Otherwise we are running into weired problems as downstream modules
refuse to set the version information correctly. This leads to
unset DUNE_MODULE_VERSION_{MINOR,MAJOR,REVISION} in config.h.
This only happened for dune-istl which was first requested by opm-core
but that module did not set the version information. Opm-grid requested
it, too. But this time this information was not correctly put into
opm-grid-config.cmake leading to erroneous checks like:
if (DEFINED DUNE_ISTL_VERSION_MAJOR AND NOT "${DUNE_ISTL_VERSION_MAJOR}" STREQUAL "")
message (WARNING "Incompatible value \"${DUNE_ISTL_VERSION_MAJOR}\" of variable \"DUNE_ISTL_VERSION_MAJOR\"")
endif ()
Note there should have been a number instead of an empty string. The version
numbers were also undefined in config.h of opm-grid. Same picture for opm-simulators.
My assumption is that for these modules Finddune-istl.cmake was implicitly included
when searching for opm-core. That module sets the version in the parent scope only
which might cause this problem. I am not sure though.
With this commit the version is always added to CONFIG_VARS for the first most upstream
OPM module that requests a dune module. This fixes the issue at least for me on my system.
Also note that this problem did not occur in all configurations. I experienced breaking
builds with the installed dune module 2.4.1 of Debian backports while using dunecontrol.
The warnings about unmatching versions happened with uninstalled 2.3, too.
newer petsc changed layout a bit. as of petsc 3.6.4 a
proper pkg-config file is ready for use.
additionally, align the find module with how petsc is expected to be
configured. the environment variable PETSC_DIR is the root of the source
tree and PETSC_ARCH is the build configuration, located in
$PETSC_DIR/$PETSC_ARCH
the old approach is kept for backwards compatibility but will not work
with newer petsc, in particular not if petsc is built static.
This commit brings the system probe for module
opm-flowdiagnostics
up-to-date with recent changes in the public interface of the
module's main entry point. Commit OPM/opm-flowdiagnostics@8456e81
removed the nested type
Opm::FlowDiagnostics::Toolbox::PoreVolume
so we must drop all references to this type.
the purpose is to make CI systems like travis happy because ctest
counts tests with status "(Not Run)" as failed and returns a non-zero
exit code.
hiding disabled ctests can be achived by adding the
-DADD_DISABLED_CTESTS=OFF cmake parameter.
This commit recognizes the opm-flowdiagnostics-applications module
in OPM's centralised build system. We make requisite changes to the
FIND support for module opm-flowdiagnostics (header 'tarjan.h' is
currently not public) and make a first cut at explicit prerequisites
for FD-Applications.
Added a new optional dependency the ERT Python distribution to
opm-simulators, and a cmake function 'opm_add_python_test' which can be
used to simplify integration tests of the type:
1. Run flow on a model.
2. Compare the results with a well known reference solution.
opm-output is about to remove the OutputWriter.hpp header, which means
this test program will break. Wells.hpp and default-constructed
data::Rates is a better fit for the future.
Andreas Lauser
Markus Blatt
Atgeirr Flø Rasmussen
Robert Kloefkorn
Arne Morten Kvarving
Joakim Hove
Pål Grønås Drange
Kristian Hole-Drabløs
Bård Skaflestad
Jørgen Kvalsvik