New Coverage Analysis Reporter and Runs

[Joel Sherrill]

Hi,

Glenn Humphrey of OAR has spent the past month or so working
on converting the coverage analysis from "chunk or blob" oriented to
function oriented.  This means you don't have to link the
RTEMS tests in a special way.  You just provide a list of
methods you want to analyse.  He also added branch taken/not taken
tracking but no simulator reports correctly on branches
yet so we don't trust that information.  When qemu gets that
right, we will trust it.

The first run of all BSPs was overnight. This is how long
it took on britain.  Britain is a Intel(R) Core(TM)2 Duo CPU
E8500  @ 3.16GHz so this is obviously a CPU intensive process.

STARTED AT:  Wed Mar 10 16:45:17 CST 2010
STOPPED AT:  Thu Mar 11 07:09:51 CST 2010

real    864m34.093s
user    638m59.196s
sys    163m17.980s

Yep. ~14 hours. :)

This generated 6 report sets for 9 BSPs for a total of 54
coverage configurations.  The BSPs tested are:

arm/edb7312
arm/gumstix
arm/smdk2410
i386/pc386
m68k/uC5282
powerpc/qemuppc
sparc/erc32
sparc/leon2
sparc/leon3

As we get other simulators that support coverage or
trace output, we will add them.  The plan is for this
to be run from cron once a week so we can track changes
in test coverage.

For those interested in performing coverage analysis on
their own applications, we are now in a position to
help you do that.  If interested, please contact me.

--joel