Re: FreeBSD 10.0-CURRENT #0 r232730: buildworld broken with CLANG?

From: O. Hartmann <ohartman_at_zedat.fu-berlin.de>
Date: Sun, 11 Mar 2012 16:54:05 +0100
On 03/10/12 19:09, Dimitry Andric wrote:
> On 2012-03-10 17:11, Ivan Klymenko wrote:
>> В Sat, 10 Mar 2012 14:26:42 +0100
>> Dimitry Andric <dim_at_FreeBSD.org> пишет:
> ...
>>> Unfortunately, you did a -j build, which makes the actual errors
>>> difficult to find, and if you show only the last few lines, as you
>>> have done here, those errors are not visible at all.
>>>
>>> Try doing a single-threaded build instead.  Save the entire log, using
>>> script(1) for example, compress it and upload it somewhere.
>>
>> Full buildworld log:
>> http://pazzle.otdux.com.ua/logs/buildworld.log
> 
> This is, again, a multithreaded build log, so it is very difficult to
> see where the actual error is.  Moreover, it seems to be using ccache,
> which almost certainly result in problems, and non-standard CFLAGS. 
> 
> Try disabling all of these, deleting /usr/obj, and rebuild.
> 

Sorry for the noise I made especially. Reporting here a parallel make
via make -jX happened by my stupidity, I'm sorry. Mea culpa!

A "make buildworld" works on a six-core Intel i7-3930X, but it fails
while performing a "make -j24 buildworld", "make -j12 buildworld" and
"make -j6 buildworld". It fails only on this box.

So far, the initail problem has vanished, but a parallel build isn't
possible on a even more powerful architecture.

It seems that a "make buildworl" on a Core2Duao E8500 (2 cores/threads,
3 GHz, 8 GB RAM, P45/ICH10 SATA 3GB harddrive) takes approximately 100
minutes to compile world while the new box (Sandy-Bridge-E Core
i7-3930X, 6 cores/12 threads, 32 GB RAM, 3.2GHz, 460GB WD Caviar Black
HD attached to the SATA 6GB port) takes more than 125 minutes to compile
the same sources. Something is very fishy ...

Regards,
Oliver


Received on Sun Mar 11 2012 - 15:10:47 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:24 UTC