cmake/Modules/OpmInstall.cmake | 2 1 + 1 - 0 !
cmake/Modules/OpmProject.cmake | 19 17 + 2 - 0 !
2 files changed, 18 insertions(+), 3 deletions(-)

 install cmake package configuration to correct location with lib.

If we ship a library than we are architecture dependent and should
allow for installations for multiple architectures. For this to work
the package configuration files need to be installed in an
architecture aware location (e.g. lib/arch/cmake/module_name on
Debian). Just like the pkg-config files. This is done with this patch.

cmake/Modules/OpmProject.cmake | 12 4 + 8 - 0 !
1 file changed, 4 insertions(+), 8 deletions(-)

 always install pkg-config files in multiarch library directory.

Even if we do not ship a library, our pkg-config file will include
architecture dependent library paths and libraries that we depend on.
Hence we still need to install into the multiarch library directory.

tests/test_calculateCellVol.cpp | 3 2 + 1 - 0 !
1 file changed, 2 insertions(+), 1 deletion(-)

 allow more deviations when calculating volume to fix tests on i386.

tests/test_OpmLog.cpp | 2 1 + 1 - 0 !
1 file changed, 1 insertion(+), 1 deletion(-)

 prevent 32bit overflow when testing opmlog.

tests/parser/ScheduleTests.cpp | 4 2 + 2 - 0 !
tests/test_nonuniformtablelinear.cpp | 2 1 + 1 - 0 !
2 files changed, 3 insertions(+), 3 deletions(-)

 use boost_check_close for floating point comparisons.

BOOST_CHECK_EQUAL will only reliably work if the computations on the
left and right hand side were made in the exact same order, which is
often not the case.

tests/test_ESmry.cpp | 2 1 + 1 - 0 !
1 file changed, 1 insertion(+), 1 deletion(-)

 make test_esmry.cpp usable for autopkgtests.

In 2021.10 it uses the header tests/WorkArea.cpp that gets copied
to the same directory as tests/test_ESMry.cpp and we need to make in includable
from the same source directory (which will not be tests but the
uppermost source directory).

CMakeLists.txt | 3 2 + 1 - 0 !
1 file changed, 2 insertions(+), 1 deletion(-)

 skip rpath for python lib.

This will prevent lintian error:
```
E: python3-opm-common: custom-library-search-path RUNPATH /build/opm-common-2023.04+ds/obj-x86_64-linux-gnu/lib [usr/lib/python3/dist-packages/opm/opmcommon_python.cpython-311-x86_64-linux-gnu.so]
```

tests/material/test_tabulation.cpp | 4 3 + 1 - 0 !
1 file changed, 3 insertions(+), 1 deletion(-)

 [hack] restrict tabulation tests to double to prevent ftbfs.

Theses tests have been failing all the time, but im previous
opm-common releases these fails have not been fatal. They are now,
because upstream changed this (but without fixing architectures).

While this is a hack, we do because the errors were actually there in
previous releases, but did not make the build failing. Also this will
allow binary packages to actually migrate to testing.

CMakeLists.txt | 6 6 + 0 - 0 !
1 file changed, 6 insertions(+)

 prevent pybind11 from stripping debug symbols

 pybind11_add_module will strip debug symbols if CMAKE_BUILD_TYPE is
 neither Debug nor RelWithDebInfo) and there is no option to prevent
 this. To fix this we temporarily reset the CMAKE_BUILD_TYPE to Debug

We noticed the problem in the build logs:

   dh_dwz -a -O--max-parallel=2
	dwz -- debian/libopm-common/usr/lib/x86_64-linux-gnu/libopmcommon.so.2024.04
	install -m0755 -d debian/python3-opm-common/usr/lib/debug/.dwz/x86_64-linux-gnu
	dwz -mdebian/python3-opm-common/usr/lib/debug/.dwz/x86_64-linux-gnu/python3-opm-common.debug -M/usr/lib/debug/.dwz/x86_64-linux-gnu/python3-opm-common.debug -- debian/python3-opm-common/usr/lib/python3/dist-packages/opm/opmcommon_python.cpython-311-x86_64-linux-gnu.so debian/python3-opm-common/usr/lib/python3/dist-packages/opm_embedded.cpython-311-x86_64-linux-gnu.so
	install -m0755 -d debian/libopm-common-bin/usr/lib/debug/.dwz/x86_64-linux-gnu
	dwz -mdebian/libopm-common-bin/usr/lib/debug/.dwz/x86_64-linux-gnu/libopm-common-bin.debug -M/usr/lib/debug/.dwz/x86_64-linux-gnu/libopm-common-bin.debug -- debian/libopm-common-bin/usr/bin/arraylist debian/libopm-common-bin/usr/bin/co2brinepvt debian/libopm-common-bin/usr/bin/compareECL debian/libopm-common-bin/usr/bin/convertECL debian/libopm-common-bin/usr/bin/hysteresis debian/libopm-common-bin/usr/bin/opmhash debian/libopm-common-bin/usr/bin/opmpack debian/libopm-common-bin/usr/bin/rst_deck debian/libopm-common-bin/usr/bin/summary
dwz: debian/python3-opm-common/usr/lib/python3/dist-packages/opm/opmcommon_python.cpython-311-x86_64-linux-gnu.so: .debug_info section not present
dwz: debian/python3-opm-common/usr/lib/python3/dist-packages/opm_embedded.cpython-311-x86_64-linux-gnu.so: .debug_info section not present

msim/include/opm/msim/msim.hpp | 3 3 + 0 - 0 !
msim/src/msim.cpp | 3 2 + 1 - 0 !
tests/msim/test_msim_ACTIONX.cpp | 5 2 + 3 - 0 !
3 files changed, 7 insertions(+), 4 deletions(-)

 fix python usage in msim and its tests.

