From version 4.5, Metview uses CMake for its compilation and installation. This is part of the process of homogenising the installation procedures for all ECMWF packages.
CMake installation instructions
The CMake build system is used to build ECMWF software. The build process comprises two stages:
- CMake runs some tests on the system and finds out if required software libraries and headers are available. It uses this information to create native build tools (e.g. Makefiles) for the current platform.
- The actual build can take place, for example by typing '
To install any ECMWF software package, CMake needs to be installed on your system. On most systems it will be already installed or this can be done through the standard package manager to install software. For further information to install CMake see
During a build with CMake there are three different directories involved: The source dir, the build dir and the install dir.
|Source||Contains the software's source code. This is where a source tarball should be extracted to.|
|Build||Configuration and compiler outputs are generated here, including libraries and executables.|
|Install||Where the software will actually be used from. Installation to this directory is the final stage.|
Of these, the source and build directories can be anywhere on the system. The installation directory is usually left at its default, which is
/usr/local. Installing software here ensures that it is automatically available to users. It is possible to specify a different installation directory by adding
-DCMAKE_INSTALL_PREFIX=/path/to/install/dir to the CMake command line.
ECMWF software does not support in-source builds. Therefore the build directory cannot be (a subdirectory of) the source directory.
Quick Build Example
Here is an example set of commands to set up and build a software package using default settings. More detail for a customised build is given below.
On a machine with multiple cores, compilation will be faster by specifying the number of cores to be used simultaneously for the build, for example:
make command fails, you can get more output by typing:
The software distribution will include a small set of tests which can help ensure that the build was successful. To start the tests, type:
As before if you have multiple cores, you can run the tests in parallel by:
Some projects might not be set up to run tests in parallel. If you experience test failures, run the tests sequentially.
If the tests are successful, you can install the software:
General CMake options
Various options can be passed to the CMake command. The following table gives an overview of some of the general options that can be used. Options are passed to the
cmake command by prefixing them with -D, for example -DCMAKE_INSTALL_PREFIX=/path/to/dir.
|where to install the software|
to select the type of compilation:
(release with debug info)
|Additional flags to pass to the C++ compiler|
|Additional flags to pass to the C compiler|
|Additional flags to pass to the Fortran compiler|
The C, C++ and Fortran compilers are chosen by CMake. This can be overwritten by setting the environment variables CC, CXX and F77, before the call to
cmake, to set the preferred compiler. Further the variable
CMAKE_CXX_FLAGS can be used to set compiler flags for optimisation or debugging. For example, using
CMAKE_CXX_FLAGS="-O2 -mtune=native" sets options for better optimisation.
Finding support libraries
If any support libraries are installed in non-default locations, CMake can be instructed where to find them by one of the following methods. First, the option CMAKE_PREFIX_PATH can be set to a colon-separated list of base directories where the libraries are installed, for example
-DCMAKE_PREFIX_PATH=/path/where/my/sw/is/installed. CMake will check these directories for any package it requires. This method is therefore useful if many support libraries are installed into the same location.
Debugging configure failures
If CMake fails to configure your project, run with debug logging first:
This will output lots of diagnostic information (in blue) on discovery of dependencies and much more.
Requirements to build Metview
The following table lists the dependencies Metview requires to be built from source. Please note, if you install these package from source you also might have to install the respective "-devel" packages.
|Fortran||http://gcc.gnu.org/fortran/||If Fortran or BUFR support needed|
|Third party libraries|
if Metview's user interface is required (version 4.6.2 or later of Qt is needed).
Note that on some systems it is also necessary to install the libQtWebKit-devel development package (it may have different names on different systems)
|ImageMagick||http://www.imagemagick.org/script/index.php||if Metview's user interface is required|
|proj||http://trac.osgeo.org/proj/||if OPERA radar data support is required|
|odb-api||ODB-API Home||if ODB support needed|
|magics||Magics||if plotting support is needed. Note that Magics should be configured with the -DENABLE_METVIEW=ON |
For a 'pure batch' installation of Metview with no user interface, it is possible to supply Magics with the option -DENABLE_METVIEW_NO_QT=ON
CMake options used in Metview
CMake options are passed to the
cmake command by prefixing them with -D, for example -DENABLE_UI=OFF.
|ENABLE_UI||enables the Qt-based user interface||ON|
|ENABLE_PLOTTING||enables plotting capabilities using Magics||ON|
|ENABLE_OPERA_RADAR||enables the Opera Radar Filter module (requires the proj4 library and headers)||OFF|
|ENABLE_MARS||enables MARS access (not required if using through the Web API)||OFF|
|MARS_LOCAL_HOME||sets the path to where local MARS is installed|
|ENABLE_ODB||enables processing and plotting of ODB data||OFF|
|ENABLE_MARS_ODB||enables ODB capabilities in MARS client||OFF|
|ENABLE_USAGE_LOG||enables logging of Metview startup calls||OFF|
|LOG_DIR||path to where to log the Metview startup calls|
|METVIEW_SCRIPT||name of the generated Metview startup script||metview|
|EXTRA_CONFIG_PATH||path to optional directory containing metview_local* script files|
|ENABLE_QT_DEBUG||outputs additional log messages from Qt-based modules||OFF|
|EXTRA_TITLE||build-specific title to add to the log entries|
enables the deprecated Motif-based user interface
|ENABLE_INPE||enables INPE modules||ON|
|Path options - only required when support libraries are not installed in default locations|
|GRIB_API_PATH||path to where GRIB_API has been installed|
|MAGICS_PATH||path to where Magics has been installed||Only required if plotting is enabled|
|NETCDF_PATH||path to where netCDF has been installed|
|ODB_API_PATH||path to where ODB_API has been installed||Only required if ODB is enabled|
|ODB_PATH||path to where the original ODB has been installed||Optional if ODB is enabled|
|EMOS_PATH||path to where Emoslib has been installed||Also set EMOS_LIB_NAME|
|FDB_PATH||path to where fdb has been installed||Only required if MARS is enabled|
|PROJ4_PATH||path to where proj4 has been installed||Only required if OPERA Radar is enabled|
|FLEXTRA_PATH||path to where the FLEXTRA executable has been installed||See Tutorials for more on FLEXTRA|
Notes for installers of Metview 3
If you have installed Metview 3 before, then here are some things to note. Metview 4 does not use directly OpenGL for its on-screen graphics; therefore, it is not necessary to build your own Mesa library anymore. However, Metview 4 does not come with its own Emoslib; therefore, it will be necessary to install your own.
Metview 4 can be installed side-by-side with an existing Metview 3 installation. However, note that the default startup script will be
so make sure this will not clash with an existing installation. See the table of CMake options for the flag which will allow you to change this.
See also the Installation FAQ.