I was getting an configure time error from CMake 3.19.1 with spaces before the
math library in Trilinos_EXTRA_LINK_FLAGS with the
sems-rhel7-clang-7.0.1-openmp-shared-release-debug build on by COE RHEL7
machine.

Calling string(STRIP ...) takes care of the problem.

…/TriBITS#299)

New TriBITS using modern CMake targets no longer sets the INCLUDE_DIRECTORIES
directory property but instead passes along include dirs using
INTERFACE_INCLUDE_DIRECTORIES target properties (like a good CMake project).
This was a strange test that does not actually link to the Sacado library but
yet needs a bunch of Sacado include dirs?

We need to look into this and see how to make this work more cleanly either in
Sacado or TriBITS because this is unfortunate.

…/TriBITS#464)

Before, this was passing in:

   FAIL_REGULAR_EXPRESSION "FAIL;BUMMER"

which just happened to be added as a list to the test property with the old
TriBITS.  But with the new TriBITS in TriBITSPub/TriBITS#464 which more
correctly deals with semi-colons and list arguments, this was being set as a
single property value "FAIL\\;BUMMER" which did not match the output.

Passing in the list with:

   FAIL_REGULAR_EXPRESSION FAIL BUMMER

works both with old and new TriBITS and is more logical given how CMake/CTest
is supposed to be used.  (The old TriBITS documentation was incorrect.)

…BITS#464)

Tt turns out that these tests really did not need to pass a list of values to
FAIL_REGULAR_EXPRESSION.  Each test was really only looking for a single value
of 'FAILED' or 'BUMMER'.  And you can argue that the tests are stronger and
better by taking out the other value that is not supposed to be printed in
that test case.

…rilinos#10534)

This test code has not been enabled or built in over 10 years due to a defect
added in commit:

  7aabd5d "Isorropia: Make EpetraExt a required dependency"
  Author: Brent Perschbacher <bmpersc@sandia.gov>
  Date:   Tue Feb 21 13:37:59 2012 -0700 (10 years ago)

  M       packages/isorropia/cmake/Dependencies.cmake

But note that HAVE_EPETRA was getting set to true unconditionally due to a
defect in how it was set as:

  SET(HAVE_EPETRAEXT ${PACKAGE_NAME}_ENABLE_EpetraExt)

instead of as dereferencing the variable with:

  SET(HAVE_EPETRAEXT ${${PACKAGE_NAME}_ENABLE_EpetraExt})

So the value "Isorropia_ENABLE_EpetraExt" evaluates to true in CMake so
EpetraExt support was always required, even if EpetraExt support in Isorropia
was disabled!

So I just fixed this to make it clear that this was **always** on by changing
this to:

  SET(HAVE_EPETRAEXT ON)

This test code was getting switched on again with the new TriBITS that sets
Isorropia_ENABLE_EpetraExt to TRUE, even for required dependencies (see PR