Tools » teuthology

gitbuilder-ceph-deb-precise-amd64-notcmalloc is made. It looks like it had a few problems due to using master preventing the upstart job from starting.... Adding build-auto.sh from the wip-auto branch made the upstart job start but then it was failing.

It looks like the problem was with a link pointing to a link and dirname was getting the wrong dir:

+ pwd
+ pwd=/srv/autobuild-ceph/gitbuilder.git/build
+ dirname ../build.sh
+ mydir=..
+ hostname
+ grep q ^gitbuilder
+ hostname
+ grep q - notcmalloc
+ hostname
+ grep q - gcov
+ hostname
+ grep q - deb
+ exec ../build-ceph-deb-native.sh
../build.sh: 20: exec: ../build-ceph-deb-native.sh: not found

Changing auto-build.sh from:

mydir=`dirname $0`

to:

if [ -L $0 ]; then
    myfile=`readlink -f $0`
    mydir=`dirname $myfile`
else
    mydir=`dirname $0`
fi

Fixes the problem and a build was going.

gitbuilder-ceph-deb-precise-amd64-gcov gave me problems. During the fab process it died in the middle:

[ubuntu@gitbuilder-ceph-deb-precise-amd64-gcov.front.sepia.ceph.com] out:  * Starting NTP server ntpd
[ubuntu@gitbuilder-ceph-deb-precise-amd64-gcov.front.sepia.ceph.com] out: * glibc detected  lockfile-create: malloc(): memory corruption (fast): 0x0000000001c8d0e0 **

And ssh'ing into the machine said it required a reboot. Upon rebooting it would no longer boot due to hanging during cloud-init. Had to re-create the guest and giving this another try. I will not reboot if it has the same problem again.

Added both to the proxy and gitbuilder.cgi in the meantime.

New guest ran into the same stuff but I was able to get past it thistime with some dpkg foo. The gitbuilder-ceph-deb-precise-amd64-gcov is up and building.