Maria Haupt / ctools

…alues to the vector when called several times)

…(see #1360 for details).

…)-function of individual ctools

Conflicts:
	ChangeLog
	NEWS

As the GCTAObservation class now handles transparently the disposal and re-loading of information it is not needed anymore to re-load the event list upon call of the obs() method. We thus can put the obs() method inline as before.

The memory needs for ctobssim have been reduced by first working on a local copy of the CTA observation (this de-allocates properly and memory cache associated to the response functions), and by saving and disposing the event lists immediately if called using the execute() method (this is the method called when the tool is run from the command line).

…d scave() methods to ctool interface. Add read_ahead() method to handle reading of task parameters at the beginning of the task execution. Remove all m_read_ahead members from the derived classes.

ctobssim was not fully thread save as the models were shared among the threads which led to concurrent memory access when evaluating the Monte Carlo cache. The models are now copied within the parallel section, avoiding concurrent memory access (#1258).

True photons are now drawn over a sufficiently large energy band so that when dispersed the selected energy range is correctly filled.

A new logic has been implemented concerning the caldb parameter. If the name specified by caldb is a directory, the directory is set as the root directory of the calibration database. The file specified by irf must then be a file provided within this root directory.

If the name is not a directory it is used as the mission name that is used to build the calibration directory path. The irf is then the response name.

The instrument and id attributes provided by an observation were not considering when simulating source and background events. This has been changed now that only events will be simulated when the instrument and id are valid for a given model.

Output observation definition files are XML files, hence they should have a .xml suffix to avoid confusion with FITS files.

The time slice duration is now adapted to the source model flux. The m_max_photons member now defines the maximum number of photons that should be simulated in one slice. This prevents core dumps due to memory allocation problems in case that large fluxes have been specified.

This partially adresses feature #807.

The GCTARoi::contains() method is now used to check for ROI containment.

Background simulations can now be done for any model of type GModelData. This is thanks to the abstract mc() method that has been introduced in GModelData. A test case for the new GCTABackground model has been added to the unit tests.

…g so that also sources outside but at the edge of the ROI will be simulated.

Conflicts:
	scripts/cssens.py
	scripts/cstsdist.py
	src/ctmodel/ctmodel.cpp

GammaLib is undergoing a major interface change which requires adaption of the ctools code.