Pstream: added maxCommsSize setting to do (unstructured) parallel transfers in blocks.
Tested:
- with maxCommsSize 0 produces exactly same result as plus.develop
- compiles with label64
- with maxCommsSize e.g. 3 produces exactly same result as plus.develop
- with maxCommsSize=0 exactly the same messages (with Pstream::debug = 1) as plus.develop
See merge request !85
which provided warning about backward-compatibility issue with setting div
schemes for steady-state. It caused confusion by generating incorrect warning
messages for compressible cases for which the 'bounded' should NOT be applied to
the 'div(phid,p)'.
STYLE: only use paraview settings when actually available
- this means executing makeParaView prior to building OpenFOAM itself,
but is consistent with the instructions given by makeParaView,
and elminates anticipating the source location from the paraview
config file, which increases the build flexibilty for ThirdParty
- support gperftools-none, gperftools-system configurations
as per other third-party packages.
STYLE: clean up more environment variables
CONFIG: testing adios rule
* boost 1_62_0 (Sept 2016)
* CGAL 4.9 (Sept 2016)
- now has headers-only mode that could be interesting
* FFTW 3.3.5 (Jul 2016)
* openmpi 1.10.4 (Sept 2016)
- the first openmpi 2.x release is also available, but too early to switch
* paraview 5.2.0 (Nov 2016)
- builds without additional patching
STYLE: removed unneeded CMake environment variables
- "$FOAM_USER_APPBIN" and "$FOAM_USER_LIBBIN" have been added to
"foamOldDirs" in "etc/bashrc" and "etc/config.sh/unset"
- "$OPAL_PREFIX" is now undefined in the option "SYSTEMOPENMPI" within
"etc/config.sh/mpi", but only if the path defined in this variable
is cleaned when using "foamCleanPath".
- "$OPAL_PREFIX" is now also conditionally undefined in
"etc/config.sh/unset" when the path is picked up by "foamCleanPath".
Patch contributed by Bruno Santos
Resolved bug-report http://bugs.openfoam.org/view.php?id=2210
- Now that the metisDecomp uses the metis definition for float/double,
do not need to verify the scalar sizes.
Note:
- could drop precision qualifier for metis, scotch installation
(include, lib) as being unneeded, but it is simpler to keep them
and continue to use the FOAM_EXT_LIBBIN path
(in case other ThirdParty software is compiled with different
precisions).
- Cannot test if these older reader modules actually build,
owning largely to build issues since with paraview 3.14 (Feb 2012)
and paraview 3.98 (Dec 2012) themselves.
- Follow similar pattern as per boost/CGAL with the '*-none' to disable,
and '*-system' for system installations.
Also support central non-ThirdParty installations by specifying the
version as "*-system", but also providing the fully qualified
*_ARCH_PATH too.
--
ENH: Verify label and scalar sizes used in the metis header.
- ensure they match with expected values, which can be obtained via
the OpenFOAM definitions for WM_LABEL_SIZE and WM_PRECISION_OPTION
On 64-bit systems, the system installations of boost, cgal are under
lib64/. The behaviour for a ThirdParty build is mostly lib/ but this
can also be changing.
Boost 1_62_0 and older build into 'lib/'.
CGAL-4.9 builds into 'lib64/', older versions into 'lib/'.
Future-proof things by using lib$WM_COMPILER_LIB_ARCH for boost and
cgal build rules, and forcing these as build targets in the ThirdParty
makeCGAL as well.
--
STYLE: check for boost/version.hpp, CGAL/version.h instead their directories
- On the first call, ParaView_DIR is unset and thus the clean-path
fails with the warning "ParaView_DIR: Undefined variable."
This looks messy, but is of no _major_ consequence since paraview
doesn't need to be removed anyhow. The only slight risk is that the
path to a third-party cmake might not be cleaned.
- Patch as per Bruno's suggestion.
- It is incorrect to prefix the assignment with an 'export' since this
automatically marks the overall command as successful and circumvents
the fallback.
There is no simple way to have proper behaviour when sourced with a
local directory path, but at least it now uses the fallback.
It is still easy to wreak the mechanism with valid but confusing input.
For example,
". /path/to/openfoam/etc/././bashrc"
The only real certainty is that "${BASH_SOURCE%/*}"
should point to the 'etc/' directory. In which case,
cd ${BASH_SOURCE%/*} # <- now in the etc/ directory
pwd=$(pwd -P) # <- fully-qualified path to etc/
pwd=${pwd%/*/*} # <- up two parent levels
- This works with ". ./bashrc", but fails with ". bashrc"
(probably not so common).
- Con: The construct requires an additional sub-shell.
- In the foundation version they introduced a construct to handle
the transition from 'wmRefresh' as an alias to 'wmRefresh' as
a shell function. This transition is unnecessary for OpenFOAM+
since 1606 used wmREFRESH (not wmRefresh) as an alias.
For portability it is important to avoid this non-POSIX
"type -t". It causes issues with dash and with zsh
(mentioned in issue #277).
type -t dash -> -t: not found
type -t zsh -> zsh: bad option: -t
Note: zsh users may still noticed other problems.
For example, the POSIX 'unset -f' normally has no output, but in zsh
it reports an error and has exit code 1 if the function was not
previously defined. Whereas in POSIX (including bash, dash) it only
returns non-zero if the name(s) could not be unset.
The change from C++0x to C++11 allows all of C++11 functionality to be
used in OpenFOAM, in particular constructor delegation which avoids code
duplication or constructor helper functions. However, this also means a
change to the minimum gcc version supported which is now 4.7 rather than
4.5.
Note that gcc-4.7 does not support the entire C++11 standard but does
support all of the functionality currently needed for further OpenFOAM
development. The minimum gcc-version which supports the entire C++11
standard is 4.8 which is now the recommended minimum gcc version.
