Still Seeing Parallel Build Breakage In 10.1-BETA3

Glen Barber gjb at FreeBSD.org
Mon Sep 29 13:55:08 UTC 2014


[CC imp@]

On Mon, Sep 29, 2014 at 08:32:21AM -0500, Tim Daneliuk wrote:
> FYI -
> 
> Code Branch:
> 
> /usr/src>svn info
> Path: .
> Working Copy Root Path: /usr/src
> URL: svn://svn.freebsd.org/base/stable/10
> Relative URL: ^/stable/10
> Repository Root: svn://svn.freebsd.org/base
> Repository UUID: ccf9f872-aa2e-dd11-9fc8-001c23d0bc1f
> Revision: 272279
> Node Kind: directory
> Schedule: normal
> Last Changed Author: wblock
> Last Changed Rev: 272275
> Last Changed Date: 2014-09-28 18:15:18 -0500 (Sun, 28 Sep 2014)
> 
> 
> Buildworld is fine with -j4, but buildkernel blows out:
> 

This is fixed in head in revision r268376, and planned to be merged to
stable/10 soon (so it will be included in 10.1-RELEASE).

A random, but interesting datapoint:  the fts_read() failure seems to be
correlated to the processing speed (CPU and RAM) versus the speed of the
disks.

On one of the RE build machines, prior to r268376 in head/, I could
consistently reproduce this particular build failure when building with
'-j' values of 24+.  On one of the newer RE machines, both world and
kernel are build with '-j48', and I have yet to see this particular
build failure in the stable/10 branch.  The main difference between the
two machines is that the second machine mentioned is configured with
proper partition alignment and ashift=12 versus ashift=9.

Glen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20140929/e61304d6/attachment.sig>


More information about the freebsd-stable mailing list