was running that new kernel for a few days before it locked up.
<viric>ah
<mark_weaver>I wonder if the SysRq key stuff would work. I've never learned how to use those key sequences.
<viric>alt-sysrq-h shows help :D
<viric>and then, alt-sysrq-whateveryouwant
<viric>1) check if the help message appears
<viric>if so, 'w' prints the locked processes and their stack backtrace...
<viric>if not.. maybe you can increase the console loglevel (1, 2, 3 ...)
<mark_weaver>viric: thanks for the tips. alt-sysrq-h does work on the text console.
<viric>some people learn sequences of letters, instead of looking at help.
<viric>I always call 'h', and then decide what to do
<mark_weaver>when I type alt-sysrq-w, the output is very large, and most of it scrolls off the screen before I have a hope of seeing it.
<viric>hm
<viric>bad then.
<viric>one thing would have been to use netconsole, but it's late in your case
<mark_weaver>given how locked things are, I'm not sure anything that uses net will still work.
<mark_weaver>I can't switch virtual terminals, for example.
<mark_weaver>(when it locks up)
<viric>yet alt-sysrq-h works?
<mark_weaver>I don't know, I never tried it when the machine was locked. (I've since rebooted it)
<viric>there is u for 'unraw' the keyboard. Should bring back normal key handlers.
<viric>ah ok
<mark_weaver>alt-sysrq-h works right now, on a non-locked machine, is what I meant.
<viric>set up the netconsole
<viric>you may have relevant messages. the netconsole is quite a lowlevel thing.
<viric>for example, if you have a "null pointer resolution", it will vomit it to udp
<mark_weaver>I probably just don't have the resources to set up that kind of thing. I only have one working local machine at the moment.
<mark_weaver>I guess someone else will have to help debug this.
<mark_weaver>or perhaps I'll just do a git bisect that is likely to take several weeks or months to complete :-(
<civodul>mark_weaver: ouch, too bad the thing crashed
<civodul>i'm going back home and should be able to catch up tonight
<civodul>hey hey!
<jxself>You're back!
<jxself>I missed you.
<civodul>oh!
<civodul>:-)
*sea` walks past jxself a few times