Ltest uses 'lbuild' to make rpms. When there're multiple ltestds do 'lbuild' at the same time, there's a good chance in a kernel or lustre building race if they build the rpms in a same directory. So we should use separate buildroot to avoid such issue. Also make changes to get 'make rpms' and 'make srpm' work smoothly with this change to lustre.spec.in.

so let's pass the right macro definition to it.

the same rpm -- which ends up happening in the same directory. In order for
this race avoidance to work, the caller has to supply a unique TMPDIR which
rpmbuild will use at it's _tmpdir. Default to /var/tmp if not supplied.

on both b_new_cmd and b1_5, it's obviously not working for some cases:
http://buffalo.lustre.org:8080/display_report.pl?report_id=278455
I am not sure why though.  It would seem like lustre.spec.in is not getting
processed by configure, but all indications are that it is.  Why is the
@RPM_LIBIAM_PATH@ substitution not being done then?

b=10851

Includes new ost-survey script, fixes to obdfiler-survey from scjody, new echo.sh for obdfilter setup from eeb

- replace all memcmp of fids by lu_fid_eq;
- allocate new fid for create case in ll_revalidate_it() in case of doing "goto do_lookup";
- removed needless lmv_fld_lookup() in lmv;
- fixed lmv_obj refcount management in error cases;
- comments are adjusted to be coherent with 80 columns style.

Teach lustrecvs and buildcvs about b_release_1_4_7-test branch. sigh.

Add directories used by quilt to .cvsignore

Take two.

want packaged.

that it does has it's own set of problems.  ~sigh~

simply builds a vanilla 2.6.16.22 kernel for the client nodes.

Added scripts to parse output per Peter Braam

r=brian

Merge from b_release_1_4_6: Run update_oldconfig only if oldconfig fails;
always strace update_oldconfig so we can debug it.

use default mode for modified modprobe.conf instead of mktemp mode

remove lustre-ldap

  - change module name from "llite.ko" to "lustre.ko"
  - drop ldap and python from rpms
  - remove initd scripts from rpms

"kernel-lustre"[-smp] so that package management (i.e. yum) does not treat
our kernel equal to the vendor's kernel and allow inadvertently upgrading
from one to the other.
We still "provide" a "kernel" functionality, so in fact our
kernel-lustre[-smp] can be installed on say RHEL4 and then the vendor kernel
package can be completely removed because our kernel-lustre[-smp] will
satisfy all required dependecies with regard to "kernel".

build-core's HEAD is going to diverge for 1.6.0, so "capture" build-core's
current state in a branch mainly for b1_4.
Note that b_release_1_4_6-patchless is not really a release branch and thus
tracks b1_4.
Also, v1_2_8 and v1_4_0 pre-existed this build system so which branch to
check out for them is actually quite moot.  Let's just keep them tracking
b1_4.

aagh, broke old b1_4 stuff - wait until we branch lustre-build

- change module name from "llite.ko" to "lustre.ko" (r=adilger)
- drop ldap and python from rpms
- remove initd scripts from rpms
- remove old cluster_scripts dir