There can only be one embedded Python interpreter in a program,
as Python will never totally cleaned up if it is embedded (using
pybind11).

Unfortunately, test_msim_ACTIONX.cpp created a couple Python
interpreters and with Python 3,12 this resulted in double frees.

This commit fixes this by creating the Python interpreter in a static
shared_ptr and making the tests use that.

cmake/Modules/OpmAliases.cmake | 1 0 + 1 - 0 !
1 file changed, 1 deletion(-)

 removed have_umfpack alias.

This is not used even in DUNE 2.7 and we get annoying redefined
warnings with it.

cmake/Modules/OpmPackage.cmake | 3 3 + 0 - 0 !
1 file changed, 3 insertions(+)

 workaround possibly set have_dune_module (dune 2.10).

For 2.10 this is set by the *-config.cmake file to TRUE which
suddenly resulted in
```
```
appearing in config.h. As TRUE is not defined we would pretend that there
is no dune-istl and compilation would fail where we pack BlockVectors
for serialization.

opm/common/utility/Serializer.hpp | 2 1 + 1 - 0 !
opm/common/utility/Visitor.hpp | 2 1 + 1 - 0 !
opm/material/fluidmatrixinteractions/EclMaterialLawManagerHystParams.cpp | 1 1 + 0 - 0 !
opm/material/fluidmatrixinteractions/EclMaterialLawManagerInitParams.cpp | 1 1 + 0 - 0 !
opm/material/fluidmatrixinteractions/EclMaterialLawManagerReadEffectiveParams.cpp | 1 1 + 0 - 0 !
5 files changed, 5 insertions(+), 2 deletions(-)

 fix the obvious typos in the license specifications.

Here I am 100% sure that these are fixes.

python/setup.py.in | 2 1 + 1 - 0 !
1 file changed, 1 insertion(+), 1 deletion(-)

 corrected license of python packages to gplv3+

tests/MSW.DATA | 21 5 + 16 - 0 !
tests/MSW_2WELSEGS.DATA | 21 5 + 16 - 0 !
tests/MSW_RESTART.DATA | 21 5 + 16 - 0 !
3 files changed, 15 insertions(+), 48 deletions(-)

 use correct license (odbl instead gpl) for mswell data files

cmake/Modules/ConfigVars.cmake | 14 9 + 5 - 0 !
1 file changed, 9 insertions(+), 5 deletions(-)

 fixed handling of cmake bool in config.h

Only values 0 or 1 were handled correctly.
A value defined to TRUE, ON, OFF would have made the value appear in
the config.h as text e.g.
```
```

This caused a lot of trouble with refactored DUNE build system.

cmake/Modules/FindSuiteSparse.cmake | 31 17 + 14 - 0 !
cmake/Modules/Finddune-istl.cmake | 2 1 + 1 - 0 !
2 files changed, 18 insertions(+), 15 deletions(-)

 use uppercase for components in findsuitesparse

This is in line with DUNE since 2015 and fixes problems when trying to
use opm-grid as a DUNE module.

opm/input/eclipse/EclipseState/EclipseState.cpp | 1 1 + 0 - 0 !
opm/input/eclipse/EclipseState/Grid/FaceDir.cpp | 1 1 + 0 - 0 !
opm/input/eclipse/Schedule/Action/ASTNode.cpp | 1 1 + 0 - 0 !
opm/input/eclipse/Schedule/Well/WellKeywordHandlers.cpp | 1 1 + 0 - 0 !
opm/io/eclipse/rst/state.cpp | 2 2 + 0 - 0 !
tests/test_AggregateActionxData.cpp | 1 1 + 0 - 0 !
6 files changed, 7 insertions(+)

 fixed: fmt::join is in ranges.h using {{fmt11}}

Backported fix from upstream's master.

opm/material/densead/EvaluationFormat.hpp | 2 1 + 1 - 0 !
opm/output/eclipse/Summary.cpp | 4 2 + 2 - 0 !
2 files changed, 3 insertions(+), 3 deletions(-)

 fixed: format() must be const for custom formatters using {{fmt11}}

python/tests/test_emodel.py | 5 4 + 1 - 0 !
1 file changed, 4 insertions(+), 1 deletion(-)

 use numpy.sum instead of sum to fix test with python 3.12 (closes:
 #1095362)

The sum function in Python 3.12 switched to using Neumaier summation
for floats for better accuracy.This broke on of the unit tests.
With this change we use numpy.sum to fix the failing test,
which still uses the not so accurate summation algorithm of sum in
Python 3.11

opm/common/utility/OpmInputError.cpp | 94 54 + 40 - 0 !
opm/common/utility/OpmInputError.hpp | 138 68 + 70 - 0 !
opm/input/eclipse/Parser/Parser.cpp | 22 13 + 9 - 0 !
tests/parser/integration/ParseKEYWORD.cpp | 25 15 + 10 - 0 !
4 files changed, 150 insertions(+), 129 deletions(-)

 abide by fmt-v11 format string requirements

In particular, do not mix automatic formatting requests (empty
braces, {}) with manual formatting requests (argument IDs and/or
named arguments).  The std::format() function of C++20 has the same
limitation, so it's beneficial to prepare for this now.

While here, prefer '//'-style comments, tidy up include statements
and make objects 'const' where possible.