distcc compilation problem in kernel 3.16.0 (Debian)[SOLVED]

[bodhi]

Hi guys,

I'm using this combo for all server and client nodes, all run Debian kernel 3.16.0:
$this->bbcode_second_pass_code('', 'distcc 3.1 arm-unknown-linux-gnueabi
gcc version 4.6.3 (Debian 4.6.3-14) ')

I've noticed my kernel build lately has a lot of distcc COMPILE_ERROR in the remote nodes. The same files were compiled OK on the main server after it fell back. But it does not fail all the time, perhaps about 40% of the jobs were compiled OK on remote nodes.

Anybody seeing this? I thought it might be helpful to double check Arch to see if it's my tool chains problem!

Thanks,
bodhi

[moonman]

Are the toolchains the same version? I.E. remote and local versions?

[WarheadsSE]

Definitely sounds like something is out of sync

[bodhi]

Gcc and distcc are the same. Anything else needs to be the same? I did apt-get uprade for all nodes about the same time. What other tools should I verify? Thanks!

[bodhi]

I've made sure the same kernel, distcc, and gcc versions are the same. Still seeing this error once in a while . Would install kernel headers help in anyway (I never did on remote nodes)? I have also looked at package list with egrep -i '(gcc|libc)') and made sure they matched.

The strange thing is only perhaps 30% to %40 of files failed to compile remotely, but compile OK locally. Are you guys compiling Kirkwood kernel 3.16+ natively with distcc?

Thanks,
-bodhi

[kmihelich]

Zero errors on any kernel, all built from ARM devices, distcc to x86. The distcc logs on the x86 machine may be more enlightening, perhaps with --verbose given to distccd.

Also, Debian people would probably be more helpful when you're debugging things on Debian with a Debian toolchain. This thread's subject is misleading, as if you're using our tools or distribution.

[bodhi]

$this->bbcode_second_pass_quote('kmihelich', 'A')lso, Debian people would probably be more helpful when you're debugging things on Debian with a Debian toolchain. This thread's subject is misleading, as if you're using our tools or distribution.

[bodhi]

Thanks to all those who replied! I did what kmihelich suggested, turning --verbose on (both client and server). Still not apparent what the problem was, it just failed with internal error 0x100.

So I downgraded the kernel on the distccd boxes (client and server) to 3.14.0. And everything is working fine again.

-bodhi

[moonman]

What kernel were you running previously? It could be related to the problem people are having with 3.16

[bodhi]

moonman,

Yes it could be, I was running Debian kernel 3.16.0 on all these nodes, trying to compile the next version of 3.16.0.xx. All remote distcc nodes are booting from an NFS root which resides on the compiling node SATA HDD (GoFlex Net). The protocol is NFS v3. All of them are connecting to the same Gb switch. However, the distcc remote nodes only write to their /tmp (i.e. does not use NFS to read/write temp files or log).