Ceph

There is no context about why this partial, hand made backport was considered useful: commit:d75a9e374e6581c57461018952ed125c929aeb30 it is a partial backport that is apparently fit for firefly.

Ken Dreyer wrote:

The RHEL product has strict requirements about support, stability, ABI
guarantees, etc. Our team's able to support the gperftools-libs package
for the Ceph product, but not for all of RHEL. Since
/usr/bin/ceph-dencoder was only thing in ceph-common that links again
libtcmalloc, if we could fix the over-linking there, then we avoid the
whole issue.

The downstream BZ that corresponds to that ceph.com Redmine ticket is
https://bugzilla.redhat.com/1186953 in case anyone's interested.

Boris provided us with the following hints on how the issue was fixed in master:

• http://tracker.ceph.com/projects/ceph/repository/revisions/fb11c74d1dc3843f2f5b6dca9c76278c5ceeca1c/diff/
• http://tracker.ceph.com/projects/ceph/repository/revisions/c2b3a35fb3cbf28d46a5427f32fbaff142c85f2a/diff/ (first chunk only)