As decided in the doodle for FS#1395, really old MPI versions will be
deprecated in the 2.3 release.

This patch adds an additional check to the MPI tests for autotools and
cmake that makes sure the user has an MPI which is at least compliant
with MPI-2.1. If the user's version is older, a warning is generated.

The test uses version macros instead of the more standard runtime
MPI version query interface to avoid problems when cross-compiling.

With this patch the USE_CMAKE environment variable is queried whether or not CMake
should be used. This can be overwritten in an options file or on the command line
using either the --cmake or --no-cmakr switch.

The check for constexpr was not copied into the distribution.
Unfortunately, the autotools will only warn about a missing macro, but
give valid results otherwise. This patch adds the cxx11_constexpr.m4 to
Makefile.am, so that it is distributed on "make dist".

This fixes a Clang warning.

Make more precise statement to the user if no template specialization
of DenseMatrixAssigner is found. Also use dune_static_assert instead of
static_assert.

This implementation of an IdentityMatrix does not hold any data. It
implements a reduced version of the DenseMatrix interface including a
standard C++ cast operator to a Dune::FieldMatrix. This code illustrates
the new forward capabilities of DenseMatrixAssigner.

The default implementation of DenseMatrixAssigner replaces redundant
code in dense matrix constructors taking a primitive data type.

[DenseMatrix] Introduce DenseMatrixAssigner for initializing dense matrices from arbitrary right hand sides

In order to overcome the deficiencies of istl_assign_to_fmatrix() we
propose to always use a struct DenseMatrixAssigner for initializing a
dense matrix. This class can be specialized by users for any right hand
side to be used in an assignment.  The new mechanism can bei default
implemented using existing implementations of istl_assign_to_fmatrix()
as well.

There was one line of diff missing in that patch. Sorry!

Previously, copying a RemoteIndexListModifier resulted in a double
free of a pointer to a list of global indices. There was no need to
use a pointer to this, though. After this patch we store the list directly
by value. Thus the default copy constructor works but makes a rather cheap copy
of a (most often empty) singly linked list and we get rid off the non-default
destructor.

Update minimal required versions to these mentioned in the documentation.
Quote argument in AC_PREREQ.
(Thanks to Bård Skaflestad for the heads-up)

Configration of dune fails if the manpages are not distributed.

Professional software has man pages.

Properly quote variable GMP_FOUND.
Do not reuse variable GMP_FOUND.

Treating directories with spaces in autotools is a can of worms.
We now check whether this situation is given and issue an error when
building with autotools.

With this patch dunecontrol will work even if the path to the module
contains spaces. Unfortunately, the tests for dune modules will still
not work as I could not figure out how to use quoting in the CPP_FLAGS.

At least Clang 3.2 complains.

This is a follow up to the previous patch. It corrects the situation
for new modules created with duneproject.

Torbjörn Klatt authored 11 years ago and Markus Blatt committed 11 years ago

Previously the module path was a relative to the install
directory. Because of this dependant modules did not find
the macros of dune-common and cmake configure did not succeed.
With this the CMake package configuration file contains the
full path and the macros are found (even if there is no local
dune-common lying around).

Somehow MPI was never activated for the headercheck. This patch fixes this.

The config.h entry for ENABLE_UMFPACK should be in the same
module as the test for UMFPack, to avoid future conflicts with
multiple modules running the test for umfpack.

DenseVector is the base class of a CRTP. This is kind of an abstract
class that works only if it can be casted into a derived type. Copying
the base class is pointless.

This patch marks the copy constructor of DenseVector private (not even
implementing it), so that it cannot be called anymore. Once we have a
DUNE_DELETE macro, we should use it, here.

At the same time, the default constructor is marked protected, so that only
derived classes can instantiate a DenseVector. Notice that this does not
prevent the user from building a derivced class which does not pass
itself as implementation to DenseVector. It is (to my knowledge) not possible
to make CRTPs completely fail-safe.

Cmake by default added a dependeny variable for each headercheck
target. Given the amount of such targets and the fact, that these
dependencies should not be changed by the user, this is unwanted
behaviour. The patch fixes this by removing the variable from the
cache.

The removed lines would be necessary if the included headers
or libraries would be tested. As this is not the case, they can
be safely removed.
The way they were used, every test tried to link to the libraries
amd and umfpack causing every test to fail if the libraries are
not present. The changed declaration of CMAKE_REQUIRED_* should
be encapsulated together with the test inside
  include(CMakePushCheckState)
  cmake_push_check_state()
and
  cmake_pop_check_state()
to ensure that the definitions do not leak the test.

Dominic Kempf authored 11 years ago and Markus Blatt committed 11 years ago

Using variables as part of regex does not work with special characters.
Therefore previously the headercheck failed during configure if e.g.
the modules were located in $HOME/C++/. This patch fixes this by not
using regex to strip the location of the modules from the path any more.

This log message was amended by the pusher.