Search for modules in this library in the same style as Boost; by
specifying a list of components. If all on the list is present, then
a list of libraries to link with, is returned.
BOOST_TEST_DYN_LINK must be defined if you are *building* Boost::Test as
a shared object, and the client programs must also include this option
when they use the header files, to link to it correctly.
Normally a package returns a set of standard variables such as
Xxx_INCLUDE_DIRS. Adding this to a list that is collected for the
project amounts to a bunch of boilerplate which can rather be
encapsulated in a macro.
We want to compile with the -std=c++0x since DUNE uses this settings and
we need to use the same ABI as that one if we are going to link with it.
Find module is a copy of the one in cmake/modules in (the cmake branch of)
dune-common.
Some files are distributed with the source, but does not end up in the
final shared object. These files are captured by the glob, so we have
to remove them afterwards. (There are fewer of these files than
explicitly list all files that should be included in the build).
A fundamental problem with the configure_file() command is CMake is that
the author of the template file config.h.in must know which variables
all modules need to have defined, and this is not easily communicated.
The idea behind this function is: Each Find-module can return a list of
variables that it has filled with probed values, and that should be
written to config.h, which is again included by the header file of this
module.
Each project thus defines its own config.h file, which is common between
all module's headers.
We inherited this test from the Dune repositories, but failed to install
sufficient protection in the form of symbol renaming to guarantee that
both tests could coexist.
This issue was exposed by a recent pull request in the OPM-Porsol
repository, namely OPM/opm-porsol#16.
The test started failing in commit 7d7f62e, but this was not detected
due to no automatic test environment. The commit changed the cell
numbering from "per-column (K,I,J)-ordering" to "per-plane
(I,J,K)-ordering". Consequently, the "correct_answer" seized to be
correct.
This change restores the "correct_answer" in the ordering introduced by
commit 7d7f62e.
While here, adjust style of the DisjointColumn test case for legibility.
