The failure that's displayed for a library and toolset is the most
significant among all of the failures detected for the toolset. The order
of failures according to which the specific type is choosen is (from the
most significant):
comp > link > run > fail > cerr > time > file
Order the cells in legends according to the failure significance. This
is consistent with the older ordering of the legend elements. They are
sorted from pass to regression failure.
I am mostly interested in Boost.MPI for which we do not have a lot of testing done.
Boost.MPI depends on the underlying MPI implementation so I might have a
few different flavors.
sources) to make for easier to use build and use. Still in progress..
* Added build options and logic to set boost-root, boost-build, and
install location.
* Added general build script to install all the various testing utility
programs.
* Refactored process_jam_log so that it can be shared and incorporated
directly into other programs (i.e. library_status).
* Redoing library_status to be single program run that does the complete
build+process+generate data for a library.
